[GlobalISel] Rework more/fewer elements for vectors

[Petar.Avramovic]

Summary

Artifact combiner is not able to access individual elements after using
LCMTy style merge/unmerge, extract and insert to change vector number of
elements (pad with undef or split to sub-vector instructions).
Use unmerge to individual elements instead and then merge elements into
requested types.
Change argument lowering for vectors and moreElementsVector to use
buildPadVectorWithUndefElements and buildDeleteTrailingVectorElements.
FewerElementsVector had a few helpers that had different behaviour,
introduce new helper for most of the opcodes.
FewerElementsVector helper is more flexible since it can create leftover
instruction smaller then requested type (useful in case target wants to
avoid pad with undef and use fewer registers). If target does not want
leftover of different type it should call more elements first.
Some helpers were performing more elements first to have split without
leftover. Opcodes that used this helper use clampMaxNumElementsStrict
(does more elements first) in LegalizerInfo to avoid test changes.
Fixes failures caused by failing to combine artifacts created during
more/fewer elements vector.

[Petar.Avramovic]

Rebase and ping.

[Petar.Avramovic]

Rebase and ping.

[arsenm]

Thanks, this mostly looks less ugly than before

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

994

Why this change? I thought this previously handled arbitrary cases, and this doesn't?

3633–3639

I think the prevailing style is to use static functions instead of anonymous namespaces

3651

Not sure what the point of this check is

3816

Might as well hardcode G_PHI

6592

Use the default small size?

6870

I think it's more appropriate to lower the vector operation in terms of G_EXTRACT_VECTOR_ELT and not direct to the register extracts, G_SHUFFLE_VECTOR is not a legalization artifact

llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp

231

4 seems too small, use the default or 8?

253

Ditto

llvm/lib/CodeGen/GlobalISel/Utils.cpp

938

I'm not sure what to do about scalable vectors, but no legalization has been added for those as far as I know

llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

2248–2251

Can you split this into a separate patch? I know I have one that does the same thing somewhere

2282

Can you also split this, as it's not dependent on the larger legalizer changes or new builder functions

2284

2 seems too small, most vectors are probably 4 or 8

4696–4712

This is a lot longer, is there a simpler way?

[foad]

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

3669

is_contained(NonVecOpIndices, OpIdx)

6683

This is the same condition as two lines above.

llvm/lib/CodeGen/GlobalISel/Utils.cpp

938

Can you use LLT::scalarOrVector here, or OrigTy.changeElementCount?

[Petar.Avramovic]

Rebase and address comments.

[Petar.Avramovic]

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

994

reduceOperationWidth did number of elements splits and narrow scalar for freeze, here I extracted narrow scalar only.

3651

This is for load/store with vector pointer address. the whole function is annoying check that we don't try to split operands of non-intended Opcode.

6870

Sure, we do unmerge via custom legalize of G_EXTRACT_VECTOR_ELT

llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

4696–4712

I don't see it. Whole function is complicated, handles many cases, and packs all output operands into ResultRegs. And we no longer pad with undef in all cases.

[Petar.Avramovic]

Cleanup.

[arsenm]

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

3698

Convert if chain to else if and use llvm_unreachable instead of having to keep this list consistent

3743

Typo witout

llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp

231

Same applies elsewhere in patch

[Petar.Avramovic]

Address comments, fix more spelling mistakes.

[mehdi_amini]

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

3658

This line does not seem right? (flagged by coverity by the way)

[Petar.Avramovic]

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

3658

Thanks, that line was pasted there by mistake.

llvm/include/llvm/CodeGen/GlobalISel/LegalizerHelper.h

void extractParts(Register Reg, LLT Ty, int NumParts,
                     SmallVectorImpl<Register> &VRegs);
 
   /// Version which handles irregular splits.
   bool extractParts(Register Reg, LLT RegTy, LLT MainTy,
                     LLT &LeftoverTy,
                     SmallVectorImpl<Register> &VRegs,
                     SmallVectorImpl<Register> &LeftoverVRegs);
 
+  /// Version which handles irregular sub-vector splits.
+  void extractVectorParts(Register Reg, unsigned NumElst,
+                          SmallVectorImpl<Register> &VRegs);
+
   /// Helper function to build a wide generic register \p DstReg of type \p
   /// RegTy from smaller parts. This will produce a G_MERGE_VALUES,
   /// G_BUILD_VECTOR, G_CONCAT_VECTORS, or sequence of G_INSERT as appropriate
   /// for the types.
   ///
   /// \p PartRegs must be registers of type \p PartTy.
   ///
   /// If \p ResultTy does not evenly break into \p PartTy sized pieces, the
   /// remainder must be specified with \p LeftoverRegs of type \p LeftoverTy.
   void insertParts(Register DstReg, LLT ResultTy,
                    LLT PartTy, ArrayRef<Register> PartRegs,
                    LLT LeftoverTy = LLT(), ArrayRef<Register> LeftoverRegs = {});
 
+  /// Merge \p PartRegs with different types into \p DstReg.
+  void mergeMixedSubvectors(Register DstReg, ArrayRef<Register> PartRegs);
+
+  void appendVectorElts(SmallVectorImpl<Register> &Elts, Register Reg);
+
   /// Unmerge \p SrcReg into smaller sized values, and append them to \p
   /// Parts. The elements of \p Parts will be the greatest common divisor type
   /// of \p DstTy, \p NarrowTy and the type of \p SrcReg. This will compute and
   /// return the GCD type.
   LLT extractGCDType(SmallVectorImpl<Register> &Parts, LLT DstTy,
                      LLT NarrowTy, Register SrcReg);
 
   /// Unmerge \p SrcReg into \p GCDTy typed registers. This will append all of
MachineInstrBuilder createStackTemporary(TypeSize Bytes, Align Alignment,
                                            MachinePointerInfo &PtrInfo);
 
   /// Get a pointer to vector element \p Index located in memory for a vector of
   /// type \p VecTy starting at a base address of \p VecPtr. If \p Index is out
   /// of bounds the returned pointer is unspecified, but will be within the
   /// vector bounds.
   Register getVectorElementPointer(Register VecPtr, LLT VecTy, Register Index);
 
-  LegalizeResult fewerElementsVectorImplicitDef(MachineInstr &MI,
-                                                unsigned TypeIdx, LLT NarrowTy);
+  /// Handles most opcodes. Split \p MI into same instruction on sub-vectors or
+  /// scalars with \p NumElts elements (1 for scalar). Supports uneven splits:
+  /// there can be leftover sub-vector with fewer then \p NumElts or a leftover
+  /// scalar. To avoid this use moreElements first and set MI number of elements
+  /// to multiple of \p NumElts. Nonvector operands that should be used on all
+  /// sub-instructions without split are listed in \p NonVecOpIndices.
+  LegalizeResult fewerElementsVectorMultiEltType(
+      GenericMachineInstr &MI, unsigned NumElts,
+      std::initializer_list<unsigned> NonVecOpIndices = {});
 
-  /// Legalize a instruction with a vector type where each operand may have a
-  /// different element type. All type indexes must have the same number of
-  /// elements.
-  LegalizeResult fewerElementsVectorMultiEltType(MachineInstr &MI,
-                                                 unsigned TypeIdx, LLT NarrowTy);
-
-  LegalizeResult fewerElementsVectorCasts(MachineInstr &MI, unsigned TypeIdx,
-                                          LLT NarrowTy);
-
-  LegalizeResult
-  fewerElementsVectorCmp(MachineInstr &MI, unsigned TypeIdx, LLT NarrowTy);
-
-  LegalizeResult
-  fewerElementsVectorSelect(MachineInstr &MI, unsigned TypeIdx, LLT NarrowTy);
-
-  LegalizeResult fewerElementsVectorPhi(MachineInstr &MI,
-                                        unsigned TypeIdx, LLT NarrowTy);
+  LegalizeResult fewerElementsVectorPhi(GenericMachineInstr &MI,
+                                        unsigned NumElts);
 
   LegalizeResult moreElementsVectorPhi(MachineInstr &MI, unsigned TypeIdx,
                                        LLT MoreTy);
   LegalizeResult moreElementsVectorShuffle(MachineInstr &MI, unsigned TypeIdx,
                                            LLT MoreTy);
 
   LegalizeResult fewerElementsVectorUnmergeValues(MachineInstr &MI,
                                                   unsigned TypeIdx,
                                                   LLT NarrowTy);
   LegalizeResult fewerElementsVectorMerge(MachineInstr &MI, unsigned TypeIdx,
                                           LLT NarrowTy);
   LegalizeResult fewerElementsVectorExtractInsertVectorElt(MachineInstr &MI,
                                                            unsigned TypeIdx,
                                                            LLT NarrowTy);
 
-  LegalizeResult fewerElementsVectorMulo(MachineInstr &MI, unsigned TypeIdx,
-                                         LLT NarrowTy);
-
   LegalizeResult reduceLoadStoreWidth(GLoadStore &MI, unsigned TypeIdx,
                                       LLT NarrowTy);
 
-  /// Legalize an instruction by reducing the operation width, either by
-  /// narrowing the type of the operation or by reducing the number of elements
-  /// of a vector.
-  /// The used strategy (narrow vs. fewerElements) is decided by \p NarrowTy.
-  /// Narrow is used if the scalar type of \p NarrowTy and \p DstTy differ,
-  /// fewerElements is used when the scalar type is the same but the number of
-  /// elements between \p NarrowTy and \p DstTy differ.
-  LegalizeResult reduceOperationWidth(MachineInstr &MI, unsigned TypeIdx,
-                                      LLT NarrowTy);
+  LegalizeResult splitVectorLoadStoreToPow2NumElts(GLoadStore &MI, LLT MemTy);
 
   LegalizeResult fewerElementsVectorSextInReg(MachineInstr &MI, unsigned TypeIdx,
                                               LLT NarrowTy);
 
   LegalizeResult narrowScalarShiftByConstant(MachineInstr &MI, const APInt &Amt,
                                              LLT HalfTy, LLT ShiftAmtTy);
 
   LegalizeResult fewerElementsVectorReductions(MachineInstr &MI,

llvm/include/llvm/CodeGen/GlobalISel/LegalizerInfo.h

enum LegalizeAction : std::uint8_t {
 
   /// The operation should be implemented in terms of a wider scalar
   /// base-type. For example a <2 x s8> add could be implemented as a <2
   /// x s32> add (ignoring the high bits).
   WidenScalar,
 
   /// The (vector) operation should be implemented by splitting it into
   /// sub-vectors where the operation is legal. For example a <8 x s64> add
-  /// might be implemented as 4 separate <2 x s64> adds.
+  /// might be implemented as 4 separate <2 x s64> adds. There can be a leftover
+  /// if there are not enough elements for last sub-vector e.g. <7 x s64> add
+  /// will be implemented as 3 separate <2 x s64> adds and one s64 add. Leftover
+  /// types can be avoided by doing MoreElements first.
   FewerElements,
 
   /// The (vector) operation should be implemented by widening the input
   /// vector and ignoring the lanes added by doing so. For example <2 x i8> is
   /// rarely legal, but you might perform an <8 x i8> and then only look at
   /// the first two results.
   MoreElements,
 
return actionIf(
                  VecTy.getNumElements() < MinElements;
         },
         [=](const LegalityQuery &Query) {
           LLT VecTy = Query.Types[TypeIdx];
           return std::make_pair(
               TypeIdx, LLT::fixed_vector(MinElements, VecTy.getElementType()));
         });
   }
+
+  /// Set number of elements to nearest larger multiple of NumElts.
+  LegalizeRuleSet &alignNumElementsTo(unsigned TypeIdx, const LLT EltTy,
+                                      unsigned NumElts) {
+    typeIdx(TypeIdx);
+    return actionIf(
+        LegalizeAction::MoreElements,
+        [=](const LegalityQuery &Query) {
+          LLT VecTy = Query.Types[TypeIdx];
+          return VecTy.isVector() && VecTy.getElementType() == EltTy &&
+                 (VecTy.getNumElements() % NumElts != 0);
+        },
+        [=](const LegalityQuery &Query) {
+          LLT VecTy = Query.Types[TypeIdx];
+          unsigned NewSize = alignTo(VecTy.getNumElements(), NumElts);
+          return std::make_pair(
+              TypeIdx, LLT::fixed_vector(NewSize, VecTy.getElementType()));
+        });
+  }
+
   /// Limit the number of elements in EltTy vectors to at most MaxElements.
   LegalizeRuleSet &clampMaxNumElements(unsigned TypeIdx, const LLT EltTy,
                                        unsigned MaxElements) {
     // Mark the type index as covered:
     typeIdx(TypeIdx);
     return actionIf(
         LegalizeAction::FewerElements,
         [=](const LegalityQuery &Query) {
LegalizeRuleSet &clampNumElements(unsigned TypeIdx, const LLT MinTy,
     assert(MinTy.getElementType() == MaxTy.getElementType() &&
            "Expected element types to agree");
 
     const LLT EltTy = MinTy.getElementType();
     return clampMinNumElements(TypeIdx, EltTy, MinTy.getNumElements())
         .clampMaxNumElements(TypeIdx, EltTy, MaxTy.getNumElements());
   }
 
+  /// Express \p EltTy vectors strictly using vectors with \p NumElts elements
+  /// (or scalars when \p NumElts equals 1).
+  /// First pad with undef elements to nearest larger multiple of \p NumElts.
+  /// Then perform split with all sub-instructions having the same type.
+  /// Using clampMaxNumElements (non-strict) can result in leftover instruction
+  /// with different type (fewer elements then \p NumElts or scalar).
+  /// No effect if the type is not a vector.
+  LegalizeRuleSet &clampMaxNumElementsStrict(unsigned TypeIdx, const LLT EltTy,
+                                             unsigned NumElts) {
+    return alignNumElementsTo(TypeIdx, EltTy, NumElts)
+        .clampMaxNumElements(TypeIdx, EltTy, NumElts);
+  }
+
   /// Fallback on the previous implementation. This should only be used while
   /// porting a rule.
   LegalizeRuleSet &fallback() {
     add({always, LegalizeAction::UseLegacyRules});
     return *this;
   }
 
   /// Check if there is no type index which is obviously not handled by the

llvm/include/llvm/CodeGen/GlobalISel/MachineIRBuilder.h

public:
   ///      type.
   /// \pre \p NumBits must be an integer representing the number of low bits to
   ///      be cleared in \p Op0.
   ///
   /// \return a MachineInstrBuilder for the newly created instruction.
   MachineInstrBuilder buildMaskLowPtrBits(const DstOp &Res, const SrcOp &Op0,
                                           uint32_t NumBits);
 
+  /// Build and insert
+  /// a, b, ..., x = G_UNMERGE_VALUES \p Op0
+  /// \p Res = G_BUILD_VECTOR a, b, ..., x, undef, ..., undef
+  ///
+  /// Pad \p Op0 with undef elements to match number of elements in \p Res.
+  ///
+  /// \pre setBasicBlock or setMI must have been called.
+  /// \pre \p Res and \p Op0 must be generic virtual registers with vector type,
+  ///      same vector element type and Op0 must have fewer elements then Res.
+  ///
+  /// \return a MachineInstrBuilder for the newly created build vector instr.
+  MachineInstrBuilder buildPadVectorWithUndefElements(const DstOp &Res,
+                                                      const SrcOp &Op0);
+
+  /// Build and insert
+  /// a, b, ..., x, y, z = G_UNMERGE_VALUES \p Op0
+  /// \p Res = G_BUILD_VECTOR a, b, ..., x
+  ///
+  /// Delete trailing elements in \p Op0 to match number of elements in \p Res.
+  ///
+  /// \pre setBasicBlock or setMI must have been called.
+  /// \pre \p Res and \p Op0 must be generic virtual registers with vector type,
+  ///      same vector element type and Op0 must have more elements then Res.
+  ///
+  /// \return a MachineInstrBuilder for the newly created build vector instr.
+  MachineInstrBuilder buildDeleteTrailingVectorElements(const DstOp &Res,
+                                                        const SrcOp &Op0);
+
   /// Build and insert \p Res, \p CarryOut = G_UADDO \p Op0, \p Op1
   ///
   /// G_UADDO sets \p Res to \p Op0 + \p Op1 (truncated to the bit width) and
   /// sets \p CarryOut to 1 if the result overflowed in unsigned arithmetic.
   ///
   /// \pre setBasicBlock or setMI must have been called.
   /// \pre \p Res, \p Op0 and \p Op1 must be generic virtual registers with the
   /// same scalar type.

llvm/include/llvm/CodeGen/GlobalISel/Utils.h

 
 /// Return the least common multiple type of \p OrigTy and \p TargetTy, by changing the
 /// number of vector elements or scalar bitwidth. The intent is a
 /// G_MERGE_VALUES, G_BUILD_VECTOR, or G_CONCAT_VECTORS can be constructed from
 /// \p OrigTy elements, and unmerged into \p TargetTy
 LLVM_READNONE
 LLT getLCMType(LLT OrigTy, LLT TargetTy);
 
+LLVM_READNONE
+/// Return smallest type that covers both \p OrigTy and \p TargetTy and is
+/// multiple of TargetTy.
+LLT getCoverTy(LLT OrigTy, LLT TargetTy);
+
 /// Return a type where the total size is the greatest common divisor of \p
 /// OrigTy and \p TargetTy. This will try to either change the number of vector
 /// elements, or bitwidth of scalars. The intent is the result type can be used
 /// as the result of a G_UNMERGE_VALUES from \p OrigTy, and then some
 /// combination of G_MERGE_VALUES, G_BUILD_VECTOR and G_CONCAT_VECTORS (possibly
 /// with intermediate casts) can re-form \p TargetTy.
 ///
 /// If these are vectors with different element types, this will try to produce

llvm/lib/CodeGen/GlobalISel/CallLowering.cpp

 static MachineInstrBuilder
 mergeVectorRegsToResultRegs(MachineIRBuilder &B, ArrayRef<Register> DstRegs,
                             ArrayRef<Register> SrcRegs) {
   MachineRegisterInfo &MRI = *B.getMRI();
   LLT LLTy = MRI.getType(DstRegs[0]);
   LLT PartLLT = MRI.getType(SrcRegs[0]);
 
   // Deal with v3s16 split into v2s16
-  LLT LCMTy = getLCMType(LLTy, PartLLT);
+  LLT LCMTy = getCoverTy(LLTy, PartLLT);
   if (LCMTy == LLTy) {
     // Common case where no padding is needed.
     assert(DstRegs.size() == 1);
     return B.buildConcatVectors(DstRegs[0], SrcRegs);
   }
 
   // We need to create an unmerge to the result registers, which may require
   // widening the original value.
   Register UnmergeSrcReg;
   if (LCMTy != PartLLT) {
-    // e.g. A <3 x s16> value was split to <2 x s16>
-    // %register_value0:_(<2 x s16>)
-    // %register_value1:_(<2 x s16>)
-    // %undef:_(<2 x s16>) = G_IMPLICIT_DEF
-    // %concat:_<6 x s16>) = G_CONCAT_VECTORS %reg_value0, %reg_value1, %undef
-    // %dst_reg:_(<3 x s16>), %dead:_(<3 x s16>) = G_UNMERGE_VALUES %concat
-    const int NumWide = LCMTy.getSizeInBits() / PartLLT.getSizeInBits();
-    Register Undef = B.buildUndef(PartLLT).getReg(0);
-
-    // Build vector of undefs.
-    SmallVector<Register, 8> WidenedSrcs(NumWide, Undef);
-
-    // Replace the first sources with the real registers.
-    std::copy(SrcRegs.begin(), SrcRegs.end(), WidenedSrcs.begin());
-    UnmergeSrcReg = B.buildConcatVectors(LCMTy, WidenedSrcs).getReg(0);
+    assert(DstRegs.size() == 1);
+    return B.buildDeleteTrailingVectorElements(DstRegs[0],
+                                               B.buildMerge(LCMTy, SrcRegs));
   } else {
     // We don't need to widen anything if we're extracting a scalar which was
     // promoted to a vector e.g. s8 -> v4s8 -> s8
     assert(SrcRegs.size() == 1);
     UnmergeSrcReg = SrcRegs[0];
   }
 
   int NumDst = LCMTy.getSizeInBits() / LLTy.getSizeInBits();
 
   SmallVector<Register, 8> PadDstRegs(NumDst);
   std::copy(DstRegs.begin(), DstRegs.end(), PadDstRegs.begin());
 
   // Create the excess dead defs for the unmerge.
   for (int I = DstRegs.size(); I != NumDst; ++I)
     PadDstRegs[I] = MRI.createGenericVirtualRegister(LLTy);
 
+  if (PadDstRegs.size() == 1)
+    return B.buildDeleteTrailingVectorElements(DstRegs[0], UnmergeSrcReg);
   return B.buildUnmerge(PadDstRegs, UnmergeSrcReg);
 }
 
 /// Create a sequence of instructions to combine pieces split into register
 /// typed values to the original IR value. \p OrigRegs contains the destination
 /// value registers of type \p LLTy, and \p Regs contains the legalized pieces
 /// with type \p PartLLT. This is used for incoming values (physregs to vregs).
 static void buildCopyFromRegs(MachineIRBuilder &B, ArrayRef<Register> OrigRegs,
static void buildCopyToRegs(MachineIRBuilder &B, ArrayRef<Register> DstRegs,
   if (GCDTy == PartTy) {
     // If this already evenly divisible, we can create a simple unmerge.
     B.buildUnmerge(DstRegs, SrcReg);
     return;
   }
 
   MachineRegisterInfo &MRI = *B.getMRI();
   LLT DstTy = MRI.getType(DstRegs[0]);
-  LLT LCMTy = getLCMType(SrcTy, PartTy);
+  LLT LCMTy = getCoverTy(SrcTy, PartTy);
 
   const unsigned DstSize = DstTy.getSizeInBits();
   const unsigned SrcSize = SrcTy.getSizeInBits();
   unsigned CoveringSize = LCMTy.getSizeInBits();
 
   Register UnmergeSrc = SrcReg;
 
-  if (CoveringSize != SrcSize) {
+  if (!LCMTy.isVector() && CoveringSize != SrcSize) {
     // For scalars, it's common to be able to use a simple extension.
     if (SrcTy.isScalar() && DstTy.isScalar()) {
       CoveringSize = alignTo(SrcSize, DstSize);
       LLT CoverTy = LLT::scalar(CoveringSize);
       UnmergeSrc = B.buildInstr(ExtendOp, {CoverTy}, {SrcReg}).getReg(0);
     } else {
       // Widen to the common type.
       // FIXME: This should respect the extend type
       Register Undef = B.buildUndef(SrcTy).getReg(0);
       SmallVector<Register, 8> MergeParts(1, SrcReg);
       for (unsigned Size = SrcSize; Size != CoveringSize; Size += SrcSize)
         MergeParts.push_back(Undef);
       UnmergeSrc = B.buildMerge(LCMTy, MergeParts).getReg(0);
     }
   }
 
-  // Unmerge to the original registers and pad with dead defs.
-  SmallVector<Register, 8> UnmergeResults(DstRegs.begin(), DstRegs.end());
-  for (unsigned Size = DstSize * DstRegs.size(); Size != CoveringSize;
-       Size += DstSize) {
-    UnmergeResults.push_back(MRI.createGenericVirtualRegister(DstTy));
-  }
+  if (LCMTy.isVector() && CoveringSize != SrcSize)
+    UnmergeSrc = B.buildPadVectorWithUndefElements(LCMTy, SrcReg).getReg(0);
 
-  B.buildUnmerge(UnmergeResults, UnmergeSrc);
+  B.buildUnmerge(DstRegs, UnmergeSrc);
 }
 
 bool CallLowering::determineAndHandleAssignments(
     ValueHandler &Handler, ValueAssigner &Assigner,
     SmallVectorImpl<ArgInfo> &Args, MachineIRBuilder &MIRBuilder,
     CallingConv::ID CallConv, bool IsVarArg, Register ThisReturnReg) const {
   MachineFunction &MF = MIRBuilder.getMF();
   const Function &F = MF.getFunction();

llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp

   // Use an unmerge when possible.
   if (LeftoverSize == 0) {
     for (unsigned I = 0; I < NumParts; ++I)
       VRegs.push_back(MRI.createGenericVirtualRegister(MainTy));
     MIRBuilder.buildUnmerge(VRegs, Reg);
     return true;
   }
 
+  // Perform irregular split. Leftover is last element of RegPieces.
   if (MainTy.isVector()) {
-    unsigned EltSize = MainTy.getScalarSizeInBits();
-    if (LeftoverSize % EltSize != 0)
-      return false;
-    LeftoverTy = LLT::scalarOrVector(
-        ElementCount::getFixed(LeftoverSize / EltSize), EltSize);
-  } else {
-    LeftoverTy = LLT::scalar(LeftoverSize);
+    SmallVector<Register, 4> RegPieces;
+    extractVectorParts(Reg, MainTy.getNumElements(), RegPieces);
+    for (unsigned i = 0; i < RegPieces.size() - 1; ++i)
+      VRegs.push_back(RegPieces[i]);
+    LeftoverRegs.push_back(RegPieces[RegPieces.size() - 1]);
+    LeftoverTy = MRI.getType(LeftoverRegs[0]);
+    return true;
   }
 
+  LeftoverTy = LLT::scalar(LeftoverSize);
   // For irregular sizes, extract the individual parts.
   for (unsigned I = 0; I != NumParts; ++I) {
     Register NewReg = MRI.createGenericVirtualRegister(MainTy);
     VRegs.push_back(NewReg);
     MIRBuilder.buildExtract(NewReg, Reg, MainSize * I);
   }
 
   for (unsigned Offset = MainSize * NumParts; Offset < RegSize;
        Offset += LeftoverSize) {
     Register NewReg = MRI.createGenericVirtualRegister(LeftoverTy);
     LeftoverRegs.push_back(NewReg);
     MIRBuilder.buildExtract(NewReg, Reg, Offset);
   }
 
   return true;
 }
 
+void LegalizerHelper::extractVectorParts(Register Reg, unsigned NumElts,
+                                         SmallVectorImpl<Register> &VRegs) {
+  LLT RegTy = MRI.getType(Reg);
+  assert(RegTy.isVector() && "Expected a vector type");
+
+  LLT EltTy = RegTy.getElementType();
+  LLT NarrowTy = (NumElts == 1) ? EltTy : LLT::fixed_vector(NumElts, EltTy);
+  unsigned RegNumElts = RegTy.getNumElements();
+  unsigned LeftoverNumElts = RegNumElts % NumElts;
+  unsigned NumNarrowTyPieces = RegNumElts / NumElts;
+
+  // Perfect split without leftover
+  if (LeftoverNumElts == 0)
+    return extractParts(Reg, NarrowTy, NumNarrowTyPieces, VRegs);
+
+  // Irregular split. Provide direct access to all elements for artifact
+  // combiner using unmerge to elements. Then build vectors with NumElts
+  // elements. Remaining element(s) will be (used to build vector) Leftover.
+  SmallVector<Register, 8> Elts;
+  extractParts(Reg, EltTy, RegNumElts, Elts);
+
+  unsigned Offset = 0;
+  // Requiested sub-vectors of NarrowTy.
+  for (unsigned i = 0; i < NumNarrowTyPieces; ++i, Offset += NumElts) {
+    ArrayRef<Register> Pieces(&Elts[Offset], NumElts);
+    VRegs.push_back(MIRBuilder.buildMerge(NarrowTy, Pieces).getReg(0));
+  }
+
+  // Leftover element(s).
+  if (LeftoverNumElts == 1) {
+    VRegs.push_back(Elts[Offset]);
+  } else {
+    LLT LeftoverTy = LLT::fixed_vector(LeftoverNumElts, EltTy);
+    ArrayRef<Register> Pieces(&Elts[Offset], LeftoverNumElts);
+    VRegs.push_back(MIRBuilder.buildMerge(LeftoverTy, Pieces).getReg(0));
+  }
+}
+
 void LegalizerHelper::insertParts(Register DstReg,
                                   LLT ResultTy, LLT PartTy,
                                   ArrayRef<Register> PartRegs,
                                   LLT LeftoverTy,
                                   ArrayRef<Register> LeftoverRegs) {
   if (!LeftoverTy.isValid()) {
     assert(LeftoverRegs.empty());
 
     if (!ResultTy.isVector()) {
       MIRBuilder.buildMerge(DstReg, PartRegs);
       return;
     }
 
     if (PartTy.isVector())
       MIRBuilder.buildConcatVectors(DstReg, PartRegs);
     else
       MIRBuilder.buildBuildVector(DstReg, PartRegs);
     return;
   }
 
+  // Merge sub-vectors with different number of elements and insert into DstReg.
+  if (ResultTy.isVector()) {
+    assert(LeftoverRegs.size() == 1 && "Expected one leftover register");
+    SmallVector<Register, 4> AllRegs;
+    for (auto Reg : concat<const Register>(PartRegs, LeftoverRegs))
+      AllRegs.push_back(Reg);
+    return mergeMixedSubvectors(DstReg, AllRegs);
+  }
+
   SmallVector<Register> GCDRegs;
   LLT GCDTy = getGCDType(getGCDType(ResultTy, LeftoverTy), PartTy);
   for (auto PartReg : concat<const Register>(PartRegs, LeftoverRegs))
     extractGCDType(GCDRegs, GCDTy, PartReg);
   LLT ResultLCMTy = buildLCMMergePieces(ResultTy, LeftoverTy, GCDTy, GCDRegs);
   buildWidenedRemergeToDst(DstReg, ResultLCMTy, GCDRegs);
 }
 
+void LegalizerHelper::appendVectorElts(SmallVectorImpl<Register> &Elts,
+                                       Register Reg) {
+  LLT Ty = MRI.getType(Reg);
+  SmallVector<Register, 4> RegElts;
+  extractParts(Reg, Ty.getScalarType(), Ty.getNumElements(), RegElts);
+  Elts.append(RegElts);
+}
+
+/// Merge \p PartRegs with different types into \p DstReg.
+void LegalizerHelper::mergeMixedSubvectors(Register DstReg,
+                                           ArrayRef<Register> PartRegs) {
+  SmallVector<Register, 8> AllElts;
+  for (unsigned i = 0; i < PartRegs.size() - 1; ++i)
+    appendVectorElts(AllElts, PartRegs[i]);
+
+  Register Leftover = PartRegs[PartRegs.size() - 1];
+  if (MRI.getType(Leftover).isScalar())
+    AllElts.push_back(Leftover);
+  else
+    appendVectorElts(AllElts, Leftover);
+
+  MIRBuilder.buildMerge(DstReg, AllElts);
+  return;
+}
+
 /// Append the result registers of G_UNMERGE_VALUES \p MI to \p Regs.
 static void getUnmergeResults(SmallVectorImpl<Register> &Regs,
                               const MachineInstr &MI) {
   assert(MI.getOpcode() == TargetOpcode::G_UNMERGE_VALUES);
 
   const int StartIdx = Regs.size();
   const int NumResults = MI.getNumOperands() - 1;
   Regs.resize(Regs.size() + NumResults);

     }
 
     auto Unmerge = MIRBuilder.buildUnmerge(NarrowTy, MI.getOperand(1));
     MIRBuilder.buildCopy(MI.getOperand(0), Unmerge.getReg(0));
     MI.eraseFromParent();
     return Legalized;
   }
 
-  case TargetOpcode::G_FREEZE:
-    return reduceOperationWidth(MI, TypeIdx, NarrowTy);
+  case TargetOpcode::G_FREEZE: {
+    if (TypeIdx != 0)
+      return UnableToLegalize;
+
+    LLT Ty = MRI.getType(MI.getOperand(0).getReg());
+    // Should widen scalar first
+    if (Ty.getSizeInBits() % NarrowTy.getSizeInBits() != 0)
+      return UnableToLegalize;
+
+    auto Unmerge = MIRBuilder.buildUnmerge(NarrowTy, MI.getOperand(1).getReg());
+    SmallVector<Register, 4> Parts;
+    for (unsigned i = 0; i < Unmerge->getNumDefs(); ++i) {
+      Parts.push_back(
+          MIRBuilder.buildFreeze(NarrowTy, Unmerge.getReg(i)).getReg(0));
+    }
+
+    MIRBuilder.buildMerge(MI.getOperand(0).getReg(), Parts);
+    MI.eraseFromParent();
+    return Legalized;
+  }
   case TargetOpcode::G_ADD:
   case TargetOpcode::G_SUB:
   case TargetOpcode::G_SADDO:
   case TargetOpcode::G_SSUBO:
   case TargetOpcode::G_SADDE:
   case TargetOpcode::G_SSUBE:
   case TargetOpcode::G_UADDO:
   case TargetOpcode::G_USUBO:

   MIRBuilder.buildInstr(ExtOpcode, {MO}, {DstTrunc});
   MO.setReg(DstTrunc);
 }
 
 void LegalizerHelper::moreElementsVectorDst(MachineInstr &MI, LLT WideTy,
                                             unsigned OpIdx) {
   MachineOperand &MO = MI.getOperand(OpIdx);
   MIRBuilder.setInsertPt(MIRBuilder.getMBB(), ++MIRBuilder.getInsertPt());
-  MO.setReg(widenWithUnmerge(WideTy, MO.getReg()));
+  Register Dst = MO.getReg();
+  Register DstExt = MRI.createGenericVirtualRegister(WideTy);
+  MO.setReg(DstExt);
+  MIRBuilder.buildDeleteTrailingVectorElements(Dst, DstExt);
 }
 
 void LegalizerHelper::moreElementsVectorSrc(MachineInstr &MI, LLT MoreTy,
                                             unsigned OpIdx) {
   MachineOperand &MO = MI.getOperand(OpIdx);
-
-  LLT OldTy = MRI.getType(MO.getReg());
-  unsigned OldElts = OldTy.getNumElements();
-  unsigned NewElts = MoreTy.getNumElements();
-
-  unsigned NumParts = NewElts / OldElts;
-
-  // Use concat_vectors if the result is a multiple of the number of elements.
-  if (NumParts * OldElts == NewElts) {
-    SmallVector<Register, 8> Parts;
-    Parts.push_back(MO.getReg());
-
-    Register ImpDef = MIRBuilder.buildUndef(OldTy).getReg(0);
-    for (unsigned I = 1; I != NumParts; ++I)
-      Parts.push_back(ImpDef);
-
-    auto Concat = MIRBuilder.buildConcatVectors(MoreTy, Parts);
-    MO.setReg(Concat.getReg(0));
-    return;
-  }
-
-  Register MoreReg = MRI.createGenericVirtualRegister(MoreTy);
-  Register ImpDef = MIRBuilder.buildUndef(MoreTy).getReg(0);
-  MIRBuilder.buildInsert(MoreReg, ImpDef, MO.getReg(), 0);
-  MO.setReg(MoreReg);
+  SmallVector<Register, 4> Regs;
+  MO.setReg(MIRBuilder.buildPadVectorWithUndefElements(MoreTy, MO).getReg(0));
 }
 
 void LegalizerHelper::bitcastSrc(MachineInstr &MI, LLT CastTy, unsigned OpIdx) {
   MachineOperand &Op = MI.getOperand(OpIdx);
   Op.setReg(MIRBuilder.buildBitcast(CastTy, Op).getReg(0));
 }
 
 void LegalizerHelper::bitcastDst(MachineInstr &MI, LLT CastTy, unsigned OpIdx) {

   LLT IdxTy = MRI.getType(Index);
   auto Mul = MIRBuilder.buildMul(IdxTy, Index,
                                  MIRBuilder.buildConstant(IdxTy, EltSize));
 
   LLT PtrTy = MRI.getType(VecPtr);
   return MIRBuilder.buildPtrAdd(PtrTy, VecPtr, Mul).getReg(0);
 }
 
-LegalizerHelper::LegalizeResult LegalizerHelper::fewerElementsVectorImplicitDef(
-    MachineInstr &MI, unsigned TypeIdx, LLT NarrowTy) {
-  Register DstReg = MI.getOperand(0).getReg();
-  LLT DstTy = MRI.getType(DstReg);
-  LLT LCMTy = getLCMType(DstTy, NarrowTy);
+/// Check that all vector operands have same number of elements. Other operands
+/// should be listed in NonVecOp.
+static bool hasSameNumEltsOnAllVectorOperands(
+    GenericMachineInstr &MI, MachineRegisterInfo &MRI,
+    std::initializer_list<unsigned> NonVecOpIndices) {
+  if (MI.getNumMemOperands() != 0)
+    return false;
 
-  unsigned NumParts = LCMTy.getSizeInBits() / NarrowTy.getSizeInBits();
+  LLT VecTy = MRI.getType(MI.getReg(0));
+  if (!VecTy.isVector())
+    return false;
+  unsigned NumElts = VecTy.getNumElements();
 
-  auto NewUndef = MIRBuilder.buildUndef(NarrowTy);
-  SmallVector<Register, 8> Parts(NumParts, NewUndef.getReg(0));
+  for (unsigned OpIdx = 1; OpIdx < MI.getNumOperands(); ++OpIdx) {
+    MachineOperand &Op = MI.getOperand(OpIdx);
+    if (!Op.isReg()) {
+      if (!is_contained(NonVecOpIndices, OpIdx))
+        return false;
+      continue;
+    }
 
-  buildWidenedRemergeToDst(DstReg, LCMTy, Parts);
-  MI.eraseFromParent();
-  return Legalized;
+    LLT Ty = MRI.getType(Op.getReg());
+    if (!Ty.isVector()) {
+      if (!is_contained(NonVecOpIndices, OpIdx))
+        return false;
+      is_contained(NonVecOpIndices, OpIdx);
+      continue;
+    }
+
+    if (Ty.getNumElements() != NumElts)
+      return false;
+  }
+
+  return true;
+}
+
+/// Fill \p DstOps with DstOps that have same number of elements combined as
+/// the Ty. These DstOps have either scalar type when \p NumElts = 1 or are
+/// vectors with \p NumElts elements. When Ty.getNumElements() is not multiple
+/// of \p NumElts last DstOp (leftover) has fewer then \p NumElts elements.
+static void makeDstOps(SmallVectorImpl<DstOp> &DstOps, LLT Ty,
+                       unsigned NumElts) {
+  LLT LeftoverTy;
+  assert(Ty.isVector() && "Expected vector type");
+  LLT EltTy = Ty.getElementType();
+  LLT NarrowTy = (NumElts == 1) ? EltTy : LLT::fixed_vector(NumElts, EltTy);
+  int NumParts, NumLeftover;
+  std::tie(NumParts, NumLeftover) =
+      getNarrowTypeBreakDown(Ty, NarrowTy, LeftoverTy);
+
+  assert(NumParts > 0 && "Error in getNarrowTypeBreakDown");
+  for (int i = 0; i < NumParts; ++i) {
+    DstOps.push_back(NarrowTy);
+  }
+
+  if (LeftoverTy.isValid()) {
+    assert(NumLeftover == 1 && "expected exactly one leftover");
+    DstOps.push_back(LeftoverTy);
+  }
+}
+
+/// Operand \p Op is used on \p N sub-instructions. Fill \p Ops with \p N SrcOps
+/// made from \p Op deppending on operand type.
+static void broadcastSrcOp(SmallVectorImpl<SrcOp> &Ops, unsigned N,
+                           MachineOperand &Op) {
+  assert((Op.isReg() || Op.isImm() || Op.isPredicate()) && "Unsupported type");
+  for (unsigned i = 0; i < N; ++i) {
+    if (Op.isReg())
+      Ops.push_back(Op.getReg());
+    if (Op.isImm())
+      Ops.push_back(Op.getImm());
+    if (Op.isPredicate())
+      Ops.push_back(static_cast<CmpInst::Predicate>(Op.getPredicate()));
+  }
 }
 
 // Handle splitting vector operations which need to have the same number of
 // elements in each type index, but each type index may have a different element
 // type.
 //
 // e.g.  <4 x s64> = G_SHL <4 x s64>, <4 x s32> ->
 //       <2 x s64> = G_SHL <2 x s64>, <2 x s32>
 //       <2 x s64> = G_SHL <2 x s64>, <2 x s32>
 //
 // Also handles some irregular breakdown cases, e.g.
 // e.g.  <3 x s64> = G_SHL <3 x s64>, <3 x s32> ->
 //       <2 x s64> = G_SHL <2 x s64>, <2 x s32>
 //             s64 = G_SHL s64, s32
 LegalizerHelper::LegalizeResult
 LegalizerHelper::fewerElementsVectorMultiEltType(
-  MachineInstr &MI, unsigned TypeIdx, LLT NarrowTyArg) {
-  if (TypeIdx != 0)
-    return UnableToLegalize;
-
-  const LLT NarrowTy0 = NarrowTyArg;
-  const Register DstReg = MI.getOperand(0).getReg();
-  LLT DstTy = MRI.getType(DstReg);
-  LLT LeftoverTy0;
-
-  // All of the operands need to have the same number of elements, so if we can
-  // determine a type breakdown for the result type, we can for all of the
-  // source types.
-  int NumParts = getNarrowTypeBreakDown(DstTy, NarrowTy0, LeftoverTy0).first;
-  if (NumParts < 0)
-    return UnableToLegalize;
-
-  SmallVector<MachineInstrBuilder, 4> NewInsts;
-
-  SmallVector<Register, 4> DstRegs, LeftoverDstRegs;
-  SmallVector<Register, 4> PartRegs, LeftoverRegs;
-
-  for (unsigned I = 1, E = MI.getNumOperands(); I != E; ++I) {
-    Register SrcReg = MI.getOperand(I).getReg();
-    LLT SrcTyI = MRI.getType(SrcReg);
-    const auto NewEC = NarrowTy0.isVector() ? NarrowTy0.getElementCount()
-                                            : ElementCount::getFixed(1);
-    LLT NarrowTyI = LLT::scalarOrVector(NewEC, SrcTyI.getScalarType());
-    LLT LeftoverTyI;
-
-    // Split this operand into the requested typed registers, and any leftover
-    // required to reproduce the original type.
-    if (!extractParts(SrcReg, SrcTyI, NarrowTyI, LeftoverTyI, PartRegs,
-                      LeftoverRegs))
-      return UnableToLegalize;
-
-    if (I == 1) {
-      // For the first operand, create an instruction for each part and setup
-      // the result.
-      for (Register PartReg : PartRegs) {
-        Register PartDstReg = MRI.createGenericVirtualRegister(NarrowTy0);
-        NewInsts.push_back(MIRBuilder.buildInstrNoInsert(MI.getOpcode())
-                               .addDef(PartDstReg)
-                               .addUse(PartReg));
-        DstRegs.push_back(PartDstReg);
-      }
-
-      for (Register LeftoverReg : LeftoverRegs) {
-        Register PartDstReg = MRI.createGenericVirtualRegister(LeftoverTy0);
-        NewInsts.push_back(MIRBuilder.buildInstrNoInsert(MI.getOpcode())
-                               .addDef(PartDstReg)
-                               .addUse(LeftoverReg));
-        LeftoverDstRegs.push_back(PartDstReg);
-      }
+    GenericMachineInstr &MI, unsigned NumElts,
+    std::initializer_list<unsigned> NonVecOpIndices) {
+  assert(hasSameNumEltsOnAllVectorOperands(MI, MRI, NonVecOpIndices) &&
+         "Non-compatible opcode or not specified non-vector operands");
+  unsigned OrigNumElts = MRI.getType(MI.getReg(0)).getNumElements();
+
+  unsigned NumInputs = MI.getNumOperands() - MI.getNumDefs();
+  unsigned NumDefs = MI.getNumDefs();
+
+  // Create DstOps (NumElts sub-vectors + Leftover) for each output.
+  // Build instructions with DstOps to use instruction found by CSE directly.
+  // CSE copies found instruction into given vreg when building with vreg dest.
+  SmallVector<SmallVector<DstOp, 4>, 2> OutputOpsPieces(NumDefs);
+  // Output registers will be taken from created instructions.
+  SmallVector<SmallVector<Register, 4>, 2> OutputRegs(NumDefs);
+  for (unsigned i = 0; i < NumDefs; ++i) {
+    makeDstOps(OutputOpsPieces[i], MRI.getType(MI.getReg(i)), NumElts);
+  }
+
+  // Split vector input operands into NumElts sub-vectors + Leftover. Other
+  // operands (listed in NonVecOpIndices) will be used as is witout splitting;
+  // examples: compare predicate in icmp and fcmp (op 1), vector select with i1
+  // scalar condition (op 1), immediate in sext_inreg (op 2).
+  SmallVector<SmallVector<SrcOp, 4>, 3> InputOpsPieces(NumInputs);
+  for (unsigned UseIdx = NumDefs, UseNo = 0; UseIdx < MI.getNumOperands();
+       ++UseIdx, ++UseNo) {
+    if (is_contained(NonVecOpIndices, UseIdx)) {
+      broadcastSrcOp(InputOpsPieces[UseNo], OutputOpsPieces[0].size(),
+                     MI.getOperand(UseIdx));
     } else {
-      assert(NewInsts.size() == PartRegs.size() + LeftoverRegs.size());
-
-      // Add the newly created operand splits to the existing instructions. The
-      // odd-sized pieces are ordered after the requested NarrowTyArg sized
-      // pieces.
-      unsigned InstCount = 0;
-      for (unsigned J = 0, JE = PartRegs.size(); J != JE; ++J)
-        NewInsts[InstCount++].addUse(PartRegs[J]);
-      for (unsigned J = 0, JE = LeftoverRegs.size(); J != JE; ++J)
-        NewInsts[InstCount++].addUse(LeftoverRegs[J]);
+      SmallVector<Register, 4> SplitPieces;
+      extractVectorParts(MI.getReg(UseIdx), NumElts, SplitPieces);
+      for (auto Reg : SplitPieces)
+        InputOpsPieces[UseNo].push_back(Reg);
     }
-
-    PartRegs.clear();
-    LeftoverRegs.clear();
   }
 
-  // Insert the newly built operations and rebuild the result register.
-  for (auto &MIB : NewInsts)
-    MIRBuilder.insertInstr(MIB);
-
-  insertParts(DstReg, DstTy, NarrowTy0, DstRegs, LeftoverTy0, LeftoverDstRegs);
-
-  MI.eraseFromParent();
-  return Legalized;
-}
-
-LegalizerHelper::LegalizeResult
-LegalizerHelper::fewerElementsVectorCasts(MachineInstr &MI, unsigned TypeIdx,
-                                          LLT NarrowTy) {
-  if (TypeIdx != 0)
-    return UnableToLegalize;
+  unsigned NumLeftovers = OrigNumElts % NumElts ? 1 : 0;
 
-  Register DstReg = MI.getOperand(0).getReg();
-  Register SrcReg = MI.getOperand(1).getReg();
-  LLT DstTy = MRI.getType(DstReg);
-  LLT SrcTy = MRI.getType(SrcReg);
+  // Take i-th piece of each input operand split and build sub-vector/scalar
+  // instruction. Set i-th DstOp(s) from OutputOpsPieces as destination(s).
+  for (unsigned i = 0; i < OrigNumElts / NumElts + NumLeftovers; ++i) {
+    SmallVector<DstOp, 2> Defs;
+    for (unsigned DstNo = 0; DstNo < NumDefs; ++DstNo)
+      Defs.push_back(OutputOpsPieces[DstNo][i]);
 
-  LLT NarrowTy0 = NarrowTy;
-  LLT NarrowTy1;
-  unsigned NumParts;
-
-  if (NarrowTy.isVector()) {
-    // Uneven breakdown not handled.
-    NumParts = DstTy.getNumElements() / NarrowTy.getNumElements();
-    if (NumParts * NarrowTy.getNumElements() != DstTy.getNumElements())
-      return UnableToLegalize;
+    SmallVector<SrcOp, 3> Uses;
+    for (unsigned InputNo = 0; InputNo < NumInputs; ++InputNo)
+      Uses.push_back(InputOpsPieces[InputNo][i]);
 
-    NarrowTy1 = LLT::vector(NarrowTy.getElementCount(), SrcTy.getElementType());
-  } else {
-    NumParts = DstTy.getNumElements();
-    NarrowTy1 = SrcTy.getElementType();
+    auto I = MIRBuilder.buildInstr(MI.getOpcode(), Defs, Uses, MI.getFlags());
+    for (unsigned DstNo = 0; DstNo < NumDefs; ++DstNo)
+      OutputRegs[DstNo].push_back(I.getReg(DstNo));
   }
 
-  SmallVector<Register, 4> SrcRegs, DstRegs;
-  extractParts(SrcReg, NarrowTy1, NumParts, SrcRegs);
-
-  for (unsigned I = 0; I < NumParts; ++I) {
-    Register DstReg = MRI.createGenericVirtualRegister(NarrowTy0);
-    MachineInstr *NewInst =
-        MIRBuilder.buildInstr(MI.getOpcode(), {DstReg}, {SrcRegs[I]});
-
-    NewInst->setFlags(MI.getFlags());
-    DstRegs.push_back(DstReg);
+  // Merge small outputs into MI's output for each def operand.
+  if (NumLeftovers) {
+    for (unsigned i = 0; i < NumDefs; ++i)
+      mergeMixedSubvectors(MI.getReg(i), OutputRegs[i]);
+  } else {
+    for (unsigned i = 0; i < NumDefs; ++i)
+      MIRBuilder.buildMerge(MI.getReg(i), OutputRegs[i]);
   }
 
-  if (NarrowTy.isVector())
-    MIRBuilder.buildConcatVectors(DstReg, DstRegs);
-  else
-    MIRBuilder.buildBuildVector(DstReg, DstRegs);
-
   MI.eraseFromParent();
   return Legalized;
 }
 
 LegalizerHelper::LegalizeResult
-LegalizerHelper::fewerElementsVectorCmp(MachineInstr &MI, unsigned TypeIdx,
-                                        LLT NarrowTy) {
-  Register DstReg = MI.getOperand(0).getReg();
-  Register Src0Reg = MI.getOperand(2).getReg();
-  LLT DstTy = MRI.getType(DstReg);
-  LLT SrcTy = MRI.getType(Src0Reg);
-
-  unsigned NumParts;
-  LLT NarrowTy0, NarrowTy1;
-
-  if (TypeIdx == 0) {
-    unsigned NewElts = NarrowTy.isVector() ? NarrowTy.getNumElements() : 1;
-    unsigned OldElts = DstTy.getNumElements();
-
-    NarrowTy0 = NarrowTy;
-    NumParts = NarrowTy.isVector() ? (OldElts / NewElts) : DstTy.getNumElements();
-    NarrowTy1 = NarrowTy.isVector() ? LLT::vector(NarrowTy.getElementCount(),
-                                                  SrcTy.getScalarSizeInBits())
-                                    : SrcTy.getElementType();
-
-  } else {
-    unsigned NewElts = NarrowTy.isVector() ? NarrowTy.getNumElements() : 1;
-    unsigned OldElts = SrcTy.getNumElements();
-
-    NumParts = NarrowTy.isVector() ? (OldElts / NewElts) :
-      NarrowTy.getNumElements();
-    NarrowTy0 =
-        LLT::vector(NarrowTy.getElementCount(), DstTy.getScalarSizeInBits());
-    NarrowTy1 = NarrowTy;
+LegalizerHelper::fewerElementsVectorPhi(GenericMachineInstr &MI,
+                                        unsigned NumElts) {
+  unsigned OrigNumElts = MRI.getType(MI.getReg(0)).getNumElements();
+
+  unsigned NumInputs = MI.getNumOperands() - MI.getNumDefs();
+  unsigned NumDefs = MI.getNumDefs();
+
+  SmallVector<DstOp, 4> OutputOpsPieces;
+  SmallVector<Register, 4> OutputRegs;
+  makeDstOps(OutputOpsPieces, MRI.getType(MI.getReg(0)), NumElts);
+
+  // Instructions that perform register split will be inserted in basic block
+  // where register is defined (basic block is in the next operand).
+  SmallVector<SmallVector<Register, 4>, 3> InputOpsPieces(NumInputs / 2);
+  for (unsigned UseIdx = NumDefs, UseNo = 0; UseIdx < MI.getNumOperands();
+       UseIdx += 2, ++UseNo) {
+    MachineBasicBlock &OpMBB = *MI.getOperand(UseIdx + 1).getMBB();
+    MIRBuilder.setInsertPt(OpMBB, OpMBB.getFirstTerminator());
+    extractVectorParts(MI.getReg(UseIdx), NumElts, InputOpsPieces[UseNo]);
   }
 
-  // FIXME: Don't know how to handle the situation where the small vectors
-  // aren't all the same size yet.
-  if (NarrowTy1.isVector() &&
-      NarrowTy1.getNumElements() * NumParts != DstTy.getNumElements())
-    return UnableToLegalize;
-
-  CmpInst::Predicate Pred
-    = static_cast<CmpInst::Predicate>(MI.getOperand(1).getPredicate());
-
-  SmallVector<Register, 2> Src1Regs, Src2Regs, DstRegs;
-  extractParts(MI.getOperand(2).getReg(), NarrowTy1, NumParts, Src1Regs);
-  extractParts(MI.getOperand(3).getReg(), NarrowTy1, NumParts, Src2Regs);
-
-  for (unsigned I = 0; I < NumParts; ++I) {
-    Register DstReg = MRI.createGenericVirtualRegister(NarrowTy0);
-    DstRegs.push_back(DstReg);
+  // Build PHIs with fewer elements.
+  unsigned NumLeftovers = OrigNumElts % NumElts ? 1 : 0;
+  MIRBuilder.setInsertPt(*MI.getParent(), MI);
+  for (unsigned i = 0; i < OrigNumElts / NumElts + NumLeftovers; ++i) {
+    auto Phi = MIRBuilder.buildInstr(TargetOpcode::G_PHI);
+    Phi.addDef(
+        MRI.createGenericVirtualRegister(OutputOpsPieces[i].getLLTTy(MRI)));
+    OutputRegs.push_back(Phi.getReg(0));
 
-    if (MI.getOpcode() == TargetOpcode::G_ICMP)
-      MIRBuilder.buildICmp(Pred, DstReg, Src1Regs[I], Src2Regs[I]);
-    else {
-      MachineInstr *NewCmp
-        = MIRBuilder.buildFCmp(Pred, DstReg, Src1Regs[I], Src2Regs[I]);
-      NewCmp->setFlags(MI.getFlags());
+    for (unsigned j = 0; j < NumInputs / 2; ++j) {
+      Phi.addUse(InputOpsPieces[j][i]);
+      Phi.add(MI.getOperand(1 + j * 2 + 1));
     }
   }
 
-  if (NarrowTy1.isVector())
-    MIRBuilder.buildConcatVectors(DstReg, DstRegs);
-  else
-    MIRBuilder.buildBuildVector(DstReg, DstRegs);
-
-  MI.eraseFromParent();
-  return Legalized;
-}
-
-LegalizerHelper::LegalizeResult
-LegalizerHelper::fewerElementsVectorSelect(MachineInstr &MI, unsigned TypeIdx,
-                                           LLT NarrowTy) {
-  Register DstReg = MI.getOperand(0).getReg();
-  Register CondReg = MI.getOperand(1).getReg();
-
-  unsigned NumParts = 0;
-  LLT NarrowTy0, NarrowTy1;
-
-  LLT DstTy = MRI.getType(DstReg);
-  LLT CondTy = MRI.getType(CondReg);
-  unsigned Size = DstTy.getSizeInBits();
-
-  assert(TypeIdx == 0 || CondTy.isVector());
-
-  if (TypeIdx == 0) {
-    NarrowTy0 = NarrowTy;
-    NarrowTy1 = CondTy;
-
-    unsigned NarrowSize = NarrowTy0.getSizeInBits();
-    // FIXME: Don't know how to handle the situation where the small vectors
-    // aren't all the same size yet.
-    if (Size % NarrowSize != 0)
-      return UnableToLegalize;
-
-    NumParts = Size / NarrowSize;
-
-    // Need to break down the condition type
-    if (CondTy.isVector()) {
-      if (CondTy.getNumElements() == NumParts)
-        NarrowTy1 = CondTy.getElementType();
-      else
-        NarrowTy1 =
-            LLT::vector(CondTy.getElementCount().divideCoefficientBy(NumParts),
-                        CondTy.getScalarSizeInBits());
-    }
+  // Merge small outputs into MI's def.
+  if (NumLeftovers) {
+    mergeMixedSubvectors(MI.getReg(0), OutputRegs);
   } else {
-    NumParts = CondTy.getNumElements();
-    if (NarrowTy.isVector()) {
-      // TODO: Handle uneven breakdown.
-      if (NumParts * NarrowTy.getNumElements() != CondTy.getNumElements())
-        return UnableToLegalize;
-
-      return UnableToLegalize;
-    } else {
-      NarrowTy0 = DstTy.getElementType();
-      NarrowTy1 = NarrowTy;
-    }
-  }
-
-  SmallVector<Register, 2> DstRegs, Src0Regs, Src1Regs, Src2Regs;
-  if (CondTy.isVector())
-    extractParts(MI.getOperand(1).getReg(), NarrowTy1, NumParts, Src0Regs);
-
-  extractParts(MI.getOperand(2).getReg(), NarrowTy0, NumParts, Src1Regs);
-  extractParts(MI.getOperand(3).getReg(), NarrowTy0, NumParts, Src2Regs);
-
-  for (unsigned i = 0; i < NumParts; ++i) {
-    Register DstReg = MRI.createGenericVirtualRegister(NarrowTy0);
-    MIRBuilder.buildSelect(DstReg, CondTy.isVector() ? Src0Regs[i] : CondReg,
-                           Src1Regs[i], Src2Regs[i]);
-    DstRegs.push_back(DstReg);
-  }
-
-  if (NarrowTy0.isVector())
-    MIRBuilder.buildConcatVectors(DstReg, DstRegs);
-  else
-    MIRBuilder.buildBuildVector(DstReg, DstRegs);
-
-  MI.eraseFromParent();
-  return Legalized;
-}
-
-LegalizerHelper::LegalizeResult
-LegalizerHelper::fewerElementsVectorPhi(MachineInstr &MI, unsigned TypeIdx,
-                                        LLT NarrowTy) {
-  const Register DstReg = MI.getOperand(0).getReg();
-  LLT PhiTy = MRI.getType(DstReg);
-  LLT LeftoverTy;
-
-  // All of the operands need to have the same number of elements, so if we can
-  // determine a type breakdown for the result type, we can for all of the
-  // source types.
-  int NumParts, NumLeftover;
-  std::tie(NumParts, NumLeftover)
-    = getNarrowTypeBreakDown(PhiTy, NarrowTy, LeftoverTy);
-  if (NumParts < 0)
-    return UnableToLegalize;
-
-  SmallVector<Register, 4> DstRegs, LeftoverDstRegs;
-  SmallVector<MachineInstrBuilder, 4> NewInsts;
-
-  const int TotalNumParts = NumParts + NumLeftover;
-
-  // Insert the new phis in the result block first.
-  for (int I = 0; I != TotalNumParts; ++I) {
-    LLT Ty = I < NumParts ? NarrowTy : LeftoverTy;
-    Register PartDstReg = MRI.createGenericVirtualRegister(Ty);
-    NewInsts.push_back(MIRBuilder.buildInstr(TargetOpcode::G_PHI)
-                       .addDef(PartDstReg));
-    if (I < NumParts)
-      DstRegs.push_back(PartDstReg);
-    else
-      LeftoverDstRegs.push_back(PartDstReg);
-  }
-
-  MachineBasicBlock *MBB = MI.getParent();
-  MIRBuilder.setInsertPt(*MBB, MBB->getFirstNonPHI());
-  insertParts(DstReg, PhiTy, NarrowTy, DstRegs, LeftoverTy, LeftoverDstRegs);
-
-  SmallVector<Register, 4> PartRegs, LeftoverRegs;
-
-  // Insert code to extract the incoming values in each predecessor block.
-  for (unsigned I = 1, E = MI.getNumOperands(); I != E; I += 2) {
-    PartRegs.clear();
-    LeftoverRegs.clear();
-
-    Register SrcReg = MI.getOperand(I).getReg();
-    MachineBasicBlock &OpMBB = *MI.getOperand(I + 1).getMBB();
-    MIRBuilder.setInsertPt(OpMBB, OpMBB.getFirstTerminator());
-
-    LLT Unused;
-    if (!extractParts(SrcReg, PhiTy, NarrowTy, Unused, PartRegs,
-                      LeftoverRegs))
-      return UnableToLegalize;
-
-    // Add the newly created operand splits to the existing instructions. The
-    // odd-sized pieces are ordered after the requested NarrowTyArg sized
-    // pieces.
-    for (int J = 0; J != TotalNumParts; ++J) {
-      MachineInstrBuilder MIB = NewInsts[J];
-      MIB.addUse(J < NumParts ? PartRegs[J] : LeftoverRegs[J - NumParts]);
-      MIB.addMBB(&OpMBB);
-    }
+    MIRBuilder.buildMerge(MI.getReg(0), OutputRegs);
   }
 
   MI.eraseFromParent();
   return Legalized;
 }
 
 LegalizerHelper::LegalizeResult
 LegalizerHelper::fewerElementsVectorUnmergeValues(MachineInstr &MI,
                                                   unsigned TypeIdx,
                                                   LLT NarrowTy) {
-  if (TypeIdx != 1)
-    return UnableToLegalize;
-
   const int NumDst = MI.getNumOperands() - 1;
   const Register SrcReg = MI.getOperand(NumDst).getReg();
-  LLT SrcTy = MRI.getType(SrcReg);
-
   LLT DstTy = MRI.getType(MI.getOperand(0).getReg());
+  LLT SrcTy = MRI.getType(SrcReg);
 
-  // TODO: Create sequence of extracts.
-  if (DstTy == NarrowTy)
+  if (TypeIdx != 1 || NarrowTy == DstTy)
     return UnableToLegalize;
 
-  LLT GCDTy = getGCDType(SrcTy, NarrowTy);
-  if (DstTy == GCDTy) {
-    // This would just be a copy of the same unmerge.
-    // TODO: Create extracts, pad with undef and create intermediate merges.
+  // Requires compatible types. Otherwise SrcReg should have been defined by
+  // merge-like instruction that would get artifact combined. Most likely
+  // instruction that defines SrcReg has to perform more/fewer elements
+  // legalization compatible with NarrowTy.
+  assert(SrcTy.isVector() && NarrowTy.isVector() && "Expected vector types");
+  assert((SrcTy.getScalarType() == NarrowTy.getScalarType()) && "bad type");
+
+  if ((SrcTy.getSizeInBits() % NarrowTy.getSizeInBits() != 0) ||
+      (NarrowTy.getSizeInBits() % DstTy.getSizeInBits() != 0))
     return UnableToLegalize;
-  }
 
-  auto Unmerge = MIRBuilder.buildUnmerge(GCDTy, SrcReg);
+  // This is most likely DstTy (smaller then register size) packed in SrcTy
+  // (larger then register size) and since unmerge was not combined it will be
+  // lowered to bit sequence extracts from register. Unpack SrcTy to NarrowTy
+  // (register size) pieces first. Then unpack each of NarrowTy pieces to DstTy.
+
+  // %1:_(DstTy), %2, %3, %4 = G_UNMERGE_VALUES %0:_(SrcTy)
+  //
+  // %5:_(NarrowTy), %6 = G_UNMERGE_VALUES %0:_(SrcTy) - reg sequence
+  // %1:_(DstTy), %2 = G_UNMERGE_VALUES %5:_(NarrowTy) - sequence of bits in reg
+  // %3:_(DstTy), %4 = G_UNMERGE_VALUES %6:_(NarrowTy)
+  auto Unmerge = MIRBuilder.buildUnmerge(NarrowTy, SrcReg);
   const int NumUnmerge = Unmerge->getNumOperands() - 1;
   const int PartsPerUnmerge = NumDst / NumUnmerge;
 
   for (int I = 0; I != NumUnmerge; ++I) {
     auto MIB = MIRBuilder.buildInstr(TargetOpcode::G_UNMERGE_VALUES);
 
     for (int J = 0; J != PartsPerUnmerge; ++J)
       MIB.addDef(MI.getOperand(I * PartsPerUnmerge + J).getReg());
     MIB.addUse(Unmerge.getReg(I));
   }
 
   MI.eraseFromParent();
   return Legalized;
 }
 
-LegalizerHelper::LegalizeResult
-LegalizerHelper::fewerElementsVectorMulo(MachineInstr &MI, unsigned TypeIdx,
-                                         LLT NarrowTy) {
-  Register Result = MI.getOperand(0).getReg();
-  Register Overflow = MI.getOperand(1).getReg();
-  Register LHS = MI.getOperand(2).getReg();
-  Register RHS = MI.getOperand(3).getReg();
-
-  LLT SrcTy = MRI.getType(LHS);
-  if (!SrcTy.isVector())
-    return UnableToLegalize;
-
-  LLT ElementType = SrcTy.getElementType();
-  LLT OverflowElementTy = MRI.getType(Overflow).getElementType();
-  const ElementCount NumResult = SrcTy.getElementCount();
-  LLT GCDTy = getGCDType(SrcTy, NarrowTy);
-
-  // Unmerge the operands to smaller parts of GCD type.
-  auto UnmergeLHS = MIRBuilder.buildUnmerge(GCDTy, LHS);
-  auto UnmergeRHS = MIRBuilder.buildUnmerge(GCDTy, RHS);
-
-  const int NumOps = UnmergeLHS->getNumOperands() - 1;
-  const ElementCount PartsPerUnmerge = NumResult.divideCoefficientBy(NumOps);
-  LLT OverflowTy = LLT::scalarOrVector(PartsPerUnmerge, OverflowElementTy);
-  LLT ResultTy = LLT::scalarOrVector(PartsPerUnmerge, ElementType);
-
-  // Perform the operation over unmerged parts.
-  SmallVector<Register, 8> ResultParts;
-  SmallVector<Register, 8> OverflowParts;
-  for (int I = 0; I != NumOps; ++I) {
-    Register Operand1 = UnmergeLHS->getOperand(I).getReg();
-    Register Operand2 = UnmergeRHS->getOperand(I).getReg();
-    auto PartMul = MIRBuilder.buildInstr(MI.getOpcode(), {ResultTy, OverflowTy},
-                                         {Operand1, Operand2});
-    ResultParts.push_back(PartMul->getOperand(0).getReg());
-    OverflowParts.push_back(PartMul->getOperand(1).getReg());
-  }
-
-  LLT ResultLCMTy = buildLCMMergePieces(SrcTy, NarrowTy, GCDTy, ResultParts);
-  LLT OverflowLCMTy =
-      LLT::scalarOrVector(ResultLCMTy.getElementCount(), OverflowElementTy);
-
-  // Recombine the pieces to the original result and overflow registers.
-  buildWidenedRemergeToDst(Result, ResultLCMTy, ResultParts);
-  buildWidenedRemergeToDst(Overflow, OverflowLCMTy, OverflowParts);
-  MI.eraseFromParent();
-  return Legalized;
-}
-
-// Handle FewerElementsVector a G_BUILD_VECTOR or G_CONCAT_VECTORS that produces
-// a vector
-//
-// Create a G_BUILD_VECTOR or G_CONCAT_VECTORS of NarrowTy pieces, padding with
-// undef as necessary.
-//
-// %3:_(<3 x s16>) = G_BUILD_VECTOR %0, %1, %2
-//   -> <2 x s16>
-//
-// %4:_(s16) = G_IMPLICIT_DEF
-// %5:_(<2 x s16>) = G_BUILD_VECTOR %0, %1
-// %6:_(<2 x s16>) = G_BUILD_VECTOR %2, %4
-// %7:_(<2 x s16>) = G_IMPLICIT_DEF
-// %8:_(<6 x s16>) = G_CONCAT_VECTORS %5, %6, %7
-// %3:_(<3 x s16>), %8:_(<3 x s16>) = G_UNMERGE_VALUES %8
 LegalizerHelper::LegalizeResult
 LegalizerHelper::fewerElementsVectorMerge(MachineInstr &MI, unsigned TypeIdx,
                                           LLT NarrowTy) {
   Register DstReg = MI.getOperand(0).getReg();
   LLT DstTy = MRI.getType(DstReg);
   LLT SrcTy = MRI.getType(MI.getOperand(1).getReg());
-  LLT GCDTy = getGCDType(getGCDType(SrcTy, NarrowTy), DstTy);
+  // Requires compatible types. Otherwise user of DstReg did not peform unmerge
+  // that should have been artifact combined. Most likely instruction that uses
+  // DstReg has to do more/fewer elements legalization compatible with NarrowTy.
+  assert(DstTy.isVector() && NarrowTy.isVector() && "Expected vector types");
+  assert((DstTy.getScalarType() == NarrowTy.getScalarType()) && "bad type");
+  if (NarrowTy == SrcTy)
+    return UnableToLegalize;
 
-  // Break into a common type
-  SmallVector<Register, 16> Parts;
-  for (const MachineOperand &MO : llvm::drop_begin(MI.operands()))
-    extractGCDType(Parts, GCDTy, MO.getReg());
+  // This attempts to lower part of LCMty merge/unmerge sequence. Intended use
+  // is for old mir tests. Since the changes to more/fewer elements it should no
+  // longer be possible to generate MIR like this when starting from llvm-ir
+  // because LCMty approach was replaced with merge/unmerge to vector elements.
+  if (TypeIdx == 1) {
+    assert(SrcTy.isVector() && "Expected vector types");
+    assert((SrcTy.getScalarType() == NarrowTy.getScalarType()) && "bad type");
+    if ((DstTy.getSizeInBits() % NarrowTy.getSizeInBits() != 0) ||
+        (NarrowTy.getNumElements() >= SrcTy.getNumElements()))
+      return UnableToLegalize;
+    // %2:_(DstTy) = G_CONCAT_VECTORS %0:_(SrcTy), %1:_(SrcTy)
+    //
+    // %3:_(EltTy), %4, %5 = G_UNMERGE_VALUES %0:_(SrcTy)
+    // %6:_(EltTy), %7, %8 = G_UNMERGE_VALUES %1:_(SrcTy)
+    // %9:_(NarrowTy) = G_BUILD_VECTOR %3:_(EltTy), %4
+    // %10:_(NarrowTy) = G_BUILD_VECTOR %5:_(EltTy), %6
+    // %11:_(NarrowTy) = G_BUILD_VECTOR %7:_(EltTy), %8
+    // %2:_(DstTy) = G_CONCAT_VECTORS %9:_(NarrowTy), %10, %11
+
+    SmallVector<Register, 4> Elts;
+    LLT EltTy = MRI.getType(MI.getOperand(1).getReg()).getScalarType();
+    for (unsigned i = 1; i < MI.getNumOperands(); ++i) {
+      auto Unmerge = MIRBuilder.buildUnmerge(EltTy, MI.getOperand(i).getReg());
+      for (unsigned j = 0; j < Unmerge->getNumDefs(); ++j)
+        Elts.push_back(Unmerge.getReg(j));
+    }
 
-  // Build the requested new merge, padding with undef.
-  LLT LCMTy = buildLCMMergePieces(DstTy, NarrowTy, GCDTy, Parts,
-                                  TargetOpcode::G_ANYEXT);
+    SmallVector<Register, 4> NarrowTyElts;
+    unsigned NumNarrowTyElts = NarrowTy.getNumElements();
+    unsigned NumNarrowTyPieces = DstTy.getNumElements() / NumNarrowTyElts;
+    for (unsigned i = 0, Offset = 0; i < NumNarrowTyPieces;
+         ++i, Offset += NumNarrowTyElts) {
+      ArrayRef<Register> Pieces(&Elts[Offset], NumNarrowTyElts);
+      NarrowTyElts.push_back(MIRBuilder.buildMerge(NarrowTy, Pieces).getReg(0));
+    }
 
-  // Pack into the original result register.
-  buildWidenedRemergeToDst(DstReg, LCMTy, Parts);
+    MIRBuilder.buildMerge(DstReg, NarrowTyElts);
+    MI.eraseFromParent();
+    return Legalized;
+  }
+
+  assert(TypeIdx == 0 && "Bad type index");
+  if ((NarrowTy.getSizeInBits() % SrcTy.getSizeInBits() != 0) ||
+      (DstTy.getSizeInBits() % NarrowTy.getSizeInBits() != 0))
+    return UnableToLegalize;
+
+  // This is most likely SrcTy (smaller then register size) packed in DstTy
+  // (larger then register size) and since merge was not combined it will be
+  // lowered to bit sequence packing into register. Merge SrcTy to NarrowTy
+  // (register size) pieces first. Then merge each of NarrowTy pieces to DstTy.
+
+  // %0:_(DstTy) = G_MERGE_VALUES %1:_(SrcTy), %2, %3, %4
+  //
+  // %5:_(NarrowTy) = G_MERGE_VALUES %1:_(SrcTy), %2 - sequence of bits in reg
+  // %6:_(NarrowTy) = G_MERGE_VALUES %3:_(SrcTy), %4
+  // %0:_(DstTy)  = G_MERGE_VALUES %5:_(NarrowTy), %6 - reg sequence
+  SmallVector<Register, 4> NarrowTyElts;
+  unsigned NumParts = DstTy.getNumElements() / NarrowTy.getNumElements();
+  unsigned NumSrcElts = SrcTy.isVector() ? SrcTy.getNumElements() : 1;
+  unsigned NumElts = NarrowTy.getNumElements() / NumSrcElts;
+  for (unsigned i = 0; i < NumParts; ++i) {
+    SmallVector<Register, 4> Sources;
+    for (unsigned j = 0; j < NumElts; ++j)
+      Sources.push_back(MI.getOperand(1 + i * NumElts + j).getReg());
+    NarrowTyElts.push_back(MIRBuilder.buildMerge(NarrowTy, Sources).getReg(0));
+  }
 
+  MIRBuilder.buildMerge(DstReg, NarrowTyElts);
   MI.eraseFromParent();
   return Legalized;
 }
 
 LegalizerHelper::LegalizeResult
 LegalizerHelper::fewerElementsVectorExtractInsertVectorElt(MachineInstr &MI,
                                                            unsigned TypeIdx,
                                                            LLT NarrowVecTy) {

                 LeftoverTy, NarrowLeftoverRegs);
   }
 
   LdStMI.eraseFromParent();
   return Legalized;
 }
 
 LegalizerHelper::LegalizeResult
-LegalizerHelper::reduceOperationWidth(MachineInstr &MI, unsigned int TypeIdx,
-                                      LLT NarrowTy) {
-  assert(TypeIdx == 0 && "only one type index expected");
-
-  const unsigned Opc = MI.getOpcode();
-  const int NumDefOps = MI.getNumExplicitDefs();
-  const int NumSrcOps = MI.getNumOperands() - NumDefOps;
-  const unsigned Flags = MI.getFlags();
-  const unsigned NarrowSize = NarrowTy.getSizeInBits();
-  const LLT NarrowScalarTy = LLT::scalar(NarrowSize);
-
-  assert(MI.getNumOperands() <= 4 && "expected instruction with either 1 "
-                                     "result and 1-3 sources or 2 results and "
-                                     "1-2 sources");
-
-  SmallVector<Register, 2> DstRegs;
-  for (int I = 0; I < NumDefOps; ++I)
-    DstRegs.push_back(MI.getOperand(I).getReg());
-
-  // First of all check whether we are narrowing (changing the element type)
-  // or reducing the vector elements
-  const LLT DstTy = MRI.getType(DstRegs[0]);
-  const bool IsNarrow = NarrowTy.getScalarType() != DstTy.getScalarType();
-
-  SmallVector<Register, 8> ExtractedRegs[3];
-  SmallVector<Register, 8> Parts;
-
-  // Break down all the sources into NarrowTy pieces we can operate on. This may
-  // involve creating merges to a wider type, padded with undef.
-  for (int I = 0; I != NumSrcOps; ++I) {
-    Register SrcReg = MI.getOperand(I + NumDefOps).getReg();
-    LLT SrcTy = MRI.getType(SrcReg);
-
-    // The type to narrow SrcReg to. For narrowing, this is a smaller scalar.
-    // For fewerElements, this is a smaller vector with the same element type.
-    LLT OpNarrowTy;
-    if (IsNarrow) {
-      OpNarrowTy = NarrowScalarTy;
-
-      // In case of narrowing, we need to cast vectors to scalars for this to
-      // work properly
-      // FIXME: Can we do without the bitcast here if we're narrowing?
-      if (SrcTy.isVector()) {
-        SrcTy = LLT::scalar(SrcTy.getSizeInBits());
-        SrcReg = MIRBuilder.buildBitcast(SrcTy, SrcReg).getReg(0);
-      }
-    } else {
-      auto NarrowEC = NarrowTy.isVector() ? NarrowTy.getElementCount()
-                                          : ElementCount::getFixed(1);
-      OpNarrowTy = LLT::scalarOrVector(NarrowEC, SrcTy.getScalarType());
-    }
-
-    LLT GCDTy = extractGCDType(ExtractedRegs[I], SrcTy, OpNarrowTy, SrcReg);
-
-    // Build a sequence of NarrowTy pieces in ExtractedRegs for this operand.
-    buildLCMMergePieces(SrcTy, OpNarrowTy, GCDTy, ExtractedRegs[I],
-                        TargetOpcode::G_ANYEXT);
-  }
-
-  SmallVector<Register, 8> ResultRegs[2];
-
-  // Input operands for each sub-instruction.
-  SmallVector<SrcOp, 4> InputRegs(NumSrcOps, Register());
-
-  int NumParts = ExtractedRegs[0].size();
-  const unsigned DstSize = DstTy.getSizeInBits();
-  const LLT DstScalarTy = LLT::scalar(DstSize);
-
-  // Narrowing needs to use scalar types
-  LLT DstLCMTy, NarrowDstTy;
-  if (IsNarrow) {
-    DstLCMTy = getLCMType(DstScalarTy, NarrowScalarTy);
-    NarrowDstTy = NarrowScalarTy;
-  } else {
-    DstLCMTy = getLCMType(DstTy, NarrowTy);
-    NarrowDstTy = NarrowTy;
-  }
-
-  // We widened the source registers to satisfy merge/unmerge size
-  // constraints. We'll have some extra fully undef parts.
-  const int NumRealParts = (DstSize + NarrowSize - 1) / NarrowSize;
-
-  for (int I = 0; I != NumRealParts; ++I) {
-    // Emit this instruction on each of the split pieces.
-    for (int J = 0; J != NumSrcOps; ++J)
-      InputRegs[J] = ExtractedRegs[J][I];
-
-    MachineInstrBuilder Inst;
-    if (NumDefOps == 1)
-      Inst = MIRBuilder.buildInstr(Opc, {NarrowDstTy}, InputRegs, Flags);
-    else
-      Inst = MIRBuilder.buildInstr(Opc, {NarrowDstTy, NarrowDstTy}, InputRegs,
-                                   Flags);
-
-    for (int J = 0; J != NumDefOps; ++J)
-      ResultRegs[J].push_back(Inst.getReg(J));
-  }
-
-  // Fill out the widened result with undef instead of creating instructions
-  // with undef inputs.
-  int NumUndefParts = NumParts - NumRealParts;
-  if (NumUndefParts != 0) {
-    Register Undef = MIRBuilder.buildUndef(NarrowDstTy).getReg(0);
-    for (int I = 0; I != NumDefOps; ++I)
-      ResultRegs[I].append(NumUndefParts, Undef);
-  }
-
-  // Extract the possibly padded result. Use a scratch register if we need to do
-  // a final bitcast, otherwise use the original result register.
-  Register MergeDstReg;
-  for (int I = 0; I != NumDefOps; ++I) {
-    if (IsNarrow && DstTy.isVector())
-      MergeDstReg = MRI.createGenericVirtualRegister(DstScalarTy);
-    else
-      MergeDstReg = DstRegs[I];
-
-    buildWidenedRemergeToDst(MergeDstReg, DstLCMTy, ResultRegs[I]);
-
-    // Recast to vector if we narrowed a vector
-    if (IsNarrow && DstTy.isVector())
-      MIRBuilder.buildBitcast(DstRegs[I], MergeDstReg);
-  }
-
-  MI.eraseFromParent();
-  return Legalized;
-}
-
-LegalizerHelper::LegalizeResult
-LegalizerHelper::fewerElementsVectorSextInReg(MachineInstr &MI, unsigned TypeIdx,
-                                              LLT NarrowTy) {
-  Register DstReg = MI.getOperand(0).getReg();
-  Register SrcReg = MI.getOperand(1).getReg();
-  int64_t Imm = MI.getOperand(2).getImm();
-
-  LLT DstTy = MRI.getType(DstReg);
-
-  SmallVector<Register, 8> Parts;
-  LLT GCDTy = extractGCDType(Parts, DstTy, NarrowTy, SrcReg);
-  LLT LCMTy = buildLCMMergePieces(DstTy, NarrowTy, GCDTy, Parts);
-
-  for (Register &R : Parts)
-    R = MIRBuilder.buildSExtInReg(NarrowTy, R, Imm).getReg(0);
-
-  buildWidenedRemergeToDst(DstReg, LCMTy, Parts);
-
-  MI.eraseFromParent();
-  return Legalized;
+LegalizerHelper::splitVectorLoadStoreToPow2NumElts(GLoadStore &MI, LLT MemTy) {
+  unsigned Pow2NumElts = PowerOf2Floor(MemTy.getNumElements() - 1);
+  LLT EltTy = MemTy.getElementType();
+  LLT P2Ty = (Pow2NumElts == 1) ? EltTy : LLT::fixed_vector(Pow2NumElts, EltTy);
+  return reduceLoadStoreWidth(MI, 0, P2Ty);
 }
 
 LegalizerHelper::LegalizeResult
 LegalizerHelper::fewerElementsVector(MachineInstr &MI, unsigned TypeIdx,
                                      LLT NarrowTy) {
   using namespace TargetOpcode;
+  GenericMachineInstr &GMI = cast<GenericMachineInstr>(MI);
+  unsigned NumElts = NarrowTy.isVector() ? NarrowTy.getNumElements() : 1;
 
   switch (MI.getOpcode()) {
   case G_IMPLICIT_DEF:
-    return fewerElementsVectorImplicitDef(MI, TypeIdx, NarrowTy);
   case G_TRUNC:
   case G_AND:
   case G_OR:
   case G_XOR:
   case G_ADD:
   case G_SUB:
   case G_MUL:
   case G_PTR_ADD:

   case G_FSHR:
   case G_ROTL:
   case G_ROTR:
   case G_FREEZE:
   case G_SADDSAT:
   case G_SSUBSAT:
   case G_UADDSAT:
   case G_USUBSAT:
-    return reduceOperationWidth(MI, TypeIdx, NarrowTy);
   case G_UMULO:
   case G_SMULO:
-    return fewerElementsVectorMulo(MI, TypeIdx, NarrowTy);
   case G_SHL:
   case G_LSHR:
   case G_ASHR:
   case G_SSHLSAT:
   case G_USHLSAT:
   case G_CTLZ:
   case G_CTLZ_ZERO_UNDEF:
   case G_CTTZ:
   case G_CTTZ_ZERO_UNDEF:
   case G_CTPOP:
   case G_FCOPYSIGN:
-    return fewerElementsVectorMultiEltType(MI, TypeIdx, NarrowTy);
   case G_ZEXT:
   case G_SEXT:
   case G_ANYEXT:
   case G_FPEXT:
   case G_FPTRUNC:
   case G_SITOFP:
   case G_UITOFP:
   case G_FPTOSI:
   case G_FPTOUI:
   case G_INTTOPTR:
   case G_PTRTOINT:
   case G_ADDRSPACE_CAST:
-    return fewerElementsVectorCasts(MI, TypeIdx, NarrowTy);
+    return fewerElementsVectorMultiEltType(GMI, NumElts);
   case G_ICMP:
   case G_FCMP:
-    return fewerElementsVectorCmp(MI, TypeIdx, NarrowTy);
+    return fewerElementsVectorMultiEltType(GMI, NumElts, {1 /*cpm predicate*/});
   case G_SELECT:
-    return fewerElementsVectorSelect(MI, TypeIdx, NarrowTy);
+    if (MRI.getType(MI.getOperand(1).getReg()).isVector())
+      return fewerElementsVectorMultiEltType(GMI, NumElts);
+    return fewerElementsVectorMultiEltType(GMI, NumElts, {1 /*scalar cond*/});
   case G_PHI:
-    return fewerElementsVectorPhi(MI, TypeIdx, NarrowTy);
+    return fewerElementsVectorPhi(GMI, NumElts);
   case G_UNMERGE_VALUES:
     return fewerElementsVectorUnmergeValues(MI, TypeIdx, NarrowTy);
   case G_BUILD_VECTOR:
     assert(TypeIdx == 0 && "not a vector type index");
     return fewerElementsVectorMerge(MI, TypeIdx, NarrowTy);
   case G_CONCAT_VECTORS:
     if (TypeIdx != 1) // TODO: This probably does work as expected already.
       return UnableToLegalize;
     return fewerElementsVectorMerge(MI, TypeIdx, NarrowTy);
   case G_EXTRACT_VECTOR_ELT:
   case G_INSERT_VECTOR_ELT:
     return fewerElementsVectorExtractInsertVectorElt(MI, TypeIdx, NarrowTy);
   case G_LOAD:
   case G_STORE:
     return reduceLoadStoreWidth(cast<GLoadStore>(MI), TypeIdx, NarrowTy);
   case G_SEXT_INREG:
-    return fewerElementsVectorSextInReg(MI, TypeIdx, NarrowTy);
+    return fewerElementsVectorMultiEltType(GMI, NumElts, {2 /*imm*/});
   GISEL_VECREDUCE_CASES_NONSEQ
     return fewerElementsVectorReductions(MI, TypeIdx, NarrowTy);
   case G_SHUFFLE_VECTOR:
     return fewerElementsVectorShuffle(MI, TypeIdx, NarrowTy);
   default:
     return UnableToLegalize;
   }
 }

       return UnableToLegalize;
     Observer.changingInstr(MI);
     moreElementsVectorSrc(MI, MoreTy, 0);
     Observer.changedInstr(MI);
     return Legalized;
   case TargetOpcode::G_AND:
   case TargetOpcode::G_OR:
   case TargetOpcode::G_XOR:
+  case TargetOpcode::G_ADD:
+  case TargetOpcode::G_SUB:
+  case TargetOpcode::G_MUL:
+  case TargetOpcode::G_FADD:
+  case TargetOpcode::G_FMUL:
+  case TargetOpcode::G_UADDSAT:
+  case TargetOpcode::G_USUBSAT:
+  case TargetOpcode::G_SADDSAT:
+  case TargetOpcode::G_SSUBSAT:
   case TargetOpcode::G_SMIN:
   case TargetOpcode::G_SMAX:
   case TargetOpcode::G_UMIN:
   case TargetOpcode::G_UMAX:
   case TargetOpcode::G_FMINNUM:
   case TargetOpcode::G_FMAXNUM:
   case TargetOpcode::G_FMINNUM_IEEE:
   case TargetOpcode::G_FMAXNUM_IEEE:
   case TargetOpcode::G_FMINIMUM:
   case TargetOpcode::G_FMAXIMUM: {
     Observer.changingInstr(MI);
     moreElementsVectorSrc(MI, MoreTy, 1);
     moreElementsVectorSrc(MI, MoreTy, 2);
     moreElementsVectorDst(MI, MoreTy, 0);
     Observer.changedInstr(MI);
     return Legalized;
   }
+  case TargetOpcode::G_FMA:
+  case TargetOpcode::G_FSHR:
+  case TargetOpcode::G_FSHL: {
+    Observer.changingInstr(MI);
+    moreElementsVectorSrc(MI, MoreTy, 1);
+    moreElementsVectorSrc(MI, MoreTy, 2);
+    moreElementsVectorSrc(MI, MoreTy, 3);
+    moreElementsVectorDst(MI, MoreTy, 0);
+    Observer.changedInstr(MI);
+    return Legalized;
+  }
   case TargetOpcode::G_EXTRACT:
     if (TypeIdx != 1)
       return UnableToLegalize;
     Observer.changingInstr(MI);
     moreElementsVectorSrc(MI, MoreTy, 1);
     Observer.changedInstr(MI);
     return Legalized;
   case TargetOpcode::G_INSERT:
   case TargetOpcode::G_FREEZE:
+  case TargetOpcode::G_FNEG:
+  case TargetOpcode::G_FABS:
+  case TargetOpcode::G_BSWAP:
+  case TargetOpcode::G_FCANONICALIZE:
+  case TargetOpcode::G_SEXT_INREG:
     if (TypeIdx != 0)
       return UnableToLegalize;
     Observer.changingInstr(MI);
     moreElementsVectorSrc(MI, MoreTy, 1);
     moreElementsVectorDst(MI, MoreTy, 0);
     Observer.changedInstr(MI);
     return Legalized;
   case TargetOpcode::G_SELECT:
     if (TypeIdx != 0)
       return UnableToLegalize;
     if (MRI.getType(MI.getOperand(1).getReg()).isVector())
       return UnableToLegalize;
 
     Observer.changingInstr(MI);
     moreElementsVectorSrc(MI, MoreTy, 2);
     moreElementsVectorSrc(MI, MoreTy, 3);
     moreElementsVectorDst(MI, MoreTy, 0);
     Observer.changedInstr(MI);
     return Legalized;
-  case TargetOpcode::G_UNMERGE_VALUES: {
-    if (TypeIdx != 1)
-      return UnableToLegalize;
-
-    LLT DstTy = MRI.getType(MI.getOperand(0).getReg());
-    int NumDst = MI.getNumOperands() - 1;
-    moreElementsVectorSrc(MI, MoreTy, NumDst);
-
-    auto MIB = MIRBuilder.buildInstr(TargetOpcode::G_UNMERGE_VALUES);
-    for (int I = 0; I != NumDst; ++I)
-      MIB.addDef(MI.getOperand(I).getReg());
+  case TargetOpcode::G_UNMERGE_VALUES:
+    return UnableToLegalize;
+  case TargetOpcode::G_PHI:
+    return moreElementsVectorPhi(MI, TypeIdx, MoreTy);
+  case TargetOpcode::G_SHUFFLE_VECTOR:
+    return moreElementsVectorShuffle(MI, TypeIdx, MoreTy);
+  case TargetOpcode::G_BUILD_VECTOR: {
+    SmallVector<SrcOp, 8> Elts;
+    for (auto Op : MI.uses()) {
+      Elts.push_back(Op.getReg());
+    }
 
-    int NewNumDst = MoreTy.getSizeInBits() / DstTy.getSizeInBits();
-    for (int I = NumDst; I != NewNumDst; ++I)
-      MIB.addDef(MRI.createGenericVirtualRegister(DstTy));
+    for (unsigned i = Elts.size(); i < MoreTy.getNumElements(); ++i) {
+      Elts.push_back(MIRBuilder.buildUndef(MoreTy.getScalarType()));
+    }
 
-    MIB.addUse(MI.getOperand(NumDst).getReg());
+    MIRBuilder.buildDeleteTrailingVectorElements(
+        MI.getOperand(0).getReg(), MIRBuilder.buildInstr(Opc, {MoreTy}, Elts));
     MI.eraseFromParent();
     return Legalized;
   }
-  case TargetOpcode::G_PHI:
-    return moreElementsVectorPhi(MI, TypeIdx, MoreTy);
-  case TargetOpcode::G_SHUFFLE_VECTOR:
-    return moreElementsVectorShuffle(MI, TypeIdx, MoreTy);
+  case TargetOpcode::G_TRUNC: {
+    Observer.changingInstr(MI);
+    moreElementsVectorSrc(MI, MoreTy, 1);
+    moreElementsVectorDst(MI, MoreTy, 0);
+    Observer.changedInstr(MI);
+    return Legalized;
+  }
   default:
     return UnableToLegalize;
   }
 }
 
 LegalizerHelper::LegalizeResult
 LegalizerHelper::moreElementsVectorShuffle(MachineInstr &MI,
                                            unsigned int TypeIdx, LLT MoreTy) {

   Register InsertVal;
   if (MI.getOpcode() == TargetOpcode::G_INSERT_VECTOR_ELT)
     InsertVal = MI.getOperand(2).getReg();
 
   Register Idx = MI.getOperand(MI.getNumOperands() - 1).getReg();
 
   LLT VecTy = MRI.getType(SrcVec);
   LLT EltTy = VecTy.getElementType();
+  unsigned NumElts = VecTy.getNumElements();
+
+  int64_t IdxVal;
+  if (mi_match(Idx, MRI, m_ICst(IdxVal)) && IdxVal <= NumElts) {
+    SmallVector<Register, 8> SrcRegs;
+    extractParts(SrcVec, EltTy, NumElts, SrcRegs);
+
+    if (InsertVal) {
+      SrcRegs[IdxVal] = MI.getOperand(2).getReg();
+      MIRBuilder.buildMerge(DstReg, SrcRegs);
+    } else {
+      MIRBuilder.buildCopy(DstReg, SrcRegs[IdxVal]);
+    }
+
+    MI.eraseFromParent();
+    return Legalized;
+  }
+
   if (!EltTy.isByteSized()) { // Not implemented.
     LLVM_DEBUG(dbgs() << "Can't handle non-byte element vectors yet\n");
     return UnableToLegalize;
   }
 
   unsigned EltBytes = EltTy.getSizeInBytes();
   Align VecAlign = getStackTemporaryAlignment(VecTy);
   Align EltAlign;
 
   MachinePointerInfo PtrInfo;
   auto StackTemp = createStackTemporary(TypeSize::Fixed(VecTy.getSizeInBytes()),
                                         VecAlign, PtrInfo);
   MIRBuilder.buildStore(SrcVec, StackTemp, PtrInfo, VecAlign);
 
   // Get the pointer to the element, and be sure not to hit undefined behavior
   // if the index is out of bounds.
   Register EltPtr = getVectorElementPointer(StackTemp.getReg(0), VecTy, Idx);
 
-  int64_t IdxVal;
   if (mi_match(Idx, MRI, m_ICst(IdxVal))) {
     int64_t Offset = IdxVal * EltBytes;
     PtrInfo = PtrInfo.getWithOffset(Offset);
     EltAlign = commonAlignment(VecAlign, Offset);
   } else {
     // We lose information with a variable offset.
     EltAlign = getStackTemporaryAlignment(EltTy);
     PtrInfo = MachinePointerInfo(MRI.getType(EltPtr).getAddressSpace());

   Register Undef;
   SmallVector<Register, 32> BuildVec;
   LLT EltTy = DstTy.getElementType();
 
   for (int Idx : Mask) {
     if (Idx < 0) {
       if (!Undef.isValid())
         Undef = MIRBuilder.buildUndef(EltTy).getReg(0);
       BuildVec.push_back(Undef);
       continue;
     }
 
     if (Src0Ty.isScalar()) {
       BuildVec.push_back(Idx == 0 ? Src0Reg : Src1Reg);
     } else {
       int NumElts = Src0Ty.getNumElements();
       Register SrcVec = Idx < NumElts ? Src0Reg : Src1Reg;
       int ExtractIdx = Idx < NumElts ? Idx : Idx - NumElts;
       auto IdxK = MIRBuilder.buildConstant(IdxTy, ExtractIdx);
       auto Extract = MIRBuilder.buildExtractVectorElement(EltTy, SrcVec, IdxK);
       BuildVec.push_back(Extract.getReg(0));
     }
   }
 
   MIRBuilder.buildBuildVector(DstReg, BuildVec);
   MI.eraseFromParent();
   return Legalized;
 }

 LegalizerHelper::lowerExtract(MachineInstr &MI) {
   Register Dst = MI.getOperand(0).getReg();
   Register Src = MI.getOperand(1).getReg();
   unsigned Offset = MI.getOperand(2).getImm();
 
   LLT DstTy = MRI.getType(Dst);
   LLT SrcTy = MRI.getType(Src);
 
+  // Extract sub-vector or one element
+  if (SrcTy.isVector()) {
+    unsigned SrcEltSize = SrcTy.getElementType().getSizeInBits();
+    unsigned DstSize = DstTy.getSizeInBits();
+
+    if ((Offset % SrcEltSize == 0) && (DstSize % SrcEltSize == 0) &&
+        (Offset + DstSize <= SrcTy.getSizeInBits())) {
+      // Unmerge and allow access to each Src element for the artifact combiner.
+      auto Unmerge = MIRBuilder.buildUnmerge(SrcTy.getElementType(), Src);
+
+      // Take element(s) we need to extract and copy it (merge them).
+      SmallVector<Register, 4> SubVectorElts;
+      for (unsigned Idx = Offset / SrcEltSize;
+           Idx < (Offset + DstSize) / SrcEltSize; ++Idx) {
+        SubVectorElts.push_back(Unmerge.getReg(Idx));
+      }
+      if (SubVectorElts.size() == 1)
+        MIRBuilder.buildCopy(Dst, SubVectorElts[0]);
+      else
+        MIRBuilder.buildMerge(Dst, SubVectorElts);
+
+      MI.eraseFromParent();
+      return Legalized;
+    }
+  }
+
   if (DstTy.isScalar() &&
       (SrcTy.isScalar() ||
        (SrcTy.isVector() && DstTy == SrcTy.getElementType()))) {
     LLT SrcIntTy = SrcTy;
     if (!SrcTy.isScalar()) {
       SrcIntTy = LLT::scalar(SrcTy.getSizeInBits());
       Src = MIRBuilder.buildBitcast(SrcIntTy, Src).getReg(0);
     }

   Register Dst = MI.getOperand(0).getReg();
   Register Src = MI.getOperand(1).getReg();
   Register InsertSrc = MI.getOperand(2).getReg();
   uint64_t Offset = MI.getOperand(3).getImm();
 
   LLT DstTy = MRI.getType(Src);
   LLT InsertTy = MRI.getType(InsertSrc);
 
+  // Insert sub-vector or one element
+  if (DstTy.isVector() && !InsertTy.isPointer()) {
+    LLT EltTy = DstTy.getElementType();
+    unsigned EltSize = EltTy.getSizeInBits();
+    unsigned InsertSize = InsertTy.getSizeInBits();
+
+    if ((Offset % EltSize == 0) && (InsertSize % EltSize == 0) &&
+        (Offset + InsertSize <= DstTy.getSizeInBits())) {
+      auto UnmergeSrc = MIRBuilder.buildUnmerge(EltTy, Src);
+      SmallVector<Register, 4> DstElts;
+      unsigned Idx = 0;
+      // Elements from Src before insert start Offset
+      for (; Idx < Offset / EltSize; ++Idx) {
+        DstElts.push_back(UnmergeSrc.getReg(Idx));
+      }
+
+      // Replace elements in Src with elements from InsertSrc
+      if (InsertTy.getSizeInBits() > EltSize) {
+        auto UnmergeInsertSrc = MIRBuilder.buildUnmerge(EltTy, InsertSrc);
+        for (unsigned i = 0; Idx < (Offset + InsertSize) / EltSize;
+             ++Idx, ++i) {
+          DstElts.push_back(UnmergeInsertSrc.getReg(i));
+        }
+      } else {
+        DstElts.push_back(InsertSrc);
+        ++Idx;
+      }
+
+      // Remaining elements from Src after insert
+      for (; Idx < DstTy.getNumElements(); ++Idx) {
+        DstElts.push_back(UnmergeSrc.getReg(Idx));
+      }
+
+      MIRBuilder.buildMerge(Dst, DstElts);
+      MI.eraseFromParent();
+      return Legalized;
+    }
+  }
+
   if (InsertTy.isVector() ||
       (DstTy.isVector() && DstTy.getElementType() != InsertTy))
     return UnableToLegalize;
 
   const DataLayout &DL = MIRBuilder.getDataLayout();
   if ((DstTy.isPointer() &&
        DL.isNonIntegralAddressSpace(DstTy.getAddressSpace())) ||
       (InsertTy.isPointer() &&

llvm/lib/CodeGen/GlobalISel/MachineIRBuilder.cpp

MachineInstrBuilder MachineIRBuilder::buildMaskLowPtrBits(const DstOp &Res,
                                                           uint32_t NumBits) {
   LLT PtrTy = Res.getLLTTy(*getMRI());
   LLT MaskTy = LLT::scalar(PtrTy.getSizeInBits());
   Register MaskReg = getMRI()->createGenericVirtualRegister(MaskTy);
   buildConstant(MaskReg, maskTrailingZeros<uint64_t>(NumBits));
   return buildPtrMask(Res, Op0, MaskReg);
 }
 
+MachineInstrBuilder
+MachineIRBuilder::buildPadVectorWithUndefElements(const DstOp &Res,
+                                                  const SrcOp &Op0) {
+  LLT ResTy = Res.getLLTTy(*getMRI());
+  LLT Op0Ty = Op0.getLLTTy(*getMRI());
+
+  assert((ResTy.isVector() && Op0Ty.isVector()) && "Non vector type");
+  assert((ResTy.getElementType() == Op0Ty.getElementType()) &&
+         "Different vector element types");
+  assert((ResTy.getNumElements() > Op0Ty.getNumElements()) &&
+         "Op0 has more elements");
+
+  auto Unmerge = buildUnmerge(Op0Ty.getElementType(), Op0);
+  SmallVector<Register, 8> Regs;
+  for (auto Op : Unmerge.getInstr()->defs())
+    Regs.push_back(Op.getReg());
+  Register Undef = buildUndef(Op0Ty.getElementType()).getReg(0);
+  unsigned NumberOfPadElts = ResTy.getNumElements() - Regs.size();
+  for (unsigned i = 0; i < NumberOfPadElts; ++i)
+    Regs.push_back(Undef);
+  return buildMerge(Res, Regs);
+}
+
+MachineInstrBuilder
+MachineIRBuilder::buildDeleteTrailingVectorElements(const DstOp &Res,
+                                                    const SrcOp &Op0) {
+  LLT ResTy = Res.getLLTTy(*getMRI());
+  LLT Op0Ty = Op0.getLLTTy(*getMRI());
+
+  assert((ResTy.isVector() && Op0Ty.isVector()) && "Non vector type");
+  assert((ResTy.getElementType() == Op0Ty.getElementType()) &&
+         "Different vector element types");
+  assert((ResTy.getNumElements() < Op0Ty.getNumElements()) &&
+         "Op0 has fewer elements");
+
+  SmallVector<Register, 8> Regs;
+  auto Unmerge = buildUnmerge(Op0Ty.getElementType(), Op0);
+  for (unsigned i = 0; i < ResTy.getNumElements(); ++i)
+    Regs.push_back(Unmerge.getReg(i));
+  return buildMerge(Res, Regs);
+}
+
 MachineInstrBuilder MachineIRBuilder::buildBr(MachineBasicBlock &Dest) {
   return buildInstr(TargetOpcode::G_BR).addMBB(&Dest);
 }
 
 MachineInstrBuilder MachineIRBuilder::buildBrIndirect(Register Tgt) {
   assert(getMRI()->getType(Tgt).isPointer() && "invalid branch destination");
   return buildInstr(TargetOpcode::G_BRINDIRECT).addUse(Tgt);
 }
MachineInstrBuilder MachineIRBuilder::buildUnmerge(ArrayRef<LLT> Res,
   SmallVector<DstOp, 8> TmpVec(Res.begin(), Res.end());
   assert(TmpVec.size() > 1);
   return buildInstr(TargetOpcode::G_UNMERGE_VALUES, TmpVec, Op);
 }
 
 MachineInstrBuilder MachineIRBuilder::buildUnmerge(LLT Res,
                                                    const SrcOp &Op) {
   unsigned NumReg = Op.getLLTTy(*getMRI()).getSizeInBits() / Res.getSizeInBits();
-  SmallVector<Register, 8> TmpVec;
-  for (unsigned I = 0; I != NumReg; ++I)
-    TmpVec.push_back(getMRI()->createGenericVirtualRegister(Res));
-  return buildUnmerge(TmpVec, Op);
+  SmallVector<DstOp, 8> TmpVec(NumReg, Res);
+  return buildInstr(TargetOpcode::G_UNMERGE_VALUES, TmpVec, Op);
 }
 
 MachineInstrBuilder MachineIRBuilder::buildUnmerge(ArrayRef<Register> Res,
                                                    const SrcOp &Op) {
   // Unfortunately to convert from ArrayRef<Register> to ArrayRef<DstOp>,
   // we need some temporary storage for the DstOp objects. Here we use a
   // sufficiently large SmallVector to not go through the heap.
   SmallVector<DstOp, 8> TmpVec(Res.begin(), Res.end());

llvm/lib/CodeGen/GlobalISel/Utils.cpp

   if (LCMSize == OrigSize)
     return OrigTy;
   if (LCMSize == TargetSize)
     return TargetTy;
 
   return LLT::scalar(LCMSize);
 }
 
+LLT llvm::getCoverTy(LLT OrigTy, LLT TargetTy) {
+  if (!OrigTy.isVector() || !TargetTy.isVector() || OrigTy == TargetTy ||
+      (OrigTy.getScalarSizeInBits() != TargetTy.getScalarSizeInBits()))
+    return getLCMType(OrigTy, TargetTy);
+
+  unsigned OrigTyNumElts = OrigTy.getNumElements();
+  unsigned TargetTyNumElts = TargetTy.getNumElements();
+  if (OrigTyNumElts % TargetTyNumElts == 0)
+    return OrigTy;
+
+  unsigned NumElts = alignTo(OrigTyNumElts, TargetTyNumElts);
+  return LLT::scalarOrVector(ElementCount::getFixed(NumElts),
+                             OrigTy.getElementType());
+}
+
 LLT llvm::getGCDType(LLT OrigTy, LLT TargetTy) {
   const unsigned OrigSize = OrigTy.getSizeInBits();
   const unsigned TargetSize = TargetTy.getSizeInBits();
 
   if (OrigSize == TargetSize)
     return OrigTy;
 
   if (OrigTy.isVector()) {

llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp

     .moreElementsIf(isSmallOddVector(0), oneMoreElement(0))
     .scalarize(0);
 
   if (ST.hasVOP3PInsts() && ST.hasAddNoCarry() && ST.hasIntClamp()) {
     // Full set of gfx9 features.
     getActionDefinitionsBuilder({G_ADD, G_SUB, G_MUL})
       .legalFor({S32, S16, V2S16})
       .minScalar(0, S16)
-      .clampMaxNumElements(0, S16, 2)
+      .clampMaxNumElementsStrict(0, S16, 2)
       .widenScalarToNextMultipleOf(0, 32)
       .maxScalar(0, S32)
       .scalarize(0);
 
     getActionDefinitionsBuilder({G_UADDSAT, G_USUBSAT, G_SADDSAT, G_SSUBSAT})
       .legalFor({S32, S16, V2S16}) // Clamp modifier
       .minScalarOrElt(0, S16)
-      .clampMaxNumElements(0, S16, 2)
+      .clampMaxNumElementsStrict(0, S16, 2)
       .scalarize(0)
       .widenScalarToNextPow2(0, 32)
       .lower();
   } else if (ST.has16BitInsts()) {
     getActionDefinitionsBuilder({G_ADD, G_SUB, G_MUL})
       .legalFor({S32, S16})
       .minScalar(0, S16)
       .widenScalarToNextMultipleOf(0, 32)

       .scalarize(0);
   } else {
     MinNumMaxNum.customFor(FPTypesBase)
       .clampScalar(0, S32, S64)
       .scalarize(0);
   }
 
   if (ST.hasVOP3PInsts())
-    FPOpActions.clampMaxNumElements(0, S16, 2);
+    FPOpActions.clampMaxNumElementsStrict(0, S16, 2);
 
   FPOpActions
     .scalarize(0)
     .clampScalar(0, ST.has16BitInsts() ? S16 : S32, S64);
 
   TrigActions
     .scalarize(0)
     .clampScalar(0, ST.has16BitInsts() ? S16 : S32, S64);
 
   FDIVActions
     .scalarize(0)
     .clampScalar(0, ST.has16BitInsts() ? S16 : S32, S64);
 
   getActionDefinitionsBuilder({G_FNEG, G_FABS})
     .legalFor(FPTypesPK16)
-    .clampMaxNumElements(0, S16, 2)
+    .clampMaxNumElementsStrict(0, S16, 2)
     .scalarize(0)
     .clampScalar(0, S16, S64);
 
   if (ST.has16BitInsts()) {
     getActionDefinitionsBuilder({G_FSQRT, G_FFLOOR})
       .legalFor({S32, S64, S16})
       .scalarize(0)
       .clampScalar(0, S16, S64);

     .legalFor({S32, S64})
     .clampScalar(0, S32, S64)
     .scalarize(0)
     .widenScalarToNextPow2(0);
 
   if (ST.has16BitInsts()) {
     getActionDefinitionsBuilder(G_BSWAP)
       .legalFor({S16, S32, V2S16})
-      .clampMaxNumElements(0, S16, 2)
+      .clampMaxNumElementsStrict(0, S16, 2)
       // FIXME: Fixing non-power-of-2 before clamp is workaround for
       // narrowScalar limitation.
       .widenScalarToNextPow2(0)
       .clampScalar(0, S16, S32)
       .scalarize(0);
 
     if (ST.hasVOP3PInsts()) {
       getActionDefinitionsBuilder({G_SMIN, G_SMAX, G_UMIN, G_UMAX, G_ABS})

 
   for (unsigned Op : {G_EXTRACT, G_INSERT}) {
     unsigned BigTyIdx = Op == G_EXTRACT ? 1 : 0;
     unsigned LitTyIdx = Op == G_EXTRACT ? 0 : 1;
 
     // FIXME: Doesn't handle extract of illegal sizes.
     getActionDefinitionsBuilder(Op)
       .lowerIf(all(typeIs(LitTyIdx, S16), sizeIs(BigTyIdx, 32)))
+      .lowerIf([=](const LegalityQuery &Query) {
+          // Sub-vector(or single element) insert and extract. 
+          // TODO: verify immediate offset here since lower only works with
+          // whole elements.
+          const LLT BigTy = Query.Types[BigTyIdx];
+          return BigTy.isVector();
+        })
       // FIXME: Multiples of 16 should not be legal.
       .legalIf([=](const LegalityQuery &Query) {
           const LLT BigTy = Query.Types[BigTyIdx];
           const LLT LitTy = Query.Types[LitTyIdx];
           return (BigTy.getSizeInBits() % 32 == 0) &&
                  (LitTy.getSizeInBits() % 16 == 0);
         })
       .widenScalarIf(

   auto &SextInReg = getActionDefinitionsBuilder(G_SEXT_INREG)
     .legalFor({{S32}, {S64}});
 
   if (ST.hasVOP3PInsts()) {
     SextInReg.lowerFor({{V2S16}})
       // Prefer to reduce vector widths for 16-bit vectors before lowering, to
       // get more vector shift opportunities, since we'll get those when
       // expanded.
-      .fewerElementsIf(elementTypeIs(0, S16), changeTo(0, V2S16));
+      .clampMaxNumElementsStrict(0, S16, 2);
   } else if (ST.has16BitInsts()) {
     SextInReg.lowerFor({{S32}, {S64}, {S16}});
   } else {
     // Prefer to promote to s32 before lowering if we don't have 16-bit
     // shifts. This avoid a lot of intermediate truncate and extend operations.
     SextInReg.lowerFor({{S32}, {S64}});
   }
 
   SextInReg
     .scalarize(0)
     .clampScalar(0, S32, S64)
     .lower();
 
   getActionDefinitionsBuilder({G_ROTR, G_ROTL})
     .scalarize(0)
     .lower();
 
   // TODO: Only Try to form v2s16 with legal packed instructions.
   getActionDefinitionsBuilder(G_FSHR)
     .legalFor({{S32, S32}})
     .lowerFor({{V2S16, V2S16}})
-    .fewerElementsIf(elementTypeIs(0, S16), changeTo(0, V2S16))
+    .clampMaxNumElementsStrict(0, S16, 2)
     .scalarize(0)
     .lower();
 
   if (ST.hasVOP3PInsts()) {
     getActionDefinitionsBuilder(G_FSHL)
       .lowerFor({{V2S16, V2S16}})
-      .fewerElementsIf(elementTypeIs(0, S16), changeTo(0, V2S16))
+      .clampMaxNumElementsStrict(0, S16, 2)
       .scalarize(0)
       .lower();
   } else {
     getActionDefinitionsBuilder(G_FSHL)
       .scalarize(0)
       .lower();
   }
 

 
   int64_t IdxVal = MaybeIdxVal->Value.getSExtValue();
   Register Dst = MI.getOperand(0).getReg();
   Register Vec = MI.getOperand(1).getReg();
   Register Ins = MI.getOperand(2).getReg();
 
   LLT VecTy = MRI.getType(Vec);
   LLT EltTy = VecTy.getElementType();
   assert(EltTy == MRI.getType(Ins));
 
   unsigned NumElts = VecTy.getNumElements();
   if (IdxVal < NumElts) {
     SmallVector<Register, 8> SrcRegs;
     for (unsigned i = 0; i < NumElts; ++i)
       SrcRegs.push_back(MRI.createGenericVirtualRegister(EltTy));
     B.buildUnmerge(SrcRegs, Vec);
 
     SrcRegs[IdxVal] = MI.getOperand(2).getReg();
     B.buildMerge(Dst, SrcRegs);
   } else {

   LLT SrcTy = MRI.getType(Src0);
 
   if (SrcTy == V2S16 && DstTy == V2S16 &&
       AMDGPU::isLegalVOP3PShuffleMask(MI.getOperand(3).getShuffleMask()))
     return true;
 
   MachineIRBuilder HelperBuilder(MI);
   GISelObserverWrapper DummyObserver;
   LegalizerHelper Helper(B.getMF(), DummyObserver, HelperBuilder);
   return Helper.lowerShuffleVector(MI) == LegalizerHelper::Legalized;
 }
 
 bool AMDGPULegalizerInfo::legalizeSinCos(
   MachineInstr &MI, MachineRegisterInfo &MRI,
   MachineIRBuilder &B) const {
 
   Register DstReg = MI.getOperand(0).getReg();
   Register SrcReg = MI.getOperand(1).getReg();
   LLT Ty = MRI.getType(DstReg);

       if (isRegisterType(ValTy)) {
         // If this a case where G_EXTRACT is legal, use it.
         // (e.g. <3 x s32> -> <4 x s32>)
         WideLoad = B.buildLoadFromOffset(WideTy, PtrReg, *MMO, 0).getReg(0);
         B.buildExtract(ValReg, WideLoad, 0);
       } else {
         // For cases where the widened type isn't a nice register value, unmerge
         // from a widened register (e.g. <3 x s16> -> <4 x s16>)
-        B.setInsertPt(B.getMBB(), ++B.getInsertPt());
-        WideLoad = Helper.widenWithUnmerge(WideTy, ValReg);
-        B.setInsertPt(B.getMBB(), MI.getIterator());
-        B.buildLoadFromOffset(WideLoad, PtrReg, *MMO, 0);
+        WideLoad = B.buildLoadFromOffset(WideTy, PtrReg, *MMO, 0).getReg(0);
+        B.buildDeleteTrailingVectorElements(ValReg, WideLoad);
       }
     }
 
     MI.eraseFromParent();
     return true;
   }
 
   return false;

       PackedRegs.resize(4, B.buildUndef(S32).getReg(0));
       return B.buildBuildVector(LLT::fixed_vector(4, S32), PackedRegs)
           .getReg(0);
     }
 
     llvm_unreachable("invalid data type");
   }
 
+  if (StoreVT == LLT::fixed_vector(3, S16)) {
+    Reg = B.buildPadVectorWithUndefElements(LLT::fixed_vector(4, S16), Reg)
+              .getReg(0);
+  }
   return Reg;
 }
 
 Register AMDGPULegalizerInfo::fixStoreSourceType(
   MachineIRBuilder &B, Register VData, bool IsFormat) const {
   MachineRegisterInfo *MRI = B.getMRI();
   LLT Ty = MRI->getType(VData);
 

 /// to exposes all register repacking to the legalizer/combiners. We also don't
 /// want a selected instrution entering RegBankSelect. In order to avoid
 /// defining a multitude of intermediate image instructions, directly hack on
 /// the intrinsic's arguments. In cases like a16 addresses, this requires
 /// padding now unnecessary arguments with $noreg.
 bool AMDGPULegalizerInfo::legalizeImageIntrinsic(
     MachineInstr &MI, MachineIRBuilder &B, GISelChangeObserver &Observer,
     const AMDGPU::ImageDimIntrinsicInfo *Intr) const {
+  dbgs() << "\n\n";
+  MI.dump();
 
   const unsigned NumDefs = MI.getNumExplicitDefs();
   const unsigned ArgOffset = NumDefs + 1;
   bool IsTFE = NumDefs == 2;
   // We are only processing the operands of d16 image operations on subtargets
   // that use the unpacked register layout, or need to repack the TFE result.
 
   // TODO: Do we need to guard against already legalized intrinsics?

   }
 
   assert(!ST.hasUnpackedD16VMem() && ResTy == V2S16);
   const int RegsToCover = (Ty.getSizeInBits() + 31) / 32;
 
   // Deal with the one annoying legal case.
   const LLT V3S16 = LLT::fixed_vector(3, 16);
   if (Ty == V3S16) {
-    padWithUndef(ResTy, RegsToCover - ResultRegs.size() + 1);
-    auto Concat = B.buildConcatVectors(LLT::fixed_vector(6, 16), ResultRegs);
-    B.buildUnmerge({DstReg, MRI->createGenericVirtualRegister(V3S16)}, Concat);
+    if (IsTFE) {
+      if (ResultRegs.size() == 1) {
+        NewResultReg = ResultRegs[0];
+      } else if (ResultRegs.size() == 2) {
+        LLT V4S16 = LLT::fixed_vector(4, 16);
+        NewResultReg = B.buildConcatVectors(V4S16, ResultRegs).getReg(0);
+      } else {
+        return false;
+      }
+    }
+
+    if (MRI->getType(DstReg).getNumElements() <
+        MRI->getType(NewResultReg).getNumElements()) {
+      B.buildDeleteTrailingVectorElements(DstReg, NewResultReg);
+    } else {
+      B.buildPadVectorWithUndefElements(DstReg, NewResultReg);
+    }
     return true;
   }
 
   padWithUndef(ResTy, RegsToCover - ResultRegs.size());
   B.buildConcatVectors(DstReg, ResultRegs);
   return true;
 }
 
 bool AMDGPULegalizerInfo::legalizeSBufferLoad(
   LegalizerHelper &Helper, MachineInstr &MI) const {
   MachineIRBuilder &B = Helper.MIRBuilder;
   GISelChangeObserver &Observer = Helper.Observer;
 
   Register Dst = MI.getOperand(0).getReg();
   LLT Ty = B.getMRI()->getType(Dst);
   unsigned Size = Ty.getSizeInBits();
   MachineFunction &MF = B.getMF();
 
   Observer.changingInstr(MI);
 
   if (shouldBitcastLoadStoreType(ST, Ty, LLT::scalar(Size))) {
     Ty = getBitcastRegisterType(Ty);
     Helper.bitcastDst(MI, Ty, 0);
     Dst = MI.getOperand(0).getReg();
     B.setInsertPt(B.getMBB(), MI);
   }
 
   // FIXME: We don't really need this intermediate instruction. The intrinsic
   // should be fixed to have a memory operand. Since it's readnone, we're not

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

         B.buildZExtInReg(MI.getOperand(0), WideLoad, MemSize);
       } else
         // We do not need to touch the higher bits for regular loads.
         B.buildLoadFromOffset(MI.getOperand(0), PtrReg, *MMO, 0);
     } else {
       // 96-bit loads are only available for vector loads. We need to split this
       // into a 64-bit part, and 32 (unless we can widen to a 128-bit load).
       if (MMO->getAlign() < Align(16)) {
+        MachineFunction *MF = MI.getParent()->getParent();
+        ApplyRegBankMapping ApplyBank(*this, MRI, DstBank);
+        MachineIRBuilder B(MI, ApplyBank);
+        LegalizerHelper Helper(*MF, ApplyBank, B);
         LLT Part64, Part32;
         std::tie(Part64, Part32) = splitUnequalType(LoadTy, 64);
-        auto Load0 = B.buildLoadFromOffset(Part64, PtrReg, *MMO, 0);
-        auto Load1 = B.buildLoadFromOffset(Part32, PtrReg, *MMO, 8);
-
-        auto Undef = B.buildUndef(LoadTy);
-        auto Ins0 = B.buildInsert(LoadTy, Undef, Load0, 0);
-        B.buildInsert(MI.getOperand(0), Ins0, Load1, 64);
+        if (Helper.reduceLoadStoreWidth(cast<GAnyLoad>(MI), 0, Part64) !=
+            LegalizerHelper::Legalized)
+          return false;
+        return true;
       } else {
         LLT WiderTy = widen96To128(LoadTy);
         auto WideLoad = B.buildLoadFromOffset(WiderTy, PtrReg, *MMO, 0);
-        B.buildExtract(MI.getOperand(0), WideLoad, 0);
+        if (WiderTy.isScalar())
+          B.buildTrunc(MI.getOperand(0), WideLoad);
+        else {
+          B.buildDeleteTrailingVectorElements(MI.getOperand(0).getReg(),
+                                              WideLoad);
+        }
       }
     }
 
     MI.eraseFromParent();
     return true;
   }
 
   // 128-bit loads are supported for all instruction types.

llvm/test/CodeGen/AArch64/GlobalISel/arm64-fallback.ll

 ; RUN: llc -O0 -global-isel -global-isel-abort=2 -pass-remarks-missed='gisel*' -verify-machineinstrs %s -o %t.out 2> %t.err
 ; RUN: FileCheck %s --check-prefix=FALLBACK-WITH-REPORT-OUT < %t.out
 ; RUN: FileCheck %s --check-prefix=FALLBACK-WITH-REPORT-ERR < %t.err
 ; RUN: not --crash llc -global-isel -mtriple aarch64_be %s -o - 2>&1 | FileCheck %s --check-prefix=BIG-ENDIAN
 ; This file checks that the fallback path to selection dag works.
 ; The test is fragile in the sense that it must be updated to expose
 ; something that fails with global-isel.
 ; When we cannot produce a test case anymore, that means we can remove
 ; the fallback path.
 
 target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
 target triple = "aarch64--"
 
 ; BIG-ENDIAN: unable to translate in big endian mode
-; FALLBACK-WITH-REPORT-ERR: remark: <unknown>:0:0: unable to legalize instruction: %{{[0-9]+}}:_(<28 x s32>) = G_CONCAT_VECTORS %{{[0-9]+}}:_(<4 x s32>), %{{[0-9]+}}:_(<4 x s32>), %{{[0-9]+}}:_(<4 x s32>), %{{[0-9]+}}:_(<4 x s32>), %{{[0-9]+}}:_(<4 x s32>), %{{[0-9]+}}:_(<4 x s32>), %{{[0-9]+}}:_(<4 x s32>) (in function: odd_vector)
-; FALLBACK-WITH-REPORT-ERR: warning: Instruction selection used fallback path for odd_vector
-; FALLBACK-WITH-REPORT-OUT-LABEL: odd_vector:
-define void @odd_vector(<7 x i32>* %addr) {
-  %vec = load <7 x i32>, <7 x i32>* %addr
-  store <7 x i32> %vec, <7 x i32>* %addr
-  ret void
-}
 
 ; Make sure we don't mess up metadata arguments.
 declare void @llvm.write_register.i64(metadata, i64)
 
 ; FALLBACK-WITH-REPORT-ERR: remark: <unknown>:0:0: unable to legalize instruction: G_WRITE_REGISTER !0, %0:_(s64) (in function: test_write_register_intrin)
 ; FALLBACK-WITH-REPORT-ERR: warning: Instruction selection used fallback path for test_write_register_intrin
 ; FALLBACK-WITH-REPORT-LABEL: test_write_register_intrin:
 define void @test_write_register_intrin() {

llvm/test/CodeGen/AArch64/GlobalISel/call-lowering-vectors.ll

define void @test_return_v3f32() {
   ; CHECK-LABEL: name: test_return_v3f32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
   ; CHECK-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $sp, implicit $sp
   ; CHECK-NEXT:   $s0 = COPY [[DEF]](s32)
   ; CHECK-NEXT:   BL @bar, csr_aarch64_aapcs, implicit-def $lr, implicit $sp, implicit $s0, implicit-def $q0
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<2 x s64>) = COPY $q0
   ; CHECK-NEXT:   [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[COPY]](<2 x s64>)
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[BITCAST]](<4 x s32>), [[DEF1]](<4 x s32>), [[DEF1]](<4 x s32>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<4 x s32>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
   ; CHECK-NEXT:   ADJCALLSTACKUP 0, 0, implicit-def $sp, implicit $sp
   ; CHECK-NEXT:   RET_ReallyLR
   %call = call <3 x float> @bar(float undef)
   ret void
 }

llvm/test/CodeGen/AArch64/GlobalISel/legalize-saddsat.mir

bb.0:
     ; CHECK-NEXT: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[UADDE]], 24
     ; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s32) = G_ICMP intpred(ne), [[UADDE]](s32), [[SEXT_INREG2]]
     ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s1) = G_TRUNC [[ICMP]](s32)
     ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8), [[UV8:%[0-9]+]]:_(s8), [[UV9:%[0-9]+]]:_(s8), [[UV10:%[0-9]+]]:_(s8), [[UV11:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[UADDO]](s64)
     ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(s8), [[UV13:%[0-9]+]]:_(s8), [[UV14:%[0-9]+]]:_(s8), [[UV15:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[UADDE]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s8) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s8), [[UV17:%[0-9]+]]:_(s8), [[UV18:%[0-9]+]]:_(s8), [[UV19:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[DEF1]](s32)
-    ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(s8), [[UV21:%[0-9]+]]:_(s8), [[UV22:%[0-9]+]]:_(s8), [[UV23:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[DEF1]](s32)
     ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV4]](s8), [[UV5]](s8), [[UV6]](s8), [[UV7]](s8)
     ; CHECK-NEXT: [[MV1:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV8]](s8), [[UV9]](s8), [[UV10]](s8), [[UV11]](s8)
     ; CHECK-NEXT: [[MV2:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV12]](s8), [[UV13]](s8), [[UV14]](s8), [[DEF]](s8)
-    ; CHECK-NEXT: [[MV3:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV16]](s8), [[UV17]](s8), [[UV18]](s8), [[UV20]](s8)
+    ; CHECK-NEXT: [[MV3:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV16]](s8), [[UV17]](s8), [[UV18]](s8), [[UV16]](s8)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 40
     ; CHECK-NEXT: [[MV4:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[MV]](s32), [[MV1]](s32)
     ; CHECK-NEXT: [[MV5:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[MV2]](s32), [[MV3]](s32)
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[MV5]], [[C1]](s64)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[MV4]], [[C2]](s64)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s64) = G_OR [[SHL]], [[LSHR]]
     ; CHECK-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[OR]], [[C1]](s64)

llvm/test/CodeGen/AArch64/GlobalISel/legalize-shuffle-vector.mir

bb.1:
     ; CHECK: liveins: $d0, $d1, $d2, $d3, $d4, $d5, $d6, $d7, $x0
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $d0
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $d1
     ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $d2
     ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY $d3
     ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s64) = COPY $d4
     ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s64) = COPY $d5
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY2]](s64), [[COPY3]](s64)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY4]](s64), [[COPY5]](s64)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY]](s64), [[COPY1]](s64)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY]](s64), [[COPY1]](s64)
     ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s64) = COPY $d6
     ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s64) = COPY $d7
     ; CHECK-NEXT: [[FRAME_INDEX:%[0-9]+]]:_(p0) = G_FRAME_INDEX %fixed-stack.2
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[FRAME_INDEX]](p0) :: (invariant load (s64) from %fixed-stack.2, align 16)
     ; CHECK-NEXT: [[FRAME_INDEX1:%[0-9]+]]:_(p0) = G_FRAME_INDEX %fixed-stack.3
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[FRAME_INDEX1]](p0) :: (invariant load (s64) from %fixed-stack.3)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(p0) = COPY $x0
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY2]](s64), [[COPY3]](s64)
+    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY4]](s64), [[COPY5]](s64)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY]](s64), [[COPY1]](s64)
+    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY]](s64), [[COPY1]](s64)
     ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY6]](s64), [[COPY7]](s64)
     ; CHECK-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[LOAD]](s64), [[LOAD1]](s64)
-    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(p0) = COPY $x0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; CHECK-NEXT: [[EVEC:%[0-9]+]]:_(s64) = G_EXTRACT_VECTOR_ELT [[BUILD_VECTOR]](<2 x s64>), [[C]](s64)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
     ; CHECK-NEXT: [[EVEC1:%[0-9]+]]:_(s64) = G_EXTRACT_VECTOR_ELT [[BUILD_VECTOR1]](<2 x s64>), [[C1]](s64)
     ; CHECK-NEXT: [[EVEC2:%[0-9]+]]:_(s64) = G_EXTRACT_VECTOR_ELT [[BUILD_VECTOR4]](<2 x s64>), [[C]](s64)
     ; CHECK-NEXT: [[EVEC3:%[0-9]+]]:_(s64) = G_EXTRACT_VECTOR_ELT [[BUILD_VECTOR2]](<2 x s64>), [[C1]](s64)
+    ; CHECK-NEXT: [[SHUF:%[0-9]+]]:_(<2 x s64>) = G_SHUFFLE_VECTOR [[BUILD_VECTOR3]](<2 x s64>), [[BUILD_VECTOR5]], shufflemask(1, 3)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[SHUF]](<2 x s64>)
     ; CHECK-NEXT: [[BUILD_VECTOR6:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[EVEC]](s64), [[EVEC1]](s64)
     ; CHECK-NEXT: [[BUILD_VECTOR7:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[EVEC2]](s64), [[EVEC3]](s64)
-    ; CHECK-NEXT: [[SHUF:%[0-9]+]]:_(<2 x s64>) = G_SHUFFLE_VECTOR [[BUILD_VECTOR3]](<2 x s64>), [[BUILD_VECTOR5]], shufflemask(1, 3)
+    ; CHECK-NEXT: [[BUILD_VECTOR8:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64)
     ; CHECK-NEXT: G_STORE [[BUILD_VECTOR6]](<2 x s64>), [[COPY8]](p0) :: (store (<2 x s64>), align 64)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY8]], [[C2]](s64)
     ; CHECK-NEXT: G_STORE [[BUILD_VECTOR7]](<2 x s64>), [[PTR_ADD]](p0) :: (store (<2 x s64>) into unknown-address + 16)
     ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; CHECK-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY8]], [[C3]](s64)
-    ; CHECK-NEXT: G_STORE [[SHUF]](<2 x s64>), [[PTR_ADD1]](p0) :: (store (<2 x s64>) into unknown-address + 32, align 32)
+    ; CHECK-NEXT: G_STORE [[BUILD_VECTOR8]](<2 x s64>), [[PTR_ADD1]](p0) :: (store (<2 x s64>) into unknown-address + 32, align 32)
     ; CHECK-NEXT: RET_ReallyLR
     %3:_(s64) = COPY $d0
     %4:_(s64) = COPY $d1
     %5:_(s64) = COPY $d2
     %6:_(s64) = COPY $d3
     %7:_(s64) = COPY $d4
     %8:_(s64) = COPY $d5
     %0:_(<6 x s64>) = G_BUILD_VECTOR %3(s64), %4(s64), %5(s64), %6(s64), %7(s64), %8(s64)

llvm/test/CodeGen/AArch64/GlobalISel/legalize-ssubsat.mir

bb.0:
     ; CHECK-NEXT: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[USUBE]], 24
     ; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s32) = G_ICMP intpred(ne), [[USUBE]](s32), [[SEXT_INREG2]]
     ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s1) = G_TRUNC [[ICMP]](s32)
     ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8), [[UV8:%[0-9]+]]:_(s8), [[UV9:%[0-9]+]]:_(s8), [[UV10:%[0-9]+]]:_(s8), [[UV11:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[USUBO]](s64)
     ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(s8), [[UV13:%[0-9]+]]:_(s8), [[UV14:%[0-9]+]]:_(s8), [[UV15:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[USUBE]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s8) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s8), [[UV17:%[0-9]+]]:_(s8), [[UV18:%[0-9]+]]:_(s8), [[UV19:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[DEF1]](s32)
-    ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(s8), [[UV21:%[0-9]+]]:_(s8), [[UV22:%[0-9]+]]:_(s8), [[UV23:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[DEF1]](s32)
     ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV4]](s8), [[UV5]](s8), [[UV6]](s8), [[UV7]](s8)
     ; CHECK-NEXT: [[MV1:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV8]](s8), [[UV9]](s8), [[UV10]](s8), [[UV11]](s8)
     ; CHECK-NEXT: [[MV2:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV12]](s8), [[UV13]](s8), [[UV14]](s8), [[DEF]](s8)
-    ; CHECK-NEXT: [[MV3:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV16]](s8), [[UV17]](s8), [[UV18]](s8), [[UV20]](s8)
+    ; CHECK-NEXT: [[MV3:%[0-9]+]]:_(s32) = G_MERGE_VALUES [[UV16]](s8), [[UV17]](s8), [[UV18]](s8), [[UV16]](s8)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 40
     ; CHECK-NEXT: [[MV4:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[MV]](s32), [[MV1]](s32)
     ; CHECK-NEXT: [[MV5:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[MV2]](s32), [[MV3]](s32)
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[MV5]], [[C1]](s64)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 24
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[MV4]], [[C2]](s64)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s64) = G_OR [[SHL]], [[LSHR]]
     ; CHECK-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[OR]], [[C1]](s64)

llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-extract.mir

 name: extract_s64_build_vector_v2s64_s64_s64_offset32
 
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s64_build_vector_v2s64_s64_s64_offset32
     ; CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C1]](s64)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s64) = G_EXTRACT [[BUILD_VECTOR]](<2 x s64>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](s64)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s128) = G_BITCAST [[BUILD_VECTOR]](<2 x s64>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[BITCAST]](s128)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[C2]](s32)
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[C2]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s64) = G_OR [[LSHR]], [[SHL]]
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[OR]](s64)
     %0:_(s64) = G_CONSTANT i64 0
     %1:_(s64) = G_CONSTANT i64 1
     %2:_(<2 x s64>) = G_BUILD_VECTOR %0, %1
     %3:_(s64) = G_EXTRACT %2, 32
     $vgpr0_vgpr1 = COPY %3
 ...
 
 # Test extracting something smaller than the element size

llvm/test/CodeGen/AMDGPU/GlobalISel/artifact-combiner-zext.mir

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2
 
     ; CHECK-LABEL: name: test_zext_trunc_v3s32_to_v3s16_to_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[C]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s32>) = G_CONCAT_VECTORS [[AND]](<2 x s32>), [[AND1]](<2 x s32>), [[DEF1]](<2 x s32>)
-    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<3 x s32>), [[UV4:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s32>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[UV3]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR]], [[BUILD_VECTOR1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[AND1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR2]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>) = G_TRUNC %0
     %2:_(<3 x s32>) = G_ZEXT %1
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 
 # Test for "Too many bits for uint64_t" assertion when combining
 # zexts with wide sources.

llvm/test/CodeGen/AMDGPU/GlobalISel/bug-legalization-artifact-combiner-dead-def.mir

bb.0:
     liveins: $vgpr0, $vgpr1, $vgpr2
 
     ; GFX10-LABEL: name: value_finder_bug_before_artifact_combine
     ; GFX10: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
     ; GFX10-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
     ; GFX10-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
     ; GFX10-NEXT: [[MV:%[0-9]+]]:_(p4) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
     ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[MV]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
-    ; GFX10-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[LOAD]](<4 x s32>), 96
-    ; GFX10-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[LOAD]](<4 x s32>), 64
-    ; GFX10-NEXT: G_STORE [[EXTRACT1]](s32), [[COPY]](p5) :: (store (s32), align 8, addrspace 5)
+    ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; GFX10-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
+    ; GFX10-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; GFX10-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV6]](s32)
+    ; GFX10-NEXT: G_STORE [[COPY4]](s32), [[COPY]](p5) :: (store (s32), align 8, addrspace 5)
     ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
     ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C]](s32)
-    ; GFX10-NEXT: G_STORE [[EXTRACT]](s32), [[PTR_ADD]](p5) :: (store (s32) into unknown-address + 4, addrspace 5)
+    ; GFX10-NEXT: G_STORE [[COPY3]](s32), [[PTR_ADD]](p5) :: (store (s32) into unknown-address + 4, addrspace 5)
     %0:_(p5) = COPY $vgpr0
     %1:_(s32) = COPY $vgpr1
     %2:_(s32) = COPY $vgpr2
     %3:_(p4) = G_MERGE_VALUES %1(s32), %2(s32)
     %4:_(<4 x s32>) = G_LOAD %3(p4) :: (load (<4 x s32>), align 4, addrspace 4)
     %5:_(s32) = G_EXTRACT %4(<4 x s32>), 96
     %6:_(s32) = G_EXTRACT %4(<4 x s32>), 64
     %7:_(s32) = G_IMPLICIT_DEF

llvm/test/CodeGen/AMDGPU/GlobalISel/combine-fma-add-mul.ll

 ; GFX9:       ; %bb.0: ; %.entry
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX9-NEXT:    v_mov_b32_e32 v9, 0xffff
 ; GFX9-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
 ; GFX9-NEXT:    v_and_or_b32 v0, v0, v9, v6
 ; GFX9-NEXT:    v_lshlrev_b32_e32 v6, 16, v7
-; GFX9-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
-; GFX9-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX9-NEXT:    v_and_or_b32 v2, v2, v9, v6
+; GFX9-NEXT:    s_lshl_b32 s4, s4, 16
+; GFX9-NEXT:    v_pk_mul_f16 v0, v0, v2
+; GFX9-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
 ; GFX9-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-NEXT:    v_and_or_b32 v3, v3, v9, s4
-; GFX9-NEXT:    v_pk_mul_f16 v0, v0, v2
-; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v8
+; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GFX9-NEXT:    v_pk_mul_f16 v1, v1, v3
-; GFX9-NEXT:    v_and_or_b32 v2, v4, v9, v2
-; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
-; GFX9-NEXT:    v_and_or_b32 v3, v5, v9, s4
-; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v8
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX9-NEXT:    v_and_or_b32 v3, v4, v9, v3
+; GFX9-NEXT:    v_and_or_b32 v0, v0, v9, v2
+; GFX9-NEXT:    v_pk_add_f16 v0, v3, v0
+; GFX9-NEXT:    v_and_or_b32 v4, v5, v9, s4
 ; GFX9-NEXT:    v_and_or_b32 v1, v1, v9, s4
-; GFX9-NEXT:    v_and_or_b32 v0, v0, v9, v4
-; GFX9-NEXT:    v_pk_add_f16 v1, v3, v1
-; GFX9-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX9-NEXT:    v_pk_add_f16 v0, v2, v0
+; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-NEXT:    v_pk_add_f16 v1, v4, v1
+; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX9-NEXT:    v_and_or_b32 v0, v0, v9, v2
 ; GFX9-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-CONTRACT-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX9-CONTRACT:       ; %bb.0: ; %.entry
 ; GFX9-CONTRACT-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-CONTRACT-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
 ; GFX9-CONTRACT-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX9-CONTRACT-NEXT:    v_mov_b32_e32 v9, 0xffff
 ; GFX9-CONTRACT-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
 ; GFX9-CONTRACT-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
 ; GFX9-CONTRACT-NEXT:    v_and_or_b32 v0, v0, v9, v6
-; GFX9-CONTRACT-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX9-CONTRACT-NEXT:    v_lshlrev_b32_e32 v6, 16, v7
-; GFX9-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-CONTRACT-NEXT:    v_and_or_b32 v2, v2, v9, v6
-; GFX9-CONTRACT-NEXT:    v_and_or_b32 v3, v3, v9, s4
 ; GFX9-CONTRACT-NEXT:    v_lshlrev_b32_e32 v6, 16, v8
-; GFX9-CONTRACT-NEXT:    v_and_or_b32 v5, v5, v9, s4
 ; GFX9-CONTRACT-NEXT:    v_and_or_b32 v4, v4, v9, v6
-; GFX9-CONTRACT-NEXT:    v_pk_fma_f16 v1, v1, v3, v5
 ; GFX9-CONTRACT-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX9-CONTRACT-NEXT:    v_pk_fma_f16 v0, v0, v2, v4
 ; GFX9-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v9, s4
+; GFX9-CONTRACT-NEXT:    v_and_or_b32 v3, v3, v9, s4
+; GFX9-CONTRACT-NEXT:    v_and_or_b32 v5, v5, v9, s4
+; GFX9-CONTRACT-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-CONTRACT-NEXT:    v_pk_fma_f16 v1, v1, v3, v5
+; GFX9-CONTRACT-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX9-CONTRACT-NEXT:    v_and_or_b32 v0, v0, v9, v2
+; GFX9-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-CONTRACT-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-DENORM-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX9-DENORM:       ; %bb.0: ; %.entry
 ; GFX9-DENORM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
 ; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX9-DENORM-NEXT:    v_mov_b32_e32 v9, 0xffff
 ; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
 ; GFX9-DENORM-NEXT:    v_and_or_b32 v0, v0, v9, v6
 ; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v6, 16, v7
-; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
-; GFX9-DENORM-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX9-DENORM-NEXT:    v_and_or_b32 v2, v2, v9, v6
+; GFX9-DENORM-NEXT:    s_lshl_b32 s4, s4, 16
+; GFX9-DENORM-NEXT:    v_pk_mul_f16 v0, v0, v2
+; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
 ; GFX9-DENORM-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-DENORM-NEXT:    v_and_or_b32 v3, v3, v9, s4
-; GFX9-DENORM-NEXT:    v_pk_mul_f16 v0, v0, v2
-; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v2, 16, v8
+; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GFX9-DENORM-NEXT:    v_pk_mul_f16 v1, v1, v3
-; GFX9-DENORM-NEXT:    v_and_or_b32 v2, v4, v9, v2
-; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
-; GFX9-DENORM-NEXT:    v_and_or_b32 v3, v5, v9, s4
-; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v3, 16, v8
+; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX9-DENORM-NEXT:    v_and_or_b32 v3, v4, v9, v3
+; GFX9-DENORM-NEXT:    v_and_or_b32 v0, v0, v9, v2
+; GFX9-DENORM-NEXT:    v_pk_add_f16 v0, v3, v0
+; GFX9-DENORM-NEXT:    v_and_or_b32 v4, v5, v9, s4
 ; GFX9-DENORM-NEXT:    v_and_or_b32 v1, v1, v9, s4
-; GFX9-DENORM-NEXT:    v_and_or_b32 v0, v0, v9, v4
-; GFX9-DENORM-NEXT:    v_pk_add_f16 v1, v3, v1
-; GFX9-DENORM-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX9-DENORM-NEXT:    v_pk_add_f16 v0, v2, v0
+; GFX9-DENORM-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-DENORM-NEXT:    v_pk_add_f16 v1, v4, v1
+; GFX9-DENORM-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX9-DENORM-NEXT:    v_and_or_b32 v0, v0, v9, v2
 ; GFX9-DENORM-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-DENORM-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-UNSAFE-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX9-UNSAFE:       ; %bb.0: ; %.entry
 ; GFX9-UNSAFE-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-UNSAFE-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
 ; GFX9-UNSAFE-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX9-UNSAFE-NEXT:    v_mov_b32_e32 v9, 0xffff
 ; GFX9-UNSAFE-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
 ; GFX9-UNSAFE-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
 ; GFX9-UNSAFE-NEXT:    v_and_or_b32 v0, v0, v9, v6
-; GFX9-UNSAFE-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX9-UNSAFE-NEXT:    v_lshlrev_b32_e32 v6, 16, v7
-; GFX9-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-UNSAFE-NEXT:    v_and_or_b32 v2, v2, v9, v6
-; GFX9-UNSAFE-NEXT:    v_and_or_b32 v3, v3, v9, s4
 ; GFX9-UNSAFE-NEXT:    v_lshlrev_b32_e32 v6, 16, v8
-; GFX9-UNSAFE-NEXT:    v_and_or_b32 v5, v5, v9, s4
 ; GFX9-UNSAFE-NEXT:    v_and_or_b32 v4, v4, v9, v6
-; GFX9-UNSAFE-NEXT:    v_pk_fma_f16 v1, v1, v3, v5
 ; GFX9-UNSAFE-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX9-UNSAFE-NEXT:    v_pk_fma_f16 v0, v0, v2, v4
 ; GFX9-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v9, s4
+; GFX9-UNSAFE-NEXT:    v_and_or_b32 v3, v3, v9, s4
+; GFX9-UNSAFE-NEXT:    v_and_or_b32 v5, v5, v9, s4
+; GFX9-UNSAFE-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX9-UNSAFE-NEXT:    v_pk_fma_f16 v1, v1, v3, v5
+; GFX9-UNSAFE-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX9-UNSAFE-NEXT:    v_and_or_b32 v0, v0, v9, v2
+; GFX9-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v9, s4
 ; GFX9-UNSAFE-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX10:       ; %bb.0: ; %.entry
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v8, 0xffff
 ; GFX10-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX10-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
 ; GFX10-NEXT:    v_lshlrev_b32_e32 v7, 16, v7
 ; GFX10-NEXT:    v_and_or_b32 v1, v1, v8, s4
+; GFX10-NEXT:    v_and_or_b32 v3, v3, v8, s4
 ; GFX10-NEXT:    v_and_or_b32 v0, v0, v8, v6
 ; GFX10-NEXT:    v_and_or_b32 v2, v2, v8, v7
+; GFX10-NEXT:    v_pk_mul_f16 v1, v1, v3
 ; GFX10-NEXT:    v_pk_mul_f16 v0, v0, v2
-; GFX10-NEXT:    v_and_or_b32 v2, v3, v8, s4
-; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 16, v4
+; GFX10-NEXT:    v_lshrrev_b32_e32 v2, 16, v4
+; GFX10-NEXT:    v_and_or_b32 v1, v1, v8, s4
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
-; GFX10-NEXT:    v_pk_mul_f16 v1, v1, v2
+; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX10-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
+; GFX10-NEXT:    v_and_or_b32 v2, v4, v8, v2
+; GFX10-NEXT:    v_and_or_b32 v0, v0, v8, v6
+; GFX10-NEXT:    v_pk_add_f16 v0, v2, v0
 ; GFX10-NEXT:    v_and_or_b32 v2, v5, v8, s4
-; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
-; GFX10-NEXT:    v_lshlrev_b32_e32 v5, 16, v6
-; GFX10-NEXT:    v_and_or_b32 v1, v1, v8, s4
-; GFX10-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX10-NEXT:    v_and_or_b32 v3, v4, v8, v3
-; GFX10-NEXT:    v_and_or_b32 v0, v0, v8, v5
+; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
 ; GFX10-NEXT:    v_pk_add_f16 v1, v2, v1
-; GFX10-NEXT:    v_pk_add_f16 v0, v3, v0
+; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
 ; GFX10-NEXT:    v_and_or_b32 v1, v1, v8, s4
+; GFX10-NEXT:    v_and_or_b32 v0, v0, v8, v3
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-CONTRACT-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX10-CONTRACT:       ; %bb.0: ; %.entry
 ; GFX10-CONTRACT-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-CONTRACT-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-CONTRACT-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
-; GFX10-CONTRACT-NEXT:    v_mov_b32_e32 v7, 0xffff
-; GFX10-CONTRACT-NEXT:    v_lshrrev_b32_e32 v8, 16, v2
-; GFX10-CONTRACT-NEXT:    v_lshrrev_b32_e32 v9, 16, v4
+; GFX10-CONTRACT-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
+; GFX10-CONTRACT-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
+; GFX10-CONTRACT-NEXT:    v_mov_b32_e32 v9, 0xffff
 ; GFX10-CONTRACT-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX10-CONTRACT-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v7, s4
+; GFX10-CONTRACT-NEXT:    v_lshlrev_b32_e32 v7, 16, v7
 ; GFX10-CONTRACT-NEXT:    v_lshlrev_b32_e32 v8, 16, v8
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v3, v3, v7, s4
-; GFX10-CONTRACT-NEXT:    v_lshlrev_b32_e32 v9, 16, v9
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v5, v5, v7, s4
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v0, v0, v7, v6
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v2, v2, v7, v8
-; GFX10-CONTRACT-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v4, v4, v7, v9
-; GFX10-CONTRACT-NEXT:    v_pk_fma_f16 v1, v1, v3, v5
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v9, s4
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v0, v0, v9, v6
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v2, v2, v9, v7
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v4, v4, v9, v8
 ; GFX10-CONTRACT-NEXT:    v_pk_fma_f16 v0, v0, v2, v4
-; GFX10-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v7, s4
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v2, v3, v9, s4
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v4, v5, v9, s4
+; GFX10-CONTRACT-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
+; GFX10-CONTRACT-NEXT:    v_pk_fma_f16 v1, v1, v2, v4
+; GFX10-CONTRACT-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v1, v1, v9, s4
+; GFX10-CONTRACT-NEXT:    v_and_or_b32 v0, v0, v9, v3
 ; GFX10-CONTRACT-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-DENORM-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX10-DENORM:       ; %bb.0: ; %.entry
 ; GFX10-DENORM-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-DENORM-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-DENORM-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
 ; GFX10-DENORM-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX10-DENORM-NEXT:    v_mov_b32_e32 v8, 0xffff
 ; GFX10-DENORM-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
 ; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v7, 16, v7
 ; GFX10-DENORM-NEXT:    v_and_or_b32 v1, v1, v8, s4
+; GFX10-DENORM-NEXT:    v_and_or_b32 v3, v3, v8, s4
 ; GFX10-DENORM-NEXT:    v_and_or_b32 v0, v0, v8, v6
 ; GFX10-DENORM-NEXT:    v_and_or_b32 v2, v2, v8, v7
+; GFX10-DENORM-NEXT:    v_pk_mul_f16 v1, v1, v3
 ; GFX10-DENORM-NEXT:    v_pk_mul_f16 v0, v0, v2
-; GFX10-DENORM-NEXT:    v_and_or_b32 v2, v3, v8, s4
-; GFX10-DENORM-NEXT:    v_lshrrev_b32_e32 v3, 16, v4
+; GFX10-DENORM-NEXT:    v_lshrrev_b32_e32 v2, 16, v4
+; GFX10-DENORM-NEXT:    v_and_or_b32 v1, v1, v8, s4
 ; GFX10-DENORM-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
-; GFX10-DENORM-NEXT:    v_pk_mul_f16 v1, v1, v2
+; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
+; GFX10-DENORM-NEXT:    v_and_or_b32 v2, v4, v8, v2
+; GFX10-DENORM-NEXT:    v_and_or_b32 v0, v0, v8, v6
+; GFX10-DENORM-NEXT:    v_pk_add_f16 v0, v2, v0
 ; GFX10-DENORM-NEXT:    v_and_or_b32 v2, v5, v8, s4
-; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
-; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v5, 16, v6
-; GFX10-DENORM-NEXT:    v_and_or_b32 v1, v1, v8, s4
-; GFX10-DENORM-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX10-DENORM-NEXT:    v_and_or_b32 v3, v4, v8, v3
-; GFX10-DENORM-NEXT:    v_and_or_b32 v0, v0, v8, v5
+; GFX10-DENORM-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
 ; GFX10-DENORM-NEXT:    v_pk_add_f16 v1, v2, v1
-; GFX10-DENORM-NEXT:    v_pk_add_f16 v0, v3, v0
+; GFX10-DENORM-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
 ; GFX10-DENORM-NEXT:    v_and_or_b32 v1, v1, v8, s4
+; GFX10-DENORM-NEXT:    v_and_or_b32 v0, v0, v8, v3
 ; GFX10-DENORM-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-UNSAFE-LABEL: test_3xhalf_add_mul_rhs:
 ; GFX10-UNSAFE:       ; %bb.0: ; %.entry
 ; GFX10-UNSAFE-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-UNSAFE-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-UNSAFE-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
-; GFX10-UNSAFE-NEXT:    v_mov_b32_e32 v7, 0xffff
-; GFX10-UNSAFE-NEXT:    v_lshrrev_b32_e32 v8, 16, v2
-; GFX10-UNSAFE-NEXT:    v_lshrrev_b32_e32 v9, 16, v4
+; GFX10-UNSAFE-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
+; GFX10-UNSAFE-NEXT:    v_lshrrev_b32_e32 v8, 16, v4
+; GFX10-UNSAFE-NEXT:    v_mov_b32_e32 v9, 0xffff
 ; GFX10-UNSAFE-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX10-UNSAFE-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v7, s4
+; GFX10-UNSAFE-NEXT:    v_lshlrev_b32_e32 v7, 16, v7
 ; GFX10-UNSAFE-NEXT:    v_lshlrev_b32_e32 v8, 16, v8
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v3, v3, v7, s4
-; GFX10-UNSAFE-NEXT:    v_lshlrev_b32_e32 v9, 16, v9
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v5, v5, v7, s4
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v0, v0, v7, v6
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v2, v2, v7, v8
-; GFX10-UNSAFE-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v4, v4, v7, v9
-; GFX10-UNSAFE-NEXT:    v_pk_fma_f16 v1, v1, v3, v5
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v9, s4
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v0, v0, v9, v6
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v2, v2, v9, v7
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v4, v4, v9, v8
 ; GFX10-UNSAFE-NEXT:    v_pk_fma_f16 v0, v0, v2, v4
-; GFX10-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v7, s4
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v2, v3, v9, s4
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v4, v5, v9, s4
+; GFX10-UNSAFE-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
+; GFX10-UNSAFE-NEXT:    v_pk_fma_f16 v1, v1, v2, v4
+; GFX10-UNSAFE-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v1, v1, v9, s4
+; GFX10-UNSAFE-NEXT:    v_and_or_b32 v0, v0, v9, v3
 ; GFX10-UNSAFE-NEXT:    s_setpc_b64 s[30:31]
 .entry:
   %a = fmul <3 x half> %x, %y
   %b = fadd <3 x half> %z, %a
   ret <3 x half> %b
 }
 
 define <4 x double> @test_4xdouble_add_mul(<4 x double> %x, <4 x double> %y, <4 x double> %z) {

llvm/test/CodeGen/AMDGPU/GlobalISel/function-returns.ll

 define <3 x i16> @v3i16_func_void() #0 {
   ; CHECK-LABEL: name: v3i16_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $sgpr30_sgpr31
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(<3 x s16>) = G_LOAD [[DEF]](p1) :: (load (<3 x s16>) from `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<3 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[LOAD]](<3 x s16>), [[DEF1]](<3 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<3 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s16>)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV3]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV4]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK-NEXT:   S_SETPC_B64_return [[COPY1]], implicit $vgpr0, implicit $vgpr1
   %val = load <3 x i16>, <3 x i16> addrspace(1)* undef
   ret <3 x i16> %val
 }
 
 define <4 x i16> @v4i16_func_void() #0 {
   ; CHECK-LABEL: name: v4i16_func_void

   ; CHECK-LABEL: name: v5i16_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $sgpr30_sgpr31
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (volatile load (p1) from `<5 x i16> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:_(<5 x s16>) = G_LOAD [[LOAD]](p1) :: (load (<5 x s16>) from %ir.ptr, align 16, addrspace 1)
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<5 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[LOAD1]](<5 x s16>), [[DEF1]](<5 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s16>)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr2 = COPY [[UV2]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD1]](<5 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<6 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<6 x s16>)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV5]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV6]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr2 = COPY [[UV7]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK-NEXT:   S_SETPC_B64_return [[COPY1]], implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
   %ptr = load volatile <5 x i16> addrspace(1)*, <5 x i16> addrspace(1)* addrspace(4)* undef
   %val = load <5 x i16>, <5 x i16> addrspace(1)* %ptr
   ret <5 x i16> %val
 }
 
 define <8 x i16> @v8i16_func_void() #0 {

llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-extract.mir

bb.0:
     ; CHECK: [[COPY3:%[0-9]+]]:sgpr_96 = COPY [[COPY]].sub1_sub2_sub3
     ; CHECK: S_ENDPGM 0, implicit [[COPY2]], implicit [[COPY3]]
     %0:sgpr(s128) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:sgpr(s96) = G_EXTRACT %0, 0
     %2:sgpr(s96) = G_EXTRACT %0, 32
     S_ENDPGM 0, implicit %1, implicit %2
 
 ...
-
----
-name:            extract_sgpr_s32_from_v3s32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1_sgpr2
-    ; CHECK-LABEL: name: extract_sgpr_s32_from_v3s32
-    ; CHECK: [[COPY:%[0-9]+]]:sgpr_96 = COPY $sgpr0_sgpr1_sgpr2
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1
-    ; CHECK: [[COPY2:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub2
-    ; CHECK: S_ENDPGM 0, implicit [[COPY]], implicit [[COPY1]], implicit [[COPY2]]
-    %0:sgpr(<3 x s32>) = COPY $sgpr0_sgpr1_sgpr2
-    %1:sgpr(s32) = G_EXTRACT %0, 0
-    %2:sgpr(s32) = G_EXTRACT %0, 32
-    %3:sgpr(s32) = G_EXTRACT %0, 64
-    S_ENDPGM 0, implicit %0, implicit %2, implicit %3
-
-...
-
----
-name:            extract_sgpr_v2s32_from_v3s32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1_sgpr2
-    ; CHECK-LABEL: name: extract_sgpr_v2s32_from_v3s32
-    ; CHECK: [[COPY:%[0-9]+]]:sgpr_96_with_sub0_sub1 = COPY $sgpr0_sgpr1_sgpr2
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_64 = COPY [[COPY]].sub0_sub1
-    ; CHECK: S_ENDPGM 0, implicit [[COPY1]]
-    %0:sgpr(<3 x s32>) = COPY $sgpr0_sgpr1_sgpr2
-    %1:sgpr(<2 x s32>) = G_EXTRACT %0, 0
-    S_ENDPGM 0, implicit %1
-
-...
-
----
-name:            extract_sgpr_v3s32_from_v4s32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1_sgpr2_sgpr3
-    ; CHECK-LABEL: name: extract_sgpr_v3s32_from_v4s32
-    ; CHECK: [[COPY:%[0-9]+]]:sgpr_128_with_sub1_sub2_sub3 = COPY $sgpr0_sgpr1_sgpr2_sgpr3
-    ; CHECK: [[COPY1:%[0-9]+]]:sgpr_128_with_sub0_sub1_sub2 = COPY [[COPY]]
-    ; CHECK: [[COPY2:%[0-9]+]]:sgpr_96 = COPY [[COPY1]].sub0_sub1_sub2
-    ; CHECK: [[COPY3:%[0-9]+]]:sgpr_96 = COPY [[COPY]].sub1_sub2_sub3
-    ; CHECK: S_ENDPGM 0, implicit [[COPY2]], implicit [[COPY3]]
-    %0:sgpr(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
-    %1:sgpr(<3 x s32>) = G_EXTRACT %0, 0
-    %2:sgpr(<3 x s32>) = G_EXTRACT %0, 32
-    S_ENDPGM 0, implicit %1, implicit %2
-
-...
-
----
-name:            extract_sgpr_v2s16_from_v4s16_offset0
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1
-    ; CHECK-LABEL: name: extract_sgpr_v2s16_from_v4s16_offset0
-    ; CHECK: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub0
-    ; CHECK: S_ENDPGM 0, implicit [[COPY1]]
-    %0:sgpr(<4 x s16>) = COPY $sgpr0_sgpr1
-    %1:sgpr(<2 x s16>) = G_EXTRACT %0, 0
-    S_ENDPGM 0, implicit %1
-
-...
-
----
-name:            extract_sgpr_v2s16_from_v4s16_offset32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1
-    ; CHECK-LABEL: name: extract_sgpr_v2s16_from_v4s16_offset32
-    ; CHECK: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1
-    ; CHECK: S_ENDPGM 0, implicit [[COPY1]]
-    %0:sgpr(<4 x s16>) = COPY $sgpr0_sgpr1
-    %1:sgpr(<2 x s16>) = G_EXTRACT %0, 32
-    S_ENDPGM 0, implicit %1
-
-...
-
-# FIXME: Probably should not be legal
----
-name:            extract_sgpr_s16_from_v4s16_offset0
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1
-    ; CHECK-LABEL: name: extract_sgpr_s16_from_v4s16_offset0
-    ; CHECK: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub0
-    ; CHECK: S_ENDPGM 0, implicit [[COPY1]]
-    %0:sgpr(<4 x s16>) = COPY $sgpr0_sgpr1
-    %1:sgpr(s16) = G_EXTRACT %0, 0
-    S_ENDPGM 0, implicit %1
-
-...
-
-# FIXME: Probably should not be legal
----
-name:            extract_sgpr_s16_from_v4s16_offset32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1
-    ; CHECK-LABEL: name: extract_sgpr_s16_from_v4s16_offset32
-    ; CHECK: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1
-    ; CHECK: S_ENDPGM 0, implicit [[COPY1]]
-    %0:sgpr(<4 x s16>) = COPY $sgpr0_sgpr1
-    %1:sgpr(s16) = G_EXTRACT %0, 32
-    S_ENDPGM 0, implicit %1
-
-...
-
-# FIXME: Probably should not be legal
----
-name:            extract_sgpr_s16_from_v6s16_offset32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1_sgpr2
-    ; CHECK-LABEL: name: extract_sgpr_s16_from_v6s16_offset32
-    ; CHECK: [[COPY:%[0-9]+]]:sgpr_96 = COPY $sgpr0_sgpr1_sgpr2
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY [[COPY]].sub1
-    ; CHECK: S_ENDPGM 0, implicit [[COPY1]]
-    %0:sgpr(<6 x s16>) = COPY $sgpr0_sgpr1_sgpr2
-    %1:sgpr(s16) = G_EXTRACT %0, 32
-    S_ENDPGM 0, implicit %1
-
-...

llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-fract.f64.mir

bb.1:
     ; CHECK: %15:vreg_64 = nofpexcept V_FRACT_F64_e64 0, %12, 0, 0, implicit $mode, implicit $exec
     ; CHECK: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
     ; CHECK: GLOBAL_STORE_DWORDX2_SADDR [[V_MOV_B32_e32_]], %15, [[COPY1]], 0, 0, implicit $exec :: (store (s64), addrspace 1)
     ; CHECK: S_ENDPGM 0
     %2:sgpr(p4) = COPY $sgpr0_sgpr1
     %7:sgpr(s64) = G_CONSTANT i64 36
     %8:sgpr(p4) = G_PTR_ADD %2, %7(s64)
     %9:sgpr(<2 x s64>) = G_LOAD %8(p4) :: (dereferenceable invariant load (<2 x s64>), align 4, addrspace 4)
-    %10:sgpr(s64) = G_EXTRACT %9(<2 x s64>), 0
-    %13:sgpr(s64) = G_EXTRACT %9(<2 x s64>), 64
+    %10:sgpr(s64), %13:sgpr(s64) = G_UNMERGE_VALUES %9(<2 x s64>)
     %15:sgpr(p1) = G_INTTOPTR %13(s64)
     %18:sgpr(s64) = G_LOAD %15(p1) :: (load (s64), addrspace 1)
     %19:sgpr(s64) = G_FCONSTANT double -0.000000e+00
     %24:sgpr(s64) = G_FNEG %18
     %25:vgpr(s64) = COPY %19(s64)
     %26:vgpr(s64) = COPY %24(s64)
     %20:vgpr(s64) = G_FADD %25, %26
     %21:vgpr(s64) = G_FFLOOR %20
bb.1:
     ; CHECK: %16:vreg_64 = nofpexcept V_FRACT_F64_e64 0, %13, 0, 0, implicit $mode, implicit $exec
     ; CHECK: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
     ; CHECK: GLOBAL_STORE_DWORDX2_SADDR [[V_MOV_B32_e32_]], %16, [[COPY1]], 0, 0, implicit $exec :: (store (s64), addrspace 1)
     ; CHECK: S_ENDPGM 0
     %2:sgpr(p4) = COPY $sgpr0_sgpr1
     %7:sgpr(s64) = G_CONSTANT i64 36
     %8:sgpr(p4) = G_PTR_ADD %2, %7(s64)
     %9:sgpr(<2 x s64>) = G_LOAD %8(p4) :: (dereferenceable invariant load (<2 x s64>), align 4, addrspace 4)
-    %10:sgpr(s64) = G_EXTRACT %9(<2 x s64>), 0
-    %13:sgpr(s64) = G_EXTRACT %9(<2 x s64>), 64
+    %10:sgpr(s64), %13:sgpr(s64) = G_UNMERGE_VALUES %9(<2 x s64>)
     %15:sgpr(p1) = G_INTTOPTR %13(s64)
     %18:sgpr(s64) = G_LOAD %15(p1) :: (load (s64), addrspace 1)
     %19:sgpr(s64) = G_FABS %18
     %20:sgpr(s64) = G_FCONSTANT double -0.000000e+00
     %25:sgpr(s64) = G_FNEG %19
     %26:vgpr(s64) = COPY %20(s64)
     %27:vgpr(s64) = COPY %25(s64)
     %21:vgpr(s64) = G_FADD %26, %27
     %22:vgpr(s64) = G_FFLOOR %21
     %24:vgpr(s64) = G_FNEG %22
     %23:vgpr(s64) = G_FADD %21, %24
     %12:sgpr(p1) = G_INTTOPTR %10(s64)
     %28:vgpr(p1) = COPY %12(p1)
     G_STORE %23(s64), %28(p1) :: (store (s64), addrspace 1)
     S_ENDPGM 0
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-insert.mir

     ; CHECK: [[COPY1:%[0-9]+]]:vreg_128 = COPY $vgpr8_vgpr9_vgpr10_vgpr11
     ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:vreg_256 = INSERT_SUBREG [[COPY]], [[COPY1]], %subreg.sub4_sub5_sub6_sub7
     ; CHECK: S_ENDPGM 0, implicit [[INSERT_SUBREG]]
     %0:vgpr(s256) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:vgpr(s128) = COPY $vgpr8_vgpr9_vgpr10_vgpr11
     %2:vgpr(s256) = G_INSERT %0, %1, 128
     S_ENDPGM 0, implicit %2
 ...
-
----
-name:            insert_sgpr_v2s16_to_v4s16_offset0
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1, $sgpr2
-    ; CHECK-LABEL: name: insert_sgpr_v2s16_to_v4s16_offset0
-    ; CHECK: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
-    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:sreg_64 = INSERT_SUBREG [[COPY]], [[COPY1]], %subreg.sub0
-    ; CHECK: S_ENDPGM 0, implicit [[INSERT_SUBREG]]
-    %0:sgpr(<4 x s16>) = COPY $sgpr0_sgpr1
-    %1:sgpr(<2 x s16>) = COPY $sgpr2
-    %2:sgpr(<4 x s16>) = G_INSERT %0, %1, 0
-    S_ENDPGM 0, implicit %2
-
-...
-
----
-name:            insert_sgpr_v2s16_to_v4s16_offset32
-legalized:       true
-regBankSelected: true
-
-body: |
-  bb.0:
-    liveins: $sgpr0_sgpr1, $sgpr2
-    ; CHECK-LABEL: name: insert_sgpr_v2s16_to_v4s16_offset32
-    ; CHECK: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr0_sgpr1
-    ; CHECK: [[COPY1:%[0-9]+]]:sreg_32 = COPY $sgpr2
-    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:sreg_64 = INSERT_SUBREG [[COPY]], [[COPY1]], %subreg.sub1
-    ; CHECK: S_ENDPGM 0, implicit [[INSERT_SUBREG]]
-    %0:sgpr(<4 x s16>) = COPY $sgpr0_sgpr1
-    %1:sgpr(<2 x s16>) = COPY $sgpr2
-    %2:sgpr(<4 x s16>) = G_INSERT %0, %1, 32
-    S_ENDPGM 0, implicit %2
-...

llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-insert.xfail.mir

 # RUN: llc -march=amdgcn -mcpu=gfx900 -run-pass=instruction-select -global-isel-abort=2 -pass-remarks-missed='gisel*' -verify-machineinstrs -o /dev/null %s 2>&1 | FileCheck -check-prefix=ERR %s
 
-# ERR: remark: <unknown>:0:0: cannot select: %3:sgpr(<4 x s16>) = G_INSERT %0:sgpr, %2:sgpr(s16), 0 (in function: insert_sgpr_s16_to_v4s16_offset0)
-# ERR-NEXT: remark: <unknown>:0:0: cannot select: %2:sgpr(<16 x s32>) = G_INSERT %0:sgpr, %1:sgpr(<8 x s32>), 0 (in function: insert_sgpr_v8s32_to_v16s32_offset0)
+# ERR: remark: <unknown>:0:0: instruction is not legal: %3:sgpr(<4 x s16>) = G_INSERT %0:sgpr, %2:sgpr(s16), 0 (in function: insert_sgpr_s16_to_v4s16_offset0)
+# ERR-NEXT: <unknown>:0:0: instruction is not legal: %2:sgpr(<16 x s32>) = G_INSERT %0:sgpr, %1:sgpr(<8 x s32>), 0 (in function: insert_sgpr_v8s32_to_v16s32_offset0)
 # ERR-NOT: remark
 
 # FIXME: This 16-bit insert source should not be legal and this test
 # should be deleted
 ---
 name:            insert_sgpr_s16_to_v4s16_offset0
 legalized:       true
 regBankSelected: true

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll

   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
   ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-  ; GCN-NEXT:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>), [[DEF1]](<2 x s16>)
-  ; GCN-NEXT:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
+  ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>)
+  ; GCN-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
+  ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
-  ; GCN-NEXT:   G_STORE [[UV]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
+  ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <3 x i16> @external_v3i16_func_void()
   store volatile <3 x i16> %val, <3 x i16> addrspace(1)* undef
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_v4i16_func_void() #0 {
   ; GCN-LABEL: name: test_call_external_v4i16_func_void

   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN-NEXT:   $vgpr31 = COPY [[OR1]](s32)
   ; GCN-NEXT:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN-NEXT:   [[COPY21:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY22:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-  ; GCN-NEXT:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>), [[DEF1]](<2 x s16>)
-  ; GCN-NEXT:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
+  ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>)
+  ; GCN-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
+  ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; GCN-NEXT:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
-  ; GCN-NEXT:   G_STORE [[UV]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
+  ; GCN-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN-NEXT:   S_ENDPGM 0
   %val = call <3 x half> @external_v3f16_func_void()
   store volatile <3 x half> %val, <3 x half> addrspace(1)* undef
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_v4f16_func_void() #0 {
   ; GCN-LABEL: name: test_call_external_v4f16_func_void

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call.ll

   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<3 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[LOAD]](<3 x s16>), [[DEF1]](<3 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<3 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s16>)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV3]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV4]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)

   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<3 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[LOAD]](<3 x s16>), [[DEF1]](<3 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<3 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s16>)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV3]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV4]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)

   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<5 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[LOAD]](<5 x s16>), [[DEF1]](<5 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s16>)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr2 = COPY [[UV2]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<5 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<6 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<6 x s16>)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV5]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV6]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr2 = COPY [[UV7]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)

   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<7 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<14 x s16>) = G_CONCAT_VECTORS [[LOAD]](<7 x s16>), [[DEF1]](<7 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<14 x s16>)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr2 = COPY [[UV2]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr3 = COPY [[UV3]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<7 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16), [[UV6]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s16>)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV7]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV8]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr2 = COPY [[UV9]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr3 = COPY [[UV10]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)

   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<63 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<126 x s16>) = G_CONCAT_VECTORS [[LOAD]](<63 x s16>), [[DEF1]](<63 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>), [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>), [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>), [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>), [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>), [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>), [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>), [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<126 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16), [[UV8:%[0-9]+]]:_(s16), [[UV9:%[0-9]+]]:_(s16), [[UV10:%[0-9]+]]:_(s16), [[UV11:%[0-9]+]]:_(s16), [[UV12:%[0-9]+]]:_(s16), [[UV13:%[0-9]+]]:_(s16), [[UV14:%[0-9]+]]:_(s16), [[UV15:%[0-9]+]]:_(s16), [[UV16:%[0-9]+]]:_(s16), [[UV17:%[0-9]+]]:_(s16), [[UV18:%[0-9]+]]:_(s16), [[UV19:%[0-9]+]]:_(s16), [[UV20:%[0-9]+]]:_(s16), [[UV21:%[0-9]+]]:_(s16), [[UV22:%[0-9]+]]:_(s16), [[UV23:%[0-9]+]]:_(s16), [[UV24:%[0-9]+]]:_(s16), [[UV25:%[0-9]+]]:_(s16), [[UV26:%[0-9]+]]:_(s16), [[UV27:%[0-9]+]]:_(s16), [[UV28:%[0-9]+]]:_(s16), [[UV29:%[0-9]+]]:_(s16), [[UV30:%[0-9]+]]:_(s16), [[UV31:%[0-9]+]]:_(s16), [[UV32:%[0-9]+]]:_(s16), [[UV33:%[0-9]+]]:_(s16), [[UV34:%[0-9]+]]:_(s16), [[UV35:%[0-9]+]]:_(s16), [[UV36:%[0-9]+]]:_(s16), [[UV37:%[0-9]+]]:_(s16), [[UV38:%[0-9]+]]:_(s16), [[UV39:%[0-9]+]]:_(s16), [[UV40:%[0-9]+]]:_(s16), [[UV41:%[0-9]+]]:_(s16), [[UV42:%[0-9]+]]:_(s16), [[UV43:%[0-9]+]]:_(s16), [[UV44:%[0-9]+]]:_(s16), [[UV45:%[0-9]+]]:_(s16), [[UV46:%[0-9]+]]:_(s16), [[UV47:%[0-9]+]]:_(s16), [[UV48:%[0-9]+]]:_(s16), [[UV49:%[0-9]+]]:_(s16), [[UV50:%[0-9]+]]:_(s16), [[UV51:%[0-9]+]]:_(s16), [[UV52:%[0-9]+]]:_(s16), [[UV53:%[0-9]+]]:_(s16), [[UV54:%[0-9]+]]:_(s16), [[UV55:%[0-9]+]]:_(s16), [[UV56:%[0-9]+]]:_(s16), [[UV57:%[0-9]+]]:_(s16), [[UV58:%[0-9]+]]:_(s16), [[UV59:%[0-9]+]]:_(s16), [[UV60:%[0-9]+]]:_(s16), [[UV61:%[0-9]+]]:_(s16), [[UV62:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<63 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<64 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16), [[UV6]](s16), [[UV7]](s16), [[UV8]](s16), [[UV9]](s16), [[UV10]](s16), [[UV11]](s16), [[UV12]](s16), [[UV13]](s16), [[UV14]](s16), [[UV15]](s16), [[UV16]](s16), [[UV17]](s16), [[UV18]](s16), [[UV19]](s16), [[UV20]](s16), [[UV21]](s16), [[UV22]](s16), [[UV23]](s16), [[UV24]](s16), [[UV25]](s16), [[UV26]](s16), [[UV27]](s16), [[UV28]](s16), [[UV29]](s16), [[UV30]](s16), [[UV31]](s16), [[UV32]](s16), [[UV33]](s16), [[UV34]](s16), [[UV35]](s16), [[UV36]](s16), [[UV37]](s16), [[UV38]](s16), [[UV39]](s16), [[UV40]](s16), [[UV41]](s16), [[UV42]](s16), [[UV43]](s16), [[UV44]](s16), [[UV45]](s16), [[UV46]](s16), [[UV47]](s16), [[UV48]](s16), [[UV49]](s16), [[UV50]](s16), [[UV51]](s16), [[UV52]](s16), [[UV53]](s16), [[UV54]](s16), [[UV55]](s16), [[UV56]](s16), [[UV57]](s16), [[UV58]](s16), [[UV59]](s16), [[UV60]](s16), [[UV61]](s16), [[UV62]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>), [[UV65:%[0-9]+]]:_(<2 x s16>), [[UV66:%[0-9]+]]:_(<2 x s16>), [[UV67:%[0-9]+]]:_(<2 x s16>), [[UV68:%[0-9]+]]:_(<2 x s16>), [[UV69:%[0-9]+]]:_(<2 x s16>), [[UV70:%[0-9]+]]:_(<2 x s16>), [[UV71:%[0-9]+]]:_(<2 x s16>), [[UV72:%[0-9]+]]:_(<2 x s16>), [[UV73:%[0-9]+]]:_(<2 x s16>), [[UV74:%[0-9]+]]:_(<2 x s16>), [[UV75:%[0-9]+]]:_(<2 x s16>), [[UV76:%[0-9]+]]:_(<2 x s16>), [[UV77:%[0-9]+]]:_(<2 x s16>), [[UV78:%[0-9]+]]:_(<2 x s16>), [[UV79:%[0-9]+]]:_(<2 x s16>), [[UV80:%[0-9]+]]:_(<2 x s16>), [[UV81:%[0-9]+]]:_(<2 x s16>), [[UV82:%[0-9]+]]:_(<2 x s16>), [[UV83:%[0-9]+]]:_(<2 x s16>), [[UV84:%[0-9]+]]:_(<2 x s16>), [[UV85:%[0-9]+]]:_(<2 x s16>), [[UV86:%[0-9]+]]:_(<2 x s16>), [[UV87:%[0-9]+]]:_(<2 x s16>), [[UV88:%[0-9]+]]:_(<2 x s16>), [[UV89:%[0-9]+]]:_(<2 x s16>), [[UV90:%[0-9]+]]:_(<2 x s16>), [[UV91:%[0-9]+]]:_(<2 x s16>), [[UV92:%[0-9]+]]:_(<2 x s16>), [[UV93:%[0-9]+]]:_(<2 x s16>), [[UV94:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<64 x s16>)
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(p5) = COPY $sp_reg
   ; CHECK-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; CHECK-NEXT:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C3]](s32)
-  ; CHECK-NEXT:   G_STORE [[UV31]](<2 x s16>), [[PTR_ADD1]](p5) :: (store (<2 x s16>) into stack, align 16, addrspace 5)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr2 = COPY [[UV2]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr3 = COPY [[UV3]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr4 = COPY [[UV4]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr5 = COPY [[UV5]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr6 = COPY [[UV6]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr7 = COPY [[UV7]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr8 = COPY [[UV8]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr9 = COPY [[UV9]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr10 = COPY [[UV10]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr11 = COPY [[UV11]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr12 = COPY [[UV12]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr13 = COPY [[UV13]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr14 = COPY [[UV14]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr15 = COPY [[UV15]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr16 = COPY [[UV16]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr17 = COPY [[UV17]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr18 = COPY [[UV18]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr19 = COPY [[UV19]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr20 = COPY [[UV20]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr21 = COPY [[UV21]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr22 = COPY [[UV22]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr23 = COPY [[UV23]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr24 = COPY [[UV24]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr25 = COPY [[UV25]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr26 = COPY [[UV26]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr27 = COPY [[UV27]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr28 = COPY [[UV28]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr29 = COPY [[UV29]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr30 = COPY [[UV30]](<2 x s16>)
+  ; CHECK-NEXT:   G_STORE [[UV94]](<2 x s16>), [[PTR_ADD1]](p5) :: (store (<2 x s16>) into stack, align 16, addrspace 5)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV63]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV64]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr2 = COPY [[UV65]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr3 = COPY [[UV66]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr4 = COPY [[UV67]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr5 = COPY [[UV68]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr6 = COPY [[UV69]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr7 = COPY [[UV70]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr8 = COPY [[UV71]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr9 = COPY [[UV72]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr10 = COPY [[UV73]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr11 = COPY [[UV74]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr12 = COPY [[UV75]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr13 = COPY [[UV76]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr14 = COPY [[UV77]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr15 = COPY [[UV78]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr16 = COPY [[UV79]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr17 = COPY [[UV80]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr18 = COPY [[UV81]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr19 = COPY [[UV82]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr20 = COPY [[UV83]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr21 = COPY [[UV84]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr22 = COPY [[UV85]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr23 = COPY [[UV86]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr24 = COPY [[UV87]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr25 = COPY [[UV88]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr26 = COPY [[UV89]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr27 = COPY [[UV90]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr28 = COPY [[UV91]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr29 = COPY [[UV92]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr30 = COPY [[UV93]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)

   ; CHECK-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
   ; CHECK-NEXT:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
   ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<65 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<130 x s16>) = G_CONCAT_VECTORS [[LOAD]](<65 x s16>), [[DEF1]](<65 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>), [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>), [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>), [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>), [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>), [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>), [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>), [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>), [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<130 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16), [[UV8:%[0-9]+]]:_(s16), [[UV9:%[0-9]+]]:_(s16), [[UV10:%[0-9]+]]:_(s16), [[UV11:%[0-9]+]]:_(s16), [[UV12:%[0-9]+]]:_(s16), [[UV13:%[0-9]+]]:_(s16), [[UV14:%[0-9]+]]:_(s16), [[UV15:%[0-9]+]]:_(s16), [[UV16:%[0-9]+]]:_(s16), [[UV17:%[0-9]+]]:_(s16), [[UV18:%[0-9]+]]:_(s16), [[UV19:%[0-9]+]]:_(s16), [[UV20:%[0-9]+]]:_(s16), [[UV21:%[0-9]+]]:_(s16), [[UV22:%[0-9]+]]:_(s16), [[UV23:%[0-9]+]]:_(s16), [[UV24:%[0-9]+]]:_(s16), [[UV25:%[0-9]+]]:_(s16), [[UV26:%[0-9]+]]:_(s16), [[UV27:%[0-9]+]]:_(s16), [[UV28:%[0-9]+]]:_(s16), [[UV29:%[0-9]+]]:_(s16), [[UV30:%[0-9]+]]:_(s16), [[UV31:%[0-9]+]]:_(s16), [[UV32:%[0-9]+]]:_(s16), [[UV33:%[0-9]+]]:_(s16), [[UV34:%[0-9]+]]:_(s16), [[UV35:%[0-9]+]]:_(s16), [[UV36:%[0-9]+]]:_(s16), [[UV37:%[0-9]+]]:_(s16), [[UV38:%[0-9]+]]:_(s16), [[UV39:%[0-9]+]]:_(s16), [[UV40:%[0-9]+]]:_(s16), [[UV41:%[0-9]+]]:_(s16), [[UV42:%[0-9]+]]:_(s16), [[UV43:%[0-9]+]]:_(s16), [[UV44:%[0-9]+]]:_(s16), [[UV45:%[0-9]+]]:_(s16), [[UV46:%[0-9]+]]:_(s16), [[UV47:%[0-9]+]]:_(s16), [[UV48:%[0-9]+]]:_(s16), [[UV49:%[0-9]+]]:_(s16), [[UV50:%[0-9]+]]:_(s16), [[UV51:%[0-9]+]]:_(s16), [[UV52:%[0-9]+]]:_(s16), [[UV53:%[0-9]+]]:_(s16), [[UV54:%[0-9]+]]:_(s16), [[UV55:%[0-9]+]]:_(s16), [[UV56:%[0-9]+]]:_(s16), [[UV57:%[0-9]+]]:_(s16), [[UV58:%[0-9]+]]:_(s16), [[UV59:%[0-9]+]]:_(s16), [[UV60:%[0-9]+]]:_(s16), [[UV61:%[0-9]+]]:_(s16), [[UV62:%[0-9]+]]:_(s16), [[UV63:%[0-9]+]]:_(s16), [[UV64:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[LOAD]](<65 x s16>)
+  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<66 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16), [[UV6]](s16), [[UV7]](s16), [[UV8]](s16), [[UV9]](s16), [[UV10]](s16), [[UV11]](s16), [[UV12]](s16), [[UV13]](s16), [[UV14]](s16), [[UV15]](s16), [[UV16]](s16), [[UV17]](s16), [[UV18]](s16), [[UV19]](s16), [[UV20]](s16), [[UV21]](s16), [[UV22]](s16), [[UV23]](s16), [[UV24]](s16), [[UV25]](s16), [[UV26]](s16), [[UV27]](s16), [[UV28]](s16), [[UV29]](s16), [[UV30]](s16), [[UV31]](s16), [[UV32]](s16), [[UV33]](s16), [[UV34]](s16), [[UV35]](s16), [[UV36]](s16), [[UV37]](s16), [[UV38]](s16), [[UV39]](s16), [[UV40]](s16), [[UV41]](s16), [[UV42]](s16), [[UV43]](s16), [[UV44]](s16), [[UV45]](s16), [[UV46]](s16), [[UV47]](s16), [[UV48]](s16), [[UV49]](s16), [[UV50]](s16), [[UV51]](s16), [[UV52]](s16), [[UV53]](s16), [[UV54]](s16), [[UV55]](s16), [[UV56]](s16), [[UV57]](s16), [[UV58]](s16), [[UV59]](s16), [[UV60]](s16), [[UV61]](s16), [[UV62]](s16), [[UV63]](s16), [[UV64]](s16), [[DEF1]](s16)
+  ; CHECK-NEXT:   [[UV65:%[0-9]+]]:_(<2 x s16>), [[UV66:%[0-9]+]]:_(<2 x s16>), [[UV67:%[0-9]+]]:_(<2 x s16>), [[UV68:%[0-9]+]]:_(<2 x s16>), [[UV69:%[0-9]+]]:_(<2 x s16>), [[UV70:%[0-9]+]]:_(<2 x s16>), [[UV71:%[0-9]+]]:_(<2 x s16>), [[UV72:%[0-9]+]]:_(<2 x s16>), [[UV73:%[0-9]+]]:_(<2 x s16>), [[UV74:%[0-9]+]]:_(<2 x s16>), [[UV75:%[0-9]+]]:_(<2 x s16>), [[UV76:%[0-9]+]]:_(<2 x s16>), [[UV77:%[0-9]+]]:_(<2 x s16>), [[UV78:%[0-9]+]]:_(<2 x s16>), [[UV79:%[0-9]+]]:_(<2 x s16>), [[UV80:%[0-9]+]]:_(<2 x s16>), [[UV81:%[0-9]+]]:_(<2 x s16>), [[UV82:%[0-9]+]]:_(<2 x s16>), [[UV83:%[0-9]+]]:_(<2 x s16>), [[UV84:%[0-9]+]]:_(<2 x s16>), [[UV85:%[0-9]+]]:_(<2 x s16>), [[UV86:%[0-9]+]]:_(<2 x s16>), [[UV87:%[0-9]+]]:_(<2 x s16>), [[UV88:%[0-9]+]]:_(<2 x s16>), [[UV89:%[0-9]+]]:_(<2 x s16>), [[UV90:%[0-9]+]]:_(<2 x s16>), [[UV91:%[0-9]+]]:_(<2 x s16>), [[UV92:%[0-9]+]]:_(<2 x s16>), [[UV93:%[0-9]+]]:_(<2 x s16>), [[UV94:%[0-9]+]]:_(<2 x s16>), [[UV95:%[0-9]+]]:_(<2 x s16>), [[UV96:%[0-9]+]]:_(<2 x s16>), [[UV97:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<66 x s16>)
   ; CHECK-NEXT:   [[COPY20:%[0-9]+]]:_(p5) = COPY $sp_reg
   ; CHECK-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; CHECK-NEXT:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C3]](s32)
-  ; CHECK-NEXT:   G_STORE [[UV31]](<2 x s16>), [[PTR_ADD1]](p5) :: (store (<2 x s16>) into stack, align 16, addrspace 5)
+  ; CHECK-NEXT:   G_STORE [[UV96]](<2 x s16>), [[PTR_ADD1]](p5) :: (store (<2 x s16>) into stack, align 16, addrspace 5)
   ; CHECK-NEXT:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK-NEXT:   [[PTR_ADD2:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C4]](s32)
-  ; CHECK-NEXT:   G_STORE [[UV32]](<2 x s16>), [[PTR_ADD2]](p5) :: (store (<2 x s16>) into stack + 4, addrspace 5)
-  ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr2 = COPY [[UV2]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr3 = COPY [[UV3]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr4 = COPY [[UV4]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr5 = COPY [[UV5]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr6 = COPY [[UV6]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr7 = COPY [[UV7]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr8 = COPY [[UV8]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr9 = COPY [[UV9]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr10 = COPY [[UV10]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr11 = COPY [[UV11]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr12 = COPY [[UV12]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr13 = COPY [[UV13]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr14 = COPY [[UV14]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr15 = COPY [[UV15]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr16 = COPY [[UV16]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr17 = COPY [[UV17]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr18 = COPY [[UV18]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr19 = COPY [[UV19]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr20 = COPY [[UV20]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr21 = COPY [[UV21]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr22 = COPY [[UV22]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr23 = COPY [[UV23]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr24 = COPY [[UV24]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr25 = COPY [[UV25]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr26 = COPY [[UV26]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr27 = COPY [[UV27]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr28 = COPY [[UV28]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr29 = COPY [[UV29]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr30 = COPY [[UV30]](<2 x s16>)
+  ; CHECK-NEXT:   G_STORE [[UV97]](<2 x s16>), [[PTR_ADD2]](p5) :: (store (<2 x s16>) into stack + 4, addrspace 5)
+  ; CHECK-NEXT:   $vgpr0 = COPY [[UV65]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr1 = COPY [[UV66]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr2 = COPY [[UV67]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr3 = COPY [[UV68]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr4 = COPY [[UV69]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr5 = COPY [[UV70]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr6 = COPY [[UV71]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr7 = COPY [[UV72]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr8 = COPY [[UV73]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr9 = COPY [[UV74]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr10 = COPY [[UV75]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr11 = COPY [[UV76]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr12 = COPY [[UV77]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr13 = COPY [[UV78]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr14 = COPY [[UV79]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr15 = COPY [[UV80]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr16 = COPY [[UV81]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr17 = COPY [[UV82]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr18 = COPY [[UV83]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr19 = COPY [[UV84]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr20 = COPY [[UV85]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr21 = COPY [[UV86]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr22 = COPY [[UV87]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr23 = COPY [[UV88]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr24 = COPY [[UV89]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr25 = COPY [[UV90]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr26 = COPY [[UV91]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr27 = COPY [[UV92]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr28 = COPY [[UV93]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr29 = COPY [[UV94]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr30 = COPY [[UV95]](<2 x s16>)
   ; CHECK-NEXT:   [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
   ; CHECK-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
   ; CHECK-NEXT:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK-NEXT:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK-NEXT:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK-NEXT:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK-NEXT:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK-NEXT:   $sgpr13 = COPY [[COPY15]](s32)

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-function-args.ll

 
 define void @void_func_v3i16(<3 x i16> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v3i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   G_STORE [[UV]](<3 x s16>), [[DEF1]](p1) :: (store (<3 x s16>) into `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s16>), [[DEF]](p1) :: (store (<3 x s16>) into `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY2]]
   ; CHECK-NEXT:   S_SETPC_B64_return [[COPY3]]
   store <3 x i16> %arg0, <3 x i16> addrspace(1)* undef
   ret void
 }
 
 define void @void_func_v4i16(<4 x i16> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v4i16

 define void @void_func_v5i16(<5 x i16> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v5i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $vgpr0, $vgpr1, $vgpr2, $sgpr30_sgpr31
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[COPY2]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s16>)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[COPY2]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16)
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   G_STORE [[UV]](<5 x s16>), [[DEF1]](p1) :: (store (<5 x s16>) into `<5 x i16> addrspace(1)* undef`, align 16, addrspace 1)
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   G_STORE [[BUILD_VECTOR]](<5 x s16>), [[DEF]](p1) :: (store (<5 x s16>) into `<5 x i16> addrspace(1)* undef`, align 16, addrspace 1)
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY3]]
   ; CHECK-NEXT:   S_SETPC_B64_return [[COPY4]]
   store <5 x i16> %arg0, <5 x i16> addrspace(1)* undef
   ret void
 }
 
 define void @void_func_v8i16(<8 x i16> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v8i16

   ; CHECK-NEXT:   [[COPY27:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr27
   ; CHECK-NEXT:   [[COPY28:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr28
   ; CHECK-NEXT:   [[COPY29:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr29
   ; CHECK-NEXT:   [[COPY30:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr30
   ; CHECK-NEXT:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %fixed-stack.1
   ; CHECK-NEXT:   [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[FRAME_INDEX]](p5) :: (invariant load (<2 x s16>) from %fixed-stack.1, align 16, addrspace 5)
   ; CHECK-NEXT:   [[FRAME_INDEX1:%[0-9]+]]:_(p5) = G_FRAME_INDEX %fixed-stack.0
   ; CHECK-NEXT:   [[LOAD1:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[FRAME_INDEX1]](p5) :: (invariant load (<2 x s16>) from %fixed-stack.0, addrspace 5)
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<130 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[COPY2]](<2 x s16>), [[COPY3]](<2 x s16>), [[COPY4]](<2 x s16>), [[COPY5]](<2 x s16>), [[COPY6]](<2 x s16>), [[COPY7]](<2 x s16>), [[COPY8]](<2 x s16>), [[COPY9]](<2 x s16>), [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[COPY12]](<2 x s16>), [[COPY13]](<2 x s16>), [[COPY14]](<2 x s16>), [[COPY15]](<2 x s16>), [[COPY16]](<2 x s16>), [[COPY17]](<2 x s16>), [[COPY18]](<2 x s16>), [[COPY19]](<2 x s16>), [[COPY20]](<2 x s16>), [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>), [[COPY23]](<2 x s16>), [[COPY24]](<2 x s16>), [[COPY25]](<2 x s16>), [[COPY26]](<2 x s16>), [[COPY27]](<2 x s16>), [[COPY28]](<2 x s16>), [[COPY29]](<2 x s16>), [[COPY30]](<2 x s16>), [[LOAD]](<2 x s16>), [[LOAD1]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<65 x s16>), [[UV1:%[0-9]+]]:_(<65 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<130 x s16>)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<66 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[COPY2]](<2 x s16>), [[COPY3]](<2 x s16>), [[COPY4]](<2 x s16>), [[COPY5]](<2 x s16>), [[COPY6]](<2 x s16>), [[COPY7]](<2 x s16>), [[COPY8]](<2 x s16>), [[COPY9]](<2 x s16>), [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[COPY12]](<2 x s16>), [[COPY13]](<2 x s16>), [[COPY14]](<2 x s16>), [[COPY15]](<2 x s16>), [[COPY16]](<2 x s16>), [[COPY17]](<2 x s16>), [[COPY18]](<2 x s16>), [[COPY19]](<2 x s16>), [[COPY20]](<2 x s16>), [[COPY21]](<2 x s16>), [[COPY22]](<2 x s16>), [[COPY23]](<2 x s16>), [[COPY24]](<2 x s16>), [[COPY25]](<2 x s16>), [[COPY26]](<2 x s16>), [[COPY27]](<2 x s16>), [[COPY28]](<2 x s16>), [[COPY29]](<2 x s16>), [[COPY30]](<2 x s16>), [[LOAD]](<2 x s16>), [[LOAD1]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16), [[UV8:%[0-9]+]]:_(s16), [[UV9:%[0-9]+]]:_(s16), [[UV10:%[0-9]+]]:_(s16), [[UV11:%[0-9]+]]:_(s16), [[UV12:%[0-9]+]]:_(s16), [[UV13:%[0-9]+]]:_(s16), [[UV14:%[0-9]+]]:_(s16), [[UV15:%[0-9]+]]:_(s16), [[UV16:%[0-9]+]]:_(s16), [[UV17:%[0-9]+]]:_(s16), [[UV18:%[0-9]+]]:_(s16), [[UV19:%[0-9]+]]:_(s16), [[UV20:%[0-9]+]]:_(s16), [[UV21:%[0-9]+]]:_(s16), [[UV22:%[0-9]+]]:_(s16), [[UV23:%[0-9]+]]:_(s16), [[UV24:%[0-9]+]]:_(s16), [[UV25:%[0-9]+]]:_(s16), [[UV26:%[0-9]+]]:_(s16), [[UV27:%[0-9]+]]:_(s16), [[UV28:%[0-9]+]]:_(s16), [[UV29:%[0-9]+]]:_(s16), [[UV30:%[0-9]+]]:_(s16), [[UV31:%[0-9]+]]:_(s16), [[UV32:%[0-9]+]]:_(s16), [[UV33:%[0-9]+]]:_(s16), [[UV34:%[0-9]+]]:_(s16), [[UV35:%[0-9]+]]:_(s16), [[UV36:%[0-9]+]]:_(s16), [[UV37:%[0-9]+]]:_(s16), [[UV38:%[0-9]+]]:_(s16), [[UV39:%[0-9]+]]:_(s16), [[UV40:%[0-9]+]]:_(s16), [[UV41:%[0-9]+]]:_(s16), [[UV42:%[0-9]+]]:_(s16), [[UV43:%[0-9]+]]:_(s16), [[UV44:%[0-9]+]]:_(s16), [[UV45:%[0-9]+]]:_(s16), [[UV46:%[0-9]+]]:_(s16), [[UV47:%[0-9]+]]:_(s16), [[UV48:%[0-9]+]]:_(s16), [[UV49:%[0-9]+]]:_(s16), [[UV50:%[0-9]+]]:_(s16), [[UV51:%[0-9]+]]:_(s16), [[UV52:%[0-9]+]]:_(s16), [[UV53:%[0-9]+]]:_(s16), [[UV54:%[0-9]+]]:_(s16), [[UV55:%[0-9]+]]:_(s16), [[UV56:%[0-9]+]]:_(s16), [[UV57:%[0-9]+]]:_(s16), [[UV58:%[0-9]+]]:_(s16), [[UV59:%[0-9]+]]:_(s16), [[UV60:%[0-9]+]]:_(s16), [[UV61:%[0-9]+]]:_(s16), [[UV62:%[0-9]+]]:_(s16), [[UV63:%[0-9]+]]:_(s16), [[UV64:%[0-9]+]]:_(s16), [[UV65:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<66 x s16>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<65 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16), [[UV6]](s16), [[UV7]](s16), [[UV8]](s16), [[UV9]](s16), [[UV10]](s16), [[UV11]](s16), [[UV12]](s16), [[UV13]](s16), [[UV14]](s16), [[UV15]](s16), [[UV16]](s16), [[UV17]](s16), [[UV18]](s16), [[UV19]](s16), [[UV20]](s16), [[UV21]](s16), [[UV22]](s16), [[UV23]](s16), [[UV24]](s16), [[UV25]](s16), [[UV26]](s16), [[UV27]](s16), [[UV28]](s16), [[UV29]](s16), [[UV30]](s16), [[UV31]](s16), [[UV32]](s16), [[UV33]](s16), [[UV34]](s16), [[UV35]](s16), [[UV36]](s16), [[UV37]](s16), [[UV38]](s16), [[UV39]](s16), [[UV40]](s16), [[UV41]](s16), [[UV42]](s16), [[UV43]](s16), [[UV44]](s16), [[UV45]](s16), [[UV46]](s16), [[UV47]](s16), [[UV48]](s16), [[UV49]](s16), [[UV50]](s16), [[UV51]](s16), [[UV52]](s16), [[UV53]](s16), [[UV54]](s16), [[UV55]](s16), [[UV56]](s16), [[UV57]](s16), [[UV58]](s16), [[UV59]](s16), [[UV60]](s16), [[UV61]](s16), [[UV62]](s16), [[UV63]](s16), [[UV64]](s16)
   ; CHECK-NEXT:   [[COPY31:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   G_STORE [[UV]](<65 x s16>), [[DEF1]](p1) :: (store (<65 x s16>) into `<65 x i16> addrspace(1)* undef`, align 256, addrspace 1)
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   G_STORE [[BUILD_VECTOR]](<65 x s16>), [[DEF]](p1) :: (store (<65 x s16>) into `<65 x i16> addrspace(1)* undef`, align 256, addrspace 1)
   ; CHECK-NEXT:   [[COPY32:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY31]]
   ; CHECK-NEXT:   S_SETPC_B64_return [[COPY32]]
   store <65 x i16> %arg0, <65 x i16> addrspace(1)* undef
   ret void
 }
 
 define void @void_func_v2f32(<2 x float> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v2f32

 
 define void @void_func_v3f16(<3 x half> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v3f16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   G_STORE [[UV]](<3 x s16>), [[DEF1]](p1) :: (store (<3 x s16>) into `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   G_STORE [[BUILD_VECTOR]](<3 x s16>), [[DEF]](p1) :: (store (<3 x s16>) into `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY2]]
   ; CHECK-NEXT:   S_SETPC_B64_return [[COPY3]]
   store <3 x half> %arg0, <3 x half> addrspace(1)* undef
   ret void
 }
 
 define void @void_func_v4f16(<4 x half> %arg0) #0 {
   ; CHECK-LABEL: name: void_func_v4f16

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-sibling-call.ll

   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN-NEXT:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GCN-NEXT:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN-NEXT:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN-NEXT:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN-NEXT:   [[COPY8:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN-NEXT:   [[COPY9:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-  ; GCN-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY8]](<2 x s16>), [[COPY9]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; GCN-NEXT:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
+  ; GCN-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY8]](<2 x s16>), [[COPY9]](<2 x s16>)
+  ; GCN-NEXT:   [[UV:%[0-9]+]]:_(s16), [[UV1:%[0-9]+]]:_(s16), [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<4 x s16>)
+  ; GCN-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16)
   ; GCN-NEXT:   [[COPY10:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN-NEXT:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @v3i16_fastcc_v3i16
   ; GCN-NEXT:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN-NEXT:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GCN-NEXT:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GCN-NEXT:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GCN-NEXT:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GCN-NEXT:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GCN-NEXT:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GCN-NEXT:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
-  ; GCN-NEXT:   [[DEF1:%[0-9]+]]:_(<3 x s16>) = G_IMPLICIT_DEF
-  ; GCN-NEXT:   [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[UV]](<3 x s16>), [[DEF1]](<3 x s16>)
-  ; GCN-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<6 x s16>)
-  ; GCN-NEXT:   $vgpr0 = COPY [[UV2]](<2 x s16>)
-  ; GCN-NEXT:   $vgpr1 = COPY [[UV3]](<2 x s16>)
+  ; GCN-NEXT:   [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<3 x s16>)
+  ; GCN-NEXT:   [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  ; GCN-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[UV4]](s16), [[UV5]](s16), [[UV6]](s16), [[DEF]](s16)
+  ; GCN-NEXT:   [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BUILD_VECTOR1]](<4 x s16>)
+  ; GCN-NEXT:   $vgpr0 = COPY [[UV7]](<2 x s16>)
+  ; GCN-NEXT:   $vgpr1 = COPY [[UV8]](<2 x s16>)
   ; GCN-NEXT:   [[COPY19:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN-NEXT:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY19]](<4 x s32>)
   ; GCN-NEXT:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
   ; GCN-NEXT:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; GCN-NEXT:   $sgpr8_sgpr9 = COPY [[COPY13]](p4)
   ; GCN-NEXT:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; GCN-NEXT:   $sgpr12 = COPY [[COPY15]](s32)
   ; GCN-NEXT:   $sgpr13 = COPY [[COPY16]](s32)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-and.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -allow-ginsert-as-artifact=0 -global-isel-abort=0 %s -o - | FileCheck %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck %s
 
 ---
 name: test_and_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; CHECK-LABEL: name: test_and_s32

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_and_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s32>) = G_CONCAT_VECTORS [[AND]](<2 x s32>), [[AND1]](<2 x s32>), [[DEF1]](<2 x s32>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<3 x s32>), [[UV7:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s32>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[UV6]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR]], [[BUILD_VECTOR1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[UV5]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[UV7]](s32), [[AND1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR2]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s32>) = G_AND %0, %1
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 
 ---
 name: test_and_v4s32

 name: test_and_v5s32
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_and_v5s32
     ; CHECK: [[DEF:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<5 x s32>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
     ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[DEF2]](s32)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](<5 x s32>)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV6]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV7]](s32), [[UV8]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV9]](s32), [[DEF2]](s32)
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR1]], [[BUILD_VECTOR4]]
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR2]], [[BUILD_VECTOR5]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s32>) = G_CONCAT_VECTORS [[AND]](<2 x s32>), [[AND1]](<2 x s32>), [[AND2]](<2 x s32>), [[DEF3]](<2 x s32>), [[DEF3]](<2 x s32>)
-    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(<5 x s32>), [[UV11:%[0-9]+]]:_(<5 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s32>)
-    ; CHECK-NEXT: [[DEF4:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<8 x s32>) = G_INSERT [[DEF4]], [[UV10]](<5 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<8 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV6]](s32)
+    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV7]](s32), [[UV8]](s32)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(<2 x s32>) = G_AND [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[UV9]]
+    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND]](<2 x s32>)
+    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND1]](<2 x s32>)
+    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF2]](<8 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[AND2]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR4]](<8 x s32>)
     %0:_(<5 x s32>) = G_IMPLICIT_DEF
     %1:_(<5 x s32>) = G_IMPLICIT_DEF
     %2:_(<5 x s32>) = G_AND %0, %1
     %3:_(<8 x s32>) = G_IMPLICIT_DEF
     %4:_(<8 x s32>) = G_INSERT %3, %2, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %4
 ...
 

 ---
 name: test_and_v3s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
     ; CHECK-LABEL: name: test_and_v3s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT]], [[INSERT1]]
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AND]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
-    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL2]]
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
     ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(<4 x s16>) = G_AND [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AND6]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST8]], [[C1]]
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST9]], [[C1]]
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C1]]
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND9]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C1]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND12]], [[C]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND11]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_AND %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7
     $vgpr0_vgpr1_vgpr2 = COPY %8

 
 ---
 name: test_and_v5s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_and_v5s16
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>), [[UV11:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV6]](<5 x s16>), 0
-    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>)
-    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV12]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV14]](<3 x s16>), 0
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT2]], [[INSERT3]]
-    ; CHECK-NEXT: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV13]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV15]](<3 x s16>), 0
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(<4 x s16>) = G_AND [[INSERT4]], [[INSERT5]]
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AND]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>)
-    ; CHECK-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AND1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
-    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL2]]
-    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(<5 x s16>), [[UV21:%[0-9]+]]:_(<5 x s16>), [[UV22:%[0-9]+]]:_(<5 x s16>), [[UV23:%[0-9]+]]:_(<5 x s16>), [[UV24:%[0-9]+]]:_(<5 x s16>), [[UV25:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF4:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[UV20]](<5 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT6]](<8 x s16>)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(<4 x s16>) = G_AND [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AND6]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[COPY1]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>)
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND9]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[COPY3]], [[SHL7]]
+    ; CHECK-NEXT: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST14]](<2 x s16>), [[BITCAST15]](<2 x s16>)
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(<4 x s16>) = G_AND [[CONCAT_VECTORS2]], [[CONCAT_VECTORS3]]
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AND11]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST16:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST16]], [[C]](s32)
+    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<8 x s16>)
+    ; CHECK-NEXT: [[BITCAST17:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST17]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST18:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST18]], [[C]](s32)
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C1]]
+    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
+    ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C]](s32)
+    ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL8]]
+    ; CHECK-NEXT: [[BITCAST19:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
+    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C1]]
+    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[BITCAST16]], [[C1]]
+    ; CHECK-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C]](s32)
+    ; CHECK-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND14]], [[SHL9]]
+    ; CHECK-NEXT: [[BITCAST20:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR9]](s32)
+    ; CHECK-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
+    ; CHECK-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C1]]
+    ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C]](s32)
+    ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL10]]
+    ; CHECK-NEXT: [[BITCAST21:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
+    ; CHECK-NEXT: [[AND18:%[0-9]+]]:_(s32) = G_AND [[BITCAST18]], [[C1]]
+    ; CHECK-NEXT: [[AND19:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C1]]
+    ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND19]], [[C]](s32)
+    ; CHECK-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[AND18]], [[SHL11]]
+    ; CHECK-NEXT: [[BITCAST22:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR11]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST19]](<2 x s16>), [[BITCAST20]](<2 x s16>), [[BITCAST21]](<2 x s16>), [[BITCAST22]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<8 x s16>)
     %0:_(<5 x s16>) = G_IMPLICIT_DEF
     %1:_(<5 x s16>) = G_IMPLICIT_DEF
     %2:_(<5 x s16>) = G_AND %0, %1
     %3:_(<8 x s16>) = G_IMPLICIT_DEF
     %4:_(<8 x s16>) = G_INSERT %3, %2, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %4
 ...
 
 ---
 name: test_and_v3s8
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_and_v3s8
-    ; CHECK: [[DEF:%[0-9]+]]:_(<3 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF2]], [[DEF]](<3 x s8>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF2]], [[DEF1]](<3 x s8>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8), [[UV3:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT]](<4 x s8>)
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT1]](<4 x s8>)
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s8)
-    ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV4]](s8)
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[ANYEXT]], [[ANYEXT1]]
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[AND]](s32)
-    ; CHECK-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV1]](s8)
-    ; CHECK-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[UV5]](s8)
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[ANYEXT2]], [[ANYEXT3]]
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[AND1]](s32)
-    ; CHECK-NEXT: [[ANYEXT4:%[0-9]+]]:_(s32) = G_ANYEXT [[UV2]](s8)
-    ; CHECK-NEXT: [[ANYEXT5:%[0-9]+]]:_(s32) = G_ANYEXT [[UV6]](s8)
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[ANYEXT4]], [[ANYEXT5]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[AND2]](s32)
-    ; CHECK-NEXT: [[ANYEXT6:%[0-9]+]]:_(s32) = G_ANYEXT [[UV3]](s8)
-    ; CHECK-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[UV7]](s8)
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[ANYEXT6]], [[ANYEXT7]]
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[AND3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC]](s8), [[TRUNC1]](s8), [[TRUNC2]](s8), [[TRUNC3]](s8)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF2]](<4 x s8>), [[DEF2]](<4 x s8>)
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>)
-    ; CHECK-NEXT: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>)
+    ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](<4 x s32>)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[UV4]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[UV5]]
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[UV6]]
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[AND]](s32), [[AND1]](s32), [[AND2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s8>) = G_IMPLICIT_DEF
     %1:_(<3 x s8>) = G_IMPLICIT_DEF
     %2:_(<3 x s8>) = G_AND %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     $vgpr0_vgpr1_vgpr2 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-anyext.mir

 ---
 name: test_anyext_v3s16_to_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: test_anyext_v3s16_to_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
     ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ashr.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=SI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9PLUS %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=SI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=GFX9PLUS %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=GFX9PLUS %s
 
 ---
 name: test_ashr_s32_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_ashr_s32_s32

 ---
 name: test_ashr_v3s64_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_ashr_v3s64_v3s32
     ; SI: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; SI-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[UV]], [[UV3]](s32)
-    ; SI-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[UV1]], [[UV4]](s32)
-    ; SI-NEXT: [[ASHR2:%[0-9]+]]:_(s64) = G_ASHR [[UV2]], [[UV5]](s32)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[ASHR]](s64), [[ASHR1]](s64), [[ASHR2]](s64)
+    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[UV]], [[UV4]](s32)
+    ; SI-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[UV1]], [[UV5]](s32)
+    ; SI-NEXT: [[ASHR2:%[0-9]+]]:_(s64) = G_ASHR [[UV2]], [[UV6]](s32)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[ASHR]](s64), [[ASHR1]](s64), [[ASHR2]](s64), [[UV10]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_ashr_v3s64_v3s32
     ; VI: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; VI-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[UV]], [[UV3]](s32)
-    ; VI-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[UV1]], [[UV4]](s32)
-    ; VI-NEXT: [[ASHR2:%[0-9]+]]:_(s64) = G_ASHR [[UV2]], [[UV5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[ASHR]](s64), [[ASHR1]](s64), [[ASHR2]](s64)
+    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[UV]], [[UV4]](s32)
+    ; VI-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[UV1]], [[UV5]](s32)
+    ; VI-NEXT: [[ASHR2:%[0-9]+]]:_(s64) = G_ASHR [[UV2]], [[UV6]](s32)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[ASHR]](s64), [[ASHR1]](s64), [[ASHR2]](s64), [[UV10]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9PLUS-LABEL: name: test_ashr_v3s64_v3s32
     ; GFX9PLUS: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; GFX9PLUS-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; GFX9PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; GFX9PLUS-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; GFX9PLUS-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; GFX9PLUS-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; GFX9PLUS-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[UV]], [[UV3]](s32)
-    ; GFX9PLUS-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[UV1]], [[UV4]](s32)
-    ; GFX9PLUS-NEXT: [[ASHR2:%[0-9]+]]:_(s64) = G_ASHR [[UV2]], [[UV5]](s32)
-    ; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[ASHR]](s64), [[ASHR1]](s64), [[ASHR2]](s64)
+    ; GFX9PLUS-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9PLUS-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[UV]], [[UV4]](s32)
+    ; GFX9PLUS-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[UV1]], [[UV5]](s32)
+    ; GFX9PLUS-NEXT: [[ASHR2:%[0-9]+]]:_(s64) = G_ASHR [[UV2]], [[UV6]](s32)
     ; GFX9PLUS-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9PLUS-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9PLUS-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[ASHR]](s64), [[ASHR1]](s64), [[ASHR2]](s64), [[UV10]](s64)
+    ; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:_(<3 x s64>) = G_EXTRACT %0, 0
     %2:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
     %3:_(<3 x s64>) = G_ASHR %1, %2
     %4:_(<4 x s64>) = G_IMPLICIT_DEF
     %5:_(<4 x s64>) = G_INSERT %4, %3, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %5
 ...
bb.0:
     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
     ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9PLUS-LABEL: name: test_ashr_v3s16_v3s16
     ; GFX9PLUS: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9PLUS-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; GFX9PLUS-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9PLUS-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; GFX9PLUS-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9PLUS-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9PLUS-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; GFX9PLUS-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9PLUS-NEXT: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32
-    ; GFX9PLUS-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9PLUS-NEXT: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0
-    ; GFX9PLUS-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9PLUS-NEXT: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32
-    ; GFX9PLUS-NEXT: [[ASHR:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[EXTRACT]], [[EXTRACT2]](<2 x s16>)
-    ; GFX9PLUS-NEXT: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[EXTRACT1]], [[EXTRACT3]](s16)
-    ; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[ASHR]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9PLUS-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9PLUS-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9PLUS-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
-    ; GFX9PLUS-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
     ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9PLUS-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; GFX9PLUS-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9PLUS-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9PLUS-NEXT: [[ASHR:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[TRUNC]], [[TRUNC1]](s16)
+    ; GFX9PLUS-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[ASHR]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9PLUS-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9PLUS-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9PLUS-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9PLUS-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9PLUS-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[ASHR1]](s16)
-    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST1]](s32)
-    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST2]](s32)
-    ; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST5]](s32)
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST6]](s32)
+    ; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
     ; GFX9PLUS-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_ASHR %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-bitcast.mir

     liveins: $vgpr0
 
     ; CHECK-LABEL: name: test_bitcast_v2s16_to_v4s8
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
-    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C2]](s32)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
     ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[COPY1]]
     ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
     ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC1]], [[COPY2]]
     ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s16) = COPY [[TRUNC2]](s16)

 name: test_bitcast_v4s8_to_v2s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_bitcast_v4s8_to_v2s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
-    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC1]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
-    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC3]]
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
-    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC4]], [[TRUNC5]]
-    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
-    ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[TRUNC6]], [[TRUNC7]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
+    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[COPY1]], [[TRUNC]]
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
+    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[COPY2]], [[TRUNC1]]
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s16) = COPY [[TRUNC2]](s16)
+    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[COPY3]], [[TRUNC2]]
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s16) = COPY [[TRUNC3]](s16)
+    ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[COPY4]], [[TRUNC3]]
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[ADD]], [[C]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[ADD1]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[ADD2]], [[C]]
     ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[ADD3]], [[C]]

     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
     ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
     ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C3]](s32)
     ; CHECK-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C4]](s32)
     ; CHECK-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
+    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C5]](s32)
     ; CHECK-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C3]](s32)
-    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C4]](s32)
-    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C5]](s32)
-    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C6]](s32)
+    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C6]](s32)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
     ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[COPY1]]
     ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
     ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC1]], [[COPY2]]
     ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s16) = COPY [[TRUNC2]](s16)

 name: test_bitcast_v8s4_to_v2s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_bitcast_v8s4_to_v2s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<8 x s32>)
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<8 x s32>)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV8]](s32)
-    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC1]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV9]](s32)
-    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC3]]
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV10]](s32)
-    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC4]], [[TRUNC5]]
-    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV11]](s32)
-    ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[TRUNC6]], [[TRUNC7]]
-    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
-    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[UV12]](s32)
-    ; CHECK-NEXT: [[ADD4:%[0-9]+]]:_(s16) = G_ADD [[TRUNC8]], [[TRUNC9]]
-    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
-    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[UV13]](s32)
-    ; CHECK-NEXT: [[ADD5:%[0-9]+]]:_(s16) = G_ADD [[TRUNC10]], [[TRUNC11]]
-    ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
-    ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[UV14]](s32)
-    ; CHECK-NEXT: [[ADD6:%[0-9]+]]:_(s16) = G_ADD [[TRUNC12]], [[TRUNC13]]
-    ; CHECK-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
-    ; CHECK-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV15]](s32)
-    ; CHECK-NEXT: [[ADD7:%[0-9]+]]:_(s16) = G_ADD [[TRUNC14]], [[TRUNC15]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
+    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[COPY1]], [[TRUNC]]
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
+    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[COPY2]], [[TRUNC1]]
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s16) = COPY [[TRUNC2]](s16)
+    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[COPY3]], [[TRUNC2]]
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s16) = COPY [[TRUNC3]](s16)
+    ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[COPY4]], [[TRUNC3]]
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s16) = COPY [[TRUNC4]](s16)
+    ; CHECK-NEXT: [[ADD4:%[0-9]+]]:_(s16) = G_ADD [[COPY5]], [[TRUNC4]]
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s16) = COPY [[TRUNC5]](s16)
+    ; CHECK-NEXT: [[ADD5:%[0-9]+]]:_(s16) = G_ADD [[COPY6]], [[TRUNC5]]
+    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
+    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s16) = COPY [[TRUNC6]](s16)
+    ; CHECK-NEXT: [[ADD6:%[0-9]+]]:_(s16) = G_ADD [[COPY7]], [[TRUNC6]]
+    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s16) = COPY [[TRUNC7]](s16)
+    ; CHECK-NEXT: [[ADD7:%[0-9]+]]:_(s16) = G_ADD [[COPY8]], [[TRUNC7]]
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 15
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[ADD]], [[C]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[ADD1]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 4
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[ADD2]], [[C]]
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8

     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32)
     ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
     ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC4]]
     ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC1]], [[TRUNC5]]
     ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC6]]
     ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[TRUNC3]], [[TRUNC7]]
     ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[ADD]](s16)

 name: test_bitcast_v8s8_to_v2s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
 
     ; CHECK-LABEL: name: test_bitcast_v8s8_to_v2s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<8 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<8 x s32>)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[UV3]](s32)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
     ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C]]
     ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
     ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
     ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<8 x s32>)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV12]], [[C]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[UV13]], [[C]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s8) = COPY [[TRUNC2]](s8)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s8) = COPY [[TRUNC3]](s8)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C]]
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C]]
     ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32)
     ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
-    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[UV14]], [[C]]
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C]]
     ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C2]](s32)
     ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
-    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[UV15]], [[C]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[UV7]], [[C]]
     ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C3]](s32)
     ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32)
     ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<2 x s32>)
     %0:_(<8 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:_(<8 x s8>) = G_TRUNC %0
     %2:_(<2 x s32>) = G_BITCAST %1
     S_ENDPGM 0, implicit %2

     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s4) = G_TRUNC [[BITCAST]](s32)
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s4) = G_TRUNC [[LSHR]](s32)
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s4) = G_TRUNC [[BITCAST1]](s32)
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s4) = G_TRUNC [[LSHR1]](s32)
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s4) = G_TRUNC [[BITCAST2]](s32)
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s4) = G_TRUNC [[LSHR2]](s32)
+    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s4) = G_TRUNC [[BITCAST3]](s32)
+    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s4) = G_TRUNC [[LSHR3]](s32)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 8

     ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; CHECK-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
     ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C6]](s32)
     ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
     ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
     ; CHECK-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
     ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[OR5]], [[SHL6]]
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<16 x s16>)
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV14]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s4) = COPY [[TRUNC]](s4)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s4) = COPY [[TRUNC1]](s4)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s4) = COPY [[TRUNC2]](s4)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s4) = COPY [[TRUNC3]](s4)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s4) = COPY [[TRUNC4]](s4)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s4) = COPY [[TRUNC5]](s4)
+    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s4) = COPY [[TRUNC6]](s4)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s4) = COPY [[TRUNC7]](s4)
     ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
     ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C2]](s32)
     ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL7]]
     ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
     ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C3]](s32)
     ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
     ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]

 name: test_bitcast_v6s8_to_v3s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_bitcast_v6s8_to_v3s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<6 x s32>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<6 x s32>)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
-    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC1]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
-    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC3]]
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV8]](s32)
-    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC4]], [[TRUNC5]]
-    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV9]](s32)
-    ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[TRUNC6]], [[TRUNC7]]
-    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
-    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[UV10]](s32)
-    ; CHECK-NEXT: [[ADD4:%[0-9]+]]:_(s16) = G_ADD [[TRUNC8]], [[TRUNC9]]
-    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
-    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[UV11]](s32)
-    ; CHECK-NEXT: [[ADD5:%[0-9]+]]:_(s16) = G_ADD [[TRUNC10]], [[TRUNC11]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
+    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[COPY1]], [[TRUNC]]
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
+    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[COPY2]], [[TRUNC1]]
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s16) = COPY [[TRUNC2]](s16)
+    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[COPY3]], [[TRUNC2]]
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s16) = COPY [[TRUNC3]](s16)
+    ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[COPY4]], [[TRUNC3]]
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s16) = COPY [[TRUNC4]](s16)
+    ; CHECK-NEXT: [[ADD4:%[0-9]+]]:_(s16) = G_ADD [[COPY5]], [[TRUNC4]]
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s16) = COPY [[TRUNC5]](s16)
+    ; CHECK-NEXT: [[ADD5:%[0-9]+]]:_(s16) = G_ADD [[COPY6]], [[TRUNC5]]
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[ADD]], [[C]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[ADD1]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[ADD2]], [[C]]
     ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[ADD3]], [[C]]

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2
 
     ; CHECK-LABEL: name: test_bitcast_v3s16_to_v6s8
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
     ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
     ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
-    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
-    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC3]]
-    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC1]], [[TRUNC4]]
-    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC5]]
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s16) = COPY [[TRUNC2]](s16)
+    ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[COPY1]], [[TRUNC]]
+    ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[COPY2]], [[TRUNC1]]
+    ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[COPY3]], [[TRUNC2]]
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s16) = G_LSHR [[ADD]], [[C]](s16)
     ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[ADD]](s16)
     ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR]](s16)
     ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[ADD1]], [[C]](s16)
     ; CHECK-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[ADD1]](s16)
     ; CHECK-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR1]](s16)
     ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s16) = G_LSHR [[ADD2]], [[C]](s16)

 name: test_bitcast_v16s8_to_v2s64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
 
     ; CHECK-LABEL: name: test_bitcast_v16s8_to_v2s64
     ; CHECK: [[COPY:%[0-9]+]]:_(<16 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s8) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s8) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s8) = G_TRUNC [[UV6]](s32)
+    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s8) = G_TRUNC [[UV7]](s32)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
+    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C]]
+    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC9]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
+    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC10]], [[C]]
+    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC11]], [[C]]
     ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C1]](s16)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C]]
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C]]
+    ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC12]], [[C]]
+    ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC13]], [[C]]
     ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C1]](s16)
     ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]]
-    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
-    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s16) = G_AND [[TRUNC6]], [[C]]
-    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
-    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[TRUNC7]], [[C]]
+    ; CHECK-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s16) = G_AND [[TRUNC14]], [[C]]
+    ; CHECK-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[TRUNC15]], [[C]]
     ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND7]], [[C1]](s16)
     ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND6]], [[SHL3]]
     ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
     ; CHECK-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32)
     ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
     ; CHECK-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
     ; CHECK-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
     ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C2]](s32)
     ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL5]]
     ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR4]](s32), [[OR5]](s32)
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[UV24]](s32)
-    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C]]
-    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[UV25]](s32)
-    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC9]], [[C]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s8) = COPY [[TRUNC2]](s8)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s8) = COPY [[TRUNC3]](s8)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s8) = COPY [[TRUNC6]](s8)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s8) = COPY [[TRUNC7]](s8)
+    ; CHECK-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[UV8]](s32)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[TRUNC16]], [[C]]
+    ; CHECK-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[UV9]](s32)
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC17]], [[C]]
     ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s16) = G_SHL [[AND9]], [[C1]](s16)
     ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s16) = G_OR [[AND8]], [[SHL6]]
-    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV26]](s32)
-    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[TRUNC10]], [[C]]
-    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[UV27]](s32)
-    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[TRUNC11]], [[C]]
+    ; CHECK-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[UV10]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[TRUNC18]], [[C]]
+    ; CHECK-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[UV11]](s32)
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[TRUNC19]], [[C]]
     ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s16) = G_SHL [[AND11]], [[C1]](s16)
     ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s16) = G_OR [[AND10]], [[SHL7]]
-    ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[UV28]](s32)
-    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s16) = G_AND [[TRUNC12]], [[C]]
-    ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[UV29]](s32)
-    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s16) = G_AND [[TRUNC13]], [[C]]
+    ; CHECK-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[UV12]](s32)
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s16) = G_AND [[TRUNC20]], [[C]]
+    ; CHECK-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[UV13]](s32)
+    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s16) = G_AND [[TRUNC21]], [[C]]
     ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s16) = G_SHL [[AND13]], [[C1]](s16)
     ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s16) = G_OR [[AND12]], [[SHL8]]
-    ; CHECK-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[UV30]](s32)
-    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s16) = G_AND [[TRUNC14]], [[C]]
-    ; CHECK-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV31]](s32)
-    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s16) = G_AND [[TRUNC15]], [[C]]
+    ; CHECK-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[UV14]](s32)
+    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s16) = G_AND [[TRUNC22]], [[C]]
+    ; CHECK-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[UV15]](s32)
+    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s16) = G_AND [[TRUNC23]], [[C]]
     ; CHECK-NEXT: [[SHL9:%[0-9]+]]:_(s16) = G_SHL [[AND15]], [[C1]](s16)
     ; CHECK-NEXT: [[OR9:%[0-9]+]]:_(s16) = G_OR [[AND14]], [[SHL9]]
     ; CHECK-NEXT: [[ZEXT4:%[0-9]+]]:_(s32) = G_ZEXT [[OR6]](s16)
     ; CHECK-NEXT: [[ZEXT5:%[0-9]+]]:_(s32) = G_ZEXT [[OR7]](s16)
     ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXT5]], [[C2]](s32)
     ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[ZEXT4]], [[SHL10]]
     ; CHECK-NEXT: [[ZEXT6:%[0-9]+]]:_(s32) = G_ZEXT [[OR8]](s16)
     ; CHECK-NEXT: [[ZEXT7:%[0-9]+]]:_(s32) = G_ZEXT [[OR9]](s16)

 name: test_bitcast_v16s8_to_v4s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
 
     ; CHECK-LABEL: name: test_bitcast_v16s8_to_v4s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<16 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[UV3]](s32)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C2]](s32)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
     ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C]]
     ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
     ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C3]](s32)
     ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[OR1]], [[SHL2]]
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV20]], [[C]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[UV21]], [[C]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s8) = COPY [[TRUNC2]](s8)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s8) = COPY [[TRUNC3]](s8)
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s8) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s8) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s8) = G_TRUNC [[UV6]](s32)
+    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s8) = G_TRUNC [[UV7]](s32)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C]]
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C]]
     ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32)
     ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL3]]
-    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[UV22]], [[C]]
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C]]
     ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C2]](s32)
     ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[OR3]], [[SHL4]]
-    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[UV23]], [[C]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[UV7]], [[C]]
     ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C3]](s32)
     ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[OR4]], [[SHL5]]
-    ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32), [[UV34:%[0-9]+]]:_(s32), [[UV35:%[0-9]+]]:_(s32), [[UV36:%[0-9]+]]:_(s32), [[UV37:%[0-9]+]]:_(s32), [[UV38:%[0-9]+]]:_(s32), [[UV39:%[0-9]+]]:_(s32), [[UV40:%[0-9]+]]:_(s32), [[UV41:%[0-9]+]]:_(s32), [[UV42:%[0-9]+]]:_(s32), [[UV43:%[0-9]+]]:_(s32), [[UV44:%[0-9]+]]:_(s32), [[UV45:%[0-9]+]]:_(s32), [[UV46:%[0-9]+]]:_(s32), [[UV47:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV40]], [[C]]
-    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[UV41]], [[C]]
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s8) = COPY [[TRUNC2]](s8)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s8) = COPY [[TRUNC3]](s8)
+    ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s8) = COPY [[TRUNC6]](s8)
+    ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(s8) = COPY [[TRUNC7]](s8)
+    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s8) = G_TRUNC [[UV8]](s32)
+    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s8) = G_TRUNC [[UV9]](s32)
+    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s8) = G_TRUNC [[UV10]](s32)
+    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s8) = G_TRUNC [[UV11]](s32)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV8]], [[C]]
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[UV9]], [[C]]
     ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C1]](s32)
     ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
-    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[UV42]], [[C]]
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[UV10]], [[C]]
     ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND10]], [[C2]](s32)
     ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[OR6]], [[SHL7]]
-    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[UV43]], [[C]]
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[UV11]], [[C]]
     ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C3]](s32)
     ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[OR7]], [[SHL8]]
-    ; CHECK-NEXT: [[UV48:%[0-9]+]]:_(s32), [[UV49:%[0-9]+]]:_(s32), [[UV50:%[0-9]+]]:_(s32), [[UV51:%[0-9]+]]:_(s32), [[UV52:%[0-9]+]]:_(s32), [[UV53:%[0-9]+]]:_(s32), [[UV54:%[0-9]+]]:_(s32), [[UV55:%[0-9]+]]:_(s32), [[UV56:%[0-9]+]]:_(s32), [[UV57:%[0-9]+]]:_(s32), [[UV58:%[0-9]+]]:_(s32), [[UV59:%[0-9]+]]:_(s32), [[UV60:%[0-9]+]]:_(s32), [[UV61:%[0-9]+]]:_(s32), [[UV62:%[0-9]+]]:_(s32), [[UV63:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV60]], [[C]]
-    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[UV61]], [[C]]
+    ; CHECK-NEXT: [[COPY13:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY14:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY15:%[0-9]+]]:_(s8) = COPY [[TRUNC2]](s8)
+    ; CHECK-NEXT: [[COPY16:%[0-9]+]]:_(s8) = COPY [[TRUNC3]](s8)
+    ; CHECK-NEXT: [[COPY17:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY18:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY19:%[0-9]+]]:_(s8) = COPY [[TRUNC6]](s8)
+    ; CHECK-NEXT: [[COPY20:%[0-9]+]]:_(s8) = COPY [[TRUNC7]](s8)
+    ; CHECK-NEXT: [[COPY21:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY22:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY23:%[0-9]+]]:_(s8) = COPY [[TRUNC10]](s8)
+    ; CHECK-NEXT: [[COPY24:%[0-9]+]]:_(s8) = COPY [[TRUNC11]](s8)
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[UV12]], [[C]]
+    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[UV13]], [[C]]
     ; CHECK-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C1]](s32)
     ; CHECK-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
-    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[UV62]], [[C]]
+    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[UV14]], [[C]]
     ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C2]](s32)
     ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[OR9]], [[SHL10]]
-    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[UV63]], [[C]]
+    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[UV15]], [[C]]
     ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C3]](s32)
     ; CHECK-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[OR10]], [[SHL11]]
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[OR2]](s32), [[OR5]](s32), [[OR8]](s32), [[OR11]](s32)
     ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<16 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
     %1:_(<16 x s8>) = G_TRUNC %0
     %2:_(<4 x s32>) = G_BITCAST %1
     S_ENDPGM 0, implicit %2

 name: test_bitcast_v16s8_to_v8s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
 
     ; CHECK-LABEL: name: test_bitcast_v16s8_to_v8s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<16 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[UV1]](s32)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C1]](s16)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV18]](s32)
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV19]](s32)
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s8) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s8) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC6]], [[C]]
+    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC7]], [[C]]
     ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C1]](s16)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32), [[UV34:%[0-9]+]]:_(s32), [[UV35:%[0-9]+]]:_(s32), [[UV36:%[0-9]+]]:_(s32), [[UV37:%[0-9]+]]:_(s32), [[UV38:%[0-9]+]]:_(s32), [[UV39:%[0-9]+]]:_(s32), [[UV40:%[0-9]+]]:_(s32), [[UV41:%[0-9]+]]:_(s32), [[UV42:%[0-9]+]]:_(s32), [[UV43:%[0-9]+]]:_(s32), [[UV44:%[0-9]+]]:_(s32), [[UV45:%[0-9]+]]:_(s32), [[UV46:%[0-9]+]]:_(s32), [[UV47:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV36]](s32)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C]]
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV37]](s32)
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C]]
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s8) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s8) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC10]], [[C]]
+    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC11]], [[C]]
     ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C1]](s16)
     ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]]
-    ; CHECK-NEXT: [[UV48:%[0-9]+]]:_(s32), [[UV49:%[0-9]+]]:_(s32), [[UV50:%[0-9]+]]:_(s32), [[UV51:%[0-9]+]]:_(s32), [[UV52:%[0-9]+]]:_(s32), [[UV53:%[0-9]+]]:_(s32), [[UV54:%[0-9]+]]:_(s32), [[UV55:%[0-9]+]]:_(s32), [[UV56:%[0-9]+]]:_(s32), [[UV57:%[0-9]+]]:_(s32), [[UV58:%[0-9]+]]:_(s32), [[UV59:%[0-9]+]]:_(s32), [[UV60:%[0-9]+]]:_(s32), [[UV61:%[0-9]+]]:_(s32), [[UV62:%[0-9]+]]:_(s32), [[UV63:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[UV54]](s32)
-    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s16) = G_AND [[TRUNC6]], [[C]]
-    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV55]](s32)
-    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[TRUNC7]], [[C]]
+    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s8) = G_TRUNC [[UV6]](s32)
+    ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s8) = G_TRUNC [[UV7]](s32)
+    ; CHECK-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s16) = G_AND [[TRUNC14]], [[C]]
+    ; CHECK-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[TRUNC15]], [[C]]
     ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND7]], [[C1]](s16)
     ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND6]], [[SHL3]]
-    ; CHECK-NEXT: [[UV64:%[0-9]+]]:_(s32), [[UV65:%[0-9]+]]:_(s32), [[UV66:%[0-9]+]]:_(s32), [[UV67:%[0-9]+]]:_(s32), [[UV68:%[0-9]+]]:_(s32), [[UV69:%[0-9]+]]:_(s32), [[UV70:%[0-9]+]]:_(s32), [[UV71:%[0-9]+]]:_(s32), [[UV72:%[0-9]+]]:_(s32), [[UV73:%[0-9]+]]:_(s32), [[UV74:%[0-9]+]]:_(s32), [[UV75:%[0-9]+]]:_(s32), [[UV76:%[0-9]+]]:_(s32), [[UV77:%[0-9]+]]:_(s32), [[UV78:%[0-9]+]]:_(s32), [[UV79:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[UV72]](s32)
-    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C]]
-    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[UV73]](s32)
-    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC9]], [[C]]
+    ; CHECK-NEXT: [[COPY13:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY14:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY15:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY16:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY17:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY18:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY19:%[0-9]+]]:_(s8) = COPY [[TRUNC12]](s8)
+    ; CHECK-NEXT: [[COPY20:%[0-9]+]]:_(s8) = COPY [[TRUNC13]](s8)
+    ; CHECK-NEXT: [[TRUNC16:%[0-9]+]]:_(s8) = G_TRUNC [[UV8]](s32)
+    ; CHECK-NEXT: [[TRUNC17:%[0-9]+]]:_(s8) = G_TRUNC [[UV9]](s32)
+    ; CHECK-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[UV8]](s32)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[TRUNC18]], [[C]]
+    ; CHECK-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[UV9]](s32)
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC19]], [[C]]
     ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s16) = G_SHL [[AND9]], [[C1]](s16)
     ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s16) = G_OR [[AND8]], [[SHL4]]
-    ; CHECK-NEXT: [[UV80:%[0-9]+]]:_(s32), [[UV81:%[0-9]+]]:_(s32), [[UV82:%[0-9]+]]:_(s32), [[UV83:%[0-9]+]]:_(s32), [[UV84:%[0-9]+]]:_(s32), [[UV85:%[0-9]+]]:_(s32), [[UV86:%[0-9]+]]:_(s32), [[UV87:%[0-9]+]]:_(s32), [[UV88:%[0-9]+]]:_(s32), [[UV89:%[0-9]+]]:_(s32), [[UV90:%[0-9]+]]:_(s32), [[UV91:%[0-9]+]]:_(s32), [[UV92:%[0-9]+]]:_(s32), [[UV93:%[0-9]+]]:_(s32), [[UV94:%[0-9]+]]:_(s32), [[UV95:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[UV90]](s32)
-    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[TRUNC10]], [[C]]
-    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[UV91]](s32)
-    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[TRUNC11]], [[C]]
+    ; CHECK-NEXT: [[COPY21:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY22:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY23:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY24:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY25:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY26:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY27:%[0-9]+]]:_(s8) = COPY [[TRUNC12]](s8)
+    ; CHECK-NEXT: [[COPY28:%[0-9]+]]:_(s8) = COPY [[TRUNC13]](s8)
+    ; CHECK-NEXT: [[COPY29:%[0-9]+]]:_(s8) = COPY [[TRUNC16]](s8)
+    ; CHECK-NEXT: [[COPY30:%[0-9]+]]:_(s8) = COPY [[TRUNC17]](s8)
+    ; CHECK-NEXT: [[TRUNC20:%[0-9]+]]:_(s8) = G_TRUNC [[UV10]](s32)
+    ; CHECK-NEXT: [[TRUNC21:%[0-9]+]]:_(s8) = G_TRUNC [[UV11]](s32)
+    ; CHECK-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[UV10]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[TRUNC22]], [[C]]
+    ; CHECK-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[UV11]](s32)
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[TRUNC23]], [[C]]
     ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s16) = G_SHL [[AND11]], [[C1]](s16)
     ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s16) = G_OR [[AND10]], [[SHL5]]
-    ; CHECK-NEXT: [[UV96:%[0-9]+]]:_(s32), [[UV97:%[0-9]+]]:_(s32), [[UV98:%[0-9]+]]:_(s32), [[UV99:%[0-9]+]]:_(s32), [[UV100:%[0-9]+]]:_(s32), [[UV101:%[0-9]+]]:_(s32), [[UV102:%[0-9]+]]:_(s32), [[UV103:%[0-9]+]]:_(s32), [[UV104:%[0-9]+]]:_(s32), [[UV105:%[0-9]+]]:_(s32), [[UV106:%[0-9]+]]:_(s32), [[UV107:%[0-9]+]]:_(s32), [[UV108:%[0-9]+]]:_(s32), [[UV109:%[0-9]+]]:_(s32), [[UV110:%[0-9]+]]:_(s32), [[UV111:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[UV108]](s32)
-    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s16) = G_AND [[TRUNC12]], [[C]]
-    ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[UV109]](s32)
-    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s16) = G_AND [[TRUNC13]], [[C]]
+    ; CHECK-NEXT: [[COPY31:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY32:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY33:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY34:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY35:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY36:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY37:%[0-9]+]]:_(s8) = COPY [[TRUNC12]](s8)
+    ; CHECK-NEXT: [[COPY38:%[0-9]+]]:_(s8) = COPY [[TRUNC13]](s8)
+    ; CHECK-NEXT: [[COPY39:%[0-9]+]]:_(s8) = COPY [[TRUNC16]](s8)
+    ; CHECK-NEXT: [[COPY40:%[0-9]+]]:_(s8) = COPY [[TRUNC17]](s8)
+    ; CHECK-NEXT: [[COPY41:%[0-9]+]]:_(s8) = COPY [[TRUNC20]](s8)
+    ; CHECK-NEXT: [[COPY42:%[0-9]+]]:_(s8) = COPY [[TRUNC21]](s8)
+    ; CHECK-NEXT: [[TRUNC24:%[0-9]+]]:_(s8) = G_TRUNC [[UV12]](s32)
+    ; CHECK-NEXT: [[TRUNC25:%[0-9]+]]:_(s8) = G_TRUNC [[UV13]](s32)
+    ; CHECK-NEXT: [[TRUNC26:%[0-9]+]]:_(s16) = G_TRUNC [[UV12]](s32)
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s16) = G_AND [[TRUNC26]], [[C]]
+    ; CHECK-NEXT: [[TRUNC27:%[0-9]+]]:_(s16) = G_TRUNC [[UV13]](s32)
+    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s16) = G_AND [[TRUNC27]], [[C]]
     ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s16) = G_SHL [[AND13]], [[C1]](s16)
     ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s16) = G_OR [[AND12]], [[SHL6]]
-    ; CHECK-NEXT: [[UV112:%[0-9]+]]:_(s32), [[UV113:%[0-9]+]]:_(s32), [[UV114:%[0-9]+]]:_(s32), [[UV115:%[0-9]+]]:_(s32), [[UV116:%[0-9]+]]:_(s32), [[UV117:%[0-9]+]]:_(s32), [[UV118:%[0-9]+]]:_(s32), [[UV119:%[0-9]+]]:_(s32), [[UV120:%[0-9]+]]:_(s32), [[UV121:%[0-9]+]]:_(s32), [[UV122:%[0-9]+]]:_(s32), [[UV123:%[0-9]+]]:_(s32), [[UV124:%[0-9]+]]:_(s32), [[UV125:%[0-9]+]]:_(s32), [[UV126:%[0-9]+]]:_(s32), [[UV127:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<16 x s32>)
-    ; CHECK-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[UV126]](s32)
-    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s16) = G_AND [[TRUNC14]], [[C]]
-    ; CHECK-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[UV127]](s32)
-    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s16) = G_AND [[TRUNC15]], [[C]]
+    ; CHECK-NEXT: [[COPY43:%[0-9]+]]:_(s8) = COPY [[TRUNC]](s8)
+    ; CHECK-NEXT: [[COPY44:%[0-9]+]]:_(s8) = COPY [[TRUNC1]](s8)
+    ; CHECK-NEXT: [[COPY45:%[0-9]+]]:_(s8) = COPY [[TRUNC4]](s8)
+    ; CHECK-NEXT: [[COPY46:%[0-9]+]]:_(s8) = COPY [[TRUNC5]](s8)
+    ; CHECK-NEXT: [[COPY47:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY48:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY49:%[0-9]+]]:_(s8) = COPY [[TRUNC12]](s8)
+    ; CHECK-NEXT: [[COPY50:%[0-9]+]]:_(s8) = COPY [[TRUNC13]](s8)
+    ; CHECK-NEXT: [[COPY51:%[0-9]+]]:_(s8) = COPY [[TRUNC16]](s8)
+    ; CHECK-NEXT: [[COPY52:%[0-9]+]]:_(s8) = COPY [[TRUNC17]](s8)
+    ; CHECK-NEXT: [[COPY53:%[0-9]+]]:_(s8) = COPY [[TRUNC20]](s8)
+    ; CHECK-NEXT: [[COPY54:%[0-9]+]]:_(s8) = COPY [[TRUNC21]](s8)
+    ; CHECK-NEXT: [[COPY55:%[0-9]+]]:_(s8) = COPY [[TRUNC24]](s8)
+    ; CHECK-NEXT: [[COPY56:%[0-9]+]]:_(s8) = COPY [[TRUNC25]](s8)
+    ; CHECK-NEXT: [[TRUNC28:%[0-9]+]]:_(s16) = G_TRUNC [[UV14]](s32)
+    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s16) = G_AND [[TRUNC28]], [[C]]
+    ; CHECK-NEXT: [[TRUNC29:%[0-9]+]]:_(s16) = G_TRUNC [[UV15]](s32)
+    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s16) = G_AND [[TRUNC29]], [[C]]
     ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s16) = G_SHL [[AND15]], [[C1]](s16)
     ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s16) = G_OR [[AND14]], [[SHL7]]
     ; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
     ; CHECK-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32)
     ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL8]]
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)

     ; CHECK-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
     ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL4]]
     ; CHECK-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[OR2]](s16)
     ; CHECK-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[OR3]](s16)
     ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32)
     ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL5]]
     ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR4]](s32), [[OR5]](s32)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<12 x s16>)
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32)
     ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32)
     ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR5]](s32)
-    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<12 x s16>)
-    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>)
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<12 x s16>)
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST6]](s32)
     ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
-    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST7]](s32)
     ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
     ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C1]]
     ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC9]], [[C1]]
     ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s16) = G_SHL [[AND9]], [[C2]](s16)
     ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s16) = G_OR [[AND8]], [[SHL6]]
     ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[TRUNC10]], [[C1]]

     ; CHECK-NEXT: [[ZEXT5:%[0-9]+]]:_(s32) = G_ZEXT [[OR7]](s16)
     ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[ZEXT5]], [[C]](s32)
     ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[ZEXT4]], [[SHL10]]
     ; CHECK-NEXT: [[ZEXT6:%[0-9]+]]:_(s32) = G_ZEXT [[OR8]](s16)
     ; CHECK-NEXT: [[ZEXT7:%[0-9]+]]:_(s32) = G_ZEXT [[OR9]](s16)
     ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[ZEXT7]], [[C]](s32)
     ; CHECK-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[ZEXT6]], [[SHL11]]
     ; CHECK-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR10]](s32), [[OR11]](s32)
-    ; CHECK-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<12 x s16>)
-    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV20]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST8]](s32)
     ; CHECK-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR8]](s32)
-    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV21]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST9]](s32)
     ; CHECK-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
-    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV22]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC20:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST10]](s32)
     ; CHECK-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC21:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR10]](s32)
-    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV23]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV11]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC22:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST11]](s32)
     ; CHECK-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST11]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC23:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
     ; CHECK-NEXT: [[AND16:%[0-9]+]]:_(s16) = G_AND [[TRUNC16]], [[C1]]
     ; CHECK-NEXT: [[AND17:%[0-9]+]]:_(s16) = G_AND [[TRUNC17]], [[C1]]
     ; CHECK-NEXT: [[SHL12:%[0-9]+]]:_(s16) = G_SHL [[AND17]], [[C2]](s16)
     ; CHECK-NEXT: [[OR12:%[0-9]+]]:_(s16) = G_OR [[AND16]], [[SHL12]]
     ; CHECK-NEXT: [[AND18:%[0-9]+]]:_(s16) = G_AND [[TRUNC18]], [[C1]]

     ; CHECK-LABEL: name: test_bitcast_v8s8_to_v4s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<8 x s16>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<8 x s16>)
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+    ; CHECK-NEXT: [[TRUNC8:%[0-9]+]]:_(s8) = G_TRUNC [[BITCAST]](s32)
+    ; CHECK-NEXT: [[TRUNC9:%[0-9]+]]:_(s8) = G_TRUNC [[LSHR]](s32)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C1]]
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND1]], [[C2]](s16)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<8 x s16>)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[TRUNC10:%[0-9]+]]:_(s8) = G_TRUNC [[BITCAST1]](s32)
+    ; CHECK-NEXT: [[TRUNC11:%[0-9]+]]:_(s8) = G_TRUNC [[LSHR1]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C1]]
     ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C1]]
     ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND3]], [[C2]](s16)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<8 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>)
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
-    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s8) = COPY [[TRUNC10]](s8)
+    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:_(s8) = COPY [[TRUNC11]](s8)
+    ; CHECK-NEXT: [[TRUNC12:%[0-9]+]]:_(s8) = G_TRUNC [[BITCAST2]](s32)
+    ; CHECK-NEXT: [[TRUNC13:%[0-9]+]]:_(s8) = G_TRUNC [[LSHR2]](s32)
     ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C1]]
     ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C1]]
     ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[AND5]], [[C2]](s16)
     ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[AND4]], [[SHL2]]
-    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<8 x s16>)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV15]](<2 x s16>)
-    ; CHECK-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
-    ; CHECK-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
+    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:_(s8) = COPY [[TRUNC8]](s8)
+    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:_(s8) = COPY [[TRUNC9]](s8)
+    ; CHECK-NEXT: [[COPY9:%[0-9]+]]:_(s8) = COPY [[TRUNC10]](s8)
+    ; CHECK-NEXT: [[COPY10:%[0-9]+]]:_(s8) = COPY [[TRUNC11]](s8)
+    ; CHECK-NEXT: [[COPY11:%[0-9]+]]:_(s8) = COPY [[TRUNC12]](s8)
+    ; CHECK-NEXT: [[COPY12:%[0-9]+]]:_(s8) = COPY [[TRUNC13]](s8)
     ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s16) = G_AND [[TRUNC6]], [[C1]]
     ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[TRUNC7]], [[C1]]
     ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s16) = G_SHL [[AND7]], [[C2]](s16)
     ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[AND6]], [[SHL3]]
     ; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[OR]], [[OR]]
     ; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[OR1]], [[OR1]]
     ; CHECK-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[OR2]], [[OR2]]
     ; CHECK-NEXT: [[ADD3:%[0-9]+]]:_(s16) = G_ADD [[OR3]], [[OR3]]

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract-vector-elt.mir

 name: extract_vector_elt_33_v64p3
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
 
     ; CHECK-LABEL: name: extract_vector_elt_33_v64p3
     ; CHECK: [[COPY:%[0-9]+]]:_(p1) = COPY $sgpr0_sgpr1
-    ; CHECK-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[COPY]](p1) :: (load (<16 x s32>), align 4, addrspace 4)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD]](<16 x s32>)
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 64
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 128
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<16 x s32>) from unknown-address + 64, align 4, addrspace 4)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD1]](<16 x s32>)
-    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 128
-    ; CHECK-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; CHECK-NEXT: [[LOAD2:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD1]](p1) :: (load (<16 x s32>) from unknown-address + 128, align 4, addrspace 4)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD2]](<16 x s32>)
-    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 192
-    ; CHECK-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
-    ; CHECK-NEXT: [[LOAD3:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD2]](p1) :: (load (<16 x s32>) from unknown-address + 192, align 4, addrspace 4)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD3]](<16 x s32>)
-    ; CHECK-NEXT: [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
+    ; CHECK-NEXT: [[LOAD:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<16 x s32>) from unknown-address + 128, align 4, addrspace 4)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<16 x p3>) = G_BITCAST [[LOAD]](<16 x s32>)
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(p3), [[UV1:%[0-9]+]]:_(p3), [[UV2:%[0-9]+]]:_(p3), [[UV3:%[0-9]+]]:_(p3), [[UV4:%[0-9]+]]:_(p3), [[UV5:%[0-9]+]]:_(p3), [[UV6:%[0-9]+]]:_(p3), [[UV7:%[0-9]+]]:_(p3), [[UV8:%[0-9]+]]:_(p3), [[UV9:%[0-9]+]]:_(p3), [[UV10:%[0-9]+]]:_(p3), [[UV11:%[0-9]+]]:_(p3), [[UV12:%[0-9]+]]:_(p3), [[UV13:%[0-9]+]]:_(p3), [[UV14:%[0-9]+]]:_(p3), [[UV15:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST]](<16 x p3>)
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(p3), [[UV17:%[0-9]+]]:_(p3), [[UV18:%[0-9]+]]:_(p3), [[UV19:%[0-9]+]]:_(p3), [[UV20:%[0-9]+]]:_(p3), [[UV21:%[0-9]+]]:_(p3), [[UV22:%[0-9]+]]:_(p3), [[UV23:%[0-9]+]]:_(p3), [[UV24:%[0-9]+]]:_(p3), [[UV25:%[0-9]+]]:_(p3), [[UV26:%[0-9]+]]:_(p3), [[UV27:%[0-9]+]]:_(p3), [[UV28:%[0-9]+]]:_(p3), [[UV29:%[0-9]+]]:_(p3), [[UV30:%[0-9]+]]:_(p3), [[UV31:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST1]](<16 x p3>)
-    ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(p3), [[UV33:%[0-9]+]]:_(p3), [[UV34:%[0-9]+]]:_(p3), [[UV35:%[0-9]+]]:_(p3), [[UV36:%[0-9]+]]:_(p3), [[UV37:%[0-9]+]]:_(p3), [[UV38:%[0-9]+]]:_(p3), [[UV39:%[0-9]+]]:_(p3), [[UV40:%[0-9]+]]:_(p3), [[UV41:%[0-9]+]]:_(p3), [[UV42:%[0-9]+]]:_(p3), [[UV43:%[0-9]+]]:_(p3), [[UV44:%[0-9]+]]:_(p3), [[UV45:%[0-9]+]]:_(p3), [[UV46:%[0-9]+]]:_(p3), [[UV47:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST2]](<16 x p3>)
-    ; CHECK-NEXT: [[UV48:%[0-9]+]]:_(p3), [[UV49:%[0-9]+]]:_(p3), [[UV50:%[0-9]+]]:_(p3), [[UV51:%[0-9]+]]:_(p3), [[UV52:%[0-9]+]]:_(p3), [[UV53:%[0-9]+]]:_(p3), [[UV54:%[0-9]+]]:_(p3), [[UV55:%[0-9]+]]:_(p3), [[UV56:%[0-9]+]]:_(p3), [[UV57:%[0-9]+]]:_(p3), [[UV58:%[0-9]+]]:_(p3), [[UV59:%[0-9]+]]:_(p3), [[UV60:%[0-9]+]]:_(p3), [[UV61:%[0-9]+]]:_(p3), [[UV62:%[0-9]+]]:_(p3), [[UV63:%[0-9]+]]:_(p3) = G_UNMERGE_VALUES [[BITCAST3]](<16 x p3>)
-    ; CHECK-NEXT: G_STORE [[UV]](p3), [[FRAME_INDEX]](p5) :: (store (p3) into %stack.0, align 256, addrspace 5)
-    ; CHECK-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
-    ; CHECK-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C3]](s32)
-    ; CHECK-NEXT: G_STORE [[UV1]](p3), [[PTR_ADD3]](p5) :: (store (p3) into %stack.0 + 4, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
-    ; CHECK-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C4]](s32)
-    ; CHECK-NEXT: G_STORE [[UV2]](p3), [[PTR_ADD4]](p5) :: (store (p3) into %stack.0 + 8, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
-    ; CHECK-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C5]](s32)
-    ; CHECK-NEXT: G_STORE [[UV3]](p3), [[PTR_ADD5]](p5) :: (store (p3) into %stack.0 + 12, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CHECK-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C6]](s32)
-    ; CHECK-NEXT: G_STORE [[UV4]](p3), [[PTR_ADD6]](p5) :: (store (p3) into %stack.0 + 16, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
-    ; CHECK-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C7]](s32)
-    ; CHECK-NEXT: G_STORE [[UV5]](p3), [[PTR_ADD7]](p5) :: (store (p3) into %stack.0 + 20, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
-    ; CHECK-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C8]](s32)
-    ; CHECK-NEXT: G_STORE [[UV6]](p3), [[PTR_ADD8]](p5) :: (store (p3) into %stack.0 + 24, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 28
-    ; CHECK-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C9]](s32)
-    ; CHECK-NEXT: G_STORE [[UV7]](p3), [[PTR_ADD9]](p5) :: (store (p3) into %stack.0 + 28, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C10:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
-    ; CHECK-NEXT: [[PTR_ADD10:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C10]](s32)
-    ; CHECK-NEXT: G_STORE [[UV8]](p3), [[PTR_ADD10]](p5) :: (store (p3) into %stack.0 + 32, align 32, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 36
-    ; CHECK-NEXT: [[PTR_ADD11:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C11]](s32)
-    ; CHECK-NEXT: G_STORE [[UV9]](p3), [[PTR_ADD11]](p5) :: (store (p3) into %stack.0 + 36, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 40
-    ; CHECK-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C12]](s32)
-    ; CHECK-NEXT: G_STORE [[UV10]](p3), [[PTR_ADD12]](p5) :: (store (p3) into %stack.0 + 40, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 44
-    ; CHECK-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C13]](s32)
-    ; CHECK-NEXT: G_STORE [[UV11]](p3), [[PTR_ADD13]](p5) :: (store (p3) into %stack.0 + 44, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 48
-    ; CHECK-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C14]](s32)
-    ; CHECK-NEXT: G_STORE [[UV12]](p3), [[PTR_ADD14]](p5) :: (store (p3) into %stack.0 + 48, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 52
-    ; CHECK-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C15]](s32)
-    ; CHECK-NEXT: G_STORE [[UV13]](p3), [[PTR_ADD15]](p5) :: (store (p3) into %stack.0 + 52, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C16:%[0-9]+]]:_(s32) = G_CONSTANT i32 56
-    ; CHECK-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C16]](s32)
-    ; CHECK-NEXT: G_STORE [[UV14]](p3), [[PTR_ADD16]](p5) :: (store (p3) into %stack.0 + 56, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C17:%[0-9]+]]:_(s32) = G_CONSTANT i32 60
-    ; CHECK-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C17]](s32)
-    ; CHECK-NEXT: G_STORE [[UV15]](p3), [[PTR_ADD17]](p5) :: (store (p3) into %stack.0 + 60, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C18:%[0-9]+]]:_(s32) = G_CONSTANT i32 64
-    ; CHECK-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C18]](s32)
-    ; CHECK-NEXT: G_STORE [[UV16]](p3), [[PTR_ADD18]](p5) :: (store (p3) into %stack.0 + 64, align 64, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C19:%[0-9]+]]:_(s32) = G_CONSTANT i32 68
-    ; CHECK-NEXT: [[PTR_ADD19:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C19]](s32)
-    ; CHECK-NEXT: G_STORE [[UV17]](p3), [[PTR_ADD19]](p5) :: (store (p3) into %stack.0 + 68, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C20:%[0-9]+]]:_(s32) = G_CONSTANT i32 72
-    ; CHECK-NEXT: [[PTR_ADD20:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C20]](s32)
-    ; CHECK-NEXT: G_STORE [[UV18]](p3), [[PTR_ADD20]](p5) :: (store (p3) into %stack.0 + 72, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C21:%[0-9]+]]:_(s32) = G_CONSTANT i32 76
-    ; CHECK-NEXT: [[PTR_ADD21:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C21]](s32)
-    ; CHECK-NEXT: G_STORE [[UV19]](p3), [[PTR_ADD21]](p5) :: (store (p3) into %stack.0 + 76, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C22:%[0-9]+]]:_(s32) = G_CONSTANT i32 80
-    ; CHECK-NEXT: [[PTR_ADD22:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C22]](s32)
-    ; CHECK-NEXT: G_STORE [[UV20]](p3), [[PTR_ADD22]](p5) :: (store (p3) into %stack.0 + 80, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C23:%[0-9]+]]:_(s32) = G_CONSTANT i32 84
-    ; CHECK-NEXT: [[PTR_ADD23:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C23]](s32)
-    ; CHECK-NEXT: G_STORE [[UV21]](p3), [[PTR_ADD23]](p5) :: (store (p3) into %stack.0 + 84, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C24:%[0-9]+]]:_(s32) = G_CONSTANT i32 88
-    ; CHECK-NEXT: [[PTR_ADD24:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C24]](s32)
-    ; CHECK-NEXT: G_STORE [[UV22]](p3), [[PTR_ADD24]](p5) :: (store (p3) into %stack.0 + 88, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C25:%[0-9]+]]:_(s32) = G_CONSTANT i32 92
-    ; CHECK-NEXT: [[PTR_ADD25:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C25]](s32)
-    ; CHECK-NEXT: G_STORE [[UV23]](p3), [[PTR_ADD25]](p5) :: (store (p3) into %stack.0 + 92, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C26:%[0-9]+]]:_(s32) = G_CONSTANT i32 96
-    ; CHECK-NEXT: [[PTR_ADD26:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C26]](s32)
-    ; CHECK-NEXT: G_STORE [[UV24]](p3), [[PTR_ADD26]](p5) :: (store (p3) into %stack.0 + 96, align 32, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C27:%[0-9]+]]:_(s32) = G_CONSTANT i32 100
-    ; CHECK-NEXT: [[PTR_ADD27:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C27]](s32)
-    ; CHECK-NEXT: G_STORE [[UV25]](p3), [[PTR_ADD27]](p5) :: (store (p3) into %stack.0 + 100, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C28:%[0-9]+]]:_(s32) = G_CONSTANT i32 104
-    ; CHECK-NEXT: [[PTR_ADD28:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C28]](s32)
-    ; CHECK-NEXT: G_STORE [[UV26]](p3), [[PTR_ADD28]](p5) :: (store (p3) into %stack.0 + 104, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C29:%[0-9]+]]:_(s32) = G_CONSTANT i32 108
-    ; CHECK-NEXT: [[PTR_ADD29:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C29]](s32)
-    ; CHECK-NEXT: G_STORE [[UV27]](p3), [[PTR_ADD29]](p5) :: (store (p3) into %stack.0 + 108, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C30:%[0-9]+]]:_(s32) = G_CONSTANT i32 112
-    ; CHECK-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C30]](s32)
-    ; CHECK-NEXT: G_STORE [[UV28]](p3), [[PTR_ADD30]](p5) :: (store (p3) into %stack.0 + 112, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C31:%[0-9]+]]:_(s32) = G_CONSTANT i32 116
-    ; CHECK-NEXT: [[PTR_ADD31:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C31]](s32)
-    ; CHECK-NEXT: G_STORE [[UV29]](p3), [[PTR_ADD31]](p5) :: (store (p3) into %stack.0 + 116, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C32:%[0-9]+]]:_(s32) = G_CONSTANT i32 120
-    ; CHECK-NEXT: [[PTR_ADD32:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C32]](s32)
-    ; CHECK-NEXT: G_STORE [[UV30]](p3), [[PTR_ADD32]](p5) :: (store (p3) into %stack.0 + 120, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C33:%[0-9]+]]:_(s32) = G_CONSTANT i32 124
-    ; CHECK-NEXT: [[PTR_ADD33:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C33]](s32)
-    ; CHECK-NEXT: G_STORE [[UV31]](p3), [[PTR_ADD33]](p5) :: (store (p3) into %stack.0 + 124, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C34:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
-    ; CHECK-NEXT: [[PTR_ADD34:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C34]](s32)
-    ; CHECK-NEXT: G_STORE [[UV32]](p3), [[PTR_ADD34]](p5) :: (store (p3) into %stack.0 + 128, align 128, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C35:%[0-9]+]]:_(s32) = G_CONSTANT i32 132
-    ; CHECK-NEXT: [[PTR_ADD35:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C35]](s32)
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(p5) = COPY [[PTR_ADD35]](p5)
-    ; CHECK-NEXT: G_STORE [[UV33]](p3), [[COPY1]](p5) :: (store (p3) into %stack.0 + 132, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C36:%[0-9]+]]:_(s32) = G_CONSTANT i32 136
-    ; CHECK-NEXT: [[PTR_ADD36:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C36]](s32)
-    ; CHECK-NEXT: G_STORE [[UV34]](p3), [[PTR_ADD36]](p5) :: (store (p3) into %stack.0 + 136, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C37:%[0-9]+]]:_(s32) = G_CONSTANT i32 140
-    ; CHECK-NEXT: [[PTR_ADD37:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C37]](s32)
-    ; CHECK-NEXT: G_STORE [[UV35]](p3), [[PTR_ADD37]](p5) :: (store (p3) into %stack.0 + 140, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C38:%[0-9]+]]:_(s32) = G_CONSTANT i32 144
-    ; CHECK-NEXT: [[PTR_ADD38:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C38]](s32)
-    ; CHECK-NEXT: G_STORE [[UV36]](p3), [[PTR_ADD38]](p5) :: (store (p3) into %stack.0 + 144, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C39:%[0-9]+]]:_(s32) = G_CONSTANT i32 148
-    ; CHECK-NEXT: [[PTR_ADD39:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C39]](s32)
-    ; CHECK-NEXT: G_STORE [[UV37]](p3), [[PTR_ADD39]](p5) :: (store (p3) into %stack.0 + 148, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C40:%[0-9]+]]:_(s32) = G_CONSTANT i32 152
-    ; CHECK-NEXT: [[PTR_ADD40:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C40]](s32)
-    ; CHECK-NEXT: G_STORE [[UV38]](p3), [[PTR_ADD40]](p5) :: (store (p3) into %stack.0 + 152, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C41:%[0-9]+]]:_(s32) = G_CONSTANT i32 156
-    ; CHECK-NEXT: [[PTR_ADD41:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C41]](s32)
-    ; CHECK-NEXT: G_STORE [[UV39]](p3), [[PTR_ADD41]](p5) :: (store (p3) into %stack.0 + 156, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C42:%[0-9]+]]:_(s32) = G_CONSTANT i32 160
-    ; CHECK-NEXT: [[PTR_ADD42:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C42]](s32)
-    ; CHECK-NEXT: G_STORE [[UV40]](p3), [[PTR_ADD42]](p5) :: (store (p3) into %stack.0 + 160, align 32, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C43:%[0-9]+]]:_(s32) = G_CONSTANT i32 164
-    ; CHECK-NEXT: [[PTR_ADD43:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C43]](s32)
-    ; CHECK-NEXT: G_STORE [[UV41]](p3), [[PTR_ADD43]](p5) :: (store (p3) into %stack.0 + 164, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C44:%[0-9]+]]:_(s32) = G_CONSTANT i32 168
-    ; CHECK-NEXT: [[PTR_ADD44:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C44]](s32)
-    ; CHECK-NEXT: G_STORE [[UV42]](p3), [[PTR_ADD44]](p5) :: (store (p3) into %stack.0 + 168, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C45:%[0-9]+]]:_(s32) = G_CONSTANT i32 172
-    ; CHECK-NEXT: [[PTR_ADD45:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C45]](s32)
-    ; CHECK-NEXT: G_STORE [[UV43]](p3), [[PTR_ADD45]](p5) :: (store (p3) into %stack.0 + 172, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C46:%[0-9]+]]:_(s32) = G_CONSTANT i32 176
-    ; CHECK-NEXT: [[PTR_ADD46:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C46]](s32)
-    ; CHECK-NEXT: G_STORE [[UV44]](p3), [[PTR_ADD46]](p5) :: (store (p3) into %stack.0 + 176, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C47:%[0-9]+]]:_(s32) = G_CONSTANT i32 180
-    ; CHECK-NEXT: [[PTR_ADD47:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C47]](s32)
-    ; CHECK-NEXT: G_STORE [[UV45]](p3), [[PTR_ADD47]](p5) :: (store (p3) into %stack.0 + 180, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C48:%[0-9]+]]:_(s32) = G_CONSTANT i32 184
-    ; CHECK-NEXT: [[PTR_ADD48:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C48]](s32)
-    ; CHECK-NEXT: G_STORE [[UV46]](p3), [[PTR_ADD48]](p5) :: (store (p3) into %stack.0 + 184, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C49:%[0-9]+]]:_(s32) = G_CONSTANT i32 188
-    ; CHECK-NEXT: [[PTR_ADD49:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C49]](s32)
-    ; CHECK-NEXT: G_STORE [[UV47]](p3), [[PTR_ADD49]](p5) :: (store (p3) into %stack.0 + 188, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C50:%[0-9]+]]:_(s32) = G_CONSTANT i32 192
-    ; CHECK-NEXT: [[PTR_ADD50:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C50]](s32)
-    ; CHECK-NEXT: G_STORE [[UV48]](p3), [[PTR_ADD50]](p5) :: (store (p3) into %stack.0 + 192, align 64, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C51:%[0-9]+]]:_(s32) = G_CONSTANT i32 196
-    ; CHECK-NEXT: [[PTR_ADD51:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C51]](s32)
-    ; CHECK-NEXT: G_STORE [[UV49]](p3), [[PTR_ADD51]](p5) :: (store (p3) into %stack.0 + 196, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C52:%[0-9]+]]:_(s32) = G_CONSTANT i32 200
-    ; CHECK-NEXT: [[PTR_ADD52:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C52]](s32)
-    ; CHECK-NEXT: G_STORE [[UV50]](p3), [[PTR_ADD52]](p5) :: (store (p3) into %stack.0 + 200, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C53:%[0-9]+]]:_(s32) = G_CONSTANT i32 204
-    ; CHECK-NEXT: [[PTR_ADD53:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C53]](s32)
-    ; CHECK-NEXT: G_STORE [[UV51]](p3), [[PTR_ADD53]](p5) :: (store (p3) into %stack.0 + 204, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C54:%[0-9]+]]:_(s32) = G_CONSTANT i32 208
-    ; CHECK-NEXT: [[PTR_ADD54:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C54]](s32)
-    ; CHECK-NEXT: G_STORE [[UV52]](p3), [[PTR_ADD54]](p5) :: (store (p3) into %stack.0 + 208, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C55:%[0-9]+]]:_(s32) = G_CONSTANT i32 212
-    ; CHECK-NEXT: [[PTR_ADD55:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C55]](s32)
-    ; CHECK-NEXT: G_STORE [[UV53]](p3), [[PTR_ADD55]](p5) :: (store (p3) into %stack.0 + 212, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C56:%[0-9]+]]:_(s32) = G_CONSTANT i32 216
-    ; CHECK-NEXT: [[PTR_ADD56:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C56]](s32)
-    ; CHECK-NEXT: G_STORE [[UV54]](p3), [[PTR_ADD56]](p5) :: (store (p3) into %stack.0 + 216, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C57:%[0-9]+]]:_(s32) = G_CONSTANT i32 220
-    ; CHECK-NEXT: [[PTR_ADD57:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C57]](s32)
-    ; CHECK-NEXT: G_STORE [[UV55]](p3), [[PTR_ADD57]](p5) :: (store (p3) into %stack.0 + 220, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C58:%[0-9]+]]:_(s32) = G_CONSTANT i32 224
-    ; CHECK-NEXT: [[PTR_ADD58:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C58]](s32)
-    ; CHECK-NEXT: G_STORE [[UV56]](p3), [[PTR_ADD58]](p5) :: (store (p3) into %stack.0 + 224, align 32, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C59:%[0-9]+]]:_(s32) = G_CONSTANT i32 228
-    ; CHECK-NEXT: [[PTR_ADD59:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C59]](s32)
-    ; CHECK-NEXT: G_STORE [[UV57]](p3), [[PTR_ADD59]](p5) :: (store (p3) into %stack.0 + 228, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C60:%[0-9]+]]:_(s32) = G_CONSTANT i32 232
-    ; CHECK-NEXT: [[PTR_ADD60:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C60]](s32)
-    ; CHECK-NEXT: G_STORE [[UV58]](p3), [[PTR_ADD60]](p5) :: (store (p3) into %stack.0 + 232, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C61:%[0-9]+]]:_(s32) = G_CONSTANT i32 236
-    ; CHECK-NEXT: [[PTR_ADD61:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C61]](s32)
-    ; CHECK-NEXT: G_STORE [[UV59]](p3), [[PTR_ADD61]](p5) :: (store (p3) into %stack.0 + 236, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C62:%[0-9]+]]:_(s32) = G_CONSTANT i32 240
-    ; CHECK-NEXT: [[PTR_ADD62:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C62]](s32)
-    ; CHECK-NEXT: G_STORE [[UV60]](p3), [[PTR_ADD62]](p5) :: (store (p3) into %stack.0 + 240, align 16, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C63:%[0-9]+]]:_(s32) = G_CONSTANT i32 244
-    ; CHECK-NEXT: [[PTR_ADD63:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C63]](s32)
-    ; CHECK-NEXT: G_STORE [[UV61]](p3), [[PTR_ADD63]](p5) :: (store (p3) into %stack.0 + 244, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C64:%[0-9]+]]:_(s32) = G_CONSTANT i32 248
-    ; CHECK-NEXT: [[PTR_ADD64:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C64]](s32)
-    ; CHECK-NEXT: G_STORE [[UV62]](p3), [[PTR_ADD64]](p5) :: (store (p3) into %stack.0 + 248, align 8, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[C65:%[0-9]+]]:_(s32) = G_CONSTANT i32 252
-    ; CHECK-NEXT: [[PTR_ADD65:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C65]](s32)
-    ; CHECK-NEXT: G_STORE [[UV63]](p3), [[PTR_ADD65]](p5) :: (store (p3) into %stack.0 + 252, basealign 256, addrspace 5)
-    ; CHECK-NEXT: [[LOAD4:%[0-9]+]]:_(p3) = G_LOAD [[PTR_ADD35]](p5) :: (load (p3) from %stack.0 + 132, addrspace 5)
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[LOAD4]](p3)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(p3) = COPY [[UV1]](p3)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[COPY1]](p3)
     %0:_(p1) = COPY $sgpr0_sgpr1
     %1:_(s32) = G_CONSTANT i32 33
     %2:_(<64 x p3>) = G_LOAD %0 :: (load (<64 x p3>), align 4, addrspace 4)
     %3:_(p3) = G_EXTRACT_VECTOR_ELT %2, %1
     S_ENDPGM 0, implicit %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-extract.mir

 ---
 name: test_extract_s32_v2s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: test_extract_s32_v2s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY]](<2 x s32>), 0
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
     %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_EXTRACT %0, 0
     $vgpr0 = COPY %1
 ...
 ---
 name: test_extract_s32_v2s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: test_extract_s32_v2s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY]](<2 x s32>), 32
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
     %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_EXTRACT %0, 32
      $vgpr0 = COPY %1
 ...
 ---
 name: test_extract_s32_v3s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2
 
     ; CHECK-LABEL: name: test_extract_s32_v3s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY]](<3 x s32>), 0
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(s32) = G_EXTRACT %0, 0
     $vgpr0 = COPY %1
 ...
 ---
 name: test_extract_s32_v3s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2
 
     ; CHECK-LABEL: name: test_extract_s32_v3s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY]](<3 x s32>), 32
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(s32) = G_EXTRACT %0, 32
     $vgpr0 = COPY %1
 ...
 ---
 name: test_extract_s32_v3s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2
 
     ; CHECK-LABEL: name: test_extract_s32_v3s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY]](<3 x s32>), 64
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(s32) = G_EXTRACT %0, 64
     $vgpr0 = COPY %1
 ...
 ---
 name: test_extract_s32_v4s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_s32_v4s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY]](<4 x s32>), 0
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[COPY1]](s32)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = G_EXTRACT %0, 0
     $vgpr0 = COPY %1
 ...
 ---
 name: test_extract_s32_v4s32_offset32
 body: |
   bb.0:

 ---
 name: test_extract_v2s32_v4s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_v2s32_v4s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY]](<4 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](<2 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<2 x s32>) = G_EXTRACT %0, 0
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: test_extract_v2s32_v4s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_v2s32_v4s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY]](<4 x s32>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](<2 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV1]](s32), [[UV2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<2 x s32>) = G_EXTRACT %0, 32
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: test_extract_v2s32_v4s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_v2s32_v4s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY]](<4 x s32>), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](<2 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<2 x s32>) = G_EXTRACT %0, 64
     $vgpr0_vgpr1 = COPY %1
 ...
 ---
 name: test_extract_s64_v4s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_s64_v4s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s64) = G_EXTRACT [[COPY]](<4 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](s64)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UV]](s32), [[UV1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](s64)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s64) = G_EXTRACT %0, 0
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: test_extract_s64_v4s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_s64_v4s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s64) = G_EXTRACT [[COPY]](<4 x s32>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](s64)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UV1]](s32), [[UV2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](s64)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s64) = G_EXTRACT %0, 32
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: test_extract_s64_v4s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_s64_v4s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s64) = G_EXTRACT [[COPY]](<4 x s32>), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](s64)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](s64)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s64) = G_EXTRACT %0, 64
     $vgpr0_vgpr1 = COPY %1
 ...
 ---
 name: test_extract_p0_v4s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_p0_v4s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(p0) = G_EXTRACT [[COPY]](<4 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](p0)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[UV]](s32), [[UV1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](p0)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(p0) = G_EXTRACT %0, 0
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: test_extract_p0_v4s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_p0_v4s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(p0) = G_EXTRACT [[COPY]](<4 x s32>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](p0)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[UV1]](s32), [[UV2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](p0)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(p0) = G_EXTRACT %0, 32
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: test_extract_p0_v4s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_extract_p0_v4s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(p0) = G_EXTRACT [[COPY]](<4 x s32>), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[EXTRACT]](p0)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](p0)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(p0) = G_EXTRACT %0, 64
     $vgpr0_vgpr1 = COPY %1
 ...
 
 ---
 name: extract_s8_v4s8_offset0
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s8_v4s8_offset0
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV1]](<2 x s32>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 0
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV]](s32)
     %0:_(<4 x s8>) = G_IMPLICIT_DEF
     %1:_(s8) = G_EXTRACT %0, 0
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 ---
 name: extract_s8_v4s8_offset8
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s8_v4s8_offset8
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV1]](<2 x s32>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 16
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV1]](s32)
     %0:_(<4 x s8>) = G_IMPLICIT_DEF
     %1:_(s8) = G_EXTRACT %0, 8
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 ---
 name: extract_s8_v4s8_offset16
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s8_v4s8_offset16
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV1]](<2 x s32>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 32
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV2]](s32)
     %0:_(<4 x s8>) = G_IMPLICIT_DEF
     %1:_(s8) = G_EXTRACT %0, 16
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 ---
 name: extract_s8_v4s8_offset24
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s8_v4s8_offset24
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s32>), [[UV1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[UV1]](<2 x s32>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[CONCAT_VECTORS]](<4 x s16>), 48
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV3]](s32)
     %0:_(<4 x s8>) = G_IMPLICIT_DEF
     %1:_(s8) = G_EXTRACT %0, 24
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 # FIXME: Leaves behind dead G_TRUNC
 ---
 name: extract_s8_v3s8_offset16
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s8_v3s8_offset16
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
-    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
-    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[DEF1]](<2 x s16>)
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV4]](<3 x s16>), 0
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<4 x s16>), 32
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV2]](s32)
     %0:_(<3 x s8>) = G_IMPLICIT_DEF
     %1:_(s8) = G_EXTRACT %0, 16
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 ---
 name: extract_s8_v5s1_offset4
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_s8_v5s1_offset4
     ; CHECK: [[DEF:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<5 x s32>)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[DEF1]](s32)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR1]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR2]](<2 x s32>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>), [[TRUNC2]](<2 x s16>), [[DEF2]](<2 x s16>), [[DEF2]](<2 x s16>)
-    ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(<5 x s16>), [[UV6:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s16>)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF3]], [[UV5]](<5 x s16>), 0
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<6 x s16>), 64
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[UV4]](s32)
     %0:_(<5 x s1>) = G_IMPLICIT_DEF
     %1:_(s1) = G_EXTRACT %0, 4
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 ---
 name: extract_v2s16_v4s16_offset32
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_v2s16_v4s16_offset32
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[DEF]](<4 x s16>), 32
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     %0:_(<4 x s16>) = G_IMPLICIT_DEF
     %1:_(<2 x s16>) = G_EXTRACT %0, 32
     $vgpr0 = COPY %1
 ...
 
 ---
 name: extract_v2s16_v6s16_offset32
 body: |
   bb.0:
     ; CHECK-LABEL: name: extract_v2s16_v6s16_offset32
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[DEF]](<6 x s16>), 32
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     %0:_(<6 x s16>) = G_IMPLICIT_DEF
     %1:_(<2 x s16>) = G_EXTRACT %0, 32
     $vgpr0 = COPY %1
 ...
 
 ---
 name: test_extract_s8_s16_offset0
 body: |

 
 ---
 name: extract_s16_v3s16_offset0
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: extract_s16_v3s16_offset0
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[EXTRACT]](s16)
-    ; CHECK-NEXT: $vgpr0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST]](s32)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(s16) = G_EXTRACT %0, 0
     %2:_(s32) = G_ANYEXT %1
     $vgpr0 = COPY %2
 ...
 
 ---
 name: test_extract_s8_s64_offset16

 
 ---
 name: extract_v2s16_v3s16_offset0
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: extract_v2s16_v3s16_offset0
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<2 x s16>) = G_EXTRACT %0, 0
     $vgpr0 = COPY %1
 ...
 
 ---
 name: extract_v2s16_v5s16_offset0
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: extract_v2s16_v5s16_offset0
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<6 x s16>), 0
-    ; CHECK-NEXT: $vgpr0 = COPY [[EXTRACT]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     %0:_(<5 x s16>) = G_IMPLICIT_DEF
     %1:_(<2 x s16>) = G_EXTRACT %0, 0
     $vgpr0 = COPY %1
 ...
 
 ---
 name: extract_s16_v2s16_offset0
 body: |

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fabs.mir

bb.0:
     ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; SI-NEXT: [[FABS:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST2]]
     ; SI-NEXT: [[FABS1:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST3]]
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FABS]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; SI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FABS1]](<2 x s16>)
-    ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[FABS]](<2 x s16>)
-    ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; SI-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[FABS1]](<2 x s16>)
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
-    ; SI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
-    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST6]], [[C1]]
-    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32)
+    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; SI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
-    ; SI-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
-    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
-    ; SI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST7]], [[C1]]
-    ; SI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32)
-    ; SI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]]
-    ; SI-NEXT: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>), [[BITCAST10]](<2 x s16>)
+    ; SI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; SI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; SI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]]
+    ; SI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; SI-NEXT: S_NOP 0, implicit [[CONCAT_VECTORS]](<6 x s16>)
     ; VI-LABEL: name: test_fabs_v3s16
     ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; VI-NEXT: [[FABS:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST2]]
     ; VI-NEXT: [[FABS1:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BITCAST3]]
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FABS]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FABS1]](<2 x s16>)
-    ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[FABS]](<2 x s16>)
-    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; VI-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[FABS1]](<2 x s16>)
     ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
-    ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
-    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST6]], [[C1]]
-    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32)
+    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
-    ; VI-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
-    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
-    ; VI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST7]], [[C1]]
-    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32)
-    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]]
-    ; VI-NEXT: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>), [[BITCAST10]](<2 x s16>)
+    ; VI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL4]]
+    ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: S_NOP 0, implicit [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_fabs_v3s16
     ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF1]](s32)
     ; GFX9-NEXT: [[FABS:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BUILD_VECTOR_TRUNC]]
     ; GFX9-NEXT: [[FABS1:%[0-9]+]]:_(<2 x s16>) = G_FABS [[BUILD_VECTOR_TRUNC1]]
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[FABS]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[FABS1]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FABS]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FABS1]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[BITCAST4]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR2]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[BITCAST2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
     ; GFX9-NEXT: S_NOP 0, implicit [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<3 x s16>) = G_FABS %0
     %2:_(<6 x s16>) = G_CONCAT_VECTORS %1, %1
     S_NOP 0, implicit %2
 ...
 

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir

bb.0:
     ; GFX9-LABEL: name: test_fadd_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[FADD:%[0-9]+]]:_(<2 x s16>) = G_FADD [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9-NEXT: [[FADD1:%[0-9]+]]:_(<2 x s16>) = G_FADD [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FADD]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FADD1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST7]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir

 ---
 name: test_fdiv_v3s16
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_fdiv_v3s16
     ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; SI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; SI-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)

     ; SI-NEXT: [[FPTRUNC2:%[0-9]+]]:_(s16) = G_FPTRUNC [[INT20]](s32)
     ; SI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC]](s16)
     ; SI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC1]](s16)
     ; SI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FPTRUNC2]](s16)
     ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; SI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; VI-LABEL: name: test_fdiv_v3s16
     ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)

     ; VI-NEXT: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC5]](s16), [[TRUNC2]](s16)
     ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16)
     ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16)
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16)
     ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; VI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX9-LABEL: name: test_fdiv_v3s16
     ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; GFX9-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)

     ; GFX9-NEXT: [[INT5:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.div.fixup), [[FPTRUNC2]](s16), [[TRUNC5]](s16), [[TRUNC2]](s16)
     ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[INT1]](s16)
     ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[INT3]](s16)
     ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[INT5]](s16)
     ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; GFX9-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX9-UNSAFE-LABEL: name: test_fdiv_v3s16
     ; GFX9-UNSAFE: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-UNSAFE-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-UNSAFE-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-UNSAFE-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-UNSAFE-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX9-UNSAFE-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-UNSAFE-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-UNSAFE-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX9-UNSAFE-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-UNSAFE-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; GFX9-UNSAFE-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-UNSAFE-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX9-UNSAFE-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-UNSAFE-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; GFX9-UNSAFE-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-UNSAFE-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; GFX9-UNSAFE-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-UNSAFE-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; GFX9-UNSAFE-NEXT: [[INT:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC3]](s16)
     ; GFX9-UNSAFE-NEXT: [[FMUL:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC]], [[INT]]
     ; GFX9-UNSAFE-NEXT: [[INT1:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC4]](s16)
     ; GFX9-UNSAFE-NEXT: [[FMUL1:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC1]], [[INT1]]
     ; GFX9-UNSAFE-NEXT: [[INT2:%[0-9]+]]:_(s16) = G_INTRINSIC intrinsic(@llvm.amdgcn.rcp), [[TRUNC5]](s16)
     ; GFX9-UNSAFE-NEXT: [[FMUL2:%[0-9]+]]:_(s16) = G_FMUL [[TRUNC2]], [[INT2]]
     ; GFX9-UNSAFE-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL]](s16)
     ; GFX9-UNSAFE-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL1]](s16)
     ; GFX9-UNSAFE-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[FMUL2]](s16)
     ; GFX9-UNSAFE-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; GFX9-UNSAFE-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX10-LABEL: name: test_fdiv_v3s16
     ; GFX10: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX10-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX10-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX10-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX10-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; GFX10-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX10-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX10-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX10-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; GFX10-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX10-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; GFX10-NEXT: [[FPEXT:%[0-9]+]]:_(s32) = G_FPEXT [[TRUNC]](s16)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fma.mir

     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr6_vgpr7_vgpr8
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY2]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[FMA:%[0-9]+]]:_(<2 x s16>) = G_FMA [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]], [[BUILD_VECTOR_TRUNC4]]
     ; GFX9-NEXT: [[FMA1:%[0-9]+]]:_(<2 x s16>) = G_FMA [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]], [[BUILD_VECTOR_TRUNC5]]
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[FMA]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[FMA1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST6]](s32), [[LSHR3]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST7]](s32), [[BITCAST8]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST9]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC6]](<2 x s16>), [[BUILD_VECTOR_TRUNC7]](<2 x s16>), [[BUILD_VECTOR_TRUNC8]](<2 x s16>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmaxnum.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_fmaxnum_s32_ieee_mode_on
 machineFunctionInfo:
   mode:
     ieee: true
 body: |
   bb.0:
bb.0:
     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
     ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_fmaxnum_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV4]]
-    ; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV6]]
-    ; GFX9-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
-    ; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV5]]
-    ; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV7]]
-    ; GFX9-NEXT: [[FMAXNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC]]
+    ; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
+    ; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC1]]
+    ; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[FMAXNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST7]](s32)
+    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_FMAXNUM %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fminnum.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_fminnum_s32_ieee_mode_on
 machineFunctionInfo:
   mode:
     ieee: true
 body: |
   bb.0:
bb.0:
     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
     ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_fminnum_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV4]]
-    ; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV6]]
-    ; GFX9-NEXT: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
-    ; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV5]]
-    ; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV7]]
-    ; GFX9-NEXT: [[FMINNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[FMINNUM_IEEE]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[FMINNUM_IEEE1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC]]
+    ; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[FMINNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
+    ; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC1]]
+    ; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[FMINNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMINNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FMINNUM_IEEE]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FMINNUM_IEEE1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST7]](s32)
+    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_FMINNUM %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fmul.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9PLUS %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9PLUS %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9PLUS %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9PLUS %s
 
 ---
 name: test_fmul_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_fmul_s32
bb.0:
     ; GFX9PLUS-LABEL: name: test_fmul_v3s16
     ; GFX9PLUS: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9PLUS-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     ; GFX9PLUS-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9PLUS-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9PLUS-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9PLUS-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9PLUS-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
-    ; GFX9PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
     ; GFX9PLUS-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
     ; GFX9PLUS-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9PLUS-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9PLUS-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9PLUS-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
     ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
     ; GFX9PLUS-NEXT: [[FMUL:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9PLUS-NEXT: [[FMUL1:%[0-9]+]]:_(<2 x s16>) = G_FMUL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
-    ; GFX9PLUS-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9PLUS-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[FMUL]](<2 x s16>)
     ; GFX9PLUS-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; GFX9PLUS-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[FMUL1]](<2 x s16>)
+    ; GFX9PLUS-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9PLUS-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX9PLUS-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
     ; GFX9PLUS-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
     ; GFX9PLUS-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
     ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
     ; GFX9PLUS-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST7]](s32)
     ; GFX9PLUS-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-freeze.mir

 
 ---
 name: test_freeze_s1056
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_freeze_s1056
     ; CHECK: [[COPY:%[0-9]+]]:_(s512) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](s512)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[MV:%[0-9]+]]:_(s1024) = G_MERGE_VALUES [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[UV14]](s32), [[UV15]](s32), [[UV16]](s32), [[UV17]](s32), [[UV18]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32), [[UV22]](s32), [[UV23]](s32), [[UV24]](s32), [[UV25]](s32), [[UV26]](s32), [[UV27]](s32), [[UV28]](s32), [[UV29]](s32), [[UV30]](s32), [[UV31]](s32)
-    ; CHECK-NEXT: [[MV1:%[0-9]+]]:_(s1024) = G_MERGE_VALUES [[UV32]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(s1024) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(s1024) = G_FREEZE [[MV]]
-    ; CHECK-NEXT: [[FREEZE1:%[0-9]+]]:_(s1024) = G_FREEZE [[MV1]]
-    ; CHECK-NEXT: [[MV2:%[0-9]+]]:_(s33792) = G_MERGE_VALUES [[FREEZE]](s1024), [[FREEZE1]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024), [[DEF2]](s1024)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s1056) = G_TRUNC [[MV2]](s33792)
-    ; CHECK-NEXT: S_NOP 0, implicit [[TRUNC]](s1056)
+    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s1056) = G_ANYEXT [[COPY]](s512)
+    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(s1056) = G_FREEZE [[ANYEXT]]
+    ; CHECK-NEXT: S_NOP 0, implicit [[FREEZE]](s1056)
     %0:_(s512) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
     %1:_(s1056) = G_ANYEXT %0
     %2:_(s1056) = G_FREEZE %1
     S_NOP 0, implicit %2
 ...
 
 ---
 name: test_freeze_s2048

 
 ---
 name: test_freeze_v33s32
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_freeze_v33s32
     ; CHECK: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>)
     ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>)
-    ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32), [[UV34:%[0-9]+]]:_(s32), [[UV35:%[0-9]+]]:_(s32), [[UV36:%[0-9]+]]:_(s32), [[UV37:%[0-9]+]]:_(s32), [[UV38:%[0-9]+]]:_(s32), [[UV39:%[0-9]+]]:_(s32), [[UV40:%[0-9]+]]:_(s32), [[UV41:%[0-9]+]]:_(s32), [[UV42:%[0-9]+]]:_(s32), [[UV43:%[0-9]+]]:_(s32), [[UV44:%[0-9]+]]:_(s32), [[UV45:%[0-9]+]]:_(s32), [[UV46:%[0-9]+]]:_(s32), [[UV47:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<16 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32), [[UV8]](s32), [[UV9]](s32), [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[UV14]](s32), [[UV15]](s32)
     ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<16 x s32>) = G_BUILD_VECTOR [[UV16]](s32), [[UV17]](s32), [[UV18]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32), [[UV22]](s32), [[UV23]](s32), [[UV24]](s32), [[UV25]](s32), [[UV26]](s32), [[UV27]](s32), [[UV28]](s32), [[UV29]](s32), [[UV30]](s32), [[UV31]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<16 x s32>) = G_BUILD_VECTOR [[UV32]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32), [[DEF1]](s32)
     ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[BUILD_VECTOR]]
     ; CHECK-NEXT: [[FREEZE1:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[BUILD_VECTOR1]]
-    ; CHECK-NEXT: [[FREEZE2:%[0-9]+]]:_(<16 x s32>) = G_FREEZE [[BUILD_VECTOR2]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<528 x s32>) = G_CONCAT_VECTORS [[FREEZE]](<16 x s32>), [[FREEZE1]](<16 x s32>), [[FREEZE2]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>), [[DEF]](<16 x s32>)
-    ; CHECK-NEXT: [[UV48:%[0-9]+]]:_(<33 x s32>), [[UV49:%[0-9]+]]:_(<33 x s32>), [[UV50:%[0-9]+]]:_(<33 x s32>), [[UV51:%[0-9]+]]:_(<33 x s32>), [[UV52:%[0-9]+]]:_(<33 x s32>), [[UV53:%[0-9]+]]:_(<33 x s32>), [[UV54:%[0-9]+]]:_(<33 x s32>), [[UV55:%[0-9]+]]:_(<33 x s32>), [[UV56:%[0-9]+]]:_(<33 x s32>), [[UV57:%[0-9]+]]:_(<33 x s32>), [[UV58:%[0-9]+]]:_(<33 x s32>), [[UV59:%[0-9]+]]:_(<33 x s32>), [[UV60:%[0-9]+]]:_(<33 x s32>), [[UV61:%[0-9]+]]:_(<33 x s32>), [[UV62:%[0-9]+]]:_(<33 x s32>), [[UV63:%[0-9]+]]:_(<33 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<528 x s32>)
-    ; CHECK-NEXT: S_NOP 0, implicit [[UV48]](<33 x s32>)
+    ; CHECK-NEXT: [[FREEZE2:%[0-9]+]]:_(s32) = G_FREEZE [[DEF1]]
+    ; CHECK-NEXT: [[UV32:%[0-9]+]]:_(s32), [[UV33:%[0-9]+]]:_(s32), [[UV34:%[0-9]+]]:_(s32), [[UV35:%[0-9]+]]:_(s32), [[UV36:%[0-9]+]]:_(s32), [[UV37:%[0-9]+]]:_(s32), [[UV38:%[0-9]+]]:_(s32), [[UV39:%[0-9]+]]:_(s32), [[UV40:%[0-9]+]]:_(s32), [[UV41:%[0-9]+]]:_(s32), [[UV42:%[0-9]+]]:_(s32), [[UV43:%[0-9]+]]:_(s32), [[UV44:%[0-9]+]]:_(s32), [[UV45:%[0-9]+]]:_(s32), [[UV46:%[0-9]+]]:_(s32), [[UV47:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE]](<16 x s32>)
+    ; CHECK-NEXT: [[UV48:%[0-9]+]]:_(s32), [[UV49:%[0-9]+]]:_(s32), [[UV50:%[0-9]+]]:_(s32), [[UV51:%[0-9]+]]:_(s32), [[UV52:%[0-9]+]]:_(s32), [[UV53:%[0-9]+]]:_(s32), [[UV54:%[0-9]+]]:_(s32), [[UV55:%[0-9]+]]:_(s32), [[UV56:%[0-9]+]]:_(s32), [[UV57:%[0-9]+]]:_(s32), [[UV58:%[0-9]+]]:_(s32), [[UV59:%[0-9]+]]:_(s32), [[UV60:%[0-9]+]]:_(s32), [[UV61:%[0-9]+]]:_(s32), [[UV62:%[0-9]+]]:_(s32), [[UV63:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE1]](<16 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<33 x s32>) = G_BUILD_VECTOR [[UV32]](s32), [[UV33]](s32), [[UV34]](s32), [[UV35]](s32), [[UV36]](s32), [[UV37]](s32), [[UV38]](s32), [[UV39]](s32), [[UV40]](s32), [[UV41]](s32), [[UV42]](s32), [[UV43]](s32), [[UV44]](s32), [[UV45]](s32), [[UV46]](s32), [[UV47]](s32), [[UV48]](s32), [[UV49]](s32), [[UV50]](s32), [[UV51]](s32), [[UV52]](s32), [[UV53]](s32), [[UV54]](s32), [[UV55]](s32), [[UV56]](s32), [[UV57]](s32), [[UV58]](s32), [[UV59]](s32), [[UV60]](s32), [[UV61]](s32), [[UV62]](s32), [[UV63]](s32), [[FREEZE2]](s32)
+    ; CHECK-NEXT: S_NOP 0, implicit [[BUILD_VECTOR2]](<33 x s32>)
     %0:_(<33 x s32>) = G_IMPLICIT_DEF
     %1:_(<33 x s32>) = G_FREEZE %0
     S_NOP 0, implicit %1
 ...
 
 ---
 name: test_freeze_v64s32
 body: |

 
 ---
 name: test_freeze_v3s8
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_freeze_v3s8
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF]], [[TRUNC]](<3 x s8>), 0
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(<4 x s32>) = G_ANYEXT [[INSERT]](<4 x s8>)
-    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[ANYEXT]]
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<4 x s8>) = G_TRUNC [[FREEZE]](<4 x s32>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[TRUNC1]](<4 x s8>), [[DEF]](<4 x s8>), [[DEF]](<4 x s8>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s8>), [[UV1:%[0-9]+]]:_(<3 x s8>), [[UV2:%[0-9]+]]:_(<3 x s8>), [[UV3:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>)
-    ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV]](<3 x s8>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT1]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[DEF]](s32)
+    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[BUILD_VECTOR]]
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR1]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s8>) = G_TRUNC %0
     %2:_(<3 x s8>) = G_FREEZE %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     $vgpr0_vgpr1_vgpr2 = COPY %3
 ...
 
 ---

 ---
 name: test_freeze_v3s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_freeze_v3s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR1]](<2 x s32>)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>), [[DEF1]](<2 x s16>)
-    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV3]](<3 x s16>), 0
-    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<4 x s16>) = G_FREEZE [[INSERT]]
-    ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[FREEZE]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR2]](<3 x s32>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>)
+    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<4 x s16>) = G_FREEZE [[CONCAT_VECTORS]]
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[FREEZE]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C1]](s32)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST2]](s32), [[LSHR]](s32), [[BITCAST3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>) = G_TRUNC %0
     %2:_(<3 x s16>) = G_FREEZE %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     $vgpr0_vgpr1_vgpr2 = COPY %3
 ...
 
 ---

 ---
 name: test_freeze_v5s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_freeze_v5s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<5 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR1]](<2 x s32>)
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_TRUNC [[BUILD_VECTOR2]](<2 x s32>)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[TRUNC]](<2 x s16>), [[TRUNC1]](<2 x s16>), [[TRUNC2]](<2 x s16>), [[DEF1]](<2 x s16>), [[DEF1]](<2 x s16>)
-    ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(<5 x s16>), [[UV6:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[UV5]](<5 x s16>), 0
-    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<6 x s16>) = G_FREEZE [[INSERT]]
-    ; CHECK-NEXT: [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[FREEZE]](<6 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32), [[LSHR1]](s32), [[BITCAST2]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4 = COPY [[BUILD_VECTOR3]](<5 x s32>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[UV3]], [[C]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C]]
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
+    ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(<6 x s16>) = G_FREEZE [[CONCAT_VECTORS]]
+    ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[FREEZE]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C1]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[BITCAST3]](s32), [[LSHR]](s32), [[BITCAST4]](s32), [[LSHR1]](s32), [[BITCAST5]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4 = COPY [[BUILD_VECTOR]](<5 x s32>)
     %0:_(<5 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4
     %1:_(<5 x s16>) = G_TRUNC %0
     %2:_(<5 x s16>) = G_FREEZE %1
     %3:_(<5 x s32>) = G_ANYEXT %2
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fshl.mir

     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
     ; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
     ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[COPY6]](s32)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[COPY7]](s32)
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
     ; GFX9-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[C1]](s32)
     ; GFX9-NEXT: [[AND:%[0-9]+]]:_(<2 x s16>) = G_AND [[BUILD_VECTOR_TRUNC4]], [[BUILD_VECTOR_TRUNC6]]
     ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 -1
     ; GFX9-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY [[C2]](s32)

     ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(<2 x s16>) = G_AND [[XOR1]], [[BUILD_VECTOR_TRUNC9]]
     ; GFX9-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
     ; GFX9-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY15]](s32), [[COPY16]](s32)
     ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC1]], [[AND2]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[BUILD_VECTOR_TRUNC3]], [[BUILD_VECTOR_TRUNC11]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR6:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[LSHR5]], [[AND3]](<2 x s16>)
     ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(<2 x s16>) = G_OR [[SHL1]], [[LSHR6]]
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[OR]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[OR1]](<2 x s16>)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[OR1]](<2 x s16>)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST6]](s32), [[BITCAST7]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR7]](s32), [[BITCAST8]](s32)
-    ; GFX9-NEXT: $vgpr0 = COPY [[OR]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC12]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC13]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST6]](s32), [[LSHR7]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST7]](s32), [[BITCAST8]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR8]](s32), [[BITCAST9]](s32)
+    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC12]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC13]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC14]](<2 x s16>)
     %0:_(<2 x s16>) = COPY $vgpr0
     %1:_(<2 x s16>) = COPY $vgpr1
     %2:_(<2 x s16>) = COPY $vgpr2
     %3:_(<2 x s16>) = COPY $vgpr3
     %4:_(<2 x s16>) = COPY $vgpr4
     %5:_(<2 x s16>) = COPY $vgpr5
     %6:_(<2 x s16>) = G_IMPLICIT_DEF
     %7:_(<6 x s16>) = G_CONCAT_VECTORS %0(<2 x s16>), %1(<2 x s16>), %6(<2 x s16>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fshr.mir

     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
     ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY6]], [[C]](s32)
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; SI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 1
     ; SI-NEXT: [[C4:%[0-9]+]]:_(s16) = G_CONSTANT i16 15
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[C3]], [[C4]]
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s16) = G_CONSTANT i16 -1
     ; SI-NEXT: [[XOR:%[0-9]+]]:_(s16) = G_XOR [[C3]], [[C5]]
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[XOR]], [[C4]]

     ; SI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR4]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s16) = G_OR [[TRUNC]], [[TRUNC1]]
     ; SI-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[C3]], [[C4]]
     ; SI-NEXT: [[XOR1:%[0-9]+]]:_(s16) = G_XOR [[C3]], [[C5]]
     ; SI-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[XOR1]], [[C4]]
     ; SI-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[AND7]](s16)
     ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[ZEXT2]](s32)
     ; SI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[SHL3]](s32)
-    ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
+    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
     ; SI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; SI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[AND9]], [[COPY7]](s32)
+    ; SI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[AND9]], [[COPY6]](s32)
     ; SI-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[AND8]](s16)
     ; SI-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
     ; SI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[AND10]], [[ZEXT3]](s32)
     ; SI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR6]](s32)
     ; SI-NEXT: [[OR3:%[0-9]+]]:_(s16) = G_OR [[TRUNC2]], [[TRUNC3]]
+    ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
+    ; SI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[COPY7]](s32)
     ; SI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[COPY8]](s32)
-    ; SI-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[COPY9]](s32)
+    ; SI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[COPY8]](s32)
+    ; SI-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
     ; SI-NEXT: [[COPY10:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; SI-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; SI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[COPY11]], [[C]](s32)
-    ; SI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[COPY10]], [[SHL6]]
+    ; SI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[COPY10]], [[C]](s32)
+    ; SI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[COPY9]], [[SHL6]]
     ; SI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
     ; SI-NEXT: [[XOR2:%[0-9]+]]:_(<2 x s16>) = G_XOR [[BITCAST6]], [[BITCAST8]]
     ; SI-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[XOR2]](<2 x s16>)
     ; SI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST9]](s32)
     ; SI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32)
     ; SI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR7]](s32)
     ; SI-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C4]]
     ; SI-NEXT: [[XOR3:%[0-9]+]]:_(s16) = G_XOR [[TRUNC4]], [[C5]]
     ; SI-NEXT: [[AND12:%[0-9]+]]:_(s16) = G_AND [[XOR3]], [[C4]]
     ; SI-NEXT: [[ZEXT4:%[0-9]+]]:_(s32) = G_ZEXT [[AND11]](s16)
     ; SI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[OR2]](s16)
     ; SI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[ANYEXT]], [[ZEXT4]](s32)
     ; SI-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[SHL7]](s32)
-    ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
+    ; SI-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
     ; SI-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[SHL4]], [[C1]]
-    ; SI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[AND13]], [[COPY12]](s32)
+    ; SI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[AND13]], [[COPY11]](s32)
     ; SI-NEXT: [[ZEXT5:%[0-9]+]]:_(s32) = G_ZEXT [[AND12]](s16)
     ; SI-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C1]]
     ; SI-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[AND14]], [[ZEXT5]](s32)
     ; SI-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
     ; SI-NEXT: [[OR5:%[0-9]+]]:_(s16) = G_OR [[TRUNC6]], [[TRUNC7]]
     ; SI-NEXT: [[AND15:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C4]]
     ; SI-NEXT: [[XOR4:%[0-9]+]]:_(s16) = G_XOR [[TRUNC5]], [[C5]]
     ; SI-NEXT: [[AND16:%[0-9]+]]:_(s16) = G_AND [[XOR4]], [[C4]]
     ; SI-NEXT: [[ZEXT6:%[0-9]+]]:_(s32) = G_ZEXT [[AND15]](s16)
     ; SI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[OR3]](s16)
     ; SI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[ANYEXT1]], [[ZEXT6]](s32)
     ; SI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[SHL8]](s32)
-    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
+    ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
     ; SI-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[SHL5]], [[C1]]
-    ; SI-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[AND17]], [[COPY13]](s32)
+    ; SI-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[AND17]], [[COPY12]](s32)
     ; SI-NEXT: [[ZEXT7:%[0-9]+]]:_(s32) = G_ZEXT [[AND16]](s16)
     ; SI-NEXT: [[AND18:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C1]]
     ; SI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[AND18]], [[ZEXT7]](s32)
     ; SI-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR11]](s32)
     ; SI-NEXT: [[OR6:%[0-9]+]]:_(s16) = G_OR [[TRUNC8]], [[TRUNC9]]
-    ; SI-NEXT: [[ZEXT8:%[0-9]+]]:_(s32) = G_ZEXT [[OR5]](s16)
-    ; SI-NEXT: [[ZEXT9:%[0-9]+]]:_(s32) = G_ZEXT [[OR6]](s16)
-    ; SI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXT9]], [[C]](s32)
-    ; SI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[ZEXT8]], [[SHL9]]
-    ; SI-NEXT: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
     ; SI-NEXT: [[AND19:%[0-9]+]]:_(s16) = G_AND [[C3]], [[C4]]
     ; SI-NEXT: [[XOR5:%[0-9]+]]:_(s16) = G_XOR [[C3]], [[C5]]
     ; SI-NEXT: [[AND20:%[0-9]+]]:_(s16) = G_AND [[XOR5]], [[C4]]
-    ; SI-NEXT: [[ZEXT10:%[0-9]+]]:_(s32) = G_ZEXT [[AND19]](s16)
-    ; SI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[ZEXT10]](s32)
-    ; SI-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[SHL10]](s32)
-    ; SI-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
+    ; SI-NEXT: [[ZEXT8:%[0-9]+]]:_(s32) = G_ZEXT [[AND19]](s16)
+    ; SI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[ZEXT8]](s32)
+    ; SI-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[SHL9]](s32)
+    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
     ; SI-NEXT: [[AND21:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; SI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[AND21]], [[COPY14]](s32)
-    ; SI-NEXT: [[ZEXT11:%[0-9]+]]:_(s32) = G_ZEXT [[AND20]](s16)
+    ; SI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[AND21]], [[COPY13]](s32)
+    ; SI-NEXT: [[ZEXT9:%[0-9]+]]:_(s32) = G_ZEXT [[AND20]](s16)
     ; SI-NEXT: [[AND22:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C1]]
-    ; SI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[AND22]], [[ZEXT11]](s32)
+    ; SI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[AND22]], [[ZEXT9]](s32)
     ; SI-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR13]](s32)
-    ; SI-NEXT: [[OR8:%[0-9]+]]:_(s16) = G_OR [[TRUNC10]], [[TRUNC11]]
-    ; SI-NEXT: [[AND23:%[0-9]+]]:_(s16) = G_AND [[C3]], [[C4]]
-    ; SI-NEXT: [[XOR6:%[0-9]+]]:_(s16) = G_XOR [[C3]], [[C5]]
-    ; SI-NEXT: [[AND24:%[0-9]+]]:_(s16) = G_AND [[XOR6]], [[C4]]
-    ; SI-NEXT: [[ZEXT12:%[0-9]+]]:_(s32) = G_ZEXT [[AND23]](s16)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[DEF]], [[ZEXT12]](s32)
-    ; SI-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[SHL11]](s32)
-    ; SI-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[C2]], [[COPY15]](s32)
-    ; SI-NEXT: [[ZEXT13:%[0-9]+]]:_(s32) = G_ZEXT [[AND24]](s16)
-    ; SI-NEXT: [[AND25:%[0-9]+]]:_(s32) = G_AND [[LSHR14]], [[C1]]
-    ; SI-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[AND25]], [[ZEXT13]](s32)
-    ; SI-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
-    ; SI-NEXT: [[OR9:%[0-9]+]]:_(s16) = G_OR [[TRUNC12]], [[TRUNC13]]
-    ; SI-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[COPY16]](s32)
+    ; SI-NEXT: [[OR7:%[0-9]+]]:_(s16) = G_OR [[TRUNC10]], [[TRUNC11]]
+    ; SI-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
+    ; SI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[COPY14]](s32)
+    ; SI-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
+    ; SI-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
+    ; SI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[COPY16]], [[C]](s32)
+    ; SI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[COPY15]], [[SHL11]]
+    ; SI-NEXT: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
+    ; SI-NEXT: [[XOR6:%[0-9]+]]:_(<2 x s16>) = G_XOR [[BITCAST7]], [[BITCAST10]]
+    ; SI-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[XOR6]](<2 x s16>)
+    ; SI-NEXT: [[TRUNC12:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST11]](s32)
+    ; SI-NEXT: [[AND23:%[0-9]+]]:_(s16) = G_AND [[TRUNC12]], [[C4]]
+    ; SI-NEXT: [[XOR7:%[0-9]+]]:_(s16) = G_XOR [[TRUNC12]], [[C5]]
+    ; SI-NEXT: [[AND24:%[0-9]+]]:_(s16) = G_AND [[XOR7]], [[C4]]
+    ; SI-NEXT: [[ZEXT10:%[0-9]+]]:_(s32) = G_ZEXT [[AND23]](s16)
+    ; SI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[OR7]](s16)
+    ; SI-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[ANYEXT2]], [[ZEXT10]](s32)
+    ; SI-NEXT: [[TRUNC13:%[0-9]+]]:_(s16) = G_TRUNC [[SHL12]](s32)
     ; SI-NEXT: [[COPY17:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[COPY18:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
-    ; SI-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[COPY17]](s32)
-    ; SI-NEXT: [[COPY19:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; SI-NEXT: [[COPY20:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; SI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[COPY20]], [[C]](s32)
-    ; SI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[COPY19]], [[SHL14]]
-    ; SI-NEXT: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
-    ; SI-NEXT: [[XOR7:%[0-9]+]]:_(<2 x s16>) = G_XOR [[BITCAST7]], [[BITCAST11]]
-    ; SI-NEXT: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[XOR7]](<2 x s16>)
-    ; SI-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST12]](s32)
+    ; SI-NEXT: [[AND25:%[0-9]+]]:_(s32) = G_AND [[SHL10]], [[C1]]
+    ; SI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[AND25]], [[COPY17]](s32)
+    ; SI-NEXT: [[ZEXT11:%[0-9]+]]:_(s32) = G_ZEXT [[AND24]](s16)
+    ; SI-NEXT: [[AND26:%[0-9]+]]:_(s32) = G_AND [[LSHR14]], [[C1]]
+    ; SI-NEXT: [[LSHR15:%[0-9]+]]:_(s32) = G_LSHR [[AND26]], [[ZEXT11]](s32)
+    ; SI-NEXT: [[TRUNC14:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR15]](s32)
+    ; SI-NEXT: [[OR9:%[0-9]+]]:_(s16) = G_OR [[TRUNC13]], [[TRUNC14]]
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; SI-NEXT: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST12]], [[C]](s32)
-    ; SI-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
-    ; SI-NEXT: [[AND26:%[0-9]+]]:_(s16) = G_AND [[TRUNC14]], [[C4]]
-    ; SI-NEXT: [[XOR8:%[0-9]+]]:_(s16) = G_XOR [[TRUNC14]], [[C5]]
-    ; SI-NEXT: [[AND27:%[0-9]+]]:_(s16) = G_AND [[XOR8]], [[C4]]
-    ; SI-NEXT: [[ZEXT14:%[0-9]+]]:_(s32) = G_ZEXT [[AND26]](s16)
-    ; SI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[OR8]](s16)
-    ; SI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[ANYEXT2]], [[ZEXT14]](s32)
-    ; SI-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[SHL15]](s32)
-    ; SI-NEXT: [[COPY21:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[AND28:%[0-9]+]]:_(s32) = G_AND [[SHL12]], [[C1]]
-    ; SI-NEXT: [[LSHR17:%[0-9]+]]:_(s32) = G_LSHR [[AND28]], [[COPY21]](s32)
-    ; SI-NEXT: [[ZEXT15:%[0-9]+]]:_(s32) = G_ZEXT [[AND27]](s16)
-    ; SI-NEXT: [[AND29:%[0-9]+]]:_(s32) = G_AND [[LSHR17]], [[C1]]
-    ; SI-NEXT: [[LSHR18:%[0-9]+]]:_(s32) = G_LSHR [[AND29]], [[ZEXT15]](s32)
-    ; SI-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR18]](s32)
-    ; SI-NEXT: [[OR11:%[0-9]+]]:_(s16) = G_OR [[TRUNC16]], [[TRUNC17]]
-    ; SI-NEXT: [[AND30:%[0-9]+]]:_(s16) = G_AND [[TRUNC15]], [[C4]]
-    ; SI-NEXT: [[XOR9:%[0-9]+]]:_(s16) = G_XOR [[TRUNC15]], [[C5]]
-    ; SI-NEXT: [[AND31:%[0-9]+]]:_(s16) = G_AND [[XOR9]], [[C4]]
-    ; SI-NEXT: [[ZEXT16:%[0-9]+]]:_(s32) = G_ZEXT [[AND30]](s16)
-    ; SI-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[OR9]](s16)
-    ; SI-NEXT: [[SHL16:%[0-9]+]]:_(s32) = G_SHL [[ANYEXT3]], [[ZEXT16]](s32)
-    ; SI-NEXT: [[TRUNC18:%[0-9]+]]:_(s16) = G_TRUNC [[SHL16]](s32)
-    ; SI-NEXT: [[COPY22:%[0-9]+]]:_(s32) = COPY [[C6]](s32)
-    ; SI-NEXT: [[AND32:%[0-9]+]]:_(s32) = G_AND [[SHL13]], [[C1]]
-    ; SI-NEXT: [[LSHR19:%[0-9]+]]:_(s32) = G_LSHR [[AND32]], [[COPY22]](s32)
-    ; SI-NEXT: [[ZEXT17:%[0-9]+]]:_(s32) = G_ZEXT [[AND31]](s16)
-    ; SI-NEXT: [[AND33:%[0-9]+]]:_(s32) = G_AND [[LSHR19]], [[C1]]
-    ; SI-NEXT: [[LSHR20:%[0-9]+]]:_(s32) = G_LSHR [[AND33]], [[ZEXT17]](s32)
-    ; SI-NEXT: [[TRUNC19:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR20]](s32)
-    ; SI-NEXT: [[OR12:%[0-9]+]]:_(s16) = G_OR [[TRUNC18]], [[TRUNC19]]
-    ; SI-NEXT: [[ZEXT18:%[0-9]+]]:_(s32) = G_ZEXT [[OR11]](s16)
-    ; SI-NEXT: [[ZEXT19:%[0-9]+]]:_(s32) = G_ZEXT [[OR12]](s16)
-    ; SI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXT19]], [[C]](s32)
-    ; SI-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[ZEXT18]], [[SHL17]]
-    ; SI-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR13]](s32)
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[BITCAST14:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST13]](<2 x s16>)
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; SI-NEXT: [[BITCAST15:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST15]], [[C]](s32)
-    ; SI-NEXT: [[BITCAST16:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; SI-NEXT: [[AND34:%[0-9]+]]:_(s32) = G_AND [[BITCAST14]], [[C1]]
-    ; SI-NEXT: [[AND35:%[0-9]+]]:_(s32) = G_AND [[BITCAST15]], [[C1]]
-    ; SI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[AND35]], [[C]](s32)
-    ; SI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[AND34]], [[SHL18]]
-    ; SI-NEXT: [[BITCAST17:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR14]](s32)
-    ; SI-NEXT: [[AND36:%[0-9]+]]:_(s32) = G_AND [[LSHR21]], [[C1]]
-    ; SI-NEXT: [[AND37:%[0-9]+]]:_(s32) = G_AND [[BITCAST16]], [[C1]]
-    ; SI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[AND37]], [[C]](s32)
-    ; SI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[AND36]], [[SHL19]]
-    ; SI-NEXT: [[BITCAST18:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR15]](s32)
-    ; SI-NEXT: $vgpr0 = COPY [[BITCAST10]](<2 x s16>)
-    ; SI-NEXT: $vgpr1 = COPY [[BITCAST17]](<2 x s16>)
-    ; SI-NEXT: $vgpr2 = COPY [[BITCAST18]](<2 x s16>)
+    ; SI-NEXT: [[BITCAST13:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[ZEXT12:%[0-9]+]]:_(s32) = G_ZEXT [[OR5]](s16)
+    ; SI-NEXT: [[ZEXT13:%[0-9]+]]:_(s32) = G_ZEXT [[OR6]](s16)
+    ; SI-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[ZEXT13]], [[C]](s32)
+    ; SI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[ZEXT12]], [[SHL13]]
+    ; SI-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
+    ; SI-NEXT: [[ZEXT14:%[0-9]+]]:_(s32) = G_ZEXT [[OR9]](s16)
+    ; SI-NEXT: [[AND27:%[0-9]+]]:_(s32) = G_AND [[BITCAST12]], [[C1]]
+    ; SI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[AND27]], [[C]](s32)
+    ; SI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[ZEXT14]], [[SHL14]]
+    ; SI-NEXT: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR11]](s32)
+    ; SI-NEXT: [[AND28:%[0-9]+]]:_(s32) = G_AND [[LSHR16]], [[C1]]
+    ; SI-NEXT: [[AND29:%[0-9]+]]:_(s32) = G_AND [[BITCAST13]], [[C1]]
+    ; SI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[AND29]], [[C]](s32)
+    ; SI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND28]], [[SHL15]]
+    ; SI-NEXT: [[BITCAST16:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR12]](s32)
+    ; SI-NEXT: $vgpr0 = COPY [[BITCAST14]](<2 x s16>)
+    ; SI-NEXT: $vgpr1 = COPY [[BITCAST15]](<2 x s16>)
+    ; SI-NEXT: $vgpr2 = COPY [[BITCAST16]](<2 x s16>)
     ; VI-LABEL: name: test_fshr_v3s16_v3s16
     ; VI: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; VI-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
     ; VI-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
     ; VI-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
     ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)

     ; VI-NEXT: [[OR5:%[0-9]+]]:_(s16) = G_OR [[SHL7]], [[LSHR9]]
     ; VI-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC7]], [[C4]]
     ; VI-NEXT: [[XOR4:%[0-9]+]]:_(s16) = G_XOR [[TRUNC7]], [[C5]]
     ; VI-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[XOR4]], [[C4]]
     ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s16) = G_SHL [[OR3]], [[AND9]](s16)
     ; VI-NEXT: [[LSHR10:%[0-9]+]]:_(s16) = G_LSHR [[SHL5]], [[C3]](s16)
     ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s16) = G_LSHR [[LSHR10]], [[AND10]](s16)
     ; VI-NEXT: [[OR6:%[0-9]+]]:_(s16) = G_OR [[SHL8]], [[LSHR11]]
-    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR5]](s16)
-    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR6]](s16)
-    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
-    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL9]]
-    ; VI-NEXT: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
     ; VI-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[C3]], [[C4]]
     ; VI-NEXT: [[XOR5:%[0-9]+]]:_(s16) = G_XOR [[C3]], [[C5]]
     ; VI-NEXT: [[AND12:%[0-9]+]]:_(s16) = G_AND [[XOR5]], [[C4]]
-    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[AND11]](s16)
+    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[AND11]](s16)
     ; VI-NEXT: [[LSHR12:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC5]], [[C3]](s16)
     ; VI-NEXT: [[LSHR13:%[0-9]+]]:_(s16) = G_LSHR [[LSHR12]], [[AND12]](s16)
-    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s16) = G_OR [[SHL10]], [[LSHR13]]
-    ; VI-NEXT: [[AND13:%[0-9]+]]:_(s16) = G_AND [[C3]], [[C4]]
-    ; VI-NEXT: [[XOR6:%[0-9]+]]:_(s16) = G_XOR [[C3]], [[C5]]
-    ; VI-NEXT: [[AND14:%[0-9]+]]:_(s16) = G_AND [[XOR6]], [[C4]]
-    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s16) = G_SHL [[DEF]], [[AND13]](s16)
-    ; VI-NEXT: [[LSHR14:%[0-9]+]]:_(s16) = G_LSHR [[DEF]], [[C3]](s16)
-    ; VI-NEXT: [[LSHR15:%[0-9]+]]:_(s16) = G_LSHR [[LSHR14]], [[AND14]](s16)
-    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s16) = G_OR [[SHL11]], [[LSHR15]]
-    ; VI-NEXT: [[SHL12:%[0-9]+]]:_(s16) = G_SHL [[TRUNC5]], [[C3]](s16)
-    ; VI-NEXT: [[SHL13:%[0-9]+]]:_(s16) = G_SHL [[DEF]], [[C3]](s16)
+    ; VI-NEXT: [[OR7:%[0-9]+]]:_(s16) = G_OR [[SHL9]], [[LSHR13]]
+    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s16) = G_SHL [[TRUNC5]], [[C3]](s16)
     ; VI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
     ; VI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; VI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[COPY8]], [[C]](s32)
-    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[COPY7]], [[SHL14]]
-    ; VI-NEXT: [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
-    ; VI-NEXT: [[XOR7:%[0-9]+]]:_(<2 x s16>) = G_XOR [[BITCAST7]], [[BITCAST11]]
-    ; VI-NEXT: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[XOR7]](<2 x s16>)
-    ; VI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST12]](s32)
+    ; VI-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[COPY8]], [[C]](s32)
+    ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[COPY7]], [[SHL11]]
+    ; VI-NEXT: [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
+    ; VI-NEXT: [[XOR6:%[0-9]+]]:_(<2 x s16>) = G_XOR [[BITCAST7]], [[BITCAST10]]
+    ; VI-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[XOR6]](<2 x s16>)
+    ; VI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST11]](s32)
+    ; VI-NEXT: [[AND13:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C4]]
+    ; VI-NEXT: [[XOR7:%[0-9]+]]:_(s16) = G_XOR [[TRUNC8]], [[C5]]
+    ; VI-NEXT: [[AND14:%[0-9]+]]:_(s16) = G_AND [[XOR7]], [[C4]]
+    ; VI-NEXT: [[SHL12:%[0-9]+]]:_(s16) = G_SHL [[OR7]], [[AND13]](s16)
+    ; VI-NEXT: [[LSHR14:%[0-9]+]]:_(s16) = G_LSHR [[SHL10]], [[C3]](s16)
+    ; VI-NEXT: [[LSHR15:%[0-9]+]]:_(s16) = G_LSHR [[LSHR14]], [[AND14]](s16)
+    ; VI-NEXT: [[OR9:%[0-9]+]]:_(s16) = G_OR [[SHL12]], [[LSHR15]]
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; VI-NEXT: [[BITCAST12:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[LSHR16:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST12]], [[C]](s32)
-    ; VI-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR16]](s32)
-    ; VI-NEXT: [[AND15:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C4]]
-    ; VI-NEXT: [[XOR8:%[0-9]+]]:_(s16) = G_XOR [[TRUNC8]], [[C5]]
-    ; VI-NEXT: [[AND16:%[0-9]+]]:_(s16) = G_AND [[XOR8]], [[C4]]
-    ; VI-NEXT: [[SHL15:%[0-9]+]]:_(s16) = G_SHL [[OR8]], [[AND15]](s16)
-    ; VI-NEXT: [[LSHR17:%[0-9]+]]:_(s16) = G_LSHR [[SHL12]], [[C3]](s16)
-    ; VI-NEXT: [[LSHR18:%[0-9]+]]:_(s16) = G_LSHR [[LSHR17]], [[AND16]](s16)
-    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s16) = G_OR [[SHL15]], [[LSHR18]]
-    ; VI-NEXT: [[AND17:%[0-9]+]]:_(s16) = G_AND [[TRUNC9]], [[C4]]
-    ; VI-NEXT: [[XOR9:%[0-9]+]]:_(s16) = G_XOR [[TRUNC9]], [[C5]]
-    ; VI-NEXT: [[AND18:%[0-9]+]]:_(s16) = G_AND [[XOR9]], [[C4]]
-    ; VI-NEXT: [[SHL16:%[0-9]+]]:_(s16) = G_SHL [[OR9]], [[AND17]](s16)
-    ; VI-NEXT: [[LSHR19:%[0-9]+]]:_(s16) = G_LSHR [[SHL13]], [[C3]](s16)
-    ; VI-NEXT: [[LSHR20:%[0-9]+]]:_(s16) = G_LSHR [[LSHR19]], [[AND18]](s16)
-    ; VI-NEXT: [[OR12:%[0-9]+]]:_(s16) = G_OR [[SHL16]], [[LSHR20]]
-    ; VI-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[OR11]](s16)
-    ; VI-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[OR12]](s16)
-    ; VI-NEXT: [[SHL17:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32)
-    ; VI-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL17]]
-    ; VI-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR13]](s32)
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[BITCAST14:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST13]](<2 x s16>)
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; VI-NEXT: [[BITCAST15:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[LSHR21:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST15]], [[C]](s32)
-    ; VI-NEXT: [[BITCAST16:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; VI-NEXT: [[AND19:%[0-9]+]]:_(s32) = G_AND [[BITCAST14]], [[C1]]
-    ; VI-NEXT: [[AND20:%[0-9]+]]:_(s32) = G_AND [[BITCAST15]], [[C1]]
-    ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[AND20]], [[C]](s32)
-    ; VI-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[AND19]], [[SHL18]]
-    ; VI-NEXT: [[BITCAST17:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR14]](s32)
-    ; VI-NEXT: [[AND21:%[0-9]+]]:_(s32) = G_AND [[LSHR21]], [[C1]]
-    ; VI-NEXT: [[AND22:%[0-9]+]]:_(s32) = G_AND [[BITCAST16]], [[C1]]
-    ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[AND22]], [[C]](s32)
-    ; VI-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[AND21]], [[SHL19]]
-    ; VI-NEXT: [[BITCAST18:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR15]](s32)
-    ; VI-NEXT: $vgpr0 = COPY [[BITCAST10]](<2 x s16>)
-    ; VI-NEXT: $vgpr1 = COPY [[BITCAST17]](<2 x s16>)
-    ; VI-NEXT: $vgpr2 = COPY [[BITCAST18]](<2 x s16>)
+    ; VI-NEXT: [[BITCAST13:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR5]](s16)
+    ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR6]](s16)
+    ; VI-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
+    ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL13]]
+    ; VI-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
+    ; VI-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[OR9]](s16)
+    ; VI-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[BITCAST12]], [[C1]]
+    ; VI-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C]](s32)
+    ; VI-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL14]]
+    ; VI-NEXT: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR11]](s32)
+    ; VI-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[LSHR16]], [[C1]]
+    ; VI-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[BITCAST13]], [[C1]]
+    ; VI-NEXT: [[SHL15:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C]](s32)
+    ; VI-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL15]]
+    ; VI-NEXT: [[BITCAST16:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR12]](s32)
+    ; VI-NEXT: $vgpr0 = COPY [[BITCAST14]](<2 x s16>)
+    ; VI-NEXT: $vgpr1 = COPY [[BITCAST15]](<2 x s16>)
+    ; VI-NEXT: $vgpr2 = COPY [[BITCAST16]](<2 x s16>)
     ; GFX9-LABEL: name: test_fshr_v3s16_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
     ; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
     ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[COPY6]](s32)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[DEF]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[COPY7]](s32)
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
     ; GFX9-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY8]](s32), [[C1]](s32)
     ; GFX9-NEXT: [[AND:%[0-9]+]]:_(<2 x s16>) = G_AND [[BUILD_VECTOR_TRUNC4]], [[BUILD_VECTOR_TRUNC6]]
     ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 -1
     ; GFX9-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY [[C2]](s32)

     ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(<2 x s16>) = G_AND [[XOR1]], [[BUILD_VECTOR_TRUNC9]]
     ; GFX9-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
     ; GFX9-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY [[C3]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC11:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY15]](s32), [[COPY16]](s32)
     ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC11]](<2 x s16>)
     ; GFX9-NEXT: [[SHL3:%[0-9]+]]:_(<2 x s16>) = G_SHL [[SHL2]], [[AND3]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[BUILD_VECTOR_TRUNC3]], [[AND2]](<2 x s16>)
     ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(<2 x s16>) = G_OR [[SHL3]], [[LSHR4]]
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[OR]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[OR1]](<2 x s16>)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[OR1]](<2 x s16>)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST6]](s32), [[BITCAST7]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR5]](s32), [[BITCAST8]](s32)
-    ; GFX9-NEXT: $vgpr0 = COPY [[OR]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC12]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC13]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC12:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST6]](s32), [[LSHR5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC13:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST7]](s32), [[BITCAST8]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC14:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR6]](s32), [[BITCAST9]](s32)
+    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC12]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC13]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC14]](<2 x s16>)
     %0:_(<2 x s16>) = COPY $vgpr0
     %1:_(<2 x s16>) = COPY $vgpr1
     %2:_(<2 x s16>) = COPY $vgpr2
     %3:_(<2 x s16>) = COPY $vgpr3
     %4:_(<2 x s16>) = COPY $vgpr4
     %5:_(<2 x s16>) = COPY $vgpr5
     %6:_(<2 x s16>) = G_IMPLICIT_DEF
     %7:_(<6 x s16>) = G_CONCAT_VECTORS %0(<2 x s16>), %1(<2 x s16>), %6(<2 x s16>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-implicit-def.mir

 ---
 name: test_implicit_def_v33s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: test_implicit_def_v33s32
     ; CHECK: [[DEF:%[0-9]+]]:_(<16 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>)
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<16 x s32>)
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: G_STORE [[UV]](s32), [[COPY]](p1) :: (volatile store (s32), addrspace 1)
-    ; CHECK-NEXT: G_STORE [[UV16]](s32), [[COPY]](p1) :: (volatile store (s32), addrspace 1)
+    ; CHECK-NEXT: G_STORE [[DEF1]](s32), [[COPY]](p1) :: (volatile store (s32), addrspace 1)
     %0:_(<33 x s32>) = G_IMPLICIT_DEF
     %1:_(s32), %2:_(s32), %3:_(s32), %4:_(s32), %5:_(s32), %6:_(s32), %7:_(s32), %8:_(s32), %9:_(s32), %10:_(s32), %11:_(s32), %12:_(s32), %13:_(s32), %14:_(s32), %15:_(s32), %16:_(s32), %17:_(s32), %18:_(s32), %19:_(s32), %20:_(s32), %21:_(s32), %22:_(s32), %23:_(s32), %24:_(s32), %25:_(s32), %26:_(s32), %27:_(s32), %28:_(s32), %29:_(s32), %30:_(s32), %31:_(s32), %32:_(s32), %33:_(s32) = G_UNMERGE_VALUES %0
     %34:_(p1) = COPY $vgpr0_vgpr1
     G_STORE %1, %34 :: (volatile store (s32), align 4, addrspace 1)
     G_STORE %33, %34 :: (volatile store (s32), align 4, addrspace 1)
 
 ...
 

 
 ---
 name: test_implicit_def_v3s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_implicit_def_v3s16
     ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<4 x s16>) = G_IMPLICIT_DEF
     %2:_(<4 x s16>) = G_INSERT %1, %0, 0
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name: test_implicit_def_v4s16

 
 ---
 name: test_implicit_def_v5s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_implicit_def_v5s16
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF2]], [[UV]](<5 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<8 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<8 x s16>)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<8 x s16>)
     %0:_(<5 x s16>) = G_IMPLICIT_DEF
     %1:_(<8 x s16>) = G_IMPLICIT_DEF
     %2:_(<8 x s16>) = G_INSERT %1, %0, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 
 ---
 name: test_implicit_def_v6s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_implicit_def_v6s16
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF1]], [[DEF]](<6 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<8 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<8 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS]](<8 x s16>)
     %0:_(<6 x s16>) = G_IMPLICIT_DEF
     %1:_(<8 x s16>) = G_IMPLICIT_DEF
     %2:_(<8 x s16>) = G_INSERT %1, %0, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 
 ---
 name: test_implicit_def_v8s16

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-insert.mir

 name: test_insert_v2s32_s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v2s32_s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<2 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<2 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[UV1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
     %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(<2 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v2s32_s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v2s32_s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<2 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<2 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[COPY1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
     %0:_(<2 x s32>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(<2 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v3s32_s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3
 
     ; CHECK-LABEL: name: test_insert_v3s32_s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr3
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<3 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[INSERT]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(s32) = COPY $vgpr3
     %2:_(<3 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 ---
 name: test_insert_v3s32_s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3
 
     ; CHECK-LABEL: name: test_insert_v3s32_s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr3
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<3 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[INSERT]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[COPY1]](s32), [[UV2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(s32) = COPY $vgpr3
     %2:_(<3 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 ---
 name: test_insert_v3s32_s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3
 
     ; CHECK-LABEL: name: test_insert_v3s32_s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr3
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<3 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[INSERT]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[COPY1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(s32) = COPY $vgpr3
     %2:_(<3 x s32>) = G_INSERT %0, %1, 64
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4
 
     ; CHECK-LABEL: name: test_insert_v4s32_s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr4
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = COPY $vgpr4
     %2:_(<4 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4
 
     ; CHECK-LABEL: name: test_insert_v4s32_s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr4
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[COPY1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = COPY $vgpr4
     %2:_(<4 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4
 
     ; CHECK-LABEL: name: test_insert_v4s32_s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr4
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[COPY1]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = COPY $vgpr4
     %2:_(<4 x s32>) = G_INSERT %0, %1, 64
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s32_offset96
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4
 
     ; CHECK-LABEL: name: test_insert_v4s32_s32_offset96
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr4
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s32), 96
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[COPY1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = COPY $vgpr4
     %2:_(<4 x s32>) = G_INSERT %0, %1, 96
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s64_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_insert_v4s32_s64_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr4_vgpr5
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s64), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s64) = COPY $vgpr4_vgpr5
     %2:_(<4 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s64_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_insert_v4s32_s64_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr4_vgpr5
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s64), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV4]](s32), [[UV5]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s64) = COPY $vgpr4_vgpr5
     %2:_(<4 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s64_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_insert_v4s32_s64_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr4_vgpr5
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s64), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s64)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV4]](s32), [[UV5]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s64) = COPY $vgpr4_vgpr5
     %2:_(<4 x s32>) = G_INSERT %0, %1, 64
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s96_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6
 
     ; CHECK-LABEL: name: test_insert_v4s32_s96_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr4_vgpr5_vgpr6
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s96), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s96)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s96) = COPY $vgpr4_vgpr5_vgpr6
     %2:_(<4 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_s96_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6
 
     ; CHECK-LABEL: name: test_insert_v4s32_s96_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr4_vgpr5_vgpr6
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](s96), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](s96)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s96) = COPY $vgpr4_vgpr5_vgpr6
     %2:_(<4 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_v2s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_insert_v4s32_v2s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr4_vgpr5
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](<2 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<2 x s32>) = COPY $vgpr4_vgpr5
     %2:_(<4 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_v2s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_insert_v4s32_v2s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr4_vgpr5
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](<2 x s32>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV4]](s32), [[UV5]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<2 x s32>) = COPY $vgpr4_vgpr5
     %2:_(<4 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_v2s32_offset64
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_insert_v4s32_v2s32_offset64
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s32>) = COPY $vgpr4_vgpr5
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](<2 x s32>), 64
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV4]](s32), [[UV5]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<2 x s32>) = COPY $vgpr4_vgpr5
     %2:_(<4 x s32>) = G_INSERT %0, %1, 64
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_v3s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6
 
     ; CHECK-LABEL: name: test_insert_v4s32_v3s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr4_vgpr5_vgpr6
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](<3 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV3]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<3 x s32>) = COPY $vgpr4_vgpr5_vgpr6
     %2:_(<4 x s32>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_v3s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6
 
     ; CHECK-LABEL: name: test_insert_v4s32_v3s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr4_vgpr5_vgpr6
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s32>) = G_INSERT [[COPY]], [[COPY1]](<3 x s32>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT]](<4 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<4 x s32>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<4 x s32>)
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(<3 x s32>) = COPY $vgpr4_vgpr5_vgpr6
     %2:_(<4 x s32>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %2
 ...
 ---
 name: test_insert_v4s32_p0_offset0
 body: |

 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; CHECK-LABEL: name: test_insert_v2s16_s16_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
-    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -65536
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[AND]]
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     %0:_(<2 x s16>) = COPY $vgpr0
     %1:_(s32) = COPY $vgpr1
     %2:_(s16) = G_TRUNC %1
     %3:_(<2 x s16>) = G_INSERT %0, %2, 0
     $vgpr0 = COPY %3
 ...
body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; CHECK-LABEL: name: test_insert_v2s16_s16_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C]]
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C1]](s32)
-    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C]]
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     %0:_(<2 x s16>) = COPY $vgpr0
     %1:_(s32) = COPY $vgpr1
     %2:_(s16) = G_TRUNC %1
     %3:_(<2 x s16>) = G_INSERT %0, %2, 16
     $vgpr0 = COPY %3
 ...
 ---
 name: test_insert_v3s16_s16_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_s16_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[TRUNC]](s16), 0
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(s32) = COPY $vgpr2
     %3:_(s16) = G_TRUNC %2
     %4:_(<3 x s16>) = G_INSERT %1, %3, 0
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v3s16_s16_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_s16_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[TRUNC]](s16), 16
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(s32) = COPY $vgpr2
     %3:_(s16) = G_TRUNC %2
     %4:_(<3 x s16>) = G_INSERT %1, %3, 16
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v3s16_s16_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_s16_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[TRUNC]](s16), 32
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(s32) = COPY $vgpr2
     %3:_(s16) = G_TRUNC %2
     %4:_(<3 x s16>) = G_INSERT %1, %3, 32
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v3s16_v2s16_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_v2s16_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](<2 x s16>), 0
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(<2 x s16>) = COPY $vgpr2
     %4:_(<3 x s16>) = G_INSERT %1, %2, 0
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v3s16_v2s16_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_v2s16_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](<2 x s16>), 16
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(<2 x s16>) = COPY $vgpr2
     %4:_(<3 x s16>) = G_INSERT %1, %2, 16
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v3s16_s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](s32), 0
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(s32) = COPY $vgpr2
     %4:_(<3 x s16>) = G_INSERT %1, %2, 0
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v3s16_s32_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v3s16_s32_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32)
     ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[INSERT]], [[COPY1]](s32), 16
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[INSERT1]](<4 x s16>), [[DEF]](<4 x s16>), [[DEF]](<4 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<3 x s16>) = G_EXTRACT %0, 0
     %2:_(s32) = COPY $vgpr2
     %4:_(<3 x s16>) = G_INSERT %1, %2, 16
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
 ...
 ---
 name: test_insert_v4s16_s16_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s16_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[TRUNC]](s16), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(s16) = G_TRUNC %1
     %3:_(<4 x s16>) = G_INSERT %0, %2, 0
     $vgpr0_vgpr1 = COPY %3
 ...
 ---
 name: test_insert_v4s16_s16_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s16_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[TRUNC]](s16), 16
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(s16) = G_TRUNC %1
     %3:_(<4 x s16>) = G_INSERT %0, %2, 16
     $vgpr0_vgpr1 = COPY %3
 ...
 ---
 name: test_insert_v4s16_s16_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s16_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[TRUNC]](s16), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(s16) = G_TRUNC %1
     %3:_(<4 x s16>) = G_INSERT %0, %2, 32
     $vgpr0_vgpr1 = COPY %3
 ...
 ---
 name: test_insert_v4s16_s16_offset48
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s16_offset48
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[TRUNC]](s16), 48
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(s16) = G_TRUNC %1
     %3:_(<4 x s16>) = G_INSERT %0, %2, 48
     $vgpr0_vgpr1 = COPY %3
 ...
 ---
 name: test_insert_v4s16_v2s16_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_v2s16_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[COPY1]](<2 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<2 x s16>) = COPY $vgpr2
     %2:_(<4 x s16>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v4s16_v2s16_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_v2s16_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[COPY1]](<2 x s16>), 16
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<2 x s16>) = COPY $vgpr2
     %2:_(<4 x s16>) = G_INSERT %0, %1, 16
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v4s16_v2s16_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_v2s16_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[COPY1]](<2 x s16>), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<2 x s16>) = COPY $vgpr2
     %2:_(<4 x s16>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v4s16_v3s16_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_insert_v4s16_v3s16_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<4 x s16>) = COPY $vgpr2_vgpr3
     %2:_(<3 x s16>) = G_EXTRACT %1, 0
     %3:_(<4 x s16>) = G_INSERT %0, %2, 0
     $vgpr0_vgpr1 = COPY %3
 ...
 ---
 name: test_insert_v4s16_v3s16_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
 
     ; CHECK-LABEL: name: test_insert_v4s16_v3s16_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[EXTRACT]](<3 x s16>), 16
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<4 x s16>) = COPY $vgpr2_vgpr3
     %2:_(<3 x s16>) = G_EXTRACT %1, 0
     %3:_(<4 x s16>) = G_INSERT %0, %2, 16
     $vgpr0_vgpr1 = COPY %3
 ...
 ---
 name: test_insert_v4s16_s32_offset0
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s32_offset0
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[COPY1]](s32), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(<4 x s16>) = G_INSERT %0, %1, 0
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v4s16_s32_offset16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s32_offset16
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[COPY1]](s32), 16
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(<4 x s16>) = G_INSERT %0, %1, 16
     $vgpr0_vgpr1 = COPY %2
 ...
 ---
 name: test_insert_v4s16_s32_offset32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; CHECK-LABEL: name: test_insert_v4s16_s32_offset32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[COPY]], [[COPY1]](s32), 32
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY1]], [[C]](s32)
+    ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY1]], [[C1]]
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(<4 x s16>) = G_INSERT %0, %1, 32
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name: test_insert_s64_s16_offset0

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.load.2d.d16.ll

define amdgpu_ps <3 x half> @image_load_v3f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<3 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<3 x s16>) from custom "ImageResource", align 8)
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<3 x s32>)
-  ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
+  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_v3f16
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s16>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<3 x s16>) from custom "ImageResource", align 8)
   ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s16>)
-  ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
   ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[UV]](<2 x s16>)
-  ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; PACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
+  ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST3]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret <3 x half> %tex
 }
 
 define amdgpu_ps <4 x half> @image_load_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-LABEL: name: image_load_v4f16
   ; UNPACKED: bb.1 (%ir-block.0):
define amdgpu_ps <3 x half> @image_load_tfe_v3f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (<3 x s16>) from custom "ImageResource", align 8)
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
   ; UNPACKED-NEXT:   G_STORE [[UV3]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; UNPACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C]]
+  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_tfe_v3f16
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<3 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (<3 x s16>) from custom "ImageResource", align 8)
   ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<3 x s32>)
   ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[UV]](s32)
   ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[UV1]](s32)
-  ; PACKED-NEXT:   G_STORE [[UV2]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST1]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST1]](<2 x s16>)
+  ; PACKED-NEXT:   G_STORE [[UV2]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
   ; PACKED-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
-  ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST4]](<2 x s16>)
+  ; PACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST4]](<2 x s16>)
+  ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST5]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %tex = extractvalue { <3 x half>, i32 } %res, 0
   %tfe = extractvalue { <3 x half>, i32 } %res, 1
   store i32 %tfe, i32 addrspace(1)* undef
   ret <3 x half> %tex
 }
 
define amdgpu_ps <3 x half> @image_load_v3f16_dmask_1100(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 3, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<2 x s16>) from custom "ImageResource")
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
-  ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
   ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-  ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_v3f16_dmask_1100
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s16>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 3, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<2 x s16>) from custom "ImageResource")
-  ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[DEF]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
   ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret <3 x half> %tex
 }
 
 define amdgpu_ps <3 x half> @image_load_v3f16_dmask_1000(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-LABEL: name: image_load_v3f16_dmask_1000
   ; UNPACKED: bb.1 (%ir-block.0):
   ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; UNPACKED-NEXT: {{  $}}
   ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; UNPACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; UNPACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; UNPACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(s32) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (s16) from custom "ImageResource")
-  ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[AMDGPU_INTRIN_IMAGE_LOAD]], [[C1]]
-  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[AMDGPU_INTRIN_IMAGE_LOAD]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C2]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
-  ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C1]], [[SHL]]
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_v3f16_dmask_1000
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s16>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (s16) from custom "ImageResource")
-  ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[DEF]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
   ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret <3 x half> %tex
 }
 
 define amdgpu_ps <3 x half> @image_load_v3f16_dmask_0000(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-LABEL: name: image_load_v3f16_dmask_0000
   ; UNPACKED: bb.1 (%ir-block.0):
   ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; UNPACKED-NEXT: {{  $}}
   ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
   ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
   ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
   ; UNPACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
   ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
   ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
   ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
   ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
   ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST3]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST4]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST3]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_v3f16_dmask_0000
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
   ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
   ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
   ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
   ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
   ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
   ; PACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-  ; PACKED-NEXT:   [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
   ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-  ; PACKED-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST3]](<2 x s16>)
-  ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST4]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
+  ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST3]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %tex = call <3 x half> @llvm.amdgcn.image.load.2d.v3f16.i32(i32 0, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret <3 x half> %tex
 }
 
 define amdgpu_ps <4 x half> @image_load_v4f16_dmask_1110(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-LABEL: name: image_load_v4f16_dmask_1110
   ; UNPACKED: bb.1 (%ir-block.0):
define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_1100(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<3 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 3, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (<2 x s16>) from custom "ImageResource")
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<3 x s32>)
   ; UNPACKED-NEXT:   G_STORE [[UV2]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; UNPACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
   ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-  ; UNPACKED-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C1]](s32)
   ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_tfe_v3f16_dmask_1100
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 3, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (<2 x s16>) from custom "ImageResource")
   ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
   ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[UV]](s32)
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; PACKED-NEXT:   [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[DEF1]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+  ; PACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
   ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST3]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %tex = extractvalue { <3 x half>, i32 } %res, 0
   %tfe = extractvalue { <3 x half>, i32 } %res, 1
   store i32 %tfe, i32 addrspace(1)* undef
   ret <3 x half> %tex
 }
define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_1000(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (s16) from custom "ImageResource")
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
   ; UNPACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; UNPACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
-  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C2]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
-  ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C1]], [[SHL]]
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_tfe_v3f16_dmask_1000
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (s16) from custom "ImageResource")
   ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
   ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[UV]](s32)
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; PACKED-NEXT:   [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[DEF1]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+  ; PACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
   ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST3]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %tex = extractvalue { <3 x half>, i32 } %res, 0
   %tfe = extractvalue { <3 x half>, i32 } %res, 1
   store i32 %tfe, i32 addrspace(1)* undef
   ret <3 x half> %tex
 }
define amdgpu_ps <3 x half> @image_load_tfe_v3f16_dmask_0000(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; UNPACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (s16) from custom "ImageResource")
   ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
   ; UNPACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; UNPACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
-  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; UNPACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C]]
+  ; UNPACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; UNPACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; UNPACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C2]](s32)
   ; UNPACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; UNPACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
-  ; UNPACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
-  ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
-  ; UNPACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
-  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; UNPACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C1]], [[SHL]]
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; UNPACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; UNPACKED-NEXT:   $vgpr1 = COPY [[BITCAST1]](<2 x s16>)
   ; UNPACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   ; PACKED-LABEL: name: image_load_tfe_v3f16_dmask_0000
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1
   ; PACKED-NEXT: {{  $}}
   ; PACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; PACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; PACKED-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; PACKED-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; PACKED-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; PACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; PACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; PACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; PACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; PACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; PACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; PACKED-NEXT:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; PACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; PACKED-NEXT:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:_(<2 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.2d), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 1, 0, 0 :: (dereferenceable load (s16) from custom "ImageResource")
   ; PACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_INTRIN_IMAGE_LOAD]](<2 x s32>)
   ; PACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[UV]](s32)
-  ; PACKED-NEXT:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
-  ; PACKED-NEXT:   [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[DEF1]](<2 x s16>)
-  ; PACKED-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
-  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[BITCAST]](<2 x s16>)
   ; PACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; PACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+  ; PACKED-NEXT:   G_STORE [[UV1]](s32), [[DEF]](p1) :: (store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; PACKED-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; PACKED-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+  ; PACKED-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; PACKED-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; PACKED-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST]](<2 x s16>)
+  ; PACKED-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; PACKED-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; PACKED-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+  ; PACKED-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
+  ; PACKED-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; PACKED-NEXT:   $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
   ; PACKED-NEXT:   $vgpr1 = COPY [[BITCAST3]](<2 x s16>)
   ; PACKED-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1
   %res = call { <3 x half>, i32 } @llvm.amdgcn.image.load.2d.sl_v3f16i32s.i32(i32 0, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
   %tex = extractvalue { <3 x half>, i32 } %res, 0
   %tfe = extractvalue { <3 x half>, i32 } %res, 1
   store i32 %tfe, i32 addrspace(1)* undef
   ret <3 x half> %tex
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.image.store.2d.d16.ll

 ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tonga -stop-after=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=UNPACKED %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx810 -stop-after=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX81 %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -stop-after=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9 %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -stop-after=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX10 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tonga -stop-after=legalizer -o - %s | FileCheck -check-prefix=UNPACKED %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx810 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX81 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX9 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -stop-after=legalizer -o - %s | FileCheck -check-prefix=GFX10 %s
 
 define amdgpu_ps void @image_store_f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, half %data) {
   ; UNPACKED-LABEL: name: image_store_f16
   ; UNPACKED: bb.1 (%ir-block.0):
   ; UNPACKED-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2
   ; UNPACKED-NEXT: {{  $}}
   ; UNPACKED-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; UNPACKED-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
define amdgpu_ps void @image_store_v3f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, <3 x half> %in) {
   ; UNPACKED-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; UNPACKED-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; UNPACKED-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; UNPACKED-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; UNPACKED-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; UNPACKED-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; UNPACKED-NEXT:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
   ; UNPACKED-NEXT:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-  ; UNPACKED-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; UNPACKED-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-  ; UNPACKED-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
+  ; UNPACKED-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
   ; UNPACKED-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; UNPACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+  ; UNPACKED-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY11]](<2 x s16>)
   ; UNPACKED-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
-  ; UNPACKED-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[LSHR]](s32), [[UV1]](s32)
+  ; UNPACKED-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
   ; UNPACKED-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BUILD_VECTOR2]](<3 x s32>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
   ; UNPACKED-NEXT:   S_ENDPGM 0
   ; GFX81-LABEL: name: image_store_v3f16
   ; GFX81: bb.1 (%ir-block.0):
   ; GFX81-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GFX81-NEXT: {{  $}}
   ; GFX81-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; GFX81-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; GFX81-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; GFX81-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; GFX81-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; GFX81-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; GFX81-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; GFX81-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; GFX81-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GFX81-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GFX81-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GFX81-NEXT:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
   ; GFX81-NEXT:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-  ; GFX81-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; GFX81-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; GFX81-NEXT:   [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-  ; GFX81-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
+  ; GFX81-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
   ; GFX81-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; GFX81-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+  ; GFX81-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+  ; GFX81-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY11]](<2 x s16>)
   ; GFX81-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
   ; GFX81-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; GFX81-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C1]]
+  ; GFX81-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
   ; GFX81-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
   ; GFX81-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
   ; GFX81-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; GFX81-NEXT:   [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; GFX81-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C1]]
+  ; GFX81-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; GFX81-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
   ; GFX81-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GFX81-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
   ; GFX81-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-  ; GFX81-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; GFX81-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
   ; GFX81-NEXT:   [[OR2:%[0-9]+]]:_(s32) = G_OR [[C2]], [[SHL1]]
-  ; GFX81-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-  ; GFX81-NEXT:   [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
-  ; GFX81-NEXT:   [[BITCAST4:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[CONCAT_VECTORS1]](<6 x s16>)
-  ; GFX81-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BITCAST4]](<3 x s32>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
+  ; GFX81-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+  ; GFX81-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
+  ; GFX81-NEXT:   [[BITCAST5:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
+  ; GFX81-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BITCAST5]](<3 x s32>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
   ; GFX81-NEXT:   S_ENDPGM 0
   ; GFX9-LABEL: name: image_store_v3f16
   ; GFX9: bb.1 (%ir-block.0):
   ; GFX9-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GFX9-NEXT: {{  $}}
   ; GFX9-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; GFX9-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; GFX9-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; GFX9-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; GFX9-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; GFX9-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; GFX9-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; GFX9-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; GFX9-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GFX9-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GFX9-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GFX9-NEXT:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
   ; GFX9-NEXT:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-  ; GFX9-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; GFX9-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; GFX9-NEXT:   [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-  ; GFX9-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
+  ; GFX9-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
   ; GFX9-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; GFX9-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
-  ; GFX9-NEXT:   [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV]](s32), [[LSHR]](s32)
-  ; GFX9-NEXT:   [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-  ; GFX9-NEXT:   [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV1]](s32), [[DEF1]](s32)
-  ; GFX9-NEXT:   [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; GFX9-NEXT:   [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<6 x s16>)
+  ; GFX9-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+  ; GFX9-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY11]](<2 x s16>)
   ; GFX9-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
-  ; GFX9-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[UV3]](<3 x s16>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
+  ; GFX9-NEXT:   [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+  ; GFX9-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+  ; GFX9-NEXT:   [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
+  ; GFX9-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+  ; GFX9-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[CONCAT_VECTORS]](<4 x s16>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
   ; GFX9-NEXT:   S_ENDPGM 0
   ; GFX10-LABEL: name: image_store_v3f16
   ; GFX10: bb.1 (%ir-block.0):
   ; GFX10-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GFX10-NEXT: {{  $}}
   ; GFX10-NEXT:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
   ; GFX10-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
   ; GFX10-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; GFX10-NEXT:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
   ; GFX10-NEXT:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
   ; GFX10-NEXT:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
   ; GFX10-NEXT:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
   ; GFX10-NEXT:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
   ; GFX10-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GFX10-NEXT:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GFX10-NEXT:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GFX10-NEXT:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
   ; GFX10-NEXT:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-  ; GFX10-NEXT:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
+  ; GFX10-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
   ; GFX10-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; GFX10-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
-  ; GFX10-NEXT:   [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV]](s32), [[LSHR]](s32)
-  ; GFX10-NEXT:   [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-  ; GFX10-NEXT:   [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV1]](s32), [[DEF1]](s32)
-  ; GFX10-NEXT:   [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF]](<2 x s16>)
-  ; GFX10-NEXT:   [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<6 x s16>)
+  ; GFX10-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+  ; GFX10-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY11]](<2 x s16>)
   ; GFX10-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
-  ; GFX10-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[UV3]](<3 x s16>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
+  ; GFX10-NEXT:   [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+  ; GFX10-NEXT:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+  ; GFX10-NEXT:   [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
+  ; GFX10-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+  ; GFX10-NEXT:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[CONCAT_VECTORS]](<4 x s16>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store (<3 x s16>) into custom "ImageResource", align 8)
   ; GFX10-NEXT:   S_ENDPGM 0
   call void @llvm.amdgcn.image.store.2d.v3f16.i32(<3 x half> %in, i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
 
 define amdgpu_ps void @image_store_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, <4 x half> %in) {
   ; UNPACKED-LABEL: name: image_store_v4f16
   ; UNPACKED: bb.1 (%ir-block.0):

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-llvm.amdgcn.s.buffer.load.mir

 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_v3s32
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>)
-    ; GCN-NEXT: S_ENDPGM 0, implicit [[UV]](<3 x s32>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
+    ; GCN-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; GCN-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<3 x s32>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     S_ENDPGM 0, implicit %2
 
 ...
 
 ---
 name: s_buffer_load_v3p3
 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_v3p3
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>)
-    ; GCN-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x p3>) = G_BITCAST [[UV]](<3 x s32>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
+    ; GCN-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; GCN-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x p3>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
     ; GCN-NEXT: S_ENDPGM 0, implicit [[BITCAST]](<3 x p3>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<3 x p3>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     S_ENDPGM 0, implicit %2
 
 ...
 
 ---
 name: s_buffer_load_v6s16
 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_v6s16
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>)
-    ; GCN-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[UV]](<3 x s32>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
+    ; GCN-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; GCN-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
     ; GCN-NEXT: S_ENDPGM 0, implicit [[BITCAST]](<6 x s16>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<6 x s16>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     S_ENDPGM 0, implicit %2
 
 ...
 
 ---
 name: s_buffer_load_v6s32
 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_v6s32
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<8 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s192), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<24 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<8 x s32>), [[DEF]](<8 x s32>), [[DEF]](<8 x s32>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<6 x s32>), [[UV1:%[0-9]+]]:_(<6 x s32>), [[UV2:%[0-9]+]]:_(<6 x s32>), [[UV3:%[0-9]+]]:_(<6 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<24 x s32>)
-    ; GCN-NEXT: S_ENDPGM 0, implicit [[UV]](<6 x s32>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<8 x s32>)
+    ; GCN-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<6 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
+    ; GCN-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<6 x s32>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<6 x s32>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     S_ENDPGM 0, implicit %2
 
 ...
 
 ---
 name: s_buffer_load_v3s64
 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_v3s64
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s64>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s192), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s64>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s64>), [[DEF]](<4 x s64>), [[DEF]](<4 x s64>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<3 x s64>), [[UV1:%[0-9]+]]:_(<3 x s64>), [[UV2:%[0-9]+]]:_(<3 x s64>), [[UV3:%[0-9]+]]:_(<3 x s64>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s64>)
-    ; GCN-NEXT: S_ENDPGM 0, implicit [[UV]](<3 x s64>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s64>)
+    ; GCN-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64)
+    ; GCN-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<3 x s64>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<3 x s64>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     S_ENDPGM 0, implicit %2
 
 ...
 
 ---
 name: s_buffer_load_v12s8
 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_v12s8
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>)
-    ; GCN-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](<3 x s32>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
     ; GCN-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
-    ; GCN-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C1]](s32)
+    ; GCN-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C1]](s32)
     ; GCN-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GCN-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
+    ; GCN-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C2]](s32)
     ; GCN-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 24
-    ; GCN-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C3]](s32)
-    ; GCN-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C1]](s32)
-    ; GCN-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
-    ; GCN-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C3]](s32)
-    ; GCN-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C1]](s32)
-    ; GCN-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
-    ; GCN-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C3]](s32)
+    ; GCN-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C3]](s32)
+    ; GCN-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C1]](s32)
+    ; GCN-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C2]](s32)
+    ; GCN-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C3]](s32)
+    ; GCN-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C1]](s32)
+    ; GCN-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C2]](s32)
+    ; GCN-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV2]], [[C3]](s32)
     ; GCN-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GCN-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C4]]
+    ; GCN-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[UV]], [[C4]]
     ; GCN-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C4]]
     ; GCN-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C2]](s32)
     ; GCN-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; GCN-NEXT: [[BITCAST:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GCN-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C4]]
     ; GCN-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C4]]
     ; GCN-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C2]](s32)
     ; GCN-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; GCN-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; GCN-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C4]]
+    ; GCN-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[UV1]], [[C4]]
     ; GCN-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
     ; GCN-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C2]](s32)
     ; GCN-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; GCN-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; GCN-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C4]]
     ; GCN-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C4]]
     ; GCN-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C2]](s32)
     ; GCN-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; GCN-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
-    ; GCN-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C4]]
+    ; GCN-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[UV2]], [[C4]]
     ; GCN-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C4]]
     ; GCN-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C2]](s32)
     ; GCN-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL4]]
     ; GCN-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
     ; GCN-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C4]]
     ; GCN-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C4]]
     ; GCN-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C2]](s32)
     ; GCN-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL5]]
     ; GCN-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
-    ; GCN-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>)
-    ; GCN-NEXT: S_ENDPGM 0, implicit [[CONCAT_VECTORS1]](<12 x s16>)
+    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[BITCAST]](<2 x s16>), [[BITCAST1]](<2 x s16>), [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>)
+    ; GCN-NEXT: S_ENDPGM 0, implicit [[CONCAT_VECTORS]](<12 x s16>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<12 x s8>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     %3:_(<12 x s16>) = G_ANYEXT %2
     S_ENDPGM 0, implicit %3
 
 ...
 
 ---
 name: s_buffer_load_s96
 body:             |
   bb.0:
     liveins: $sgpr0_sgpr1_sgpr2_sgpr3
 
     ; GCN-LABEL: name: s_buffer_load_s96
     ; GCN: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     ; GCN-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GCN-NEXT: [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[COPY]](<4 x s32>), [[C]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-    ; GCN-NEXT: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
-    ; GCN-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-    ; GCN-NEXT: [[UV:%[0-9]+]]:_(<3 x s32>), [[UV1:%[0-9]+]]:_(<3 x s32>), [[UV2:%[0-9]+]]:_(<3 x s32>), [[UV3:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s32>)
-    ; GCN-NEXT: S_ENDPGM 0, implicit [[UV]](<3 x s32>)
+    ; GCN-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
+    ; GCN-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; GCN-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<3 x s32>) = G_INTRINSIC intrinsic(@llvm.amdgcn.s.buffer.load), %0, %1, 0
     S_ENDPGM 0, implicit %2
 
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-constant.mir

     ; CI-LABEL: name: test_load_constant_s224_align4
     ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 4)
     ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; CI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; CI-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; CI-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; CI-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; CI-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; VI-LABEL: name: test_load_constant_s224_align4
     ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 4)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; VI-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; VI-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; GFX9-LABEL: name: test_load_constant_s224_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p4) :: (load (<4 x s32>), align 4, addrspace 4)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p4) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 4)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; GFX9-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
      %0:_(p4) = COPY $vgpr0_vgpr1
      %1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 4)
      %2:_(s256) = G_IMPLICIT_DEF
      %3:_(s256) = G_INSERT %2, %1, 0
      $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 
 ...

 name: test_load_constant_v3s16_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CI-LABEL: name: test_load_constant_v3s16_align8
     ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
-    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; VI-LABEL: name: test_load_constant_v3s16_align8
     ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_load_constant_v3s16_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p4) :: (load (<4 x s16>), addrspace 4)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)

 name: test_load_constant_v3s64_align32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CI-LABEL: name: test_load_constant_v3s64_align32
     ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_constant_v3s64_align32
     ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_constant_v3s64_align32
     ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p4) :: (load (<4 x s64>), addrspace 4)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p4) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 4)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---
 name: test_load_constant_v3s64_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CI-LABEL: name: test_load_constant_v3s64_align8
     ; CI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (load (s64) from unknown-address + 16, addrspace 4)
     ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_constant_v3s64_align8
     ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (load (s64) from unknown-address + 16, addrspace 4)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_constant_v3s64_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p4) :: (load (<2 x s64>), align 8, addrspace 4)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (load (s64) from unknown-address + 16, addrspace 4)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p4) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 8, addrspace 4)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

     ; CI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; CI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; CI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; CI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; CI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; CI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; CI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_constant_v3s64_align1
     ; VI: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; VI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; VI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; VI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; VI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_constant_v3s64_align1
     ; GFX9: [[COPY:%[0-9]+]]:_(p4) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p4) :: (load (s8), addrspace 4)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p4) :: (load (s8) from unknown-address + 1, addrspace 4)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; GFX9-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; GFX9-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; GFX9-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; GFX9-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; GFX9-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; GFX9-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p4) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 1, addrspace 4)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-flat.mir

     ; CI-LABEL: name: test_load_flat_s224_align4
     ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
     ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; CI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; CI-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; CI-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; CI-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; CI-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; CI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; VI-LABEL: name: test_load_flat_s224_align4
     ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; VI-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; VI-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; GFX9-LABEL: name: test_load_flat_s224_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p0) :: (load (<4 x s32>), align 4)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p0) :: (load (<3 x s32>) from unknown-address + 16, align 4)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; GFX9-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     %0:_(p0) = COPY $vgpr0_vgpr1
     %1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 0)
      %2:_(s256) = G_IMPLICIT_DEF
      %3:_(s256) = G_INSERT %2, %1, 0
      $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 
 ...

 name: test_load_flat_v3s16_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CI-LABEL: name: test_load_flat_v3s16_align8
     ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
-    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; VI-LABEL: name: test_load_flat_v3s16_align8
     ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_load_flat_v3s16_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p0) :: (load (<4 x s16>))
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)

 
     ; CI-LABEL: name: test_load_flat_v3s64_align32
     ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
     ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_flat_v3s64_align32
     ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_flat_v3s64_align32
     ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 32)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16, align 16)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p0) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 0)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---
 name: test_load_flat_v3s64_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CI-LABEL: name: test_load_flat_v3s64_align8
     ; CI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
     ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_flat_v3s64_align8
     ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_flat_v3s64_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p0) :: (load (<2 x s64>), align 8)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p0) :: (load (s64) from unknown-address + 16)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p0) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 8, addrspace 0)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

     ; CI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; CI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; CI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; CI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; CI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; CI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; CI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_flat_v3s64_align1
     ; VI: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; VI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; VI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; VI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; VI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_flat_v3s64_align1
     ; GFX9: [[COPY:%[0-9]+]]:_(p0) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p0) :: (load (s8))
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p0) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p0) :: (load (s8) from unknown-address + 1)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; GFX9-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; GFX9-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; GFX9-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; GFX9-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; GFX9-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; GFX9-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p0) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 1, addrspace 0)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-global.mir

 name: test_load_global_s96_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_s96_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
-    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[EXTRACT]](<3 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
     ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
     ; CI-HSA-LABEL: name: test_load_global_s96_align16
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)
     ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](s96)
     ; CI-MESA-LABEL: name: test_load_global_s96_align16
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; SI-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<2 x s32>) from unknown-address + 16, align 4, addrspace 1)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
     ; SI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s32) from unknown-address + 24, addrspace 1)
     ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[UV4]](s32)
-    ; SI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV]](s32), [[UV1]](s32)
-    ; SI-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[LOAD2]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; SI-NEXT: [[UV6:%[0-9]+]]:_(<7 x s32>), [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV6]](<7 x s32>)
-    ; SI-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
+    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; CI-HSA-LABEL: name: test_load_global_s224_align4
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; CI-HSA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; CI-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; CI-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; CI-HSA-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; CI-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; CI-HSA-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; CI-HSA-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; CI-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; CI-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; CI-MESA-LABEL: name: test_load_global_s224_align4
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; CI-MESA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; CI-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; CI-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; CI-MESA-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; CI-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; CI-MESA-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; CI-MESA-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; CI-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; CI-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; VI-LABEL: name: test_load_global_s224_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; VI-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; VI-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; GFX9-HSA-LABEL: name: test_load_global_s224_align4
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
     ; GFX9-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
     ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; GFX9-HSA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; GFX9-HSA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; GFX9-HSA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; GFX9-HSA-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; GFX9-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; GFX9-HSA-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; GFX9-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; GFX9-HSA-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; GFX9-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; GFX9-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; GFX9-HSA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     ; GFX9-MESA-LABEL: name: test_load_global_s224_align4
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), align 4, addrspace 1)
     ; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<3 x s32>) from unknown-address + 16, align 4, addrspace 1)
     ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
     ; GFX9-MESA-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<3 x s32>)
-    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
-    ; GFX9-MESA-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32), [[UV5]](s32)
-    ; GFX9-MESA-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[DEF]](s32), [[DEF]](s32)
-    ; GFX9-MESA-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s32>) = G_IMPLICIT_DEF
-    ; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<21 x s32>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<3 x s32>), [[BUILD_VECTOR1]](<3 x s32>), [[BUILD_VECTOR2]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>), [[DEF1]](<3 x s32>)
-    ; GFX9-MESA-NEXT: [[UV7:%[0-9]+]]:_(<7 x s32>), [[UV8:%[0-9]+]]:_(<7 x s32>), [[UV9:%[0-9]+]]:_(<7 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<21 x s32>)
-    ; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[UV7]](<7 x s32>)
-    ; GFX9-MESA-NEXT: [[DEF2:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
-    ; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF2]], [[BITCAST]](s224), 0
+    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<7 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32), [[UV4]](s32), [[UV5]](s32), [[UV6]](s32)
+    ; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s224) = G_BITCAST [[BUILD_VECTOR]](<7 x s32>)
+    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(s256) = G_IMPLICIT_DEF
+    ; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(s256) = G_INSERT [[DEF]], [[BITCAST]](s224), 0
     ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](s256)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(s224) = G_LOAD %0 :: (load (s224), align 4, addrspace 1)
      %2:_(s256) = G_IMPLICIT_DEF
      %3:_(s256) = G_INSERT %2, %1, 0
      $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 
 ...

 name: test_load_global_v3s16_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v3s16_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v3s16_align8
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
-    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CI-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v3s16_align8
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
-    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; VI-LABEL: name: test_load_global_v3s16_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v3s16_align8
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
-    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX9-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-HSA-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-HSA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-MESA-LABEL: name: test_load_global_v3s16_align8
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p1) :: (load (<4 x s16>), addrspace 1)
-    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-MESA-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-MESA-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v5s16_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; SI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v5s16_align16
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CI-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; CI-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; CI-HSA-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v5s16_align16
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CI-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; CI-MESA-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v5s16_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; VI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST5]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST6]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST7]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align16
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; GFX9-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; GFX9-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; GFX9-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; GFX9-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[LSHR]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[BITCAST4]](s32)
     ; GFX9-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
     ; GFX9-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
     ; GFX9-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-MESA-LABEL: name: test_load_global_v5s16_align16
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; GFX9-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; GFX9-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; GFX9-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[LSHR]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[BITCAST4]](s32)
     ; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
     ; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
     ; GFX9-MESA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>)

     ; CI-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; CI-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v5s16_align8
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; CI-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; CI-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; CI-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
     ; CI-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; CI-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; CI-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v5s16_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 8, addrspace 1)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align8
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 8, addrspace 1)

     ; CI-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; CI-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v5s16_align4
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; CI-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; CI-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; CI-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
     ; CI-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; CI-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; CI-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v5s16_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 4, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 4, addrspace 1)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align4
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 4, addrspace 1)

     ; SI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; SI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; SI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; SI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; SI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v5s16_align2
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; CI-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; CI-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; CI-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; CI-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
     ; CI-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; CI-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; CI-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v5s16_align2
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; CI-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; CI-MESA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; CI-MESA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
     ; CI-MESA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; CI-MESA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-MESA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; CI-MESA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-MESA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v5s16_align2
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, addrspace 1)
     ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, addrspace 1)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; VI-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; VI-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align2
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), addrspace 1)

     ; SI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
     ; SI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
     ; SI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
     ; SI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C7]]
-    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C7]]
-    ; SI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C6]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C6]]
+    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; SI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL5]]
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
-    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C7]]
-    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C7]]
-    ; SI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C6]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C6]]
+    ; SI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; SI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL6]]
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
-    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C7]]
-    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; SI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C6]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; SI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; SI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL7]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v5s16_align1
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; CI-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s16) from unknown-address + 2, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; CI-HSA-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; CI-HSA-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s16) from unknown-address + 4, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 6
     ; CI-HSA-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; CI-HSA-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p1) :: (load (s16) from unknown-address + 6, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; CI-HSA-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
     ; CI-HSA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p1) :: (load (s16) from unknown-address + 8, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C5]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C4]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C4]]
+    ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C5]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C4]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C5]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C5]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C4]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C4]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C4]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C5]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v5s16_align1
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)

     ; CI-MESA-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
     ; CI-MESA-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
     ; CI-MESA-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
     ; CI-MESA-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C7]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C6]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C6]]
+    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL5]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C7]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C6]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL6]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C7]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C6]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL7]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v5s16_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)

     ; VI-NEXT: [[ZEXTLOAD4:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD7]](p1) :: (load (s8) from unknown-address + 8, addrspace 1)
     ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C]](s64)
     ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD8]](p1) :: (load (s8) from unknown-address + 9, addrspace 1)
     ; VI-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[LOAD4]], [[C1]](s32)
     ; VI-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[SHL4]], [[ZEXTLOAD4]]
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C7]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C7]]
-    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C6]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C6]]
+    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; VI-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL5]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C7]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C7]]
-    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C6]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C6]]
+    ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL6]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C7]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C6]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL7]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v5s16_align1
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p1) :: (load (s16), align 1, addrspace 1)

 name: test_load_global_v6s16_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v6s16_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
-    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[EXTRACT]](<3 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
     ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v6s16_align16
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s16>) = G_BITCAST [[LOAD]](<3 x s32>)
     ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BITCAST]](<6 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v6s16_align16
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v7s16_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; SI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; SI-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
     ; SI-NEXT: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v7s16_align16
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CI-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; CI-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; CI-HSA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v7s16_align16
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CI-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; CI-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v7s16_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)

     ; VI-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: $vgpr3 = COPY [[BITCAST9]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align16
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; GFX9-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; GFX9-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-HSA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; GFX9-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; GFX9-HSA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; GFX9-HSA-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[LSHR]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR2]](s32)
     ; GFX9-HSA-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[BITCAST5]](s32)
     ; GFX9-HSA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
     ; GFX9-HSA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
     ; GFX9-HSA-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-HSA-NEXT: $vgpr3 = COPY [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
     ; GFX9-MESA-LABEL: name: test_load_global_v7s16_align16
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
     ; GFX9-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(<8 x s16>) = G_BITCAST [[LOAD]](<4 x s32>)
-    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[BITCAST]](<8 x s16>)
     ; GFX9-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-MESA-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; GFX9-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; GFX9-MESA-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>), [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; GFX9-MESA-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[LSHR]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR2]](s32)
     ; GFX9-MESA-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[BITCAST5]](s32)
     ; GFX9-MESA-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC]](<2 x s16>)
     ; GFX9-MESA-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC1]](<2 x s16>)

     ; SI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; SI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; SI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; SI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v7s16_align8
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v7s16_align8
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v7s16_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; VI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align8
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; SI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; SI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; SI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; SI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v7s16_align4
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v7s16_align4
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v7s16_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; VI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 4, addrspace 1)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align4
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; SI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; SI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; SI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; SI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; SI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; SI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v7s16_align2
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v7s16_align2
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-MESA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-MESA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; CI-MESA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-MESA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-MESA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-MESA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-MESA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v7s16_align2
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; VI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, addrspace 1)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; VI-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; VI-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align2
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; SI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
     ; SI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
     ; SI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
     ; SI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
     ; SI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; SI-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C9]]
-    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C9]]
-    ; SI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C8]](s32)
+    ; SI-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C8]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C8]]
+    ; SI-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C9]](s32)
     ; SI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL7]]
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
-    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C9]]
-    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C9]]
-    ; SI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C8]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C8]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C8]]
+    ; SI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C9]](s32)
     ; SI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL8]]
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
-    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C9]]
-    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[OR5]], [[C9]]
-    ; SI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C8]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C8]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[OR5]], [[C8]]
+    ; SI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C9]](s32)
     ; SI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL9]]
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR9]](s32)
-    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[OR6]], [[C9]]
-    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C9]]
-    ; SI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C8]](s32)
+    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[OR6]], [[C8]]
+    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C8]]
+    ; SI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C9]](s32)
     ; SI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL10]]
     ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
     ; SI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; SI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; SI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; SI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-HSA-LABEL: name: test_load_global_v7s16_align1
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-HSA-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; CI-HSA-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p1) :: (load (s16) from unknown-address + 10, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; CI-HSA-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
     ; CI-HSA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD5]](p1) :: (load (s16) from unknown-address + 12, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C7]]
-    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-HSA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LOAD]], [[C6]]
+    ; CI-HSA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LOAD1]], [[C6]]
+    ; CI-HSA-NEXT: [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-HSA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; CI-HSA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C7]]
-    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LOAD2]], [[C6]]
+    ; CI-HSA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LOAD3]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; CI-HSA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C7]]
-    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LOAD4]], [[C6]]
+    ; CI-HSA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LOAD5]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-HSA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C7]]
-    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C7]]
-    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+    ; CI-HSA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LOAD6]], [[C6]]
+    ; CI-HSA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C6]]
+    ; CI-HSA-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C7]](s32)
     ; CI-HSA-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
     ; CI-HSA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
     ; CI-HSA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-HSA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; CI-MESA-LABEL: name: test_load_global_v7s16_align1
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; CI-MESA-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
     ; CI-MESA-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
     ; CI-MESA-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
     ; CI-MESA-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; CI-MESA-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; CI-MESA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-MESA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C9]]
-    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C9]]
-    ; CI-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C8]](s32)
+    ; CI-MESA-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-MESA-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C8]]
+    ; CI-MESA-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C8]]
+    ; CI-MESA-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-MESA-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C9]](s32)
     ; CI-MESA-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL7]]
     ; CI-MESA-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
-    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C9]]
-    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C9]]
-    ; CI-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C8]](s32)
+    ; CI-MESA-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C8]]
+    ; CI-MESA-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C8]]
+    ; CI-MESA-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C9]](s32)
     ; CI-MESA-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL8]]
     ; CI-MESA-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
-    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C9]]
-    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[OR5]], [[C9]]
-    ; CI-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C8]](s32)
+    ; CI-MESA-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C8]]
+    ; CI-MESA-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[OR5]], [[C8]]
+    ; CI-MESA-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C9]](s32)
     ; CI-MESA-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL9]]
     ; CI-MESA-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR9]](s32)
-    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[OR6]], [[C9]]
-    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C9]]
-    ; CI-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C8]](s32)
+    ; CI-MESA-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[OR6]], [[C8]]
+    ; CI-MESA-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C8]]
+    ; CI-MESA-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C9]](s32)
     ; CI-MESA-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL10]]
     ; CI-MESA-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
     ; CI-MESA-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; CI-MESA-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; VI-LABEL: name: test_load_global_v7s16_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

     ; VI-NEXT: [[ZEXTLOAD6:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD11]](p1) :: (load (s8) from unknown-address + 12, addrspace 1)
     ; VI-NEXT: [[PTR_ADD12:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD11]], [[C]](s64)
     ; VI-NEXT: [[LOAD6:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD12]](p1) :: (load (s8) from unknown-address + 13, addrspace 1)
     ; VI-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[LOAD6]], [[C1]](s32)
     ; VI-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[SHL6]], [[ZEXTLOAD6]]
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<8 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
-    ; VI-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C9]]
-    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C9]]
-    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C8]](s32)
+    ; VI-NEXT: [[C8:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[OR]], [[C8]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[OR1]], [[C8]]
+    ; VI-NEXT: [[C9:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C9]](s32)
     ; VI-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL7]]
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
-    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C9]]
-    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C9]]
-    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C8]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[OR2]], [[C8]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[OR3]], [[C8]]
+    ; VI-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C9]](s32)
     ; VI-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL8]]
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
-    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C9]]
-    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[OR5]], [[C9]]
-    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C8]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[OR4]], [[C8]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[OR5]], [[C8]]
+    ; VI-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C9]](s32)
     ; VI-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL9]]
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR9]](s32)
-    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[OR6]], [[C9]]
-    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C9]]
-    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C8]](s32)
+    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[OR6]], [[C8]]
+    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C8]]
+    ; VI-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C9]](s32)
     ; VI-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL10]]
     ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
     ; VI-NEXT: $vgpr0 = COPY [[BITCAST1]](<2 x s16>)
     ; VI-NEXT: $vgpr1 = COPY [[BITCAST2]](<2 x s16>)
     ; VI-NEXT: $vgpr2 = COPY [[BITCAST3]](<2 x s16>)
     ; VI-NEXT: $vgpr3 = COPY [[BITCAST4]](<2 x s16>)
     ; GFX9-HSA-LABEL: name: test_load_global_v7s16_align1
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1

 name: test_load_global_v3s32_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v3s32_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[EXTRACT]](<3 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     ; CI-HSA-LABEL: name: test_load_global_v3s32_align16
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
     ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)
     ; CI-MESA-LABEL: name: test_load_global_v3s32_align16
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
     ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[LOAD]](<3 x s32>)

 name: test_load_global_v3s64_align32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v3s64_align32
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-HSA-LABEL: name: test_load_global_v3s64_align32
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
-    ; CI-HSA-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-HSA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-HSA-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-MESA-LABEL: name: test_load_global_v3s64_align32
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
-    ; CI-MESA-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-MESA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-MESA-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_global_v3s64_align32
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-HSA-LABEL: name: test_load_global_v3s64_align32
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
-    ; GFX9-HSA-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-HSA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-HSA-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-MESA-LABEL: name: test_load_global_v3s64_align32
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s64>) = G_LOAD [[COPY]](p1) :: (load (<4 x s64>), addrspace 1)
-    ; GFX9-MESA-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[LOAD]](<4 x s64>), 0
+    ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<4 x s64>)
     ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s64>), 0
-    ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-MESA-NEXT: [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64), [[UV6:%[0-9]+]]:_(s64), [[UV7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64), [[UV7]](s64)
+    ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 1)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---
 name: test_load_global_v3s64_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_load_global_v3s64_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; SI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
     ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-HSA-LABEL: name: test_load_global_v3s64_align8
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-HSA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-HSA-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-MESA-LABEL: name: test_load_global_v3s64_align8
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
     ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-MESA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-MESA-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_global_v3s64_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-HSA-LABEL: name: test_load_global_v3s64_align8
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
     ; GFX9-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
     ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-HSA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-HSA-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-MESA-LABEL: name: test_load_global_v3s64_align8
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 8, addrspace 1)
     ; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-MESA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, addrspace 1)
     ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-MESA-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 8, addrspace 1)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

     ; SI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; SI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; SI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; SI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; SI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; SI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; SI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-HSA-LABEL: name: test_load_global_v3s64_align1
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
     ; CI-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, align 1, addrspace 1)
     ; CI-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-HSA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-HSA-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-MESA-LABEL: name: test_load_global_v3s64_align1
     ; CI-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
     ; CI-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; CI-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CI-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
     ; CI-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CI-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; CI-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; CI-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; CI-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; CI-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; CI-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; CI-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; CI-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; CI-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; CI-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-MESA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; CI-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_global_v3s64_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; VI-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; VI-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; VI-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; VI-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; VI-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; VI-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; VI-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-HSA-LABEL: name: test_load_global_v3s64_align1
     ; GFX9-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p1) :: (load (<2 x s64>), align 1, addrspace 1)
     ; GFX9-HSA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-HSA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-HSA-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p1) :: (load (s64) from unknown-address + 16, align 1, addrspace 1)
     ; GFX9-HSA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-HSA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-HSA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-HSA-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-HSA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-HSA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-MESA-LABEL: name: test_load_global_v3s64_align1
     ; GFX9-MESA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-MESA-NEXT: [[ZEXTLOAD:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[COPY]](p1) :: (load (s8), addrspace 1)
     ; GFX9-MESA-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 1
     ; GFX9-MESA-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; GFX9-MESA-NEXT: [[ZEXTLOAD1:%[0-9]+]]:_(s32) = G_ZEXTLOAD [[PTR_ADD]](p1) :: (load (s8) from unknown-address + 1, addrspace 1)
     ; GFX9-MESA-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; GFX9-MESA-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXTLOAD1]], [[C1]](s32)

     ; GFX9-MESA-NEXT: [[SHL18:%[0-9]+]]:_(s32) = G_SHL [[LOAD5]], [[C1]](s32)
     ; GFX9-MESA-NEXT: [[OR18:%[0-9]+]]:_(s32) = G_OR [[SHL18]], [[ZEXTLOAD17]]
     ; GFX9-MESA-NEXT: [[SHL19:%[0-9]+]]:_(s32) = G_SHL [[OR18]], [[C3]](s32)
     ; GFX9-MESA-NEXT: [[OR19:%[0-9]+]]:_(s32) = G_OR [[SHL19]], [[OR17]]
     ; GFX9-MESA-NEXT: [[ANYEXT2:%[0-9]+]]:_(s64) = G_ANYEXT [[OR19]](s32)
     ; GFX9-MESA-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C5]](s32)
     ; GFX9-MESA-NEXT: [[SHL20:%[0-9]+]]:_(s64) = G_SHL [[ANYEXT2]], [[COPY2]](s32)
     ; GFX9-MESA-NEXT: [[OR20:%[0-9]+]]:_(s64) = G_OR [[SHL20]], [[ZEXT2]]
-    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64)
     ; GFX9-MESA-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-MESA-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-MESA-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-MESA-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[OR6]](s64), [[OR13]](s64), [[OR20]](s64), [[UV3]](s64)
+    ; GFX9-MESA-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 1, addrspace 1)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

 name: test_global_v2s96_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; SI-LABEL: name: test_global_v2s96_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[COPY]](p1) :: (load (<4 x s32>), addrspace 1)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
-    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[EXTRACT]](<3 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 12
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; SI-NEXT: [[LOAD1:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load (<2 x s32>) from unknown-address + 12, align 4, addrspace 1)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C1]](s64)
     ; SI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p1) :: (load (s32) from unknown-address + 20, addrspace 1)
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD2]](s32)
-    ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR]](<3 x s32>)
+    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<2 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[LOAD2]](s32)
+    ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[BUILD_VECTOR1]](<3 x s32>)
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY [[BITCAST]](s96)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s96) = COPY [[BITCAST1]](s96)
     ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[COPY1]](s96)
     ; SI-NEXT: $vgpr3_vgpr4_vgpr5 = COPY [[COPY2]](s96)
     ; CI-HSA-LABEL: name: test_global_v2s96_align16
     ; CI-HSA: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-HSA-NEXT: [[LOAD:%[0-9]+]]:_(<3 x s32>) = G_LOAD [[COPY]](p1) :: (load (<3 x s32>), align 16, addrspace 1)
     ; CI-HSA-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[LOAD]](<3 x s32>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-local.mir

 name: test_load_local_v3s16_align8
 body: |
   bb.0:
     liveins: $vgpr0
 
     ; SI-LABEL: name: test_load_local_v3s16_align8
     ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; SI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; CI-LABEL: name: test_load_local_v3s16_align8
     ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; CI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; CI-DS128-LABEL: name: test_load_local_v3s16_align8
     ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; CI-DS128-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; CI-DS128-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-DS128-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; CI-DS128-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CI-DS128-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-DS128-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CI-DS128-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-DS128-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; CI-DS128-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; CI-DS128-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; CI-DS128-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; CI-DS128-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; CI-DS128-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CI-DS128-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CI-DS128-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CI-DS128-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; CI-DS128-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; CI-DS128-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; CI-DS128-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; CI-DS128-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; CI-DS128-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; VI-LABEL: name: test_load_local_v3s16_align8
     ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)

     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
     ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_load_local_v3s16_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-UNALIGNED-LABEL: name: test_load_local_v3s16_align8
     ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX9-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; GFX9-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX9-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-UNALIGNED-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-UNALIGNED-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX9-UNALIGNED-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX9-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-UNALIGNED-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-UNALIGNED-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-UNALIGNED-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX9-UNALIGNED-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX9-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX10-LABEL: name: test_load_local_v3s16_align8
     ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX10-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX10-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX10-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX10-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX10-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX10-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX10-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX10-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX10-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX10-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX10-UNALIGNED-LABEL: name: test_load_local_v3s16_align8
     ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX10-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[COPY]](p3) :: (load (<4 x s16>), addrspace 3)
-    ; GFX10-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX10-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
     ; GFX10-UNALIGNED-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX10-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX10-UNALIGNED-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX10-UNALIGNED-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX10-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX10-UNALIGNED-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; GFX10-UNALIGNED-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; GFX10-UNALIGNED-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX10-UNALIGNED-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX10-UNALIGNED-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX10-UNALIGNED-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[BITCAST2]](s32)
     ; GFX10-UNALIGNED-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
     ; GFX10-UNALIGNED-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)

     ; SI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; SI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 32, addrspace 3)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; SI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 8, addrspace 3)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C1]](s32)
     ; SI-NEXT: [[LOAD2:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD1]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[LOAD]](s64), [[LOAD1]](s64), [[LOAD2]](s64)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[LOAD]](s64), [[LOAD1]](s64), [[LOAD2]](s64), [[UV3]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-LABEL: name: test_load_local_v3s64_align32
     ; CI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s64) = G_LOAD [[COPY]](p3) :: (load (s64), align 32, addrspace 3)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 8, addrspace 3)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C1]](s32)
     ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD1]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[LOAD]](s64), [[LOAD1]](s64), [[LOAD2]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[LOAD]](s64), [[LOAD1]](s64), [[LOAD2]](s64), [[UV3]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-DS128-LABEL: name: test_load_local_v3s64_align32
     ; CI-DS128: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; CI-DS128-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
     ; CI-DS128-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-DS128-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; CI-DS128-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
     ; CI-DS128-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; CI-DS128-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; CI-DS128-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-DS128-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-DS128-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-DS128-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-DS128-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; CI-DS128-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_local_v3s64_align32
     ; VI: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
     ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_local_v3s64_align32
     ; GFX9: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-UNALIGNED-LABEL: name: test_load_local_v3s64_align32
     ; GFX9-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX9-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
     ; GFX9-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; GFX9-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
     ; GFX9-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX9-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX9-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-UNALIGNED-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-UNALIGNED-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX9-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX10-LABEL: name: test_load_local_v3s64_align32
     ; GFX10: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX10-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
     ; GFX10-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX10-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; GFX10-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
     ; GFX10-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX10-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX10-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX10-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX10-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX10-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX10-UNALIGNED-LABEL: name: test_load_local_v3s64_align32
     ; GFX10-UNALIGNED: [[COPY:%[0-9]+]]:_(p3) = COPY $vgpr0
     ; GFX10-UNALIGNED-NEXT: [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[COPY]](p3) :: (load (<2 x s64>), align 32, addrspace 3)
     ; GFX10-UNALIGNED-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX10-UNALIGNED-NEXT: [[PTR_ADD:%[0-9]+]]:_(p3) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; GFX10-UNALIGNED-NEXT: [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p3) :: (load (s64) from unknown-address + 16, align 16, addrspace 3)
     ; GFX10-UNALIGNED-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[LOAD]](<2 x s64>)
-    ; GFX10-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64)
     ; GFX10-UNALIGNED-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX10-UNALIGNED-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX10-UNALIGNED-NEXT: [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX10-UNALIGNED-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[LOAD1]](s64), [[UV5]](s64)
+    ; GFX10-UNALIGNED-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p3) = COPY $vgpr0
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 3)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-load-private.mir

     ; SI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s32) from unknown-address + 12, addrspace 5)
     ; SI-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD2]](s32), [[LOAD3]](s32)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C2]](s32)
     ; SI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p5) :: (load (s32) from unknown-address + 16, align 16, addrspace 5)
     ; SI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
     ; SI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p5) :: (load (s32) from unknown-address + 20, addrspace 5)
     ; SI-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD4]](s32), [[LOAD5]](s32)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64), [[UV3]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; CI-LABEL: name: test_load_private_v3s64_align32
     ; CI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
     ; CI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), align 32, addrspace 5)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; CI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s32) from unknown-address + 4, addrspace 5)
     ; CI-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[LOAD1]](s32)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C1]](s32)
     ; CI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s32) from unknown-address + 8, align 8, addrspace 5)
     ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
     ; CI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s32) from unknown-address + 12, addrspace 5)
     ; CI-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD2]](s32), [[LOAD3]](s32)
     ; CI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C2]](s32)
     ; CI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p5) :: (load (s32) from unknown-address + 16, align 16, addrspace 5)
     ; CI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
     ; CI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p5) :: (load (s32) from unknown-address + 20, addrspace 5)
     ; CI-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD4]](s32), [[LOAD5]](s32)
-    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64)
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64), [[UV3]](s64)
+    ; CI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_load_private_v3s64_align32
     ; VI: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
     ; VI-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), align 32, addrspace 5)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; VI-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s32) from unknown-address + 4, addrspace 5)
     ; VI-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[LOAD1]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C1]](s32)
     ; VI-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s32) from unknown-address + 8, align 8, addrspace 5)
     ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
     ; VI-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s32) from unknown-address + 12, addrspace 5)
     ; VI-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD2]](s32), [[LOAD3]](s32)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C2]](s32)
     ; VI-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p5) :: (load (s32) from unknown-address + 16, align 16, addrspace 5)
     ; VI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
     ; VI-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p5) :: (load (s32) from unknown-address + 20, addrspace 5)
     ; VI-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD4]](s32), [[LOAD5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64), [[UV3]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_load_private_v3s64_align32
     ; GFX9: [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
     ; GFX9-NEXT: [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[COPY]](p5) :: (load (s32), align 32, addrspace 5)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C]](s32)
     ; GFX9-NEXT: [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p5) :: (load (s32) from unknown-address + 4, addrspace 5)
     ; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD]](s32), [[LOAD1]](s32)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C1]](s32)
     ; GFX9-NEXT: [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s32) from unknown-address + 8, align 8, addrspace 5)
     ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD1]], [[C]](s32)
     ; GFX9-NEXT: [[LOAD3:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD2]](p5) :: (load (s32) from unknown-address + 12, addrspace 5)
     ; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD2]](s32), [[LOAD3]](s32)
     ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C2]](s32)
     ; GFX9-NEXT: [[LOAD4:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p5) :: (load (s32) from unknown-address + 16, align 16, addrspace 5)
     ; GFX9-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p5) = G_PTR_ADD [[PTR_ADD3]], [[C]](s32)
     ; GFX9-NEXT: [[LOAD5:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD4]](p5) :: (load (s32) from unknown-address + 20, addrspace 5)
     ; GFX9-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[LOAD4]](s32), [[LOAD5]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64), [[UV3]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(p5) = COPY $vgpr0
     %1:_(<3 x s64>) = G_LOAD %0 :: (load (<3 x s64>), align 32, addrspace 5)
     %2:_(<4 x s64>) = G_IMPLICIT_DEF
     %3:_(<4 x s64>) = G_INSERT %2, %1, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-lshr.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=SI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 -o - %s | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=SI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -o - %s | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_lshr_s32_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_lshr_s32_s32

 ---
 name: test_lshr_v3s64_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_lshr_v3s64_v3s32
     ; SI: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[UV3]](s32)
-    ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s64) = G_LSHR [[UV1]], [[UV4]](s32)
-    ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s64) = G_LSHR [[UV2]], [[UV5]](s32)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[LSHR]](s64), [[LSHR1]](s64), [[LSHR2]](s64)
+    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[UV4]](s32)
+    ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s64) = G_LSHR [[UV1]], [[UV5]](s32)
+    ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s64) = G_LSHR [[UV2]], [[UV6]](s32)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[LSHR]](s64), [[LSHR1]](s64), [[LSHR2]](s64), [[UV10]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_lshr_v3s64_v3s32
     ; VI: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[UV3]](s32)
-    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s64) = G_LSHR [[UV1]], [[UV4]](s32)
-    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s64) = G_LSHR [[UV2]], [[UV5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[LSHR]](s64), [[LSHR1]](s64), [[LSHR2]](s64)
+    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[UV4]](s32)
+    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s64) = G_LSHR [[UV1]], [[UV5]](s32)
+    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s64) = G_LSHR [[UV2]], [[UV6]](s32)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[LSHR]](s64), [[LSHR1]](s64), [[LSHR2]](s64), [[UV10]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_lshr_v3s64_v3s32
     ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[UV3]](s32)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s64) = G_LSHR [[UV1]], [[UV4]](s32)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s64) = G_LSHR [[UV2]], [[UV5]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[LSHR]](s64), [[LSHR1]](s64), [[LSHR2]](s64)
+    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[UV]], [[UV4]](s32)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s64) = G_LSHR [[UV1]], [[UV5]](s32)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s64) = G_LSHR [[UV2]], [[UV6]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[LSHR]](s64), [[LSHR1]](s64), [[LSHR2]](s64), [[UV10]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:_(<3 x s64>) = G_EXTRACT %0, 0
     %2:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
     %3:_(<3 x s64>) = G_LSHR %1, %2
     %4:_(<4 x s64>) = G_IMPLICIT_DEF
     %5:_(<4 x s64>) = G_INSERT %4, %3, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %5
 ...

 name: test_lshr_v3s16_v3s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
 
     ; SI-LABEL: name: test_lshr_v3s16_v3s16
     ; SI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; SI-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
+    ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[AND1]], [[AND]](s32)
     ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[AND3]], [[AND2]](s32)
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; SI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[AND5]], [[AND4]](s32)
+    ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; SI-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; SI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
     ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
     ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
     ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL]]
-    ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; SI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; SI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
-    ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; SI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
     ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL1]]
-    ; SI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[DEF1]](<2 x s16>)
-    ; SI-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-    ; SI-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0
-    ; SI-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
+    ; SI-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     ; VI-LABEL: name: test_lshr_v3s16_v3s16
     ; VI: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; VI-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
+    ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC3]](s16)
     ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC1]], [[TRUNC4]](s16)
     ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC2]], [[TRUNC5]](s16)
+    ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; VI-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; VI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; VI-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[LSHR2]](s16)
     ; VI-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[LSHR3]](s16)
     ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; VI-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[LSHR4]](s16)
-    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
+    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[DEF1]](<2 x s16>)
-    ; VI-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-    ; VI-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV4]](<3 x s16>), 0
-    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[INSERT2]](<4 x s16>)
+    ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
+    ; VI-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     ; GFX9-LABEL: name: test_lshr_v3s16_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32
-    ; GFX9-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT4:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0
-    ; GFX9-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT5:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[EXTRACT2]], [[EXTRACT4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[EXTRACT3]], [[EXTRACT5]](s16)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR2]](s32)
-    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR1]](s16)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[DEF1]](s32)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF2]](<2 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-    ; GFX9-NEXT: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[INSERT4]](<4 x s16>)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC1]](s16)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR4]](s32)
+    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR3]](s16)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[LSHR5]](s32)
+    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<4 x s16>) = COPY $vgpr2_vgpr3
     %2:_(<3 x s16>) = G_EXTRACT %0, 0
     %3:_(<3 x s16>) = G_EXTRACT %1, 0
     %4:_(<3 x s16>) = G_LSHR %2, %3
     %5:_(<4 x s16>) = G_IMPLICIT_DEF
     %6:_(<4 x s16>) = G_INSERT %5, %4, 0
     $vgpr0_vgpr1 = COPY %6
bb.0:
     ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
     ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_ashr_v3s16_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32
-    ; GFX9-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0
-    ; GFX9-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[EXTRACT]], [[EXTRACT2]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[EXTRACT1]], [[EXTRACT3]](s16)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR2]](s32)
-    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR1]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST2]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[TRUNC1]](s16)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR4]](s32)
+    ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR3]](s16)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR5]](s32), [[BITCAST6]](s32)
+    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_LSHR %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-or.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -allow-ginsert-as-artifact=0 -global-isel-abort=0 %s -o - | FileCheck %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck %s
 
 ---
 name: test_or_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; CHECK-LABEL: name: test_or_s32

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_or_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s32>) = G_CONCAT_VECTORS [[OR]](<2 x s32>), [[OR1]](<2 x s32>), [[DEF1]](<2 x s32>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<3 x s32>), [[UV7:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s32>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[UV6]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR]], [[BUILD_VECTOR1]]
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[UV2]], [[UV5]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[OR]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[UV7]](s32), [[OR1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR2]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s32>) = G_OR %0, %1
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 
 ---
 name: test_or_v4s32

 name: test_or_v5s32
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_or_v5s32
     ; CHECK: [[DEF:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<5 x s32>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
     ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[DEF2]](s32)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](<5 x s32>)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV6]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV7]](s32), [[UV8]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV9]](s32), [[DEF2]](s32)
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR1]], [[BUILD_VECTOR4]]
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR2]], [[BUILD_VECTOR5]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s32>) = G_CONCAT_VECTORS [[OR]](<2 x s32>), [[OR1]](<2 x s32>), [[OR2]](<2 x s32>), [[DEF3]](<2 x s32>), [[DEF3]](<2 x s32>)
-    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(<5 x s32>), [[UV11:%[0-9]+]]:_(<5 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s32>)
-    ; CHECK-NEXT: [[DEF4:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<8 x s32>) = G_INSERT [[DEF4]], [[UV10]](<5 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<8 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV6]](s32)
+    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV7]](s32), [[UV8]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(<2 x s32>) = G_OR [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[UV4]], [[UV9]]
+    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[OR]](<2 x s32>)
+    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[OR1]](<2 x s32>)
+    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF2]](<8 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[OR2]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR4]](<8 x s32>)
     %0:_(<5 x s32>) = G_IMPLICIT_DEF
     %1:_(<5 x s32>) = G_IMPLICIT_DEF
     %2:_(<5 x s32>) = G_OR %0, %1
     %3:_(<8 x s32>) = G_IMPLICIT_DEF
     %4:_(<8 x s32>) = G_INSERT %3, %2, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %4
 ...
 

 ---
 name: test_or_v3s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
     ; CHECK-LABEL: name: test_or_v3s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT]], [[INSERT1]]
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[OR]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
-    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
-    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(<4 x s16>) = G_OR [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[OR4]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST8]], [[C1]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST9]], [[C1]]
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C1]]
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C1]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C]](s32)
+    ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_OR %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7
     $vgpr0_vgpr1_vgpr2 = COPY %8

 
 ---
 name: test_or_v5s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_or_v5s16
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>), [[UV11:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV6]](<5 x s16>), 0
-    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>)
-    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV12]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV14]](<3 x s16>), 0
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT2]], [[INSERT3]]
-    ; CHECK-NEXT: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV13]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV15]](<3 x s16>), 0
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(<4 x s16>) = G_OR [[INSERT4]], [[INSERT5]]
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[OR]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>)
-    ; CHECK-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[OR1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
-    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
-    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
-    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(<5 x s16>), [[UV21:%[0-9]+]]:_(<5 x s16>), [[UV22:%[0-9]+]]:_(<5 x s16>), [[UV23:%[0-9]+]]:_(<5 x s16>), [[UV24:%[0-9]+]]:_(<5 x s16>), [[UV25:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF4:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[UV20]](<5 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT6]](<8 x s16>)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C]](s32)
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(<4 x s16>) = G_OR [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[OR4]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[COPY1]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
+    ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[COPY3]], [[SHL7]]
+    ; CHECK-NEXT: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST14]](<2 x s16>), [[BITCAST15]](<2 x s16>)
+    ; CHECK-NEXT: [[OR9:%[0-9]+]]:_(<4 x s16>) = G_OR [[CONCAT_VECTORS2]], [[CONCAT_VECTORS3]]
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[OR9]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST16:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST16]], [[C]](s32)
+    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<8 x s16>)
+    ; CHECK-NEXT: [[BITCAST17:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST17]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST18:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST18]], [[C]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C1]]
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
+    ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C]](s32)
+    ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL8]]
+    ; CHECK-NEXT: [[BITCAST19:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C1]]
+    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[BITCAST16]], [[C1]]
+    ; CHECK-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C]](s32)
+    ; CHECK-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
+    ; CHECK-NEXT: [[BITCAST20:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR11]](s32)
+    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
+    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C1]]
+    ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C]](s32)
+    ; CHECK-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND14]], [[SHL10]]
+    ; CHECK-NEXT: [[BITCAST21:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR12]](s32)
+    ; CHECK-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[BITCAST18]], [[C1]]
+    ; CHECK-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C1]]
+    ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C]](s32)
+    ; CHECK-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL11]]
+    ; CHECK-NEXT: [[BITCAST22:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR13]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST19]](<2 x s16>), [[BITCAST20]](<2 x s16>), [[BITCAST21]](<2 x s16>), [[BITCAST22]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<8 x s16>)
     %0:_(<5 x s16>) = G_IMPLICIT_DEF
     %1:_(<5 x s16>) = G_IMPLICIT_DEF
     %2:_(<5 x s16>) = G_OR %0, %1
     %3:_(<8 x s16>) = G_IMPLICIT_DEF
     %4:_(<8 x s16>) = G_INSERT %3, %2, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %4
 ...
 
 ---
 name: test_or_v3s8
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_or_v3s8
-    ; CHECK: [[DEF:%[0-9]+]]:_(<3 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF2]], [[DEF]](<3 x s8>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF2]], [[DEF1]](<3 x s8>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8), [[UV3:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT]](<4 x s8>)
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT1]](<4 x s8>)
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s8)
-    ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV4]](s8)
-    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ANYEXT]], [[ANYEXT1]]
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[OR]](s32)
-    ; CHECK-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV1]](s8)
-    ; CHECK-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[UV5]](s8)
-    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ANYEXT2]], [[ANYEXT3]]
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[OR1]](s32)
-    ; CHECK-NEXT: [[ANYEXT4:%[0-9]+]]:_(s32) = G_ANYEXT [[UV2]](s8)
-    ; CHECK-NEXT: [[ANYEXT5:%[0-9]+]]:_(s32) = G_ANYEXT [[UV6]](s8)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ANYEXT4]], [[ANYEXT5]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[OR2]](s32)
-    ; CHECK-NEXT: [[ANYEXT6:%[0-9]+]]:_(s32) = G_ANYEXT [[UV3]](s8)
-    ; CHECK-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[UV7]](s8)
-    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[ANYEXT6]], [[ANYEXT7]]
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[OR3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC]](s8), [[TRUNC1]](s8), [[TRUNC2]](s8), [[TRUNC3]](s8)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF2]](<4 x s8>), [[DEF2]](<4 x s8>)
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>)
-    ; CHECK-NEXT: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>)
+    ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](<4 x s32>)
+    ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[UV]], [[UV4]]
+    ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[UV1]], [[UV5]]
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[UV2]], [[UV6]]
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[OR]](s32), [[OR1]](s32), [[OR2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s8>) = G_IMPLICIT_DEF
     %1:_(<3 x s8>) = G_IMPLICIT_DEF
     %2:_(<3 x s8>) = G_OR %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     $vgpr0_vgpr1_vgpr2 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-phi.mir

body: |
   ; CHECK: bb.0:
   ; CHECK-NEXT:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
   ; CHECK-NEXT:   liveins: $vgpr0_vgpr1, $vgpr2
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; CHECK-NEXT:   [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[COPY1]](s32), [[C]]
-  ; CHECK-NEXT:   [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-  ; CHECK-NEXT:   G_BRCOND [[ICMP]](s1), %bb.1
-  ; CHECK-NEXT:   G_BR %bb.2
-  ; CHECK-NEXT: {{  $}}
-  ; CHECK-NEXT: bb.1:
-  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
-  ; CHECK-NEXT: {{  $}}
-  ; CHECK-NEXT:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
   ; CHECK-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
   ; CHECK-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32)
   ; CHECK-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-  ; CHECK-NEXT:   [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[EXTRACT]](<3 x s16>), 0
-  ; CHECK-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>)
-  ; CHECK-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-  ; CHECK-NEXT:   [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C1]](s32)
-  ; CHECK-NEXT:   [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
-  ; CHECK-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[BITCAST]], [[BITCAST2]]
-  ; CHECK-NEXT:   [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[LSHR]], [[LSHR1]]
-  ; CHECK-NEXT:   [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[BITCAST1]], [[BITCAST3]]
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; CHECK-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[ADD]], [[C2]]
-  ; CHECK-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[ADD1]], [[C2]]
+  ; CHECK-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C2]]
+  ; CHECK-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C2]]
   ; CHECK-NEXT:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-  ; CHECK-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
-  ; CHECK-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[ADD2]], [[C2]]
-  ; CHECK-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
-  ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C3]], [[C1]](s32)
+  ; CHECK-NEXT:   [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+  ; CHECK-NEXT:   [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C2]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:_(s32) = COPY [[C]](s32)
+  ; CHECK-NEXT:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C1]](s32)
   ; CHECK-NEXT:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-  ; CHECK-NEXT:   [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-  ; CHECK-NEXT:   [[DEF2:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[DEF2]](<2 x s16>)
-  ; CHECK-NEXT:   [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
-  ; CHECK-NEXT:   [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0
+  ; CHECK-NEXT:   [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+  ; CHECK-NEXT:   G_BRCOND [[ICMP]](s1), %bb.1
   ; CHECK-NEXT:   G_BR %bb.2
   ; CHECK-NEXT: {{  $}}
-  ; CHECK-NEXT: bb.2:
-  ; CHECK-NEXT:   [[PHI:%[0-9]+]]:_(<4 x s16>) = G_PHI [[INSERT]](<4 x s16>), %bb.0, [[INSERT3]](<4 x s16>), %bb.1
-  ; CHECK-NEXT:   [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[PHI]](<4 x s16>)
-  ; CHECK-NEXT:   [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[BITCAST]], [[BITCAST]]
+  ; CHECK-NEXT:   [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[LSHR]], [[LSHR]]
+  ; CHECK-NEXT:   [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[BITCAST1]], [[BITCAST1]]
+  ; CHECK-NEXT:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; CHECK-NEXT:   [[AND3:%[0-9]+]]:_(s32) = G_AND [[ADD]], [[C3]]
+  ; CHECK-NEXT:   [[AND4:%[0-9]+]]:_(s32) = G_AND [[ADD1]], [[C3]]
   ; CHECK-NEXT:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-  ; CHECK-NEXT:   [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C4]](s32)
-  ; CHECK-NEXT:   [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
-  ; CHECK-NEXT:   [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF3]](<4 x s16>)
-  ; CHECK-NEXT:   [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
-  ; CHECK-NEXT:   [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C4]](s32)
-  ; CHECK-NEXT:   [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
-  ; CHECK-NEXT:   [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-  ; CHECK-NEXT:   [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST6]], [[C5]]
-  ; CHECK-NEXT:   [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C5]]
   ; CHECK-NEXT:   [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C4]](s32)
   ; CHECK-NEXT:   [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
-  ; CHECK-NEXT:   [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-  ; CHECK-NEXT:   [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST7]], [[C5]]
-  ; CHECK-NEXT:   [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST8]], [[C5]]
-  ; CHECK-NEXT:   [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C4]](s32)
+  ; CHECK-NEXT:   [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+  ; CHECK-NEXT:   [[AND5:%[0-9]+]]:_(s32) = G_AND [[ADD2]], [[C3]]
+  ; CHECK-NEXT:   [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+  ; CHECK-NEXT:   [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C5]], [[C4]](s32)
   ; CHECK-NEXT:   [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
-  ; CHECK-NEXT:   [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
-  ; CHECK-NEXT:   [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C5]]
-  ; CHECK-NEXT:   [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST9]], [[C5]]
-  ; CHECK-NEXT:   [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C4]](s32)
-  ; CHECK-NEXT:   [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL4]]
-  ; CHECK-NEXT:   [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
-  ; CHECK-NEXT:   [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>)
-  ; CHECK-NEXT:   $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS1]](<6 x s16>)
+  ; CHECK-NEXT:   [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>)
+  ; CHECK-NEXT:   G_BR %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   [[PHI:%[0-9]+]]:_(<4 x s16>) = G_PHI [[CONCAT_VECTORS]](<4 x s16>), %bb.0, [[CONCAT_VECTORS1]](<4 x s16>), %bb.1
+  ; CHECK-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[PHI]](<4 x s16>)
+  ; CHECK-NEXT:   [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+  ; CHECK-NEXT:   [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+  ; CHECK-NEXT:   [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C6]](s32)
+  ; CHECK-NEXT:   [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+  ; CHECK-NEXT:   [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+  ; CHECK-NEXT:   [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+  ; CHECK-NEXT:   [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C6]](s32)
+  ; CHECK-NEXT:   [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+  ; CHECK-NEXT:   [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+  ; CHECK-NEXT:   [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST6]], [[C7]]
+  ; CHECK-NEXT:   [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C7]]
+  ; CHECK-NEXT:   [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C6]](s32)
+  ; CHECK-NEXT:   [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL4]]
+  ; CHECK-NEXT:   [[BITCAST10:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+  ; CHECK-NEXT:   [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST7]], [[C7]]
+  ; CHECK-NEXT:   [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST8]], [[C7]]
+  ; CHECK-NEXT:   [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C6]](s32)
+  ; CHECK-NEXT:   [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL5]]
+  ; CHECK-NEXT:   [[BITCAST11:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+  ; CHECK-NEXT:   [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C7]]
+  ; CHECK-NEXT:   [[AND11:%[0-9]+]]:_(s32) = G_AND [[BITCAST9]], [[C7]]
+  ; CHECK-NEXT:   [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C6]](s32)
+  ; CHECK-NEXT:   [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL6]]
+  ; CHECK-NEXT:   [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+  ; CHECK-NEXT:   [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST10]](<2 x s16>), [[BITCAST11]](<2 x s16>), [[BITCAST12]](<2 x s16>)
+  ; CHECK-NEXT:   $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>)
   ; CHECK-NEXT:   S_SETPC_B64 undef $sgpr30_sgpr31
   bb.0:
     successors: %bb.1, %bb.2
     liveins: $vgpr0_vgpr1, $vgpr2
 
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(s32) = COPY $vgpr2
     %2:_(s32) = G_CONSTANT i32 0
body: |
   ; CHECK-NEXT:   successors: %bb.2(0x80000000)
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
   ; CHECK-NEXT:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
   ; CHECK-NEXT:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32)
   ; CHECK-NEXT:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
   ; CHECK-NEXT:   [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
-  ; CHECK-NEXT:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
-  ; CHECK-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+  ; CHECK-NEXT:   [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
   ; CHECK-NEXT:   [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C1]](s32)
-  ; CHECK-NEXT:   [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+  ; CHECK-NEXT:   [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
   ; CHECK-NEXT:   [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C1]](s32)
   ; CHECK-NEXT:   [[ADD:%[0-9]+]]:_(s32) = G_ADD [[BITCAST]], [[BITCAST2]]
   ; CHECK-NEXT:   [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[LSHR]], [[LSHR2]]
   ; CHECK-NEXT:   [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[BITCAST1]], [[BITCAST3]]
   ; CHECK-NEXT:   [[ADD3:%[0-9]+]]:_(s32) = G_ADD [[LSHR1]], [[LSHR3]]
   ; CHECK-NEXT:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
   ; CHECK-NEXT:   [[AND:%[0-9]+]]:_(s32) = G_AND [[ADD]], [[C2]]
   ; CHECK-NEXT:   [[AND1:%[0-9]+]]:_(s32) = G_AND [[ADD1]], [[C2]]
body: |
   ; CHECK: bb.0:
   ; CHECK-NEXT:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
   ; CHECK-NEXT:   liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, $vgpr8
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr8
   ; CHECK-NEXT:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; CHECK-NEXT:   [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[COPY1]](s32), [[C]]
-  ; CHECK-NEXT:   [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[UV]](s64), [[UV1]](s64), [[UV2]](s64)
   ; CHECK-NEXT:   G_BRCOND [[ICMP]](s1), %bb.1
   ; CHECK-NEXT:   G_BR %bb.2
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT: bb.1:
   ; CHECK-NEXT:   successors: %bb.2(0x80000000)
   ; CHECK-NEXT: {{  $}}
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-  ; CHECK-NEXT:   [[UV3:%[0-9]+]]:_(s64), [[UV4:%[0-9]+]]:_(s64), [[UV5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
+  ; CHECK-NEXT:   [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64)
   ; CHECK-NEXT:   [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV]](s64)
-  ; CHECK-NEXT:   [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
-  ; CHECK-NEXT:   [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV6]], [[UV8]]
-  ; CHECK-NEXT:   [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV7]], [[UV9]], [[UADDO1]]
+  ; CHECK-NEXT:   [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
+  ; CHECK-NEXT:   [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
   ; CHECK-NEXT:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+  ; CHECK-NEXT:   [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
   ; CHECK-NEXT:   [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
-  ; CHECK-NEXT:   [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV4]](s64)
-  ; CHECK-NEXT:   [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV10]], [[UV12]]
-  ; CHECK-NEXT:   [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV11]], [[UV13]], [[UADDO3]]
+  ; CHECK-NEXT:   [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
+  ; CHECK-NEXT:   [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO3]]
   ; CHECK-NEXT:   [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
+  ; CHECK-NEXT:   [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64)
   ; CHECK-NEXT:   [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV2]](s64)
-  ; CHECK-NEXT:   [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV5]](s64)
-  ; CHECK-NEXT:   [[UADDO4:%[0-9]+]]:_(s32), [[UADDO5:%[0-9]+]]:_(s1) = G_UADDO [[UV14]], [[UV16]]
-  ; CHECK-NEXT:   [[UADDE4:%[0-9]+]]:_(s32), [[UADDE5:%[0-9]+]]:_(s1) = G_UADDE [[UV15]], [[UV17]], [[UADDO5]]
+  ; CHECK-NEXT:   [[UADDO4:%[0-9]+]]:_(s32), [[UADDO5:%[0-9]+]]:_(s1) = G_UADDO [[UV12]], [[UV14]]
+  ; CHECK-NEXT:   [[UADDE4:%[0-9]+]]:_(s32), [[UADDE5:%[0-9]+]]:_(s1) = G_UADDE [[UV13]], [[UV15]], [[UADDO5]]
   ; CHECK-NEXT:   [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO4]](s32), [[UADDE4]](s32)
-  ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64)
+  ; CHECK-NEXT:   [[BUILD_VECTOR1:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64), [[MV2]](s64)
   ; CHECK-NEXT:   G_BR %bb.2
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT: bb.2:
-  ; CHECK-NEXT:   [[PHI:%[0-9]+]]:_(<3 x s64>) = G_PHI [[EXTRACT]](<3 x s64>), %bb.0, [[BUILD_VECTOR]](<3 x s64>), %bb.1
+  ; CHECK-NEXT:   [[PHI:%[0-9]+]]:_(<3 x s64>) = G_PHI [[BUILD_VECTOR]](<3 x s64>), %bb.0, [[BUILD_VECTOR1]](<3 x s64>), %bb.1
   ; CHECK-NEXT:   [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[PHI]](<3 x s64>), 0
-  ; CHECK-NEXT:   $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+  ; CHECK-NEXT:   [[UV16:%[0-9]+]]:_(s64), [[UV17:%[0-9]+]]:_(s64), [[UV18:%[0-9]+]]:_(s64), [[UV19:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+  ; CHECK-NEXT:   [[UV20:%[0-9]+]]:_(s64), [[UV21:%[0-9]+]]:_(s64), [[UV22:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[PHI]](<3 x s64>)
+  ; CHECK-NEXT:   [[BUILD_VECTOR2:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[UV20]](s64), [[UV21]](s64), [[UV22]](s64), [[UV19]](s64)
+  ; CHECK-NEXT:   $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR2]](<4 x s64>)
   ; CHECK-NEXT:   S_SETPC_B64 undef $sgpr30_sgpr31
   bb.0:
     successors: %bb.1, %bb.2
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, $vgpr8
 
     %0:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:_(s32) = COPY $vgpr8
     %2:_(s32) = G_CONSTANT i32 0

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir

bb.0:
 
     ; GFX6-LABEL: name: saddsat_v3s16
     ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32)
     ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2147483647
     ; GFX6-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 -2147483648
     ; GFX6-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GFX6-NEXT: [[SMAX:%[0-9]+]]:_(s32) = G_SMAX [[SHL]], [[C3]]
     ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[C1]], [[SMAX]]
     ; GFX6-NEXT: [[SMIN:%[0-9]+]]:_(s32) = G_SMIN [[SHL]], [[C3]]
     ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[C2]], [[SMIN]]
     ; GFX6-NEXT: [[SMAX1:%[0-9]+]]:_(s32) = G_SMAX [[SUB1]], [[SHL1]]
     ; GFX6-NEXT: [[SMIN1:%[0-9]+]]:_(s32) = G_SMIN [[SMAX1]], [[SUB]]
     ; GFX6-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[SHL]], [[SMIN1]]
     ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s32) = G_ASHR [[ADD]], [[C]](s32)
     ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
-    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[SMAX2:%[0-9]+]]:_(s32) = G_SMAX [[SHL2]], [[C3]]
     ; GFX6-NEXT: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[C1]], [[SMAX2]]
     ; GFX6-NEXT: [[SMIN2:%[0-9]+]]:_(s32) = G_SMIN [[SHL2]], [[C3]]
     ; GFX6-NEXT: [[SUB3:%[0-9]+]]:_(s32) = G_SUB [[C2]], [[SMIN2]]
     ; GFX6-NEXT: [[SMAX3:%[0-9]+]]:_(s32) = G_SMAX [[SUB3]], [[SHL3]]
     ; GFX6-NEXT: [[SMIN3:%[0-9]+]]:_(s32) = G_SMIN [[SMAX3]], [[SUB2]]
     ; GFX6-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[SHL2]], [[SMIN3]]
     ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s32) = G_ASHR [[ADD1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR2]], [[C]](s32)
     ; GFX6-NEXT: [[SMAX4:%[0-9]+]]:_(s32) = G_SMAX [[SHL4]], [[C3]]
     ; GFX6-NEXT: [[SUB4:%[0-9]+]]:_(s32) = G_SUB [[C1]], [[SMAX4]]
     ; GFX6-NEXT: [[SMIN4:%[0-9]+]]:_(s32) = G_SMIN [[SHL4]], [[C3]]
     ; GFX6-NEXT: [[SUB5:%[0-9]+]]:_(s32) = G_SUB [[C2]], [[SMIN4]]
     ; GFX6-NEXT: [[SMAX5:%[0-9]+]]:_(s32) = G_SMAX [[SUB5]], [[SHL5]]
     ; GFX6-NEXT: [[SMIN5:%[0-9]+]]:_(s32) = G_SMIN [[SMAX5]], [[SUB4]]
     ; GFX6-NEXT: [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[SHL4]], [[SMIN5]]
     ; GFX6-NEXT: [[ASHR2:%[0-9]+]]:_(s32) = G_ASHR [[ADD2]], [[C]](s32)
     ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX6-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[ASHR]], [[C4]]
     ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[ASHR1]], [[C4]]
     ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL6]]
-    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX6-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[ASHR2]], [[C4]]
-    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
+    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C4]]
     ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL7]]
-    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX6-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
-    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C4]]
+    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
     ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL8]]
-    ; GFX6-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX8-LABEL: name: saddsat_v3s16
     ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 32767
     ; GFX8-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 -32768
     ; GFX8-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 0
     ; GFX8-NEXT: [[SMAX:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC]], [[C3]]
     ; GFX8-NEXT: [[SUB:%[0-9]+]]:_(s16) = G_SUB [[C1]], [[SMAX]]
     ; GFX8-NEXT: [[SMIN:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC]], [[C3]]
     ; GFX8-NEXT: [[SUB1:%[0-9]+]]:_(s16) = G_SUB [[C2]], [[SMIN]]
bb.0:
     ; GFX8-NEXT: [[SMAX4:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC2]], [[C3]]
     ; GFX8-NEXT: [[SUB4:%[0-9]+]]:_(s16) = G_SUB [[C1]], [[SMAX4]]
     ; GFX8-NEXT: [[SMIN4:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC2]], [[C3]]
     ; GFX8-NEXT: [[SUB5:%[0-9]+]]:_(s16) = G_SUB [[C2]], [[SMIN4]]
     ; GFX8-NEXT: [[SMAX5:%[0-9]+]]:_(s16) = G_SMAX [[SUB5]], [[TRUNC5]]
     ; GFX8-NEXT: [[SMIN5:%[0-9]+]]:_(s16) = G_SMIN [[SMAX5]], [[SUB4]]
     ; GFX8-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[SMIN5]]
     ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[ADD]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[ADD1]](s16)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[ADD2]](s16)
     ; GFX8-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
+    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C4]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
-    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C4]]
+    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
-    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: saddsat_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR2]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[SADDSAT:%[0-9]+]]:_(<2 x s16>) = G_SADDSAT [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9-NEXT: [[SADDSAT1:%[0-9]+]]:_(<2 x s16>) = G_SADDSAT [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[SADDSAT]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SADDSAT1]](<2 x s16>)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SADDSAT]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[SADDSAT1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR3]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST7]](s32)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST6]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %3:_(<3 x s16>) = G_SADDSAT %1, %2
     %4:_(<3 x s16>) = G_IMPLICIT_DEF
     %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4
     $vgpr0_vgpr1_vgpr2 = COPY %5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-select.mir

body: |
   bb.0:
     liveins: $vgpr0, $vgpr1_vgpr2_vgpr3, $vgpr4_vgpr5_vgpr6
 
     ; CHECK-LABEL: name: test_select_v3s8
     ; CHECK: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr1_vgpr2_vgpr3
     ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr4_vgpr5_vgpr6
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY1]](<3 x s32>)
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(<3 x s8>) = G_TRUNC [[COPY2]](<3 x s32>)
     ; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[COPY]](s32), [[C]]
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF]], [[TRUNC]](<3 x s8>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF]], [[TRUNC1]](<3 x s8>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8), [[UV3:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT]](<4 x s8>)
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT1]](<4 x s8>)
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[UV]](s8)
-    ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s16) = G_ANYEXT [[UV4]](s8)
-    ; CHECK-NEXT: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[ANYEXT]], [[ANYEXT1]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[SELECT]](s16)
-    ; CHECK-NEXT: [[ANYEXT2:%[0-9]+]]:_(s16) = G_ANYEXT [[UV1]](s8)
-    ; CHECK-NEXT: [[ANYEXT3:%[0-9]+]]:_(s16) = G_ANYEXT [[UV5]](s8)
-    ; CHECK-NEXT: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[ANYEXT2]], [[ANYEXT3]]
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[SELECT1]](s16)
-    ; CHECK-NEXT: [[ANYEXT4:%[0-9]+]]:_(s16) = G_ANYEXT [[UV2]](s8)
-    ; CHECK-NEXT: [[ANYEXT5:%[0-9]+]]:_(s16) = G_ANYEXT [[UV6]](s8)
-    ; CHECK-NEXT: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[ANYEXT4]], [[ANYEXT5]]
-    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s8) = G_TRUNC [[SELECT2]](s16)
-    ; CHECK-NEXT: [[ANYEXT6:%[0-9]+]]:_(s16) = G_ANYEXT [[UV3]](s8)
-    ; CHECK-NEXT: [[ANYEXT7:%[0-9]+]]:_(s16) = G_ANYEXT [[UV7]](s8)
-    ; CHECK-NEXT: [[SELECT3:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[ANYEXT6]], [[ANYEXT7]]
-    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s8) = G_TRUNC [[SELECT3]](s16)
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC2]](s8), [[TRUNC3]](s8), [[TRUNC4]](s8), [[TRUNC5]](s8)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF]](<4 x s8>), [[DEF]](<4 x s8>)
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>)
-    ; CHECK-NEXT: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
+    ; CHECK-NEXT: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[TRUNC]], [[TRUNC1]]
+    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; CHECK-NEXT: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[TRUNC2]], [[TRUNC3]]
+    ; CHECK-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; CHECK-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
+    ; CHECK-NEXT: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[TRUNC4]], [[TRUNC5]]
+    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT]](s16)
+    ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT1]](s16)
+    ; CHECK-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT2]](s16)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(s32) = COPY $vgpr0
     %1:_(s32) = G_CONSTANT i32 0
     %2:_(<3 x s32>) = COPY $vgpr1_vgpr2_vgpr3
     %3:_(<3 x s32>) = COPY $vgpr4_vgpr5_vgpr6
     %4:_(<3 x s8>) = G_TRUNC %2
     %5:_(<3 x s8>) = G_TRUNC %3
 
     %6:_(s1) = G_ICMP intpred(ne), %0, %1
body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5, $vgpr6
     ; CHECK-LABEL: name: test_select_v3s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr6
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[COPY2]](s32), [[C]]
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; CHECK-NEXT: [[SELECT:%[0-9]+]]:_(<4 x s16>) = G_SELECT [[ICMP]](s1), [[INSERT]], [[INSERT1]]
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[SELECT]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C1]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C1]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C2]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C2]]
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C1]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C2]]
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C2]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C1]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY3]], [[C1]](s32)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C1]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C2]]
     ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C2]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C2]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C1]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
     ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C2]]
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[C]](s32)
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[SELECT:%[0-9]+]]:_(<4 x s16>) = G_SELECT [[ICMP]](s1), [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[SELECT]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C1]](s32)
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C1]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST8]], [[C2]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C2]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST9]], [[C2]]
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C2]]
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C2]]
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C2]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C1]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(s32) = COPY $vgpr6
     %3:_(s32) = G_CONSTANT i32 0
     %4:_(s1) = G_ICMP intpred(ne), %2, %3
     %5:_(<3 x s16>), %6:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %7:_(<3 x s16>), %8:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %9:_(<3 x s16>) = G_SELECT %4, %5, %7

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-sext-inreg.mir

     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY1]](s32), [[C1]](s32)
     ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[SHL]], [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32)
     ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
     ; GFX9-NEXT: [[ASHR1:%[0-9]+]]:_(<2 x s16>) = G_ASHR [[SHL1]], [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[ASHR]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[ASHR1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[BITCAST4]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR2]](s32), [[BITCAST5]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-sext.mir

 ---
 name: test_sext_v3s16_to_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: test_sext_v3s16_to_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST]], 16
     ; CHECK-NEXT: [[SEXT_INREG1:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR]], 16
     ; CHECK-NEXT: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST1]], 16
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[SEXT_INREG]](s32), [[SEXT_INREG1]](s32), [[SEXT_INREG2]](s32)

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shl.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_shl_s32_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_shl_s32_s32

 ---
 name: test_shl_v3s64_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_shl_v3s64_v3s32
     ; SI: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV]], [[UV3]](s32)
-    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[UV4]](s32)
-    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[UV2]], [[UV5]](s32)
-    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[SHL]](s64), [[SHL1]](s64), [[SHL2]](s64)
+    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV]], [[UV4]](s32)
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[UV5]](s32)
+    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[UV2]], [[UV6]](s32)
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; SI-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[SHL]](s64), [[SHL1]](s64), [[SHL2]](s64), [[UV10]](s64)
+    ; SI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; VI-LABEL: name: test_shl_v3s64_v3s32
     ; VI: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV]], [[UV3]](s32)
-    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[UV4]](s32)
-    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[UV2]], [[UV5]](s32)
-    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[SHL]](s64), [[SHL1]](s64), [[SHL2]](s64)
+    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV]], [[UV4]](s32)
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[UV5]](s32)
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[UV2]], [[UV6]](s32)
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; VI-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[SHL]](s64), [[SHL1]](s64), [[SHL2]](s64), [[UV10]](s64)
+    ; VI-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     ; GFX9-LABEL: name: test_shl_v3s64_v3s32
     ; GFX9: [[COPY:%[0-9]+]]:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s64>) = G_EXTRACT [[COPY]](<4 x s64>), 0
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64), [[UV3:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[COPY]](<4 x s64>)
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s64), [[UV1:%[0-9]+]]:_(s64), [[UV2:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[EXTRACT]](<3 x s64>)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV]], [[UV3]](s32)
-    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[UV4]](s32)
-    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[UV2]], [[UV5]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[SHL]](s64), [[SHL1]](s64), [[SHL2]](s64)
+    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[UV]], [[UV4]](s32)
+    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s64) = G_SHL [[UV1]], [[UV5]](s32)
+    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s64) = G_SHL [[UV2]], [[UV6]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s64>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s64>) = G_INSERT [[DEF]], [[BUILD_VECTOR]](<3 x s64>), 0
-    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<4 x s64>)
+    ; GFX9-NEXT: [[UV7:%[0-9]+]]:_(s64), [[UV8:%[0-9]+]]:_(s64), [[UV9:%[0-9]+]]:_(s64), [[UV10:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[DEF]](<4 x s64>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s64>) = G_BUILD_VECTOR [[SHL]](s64), [[SHL1]](s64), [[SHL2]](s64), [[UV10]](s64)
+    ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR]](<4 x s64>)
     %0:_(<4 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
     %1:_(<3 x s64>) = G_EXTRACT %0, 0
     %2:_(<3 x s32>) = COPY $vgpr8_vgpr9_vgpr10
     %3:_(<3 x s64>) = G_SHL %1, %2
     %4:_(<4 x s64>) = G_IMPLICIT_DEF
     %5:_(<4 x s64>) = G_INSERT %4, %3, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %5
 ...
bb.0:
     ; VI-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL5]]
     ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
     ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
     ; VI-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: test_shl_v3s16_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT1]](<4 x s16>), 32
-    ; GFX9-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT2]](<4 x s16>), 0
-    ; GFX9-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT3:%[0-9]+]]:_(s16) = G_EXTRACT [[INSERT3]](<4 x s16>), 32
-    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[EXTRACT]], [[EXTRACT2]](<2 x s16>)
-    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[EXTRACT1]], [[EXTRACT3]](s16)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[SHL]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC1]](s16)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SHL]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SHL1]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST2]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST6]](s32)
+    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_SHL %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shuffle-vector.mir

body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
 
     ; CHECK-LABEL: name: shufflevector_v3s16_2_0
     ; CHECK: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; CHECK-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
+    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
     ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
-    ; CHECK-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT3]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST3]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]]
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]]
     ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C1]]
     ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
     ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<4 x s16>) = COPY $vgpr2_vgpr3
     %2:_(<3 x s16>) = G_EXTRACT %0, 0
     %3:_(<3 x s16>) = G_EXTRACT %1, 0
     %4:_(<4 x s16>) = G_SHUFFLE_VECTOR %2, %3, shufflemask(5, 1, 3, 0)
     $vgpr0_vgpr1 = COPY %4
 
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-shuffle-vector.s16.mir

   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
 
     ; GFX8-LABEL: name: shufflevector_v4s16_v3s16_2_0
     ; GFX8: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
     ; GFX8-NEXT: {{  $}}
     ; GFX8-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; GFX8-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; GFX8-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; GFX8-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX8-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
     ; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
-    ; GFX8-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
-    ; GFX8-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
+    ; GFX8-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C1]]
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[COPY3]], [[C1]]
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
     ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C1]]
     ; GFX8-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[COPY5]], [[C1]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
     ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST3]](<2 x s16>), [[BITCAST4]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     ; GFX9-LABEL: name: shufflevector_v4s16_v3s16_2_0
     ; GFX9: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
     ; GFX9-NEXT: {{  $}}
     ; GFX9-NEXT: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr2_vgpr3
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY1]](<4 x s16>), 0
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT1]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
-    ; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
-    ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST2]](s32)
+    ; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY2]](s32), [[COPY3]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY4]](s32), [[COPY5]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
     %0:_(<4 x s16>) = COPY $vgpr0_vgpr1
     %1:_(<4 x s16>) = COPY $vgpr2_vgpr3
     %2:_(<3 x s16>) = G_EXTRACT %0, 0
     %3:_(<3 x s16>) = G_EXTRACT %1, 0

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smax.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_smax_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_smax_s32

 ---
 name: test_smax_v3s16
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_smax_v3s16
     ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST]], 16
     ; SI-NEXT: [[SEXT_INREG1:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST2]], 16
     ; SI-NEXT: [[SMAX:%[0-9]+]]:_(s32) = G_SMAX [[SEXT_INREG]], [[SEXT_INREG1]]
     ; SI-NEXT: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR]], 16
     ; SI-NEXT: [[SEXT_INREG3:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR1]], 16
     ; SI-NEXT: [[SMAX1:%[0-9]+]]:_(s32) = G_SMAX [[SEXT_INREG2]], [[SEXT_INREG3]]
     ; SI-NEXT: [[SEXT_INREG4:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST1]], 16
     ; SI-NEXT: [[SEXT_INREG5:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST3]], 16
     ; SI-NEXT: [[SMAX2:%[0-9]+]]:_(s32) = G_SMAX [[SEXT_INREG4]], [[SEXT_INREG5]]
     ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[SMAX]](s32), [[SMAX1]](s32), [[SMAX2]](s32)
     ; SI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; VI-LABEL: name: test_smax_v3s16
     ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[SMAX:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC]], [[TRUNC3]]
     ; VI-NEXT: [[SMAX1:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC1]], [[TRUNC4]]
     ; VI-NEXT: [[SMAX2:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC2]], [[TRUNC5]]
     ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SMAX]](s16)
     ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[SMAX1]](s16)
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[SMAX2]](s16)
     ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; VI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX9-LABEL: name: test_smax_v3s16
     ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; GFX9-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[SMAX:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[UV8]], [[UV10]]
-    ; GFX9-NEXT: [[SMAX1:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[UV9]], [[UV11]]
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[SMAX]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[SMAX1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[SMAX:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[SMAX1:%[0-9]+]]:_(<2 x s16>) = G_SMAX [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SMAX]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[SMAX1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST4]](s32), [[LSHR2]](s32), [[BITCAST5]](s32)
     ; GFX9-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<3 x s16>) = G_IMPLICIT_DEF
     %2:_(<3 x s16>) = G_SMAX %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     S_NOP 0, implicit %3
 ...
 

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-smin.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_smin_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_smin_s32

 ---
 name: test_smin_v3s16
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_smin_v3s16
     ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST]], 16
     ; SI-NEXT: [[SEXT_INREG1:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST2]], 16
     ; SI-NEXT: [[SMIN:%[0-9]+]]:_(s32) = G_SMIN [[SEXT_INREG]], [[SEXT_INREG1]]
     ; SI-NEXT: [[SEXT_INREG2:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR]], 16
     ; SI-NEXT: [[SEXT_INREG3:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR1]], 16
     ; SI-NEXT: [[SMIN1:%[0-9]+]]:_(s32) = G_SMIN [[SEXT_INREG2]], [[SEXT_INREG3]]
     ; SI-NEXT: [[SEXT_INREG4:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST1]], 16
     ; SI-NEXT: [[SEXT_INREG5:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST3]], 16
     ; SI-NEXT: [[SMIN2:%[0-9]+]]:_(s32) = G_SMIN [[SEXT_INREG4]], [[SEXT_INREG5]]
     ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[SMIN]](s32), [[SMIN1]](s32), [[SMIN2]](s32)
     ; SI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; VI-LABEL: name: test_smin_v3s16
     ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[SMIN:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC]], [[TRUNC3]]
     ; VI-NEXT: [[SMIN1:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC1]], [[TRUNC4]]
     ; VI-NEXT: [[SMIN2:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC2]], [[TRUNC5]]
     ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SMIN]](s16)
     ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[SMIN1]](s16)
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[SMIN2]](s16)
     ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; VI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX9-LABEL: name: test_smin_v3s16
     ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; GFX9-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[SMIN:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[UV8]], [[UV10]]
-    ; GFX9-NEXT: [[SMIN1:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[UV9]], [[UV11]]
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[SMIN]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[SMIN1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[SMIN:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[SMIN1:%[0-9]+]]:_(<2 x s16>) = G_SMIN [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SMIN]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[SMIN1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST4]](s32), [[LSHR2]](s32), [[BITCAST5]](s32)
     ; GFX9-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<3 x s16>) = G_IMPLICIT_DEF
     %2:_(<3 x s16>) = G_SMIN %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     S_NOP 0, implicit %3
 ...
 

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-sshlsat.mir

bb.0:
 
     ; GFX6-LABEL: name: sshlsat_v3s16
     ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[SHL]], [[AND]](s32)
     ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s32) = G_ASHR [[SHL1]], [[AND]](s32)
     ; GFX6-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 -2147483648
     ; GFX6-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 2147483647
     ; GFX6-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
     ; GFX6-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SHL]](s32), [[C4]]
     ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[ICMP]](s1), [[C2]], [[C3]]
     ; GFX6-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[SHL]](s32), [[ASHR]]
     ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s32) = G_SELECT [[ICMP1]](s1), [[SELECT]], [[SHL1]]
     ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s32) = G_ASHR [[SELECT1]], [[C]](s32)
-    ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
     ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
     ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[SHL2]], [[AND1]](s32)
     ; GFX6-NEXT: [[ASHR2:%[0-9]+]]:_(s32) = G_ASHR [[SHL3]], [[AND1]](s32)
     ; GFX6-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SHL2]](s32), [[C4]]
     ; GFX6-NEXT: [[SELECT2:%[0-9]+]]:_(s32) = G_SELECT [[ICMP2]](s1), [[C2]], [[C3]]
     ; GFX6-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[SHL2]](s32), [[ASHR2]]
     ; GFX6-NEXT: [[SELECT3:%[0-9]+]]:_(s32) = G_SELECT [[ICMP3]](s1), [[SELECT2]], [[SHL3]]
     ; GFX6-NEXT: [[ASHR3:%[0-9]+]]:_(s32) = G_ASHR [[SELECT3]], [[C]](s32)
     ; GFX6-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
     ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[SHL4]], [[AND2]](s32)
     ; GFX6-NEXT: [[ASHR4:%[0-9]+]]:_(s32) = G_ASHR [[SHL5]], [[AND2]](s32)
     ; GFX6-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SHL4]](s32), [[C4]]
     ; GFX6-NEXT: [[SELECT4:%[0-9]+]]:_(s32) = G_SELECT [[ICMP4]](s1), [[C2]], [[C3]]
     ; GFX6-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[SHL4]](s32), [[ASHR4]]
     ; GFX6-NEXT: [[SELECT5:%[0-9]+]]:_(s32) = G_SELECT [[ICMP5]](s1), [[SELECT4]], [[SHL5]]
     ; GFX6-NEXT: [[ASHR5:%[0-9]+]]:_(s32) = G_ASHR [[SELECT5]], [[C]](s32)
     ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[ASHR1]], [[C1]]
     ; GFX6-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[ASHR3]], [[C1]]
     ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
     ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL6]]
-    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[ASHR5]], [[C1]]
-    ; GFX6-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; GFX6-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32)
     ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL7]]
-    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX6-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
-    ; GFX6-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; GFX6-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32)
     ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL8]]
-    ; GFX6-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX8-LABEL: name: sshlsat_v3s16
     ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16)
     ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s16) = G_ASHR [[SHL]], [[TRUNC3]](s16)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 -32768
     ; GFX8-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 32767
     ; GFX8-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 0
     ; GFX8-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[TRUNC]](s16), [[C3]]
     ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[C1]], [[C2]]
     ; GFX8-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC]](s16), [[ASHR]]
     ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP1]](s1), [[SELECT]], [[SHL]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[TRUNC4]](s16)
     ; GFX8-NEXT: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[SHL1]], [[TRUNC4]](s16)
     ; GFX8-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[TRUNC1]](s16), [[C3]]
     ; GFX8-NEXT: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP2]](s1), [[C1]], [[C2]]
     ; GFX8-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC1]](s16), [[ASHR1]]
     ; GFX8-NEXT: [[SELECT3:%[0-9]+]]:_(s16) = G_SELECT [[ICMP3]](s1), [[SELECT2]], [[SHL1]]
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[TRUNC5]](s16)
     ; GFX8-NEXT: [[ASHR2:%[0-9]+]]:_(s16) = G_ASHR [[SHL2]], [[TRUNC5]](s16)
     ; GFX8-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[TRUNC2]](s16), [[C3]]
     ; GFX8-NEXT: [[SELECT4:%[0-9]+]]:_(s16) = G_SELECT [[ICMP4]](s1), [[C1]], [[C2]]
     ; GFX8-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC2]](s16), [[ASHR2]]
     ; GFX8-NEXT: [[SELECT5:%[0-9]+]]:_(s16) = G_SELECT [[ICMP5]](s1), [[SELECT4]], [[SHL2]]
     ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT1]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT3]](s16)
     ; GFX8-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
-    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT5]](s16)
     ; GFX8-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
+    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C4]]
     ; GFX8-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL4]]
-    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
-    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C4]]
+    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
     ; GFX8-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL5]]
-    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: sshlsat_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16)
     ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s16) = G_ASHR [[SHL]], [[TRUNC3]](s16)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 -32768
     ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 32767
     ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 0
     ; GFX9-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[TRUNC]](s16), [[C3]]
     ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[C1]], [[C2]]
     ; GFX9-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC]](s16), [[ASHR]]
     ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP1]](s1), [[SELECT]], [[SHL]]
     ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[TRUNC4]](s16)
     ; GFX9-NEXT: [[ASHR1:%[0-9]+]]:_(s16) = G_ASHR [[SHL1]], [[TRUNC4]](s16)
     ; GFX9-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[TRUNC1]](s16), [[C3]]
     ; GFX9-NEXT: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP2]](s1), [[C1]], [[C2]]
     ; GFX9-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC1]](s16), [[ASHR1]]
     ; GFX9-NEXT: [[SELECT3:%[0-9]+]]:_(s16) = G_SELECT [[ICMP3]](s1), [[SELECT2]], [[SHL1]]
     ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[TRUNC5]](s16)
     ; GFX9-NEXT: [[ASHR2:%[0-9]+]]:_(s16) = G_ASHR [[SHL2]], [[TRUNC5]](s16)
     ; GFX9-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[TRUNC2]](s16), [[C3]]
     ; GFX9-NEXT: [[SELECT4:%[0-9]+]]:_(s16) = G_SELECT [[ICMP4]](s1), [[C1]], [[C2]]
     ; GFX9-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC2]](s16), [[ASHR2]]
     ; GFX9-NEXT: [[SELECT5:%[0-9]+]]:_(s16) = G_SELECT [[ICMP5]](s1), [[SELECT4]], [[SHL2]]
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT1]](s16)
     ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT3]](s16)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32)
     ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT5]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[BITCAST4]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR3]](s32), [[BITCAST4]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %3:_(<3 x s16>) = G_SSHLSAT %1, %2
     %4:_(<3 x s16>) = G_IMPLICIT_DEF
     %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4
     $vgpr0_vgpr1_vgpr2 = COPY %5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir

bb.0:
 
     ; GFX6-LABEL: name: ssubsat_v3s16
     ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32)
     ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2147483647
     ; GFX6-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 -2147483648
     ; GFX6-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 -1
     ; GFX6-NEXT: [[SMAX:%[0-9]+]]:_(s32) = G_SMAX [[SHL]], [[C3]]
     ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SMAX]], [[C1]]
     ; GFX6-NEXT: [[SMIN:%[0-9]+]]:_(s32) = G_SMIN [[SHL]], [[C3]]
     ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SMIN]], [[C2]]
     ; GFX6-NEXT: [[SMAX1:%[0-9]+]]:_(s32) = G_SMAX [[SUB]], [[SHL1]]
     ; GFX6-NEXT: [[SMIN1:%[0-9]+]]:_(s32) = G_SMIN [[SMAX1]], [[SUB1]]
     ; GFX6-NEXT: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[SMIN1]]
     ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s32) = G_ASHR [[SUB2]], [[C]](s32)
     ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
-    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[SMAX2:%[0-9]+]]:_(s32) = G_SMAX [[SHL2]], [[C3]]
     ; GFX6-NEXT: [[SUB3:%[0-9]+]]:_(s32) = G_SUB [[SMAX2]], [[C1]]
     ; GFX6-NEXT: [[SMIN2:%[0-9]+]]:_(s32) = G_SMIN [[SHL2]], [[C3]]
     ; GFX6-NEXT: [[SUB4:%[0-9]+]]:_(s32) = G_SUB [[SMIN2]], [[C2]]
     ; GFX6-NEXT: [[SMAX3:%[0-9]+]]:_(s32) = G_SMAX [[SUB3]], [[SHL3]]
     ; GFX6-NEXT: [[SMIN3:%[0-9]+]]:_(s32) = G_SMIN [[SMAX3]], [[SUB4]]
     ; GFX6-NEXT: [[SUB5:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[SMIN3]]
     ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s32) = G_ASHR [[SUB5]], [[C]](s32)
     ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR2]], [[C]](s32)
     ; GFX6-NEXT: [[SMAX4:%[0-9]+]]:_(s32) = G_SMAX [[SHL4]], [[C3]]
     ; GFX6-NEXT: [[SUB6:%[0-9]+]]:_(s32) = G_SUB [[SMAX4]], [[C1]]
     ; GFX6-NEXT: [[SMIN4:%[0-9]+]]:_(s32) = G_SMIN [[SHL4]], [[C3]]
     ; GFX6-NEXT: [[SUB7:%[0-9]+]]:_(s32) = G_SUB [[SMIN4]], [[C2]]
     ; GFX6-NEXT: [[SMAX5:%[0-9]+]]:_(s32) = G_SMAX [[SUB6]], [[SHL5]]
     ; GFX6-NEXT: [[SMIN5:%[0-9]+]]:_(s32) = G_SMIN [[SMAX5]], [[SUB7]]
     ; GFX6-NEXT: [[SUB8:%[0-9]+]]:_(s32) = G_SUB [[SHL4]], [[SMIN5]]
     ; GFX6-NEXT: [[ASHR2:%[0-9]+]]:_(s32) = G_ASHR [[SUB8]], [[C]](s32)
     ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX6-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[ASHR]], [[C4]]
     ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[ASHR1]], [[C4]]
     ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL6]]
-    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX6-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[ASHR2]], [[C4]]
-    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
+    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C4]]
     ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL7]]
-    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX6-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
-    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C4]]
+    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
     ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL8]]
-    ; GFX6-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX8-LABEL: name: ssubsat_v3s16
     ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 32767
     ; GFX8-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 -32768
     ; GFX8-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 -1
     ; GFX8-NEXT: [[SMAX:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC]], [[C3]]
     ; GFX8-NEXT: [[SUB:%[0-9]+]]:_(s16) = G_SUB [[SMAX]], [[C1]]
     ; GFX8-NEXT: [[SMIN:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC]], [[C3]]
     ; GFX8-NEXT: [[SUB1:%[0-9]+]]:_(s16) = G_SUB [[SMIN]], [[C2]]
bb.0:
     ; GFX8-NEXT: [[SMAX4:%[0-9]+]]:_(s16) = G_SMAX [[TRUNC2]], [[C3]]
     ; GFX8-NEXT: [[SUB6:%[0-9]+]]:_(s16) = G_SUB [[SMAX4]], [[C1]]
     ; GFX8-NEXT: [[SMIN4:%[0-9]+]]:_(s16) = G_SMIN [[TRUNC2]], [[C3]]
     ; GFX8-NEXT: [[SUB7:%[0-9]+]]:_(s16) = G_SUB [[SMIN4]], [[C2]]
     ; GFX8-NEXT: [[SMAX5:%[0-9]+]]:_(s16) = G_SMAX [[SUB6]], [[TRUNC5]]
     ; GFX8-NEXT: [[SMIN5:%[0-9]+]]:_(s16) = G_SMIN [[SMAX5]], [[SUB7]]
     ; GFX8-NEXT: [[SUB8:%[0-9]+]]:_(s16) = G_SUB [[TRUNC2]], [[SMIN5]]
     ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[SUB2]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[SUB5]](s16)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[SUB8]](s16)
     ; GFX8-NEXT: [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
+    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C4]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C4]]
-    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C4]]
+    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C4]]
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
-    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: ssubsat_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR2]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[SSUBSAT:%[0-9]+]]:_(<2 x s16>) = G_SSUBSAT [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9-NEXT: [[SSUBSAT1:%[0-9]+]]:_(<2 x s16>) = G_SSUBSAT [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[SSUBSAT]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SSUBSAT1]](<2 x s16>)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SSUBSAT]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[SSUBSAT1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR3]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST7]](s32)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST6]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %3:_(<3 x s16>) = G_SSUBSAT %1, %2
     %4:_(<3 x s16>) = G_IMPLICIT_DEF
     %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4
     $vgpr0_vgpr1_vgpr2 = COPY %5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store-global.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer -global-isel-abort=0 -pass-remarks-missed='gisel.*'  %s 2> %t.err -o - | FileCheck -check-prefix=SI %s
-# RUN: FileCheck -check-prefix=ERR %s < %t.err
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -O0 -run-pass=legalizer -global-isel-abort=0  %s -o - | FileCheck -check-prefix=CI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer  -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer  -global-isel-abort=0 -pass-remarks-missed='gisel.*' 2> %t.err %s -o - | FileCheck -check-prefix=GFX9 %s
-# RUN: FileCheck -check-prefix=ERR %s < %t.err
-# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer  -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9 %s
-
-
-# ERR-NOT: remark
-# ERR: remark: <unknown>:0:0: unable to legalize instruction: %{{[0-9]+}}:_(<132 x s16>) = G_CONCAT_VECTORS {{.*}} (in function: test_store_global_v11s16_align4)
-# ERR-NOT: remark
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -O0 -run-pass=legalizer  %s -o - | FileCheck -check-prefix=SI %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=bonaire -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=CI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O0 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9 %s
 
 ---
 name: test_store_global_s1_align1
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2
 
     ; SI-LABEL: name: test_store_global_s1_align1

 name: test_store_global_v3s32_align1
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_v3s32_align1
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY1]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<3 x s32>), 64
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]

     ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
     ; SI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[AND3]], [[COPY6]](s32)
     ; SI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD4]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD4]](p1) :: (store (s8) into unknown-address + 6, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR5]](s32), [[PTR_ADD6]](p1) :: (store (s8) into unknown-address + 7, addrspace 1)
     ; SI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C6]](s64)
-    ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
     ; SI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[COPY7]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; SI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C3]]
     ; SI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[AND4]], [[COPY8]](s32)
     ; SI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY7]](s32), [[PTR_ADD7]](p1) :: (store (s8) into unknown-address + 8, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR7]](s32), [[PTR_ADD9]](p1) :: (store (s8) into unknown-address + 9, addrspace 1)

 name: test_store_global_v3s32_align2
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_v3s32_align2
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY1]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<3 x s32>), 64
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY3]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD1]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY3]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 6, addrspace 1)
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
-    ; SI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
     ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[COPY4]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD3]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY4]](s32), [[PTR_ADD3]](p1) :: (store (s16) into unknown-address + 8, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR2]](s32), [[PTR_ADD4]](p1) :: (store (s16) into unknown-address + 10, addrspace 1)
     ; CI-LABEL: name: test_store_global_v3s32_align2
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 2, addrspace 1)

 name: test_store_global_v3s32_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_v3s32_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY1]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_v3s32_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
     ; VI-LABEL: name: test_store_global_v3s32_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)

 name: test_store_global_v3s32_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_v3s32_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY1]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_v3s32_align8
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_v3s32_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)

 name: test_store_global_v3s32_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_v3s32_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY1]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_v3s32_align16
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v3s32_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_s96_align1
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<3 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]

     ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C3]]
     ; SI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[AND3]], [[COPY6]](s32)
     ; SI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD4]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD4]](p1) :: (store (s8) into unknown-address + 6, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR5]](s32), [[PTR_ADD6]](p1) :: (store (s8) into unknown-address + 7, addrspace 1)
     ; SI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C6]](s64)
-    ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
     ; SI-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[COPY7]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; SI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[COPY7]], [[C3]]
     ; SI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[AND4]], [[COPY8]](s32)
     ; SI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY7]](s32), [[PTR_ADD7]](p1) :: (store (s8) into unknown-address + 8, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR7]](s32), [[PTR_ADD9]](p1) :: (store (s8) into unknown-address + 9, addrspace 1)

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_s96_align2
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<2 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<3 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[UV1]](s32)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[COPY3]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD1]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY3]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 4, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR1]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 6, addrspace 1)
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD3:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C3]](s64)
-    ; SI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV2]](s32)
     ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[COPY4]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD4:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD3]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY4]](s32), [[PTR_ADD3]](p1) :: (store (s16) into unknown-address + 8, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR2]](s32), [[PTR_ADD4]](p1) :: (store (s16) into unknown-address + 10, addrspace 1)
     ; CI-LABEL: name: test_store_global_s96_align2
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_s96_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_s96_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
     ; CI-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
     ; VI-LABEL: name: test_store_global_s96_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_s96_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_s96_align8
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
     ; CI-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_s96_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_s96_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_s96_align16
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY1]](s96)
     ; CI-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_s96_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s96) = COPY $vgpr2_vgpr3_vgpr4

 name: test_store_global_v5s32_align1
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5s32_align1
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]

     ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
     ; SI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[AND7]], [[COPY12]](s32)
     ; SI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR11]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; SI-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C8]](s64)
-    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY13]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; SI-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C3]]
     ; SI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[AND8]], [[COPY14]](s32)
     ; SI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY13]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR13]](s32), [[PTR_ADD17]](p1) :: (store (s8) into unknown-address + 17, addrspace 1)
     ; SI-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C3]]
     ; SI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[AND9]], [[COPY15]](s32)
     ; SI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD16]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR12]](s32), [[PTR_ADD16]](p1) :: (store (s8) into unknown-address + 18, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR14]](s32), [[PTR_ADD18]](p1) :: (store (s8) into unknown-address + 19, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5s32_align1
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5s32_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[C2]](s16)

     ; VI-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
     ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC7]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR11]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT7]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; VI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C7]](s64)
-    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[COPY6]](s32)
     ; VI-NEXT: [[LSHR13:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC8]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: [[ANYEXT8:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR13]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT8]](s32), [[PTR_ADD17]](p1) :: (store (s8) into unknown-address + 17, addrspace 1)
     ; VI-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
     ; VI-NEXT: [[LSHR14:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC9]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD16]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR12]](s32), [[PTR_ADD16]](p1) :: (store (s8) into unknown-address + 18, addrspace 1)
     ; VI-NEXT: [[ANYEXT9:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR14]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT9]](s32), [[PTR_ADD18]](p1) :: (store (s8) into unknown-address + 19, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5s32_align1
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x s32>), align 1, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5s32_align2
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5s32_align2
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; SI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY5]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY5]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5s32_align2
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5s32_align2
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY5]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY5]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5s32_align2
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x s32>), align 2, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5s32_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5s32_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5s32_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5s32_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5s32_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x s32>), align 4, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5s32_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5s32_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5s32_align8
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5s32_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5s32_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x s32>), align 8, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5s32_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5s32_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5s32_align16
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5s32_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5s32_align16
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[COPY1]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x s32>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x s32>), align 16, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5p3_align1
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-LABEL: name: test_store_global_v5p3_align1
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]

     ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
     ; SI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[AND7]], [[COPY12]](s32)
     ; SI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR11]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; SI-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C8]](s64)
-    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY13]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; SI-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C3]]
     ; SI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[AND8]], [[COPY14]](s32)
     ; SI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY13]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR13]](s32), [[PTR_ADD17]](p1) :: (store (s8) into unknown-address + 17, addrspace 1)
     ; SI-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C3]]
     ; SI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[AND9]], [[COPY15]](s32)
     ; SI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD16]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR12]](s32), [[PTR_ADD16]](p1) :: (store (s8) into unknown-address + 18, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR14]](s32), [[PTR_ADD18]](p1) :: (store (s8) into unknown-address + 19, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5p3_align1
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5p3_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[C2]](s16)

     ; VI-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
     ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC7]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR11]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT7]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; VI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C7]](s64)
-    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[COPY6]](s32)
     ; VI-NEXT: [[LSHR13:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC8]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: [[ANYEXT8:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR13]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT8]](s32), [[PTR_ADD17]](p1) :: (store (s8) into unknown-address + 17, addrspace 1)
     ; VI-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
     ; VI-NEXT: [[LSHR14:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC9]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD16]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR12]](s32), [[PTR_ADD16]](p1) :: (store (s8) into unknown-address + 18, addrspace 1)
     ; VI-NEXT: [[ANYEXT9:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR14]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT9]](s32), [[PTR_ADD18]](p1) :: (store (s8) into unknown-address + 19, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5p3_align1
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x p3>), align 1, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5p3_align2
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5p3_align2
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; SI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY5]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY5]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5p3_align2
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5p3_align2
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY5]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY5]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5p3_align2
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x p3>), align 2, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5p3_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5p3_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5p3_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5p3_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5p3_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x p3>), align 4, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5p3_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5p3_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5p3_align8
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5p3_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5p3_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x p3>), align 8, addrspace 1)
 ...
 
 ---
 name: test_store_global_v5p3_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v5p3_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v5p3_align16
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v5p3_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v5p3_align16
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](<5 x p3>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<5 x p3>) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (<5 x p3>), align 16, addrspace 1)
 ...
 
 ---
 name: test_store_global_v10s16_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v10s16_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v10s16_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v10s16_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v10s16_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<10 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[DEF]](<10 x s16>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<10 x s16>) = G_IMPLICIT_DEF
     G_STORE %1, %0 :: (store (<10 x s16>), align 16, addrspace 1)
 ...
 
 ---
 name: test_store_global_v11s16_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v11s16_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<12 x s16>)
-    ; SI-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>), [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>), [[UV65:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<132 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[UV2]](<2 x s16>), [[UV3]](<2 x s16>), [[UV4]](<2 x s16>), [[UV5]](<2 x s16>), [[UV6]](<2 x s16>), [[UV7]](<2 x s16>), [[UV8]](<2 x s16>), [[UV9]](<2 x s16>), [[UV10]](<2 x s16>), [[UV11]](<2 x s16>), [[UV12]](<2 x s16>), [[UV13]](<2 x s16>), [[UV14]](<2 x s16>), [[UV15]](<2 x s16>), [[UV16]](<2 x s16>), [[UV17]](<2 x s16>), [[UV18]](<2 x s16>), [[UV19]](<2 x s16>), [[UV20]](<2 x s16>), [[UV21]](<2 x s16>), [[UV22]](<2 x s16>), [[UV23]](<2 x s16>), [[UV24]](<2 x s16>), [[UV25]](<2 x s16>), [[UV26]](<2 x s16>), [[UV27]](<2 x s16>), [[UV28]](<2 x s16>), [[UV29]](<2 x s16>), [[UV30]](<2 x s16>), [[UV31]](<2 x s16>), [[UV32]](<2 x s16>), [[UV33]](<2 x s16>), [[UV34]](<2 x s16>), [[UV35]](<2 x s16>), [[UV36]](<2 x s16>), [[UV37]](<2 x s16>), [[UV38]](<2 x s16>), [[UV39]](<2 x s16>), [[UV40]](<2 x s16>), [[UV41]](<2 x s16>), [[UV42]](<2 x s16>), [[UV43]](<2 x s16>), [[UV44]](<2 x s16>), [[UV45]](<2 x s16>), [[UV46]](<2 x s16>), [[UV47]](<2 x s16>), [[UV48]](<2 x s16>), [[UV49]](<2 x s16>), [[UV50]](<2 x s16>), [[UV51]](<2 x s16>), [[UV52]](<2 x s16>), [[UV53]](<2 x s16>), [[UV54]](<2 x s16>), [[UV55]](<2 x s16>), [[UV56]](<2 x s16>), [[UV57]](<2 x s16>), [[UV58]](<2 x s16>), [[UV59]](<2 x s16>), [[UV60]](<2 x s16>), [[UV61]](<2 x s16>), [[UV62]](<2 x s16>), [[UV63]](<2 x s16>), [[UV64]](<2 x s16>), [[UV65]](<2 x s16>)
-    ; SI-NEXT: [[UV66:%[0-9]+]]:_(<11 x s16>), [[UV67:%[0-9]+]]:_(<11 x s16>), [[UV68:%[0-9]+]]:_(<11 x s16>), [[UV69:%[0-9]+]]:_(<11 x s16>), [[UV70:%[0-9]+]]:_(<11 x s16>), [[UV71:%[0-9]+]]:_(<11 x s16>), [[UV72:%[0-9]+]]:_(<11 x s16>), [[UV73:%[0-9]+]]:_(<11 x s16>), [[UV74:%[0-9]+]]:_(<11 x s16>), [[UV75:%[0-9]+]]:_(<11 x s16>), [[UV76:%[0-9]+]]:_(<11 x s16>), [[UV77:%[0-9]+]]:_(<11 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<132 x s16>)
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<8 x s16>) = G_EXTRACT [[INSERT]](<12 x s16>), 0
-    ; SI-NEXT: [[INSERT1:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<12 x s16>), 128
-    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[EXTRACT]](<8 x s16>)
-    ; SI-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
-    ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
-    ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0
-    ; SI-NEXT: [[UV78:%[0-9]+]]:_(<2 x s16>), [[UV79:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>)
-    ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV78]](<2 x s16>)
-    ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
-    ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV79]](<2 x s16>)
-    ; SI-NEXT: G_STORE [[BITCAST1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
-    ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
-    ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
-    ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
-    ; SI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
-    ; SI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
-    ; SI-NEXT: G_STORE [[BITCAST2]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+    ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; SI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; SI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; SI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; SI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; SI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; SI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; SI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; SI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; SI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; SI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; SI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; SI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; SI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; SI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; SI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; SI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; SI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
+    ; SI-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>)
+    ; SI-NEXT: [[BITCAST10:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+    ; SI-NEXT: G_STORE [[BITCAST10]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+    ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
+    ; SI-NEXT: G_STORE [[BITCAST4]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+    ; SI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+    ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
+    ; SI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+    ; SI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+    ; SI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C4]](s64)
+    ; SI-NEXT: G_STORE [[BITCAST5]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
     ; CI-LABEL: name: test_store_global_v11s16_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[DEF1:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<12 x s16>)
-    ; CI-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>), [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>), [[UV65:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<132 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[UV2]](<2 x s16>), [[UV3]](<2 x s16>), [[UV4]](<2 x s16>), [[UV5]](<2 x s16>), [[UV6]](<2 x s16>), [[UV7]](<2 x s16>), [[UV8]](<2 x s16>), [[UV9]](<2 x s16>), [[UV10]](<2 x s16>), [[UV11]](<2 x s16>), [[UV12]](<2 x s16>), [[UV13]](<2 x s16>), [[UV14]](<2 x s16>), [[UV15]](<2 x s16>), [[UV16]](<2 x s16>), [[UV17]](<2 x s16>), [[UV18]](<2 x s16>), [[UV19]](<2 x s16>), [[UV20]](<2 x s16>), [[UV21]](<2 x s16>), [[UV22]](<2 x s16>), [[UV23]](<2 x s16>), [[UV24]](<2 x s16>), [[UV25]](<2 x s16>), [[UV26]](<2 x s16>), [[UV27]](<2 x s16>), [[UV28]](<2 x s16>), [[UV29]](<2 x s16>), [[UV30]](<2 x s16>), [[UV31]](<2 x s16>), [[UV32]](<2 x s16>), [[UV33]](<2 x s16>), [[UV34]](<2 x s16>), [[UV35]](<2 x s16>), [[UV36]](<2 x s16>), [[UV37]](<2 x s16>), [[UV38]](<2 x s16>), [[UV39]](<2 x s16>), [[UV40]](<2 x s16>), [[UV41]](<2 x s16>), [[UV42]](<2 x s16>), [[UV43]](<2 x s16>), [[UV44]](<2 x s16>), [[UV45]](<2 x s16>), [[UV46]](<2 x s16>), [[UV47]](<2 x s16>), [[UV48]](<2 x s16>), [[UV49]](<2 x s16>), [[UV50]](<2 x s16>), [[UV51]](<2 x s16>), [[UV52]](<2 x s16>), [[UV53]](<2 x s16>), [[UV54]](<2 x s16>), [[UV55]](<2 x s16>), [[UV56]](<2 x s16>), [[UV57]](<2 x s16>), [[UV58]](<2 x s16>), [[UV59]](<2 x s16>), [[UV60]](<2 x s16>), [[UV61]](<2 x s16>), [[UV62]](<2 x s16>), [[UV63]](<2 x s16>), [[UV64]](<2 x s16>), [[UV65]](<2 x s16>)
-    ; CI-NEXT: [[UV66:%[0-9]+]]:_(<11 x s16>), [[UV67:%[0-9]+]]:_(<11 x s16>), [[UV68:%[0-9]+]]:_(<11 x s16>), [[UV69:%[0-9]+]]:_(<11 x s16>), [[UV70:%[0-9]+]]:_(<11 x s16>), [[UV71:%[0-9]+]]:_(<11 x s16>), [[UV72:%[0-9]+]]:_(<11 x s16>), [[UV73:%[0-9]+]]:_(<11 x s16>), [[UV74:%[0-9]+]]:_(<11 x s16>), [[UV75:%[0-9]+]]:_(<11 x s16>), [[UV76:%[0-9]+]]:_(<11 x s16>), [[UV77:%[0-9]+]]:_(<11 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<132 x s16>)
-    ; CI-NEXT: [[INSERT:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<8 x s16>) = G_EXTRACT [[INSERT]](<12 x s16>), 0
-    ; CI-NEXT: [[INSERT1:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<12 x s16>), 128
-    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[EXTRACT]](<8 x s16>)
-    ; CI-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
-    ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
-    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CI-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0
-    ; CI-NEXT: [[UV78:%[0-9]+]]:_(<2 x s16>), [[UV79:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>)
-    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV78]](<2 x s16>)
-    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
-    ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV79]](<2 x s16>)
-    ; CI-NEXT: G_STORE [[BITCAST1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
-    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
-    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
-    ; CI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
-    ; CI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
-    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
-    ; CI-NEXT: G_STORE [[BITCAST2]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+    ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; CI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; CI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; CI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; CI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; CI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; CI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; CI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; CI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; CI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; CI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; CI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; CI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; CI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; CI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; CI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
+    ; CI-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>)
+    ; CI-NEXT: [[BITCAST10:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+    ; CI-NEXT: G_STORE [[BITCAST10]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+    ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
+    ; CI-NEXT: G_STORE [[BITCAST4]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+    ; CI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+    ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
+    ; CI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+    ; CI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+    ; CI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C4]](s64)
+    ; CI-NEXT: G_STORE [[BITCAST5]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
     ; VI-LABEL: name: test_store_global_v11s16_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<12 x s16>)
-    ; VI-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>), [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>), [[UV65:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<132 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[UV2]](<2 x s16>), [[UV3]](<2 x s16>), [[UV4]](<2 x s16>), [[UV5]](<2 x s16>), [[UV6]](<2 x s16>), [[UV7]](<2 x s16>), [[UV8]](<2 x s16>), [[UV9]](<2 x s16>), [[UV10]](<2 x s16>), [[UV11]](<2 x s16>), [[UV12]](<2 x s16>), [[UV13]](<2 x s16>), [[UV14]](<2 x s16>), [[UV15]](<2 x s16>), [[UV16]](<2 x s16>), [[UV17]](<2 x s16>), [[UV18]](<2 x s16>), [[UV19]](<2 x s16>), [[UV20]](<2 x s16>), [[UV21]](<2 x s16>), [[UV22]](<2 x s16>), [[UV23]](<2 x s16>), [[UV24]](<2 x s16>), [[UV25]](<2 x s16>), [[UV26]](<2 x s16>), [[UV27]](<2 x s16>), [[UV28]](<2 x s16>), [[UV29]](<2 x s16>), [[UV30]](<2 x s16>), [[UV31]](<2 x s16>), [[UV32]](<2 x s16>), [[UV33]](<2 x s16>), [[UV34]](<2 x s16>), [[UV35]](<2 x s16>), [[UV36]](<2 x s16>), [[UV37]](<2 x s16>), [[UV38]](<2 x s16>), [[UV39]](<2 x s16>), [[UV40]](<2 x s16>), [[UV41]](<2 x s16>), [[UV42]](<2 x s16>), [[UV43]](<2 x s16>), [[UV44]](<2 x s16>), [[UV45]](<2 x s16>), [[UV46]](<2 x s16>), [[UV47]](<2 x s16>), [[UV48]](<2 x s16>), [[UV49]](<2 x s16>), [[UV50]](<2 x s16>), [[UV51]](<2 x s16>), [[UV52]](<2 x s16>), [[UV53]](<2 x s16>), [[UV54]](<2 x s16>), [[UV55]](<2 x s16>), [[UV56]](<2 x s16>), [[UV57]](<2 x s16>), [[UV58]](<2 x s16>), [[UV59]](<2 x s16>), [[UV60]](<2 x s16>), [[UV61]](<2 x s16>), [[UV62]](<2 x s16>), [[UV63]](<2 x s16>), [[UV64]](<2 x s16>), [[UV65]](<2 x s16>)
-    ; VI-NEXT: [[UV66:%[0-9]+]]:_(<11 x s16>), [[UV67:%[0-9]+]]:_(<11 x s16>), [[UV68:%[0-9]+]]:_(<11 x s16>), [[UV69:%[0-9]+]]:_(<11 x s16>), [[UV70:%[0-9]+]]:_(<11 x s16>), [[UV71:%[0-9]+]]:_(<11 x s16>), [[UV72:%[0-9]+]]:_(<11 x s16>), [[UV73:%[0-9]+]]:_(<11 x s16>), [[UV74:%[0-9]+]]:_(<11 x s16>), [[UV75:%[0-9]+]]:_(<11 x s16>), [[UV76:%[0-9]+]]:_(<11 x s16>), [[UV77:%[0-9]+]]:_(<11 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<132 x s16>)
-    ; VI-NEXT: [[INSERT:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<8 x s16>) = G_EXTRACT [[INSERT]](<12 x s16>), 0
-    ; VI-NEXT: [[INSERT1:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<12 x s16>), 128
-    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[EXTRACT]](<8 x s16>)
-    ; VI-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
-    ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
-    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0
-    ; VI-NEXT: [[UV78:%[0-9]+]]:_(<2 x s16>), [[UV79:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>)
-    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV78]](<2 x s16>)
-    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
-    ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV79]](<2 x s16>)
-    ; VI-NEXT: G_STORE [[BITCAST1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
-    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
-    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
-    ; VI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
-    ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
-    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
-    ; VI-NEXT: G_STORE [[BITCAST2]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+    ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; VI-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; VI-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; VI-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; VI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
+    ; VI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
+    ; VI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
+    ; VI-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
+    ; VI-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
+    ; VI-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; VI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
+    ; VI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; VI-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; VI-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
+    ; VI-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; VI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; VI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; VI-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
+    ; VI-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; VI-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; VI-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; VI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; VI-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; VI-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL3]]
+    ; VI-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>)
+    ; VI-NEXT: [[BITCAST10:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+    ; VI-NEXT: G_STORE [[BITCAST10]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+    ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C2]](s64)
+    ; VI-NEXT: G_STORE [[BITCAST4]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+    ; VI-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
+    ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
+    ; VI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+    ; VI-NEXT: [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
+    ; VI-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C4]](s64)
+    ; VI-NEXT: G_STORE [[BITCAST5]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v11s16_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<12 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>), [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>), [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>), [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>), [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>), [[UV23:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV24:%[0-9]+]]:_(<2 x s16>), [[UV25:%[0-9]+]]:_(<2 x s16>), [[UV26:%[0-9]+]]:_(<2 x s16>), [[UV27:%[0-9]+]]:_(<2 x s16>), [[UV28:%[0-9]+]]:_(<2 x s16>), [[UV29:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV30:%[0-9]+]]:_(<2 x s16>), [[UV31:%[0-9]+]]:_(<2 x s16>), [[UV32:%[0-9]+]]:_(<2 x s16>), [[UV33:%[0-9]+]]:_(<2 x s16>), [[UV34:%[0-9]+]]:_(<2 x s16>), [[UV35:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV36:%[0-9]+]]:_(<2 x s16>), [[UV37:%[0-9]+]]:_(<2 x s16>), [[UV38:%[0-9]+]]:_(<2 x s16>), [[UV39:%[0-9]+]]:_(<2 x s16>), [[UV40:%[0-9]+]]:_(<2 x s16>), [[UV41:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV42:%[0-9]+]]:_(<2 x s16>), [[UV43:%[0-9]+]]:_(<2 x s16>), [[UV44:%[0-9]+]]:_(<2 x s16>), [[UV45:%[0-9]+]]:_(<2 x s16>), [[UV46:%[0-9]+]]:_(<2 x s16>), [[UV47:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV48:%[0-9]+]]:_(<2 x s16>), [[UV49:%[0-9]+]]:_(<2 x s16>), [[UV50:%[0-9]+]]:_(<2 x s16>), [[UV51:%[0-9]+]]:_(<2 x s16>), [[UV52:%[0-9]+]]:_(<2 x s16>), [[UV53:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV54:%[0-9]+]]:_(<2 x s16>), [[UV55:%[0-9]+]]:_(<2 x s16>), [[UV56:%[0-9]+]]:_(<2 x s16>), [[UV57:%[0-9]+]]:_(<2 x s16>), [[UV58:%[0-9]+]]:_(<2 x s16>), [[UV59:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[UV60:%[0-9]+]]:_(<2 x s16>), [[UV61:%[0-9]+]]:_(<2 x s16>), [[UV62:%[0-9]+]]:_(<2 x s16>), [[UV63:%[0-9]+]]:_(<2 x s16>), [[UV64:%[0-9]+]]:_(<2 x s16>), [[UV65:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<12 x s16>)
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<132 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[UV1]](<2 x s16>), [[UV2]](<2 x s16>), [[UV3]](<2 x s16>), [[UV4]](<2 x s16>), [[UV5]](<2 x s16>), [[UV6]](<2 x s16>), [[UV7]](<2 x s16>), [[UV8]](<2 x s16>), [[UV9]](<2 x s16>), [[UV10]](<2 x s16>), [[UV11]](<2 x s16>), [[UV12]](<2 x s16>), [[UV13]](<2 x s16>), [[UV14]](<2 x s16>), [[UV15]](<2 x s16>), [[UV16]](<2 x s16>), [[UV17]](<2 x s16>), [[UV18]](<2 x s16>), [[UV19]](<2 x s16>), [[UV20]](<2 x s16>), [[UV21]](<2 x s16>), [[UV22]](<2 x s16>), [[UV23]](<2 x s16>), [[UV24]](<2 x s16>), [[UV25]](<2 x s16>), [[UV26]](<2 x s16>), [[UV27]](<2 x s16>), [[UV28]](<2 x s16>), [[UV29]](<2 x s16>), [[UV30]](<2 x s16>), [[UV31]](<2 x s16>), [[UV32]](<2 x s16>), [[UV33]](<2 x s16>), [[UV34]](<2 x s16>), [[UV35]](<2 x s16>), [[UV36]](<2 x s16>), [[UV37]](<2 x s16>), [[UV38]](<2 x s16>), [[UV39]](<2 x s16>), [[UV40]](<2 x s16>), [[UV41]](<2 x s16>), [[UV42]](<2 x s16>), [[UV43]](<2 x s16>), [[UV44]](<2 x s16>), [[UV45]](<2 x s16>), [[UV46]](<2 x s16>), [[UV47]](<2 x s16>), [[UV48]](<2 x s16>), [[UV49]](<2 x s16>), [[UV50]](<2 x s16>), [[UV51]](<2 x s16>), [[UV52]](<2 x s16>), [[UV53]](<2 x s16>), [[UV54]](<2 x s16>), [[UV55]](<2 x s16>), [[UV56]](<2 x s16>), [[UV57]](<2 x s16>), [[UV58]](<2 x s16>), [[UV59]](<2 x s16>), [[UV60]](<2 x s16>), [[UV61]](<2 x s16>), [[UV62]](<2 x s16>), [[UV63]](<2 x s16>), [[UV64]](<2 x s16>), [[UV65]](<2 x s16>)
-    ; GFX9-NEXT: [[UV66:%[0-9]+]]:_(<11 x s16>), [[UV67:%[0-9]+]]:_(<11 x s16>), [[UV68:%[0-9]+]]:_(<11 x s16>), [[UV69:%[0-9]+]]:_(<11 x s16>), [[UV70:%[0-9]+]]:_(<11 x s16>), [[UV71:%[0-9]+]]:_(<11 x s16>), [[UV72:%[0-9]+]]:_(<11 x s16>), [[UV73:%[0-9]+]]:_(<11 x s16>), [[UV74:%[0-9]+]]:_(<11 x s16>), [[UV75:%[0-9]+]]:_(<11 x s16>), [[UV76:%[0-9]+]]:_(<11 x s16>), [[UV77:%[0-9]+]]:_(<11 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<132 x s16>)
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<8 x s16>) = G_EXTRACT [[INSERT]](<12 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<12 x s16>) = G_INSERT [[DEF1]], [[UV66]](<11 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[INSERT1]](<12 x s16>), 128
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[EXTRACT]](<8 x s16>)
-    ; GFX9-NEXT: G_STORE [[BITCAST]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
-    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
-    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[EXTRACT1]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV78:%[0-9]+]]:_(<2 x s16>), [[UV79:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT2]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV78]](<2 x s16>)
-    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C1]](s32)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV79]](<2 x s16>)
-    ; GFX9-NEXT: G_STORE [[BITCAST1]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
+    ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR3]](s32)
+    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(<4 x s32>) = G_BITCAST [[CONCAT_VECTORS]](<8 x s16>)
+    ; GFX9-NEXT: G_STORE [[BITCAST6]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+    ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
+    ; GFX9-NEXT: G_STORE [[BITCAST4]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 16, align 16, addrspace 1)
     ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C2]](s64)
-    ; GFX9-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD1]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
     ; GFX9-NEXT: [[PTR_ADD2:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD]], [[C3]](s64)
-    ; GFX9-NEXT: G_STORE [[BITCAST2]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BITCAST5]](s32), [[PTR_ADD2]](p1) :: (store (s16) into unknown-address + 20, align 4, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<11 x s16>) = G_IMPLICIT_DEF
     G_STORE %1, %0 :: (store (<11 x s16>), align 16, addrspace 1)
 ...
 
 ---
 name: test_store_global_v12s16_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_v12s16_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v12s16_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v12s16_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v12s16_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<12 x s16>) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<6 x s32>) = G_BITCAST [[DEF]](<12 x s16>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<6 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<6 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR1]](<2 x s32>), [[PTR_ADD]](p1) :: (store (<2 x s32>) into unknown-address + 16, align 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<12 x s16>) = G_IMPLICIT_DEF
     G_STORE %1, %0 :: (store (<12 x s16>), align 16, addrspace 1)
 ...
 
 ---
 name: test_store_global_s160_align1
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_s160_align1
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY2]], [[C3]]

     ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
     ; SI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[AND7]], [[COPY12]](s32)
     ; SI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR11]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; SI-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C8]](s64)
-    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY13:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY13]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; SI-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY13]], [[C3]]
     ; SI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[AND8]], [[COPY14]](s32)
     ; SI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY13]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR13]](s32), [[PTR_ADD17]](p1) :: (store (s8) into unknown-address + 17, addrspace 1)
     ; SI-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[LSHR12]], [[C3]]
     ; SI-NEXT: [[LSHR14:%[0-9]+]]:_(s32) = G_LSHR [[AND9]], [[COPY15]](s32)
     ; SI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD16]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR12]](s32), [[PTR_ADD16]](p1) :: (store (s8) into unknown-address + 18, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR14]](s32), [[PTR_ADD18]](p1) :: (store (s8) into unknown-address + 19, addrspace 1)
     ; CI-LABEL: name: test_store_global_s160_align1
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
     ; VI-LABEL: name: test_store_global_s160_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[C2]](s16)

     ; VI-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
     ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC7]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR11]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT7]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; VI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C7]](s64)
-    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[COPY6]](s32)
     ; VI-NEXT: [[LSHR13:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC8]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: [[ANYEXT8:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR13]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT8]](s32), [[PTR_ADD17]](p1) :: (store (s8) into unknown-address + 17, addrspace 1)
     ; VI-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR12]](s32)
     ; VI-NEXT: [[LSHR14:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC9]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD18:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD16]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR12]](s32), [[PTR_ADD16]](p1) :: (store (s8) into unknown-address + 18, addrspace 1)
     ; VI-NEXT: [[ANYEXT9:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR14]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT9]](s32), [[PTR_ADD18]](p1) :: (store (s8) into unknown-address + 19, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_s160_align1
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 1, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (s160), align 1, addrspace 1)
 ...
 
 ---
 name: test_store_global_s160_align2
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_s160_align2
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; SI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY5]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY5]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; SI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; CI-LABEL: name: test_store_global_s160_align2
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
     ; VI-LABEL: name: test_store_global_s160_align2
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY2]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY2]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY5]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY5]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[EXTRACT1]](s32)
+    ; VI-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY6]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY6]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_s160_align2
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 2, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (s160), align 2, addrspace 1)
 ...
 
 ---
 name: test_store_global_s160_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_s160_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_s160_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_s160_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_s160_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (s160), align 4, addrspace 1)
 ...
 
 ---
 name: test_store_global_s160_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_s160_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_s160_align8
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_s160_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_s160_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 8, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (s160), align 8, addrspace 1)
 ...
 
 ---
 name: test_store_global_s160_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
 
     ; SI-LABEL: name: test_store_global_s160_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_s160_align16
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; CI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_s160_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_s160_align16
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(<5 x s32>) = G_BITCAST [[COPY1]](s160)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[BITCAST]](<5 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<5 x s32>), 128
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<5 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV4]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 16, align 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(s160) = COPY $vgpr2_vgpr3_vgpr4_vgpr5_vgpr6
     G_STORE %1, %0 :: (store (s160), align 16, addrspace 1)
 ...
 
 ---
 name: test_store_global_v8s32_align1
 body: |

   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_store_global_v9s32_align1
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; SI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; SI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
     ; SI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY4]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
     ; SI-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[COPY4]], [[C3]]

     ; SI-NEXT: [[COPY14:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
     ; SI-NEXT: [[LSHR11:%[0-9]+]]:_(s32) = G_LSHR [[AND7]], [[COPY14]](s32)
     ; SI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR11]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; SI-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C8]](s64)
-    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT1]](<4 x s32>)
     ; SI-NEXT: [[COPY15:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY15]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; SI-NEXT: [[COPY16:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[COPY15]], [[C3]]
     ; SI-NEXT: [[LSHR13:%[0-9]+]]:_(s32) = G_LSHR [[AND8]], [[COPY16]](s32)
     ; SI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY15]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)

     ; SI-NEXT: [[COPY26:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR21]], [[C3]]
     ; SI-NEXT: [[LSHR23:%[0-9]+]]:_(s32) = G_LSHR [[AND15]], [[COPY26]](s32)
     ; SI-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD28]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR21]](s32), [[PTR_ADD28]](p1) :: (store (s8) into unknown-address + 30, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR23]](s32), [[PTR_ADD30]](p1) :: (store (s8) into unknown-address + 31, addrspace 1)
     ; SI-NEXT: [[C9:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; SI-NEXT: [[PTR_ADD31:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C9]](s64)
-    ; SI-NEXT: [[COPY27:%[0-9]+]]:_(s32) = COPY [[EXTRACT2]](s32)
+    ; SI-NEXT: [[COPY27:%[0-9]+]]:_(s32) = COPY [[UV8]](s32)
     ; SI-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[COPY27]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD32:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD31]], [[C1]](s64)
     ; SI-NEXT: [[COPY28:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[COPY27]], [[C3]]
     ; SI-NEXT: [[LSHR25:%[0-9]+]]:_(s32) = G_LSHR [[AND16]], [[COPY28]](s32)
     ; SI-NEXT: [[PTR_ADD33:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD31]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[COPY27]](s32), [[PTR_ADD31]](p1) :: (store (s8) into unknown-address + 32, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR25]](s32), [[PTR_ADD33]](p1) :: (store (s8) into unknown-address + 33, addrspace 1)
     ; SI-NEXT: [[COPY29:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
     ; SI-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR24]], [[C3]]
     ; SI-NEXT: [[LSHR26:%[0-9]+]]:_(s32) = G_LSHR [[AND17]], [[COPY29]](s32)
     ; SI-NEXT: [[PTR_ADD34:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD32]], [[C4]](s64)
     ; SI-NEXT: G_STORE [[LSHR24]](s32), [[PTR_ADD32]](p1) :: (store (s8) into unknown-address + 34, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR26]](s32), [[PTR_ADD34]](p1) :: (store (s8) into unknown-address + 35, addrspace 1)
     ; CI-LABEL: name: test_store_global_v9s32_align1
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; CI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; CI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; CI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
     ; VI-LABEL: name: test_store_global_v9s32_align1
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; VI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; VI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; VI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
     ; VI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY4]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY4]](s32)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC]], [[C2]](s16)

     ; VI-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR9]](s32)
     ; VI-NEXT: [[LSHR11:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC7]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD12]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR9]](s32), [[PTR_ADD12]](p1) :: (store (s8) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR11]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT7]](s32), [[PTR_ADD14]](p1) :: (store (s8) into unknown-address + 15, addrspace 1)
     ; VI-NEXT: [[C7:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C7]](s64)
-    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT1]](<4 x s32>)
     ; VI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR12:%[0-9]+]]:_(s32) = G_LSHR [[COPY8]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[COPY8]](s32)
     ; VI-NEXT: [[LSHR13:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC8]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD17:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[COPY8]](s32), [[PTR_ADD15]](p1) :: (store (s8) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: [[ANYEXT8:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR13]](s16)

     ; VI-NEXT: [[TRUNC15:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR21]](s32)
     ; VI-NEXT: [[LSHR23:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC15]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD30:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD28]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR21]](s32), [[PTR_ADD28]](p1) :: (store (s8) into unknown-address + 30, addrspace 1)
     ; VI-NEXT: [[ANYEXT15:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR23]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT15]](s32), [[PTR_ADD30]](p1) :: (store (s8) into unknown-address + 31, addrspace 1)
     ; VI-NEXT: [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; VI-NEXT: [[PTR_ADD31:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C8]](s64)
-    ; VI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[EXTRACT2]](s32)
+    ; VI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[UV8]](s32)
     ; VI-NEXT: [[LSHR24:%[0-9]+]]:_(s32) = G_LSHR [[COPY12]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD32:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD31]], [[C1]](s64)
     ; VI-NEXT: [[TRUNC16:%[0-9]+]]:_(s16) = G_TRUNC [[COPY12]](s32)
     ; VI-NEXT: [[LSHR25:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC16]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD33:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD31]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[COPY12]](s32), [[PTR_ADD31]](p1) :: (store (s8) into unknown-address + 32, addrspace 1)
     ; VI-NEXT: [[ANYEXT16:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR25]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT16]](s32), [[PTR_ADD33]](p1) :: (store (s8) into unknown-address + 33, addrspace 1)
     ; VI-NEXT: [[TRUNC17:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR24]](s32)
     ; VI-NEXT: [[LSHR26:%[0-9]+]]:_(s16) = G_LSHR [[TRUNC17]], [[C2]](s16)
     ; VI-NEXT: [[PTR_ADD34:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD32]], [[C3]](s64)
     ; VI-NEXT: G_STORE [[LSHR24]](s32), [[PTR_ADD32]](p1) :: (store (s8) into unknown-address + 34, addrspace 1)
     ; VI-NEXT: [[ANYEXT17:%[0-9]+]]:_(s32) = G_ANYEXT [[LSHR26]](s16)
     ; VI-NEXT: G_STORE [[ANYEXT17]](s32), [[PTR_ADD34]](p1) :: (store (s8) into unknown-address + 35, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v9s32_align1
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 1, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 1, addrspace 1)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 1, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     %2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     %3:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
     %4:_(<9 x s32>) = G_CONCAT_VECTORS %1, %2, %3
     G_STORE %4, %0 :: (store (<9 x s32>), align 1, addrspace 1)
 ...
 
 ---
 name: test_store_global_v9s32_align2
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_store_global_v9s32_align2
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; SI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; SI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
     ; SI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY4]], [[C]](s32)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY4]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; SI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; SI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; SI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; SI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY7]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY7]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; SI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; SI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT1]](<4 x s32>)
     ; SI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; SI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY8]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY8]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; SI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
     ; SI-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY [[UV5]](s32)
     ; SI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[COPY9]], [[C]](s32)

     ; SI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
     ; SI-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[UV7]](s32)
     ; SI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[COPY11]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD13]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY11]](s32), [[PTR_ADD13]](p1) :: (store (s16) into unknown-address + 28, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR7]](s32), [[PTR_ADD14]](p1) :: (store (s16) into unknown-address + 30, addrspace 1)
     ; SI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; SI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C6]](s64)
-    ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[EXTRACT2]](s32)
+    ; SI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[UV8]](s32)
     ; SI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[COPY12]], [[C]](s32)
     ; SI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; SI-NEXT: G_STORE [[COPY12]](s32), [[PTR_ADD15]](p1) :: (store (s16) into unknown-address + 32, addrspace 1)
     ; SI-NEXT: G_STORE [[LSHR8]](s32), [[PTR_ADD16]](p1) :: (store (s16) into unknown-address + 34, addrspace 1)
     ; CI-LABEL: name: test_store_global_v9s32_align2
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; CI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; CI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; CI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
     ; VI-LABEL: name: test_store_global_v9s32_align2
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; VI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; VI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT]](<4 x s32>)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; VI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
     ; VI-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[COPY4]], [[C]](s32)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 2
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY4]](s32), [[COPY]](p1) :: (store (s16), addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR]](s32), [[PTR_ADD]](p1) :: (store (s16) into unknown-address + 2, addrspace 1)
     ; VI-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4

     ; VI-NEXT: [[PTR_ADD5:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C4]](s64)
     ; VI-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[UV3]](s32)
     ; VI-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[COPY7]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD6:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD5]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY7]](s32), [[PTR_ADD5]](p1) :: (store (s16) into unknown-address + 12, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR3]](s32), [[PTR_ADD6]](p1) :: (store (s16) into unknown-address + 14, addrspace 1)
     ; VI-NEXT: [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD7:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C5]](s64)
-    ; VI-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[EXTRACT1]](<4 x s32>)
     ; VI-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[UV4]](s32)
     ; VI-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[COPY8]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD8:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY8]](s32), [[PTR_ADD7]](p1) :: (store (s16) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR4]](s32), [[PTR_ADD8]](p1) :: (store (s16) into unknown-address + 18, addrspace 1)
     ; VI-NEXT: [[PTR_ADD9:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C2]](s64)
     ; VI-NEXT: [[COPY9:%[0-9]+]]:_(s32) = COPY [[UV5]](s32)
     ; VI-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[COPY9]], [[C]](s32)

     ; VI-NEXT: [[PTR_ADD13:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD7]], [[C4]](s64)
     ; VI-NEXT: [[COPY11:%[0-9]+]]:_(s32) = COPY [[UV7]](s32)
     ; VI-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[COPY11]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD14:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD13]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY11]](s32), [[PTR_ADD13]](p1) :: (store (s16) into unknown-address + 28, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR7]](s32), [[PTR_ADD14]](p1) :: (store (s16) into unknown-address + 30, addrspace 1)
     ; VI-NEXT: [[C6:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; VI-NEXT: [[PTR_ADD15:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C6]](s64)
-    ; VI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[EXTRACT2]](s32)
+    ; VI-NEXT: [[COPY12:%[0-9]+]]:_(s32) = COPY [[UV8]](s32)
     ; VI-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[COPY12]], [[C]](s32)
     ; VI-NEXT: [[PTR_ADD16:%[0-9]+]]:_(p1) = G_PTR_ADD [[PTR_ADD15]], [[C1]](s64)
     ; VI-NEXT: G_STORE [[COPY12]](s32), [[PTR_ADD15]](p1) :: (store (s16) into unknown-address + 32, addrspace 1)
     ; VI-NEXT: G_STORE [[LSHR8]](s32), [[PTR_ADD16]](p1) :: (store (s16) into unknown-address + 34, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v9s32_align2
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 2, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 2, addrspace 1)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 2, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     %2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     %3:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
     %4:_(<9 x s32>) = G_CONCAT_VECTORS %1, %2, %3
     G_STORE %4, %0 :: (store (<9 x s32>), align 2, addrspace 1)
 ...
 
 ---
 name: test_store_global_v9s32_align4
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_store_global_v9s32_align4
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; SI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; SI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
     ; CI-LABEL: name: test_store_global_v9s32_align4
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; CI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; CI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; CI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
     ; VI-LABEL: name: test_store_global_v9s32_align4
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; VI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; VI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; VI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v9s32_align4
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 4, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 4, addrspace 1)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     %2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     %3:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
     %4:_(<9 x s32>) = G_CONCAT_VECTORS %1, %2, %3
     G_STORE %4, %0 :: (store (<9 x s32>), align 4, addrspace 1)
 ...
 
 ---
 name: test_store_global_v9s32_align8
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_store_global_v9s32_align8
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; SI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; SI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
     ; CI-LABEL: name: test_store_global_v9s32_align8
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; CI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; CI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; CI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_v9s32_align8
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; VI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; VI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; VI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v9s32_align8
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), align 8, addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, align 8, addrspace 1)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 8, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     %2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     %3:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
     %4:_(<9 x s32>) = G_CONCAT_VECTORS %1, %2, %3
     G_STORE %4, %0 :: (store (<9 x s32>), align 8, addrspace 1)
 ...
 
 ---
 name: test_store_global_v9s32_align16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4, $vgpr5_vgpr6_vgpr7, $vgpr8_vgpr9_vgpr10
 
     ; SI-LABEL: name: test_store_global_v9s32_align16
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; SI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; SI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; SI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; SI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; SI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; SI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; SI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
     ; CI-LABEL: name: test_store_global_v9s32_align16
     ; CI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; CI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; CI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; CI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; CI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; CI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; CI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; CI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; CI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; CI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; CI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; CI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; CI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; CI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; CI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; CI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
     ; CI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; CI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; CI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+    ; CI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
     ; VI-LABEL: name: test_store_global_v9s32_align16
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; VI-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; VI-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; VI-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; VI-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; VI-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; VI-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; VI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; VI-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; VI-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; VI-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; VI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
     ; VI-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; VI-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; VI-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+    ; VI-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
     ; GFX9-LABEL: name: test_store_global_v9s32_align16
     ; GFX9: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<9 x s32>) = G_CONCAT_VECTORS [[COPY1]](<3 x s32>), [[COPY2]](<3 x s32>), [[COPY3]](<3 x s32>)
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<4 x s32>) = G_EXTRACT [[CONCAT_VECTORS]](<9 x s32>), 128
-    ; GFX9-NEXT: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY3]](<3 x s32>), 64
-    ; GFX9-NEXT: G_STORE [[EXTRACT]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY2]](<3 x s32>)
+    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY3]](<3 x s32>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32), [[UV3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[UV5]](s32), [[UV6]](s32), [[UV7]](s32)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR]](<4 x s32>), [[COPY]](p1) :: (store (<4 x s32>), addrspace 1)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
     ; GFX9-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[BUILD_VECTOR1]](<4 x s32>), [[PTR_ADD]](p1) :: (store (<4 x s32>) into unknown-address + 16, addrspace 1)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 32
     ; GFX9-NEXT: [[PTR_ADD1:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C1]](s64)
-    ; GFX9-NEXT: G_STORE [[EXTRACT2]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
+    ; GFX9-NEXT: G_STORE [[UV8]](s32), [[PTR_ADD1]](p1) :: (store (s32) into unknown-address + 32, align 16, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     %2:_(<3 x s32>) = COPY $vgpr5_vgpr6_vgpr7
     %3:_(<3 x s32>) = COPY $vgpr7_vgpr8_vgpr9
     %4:_(<9 x s32>) = G_CONCAT_VECTORS %1, %2, %3
     G_STORE %4, %0 :: (store (<9 x s32>), align 16, addrspace 1)
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-store.mir

 name: test_store_global_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1, $vgpr2_vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_v3s32
     ; SI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[COPY1]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[COPY1]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY]](p1) :: (store (<2 x s32>), align 4, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_v3s32
     ; VI: [[COPY:%[0-9]+]]:_(p1) = COPY $vgpr0_vgpr1
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     ; VI-NEXT: G_STORE [[COPY1]](<3 x s32>), [[COPY]](p1) :: (store (<3 x s32>), align 4, addrspace 1)
     %0:_(p1) = COPY $vgpr0_vgpr1
     %1:_(<3 x s32>) = COPY $vgpr2_vgpr3_vgpr4
     G_STORE %1, %0 :: (store (<3 x s32>), align 4, addrspace 1)
 ...

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4
 
     ; SI-LABEL: name: test_store_global_96
     ; SI: [[COPY:%[0-9]+]]:_(s96) = COPY $vgpr0_vgpr1_vgpr2
     ; SI-NEXT: [[COPY1:%[0-9]+]]:_(p1) = COPY $vgpr3_vgpr4
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY]](s96)
-    ; SI-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[BITCAST]](<3 x s32>), 0
-    ; SI-NEXT: [[EXTRACT1:%[0-9]+]]:_(s32) = G_EXTRACT [[BITCAST]](<3 x s32>), 64
-    ; SI-NEXT: G_STORE [[EXTRACT]](<2 x s32>), [[COPY1]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
+    ; SI-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](<3 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
+    ; SI-NEXT: G_STORE [[BUILD_VECTOR]](<2 x s32>), [[COPY1]](p1) :: (store (<2 x s32>), align 16, addrspace 1)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[COPY1]], [[C]](s64)
-    ; SI-NEXT: G_STORE [[EXTRACT1]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
+    ; SI-NEXT: G_STORE [[UV2]](s32), [[PTR_ADD]](p1) :: (store (s32) into unknown-address + 8, align 8, addrspace 1)
     ; VI-LABEL: name: test_store_global_96
     ; VI: [[COPY:%[0-9]+]]:_(s96) = COPY $vgpr0_vgpr1_vgpr2
     ; VI-NEXT: [[COPY1:%[0-9]+]]:_(p1) = COPY $vgpr3_vgpr4
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(<3 x s32>) = G_BITCAST [[COPY]](s96)
     ; VI-NEXT: G_STORE [[BITCAST]](<3 x s32>), [[COPY1]](p1) :: (store (<3 x s32>), align 16, addrspace 1)
     %0:_(s96) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(p1) = COPY $vgpr3_vgpr4
 

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-uaddsat.mir

bb.0:
 
     ; GFX6-LABEL: name: uaddsat_v3s16
     ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32)
     ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 -1
     ; GFX6-NEXT: [[XOR:%[0-9]+]]:_(s32) = G_XOR [[SHL]], [[C1]]
     ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[XOR]], [[SHL1]]
     ; GFX6-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[SHL]], [[UMIN]]
     ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[ADD]], [[C]](s32)
     ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
-    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[XOR1:%[0-9]+]]:_(s32) = G_XOR [[SHL2]], [[C1]]
     ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[XOR1]], [[SHL3]]
     ; GFX6-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[SHL2]], [[UMIN1]]
     ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[ADD1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR2]], [[C]](s32)
     ; GFX6-NEXT: [[XOR2:%[0-9]+]]:_(s32) = G_XOR [[SHL4]], [[C1]]
     ; GFX6-NEXT: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[XOR2]], [[SHL5]]
     ; GFX6-NEXT: [[ADD2:%[0-9]+]]:_(s32) = G_ADD [[SHL4]], [[UMIN2]]
     ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[ADD2]], [[C]](s32)
     ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX6-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C2]]
     ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C2]]
     ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL6]]
-    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX6-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C2]]
-    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C2]]
+    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C2]]
     ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL7]]
-    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX6-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C2]]
-    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C2]]
+    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C2]]
     ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL8]]
-    ; GFX6-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX8-LABEL: name: uaddsat_v3s16
     ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX8-NEXT: [[UADDSAT:%[0-9]+]]:_(s16) = G_UADDSAT [[TRUNC]], [[TRUNC3]]
     ; GFX8-NEXT: [[UADDSAT1:%[0-9]+]]:_(s16) = G_UADDSAT [[TRUNC1]], [[TRUNC4]]
     ; GFX8-NEXT: [[UADDSAT2:%[0-9]+]]:_(s16) = G_UADDSAT [[TRUNC2]], [[TRUNC5]]
     ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[UADDSAT]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[UADDSAT1]](s16)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[UADDSAT2]](s16)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
-    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
-    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: uaddsat_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR2]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[UADDSAT:%[0-9]+]]:_(<2 x s16>) = G_UADDSAT [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9-NEXT: [[UADDSAT1:%[0-9]+]]:_(<2 x s16>) = G_UADDSAT [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UADDSAT]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UADDSAT1]](<2 x s16>)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UADDSAT]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UADDSAT1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR3]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST7]](s32)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST6]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %3:_(<3 x s16>) = G_UADDSAT %1, %2
     %4:_(<3 x s16>) = G_IMPLICIT_DEF
     %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4
     $vgpr0_vgpr1_vgpr2 = COPY %5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umax.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_umax_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_umax_s32

 ---
 name: test_umax_v3s16
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_umax_v3s16
     ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
     ; SI-NEXT: [[UMAX:%[0-9]+]]:_(s32) = G_UMAX [[AND]], [[AND1]]
     ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; SI-NEXT: [[UMAX1:%[0-9]+]]:_(s32) = G_UMAX [[AND2]], [[AND3]]
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; SI-NEXT: [[UMAX2:%[0-9]+]]:_(s32) = G_UMAX [[AND4]], [[AND5]]
     ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UMAX]](s32), [[UMAX1]](s32), [[UMAX2]](s32)
     ; SI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; VI-LABEL: name: test_umax_v3s16
     ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[UMAX:%[0-9]+]]:_(s16) = G_UMAX [[TRUNC]], [[TRUNC3]]
     ; VI-NEXT: [[UMAX1:%[0-9]+]]:_(s16) = G_UMAX [[TRUNC1]], [[TRUNC4]]
     ; VI-NEXT: [[UMAX2:%[0-9]+]]:_(s16) = G_UMAX [[TRUNC2]], [[TRUNC5]]
     ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UMAX]](s16)
     ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UMAX1]](s16)
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UMAX2]](s16)
     ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; VI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX9-LABEL: name: test_umax_v3s16
     ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; GFX9-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[UMAX:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[UV8]], [[UV10]]
-    ; GFX9-NEXT: [[UMAX1:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[UV9]], [[UV11]]
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UMAX]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UMAX1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[UMAX:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[UMAX1:%[0-9]+]]:_(<2 x s16>) = G_UMAX [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UMAX]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UMAX1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST4]](s32), [[LSHR2]](s32), [[BITCAST5]](s32)
     ; GFX9-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<3 x s16>) = G_IMPLICIT_DEF
     %2:_(<3 x s16>) = G_UMAX %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     S_NOP 0, implicit %3
 ...
 

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umin.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=SI  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=VI %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer -global-isel-abort=0 %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -run-pass=legalizer %s -o - | FileCheck -check-prefix=SI  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer %s -o - | FileCheck -check-prefix=VI %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -run-pass=legalizer %s -o - | FileCheck -check-prefix=GFX9  %s
 
 ---
 name: test_umin_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_umin_s32

 ---
 name: test_umin_v3s16
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; SI-LABEL: name: test_umin_v3s16
     ; SI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; SI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; SI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; SI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; SI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; SI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; SI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; SI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; SI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; SI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; SI-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; SI-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; SI-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
     ; SI-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[AND]], [[AND1]]
     ; SI-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; SI-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; SI-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[AND2]], [[AND3]]
     ; SI-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
     ; SI-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; SI-NEXT: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[AND4]], [[AND5]]
     ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UMIN]](s32), [[UMIN1]](s32), [[UMIN2]](s32)
     ; SI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; VI-LABEL: name: test_umin_v3s16
     ; VI: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
     ; VI-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; VI-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; VI-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; VI-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; VI-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; VI-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; VI-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; VI-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
     ; VI-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
     ; VI-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
     ; VI-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; VI-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; VI-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
     ; VI-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
     ; VI-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; VI-NEXT: [[UMIN:%[0-9]+]]:_(s16) = G_UMIN [[TRUNC]], [[TRUNC3]]
     ; VI-NEXT: [[UMIN1:%[0-9]+]]:_(s16) = G_UMIN [[TRUNC1]], [[TRUNC4]]
     ; VI-NEXT: [[UMIN2:%[0-9]+]]:_(s16) = G_UMIN [[TRUNC2]], [[TRUNC5]]
     ; VI-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UMIN]](s16)
     ; VI-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UMIN1]](s16)
     ; VI-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UMIN2]](s16)
     ; VI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[ANYEXT]](s32), [[ANYEXT1]](s32), [[ANYEXT2]](s32)
     ; VI-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     ; GFX9-LABEL: name: test_umin_v3s16
     ; GFX9: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>), [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s16>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<12 x s16>) = G_CONCAT_VECTORS [[DEF2]](<4 x s16>), [[DEF1]](<4 x s16>), [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[UV4:%[0-9]+]]:_(<3 x s16>), [[UV5:%[0-9]+]]:_(<3 x s16>), [[UV6:%[0-9]+]]:_(<3 x s16>), [[UV7:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<12 x s16>)
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV4]](<3 x s16>), 0
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
-    ; GFX9-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<4 x s16>)
-    ; GFX9-NEXT: [[UMIN:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[UV8]], [[UV10]]
-    ; GFX9-NEXT: [[UMIN1:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[UV9]], [[UV11]]
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UMIN]](<2 x s16>)
+    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UMIN1]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST]](s32), [[LSHR]](s32), [[BITCAST1]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF2]](s32)
+    ; GFX9-NEXT: [[UMIN:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[UMIN1:%[0-9]+]]:_(<2 x s16>) = G_UMIN [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UMIN]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UMIN1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[BITCAST4]](s32), [[LSHR2]](s32), [[BITCAST5]](s32)
     ; GFX9-NEXT: S_NOP 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s16>) = G_IMPLICIT_DEF
     %1:_(<3 x s16>) = G_IMPLICIT_DEF
     %2:_(<3 x s16>) = G_UMIN %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     S_NOP 0, implicit %3
 ...
 

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-umulh.mir

bb.0:
     ; GFX8-NEXT: $vgpr0 = COPY [[OR2]](s32)
     ; GFX9-LABEL: name: test_umulh_v3s8
     ; GFX9: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr3
     ; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr4
     ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr5
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY]](s32)
     ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[COPY1]](s32)
-    ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32)
-    ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
-    ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY3]](s32)
-    ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY4]](s32)
-    ; GFX9-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[COPY5]](s32)
-    ; GFX9-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+    ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY3]](s32)
+    ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[COPY4]](s32)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 255
     ; GFX9-NEXT: [[AND:%[0-9]+]]:_(s16) = G_AND [[TRUNC]], [[C]]
     ; GFX9-NEXT: [[AND1:%[0-9]+]]:_(s16) = G_AND [[TRUNC1]], [[C]]
     ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[AND]](s16)
     ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[AND1]](s16)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32)
-    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C]]
-    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C]]
+    ; GFX9-NEXT: [[AND2:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
+    ; GFX9-NEXT: [[AND3:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
     ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[AND2]](s16)
     ; GFX9-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[AND3]](s16)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[ANYEXT3]](s32)
     ; GFX9-NEXT: [[MUL:%[0-9]+]]:_(<2 x s16>) = G_MUL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC1]]
-    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
-    ; GFX9-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[C1]](s32)
+    ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
+    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
+    ; GFX9-NEXT: [[COPY6:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY6]](s32), [[C2]](s32)
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[MUL]], [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
-    ; GFX9-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC2]], [[C]]
-    ; GFX9-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC3]], [[C]]
-    ; GFX9-NEXT: [[ANYEXT4:%[0-9]+]]:_(s32) = G_ANYEXT [[AND4]](s16)
-    ; GFX9-NEXT: [[ANYEXT5:%[0-9]+]]:_(s32) = G_ANYEXT [[AND5]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT4]](s32), [[ANYEXT5]](s32)
+    ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32)
+    ; GFX9-NEXT: [[AND4:%[0-9]+]]:_(s16) = G_AND [[TRUNC4]], [[C]]
+    ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[COPY5]](s32)
+    ; GFX9-NEXT: [[AND5:%[0-9]+]]:_(s16) = G_AND [[TRUNC5]], [[C]]
+    ; GFX9-NEXT: [[MUL1:%[0-9]+]]:_(s16) = G_MUL [[AND4]], [[AND5]]
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s16) = G_LSHR [[MUL1]], [[C1]](s16)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR]](<2 x s16>)
+    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C3]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX9-NEXT: [[AND6:%[0-9]+]]:_(s16) = G_AND [[TRUNC6]], [[C]]
+    ; GFX9-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX9-NEXT: [[AND7:%[0-9]+]]:_(s16) = G_AND [[TRUNC7]], [[C]]
-    ; GFX9-NEXT: [[ANYEXT6:%[0-9]+]]:_(s32) = G_ANYEXT [[AND6]](s16)
-    ; GFX9-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[AND7]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT6]](s32), [[ANYEXT7]](s32)
-    ; GFX9-NEXT: [[MUL1:%[0-9]+]]:_(<2 x s16>) = G_MUL [[BUILD_VECTOR_TRUNC3]], [[BUILD_VECTOR_TRUNC4]]
-    ; GFX9-NEXT: [[COPY7:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; GFX9-NEXT: [[COPY8:%[0-9]+]]:_(s32) = COPY [[C1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[COPY7]](s32), [[COPY8]](s32)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(<2 x s16>) = G_LSHR [[MUL1]], [[BUILD_VECTOR_TRUNC5]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR]](<2 x s16>)
-    ; GFX9-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C2]](s32)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[LSHR1]](<2 x s16>)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
-    ; GFX9-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C]]
-    ; GFX9-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
-    ; GFX9-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC9]], [[C]]
-    ; GFX9-NEXT: [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 8
-    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND9]], [[C3]](s16)
-    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND8]], [[SHL]]
-    ; GFX9-NEXT: [[TRUNC10:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX9-NEXT: [[AND10:%[0-9]+]]:_(s16) = G_AND [[TRUNC10]], [[C]]
-    ; GFX9-NEXT: [[TRUNC11:%[0-9]+]]:_(s16) = G_TRUNC [[DEF1]](s32)
-    ; GFX9-NEXT: [[AND11:%[0-9]+]]:_(s16) = G_AND [[TRUNC11]], [[C]]
-    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND11]], [[C3]](s16)
-    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND10]], [[SHL1]]
+    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[AND7]], [[C1]](s16)
+    ; GFX9-NEXT: [[OR:%[0-9]+]]:_(s16) = G_OR [[AND6]], [[SHL]]
+    ; GFX9-NEXT: [[AND8:%[0-9]+]]:_(s16) = G_AND [[LSHR1]], [[C]]
+    ; GFX9-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[DEF]](s32)
+    ; GFX9-NEXT: [[AND9:%[0-9]+]]:_(s16) = G_AND [[TRUNC8]], [[C]]
+    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[AND9]], [[C1]](s16)
+    ; GFX9-NEXT: [[OR1:%[0-9]+]]:_(s16) = G_OR [[AND8]], [[SHL1]]
     ; GFX9-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[OR]](s16)
     ; GFX9-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[OR1]](s16)
-    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C2]](s32)
+    ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C3]](s32)
     ; GFX9-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL2]]
     ; GFX9-NEXT: $vgpr0 = COPY [[OR2]](s32)
     %0:_(s32) = COPY $vgpr0
     %1:_(s32) = COPY $vgpr1
     %2:_(s32) = COPY $vgpr2
     %3:_(s32) = COPY $vgpr3
     %4:_(s32) = COPY $vgpr4
     %20:_(s32) = COPY $vgpr5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ushlsat.mir

bb.0:
 
     ; GFX6-LABEL: name: ushlsat_v3s16
     ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
     ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[SHL]], [[AND]](s32)
     ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[SHL1]], [[AND]](s32)
     ; GFX6-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 -1
     ; GFX6-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[SHL]](s32), [[LSHR3]]
     ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s32) = G_SELECT [[ICMP]](s1), [[C2]], [[SHL1]]
     ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[SELECT]], [[C]](s32)
-    ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
     ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
     ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[SHL2]], [[AND1]](s32)
     ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[SHL3]], [[AND1]](s32)
     ; GFX6-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[SHL2]](s32), [[LSHR5]]
     ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s32) = G_SELECT [[ICMP1]](s1), [[C2]], [[SHL3]]
     ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[SELECT1]], [[C]](s32)
     ; GFX6-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
     ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[SHL4]], [[AND2]](s32)
     ; GFX6-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[SHL5]], [[AND2]](s32)
     ; GFX6-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[SHL4]](s32), [[LSHR7]]
     ; GFX6-NEXT: [[SELECT2:%[0-9]+]]:_(s32) = G_SELECT [[ICMP2]](s1), [[C2]], [[SHL5]]
     ; GFX6-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[SELECT2]], [[C]](s32)
     ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
     ; GFX6-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C1]]
     ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
     ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL6]]
-    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C1]]
-    ; GFX6-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; GFX6-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND6]], [[C]](s32)
     ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL7]]
-    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX6-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C1]]
-    ; GFX6-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; GFX6-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND8]], [[C]](s32)
     ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND7]], [[SHL8]]
-    ; GFX6-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX8-LABEL: name: ushlsat_v3s16
     ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16)
     ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s16) = G_LSHR [[SHL]], [[TRUNC3]](s16)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 -1
     ; GFX8-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC]](s16), [[LSHR3]]
     ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[C1]], [[SHL]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[TRUNC4]](s16)
     ; GFX8-NEXT: [[LSHR4:%[0-9]+]]:_(s16) = G_LSHR [[SHL1]], [[TRUNC4]](s16)
     ; GFX8-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC1]](s16), [[LSHR4]]
     ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP1]](s1), [[C1]], [[SHL1]]
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[TRUNC5]](s16)
     ; GFX8-NEXT: [[LSHR5:%[0-9]+]]:_(s16) = G_LSHR [[SHL2]], [[TRUNC5]](s16)
     ; GFX8-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC2]](s16), [[LSHR5]]
     ; GFX8-NEXT: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP2]](s1), [[C1]], [[SHL2]]
     ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT1]](s16)
     ; GFX8-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
-    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[SELECT2]](s16)
     ; GFX8-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C2]]
+    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C2]]
     ; GFX8-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL4]]
-    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C2]]
-    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C2]]
+    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C2]]
     ; GFX8-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL5]]
-    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: ushlsat_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX9-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16)
     ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s16) = G_LSHR [[SHL]], [[TRUNC3]](s16)
     ; GFX9-NEXT: [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 -1
     ; GFX9-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC]](s16), [[LSHR3]]
     ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s16) = G_SELECT [[ICMP]](s1), [[C1]], [[SHL]]
     ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[TRUNC4]](s16)
     ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s16) = G_LSHR [[SHL1]], [[TRUNC4]](s16)
     ; GFX9-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC1]](s16), [[LSHR4]]
     ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s16) = G_SELECT [[ICMP1]](s1), [[C1]], [[SHL1]]
     ; GFX9-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[TRUNC5]](s16)
     ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s16) = G_LSHR [[SHL2]], [[TRUNC5]](s16)
     ; GFX9-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(ne), [[TRUNC2]](s16), [[LSHR5]]
     ; GFX9-NEXT: [[SELECT2:%[0-9]+]]:_(s16) = G_SELECT [[ICMP2]](s1), [[C1]], [[SHL2]]
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT]](s16)
     ; GFX9-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT1]](s16)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[ANYEXT1]](s32)
     ; GFX9-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[SELECT2]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[BITCAST4]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR6]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT2]](s32), [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR6]](s32), [[BITCAST4]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %3:_(<3 x s16>) = G_USHLSAT %1, %2
     %4:_(<3 x s16>) = G_IMPLICIT_DEF
     %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4
     $vgpr0_vgpr1_vgpr2 = COPY %5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-usubsat.mir

bb.0:
 
     ; GFX6-LABEL: name: usubsat_v3s16
     ; GFX6: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX6-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX6-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX6-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX6-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
-    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[BITCAST]], [[C]](s32)
     ; GFX6-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[LSHR1]], [[C]](s32)
     ; GFX6-NEXT: [[UMIN:%[0-9]+]]:_(s32) = G_UMIN [[SHL]], [[SHL1]]
     ; GFX6-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[SHL]], [[UMIN]]
     ; GFX6-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[SUB]], [[C]](s32)
     ; GFX6-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[LSHR]], [[C]](s32)
-    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[BITCAST2]], [[C]](s32)
     ; GFX6-NEXT: [[UMIN1:%[0-9]+]]:_(s32) = G_UMIN [[SHL2]], [[SHL3]]
     ; GFX6-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[SHL2]], [[UMIN1]]
     ; GFX6-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[SUB1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[BITCAST1]], [[C]](s32)
     ; GFX6-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[LSHR2]], [[C]](s32)
     ; GFX6-NEXT: [[UMIN2:%[0-9]+]]:_(s32) = G_UMIN [[SHL4]], [[SHL5]]
     ; GFX6-NEXT: [[SUB2:%[0-9]+]]:_(s32) = G_SUB [[SHL4]], [[UMIN2]]
     ; GFX6-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[SUB2]], [[C]](s32)
     ; GFX6-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX6-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX6-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX6-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX6-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX6-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; GFX6-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
     ; GFX6-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
     ; GFX6-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; GFX6-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL6]]
-    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX6-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX6-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
-    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; GFX6-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; GFX6-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
     ; GFX6-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL7]]
-    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX6-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX6-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C1]]
-    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; GFX6-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; GFX6-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
     ; GFX6-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL8]]
-    ; GFX6-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX6-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX6-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX6-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX8-LABEL: name: usubsat_v3s16
     ; GFX8: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX8-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
     ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
     ; GFX8-NEXT: [[USUBSAT:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC]], [[TRUNC3]]
     ; GFX8-NEXT: [[USUBSAT1:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC1]], [[TRUNC4]]
     ; GFX8-NEXT: [[USUBSAT2:%[0-9]+]]:_(s16) = G_USUBSAT [[TRUNC2]], [[TRUNC5]]
     ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT1]](s16)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[USUBSAT2]](s16)
     ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
-    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; GFX8-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND2]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND1]], [[SHL2]]
-    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>), [[BITCAST8]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
     ; GFX8-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     ; GFX9-LABEL: name: usubsat_v3s16
     ; GFX9: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; GFX9-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
     ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
     ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR1]](s32), [[BITCAST2]](s32)
     ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR2]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[USUBSAT:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9-NEXT: [[USUBSAT1:%[0-9]+]]:_(<2 x s16>) = G_USUBSAT [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT1]](<2 x s16>)
     ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[USUBSAT1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
-    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR3]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[BITCAST6]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST7]](s32)
+    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST5]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[BITCAST5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[LSHR4]](s32), [[BITCAST6]](s32)
     ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
     ; GFX9-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s16>), %2:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %3:_(<3 x s16>) = G_USUBSAT %1, %2
     %4:_(<3 x s16>) = G_IMPLICIT_DEF
     %5:_(<6 x s16>) = G_CONCAT_VECTORS %3, %4
     $vgpr0_vgpr1_vgpr2 = COPY %5

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-vector-args-gfx8-plus.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -global-isel-abort=0 -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX8 %s
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -global-isel-abort=0 -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX9 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -run-pass=legalizer -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX8 %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -run-pass=legalizer -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX9 %s
 
 --- |
 
   define <2 x i16> @and_v2i16(<2 x i16> %a, <2 x i16> %b) #0 {
     %and = and <2 x i16> %a, %b
     ret <2 x i16> %and
   }
 

 name: add_v3i16
 body: |
   bb.1:
     liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $sgpr30_sgpr31
 
     ; GFX8-LABEL: name: add_v3i16
     ; GFX8: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-    ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-    ; GFX8-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
-    ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
-    ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
     ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-    ; GFX8-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY2]](<2 x s16>), [[COPY3]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS1]](<6 x s16>)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]](s96)
-    ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; GFX8-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
     ; GFX8-NEXT: [[ADD:%[0-9]+]]:_(s16) = G_ADD [[TRUNC]], [[TRUNC3]]
     ; GFX8-NEXT: [[ADD1:%[0-9]+]]:_(s16) = G_ADD [[TRUNC1]], [[TRUNC4]]
     ; GFX8-NEXT: [[ADD2:%[0-9]+]]:_(s16) = G_ADD [[TRUNC2]], [[TRUNC5]]
-    ; GFX8-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[ADD]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[ADD1]](s16)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[ADD2]](s16)
-    ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
+    ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>)
-    ; GFX8-NEXT: $vgpr1 = COPY [[BITCAST4]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST4]](<2 x s16>)
+    ; GFX8-NEXT: $vgpr1 = COPY [[BITCAST5]](<2 x s16>)
     ; GFX8-NEXT: [[COPY5:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY4]]
     ; GFX8-NEXT: S_SETPC_B64_return [[COPY5]], implicit $vgpr0, implicit $vgpr1
     ; GFX9-LABEL: name: add_v3i16
     ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY2]](<2 x s16>), [[COPY3]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS1]](<6 x s16>)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]](s96)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
     ; GFX9-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV1]](s32), [[DEF1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV3]](s32), [[LSHR1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV4]](s32), [[DEF1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[DEF]](s32)
     ; GFX9-NEXT: [[ADD:%[0-9]+]]:_(<2 x s16>) = G_ADD [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC2]]
     ; GFX9-NEXT: [[ADD1:%[0-9]+]]:_(<2 x s16>) = G_ADD [[BUILD_VECTOR_TRUNC1]], [[BUILD_VECTOR_TRUNC3]]
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[ADD1]](<2 x s16>)
-    ; GFX9-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[BITCAST3]](s32)
-    ; GFX9-NEXT: $vgpr0 = COPY [[ADD]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[ADD]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[ADD1]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC5]](<2 x s16>)
     ; GFX9-NEXT: [[COPY5:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY4]]
     ; GFX9-NEXT: S_SETPC_B64_return [[COPY5]], implicit $vgpr0, implicit $vgpr1
     %3:_(<2 x s16>) = COPY $vgpr0
     %4:_(<2 x s16>) = COPY $vgpr1
-    %5:_(<2 x s16>) = G_IMPLICIT_DEF
-    %6:_(<6 x s16>) = G_CONCAT_VECTORS %3(<2 x s16>), %4(<2 x s16>), %5(<2 x s16>)
-    %19:_(s96) = G_BITCAST %6(<6 x s16>)
-    %20:_(s48) = G_TRUNC %19(s96)
-    %0:_(<3 x s16>) = G_BITCAST %20(s48)
-    %8:_(<2 x s16>) = COPY $vgpr2
-    %9:_(<2 x s16>) = COPY $vgpr3
-    %10:_(<6 x s16>) = G_CONCAT_VECTORS %8(<2 x s16>), %9(<2 x s16>), %5(<2 x s16>)
-    %21:_(s96) = G_BITCAST %10(<6 x s16>)
-    %22:_(s48) = G_TRUNC %21(s96)
-    %1:_(<3 x s16>) = G_BITCAST %22(s48)
+    %5:_(<4 x s16>) = G_CONCAT_VECTORS %3(<2 x s16>), %4(<2 x s16>)
+    %6:_(s16), %7:_(s16), %8:_(s16), %9:_(s16) = G_UNMERGE_VALUES %5(<4 x s16>)
+    %0:_(<3 x s16>) = G_BUILD_VECTOR %6(s16), %7(s16), %8(s16)
+    %10:_(<2 x s16>) = COPY $vgpr2
+    %11:_(<2 x s16>) = COPY $vgpr3
+    %12:_(<4 x s16>) = G_CONCAT_VECTORS %10(<2 x s16>), %11(<2 x s16>)
+    %13:_(s16), %14:_(s16), %15:_(s16), %16:_(s16) = G_UNMERGE_VALUES %12(<4 x s16>)
+    %1:_(<3 x s16>) = G_BUILD_VECTOR %13(s16), %14(s16), %15(s16)
     %2:sgpr_64 = COPY $sgpr30_sgpr31
-    %12:_(<3 x s16>) = G_ADD %0, %1
-    %16:_(<3 x s16>) = G_IMPLICIT_DEF
-    %17:_(<6 x s16>) = G_CONCAT_VECTORS %12(<3 x s16>), %16(<3 x s16>)
-    %14:_(<2 x s16>), %15:_(<2 x s16>), %18:_(<2 x s16>) = G_UNMERGE_VALUES %17(<6 x s16>)
-    $vgpr0 = COPY %14(<2 x s16>)
-    $vgpr1 = COPY %15(<2 x s16>)
-    %13:ccr_sgpr_64 = COPY %2
-    S_SETPC_B64_return %13, implicit $vgpr0, implicit $vgpr1
+    %17:_(<3 x s16>) = G_ADD %0, %1
+    %21:_(s16), %22:_(s16), %23:_(s16) = G_UNMERGE_VALUES %17(<3 x s16>)
+    %24:_(s16) = G_IMPLICIT_DEF
+    %25:_(<4 x s16>) = G_BUILD_VECTOR %21(s16), %22(s16), %23(s16), %24(s16)
+    %19:_(<2 x s16>), %20:_(<2 x s16>) = G_UNMERGE_VALUES %25(<4 x s16>)
+    $vgpr0 = COPY %19(<2 x s16>)
+    $vgpr1 = COPY %20(<2 x s16>)
+    %18:ccr_sgpr_64 = COPY %2
+    S_SETPC_B64_return %18, implicit $vgpr0, implicit $vgpr1
 ...
 
 ---
 name: shl_v3i16
 body: |
   bb.1:
     liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $sgpr30_sgpr31
 
     ; GFX8-LABEL: name: shl_v3i16
     ; GFX8: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-    ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-    ; GFX8-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
-    ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
-    ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
+    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
     ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-    ; GFX8-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY2]](<2 x s16>), [[COPY3]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS1]](<6 x s16>)
-    ; GFX8-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]](s96)
-    ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[UV3]](s32)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV3]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV4]](s32)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; GFX8-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC3]](s16)
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC1]], [[TRUNC4]](s16)
     ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s16) = G_SHL [[TRUNC2]], [[TRUNC5]](s16)
-    ; GFX8-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[SHL]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[SHL1]](s16)
     ; GFX8-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL3]]
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[SHL2]](s16)
-    ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; GFX8-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
+    ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; GFX8-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL4]]
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>)
-    ; GFX8-NEXT: $vgpr1 = COPY [[BITCAST4]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST4]](<2 x s16>)
+    ; GFX8-NEXT: $vgpr1 = COPY [[BITCAST5]](<2 x s16>)
     ; GFX8-NEXT: [[COPY5:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY4]]
     ; GFX8-NEXT: S_SETPC_B64_return [[COPY5]], implicit $vgpr0, implicit $vgpr1
     ; GFX9-LABEL: name: shl_v3i16
     ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS]](<6 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s96)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
-    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV1]](s32), [[DEF1]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[UV3:%[0-9]+]]:_(<3 x s16>), [[UV4:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<6 x s16>)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
-    ; GFX9-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY2]](<2 x s16>), [[COPY3]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s96) = G_BITCAST [[CONCAT_VECTORS2]](<6 x s16>)
-    ; GFX9-NEXT: [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]](s96)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C]](s32)
-    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV5]](s32), [[LSHR1]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV6]](s32), [[DEF1]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[UV8:%[0-9]+]]:_(<3 x s16>), [[UV9:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<6 x s16>)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX9-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
     ; GFX9-NEXT: [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV3]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT]](<4 x s16>), 0
-    ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(s16) = COPY [[TRUNC]](s16)
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF2]], [[UV8]](<3 x s16>), 0
-    ; GFX9-NEXT: [[EXTRACT1:%[0-9]+]]:_(<2 x s16>) = G_EXTRACT [[INSERT1]](<4 x s16>), 0
-    ; GFX9-NEXT: [[COPY6:%[0-9]+]]:_(s16) = COPY [[TRUNC1]](s16)
-    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[EXTRACT]], [[EXTRACT1]](<2 x s16>)
-    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[COPY5]], [[COPY6]](s16)
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[SHL]](<2 x s16>)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; GFX9-NEXT: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF3]](<4 x s16>)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[SHL:%[0-9]+]]:_(<2 x s16>) = G_SHL [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC1]](<2 x s16>)
+    ; GFX9-NEXT: [[SHL1:%[0-9]+]]:_(s16) = G_SHL [[TRUNC]], [[TRUNC1]](s16)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[SHL]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR2]](s32)
     ; GFX9-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[SHL1]](s16)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[BITCAST3]](s32)
-    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC4]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC5]](<2 x s16>)
-    ; GFX9-NEXT: [[COPY7:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY4]]
-    ; GFX9-NEXT: S_SETPC_B64_return [[COPY7]], implicit $vgpr0, implicit $vgpr1
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ANYEXT]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC2]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC3]](<2 x s16>)
+    ; GFX9-NEXT: [[COPY5:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY4]]
+    ; GFX9-NEXT: S_SETPC_B64_return [[COPY5]], implicit $vgpr0, implicit $vgpr1
     %3:_(<2 x s16>) = COPY $vgpr0
     %4:_(<2 x s16>) = COPY $vgpr1
-    %5:_(<2 x s16>) = G_IMPLICIT_DEF
-    %6:_(<6 x s16>) = G_CONCAT_VECTORS %3(<2 x s16>), %4(<2 x s16>), %5(<2 x s16>)
-    %19:_(s96) = G_BITCAST %6(<6 x s16>)
-    %20:_(s48) = G_TRUNC %19(s96)
-    %0:_(<3 x s16>) = G_BITCAST %20(s48)
-    %8:_(<2 x s16>) = COPY $vgpr2
-    %9:_(<2 x s16>) = COPY $vgpr3
-    %10:_(<6 x s16>) = G_CONCAT_VECTORS %8(<2 x s16>), %9(<2 x s16>), %5(<2 x s16>)
-    %21:_(s96) = G_BITCAST %10(<6 x s16>)
-    %22:_(s48) = G_TRUNC %21(s96)
-    %1:_(<3 x s16>) = G_BITCAST %22(s48)
+    %5:_(<4 x s16>) = G_CONCAT_VECTORS %3(<2 x s16>), %4(<2 x s16>)
+    %6:_(s16), %7:_(s16), %8:_(s16), %9:_(s16) = G_UNMERGE_VALUES %5(<4 x s16>)
+    %0:_(<3 x s16>) = G_BUILD_VECTOR %6(s16), %7(s16), %8(s16)
+    %10:_(<2 x s16>) = COPY $vgpr2
+    %11:_(<2 x s16>) = COPY $vgpr3
+    %12:_(<4 x s16>) = G_CONCAT_VECTORS %10(<2 x s16>), %11(<2 x s16>)
+    %13:_(s16), %14:_(s16), %15:_(s16), %16:_(s16) = G_UNMERGE_VALUES %12(<4 x s16>)
+    %1:_(<3 x s16>) = G_BUILD_VECTOR %13(s16), %14(s16), %15(s16)
     %2:sgpr_64 = COPY $sgpr30_sgpr31
-    %12:_(<3 x s16>) = G_SHL %0, %1(<3 x s16>)
-    %16:_(<3 x s16>) = G_IMPLICIT_DEF
-    %17:_(<6 x s16>) = G_CONCAT_VECTORS %12(<3 x s16>), %16(<3 x s16>)
-    %14:_(<2 x s16>), %15:_(<2 x s16>), %18:_(<2 x s16>) = G_UNMERGE_VALUES %17(<6 x s16>)
-    $vgpr0 = COPY %14(<2 x s16>)
-    $vgpr1 = COPY %15(<2 x s16>)
-    %13:ccr_sgpr_64 = COPY %2
-    S_SETPC_B64_return %13, implicit $vgpr0, implicit $vgpr1
+    %17:_(<3 x s16>) = G_SHL %0, %1(<3 x s16>)
+    %21:_(s16), %22:_(s16), %23:_(s16) = G_UNMERGE_VALUES %17(<3 x s16>)
+    %24:_(s16) = G_IMPLICIT_DEF
+    %25:_(<4 x s16>) = G_BUILD_VECTOR %21(s16), %22(s16), %23(s16), %24(s16)
+    %19:_(<2 x s16>), %20:_(<2 x s16>) = G_UNMERGE_VALUES %25(<4 x s16>)
+    $vgpr0 = COPY %19(<2 x s16>)
+    $vgpr1 = COPY %20(<2 x s16>)
+    %18:ccr_sgpr_64 = COPY %2
+    S_SETPC_B64_return %18, implicit $vgpr0, implicit $vgpr1
 ...
 
 ---
 name: fma_v4f16
 body: |
   bb.1:
     liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $sgpr30_sgpr31
 

 body: |
   bb.1:
     liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
 
     ; GFX8-LABEL: name: maxnum_v5i16
     ; GFX8: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX8-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
     ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
-    ; GFX8-NEXT: [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[COPY2]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[CONCAT_VECTORS]](<10 x s16>)
-    ; GFX8-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s160)
-    ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[UV]](s32)
+    ; GFX8-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST]](s32)
     ; GFX8-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
+    ; GFX8-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC1:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR]](s32)
-    ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[UV1]](s32)
-    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST1]](s32)
+    ; GFX8-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC3:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR1]](s32)
-    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[UV2]](s32)
+    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST2]](s32)
     ; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
     ; GFX8-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
     ; GFX8-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
-    ; GFX8-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[COPY3]](<2 x s16>), [[COPY4]](<2 x s16>), [[COPY5]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX8-NEXT: [[BITCAST1:%[0-9]+]]:_(s160) = G_BITCAST [[CONCAT_VECTORS1]](<10 x s16>)
-    ; GFX8-NEXT: [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]](s160)
-    ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[UV5]](s32)
-    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC5:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST3]](s32)
+    ; GFX8-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC6:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR2]](s32)
-    ; GFX8-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[UV6]](s32)
-    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C]](s32)
+    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC7:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST4]](s32)
+    ; GFX8-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
     ; GFX8-NEXT: [[TRUNC8:%[0-9]+]]:_(s16) = G_TRUNC [[LSHR3]](s32)
-    ; GFX8-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[UV7]](s32)
+    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
+    ; GFX8-NEXT: [[TRUNC9:%[0-9]+]]:_(s16) = G_TRUNC [[BITCAST5]](s32)
     ; GFX8-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC]]
     ; GFX8-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC5]]
     ; GFX8-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
     ; GFX8-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC1]]
     ; GFX8-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC6]]
     ; GFX8-NEXT: [[FMAXNUM_IEEE1:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
     ; GFX8-NEXT: [[FCANONICALIZE4:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC2]]
     ; GFX8-NEXT: [[FCANONICALIZE5:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC7]]
     ; GFX8-NEXT: [[FMAXNUM_IEEE2:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE4]], [[FCANONICALIZE5]]
     ; GFX8-NEXT: [[FCANONICALIZE6:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC3]]
     ; GFX8-NEXT: [[FCANONICALIZE7:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC8]]
     ; GFX8-NEXT: [[FMAXNUM_IEEE3:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE6]], [[FCANONICALIZE7]]
     ; GFX8-NEXT: [[FCANONICALIZE8:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC4]]
     ; GFX8-NEXT: [[FCANONICALIZE9:%[0-9]+]]:_(s16) = G_FCANONICALIZE [[TRUNC9]]
     ; GFX8-NEXT: [[FMAXNUM_IEEE4:%[0-9]+]]:_(s16) = G_FMAXNUM_IEEE [[FCANONICALIZE8]], [[FCANONICALIZE9]]
-    ; GFX8-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; GFX8-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<6 x s16>)
-    ; GFX8-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV10]](<2 x s16>)
     ; GFX8-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE]](s16)
     ; GFX8-NEXT: [[ZEXT1:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE1]](s16)
     ; GFX8-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[ZEXT1]], [[C]](s32)
     ; GFX8-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[ZEXT]], [[SHL]]
-    ; GFX8-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; GFX8-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; GFX8-NEXT: [[ZEXT2:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE2]](s16)
     ; GFX8-NEXT: [[ZEXT3:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE3]](s16)
     ; GFX8-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[ZEXT3]], [[C]](s32)
     ; GFX8-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[ZEXT2]], [[SHL1]]
-    ; GFX8-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; GFX8-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
     ; GFX8-NEXT: [[ZEXT4:%[0-9]+]]:_(s32) = G_ZEXT [[FMAXNUM_IEEE4]](s16)
-    ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
-    ; GFX8-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND]], [[C]](s32)
+    ; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; GFX8-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[C1]], [[C]](s32)
     ; GFX8-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[ZEXT4]], [[SHL2]]
-    ; GFX8-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST3]](<2 x s16>)
-    ; GFX8-NEXT: $vgpr1 = COPY [[BITCAST4]](<2 x s16>)
-    ; GFX8-NEXT: $vgpr2 = COPY [[BITCAST5]](<2 x s16>)
+    ; GFX8-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; GFX8-NEXT: $vgpr0 = COPY [[BITCAST6]](<2 x s16>)
+    ; GFX8-NEXT: $vgpr1 = COPY [[BITCAST7]](<2 x s16>)
+    ; GFX8-NEXT: $vgpr2 = COPY [[BITCAST8]](<2 x s16>)
     ; GFX8-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
     ; GFX9-LABEL: name: maxnum_v5i16
     ; GFX9: [[COPY:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
     ; GFX9-NEXT: [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
     ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
-    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[COPY]](<2 x s16>), [[COPY1]](<2 x s16>), [[COPY2]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s160) = G_BITCAST [[CONCAT_VECTORS]](<10 x s16>)
-    ; GFX9-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]](s160)
+    ; GFX9-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
     ; GFX9-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
-    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[UV]], [[C]](s32)
-    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[UV1]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV]](s32), [[LSHR]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV1]](s32), [[LSHR1]](s32)
-    ; GFX9-NEXT: [[DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV2]](s32), [[DEF1]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC]](<2 x s16>), [[BUILD_VECTOR_TRUNC1]](<2 x s16>), [[BUILD_VECTOR_TRUNC2]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[UV5:%[0-9]+]]:_(<5 x s16>), [[UV6:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<10 x s16>)
+    ; GFX9-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[COPY2]](<2 x s16>)
     ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
     ; GFX9-NEXT: [[COPY4:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr4
     ; GFX9-NEXT: [[COPY5:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr5
-    ; GFX9-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[COPY3]](<2 x s16>), [[COPY4]](<2 x s16>), [[COPY5]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[BITCAST1:%[0-9]+]]:_(s160) = G_BITCAST [[CONCAT_VECTORS2]](<10 x s16>)
-    ; GFX9-NEXT: [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]](s160)
-    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[UV7]], [[C]](s32)
-    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[UV8]], [[C]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV7]](s32), [[LSHR2]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV8]](s32), [[LSHR3]](s32)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[UV9]](s32), [[DEF1]](s32)
-    ; GFX9-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<10 x s16>) = G_CONCAT_VECTORS [[BUILD_VECTOR_TRUNC3]](<2 x s16>), [[BUILD_VECTOR_TRUNC4]](<2 x s16>), [[BUILD_VECTOR_TRUNC5]](<2 x s16>), [[DEF]](<2 x s16>), [[DEF]](<2 x s16>)
-    ; GFX9-NEXT: [[UV12:%[0-9]+]]:_(<5 x s16>), [[UV13:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<10 x s16>)
-    ; GFX9-NEXT: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[UV5]](<5 x s16>), 0
-    ; GFX9-NEXT: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF2]], [[UV12]](<5 x s16>), 0
-    ; GFX9-NEXT: [[UV14:%[0-9]+]]:_(<2 x s16>), [[UV15:%[0-9]+]]:_(<2 x s16>), [[UV16:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>)
-    ; GFX9-NEXT: [[UV17:%[0-9]+]]:_(<2 x s16>), [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>)
-    ; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV14]]
-    ; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV17]]
+    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY3]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[COPY4]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[COPY5]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST]](s32), [[LSHR]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC1:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST1]](s32), [[LSHR1]](s32)
+    ; GFX9-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC2:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC3:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST3]](s32), [[LSHR2]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC4:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST4]](s32), [[LSHR3]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC5:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST5]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: [[FCANONICALIZE:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC]]
+    ; GFX9-NEXT: [[FCANONICALIZE1:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC3]]
     ; GFX9-NEXT: [[FMAXNUM_IEEE:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE]], [[FCANONICALIZE1]]
-    ; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV15]]
-    ; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV18]]
+    ; GFX9-NEXT: [[FCANONICALIZE2:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC1]]
+    ; GFX9-NEXT: [[FCANONICALIZE3:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC4]]
     ; GFX9-NEXT: [[FMAXNUM_IEEE1:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE2]], [[FCANONICALIZE3]]
-    ; GFX9-NEXT: [[FCANONICALIZE4:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV16]]
-    ; GFX9-NEXT: [[FCANONICALIZE5:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[UV19]]
+    ; GFX9-NEXT: [[FCANONICALIZE4:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC2]]
+    ; GFX9-NEXT: [[FCANONICALIZE5:%[0-9]+]]:_(<2 x s16>) = G_FCANONICALIZE [[BUILD_VECTOR_TRUNC5]]
     ; GFX9-NEXT: [[FMAXNUM_IEEE2:%[0-9]+]]:_(<2 x s16>) = G_FMAXNUM_IEEE [[FCANONICALIZE4]], [[FCANONICALIZE5]]
-    ; GFX9-NEXT: [[DEF3:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; GFX9-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE2]](<2 x s16>)
-    ; GFX9-NEXT: [[UV20:%[0-9]+]]:_(<2 x s16>), [[UV21:%[0-9]+]]:_(<2 x s16>), [[UV22:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF3]](<6 x s16>)
-    ; GFX9-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV20]](<2 x s16>)
-    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST2]](s32), [[BITCAST3]](s32)
-    ; GFX9-NEXT: $vgpr0 = COPY [[FMAXNUM_IEEE]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr1 = COPY [[FMAXNUM_IEEE1]](<2 x s16>)
-    ; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
+    ; GFX9-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST7:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE1]](<2 x s16>)
+    ; GFX9-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST7]], [[C]](s32)
+    ; GFX9-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[FMAXNUM_IEEE2]](<2 x s16>)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC6:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST6]](s32), [[LSHR4]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC7:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST7]](s32), [[LSHR5]](s32)
+    ; GFX9-NEXT: [[BUILD_VECTOR_TRUNC8:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[BITCAST8]](s32), [[DEF]](s32)
+    ; GFX9-NEXT: $vgpr0 = COPY [[BUILD_VECTOR_TRUNC6]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr1 = COPY [[BUILD_VECTOR_TRUNC7]](<2 x s16>)
+    ; GFX9-NEXT: $vgpr2 = COPY [[BUILD_VECTOR_TRUNC8]](<2 x s16>)
     ; GFX9-NEXT: SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
     %2:_(<2 x s16>) = COPY $vgpr0
     %3:_(<2 x s16>) = COPY $vgpr1
     %4:_(<2 x s16>) = COPY $vgpr2
-    %5:_(<2 x s16>) = G_IMPLICIT_DEF
-    %6:_(<10 x s16>) = G_CONCAT_VECTORS %2(<2 x s16>), %3(<2 x s16>), %4(<2 x s16>), %5(<2 x s16>), %5(<2 x s16>)
-    %22:_(s160) = G_BITCAST %6(<10 x s16>)
-    %23:_(s80) = G_TRUNC %22(s160)
-    %0:_(<5 x s16>) = G_BITCAST %23(s80)
-    %8:_(<2 x s16>) = COPY $vgpr3
-    %9:_(<2 x s16>) = COPY $vgpr4
-    %10:_(<2 x s16>) = COPY $vgpr5
-    %11:_(<10 x s16>) = G_CONCAT_VECTORS %8(<2 x s16>), %9(<2 x s16>), %10(<2 x s16>), %5(<2 x s16>), %5(<2 x s16>)
-    %24:_(s160) = G_BITCAST %11(<10 x s16>)
-    %25:_(s80) = G_TRUNC %24(s160)
-    %1:_(<5 x s16>) = G_BITCAST %25(s80)
-    %14:_(<5 x s16>) = G_FMAXNUM %0, %1
-    %18:_(<5 x s16>) = G_IMPLICIT_DEF
-    %19:_(<10 x s16>) = G_CONCAT_VECTORS %14(<5 x s16>), %18(<5 x s16>)
-    %15:_(<2 x s16>), %16:_(<2 x s16>), %17:_(<2 x s16>), %20:_(<2 x s16>), %21:_(<2 x s16>) = G_UNMERGE_VALUES %19(<10 x s16>)
-    $vgpr0 = COPY %15(<2 x s16>)
-    $vgpr1 = COPY %16(<2 x s16>)
-    $vgpr2 = COPY %17(<2 x s16>)
+    %5:_(<6 x s16>) = G_CONCAT_VECTORS %2(<2 x s16>), %3(<2 x s16>), %4(<2 x s16>)
+    %6:_(s16), %7:_(s16), %8:_(s16), %9:_(s16), %10:_(s16), %11:_(s16) = G_UNMERGE_VALUES %5(<6 x s16>)
+    %0:_(<5 x s16>) = G_BUILD_VECTOR %6(s16), %7(s16), %8(s16), %9(s16), %10(s16)
+    %12:_(<2 x s16>) = COPY $vgpr3
+    %13:_(<2 x s16>) = COPY $vgpr4
+    %14:_(<2 x s16>) = COPY $vgpr5
+    %15:_(<6 x s16>) = G_CONCAT_VECTORS %12(<2 x s16>), %13(<2 x s16>), %14(<2 x s16>)
+    %16:_(s16), %17:_(s16), %18:_(s16), %19:_(s16), %20:_(s16), %21:_(s16) = G_UNMERGE_VALUES %15(<6 x s16>)
+    %1:_(<5 x s16>) = G_BUILD_VECTOR %16(s16), %17(s16), %18(s16), %19(s16), %20(s16)
+    %23:_(<5 x s16>) = G_FMAXNUM %0, %1
+    %27:_(s16), %28:_(s16), %29:_(s16), %30:_(s16), %31:_(s16) = G_UNMERGE_VALUES %23(<5 x s16>)
+    %32:_(s16) = G_IMPLICIT_DEF
+    %33:_(<6 x s16>) = G_BUILD_VECTOR %27(s16), %28(s16), %29(s16), %30(s16), %31(s16), %32(s16)
+    %24:_(<2 x s16>), %25:_(<2 x s16>), %26:_(<2 x s16>) = G_UNMERGE_VALUES %33(<6 x s16>)
+    $vgpr0 = COPY %24(<2 x s16>)
+    $vgpr1 = COPY %25(<2 x s16>)
+    $vgpr2 = COPY %26(<2 x s16>)
     SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-xor.mir

 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer -allow-ginsert-as-artifact=0 -global-isel-abort=0 %s -o - | FileCheck %s
+# RUN: llc -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -O0 -run-pass=legalizer %s -o - | FileCheck %s
 
 ---
 name: test_xor_s32
 body: |
   bb.0:
     liveins: $vgpr0, $vgpr1
 
     ; CHECK-LABEL: name: test_xor_s32

 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
 
     ; CHECK-LABEL: name: test_xor_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<3 x s32>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY1]](<3 x s32>)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[DEF]](s32)
-    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
-    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s32>) = G_CONCAT_VECTORS [[XOR]](<2 x s32>), [[XOR1]](<2 x s32>), [[DEF1]](<2 x s32>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<3 x s32>), [[UV7:%[0-9]+]]:_(<3 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s32>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[UV6]](<3 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV3]](s32), [[UV4]](s32)
+    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR]], [[BUILD_VECTOR1]]
+    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(s32) = G_XOR [[UV2]], [[UV5]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[XOR]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[UV6]](s32), [[UV7]](s32), [[XOR1]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR2]](<3 x s32>)
     %0:_(<3 x s32>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<3 x s32>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s32>) = G_XOR %0, %1
     $vgpr0_vgpr1_vgpr2 = COPY %2
 ...
 
 ---
 name: test_xor_v4s32

 name: test_xor_v5s32
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_xor_v5s32
     ; CHECK: [[DEF:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<5 x s32>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32)
     ; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV2]](s32), [[UV3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV4]](s32), [[DEF2]](s32)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](<5 x s32>)
-    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV6]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV7]](s32), [[UV8]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR5:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV9]](s32), [[DEF2]](s32)
-    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR]], [[BUILD_VECTOR3]]
-    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR1]], [[BUILD_VECTOR4]]
-    ; CHECK-NEXT: [[XOR2:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR2]], [[BUILD_VECTOR5]]
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<10 x s32>) = G_CONCAT_VECTORS [[XOR]](<2 x s32>), [[XOR1]](<2 x s32>), [[XOR2]](<2 x s32>), [[DEF3]](<2 x s32>), [[DEF3]](<2 x s32>)
-    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(<5 x s32>), [[UV11:%[0-9]+]]:_(<5 x s32>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<10 x s32>)
-    ; CHECK-NEXT: [[DEF4:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<8 x s32>) = G_INSERT [[DEF4]], [[UV10]](<5 x s32>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[INSERT]](<8 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV5]](s32), [[UV6]](s32)
+    ; CHECK-NEXT: [[BUILD_VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV7]](s32), [[UV8]](s32)
+    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR]], [[BUILD_VECTOR2]]
+    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(<2 x s32>) = G_XOR [[BUILD_VECTOR1]], [[BUILD_VECTOR3]]
+    ; CHECK-NEXT: [[XOR2:%[0-9]+]]:_(s32) = G_XOR [[UV4]], [[UV9]]
+    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[XOR]](<2 x s32>)
+    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[XOR1]](<2 x s32>)
+    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF2]](<8 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR4:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[UV10]](s32), [[UV11]](s32), [[UV12]](s32), [[UV13]](s32), [[XOR2]](s32), [[UV19]](s32), [[UV20]](s32), [[UV21]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY [[BUILD_VECTOR4]](<8 x s32>)
     %0:_(<5 x s32>) = G_IMPLICIT_DEF
     %1:_(<5 x s32>) = G_IMPLICIT_DEF
     %2:_(<5 x s32>) = G_XOR %0, %1
     %3:_(<8 x s32>) = G_IMPLICIT_DEF
     %4:_(<8 x s32>) = G_INSERT %3, %2, 0
     $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7 = COPY %4
 ...
 

 ---
 name: test_xor_v3s16
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1_vgpr2, $vgpr3_vgpr4_vgpr5
     ; CHECK-LABEL: name: test_xor_v3s16
     ; CHECK: [[COPY:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
-    ; CHECK-NEXT: [[UV2:%[0-9]+]]:_(<3 x s16>), [[UV3:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[UV2]](<3 x s16>), 0
-    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT]], [[INSERT1]]
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[XOR]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST2]](<2 x s16>), [[BITCAST3]](<2 x s16>)
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
     ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
     ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS]](<6 x s16>)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[XOR]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST8]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV9]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[BITCAST8]], [[C1]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[BITCAST9]], [[C1]]
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C1]]
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C1]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>), [[BITCAST14]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[CONCAT_VECTORS2]](<6 x s16>)
     %0:_(<6 x s16>) = COPY $vgpr0_vgpr1_vgpr2
     %1:_(<6 x s16>) = COPY $vgpr3_vgpr4_vgpr5
     %2:_(<3 x s16>), %3:_(<3 x s16>) = G_UNMERGE_VALUES %0
     %4:_(<3 x s16>), %5:_(<3 x s16>) = G_UNMERGE_VALUES %1
     %6:_(<3 x s16>) = G_XOR %2, %4
     %7:_(<3 x s16>) = G_IMPLICIT_DEF
     %8:_(<6 x s16>) = G_CONCAT_VECTORS %6, %7
     $vgpr0_vgpr1_vgpr2 = COPY %8

 
 ---
 name: test_xor_v5s16
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_xor_v5s16
     ; CHECK: [[DEF:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<5 x s16>), [[UV1:%[0-9]+]]:_(<5 x s16>), [[UV2:%[0-9]+]]:_(<5 x s16>), [[UV3:%[0-9]+]]:_(<5 x s16>), [[UV4:%[0-9]+]]:_(<5 x s16>), [[UV5:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[DEF2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<5 x s16>), [[UV7:%[0-9]+]]:_(<5 x s16>), [[UV8:%[0-9]+]]:_(<5 x s16>), [[UV9:%[0-9]+]]:_(<5 x s16>), [[UV10:%[0-9]+]]:_(<5 x s16>), [[UV11:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS1]](<30 x s16>)
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV]](<5 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<6 x s16>) = G_INSERT [[DEF1]], [[UV6]](<5 x s16>), 0
-    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(<3 x s16>), [[UV13:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT]](<6 x s16>)
-    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(<3 x s16>), [[UV15:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[INSERT1]](<6 x s16>)
-    ; CHECK-NEXT: [[DEF3:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT2:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV12]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT3:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV14]](<3 x s16>), 0
-    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT2]], [[INSERT3]]
-    ; CHECK-NEXT: [[INSERT4:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV13]](<3 x s16>), 0
-    ; CHECK-NEXT: [[INSERT5:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF3]], [[UV15]](<3 x s16>), 0
-    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(<4 x s16>) = G_XOR [[INSERT4]], [[INSERT5]]
-    ; CHECK-NEXT: [[UV16:%[0-9]+]]:_(<2 x s16>), [[UV17:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[XOR]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV16]](<2 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>), [[UV2:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV17]](<2 x s16>)
-    ; CHECK-NEXT: [[UV18:%[0-9]+]]:_(<2 x s16>), [[UV19:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[XOR1]](<4 x s16>)
-    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV18]](<2 x s16>)
-    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST2]], [[C]](s32)
-    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV19]](<2 x s16>)
+    ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<6 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV3:%[0-9]+]]:_(<2 x s16>), [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF1]](<6 x s16>)
+    ; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(s32) = G_BITCAST [[UV5]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
     ; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
-    ; CHECK-NEXT: [[BITCAST4:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
-    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
-    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[AND3]], [[C]](s32)
+    ; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY]], [[C]](s32)
     ; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
-    ; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
-    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
-    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
-    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND5]], [[C]](s32)
-    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND4]], [[SHL2]]
-    ; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
-    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>), [[BITCAST6]](<2 x s16>)
-    ; CHECK-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<30 x s16>) = G_CONCAT_VECTORS [[CONCAT_VECTORS2]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>), [[DEF1]](<6 x s16>)
-    ; CHECK-NEXT: [[UV20:%[0-9]+]]:_(<5 x s16>), [[UV21:%[0-9]+]]:_(<5 x s16>), [[UV22:%[0-9]+]]:_(<5 x s16>), [[UV23:%[0-9]+]]:_(<5 x s16>), [[UV24:%[0-9]+]]:_(<5 x s16>), [[UV25:%[0-9]+]]:_(<5 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS3]](<30 x s16>)
-    ; CHECK-NEXT: [[DEF4:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT6:%[0-9]+]]:_(<8 x s16>) = G_INSERT [[DEF4]], [[UV20]](<5 x s16>), 0
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[INSERT6]](<8 x s16>)
+    ; CHECK-NEXT: [[BITCAST7:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST6]](<2 x s16>), [[BITCAST7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[BITCAST3]], [[C1]]
+    ; CHECK-NEXT: [[AND4:%[0-9]+]]:_(s32) = G_AND [[LSHR2]], [[C1]]
+    ; CHECK-NEXT: [[SHL2:%[0-9]+]]:_(s32) = G_SHL [[AND4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR2:%[0-9]+]]:_(s32) = G_OR [[AND3]], [[SHL2]]
+    ; CHECK-NEXT: [[BITCAST8:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR2]](s32)
+    ; CHECK-NEXT: [[AND5:%[0-9]+]]:_(s32) = G_AND [[BITCAST4]], [[C1]]
+    ; CHECK-NEXT: [[SHL3:%[0-9]+]]:_(s32) = G_SHL [[C2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR3:%[0-9]+]]:_(s32) = G_OR [[AND5]], [[SHL3]]
+    ; CHECK-NEXT: [[BITCAST9:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR3]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS1:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST8]](<2 x s16>), [[BITCAST9]](<2 x s16>)
+    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(<4 x s16>) = G_XOR [[CONCAT_VECTORS]], [[CONCAT_VECTORS1]]
+    ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(<2 x s16>), [[UV7:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[XOR]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST10:%[0-9]+]]:_(s32) = G_BITCAST [[UV6]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST10]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST11:%[0-9]+]]:_(s32) = G_BITCAST [[UV7]](<2 x s16>)
+    ; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[LSHR1]], [[C1]]
+    ; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[BITCAST2]], [[C1]]
+    ; CHECK-NEXT: [[SHL4:%[0-9]+]]:_(s32) = G_SHL [[AND7]], [[C]](s32)
+    ; CHECK-NEXT: [[OR4:%[0-9]+]]:_(s32) = G_OR [[AND6]], [[SHL4]]
+    ; CHECK-NEXT: [[BITCAST12:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR4]](s32)
+    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL5:%[0-9]+]]:_(s32) = G_SHL [[COPY2]], [[C]](s32)
+    ; CHECK-NEXT: [[OR5:%[0-9]+]]:_(s32) = G_OR [[COPY1]], [[SHL5]]
+    ; CHECK-NEXT: [[BITCAST13:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR5]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS2:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST12]](<2 x s16>), [[BITCAST13]](<2 x s16>)
+    ; CHECK-NEXT: [[AND8:%[0-9]+]]:_(s32) = G_AND [[LSHR3]], [[C1]]
+    ; CHECK-NEXT: [[AND9:%[0-9]+]]:_(s32) = G_AND [[BITCAST5]], [[C1]]
+    ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND9]], [[C]](s32)
+    ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND8]], [[SHL6]]
+    ; CHECK-NEXT: [[BITCAST14:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR6]](s32)
+    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+    ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[COPY4]], [[C]](s32)
+    ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[COPY3]], [[SHL7]]
+    ; CHECK-NEXT: [[BITCAST15:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR7]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS3:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST14]](<2 x s16>), [[BITCAST15]](<2 x s16>)
+    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(<4 x s16>) = G_XOR [[CONCAT_VECTORS2]], [[CONCAT_VECTORS3]]
+    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<2 x s16>), [[UV9:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[XOR1]](<4 x s16>)
+    ; CHECK-NEXT: [[BITCAST16:%[0-9]+]]:_(s32) = G_BITCAST [[UV8]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST16]], [[C]](s32)
+    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<8 x s16>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(<2 x s16>), [[UV11:%[0-9]+]]:_(<2 x s16>), [[UV12:%[0-9]+]]:_(<2 x s16>), [[UV13:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[DEF2]](<8 x s16>)
+    ; CHECK-NEXT: [[BITCAST17:%[0-9]+]]:_(s32) = G_BITCAST [[UV12]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST17]], [[C]](s32)
+    ; CHECK-NEXT: [[BITCAST18:%[0-9]+]]:_(s32) = G_BITCAST [[UV13]](<2 x s16>)
+    ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST18]], [[C]](s32)
+    ; CHECK-NEXT: [[AND10:%[0-9]+]]:_(s32) = G_AND [[BITCAST10]], [[C1]]
+    ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C1]]
+    ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND11]], [[C]](s32)
+    ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[AND10]], [[SHL8]]
+    ; CHECK-NEXT: [[BITCAST19:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR8]](s32)
+    ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[BITCAST11]], [[C1]]
+    ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[BITCAST16]], [[C1]]
+    ; CHECK-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND13]], [[C]](s32)
+    ; CHECK-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND12]], [[SHL9]]
+    ; CHECK-NEXT: [[BITCAST20:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR9]](s32)
+    ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C1]]
+    ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C1]]
+    ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND15]], [[C]](s32)
+    ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[AND14]], [[SHL10]]
+    ; CHECK-NEXT: [[BITCAST21:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR10]](s32)
+    ; CHECK-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[BITCAST18]], [[C1]]
+    ; CHECK-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C1]]
+    ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND17]], [[C]](s32)
+    ; CHECK-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[AND16]], [[SHL11]]
+    ; CHECK-NEXT: [[BITCAST22:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR11]](s32)
+    ; CHECK-NEXT: [[CONCAT_VECTORS4:%[0-9]+]]:_(<8 x s16>) = G_CONCAT_VECTORS [[BITCAST19]](<2 x s16>), [[BITCAST20]](<2 x s16>), [[BITCAST21]](<2 x s16>), [[BITCAST22]](<2 x s16>)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[CONCAT_VECTORS4]](<8 x s16>)
     %0:_(<5 x s16>) = G_IMPLICIT_DEF
     %1:_(<5 x s16>) = G_IMPLICIT_DEF
     %2:_(<5 x s16>) = G_XOR %0, %1
     %4:_(<8 x s16>) = G_IMPLICIT_DEF
     %5:_(<8 x s16>) = G_INSERT %4, %2, 0
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY %5
 ...
 
 ---
 name: test_xor_v3s8
 body: |
   bb.0:
 
     ; CHECK-LABEL: name: test_xor_v3s8
-    ; CHECK: [[DEF:%[0-9]+]]:_(<3 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<3 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[DEF2:%[0-9]+]]:_(<4 x s8>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF2]], [[DEF]](<3 x s8>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:_(<4 x s8>) = G_INSERT [[DEF2]], [[DEF1]](<3 x s8>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8), [[UV3:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT]](<4 x s8>)
-    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[INSERT1]](<4 x s8>)
-    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s8)
-    ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[UV4]](s8)
-    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s32) = G_XOR [[ANYEXT]], [[ANYEXT1]]
-    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s8) = G_TRUNC [[XOR]](s32)
-    ; CHECK-NEXT: [[ANYEXT2:%[0-9]+]]:_(s32) = G_ANYEXT [[UV1]](s8)
-    ; CHECK-NEXT: [[ANYEXT3:%[0-9]+]]:_(s32) = G_ANYEXT [[UV5]](s8)
-    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(s32) = G_XOR [[ANYEXT2]], [[ANYEXT3]]
-    ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[XOR1]](s32)
-    ; CHECK-NEXT: [[ANYEXT4:%[0-9]+]]:_(s32) = G_ANYEXT [[UV2]](s8)
-    ; CHECK-NEXT: [[ANYEXT5:%[0-9]+]]:_(s32) = G_ANYEXT [[UV6]](s8)
-    ; CHECK-NEXT: [[XOR2:%[0-9]+]]:_(s32) = G_XOR [[ANYEXT4]], [[ANYEXT5]]
-    ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:_(s8) = G_TRUNC [[XOR2]](s32)
-    ; CHECK-NEXT: [[ANYEXT6:%[0-9]+]]:_(s32) = G_ANYEXT [[UV3]](s8)
-    ; CHECK-NEXT: [[ANYEXT7:%[0-9]+]]:_(s32) = G_ANYEXT [[UV7]](s8)
-    ; CHECK-NEXT: [[XOR3:%[0-9]+]]:_(s32) = G_XOR [[ANYEXT6]], [[ANYEXT7]]
-    ; CHECK-NEXT: [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[XOR3]](s32)
-    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s8>) = G_BUILD_VECTOR [[TRUNC]](s8), [[TRUNC1]](s8), [[TRUNC2]](s8), [[TRUNC3]](s8)
-    ; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<12 x s8>) = G_CONCAT_VECTORS [[BUILD_VECTOR]](<4 x s8>), [[DEF2]](<4 x s8>), [[DEF2]](<4 x s8>)
-    ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(<3 x s8>), [[UV9:%[0-9]+]]:_(<3 x s8>), [[UV10:%[0-9]+]]:_(<3 x s8>), [[UV11:%[0-9]+]]:_(<3 x s8>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<12 x s8>)
-    ; CHECK-NEXT: [[ANYEXT8:%[0-9]+]]:_(<3 x s32>) = G_ANYEXT [[UV8]](<3 x s8>)
-    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[ANYEXT8]](<3 x s32>)
+    ; CHECK: [[DEF:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF]](<4 x s32>)
+    ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](<4 x s32>)
+    ; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s32) = G_XOR [[UV]], [[UV4]]
+    ; CHECK-NEXT: [[XOR1:%[0-9]+]]:_(s32) = G_XOR [[UV1]], [[UV5]]
+    ; CHECK-NEXT: [[XOR2:%[0-9]+]]:_(s32) = G_XOR [[UV2]], [[UV6]]
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[XOR]](s32), [[XOR1]](s32), [[XOR2]](s32)
+    ; CHECK-NEXT: $vgpr0_vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(<3 x s8>) = G_IMPLICIT_DEF
     %1:_(<3 x s8>) = G_IMPLICIT_DEF
     %2:_(<3 x s8>) = G_XOR %0, %1
     %3:_(<3 x s32>) = G_ANYEXT %2
     $vgpr0_vgpr1_vgpr2 = COPY %3
 ...
 
 ---

llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-zext.mir

 ---
 name: test_zext_v3s16_to_v3s32
 body: |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: test_zext_v3s16_to_v3s32
     ; CHECK: [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:_(<3 x s16>) = G_EXTRACT [[COPY]](<4 x s16>), 0
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF]], [[EXTRACT]](<3 x s16>), 0
-    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
     ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV]](<2 x s16>)
     ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
     ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
     ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV1]](<2 x s16>)
     ; CHECK-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 65535
     ; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[BITCAST]], [[C1]]
     ; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[LSHR]], [[C1]]
     ; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[BITCAST1]], [[C1]]
bb.0:
     ; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND9]](s64)
     ; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[UV4]], [[C2]](s32)
     ; CHECK-NEXT: [[LSHR5:%[0-9]+]]:_(s32) = G_LSHR [[UV5]], [[C2]](s32)
     ; CHECK-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND10]](s64)
     ; CHECK-NEXT: [[LSHR6:%[0-9]+]]:_(s32) = G_LSHR [[UV6]], [[C2]](s32)
     ; CHECK-NEXT: [[DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
     ; CHECK-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
     ; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[UV8]], [[C2]](s32)
-    ; CHECK-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[LSHR8:%[0-9]+]]:_(s32) = G_LSHR [[UV10]], [[C2]](s32)
-    ; CHECK-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[LSHR9:%[0-9]+]]:_(s32) = G_LSHR [[UV12]], [[C2]](s32)
-    ; CHECK-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[DEF1]](s64)
-    ; CHECK-NEXT: [[LSHR10:%[0-9]+]]:_(s32) = G_LSHR [[UV14]], [[C2]](s32)
     ; CHECK-NEXT: [[AND11:%[0-9]+]]:_(s32) = G_AND [[UV4]], [[C3]]
     ; CHECK-NEXT: [[AND12:%[0-9]+]]:_(s32) = G_AND [[LSHR4]], [[C3]]
     ; CHECK-NEXT: [[SHL6:%[0-9]+]]:_(s32) = G_SHL [[AND12]], [[C2]](s32)
     ; CHECK-NEXT: [[OR6:%[0-9]+]]:_(s32) = G_OR [[AND11]], [[SHL6]]
     ; CHECK-NEXT: [[AND13:%[0-9]+]]:_(s32) = G_AND [[UV5]], [[C3]]
     ; CHECK-NEXT: [[AND14:%[0-9]+]]:_(s32) = G_AND [[LSHR5]], [[C3]]
     ; CHECK-NEXT: [[SHL7:%[0-9]+]]:_(s32) = G_SHL [[AND14]], [[C2]](s32)
     ; CHECK-NEXT: [[OR7:%[0-9]+]]:_(s32) = G_OR [[AND13]], [[SHL7]]
     ; CHECK-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR6]](s32), [[OR7]](s32)
     ; CHECK-NEXT: [[AND15:%[0-9]+]]:_(s32) = G_AND [[UV6]], [[C3]]
     ; CHECK-NEXT: [[AND16:%[0-9]+]]:_(s32) = G_AND [[LSHR6]], [[C3]]
     ; CHECK-NEXT: [[SHL8:%[0-9]+]]:_(s32) = G_SHL [[AND16]], [[C2]](s32)
     ; CHECK-NEXT: [[OR8:%[0-9]+]]:_(s32) = G_OR [[AND15]], [[SHL8]]
     ; CHECK-NEXT: [[AND17:%[0-9]+]]:_(s32) = G_AND [[UV7]], [[C3]]
     ; CHECK-NEXT: [[OR9:%[0-9]+]]:_(s32) = G_OR [[AND17]], [[SHL3]]
     ; CHECK-NEXT: [[MV4:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR8]](s32), [[OR9]](s32)
     ; CHECK-NEXT: [[AND18:%[0-9]+]]:_(s32) = G_AND [[UV8]], [[C3]]
     ; CHECK-NEXT: [[SHL9:%[0-9]+]]:_(s32) = G_SHL [[AND18]], [[C2]](s32)
     ; CHECK-NEXT: [[OR10:%[0-9]+]]:_(s32) = G_OR [[COPY1]], [[SHL9]]
     ; CHECK-NEXT: [[AND19:%[0-9]+]]:_(s32) = G_AND [[LSHR7]], [[C3]]
     ; CHECK-NEXT: [[AND20:%[0-9]+]]:_(s32) = G_AND [[UV9]], [[C3]]
     ; CHECK-NEXT: [[SHL10:%[0-9]+]]:_(s32) = G_SHL [[AND20]], [[C2]](s32)
     ; CHECK-NEXT: [[OR11:%[0-9]+]]:_(s32) = G_OR [[AND19]], [[SHL10]]
     ; CHECK-NEXT: [[MV5:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR10]](s32), [[OR11]](s32)
-    ; CHECK-NEXT: [[AND21:%[0-9]+]]:_(s32) = G_AND [[UV10]], [[C3]]
-    ; CHECK-NEXT: [[AND22:%[0-9]+]]:_(s32) = G_AND [[LSHR8]], [[C3]]
-    ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND22]], [[C2]](s32)
-    ; CHECK-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND21]], [[SHL11]]
-    ; CHECK-NEXT: [[AND23:%[0-9]+]]:_(s32) = G_AND [[UV11]], [[C3]]
-    ; CHECK-NEXT: [[AND24:%[0-9]+]]:_(s32) = G_AND [[UV12]], [[C3]]
-    ; CHECK-NEXT: [[SHL12:%[0-9]+]]:_(s32) = G_SHL [[AND24]], [[C2]](s32)
-    ; CHECK-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[AND23]], [[SHL12]]
+    ; CHECK-NEXT: [[SHL11:%[0-9]+]]:_(s32) = G_SHL [[AND19]], [[C2]](s32)
+    ; CHECK-NEXT: [[OR12:%[0-9]+]]:_(s32) = G_OR [[AND18]], [[SHL11]]
+    ; CHECK-NEXT: [[OR13:%[0-9]+]]:_(s32) = G_OR [[AND20]], [[SHL9]]
     ; CHECK-NEXT: [[MV6:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR12]](s32), [[OR13]](s32)
-    ; CHECK-NEXT: [[AND25:%[0-9]+]]:_(s32) = G_AND [[LSHR9]], [[C3]]
-    ; CHECK-NEXT: [[AND26:%[0-9]+]]:_(s32) = G_AND [[UV13]], [[C3]]
-    ; CHECK-NEXT: [[SHL13:%[0-9]+]]:_(s32) = G_SHL [[AND26]], [[C2]](s32)
-    ; CHECK-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[AND25]], [[SHL13]]
-    ; CHECK-NEXT: [[AND27:%[0-9]+]]:_(s32) = G_AND [[UV14]], [[C3]]
-    ; CHECK-NEXT: [[AND28:%[0-9]+]]:_(s32) = G_AND [[LSHR10]], [[C3]]
-    ; CHECK-NEXT: [[SHL14:%[0-9]+]]:_(s32) = G_SHL [[AND28]], [[C2]](s32)
-    ; CHECK-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[AND27]], [[SHL14]]
-    ; CHECK-NEXT: [[MV7:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR14]](s32), [[OR15]](s32)
-    ; CHECK-NEXT: [[AND29:%[0-9]+]]:_(s32) = G_AND [[UV15]], [[C3]]
-    ; CHECK-NEXT: [[OR16:%[0-9]+]]:_(s32) = G_OR [[AND29]], [[SHL3]]
-    ; CHECK-NEXT: [[OR17:%[0-9]+]]:_(s32) = G_OR [[COPY1]], [[SHL3]]
-    ; CHECK-NEXT: [[MV8:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR16]](s32), [[OR17]](s32)
+    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[OR11]](s32)
+    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[OR12]](s32)
+    ; CHECK-NEXT: [[MV7:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
+    ; CHECK-NEXT: [[OR14:%[0-9]+]]:_(s32) = G_OR [[AND20]], [[SHL3]]
+    ; CHECK-NEXT: [[OR15:%[0-9]+]]:_(s32) = G_OR [[COPY1]], [[SHL3]]
+    ; CHECK-NEXT: [[MV8:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[OR14]](s32), [[OR15]](s32)
     ; CHECK-NEXT: [[MV9:%[0-9]+]]:_(s384) = G_MERGE_VALUES [[MV3]](s64), [[MV4]](s64), [[MV5]](s64), [[MV6]](s64), [[MV7]](s64), [[MV8]](s64)
     ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:_(s112) = G_TRUNC [[MV9]](s384)
     ; CHECK-NEXT: S_ENDPGM 0, implicit [[TRUNC]](s112)
     %0:_(s32) = COPY $vgpr0
     %1:_(s2) = G_TRUNC %0
     %2:_(s112) = G_ZEXT %1
     S_ENDPGM 0, implicit %2
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.1d.d16.ll

 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s2, s4
 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s3, s5
 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s4, s6
 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s5, s7
 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s6, s8
 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s7, s9
 ; GFX8-UNPACKED-NEXT:    image_load v[0:2], v0, s[0:7] dmask:0x7 unorm d16
 ; GFX8-UNPACKED-NEXT:    s_mov_b32 s0, 0xffff
-; GFX8-UNPACKED-NEXT:    s_and_b32 s1, s0, s0
-; GFX8-UNPACKED-NEXT:    s_lshl_b32 s1, s1, 16
-; GFX8-UNPACKED-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-UNPACKED-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-UNPACKED-NEXT:    v_and_b32_e32 v4, s0, v1
-; GFX8-UNPACKED-NEXT:    v_or_b32_sdwa v1, v2, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX8-UNPACKED-NEXT:    v_lshlrev_b32_e32 v2, 16, v4
+; GFX8-UNPACKED-NEXT:    v_and_b32_e32 v3, s0, v1
+; GFX8-UNPACKED-NEXT:    v_and_b32_e32 v1, s0, v2
+; GFX8-UNPACKED-NEXT:    v_lshlrev_b32_e32 v2, 16, v3
 ; GFX8-UNPACKED-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-UNPACKED-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-PACKED-LABEL: load_1d_v3f16_xyz:
 ; GFX8-PACKED:       ; %bb.0:
 ; GFX8-PACKED-NEXT:    s_mov_b32 s0, s2
 ; GFX8-PACKED-NEXT:    s_mov_b32 s1, s3
 ; GFX8-PACKED-NEXT:    s_mov_b32 s2, s4
 ; GFX8-PACKED-NEXT:    s_mov_b32 s3, s5
 ; GFX8-PACKED-NEXT:    s_mov_b32 s4, s6
 ; GFX8-PACKED-NEXT:    s_mov_b32 s5, s7
 ; GFX8-PACKED-NEXT:    s_mov_b32 s6, s8
 ; GFX8-PACKED-NEXT:    s_mov_b32 s7, s9
 ; GFX8-PACKED-NEXT:    image_load v[0:1], v0, s[0:7] dmask:0x7 unorm d16
-; GFX8-PACKED-NEXT:    s_mov_b32 s0, 0xffff
-; GFX8-PACKED-NEXT:    s_and_b32 s0, s0, s0
-; GFX8-PACKED-NEXT:    s_lshl_b32 s0, s0, 16
-; GFX8-PACKED-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-PACKED-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-PACKED-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-PACKED-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX8-PACKED-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX8-PACKED-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-PACKED-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-PACKED-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: load_1d_v3f16_xyz:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_mov_b32 s0, s2
 ; GFX9-NEXT:    s_mov_b32 s1, s3
 ; GFX9-NEXT:    s_mov_b32 s2, s4
 ; GFX9-NEXT:    s_mov_b32 s3, s5
 ; GFX9-NEXT:    s_mov_b32 s4, s6
 ; GFX9-NEXT:    s_mov_b32 s5, s7
 ; GFX9-NEXT:    s_mov_b32 s6, s8
 ; GFX9-NEXT:    s_mov_b32 s7, s9
 ; GFX9-NEXT:    image_load v[0:1], v0, s[0:7] dmask:0x7 unorm d16
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0xffff
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 16, v0
+; GFX9-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
 ; GFX9-NEXT:    v_and_or_b32 v1, v1, v2, s0
+; GFX9-NEXT:    v_and_or_b32 v0, v0, v2, v3
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: load_1d_v3f16_xyz:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_mov_b32 s0, s2
 ; GFX10-NEXT:    s_mov_b32 s1, s3
 ; GFX10-NEXT:    s_mov_b32 s2, s4
 ; GFX10-NEXT:    s_mov_b32 s3, s5
 ; GFX10-NEXT:    s_mov_b32 s4, s6
 ; GFX10-NEXT:    s_mov_b32 s5, s7
 ; GFX10-NEXT:    s_mov_b32 s6, s8
 ; GFX10-NEXT:    s_mov_b32 s7, s9
+; GFX10-NEXT:    v_mov_b32_e32 v3, 0xffff
 ; GFX10-NEXT:    image_load v[0:1], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm d16
 ; GFX10-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    v_and_or_b32 v1, 0xffff, v1, s0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GFX10-NEXT:    v_and_or_b32 v1, v1, v3, s0
+; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GFX10-NEXT:    v_and_or_b32 v0, v0, v3, v2
 ; GFX10-NEXT:    ; return to shader part epilog
   %v = call <3 x half> @llvm.amdgcn.image.load.1d.v3f16.i32(i32 7, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret <3 x half> %v
 }
 
 define amdgpu_ps <4 x half> @load_1d_v4f16_xyzw(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX8-UNPACKED-LABEL: load_1d_v4f16_xyzw:
 ; GFX8-UNPACKED:       ; %bb.0:

llvm/test/CodeGen/AMDGPU/GlobalISel/load-constant.96.ll

 ; GFX7-NOUNALIGNED-NEXT:    ; return to shader part epilog
   %load = load <3 x i32>, <3 x i32> addrspace(4)* %ptr, align 2
   ret <3 x i32> %load
 }
 
 define amdgpu_ps <3 x i32> @s_load_constant_v3i32_align4(<3 x i32> addrspace(4)* inreg %ptr) {
 ; GFX9-LABEL: s_load_constant_v3i32_align4:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s4, s0
-; GFX9-NEXT:    s_mov_b32 s5, s1
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[0:1], 0x8
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b32 s0, s4
+; GFX9-NEXT:    s_mov_b32 s1, s5
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: s_load_constant_v3i32_align4:
 ; GFX7:       ; %bb.0:
-; GFX7-NEXT:    s_mov_b32 s2, s0
-; GFX7-NEXT:    s_mov_b32 s3, s1
-; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[2:3], 0x0
-; GFX7-NEXT:    s_load_dword s2, s[2:3], 0x2
+; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX7-NEXT:    s_load_dword s2, s[0:1], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    s_mov_b32 s0, s4
+; GFX7-NEXT:    s_mov_b32 s1, s5
 ; GFX7-NEXT:    ; return to shader part epilog
   %load = load <3 x i32>, <3 x i32> addrspace(4)* %ptr, align 4
   ret <3 x i32> %load
 }
 
 define amdgpu_ps i96 @s_load_constant_i96_align8(i96 addrspace(4)* inreg %ptr) {
 ; GFX9-LABEL: s_load_constant_i96_align8:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s4, s0
-; GFX9-NEXT:    s_mov_b32 s5, s1
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[0:1], 0x8
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b32 s0, s4
+; GFX9-NEXT:    s_mov_b32 s1, s5
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: s_load_constant_i96_align8:
 ; GFX7:       ; %bb.0:
-; GFX7-NEXT:    s_mov_b32 s2, s0
-; GFX7-NEXT:    s_mov_b32 s3, s1
-; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[2:3], 0x0
-; GFX7-NEXT:    s_load_dword s2, s[2:3], 0x2
+; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX7-NEXT:    s_load_dword s2, s[0:1], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    s_mov_b32 s0, s4
+; GFX7-NEXT:    s_mov_b32 s1, s5
 ; GFX7-NEXT:    ; return to shader part epilog
   %load = load i96, i96 addrspace(4)* %ptr, align 8
   ret i96 %load
 }
 
 define amdgpu_ps <3 x i32> @s_load_constant_v3i32_align8(<3 x i32> addrspace(4)* inreg %ptr) {
 ; GFX9-LABEL: s_load_constant_v3i32_align8:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s4, s0
-; GFX9-NEXT:    s_mov_b32 s5, s1
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[0:1], 0x8
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b32 s0, s4
+; GFX9-NEXT:    s_mov_b32 s1, s5
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: s_load_constant_v3i32_align8:
 ; GFX7:       ; %bb.0:
-; GFX7-NEXT:    s_mov_b32 s2, s0
-; GFX7-NEXT:    s_mov_b32 s3, s1
-; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[2:3], 0x0
-; GFX7-NEXT:    s_load_dword s2, s[2:3], 0x2
+; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX7-NEXT:    s_load_dword s2, s[0:1], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    s_mov_b32 s0, s4
+; GFX7-NEXT:    s_mov_b32 s1, s5
 ; GFX7-NEXT:    ; return to shader part epilog
   %load = load <3 x i32>, <3 x i32> addrspace(4)* %ptr, align 8
   ret <3 x i32> %load
 }
 
 define amdgpu_ps <3 x i32> @s_load_constant_v6i16_align8(<6 x i16> addrspace(4)* inreg %ptr) {
 ; GFX9-LABEL: s_load_constant_v6i16_align8:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b32 s4, s0
-; GFX9-NEXT:    s_mov_b32 s5, s1
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX9-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX9-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX9-NEXT:    s_load_dword s2, s[0:1], 0x8
 ; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    s_mov_b32 s0, s4
+; GFX9-NEXT:    s_mov_b32 s1, s5
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX7-LABEL: s_load_constant_v6i16_align8:
 ; GFX7:       ; %bb.0:
-; GFX7-NEXT:    s_mov_b32 s2, s0
-; GFX7-NEXT:    s_mov_b32 s3, s1
-; GFX7-NEXT:    s_load_dwordx2 s[0:1], s[2:3], 0x0
-; GFX7-NEXT:    s_load_dword s2, s[2:3], 0x2
+; GFX7-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX7-NEXT:    s_load_dword s2, s[0:1], 0x2
 ; GFX7-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX7-NEXT:    s_mov_b32 s0, s4
+; GFX7-NEXT:    s_mov_b32 s1, s5
 ; GFX7-NEXT:    ; return to shader part epilog
   %load = load <6 x i16>, <6 x i16> addrspace(4)* %ptr, align 8
   %cast = bitcast <6 x i16> %load to <3 x i32>
   ret <3 x i32> %cast
 }
 
 define amdgpu_ps <12 x i8> @s_load_constant_v12i8_align8(<12 x i8> addrspace(4)* inreg %ptr) {
 ; GFX9-LABEL: s_load_constant_v12i8_align8:

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.s.buffer.load.ll

define amdgpu_ps <3 x i32> @s_buffer_load_v3i32(<4 x i32> inreg %rsrc, i32 inreg %soffset) {
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr6
   ; CHECK-NEXT:   [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:sgpr(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY4]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:sgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:sgpr(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-  ; CHECK-NEXT:   [[BITCAST:%[0-9]+]]:sgpr(s384) = G_BITCAST [[CONCAT_VECTORS]](<12 x s32>)
-  ; CHECK-NEXT:   [[TRUNC:%[0-9]+]]:sgpr(s96) = G_TRUNC [[BITCAST]](s384)
-  ; CHECK-NEXT:   [[BITCAST1:%[0-9]+]]:sgpr(<3 x s32>) = G_BITCAST [[TRUNC]](s96)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32), [[UV2:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[BITCAST1]](<3 x s32>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32), [[UV2:%[0-9]+]]:sgpr(s32), [[UV3:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
   ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[UV]](s32)
   ; CHECK-NEXT:   [[INT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY5]](s32)
   ; CHECK-NEXT:   $sgpr0 = COPY [[INT]](s32)
   ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[UV1]](s32)
   ; CHECK-NEXT:   [[INT1:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY6]](s32)
   ; CHECK-NEXT:   $sgpr1 = COPY [[INT1]](s32)
   ; CHECK-NEXT:   [[COPY7:%[0-9]+]]:vgpr(s32) = COPY [[UV2]](s32)
   ; CHECK-NEXT:   [[INT2:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY7]](s32)
   ; CHECK-NEXT:   $sgpr2 = COPY [[INT2]](s32)
   ; CHECK-NEXT:   SI_RETURN_TO_EPILOG implicit $sgpr0, implicit $sgpr1, implicit $sgpr2
   ; GREEDY-LABEL: name: s_buffer_load_v3i32
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6
   ; GREEDY-NEXT: {{  $}}
   ; GREEDY-NEXT:   [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY-NEXT:   [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; GREEDY-NEXT:   [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; GREEDY-NEXT:   [[COPY3:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; GREEDY-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY-NEXT:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr6
   ; GREEDY-NEXT:   [[AMDGPU_S_BUFFER_LOAD:%[0-9]+]]:sgpr(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY4]](s32), 0 :: (dereferenceable invariant load (s96), align 4)
-  ; GREEDY-NEXT:   [[DEF:%[0-9]+]]:sgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:sgpr(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>), [[DEF]](<4 x s32>), [[DEF]](<4 x s32>)
-  ; GREEDY-NEXT:   [[BITCAST:%[0-9]+]]:sgpr(s384) = G_BITCAST [[CONCAT_VECTORS]](<12 x s32>)
-  ; GREEDY-NEXT:   [[TRUNC:%[0-9]+]]:sgpr(s96) = G_TRUNC [[BITCAST]](s384)
-  ; GREEDY-NEXT:   [[BITCAST1:%[0-9]+]]:sgpr(<3 x s32>) = G_BITCAST [[TRUNC]](s96)
-  ; GREEDY-NEXT:   [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32), [[UV2:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[BITCAST1]](<3 x s32>)
+  ; GREEDY-NEXT:   [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32), [[UV2:%[0-9]+]]:sgpr(s32), [[UV3:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[AMDGPU_S_BUFFER_LOAD]](<4 x s32>)
   ; GREEDY-NEXT:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[UV]](s32)
   ; GREEDY-NEXT:   [[INT:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY5]](s32)
   ; GREEDY-NEXT:   $sgpr0 = COPY [[INT]](s32)
   ; GREEDY-NEXT:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[UV1]](s32)
   ; GREEDY-NEXT:   [[INT1:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY6]](s32)
   ; GREEDY-NEXT:   $sgpr1 = COPY [[INT1]](s32)
   ; GREEDY-NEXT:   [[COPY7:%[0-9]+]]:vgpr(s32) = COPY [[UV2]](s32)
   ; GREEDY-NEXT:   [[INT2:%[0-9]+]]:sgpr(s32) = G_INTRINSIC intrinsic(@llvm.amdgcn.readfirstlane), [[COPY7]](s32)
define amdgpu_ps <3 x float> @s_buffer_load_v3f32_vgpr_offset(<4 x i32> inreg %rsrc, i32 %soffset) {
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; CHECK-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK-NEXT:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK-NEXT:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
   ; CHECK-NEXT:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK-NEXT:   [[DEF:%[0-9]+]]:sgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:vgpr(<4 x s32>) = COPY [[DEF]](<4 x s32>)
-  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:vgpr(<4 x s32>) = COPY [[DEF]](<4 x s32>)
-  ; CHECK-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[COPY5]](<4 x s32>), [[COPY6]](<4 x s32>)
-  ; CHECK-NEXT:   [[BITCAST:%[0-9]+]]:vgpr(s384) = G_BITCAST [[CONCAT_VECTORS]](<12 x s32>)
-  ; CHECK-NEXT:   [[TRUNC:%[0-9]+]]:vgpr(s96) = G_TRUNC [[BITCAST]](s384)
-  ; CHECK-NEXT:   [[BITCAST1:%[0-9]+]]:vgpr(<3 x s32>) = G_BITCAST [[TRUNC]](s96)
-  ; CHECK-NEXT:   [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32), [[UV2:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[BITCAST1]](<3 x s32>)
+  ; CHECK-NEXT:   [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32), [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[AMDGPU_BUFFER_LOAD]](<4 x s32>)
   ; CHECK-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; CHECK-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; CHECK-NEXT:   $vgpr2 = COPY [[UV2]](s32)
   ; CHECK-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
   ; GREEDY-LABEL: name: s_buffer_load_v3f32_vgpr_offset
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY-NEXT:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $vgpr0
   ; GREEDY-NEXT: {{  $}}
   ; GREEDY-NEXT:   [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY-NEXT:   [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; GREEDY-NEXT:   [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; GREEDY-NEXT:   [[COPY3:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; GREEDY-NEXT:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY-NEXT:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY-NEXT:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; GREEDY-NEXT:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
   ; GREEDY-NEXT:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY-NEXT:   [[DEF:%[0-9]+]]:sgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY-NEXT:   [[COPY5:%[0-9]+]]:vgpr(<4 x s32>) = COPY [[DEF]](<4 x s32>)
-  ; GREEDY-NEXT:   [[COPY6:%[0-9]+]]:vgpr(<4 x s32>) = COPY [[DEF]](<4 x s32>)
-  ; GREEDY-NEXT:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<12 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[COPY5]](<4 x s32>), [[COPY6]](<4 x s32>)
-  ; GREEDY-NEXT:   [[BITCAST:%[0-9]+]]:vgpr(s384) = G_BITCAST [[CONCAT_VECTORS]](<12 x s32>)
-  ; GREEDY-NEXT:   [[TRUNC:%[0-9]+]]:vgpr(s96) = G_TRUNC [[BITCAST]](s384)
-  ; GREEDY-NEXT:   [[BITCAST1:%[0-9]+]]:vgpr(<3 x s32>) = G_BITCAST [[TRUNC]](s96)
-  ; GREEDY-NEXT:   [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32), [[UV2:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[BITCAST1]](<3 x s32>)
+  ; GREEDY-NEXT:   [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32), [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[AMDGPU_BUFFER_LOAD]](<4 x s32>)
   ; GREEDY-NEXT:   $vgpr0 = COPY [[UV]](s32)
   ; GREEDY-NEXT:   $vgpr1 = COPY [[UV1]](s32)
   ; GREEDY-NEXT:   $vgpr2 = COPY [[UV2]](s32)
   ; GREEDY-NEXT:   SI_RETURN_TO_EPILOG implicit $vgpr0, implicit $vgpr1, implicit $vgpr2
   %val = call <3 x float> @llvm.amdgcn.s.buffer.load.v3f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret <3 x float> %val
 }
 

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-load.mir

   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_v3i32_align4
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<2 x s32>) = G_LOAD [[COPY]](p4) :: (invariant load (<2 x s32>), align 4, addrspace 4)
     ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:sgpr(s32) = G_LOAD [[PTR_ADD]](p4) :: (invariant load (s32) from unknown-address + 8, addrspace 4)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:sgpr(<3 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[DEF]], [[LOAD]](<2 x s32>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[INSERT]], [[LOAD1]](s32), 64
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[INSERT1]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<3 x s32>) = G_LOAD %0 :: (invariant load (<3 x s32>), addrspace 4, align 4)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_v3i32_align8
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_v3i32_align8
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<2 x s32>) = G_LOAD [[COPY]](p4) :: (invariant load (<2 x s32>), addrspace 4)
     ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:sgpr(s32) = G_LOAD [[PTR_ADD]](p4) :: (invariant load (s32) from unknown-address + 8, align 8, addrspace 4)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:sgpr(<3 x s32>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[DEF]], [[LOAD]](<2 x s32>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[INSERT]], [[LOAD1]](s32), 64
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[INSERT1]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<3 x s32>) = G_LOAD %0 :: (invariant load (<3 x s32>), addrspace 4, align 8)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_v3i32_align16
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_v3i32_align16
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<4 x s32>) = G_LOAD [[COPY]](p4) :: (invariant load (<4 x s32>), addrspace 4)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:sgpr(<3 x s32>) = G_EXTRACT [[LOAD]](<4 x s32>), 0
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[EXTRACT]](<3 x s32>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32), [[UV2:%[0-9]+]]:sgpr(s32), [[UV3:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](<4 x s32>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[UV2]](s32)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<3 x s32>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<3 x s32>) = G_LOAD %0 :: (invariant load (<3 x s32>), addrspace 4, align 16)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_v6i16_align4
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_v6i16_align4
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<4 x s16>) = G_LOAD [[COPY]](p4) :: (invariant load (<4 x s16>), align 4, addrspace 4)
     ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:sgpr(<2 x s16>) = G_LOAD [[PTR_ADD]](p4) :: (invariant load (<2 x s16>) from unknown-address + 8, addrspace 4)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:sgpr(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:sgpr(<6 x s16>) = G_INSERT [[DEF]], [[LOAD]](<4 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:sgpr(<6 x s16>) = G_INSERT [[INSERT]], [[LOAD1]](<2 x s16>), 64
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[INSERT1]](<6 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s16), [[UV1:%[0-9]+]]:sgpr(s16), [[UV2:%[0-9]+]]:sgpr(s16), [[UV3:%[0-9]+]]:sgpr(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:sgpr(s16), [[UV5:%[0-9]+]]:sgpr(s16) = G_UNMERGE_VALUES [[LOAD1]](<2 x s16>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<6 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<6 x s16>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<6 x s16>) = G_LOAD %0 :: (invariant load (<6 x s16>), addrspace 4, align 4)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_v6i16_align8
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_v6i16_align8
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<4 x s16>) = G_LOAD [[COPY]](p4) :: (invariant load (<4 x s16>), addrspace 4)
     ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:sgpr(<2 x s16>) = G_LOAD [[PTR_ADD]](p4) :: (invariant load (<2 x s16>) from unknown-address + 8, align 8, addrspace 4)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:sgpr(<6 x s16>) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:sgpr(<6 x s16>) = G_INSERT [[DEF]], [[LOAD]](<4 x s16>), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:sgpr(<6 x s16>) = G_INSERT [[INSERT]], [[LOAD1]](<2 x s16>), 64
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[INSERT1]](<6 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s16), [[UV1:%[0-9]+]]:sgpr(s16), [[UV2:%[0-9]+]]:sgpr(s16), [[UV3:%[0-9]+]]:sgpr(s16) = G_UNMERGE_VALUES [[LOAD]](<4 x s16>)
+    ; CHECK-NEXT: [[UV4:%[0-9]+]]:sgpr(s16), [[UV5:%[0-9]+]]:sgpr(s16) = G_UNMERGE_VALUES [[LOAD1]](<2 x s16>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<6 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<6 x s16>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<6 x s16>) = G_LOAD %0 :: (invariant load (<6 x s16>), addrspace 4, align 8)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_v6i16_align16
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_v6i16_align16
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(<8 x s16>) = G_LOAD [[COPY]](p4) :: (invariant load (<8 x s16>), addrspace 4)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:sgpr(<6 x s16>) = G_EXTRACT [[LOAD]](<8 x s16>), 0
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[EXTRACT]](<6 x s16>)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s16), [[UV1:%[0-9]+]]:sgpr(s16), [[UV2:%[0-9]+]]:sgpr(s16), [[UV3:%[0-9]+]]:sgpr(s16), [[UV4:%[0-9]+]]:sgpr(s16), [[UV5:%[0-9]+]]:sgpr(s16), [[UV6:%[0-9]+]]:sgpr(s16), [[UV7:%[0-9]+]]:sgpr(s16) = G_UNMERGE_VALUES [[LOAD]](<8 x s16>)
+    ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<6 x s16>) = G_BUILD_VECTOR [[UV]](s16), [[UV1]](s16), [[UV2]](s16), [[UV3]](s16), [[UV4]](s16), [[UV5]](s16)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR]](<6 x s16>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<6 x s16>) = G_LOAD %0 :: (invariant load (<6 x s16>), addrspace 4, align 16)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_i96_align4
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_i96_align4
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(s64) = G_LOAD [[COPY]](p4) :: (invariant load (s64), align 4, addrspace 4)
     ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:sgpr(s32) = G_LOAD [[PTR_ADD]](p4) :: (invariant load (s32) from unknown-address + 8, addrspace 4)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:sgpr(s96) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:sgpr(s96) = G_INSERT [[DEF]], [[LOAD]](s64), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:sgpr(s96) = G_INSERT [[INSERT]], [[LOAD1]](s32), 64
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[INSERT1]](s96)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:sgpr(s96) = G_MERGE_VALUES [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[MV]](s96)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(s96) = G_LOAD %0 :: (invariant load (s96), addrspace 4, align 4)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_i96_align8
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_i96_align8
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(s64) = G_LOAD [[COPY]](p4) :: (invariant load (s64), addrspace 4)
     ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; CHECK-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; CHECK-NEXT: [[LOAD1:%[0-9]+]]:sgpr(s32) = G_LOAD [[PTR_ADD]](p4) :: (invariant load (s32) from unknown-address + 8, align 8, addrspace 4)
-    ; CHECK-NEXT: [[DEF:%[0-9]+]]:sgpr(s96) = G_IMPLICIT_DEF
-    ; CHECK-NEXT: [[INSERT:%[0-9]+]]:sgpr(s96) = G_INSERT [[DEF]], [[LOAD]](s64), 0
-    ; CHECK-NEXT: [[INSERT1:%[0-9]+]]:sgpr(s96) = G_INSERT [[INSERT]], [[LOAD1]](s32), 64
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[INSERT1]](s96)
+    ; CHECK-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](s64)
+    ; CHECK-NEXT: [[MV:%[0-9]+]]:sgpr(s96) = G_MERGE_VALUES [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[MV]](s96)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(s96) = G_LOAD %0 :: (invariant load (s96), addrspace 4, align 8)
     S_ENDPGM 0, implicit %1
 ...
 
 ---
 name: load_constant_i96_align16
 legalized: true
 
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
     ; CHECK-LABEL: name: load_constant_i96_align16
     ; CHECK: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; CHECK-NEXT: [[LOAD:%[0-9]+]]:sgpr(s128) = G_LOAD [[COPY]](p4) :: (invariant load (s128), addrspace 4)
-    ; CHECK-NEXT: [[EXTRACT:%[0-9]+]]:sgpr(s96) = G_EXTRACT [[LOAD]](s128), 0
-    ; CHECK-NEXT: S_ENDPGM 0, implicit [[EXTRACT]](s96)
+    ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s96) = G_TRUNC [[LOAD]](s128)
+    ; CHECK-NEXT: S_ENDPGM 0, implicit [[TRUNC]](s96)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(s96) = G_LOAD %0 :: (invariant load (s96), addrspace 4, align 16)
     S_ENDPGM 0, implicit %1
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-split-scalar-load-metadata.mir

bb.0:
     liveins: $sgpr0_sgpr1
 
     ; SI-LABEL: name: split_smrd_load_range
     ; SI: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:sgpr(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), addrspace 4)
     ; SI-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; SI-NEXT: [[LOAD1:%[0-9]+]]:sgpr(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s32) from unknown-address + 8, align 8, addrspace 4)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:sgpr(<3 x s32>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[DEF]], [[LOAD]](<2 x s32>), 0
-    ; SI-NEXT: [[INSERT1:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[INSERT]], [[LOAD1]](s32), 64
-    ; SI-NEXT: $sgpr0_sgpr1_sgpr2 = COPY [[INSERT1]](<3 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
+    ; SI-NEXT: $sgpr0_sgpr1_sgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 8, addrspace 4, !range !0)
     $sgpr0_sgpr1_sgpr2 = COPY %1
 
 ...
 
 ---
 name: split_smrd_load_tbaa
 legalized: true
 body: |
   bb.0:
     liveins: $sgpr0_sgpr1
 
     ; SI-LABEL: name: split_smrd_load_tbaa
     ; SI: [[COPY:%[0-9]+]]:sgpr(p4) = COPY $sgpr0_sgpr1
     ; SI-NEXT: [[LOAD:%[0-9]+]]:sgpr(<2 x s32>) = G_LOAD [[COPY]](p4) :: (load (<2 x s32>), !tbaa !2, addrspace 4)
     ; SI-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 8
     ; SI-NEXT: [[PTR_ADD:%[0-9]+]]:sgpr(p4) = G_PTR_ADD [[COPY]], [[C]](s64)
     ; SI-NEXT: [[LOAD1:%[0-9]+]]:sgpr(s32) = G_LOAD [[PTR_ADD]](p4) :: (load (s32) from unknown-address + 8, align 8, !tbaa !2, addrspace 4)
-    ; SI-NEXT: [[DEF:%[0-9]+]]:sgpr(<3 x s32>) = G_IMPLICIT_DEF
-    ; SI-NEXT: [[INSERT:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[DEF]], [[LOAD]](<2 x s32>), 0
-    ; SI-NEXT: [[INSERT1:%[0-9]+]]:sgpr(<3 x s32>) = G_INSERT [[INSERT]], [[LOAD1]](s32), 64
-    ; SI-NEXT: $sgpr0_sgpr1_sgpr2 = COPY [[INSERT1]](<3 x s32>)
+    ; SI-NEXT: [[UV:%[0-9]+]]:sgpr(s32), [[UV1:%[0-9]+]]:sgpr(s32) = G_UNMERGE_VALUES [[LOAD]](<2 x s32>)
+    ; SI-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<3 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[UV1]](s32), [[LOAD1]](s32)
+    ; SI-NEXT: $sgpr0_sgpr1_sgpr2 = COPY [[BUILD_VECTOR]](<3 x s32>)
     %0:_(p4) = COPY $sgpr0_sgpr1
     %1:_(<3 x s32>) = G_LOAD %0 :: (load (<3 x s32>), align 8, addrspace 4, !tbaa !1)
     $sgpr0_sgpr1_sgpr2 = COPY %1
 
 ...

llvm/test/CodeGen/AMDGPU/GlobalISel/sdiv.i64.ll

   %result = sdiv <2 x i64> %num, %den
   ret <2 x i64> %result
 }
 
 define i64 @v_sdiv_i64_pow2k_denom(i64 %num) {
 ; CHECK-LABEL: v_sdiv_i64_pow2k_denom:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, 0x1000
+; CHECK-NEXT:    s_movk_i32 s4, 0x1000
+; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, s4
 ; CHECK-NEXT:    v_cvt_f32_ubyte0_e32 v3, 0
-; CHECK-NEXT:    s_movk_i32 s4, 0xf000
-; CHECK-NEXT:    s_movk_i32 s6, 0x1000
+; CHECK-NEXT:    s_movk_i32 s5, 0xf000
+; CHECK-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0x4f800000, v3
 ; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; CHECK-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
 ; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v3
 ; CHECK-NEXT:    v_mul_f32_e32 v2, 0x5f7ffffc, v2
 ; CHECK-NEXT:    v_mul_f32_e32 v4, 0x2f800000, v2
 ; CHECK-NEXT:    v_trunc_f32_e32 v4, v4
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v3

 ; CHECK-NEXT:    v_mul_hi_u32 v5, v4, v5
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
-; CHECK-NEXT:    s_bfe_i32 s7, -1, 0x10000
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc

 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, v1, v2
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v0, v4
-; CHECK-NEXT:    v_mul_hi_u32 v7, v0, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; CHECK-NEXT:    v_mul_hi_u32 v2, v1, v2
+; CHECK-NEXT:    v_mov_b32_e32 v7, 0x1000
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
+; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v7, v1, v4
+; CHECK-NEXT:    v_mul_lo_u32 v8, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v6, v0, v4
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v7, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v8, v2
+; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
+; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
 ; CHECK-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v5, 0, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s6, v4
-; CHECK-NEXT:    v_mul_hi_u32 v8, s6, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s6, v2
+; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
+; CHECK-NEXT:    v_mul_hi_u32 v9, s4, v2
+; CHECK-NEXT:    v_mul_lo_u32 v8, s4, v2
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
+; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
+; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v8
 ; CHECK-NEXT:    v_subb_u32_e64 v6, s[4:5], v1, v5, vcc
 ; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v5
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_cmp_le_u32_e64 s[4:5], s6, v0
-; CHECK-NEXT:    v_subrev_i32_e32 v0, vcc, s6, v0
+; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v7
+; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, -1, s[4:5]
-; CHECK-NEXT:    v_mov_b32_e32 v7, s7
+; CHECK-NEXT:    v_mov_b32_e32 v8, s6
 ; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v6
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, 1, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v5, v7, v5, s[4:5]
-; CHECK-NEXT:    v_addc_u32_e32 v7, vcc, 0, v4, vcc
+; CHECK-NEXT:    v_cndmask_b32_e64 v5, v8, v5, s[4:5]
+; CHECK-NEXT:    v_addc_u32_e32 v8, vcc, 0, v4, vcc
 ; CHECK-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CHECK-NEXT:    v_cmp_le_u32_e32 vcc, s6, v0
+; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
-; CHECK-NEXT:    v_mov_b32_e32 v8, s4
+; CHECK-NEXT:    v_mov_b32_e32 v7, s4
 ; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, v8, v0, vcc
+; CHECK-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v1, vcc, 1, v6
-; CHECK-NEXT:    v_addc_u32_e32 v8, vcc, 0, v7, vcc
+; CHECK-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
 ; CHECK-NEXT:    v_cndmask_b32_e32 v0, v6, v1, vcc
-; CHECK-NEXT:    v_cndmask_b32_e32 v1, v7, v8, vcc
+; CHECK-NEXT:    v_cndmask_b32_e32 v1, v8, v7, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v5
 ; CHECK-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v4, v1, vcc
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3
 ; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v3
 ; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v3
 ; CHECK-NEXT:    v_subb_u32_e32 v1, vcc, v1, v3, vcc
 ; CHECK-NEXT:    s_setpc_b64 s[30:31]

 ; GISEL-NEXT:    v_xor_b32_e32 v3, v3, v4
 ; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v2, v4
 ; GISEL-NEXT:    v_subb_u32_e32 v3, vcc, v3, v4, vcc
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; CGP-LABEL: v_sdiv_v2i64_pow2k_denom:
 ; CGP:       ; %bb.0:
 ; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_cvt_f32_u32_e32 v5, 0x1000
-; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
-; CGP-NEXT:    s_movk_i32 s6, 0xf000
-; CGP-NEXT:    s_movk_i32 s7, 0x1000
-; CGP-NEXT:    v_mov_b32_e32 v4, v5
-; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v6
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v7, v4
+; CGP-NEXT:    s_movk_i32 s6, 0x1000
+; CGP-NEXT:    v_cvt_f32_u32_e32 v4, s6
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
+; CGP-NEXT:    s_movk_i32 s7, 0xf000
+; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v4
 ; CGP-NEXT:    v_ashrrev_i32_e32 v4, 31, v1
 ; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x5f7ffffc, v7
-; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v7
+; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x2f800000, v5
+; CGP-NEXT:    v_trunc_f32_e32 v6, v6
+; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v6
+; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
+; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
+; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, v1, v8
+; CGP-NEXT:    v_mul_lo_u32 v9, v0, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
+; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
+; CGP-NEXT:    v_mov_b32_e32 v5, 0x1000
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v10, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v9, v7
+; CGP-NEXT:    v_mul_hi_u32 v9, v0, v6
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_mul_hi_u32 v6, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v7
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v6
+; CGP-NEXT:    v_mul_hi_u32 v11, s6, v7
+; CGP-NEXT:    v_mul_lo_u32 v10, s6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v10
+; CGP-NEXT:    v_subb_u32_e64 v9, s[4:5], v1, v8, vcc
+; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v8
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v5
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, -1, s[4:5]
+; CGP-NEXT:    v_mov_b32_e32 v10, s8
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v9
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, 1, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v8, v10, v8, s[4:5]
+; CGP-NEXT:    v_addc_u32_e32 v10, vcc, 0, v6, vcc
+; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
+; CGP-NEXT:    v_mov_b32_e32 v11, s4
+; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v11, v0, vcc
+; CGP-NEXT:    v_add_i32_e32 v1, vcc, 1, v9
+; CGP-NEXT:    v_addc_u32_e32 v11, vcc, 0, v10, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v9, v1, vcc
+; CGP-NEXT:    v_cvt_f32_u32_e32 v9, v5
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v10, v11, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v8
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v6, v1, vcc
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
+; CGP-NEXT:    v_mac_f32_e32 v9, 0x4f800000, v6
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v6, v9
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
+; CGP-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
+; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x5f7ffffc, v6
+; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v6
 ; CGP-NEXT:    v_trunc_f32_e32 v8, v8
-; CGP-NEXT:    v_mac_f32_e32 v7, 0xcf800000, v8
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
+; CGP-NEXT:    v_mac_f32_e32 v6, 0xcf800000, v8
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
 ; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v8
-; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
+; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
 ; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
-; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_mac_f32_e32 v5, 0x4f800000, v6
+; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, v1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, v0, v8
-; CGP-NEXT:    v_mul_hi_u32 v11, v0, v7
-; CGP-NEXT:    v_mul_hi_u32 v7, v1, v7
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, v2, v8
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
+; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v4, vcc
+; CGP-NEXT:    v_mul_hi_u32 v4, v2, v6
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v8
+; CGP-NEXT:    v_mul_hi_u32 v6, v3, v6
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v10, v4
+; CGP-NEXT:    v_mul_hi_u32 v10, v2, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
 ; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v11, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_lo_u32 v9, 0, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s7, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s7, v7
-; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v11
-; CGP-NEXT:    v_subb_u32_e64 v10, s[4:5], v1, v9, vcc
-; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v9
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v0
-; CGP-NEXT:    v_subrev_i32_e32 v0, vcc, s7, v0
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v11, s8
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v10
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, 1, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v9, v11, v9, s[4:5]
-; CGP-NEXT:    v_addc_u32_e32 v11, vcc, 0, v8, vcc
-; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v12, s4
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v12, v0, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, 1, v10
-; CGP-NEXT:    v_addc_u32_e32 v12, vcc, 0, v11, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v10, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v11, v12, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x2f800000, v5
-; CGP-NEXT:    v_trunc_f32_e32 v7, v7
-; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v8, v1, vcc
-; CGP-NEXT:    v_ashrrev_i32_e32 v6, 31, v3
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v6, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
+; CGP-NEXT:    v_mul_hi_u32 v8, v3, v8
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v4
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v6
+; CGP-NEXT:    v_mul_hi_u32 v11, s6, v4
+; CGP-NEXT:    v_mul_lo_u32 v10, s6, v4
+; CGP-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, v2, v7
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_hi_u32 v4, v2, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v8, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v7
-; CGP-NEXT:    v_mul_hi_u32 v5, v3, v5
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
-; CGP-NEXT:    v_mul_hi_u32 v9, v2, v7
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_hi_u32 v7, v3, v7
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
-; CGP-NEXT:    v_mul_lo_u32 v7, 0, v4
-; CGP-NEXT:    v_mul_lo_u32 v8, s7, v5
-; CGP-NEXT:    v_mul_hi_u32 v10, s7, v4
-; CGP-NEXT:    v_mul_lo_u32 v9, s7, v4
-; CGP-NEXT:    s_bfe_i32 s6, -1, 0x10000
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v9
-; CGP-NEXT:    v_subb_u32_e64 v8, s[4:5], v3, v7, vcc
-; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v7
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v10
+; CGP-NEXT:    v_subb_u32_e64 v9, s[4:5], v3, v8, vcc
+; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v8
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v2
-; CGP-NEXT:    v_subrev_i32_e32 v2, vcc, s7, v2
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v5
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v5
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v9, s6
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v8
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, 1, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v7, v9, v7, s[4:5]
-; CGP-NEXT:    v_addc_u32_e32 v9, vcc, 0, v5, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, -1, s[4:5]
+; CGP-NEXT:    v_mov_b32_e32 v10, s6
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v9
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, 1, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v8, v10, v8, s[4:5]
+; CGP-NEXT:    v_addc_u32_e32 v10, vcc, 0, v6, vcc
 ; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v2
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v5
 ; CGP-NEXT:    v_cndmask_b32_e64 v2, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v10, s4
+; CGP-NEXT:    v_mov_b32_e32 v5, s4
 ; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v10, v2, vcc
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, 1, v8
-; CGP-NEXT:    v_addc_u32_e32 v10, vcc, 0, v9, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v2, v5, v2, vcc
+; CGP-NEXT:    v_add_i32_e32 v3, vcc, 1, v9
+; CGP-NEXT:    v_addc_u32_e32 v5, vcc, 0, v10, vcc
 ; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v8, v3, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v9, v10, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v7
+; CGP-NEXT:    v_cndmask_b32_e32 v2, v9, v3, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v3, v10, v5, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v8
 ; CGP-NEXT:    v_cndmask_b32_e32 v2, v4, v2, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v5, v3, vcc
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v6, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v3, v6, v3, vcc
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
 ; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv <2 x i64> %num, <i64 4096, i64 4096>
   ret <2 x i64> %result
 }
 
 define i64 @v_sdiv_i64_oddk_denom(i64 %num) {
 ; CHECK-LABEL: v_sdiv_i64_oddk_denom:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, 0x12d8fb
+; CHECK-NEXT:    s_mov_b32 s4, 0x12d8fb
+; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, s4
 ; CHECK-NEXT:    v_cvt_f32_ubyte0_e32 v3, 0
-; CHECK-NEXT:    s_mov_b32 s4, 0xffed2705
-; CHECK-NEXT:    s_mov_b32 s6, 0x12d8fb
+; CHECK-NEXT:    s_mov_b32 s5, 0xffed2705
+; CHECK-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0x4f800000, v3
 ; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; CHECK-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
 ; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v3
 ; CHECK-NEXT:    v_mul_f32_e32 v2, 0x5f7ffffc, v2
 ; CHECK-NEXT:    v_mul_f32_e32 v4, 0x2f800000, v2
 ; CHECK-NEXT:    v_trunc_f32_e32 v4, v4
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v3

 ; CHECK-NEXT:    v_mul_hi_u32 v5, v4, v5
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
-; CHECK-NEXT:    s_bfe_i32 s7, -1, 0x10000
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc

 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, v1, v2
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v0, v4
-; CHECK-NEXT:    v_mul_hi_u32 v7, v0, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; CHECK-NEXT:    v_mul_hi_u32 v2, v1, v2
+; CHECK-NEXT:    v_mov_b32_e32 v7, 0x12d8fb
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
+; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v7, v1, v4
+; CHECK-NEXT:    v_mul_lo_u32 v8, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v6, v0, v4
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v7, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v8, v2
+; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
+; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
 ; CHECK-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v5, 0, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s6, v4
-; CHECK-NEXT:    v_mul_hi_u32 v8, s6, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s6, v2
+; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
+; CHECK-NEXT:    v_mul_hi_u32 v9, s4, v2
+; CHECK-NEXT:    v_mul_lo_u32 v8, s4, v2
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
+; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
+; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v8
 ; CHECK-NEXT:    v_subb_u32_e64 v6, s[4:5], v1, v5, vcc
 ; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v5
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_cmp_le_u32_e64 s[4:5], s6, v0
-; CHECK-NEXT:    v_subrev_i32_e32 v0, vcc, s6, v0
+; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v7
+; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, -1, s[4:5]
-; CHECK-NEXT:    v_mov_b32_e32 v7, s7
+; CHECK-NEXT:    v_mov_b32_e32 v8, s6
 ; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v6
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, 1, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v5, v7, v5, s[4:5]
-; CHECK-NEXT:    v_addc_u32_e32 v7, vcc, 0, v4, vcc
+; CHECK-NEXT:    v_cndmask_b32_e64 v5, v8, v5, s[4:5]
+; CHECK-NEXT:    v_addc_u32_e32 v8, vcc, 0, v4, vcc
 ; CHECK-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CHECK-NEXT:    v_cmp_le_u32_e32 vcc, s6, v0
+; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
-; CHECK-NEXT:    v_mov_b32_e32 v8, s4
+; CHECK-NEXT:    v_mov_b32_e32 v7, s4
 ; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, v8, v0, vcc
+; CHECK-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v1, vcc, 1, v6
-; CHECK-NEXT:    v_addc_u32_e32 v8, vcc, 0, v7, vcc
+; CHECK-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
 ; CHECK-NEXT:    v_cndmask_b32_e32 v0, v6, v1, vcc
-; CHECK-NEXT:    v_cndmask_b32_e32 v1, v7, v8, vcc
+; CHECK-NEXT:    v_cndmask_b32_e32 v1, v8, v7, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v5
 ; CHECK-NEXT:    v_cndmask_b32_e32 v0, v2, v0, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v4, v1, vcc
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3
 ; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v3
 ; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v3
 ; CHECK-NEXT:    v_subb_u32_e32 v1, vcc, v1, v3, vcc
 ; CHECK-NEXT:    s_setpc_b64 s[30:31]

 ; GISEL-NEXT:    v_xor_b32_e32 v3, v3, v4
 ; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v2, v4
 ; GISEL-NEXT:    v_subb_u32_e32 v3, vcc, v3, v4, vcc
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; CGP-LABEL: v_sdiv_v2i64_oddk_denom:
 ; CGP:       ; %bb.0:
 ; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_cvt_f32_u32_e32 v5, 0x12d8fb
-; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
-; CGP-NEXT:    s_mov_b32 s6, 0xffed2705
-; CGP-NEXT:    s_mov_b32 s7, 0x12d8fb
-; CGP-NEXT:    v_mov_b32_e32 v4, v5
-; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v6
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v7, v4
+; CGP-NEXT:    s_mov_b32 s6, 0x12d8fb
+; CGP-NEXT:    v_cvt_f32_u32_e32 v4, s6
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
+; CGP-NEXT:    s_mov_b32 s7, 0xffed2705
+; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v4
 ; CGP-NEXT:    v_ashrrev_i32_e32 v4, 31, v1
 ; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x5f7ffffc, v7
-; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v7
+; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x2f800000, v5
+; CGP-NEXT:    v_trunc_f32_e32 v6, v6
+; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v6
+; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
+; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
+; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, v1, v8
+; CGP-NEXT:    v_mul_lo_u32 v9, v0, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
+; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
+; CGP-NEXT:    v_mov_b32_e32 v5, 0x12d8fb
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v10, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v9, v7
+; CGP-NEXT:    v_mul_hi_u32 v9, v0, v6
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_mul_hi_u32 v6, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v7
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v6
+; CGP-NEXT:    v_mul_hi_u32 v11, s6, v7
+; CGP-NEXT:    v_mul_lo_u32 v10, s6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v10
+; CGP-NEXT:    v_subb_u32_e64 v9, s[4:5], v1, v8, vcc
+; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v8
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v5
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, -1, s[4:5]
+; CGP-NEXT:    v_mov_b32_e32 v10, s8
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v9
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, 1, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v8, v10, v8, s[4:5]
+; CGP-NEXT:    v_addc_u32_e32 v10, vcc, 0, v6, vcc
+; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
+; CGP-NEXT:    v_mov_b32_e32 v11, s4
+; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v11, v0, vcc
+; CGP-NEXT:    v_add_i32_e32 v1, vcc, 1, v9
+; CGP-NEXT:    v_addc_u32_e32 v11, vcc, 0, v10, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v9, v1, vcc
+; CGP-NEXT:    v_cvt_f32_u32_e32 v9, v5
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v10, v11, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v8
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v6, v1, vcc
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
+; CGP-NEXT:    v_mac_f32_e32 v9, 0x4f800000, v6
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v6, v9
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
+; CGP-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
+; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x5f7ffffc, v6
+; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v6
 ; CGP-NEXT:    v_trunc_f32_e32 v8, v8
-; CGP-NEXT:    v_mac_f32_e32 v7, 0xcf800000, v8
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
+; CGP-NEXT:    v_mac_f32_e32 v6, 0xcf800000, v8
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
 ; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v8
-; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
+; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
 ; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
-; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_mac_f32_e32 v5, 0x4f800000, v6
+; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, v1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, v0, v8
-; CGP-NEXT:    v_mul_hi_u32 v11, v0, v7
-; CGP-NEXT:    v_mul_hi_u32 v7, v1, v7
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, v2, v8
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
+; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v4, vcc
+; CGP-NEXT:    v_mul_hi_u32 v4, v2, v6
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v8
+; CGP-NEXT:    v_mul_hi_u32 v6, v3, v6
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v10, v4
+; CGP-NEXT:    v_mul_hi_u32 v10, v2, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
 ; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v11, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_lo_u32 v9, 0, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s7, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s7, v7
-; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v11
-; CGP-NEXT:    v_subb_u32_e64 v10, s[4:5], v1, v9, vcc
-; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v9
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v0
-; CGP-NEXT:    v_subrev_i32_e32 v0, vcc, s7, v0
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v11, s8
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v10
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, 1, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v9, v11, v9, s[4:5]
-; CGP-NEXT:    v_addc_u32_e32 v11, vcc, 0, v8, vcc
-; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v0, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v12, s4
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v12, v0, vcc
-; CGP-NEXT:    v_add_i32_e32 v1, vcc, 1, v10
-; CGP-NEXT:    v_addc_u32_e32 v12, vcc, 0, v11, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v0
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v10, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v11, v12, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v7, v0, vcc
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x2f800000, v5
-; CGP-NEXT:    v_trunc_f32_e32 v7, v7
-; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v8, v1, vcc
-; CGP-NEXT:    v_ashrrev_i32_e32 v6, 31, v3
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v6, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
+; CGP-NEXT:    v_mul_hi_u32 v8, v3, v8
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v4
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v6
+; CGP-NEXT:    v_mul_hi_u32 v11, s6, v4
+; CGP-NEXT:    v_mul_lo_u32 v10, s6, v4
+; CGP-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, v2, v7
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_hi_u32 v4, v2, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v8, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v7
-; CGP-NEXT:    v_mul_hi_u32 v5, v3, v5
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
-; CGP-NEXT:    v_mul_hi_u32 v9, v2, v7
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_hi_u32 v7, v3, v7
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
-; CGP-NEXT:    v_mul_lo_u32 v7, 0, v4
-; CGP-NEXT:    v_mul_lo_u32 v8, s7, v5
-; CGP-NEXT:    v_mul_hi_u32 v10, s7, v4
-; CGP-NEXT:    v_mul_lo_u32 v9, s7, v4
-; CGP-NEXT:    s_bfe_i32 s6, -1, 0x10000
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v9
-; CGP-NEXT:    v_subb_u32_e64 v8, s[4:5], v3, v7, vcc
-; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v7
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v10
+; CGP-NEXT:    v_subb_u32_e64 v9, s[4:5], v3, v8, vcc
+; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v8
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v2
-; CGP-NEXT:    v_subrev_i32_e32 v2, vcc, s7, v2
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v5
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v5
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v9, s6
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v8
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, 1, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v7, v9, v7, s[4:5]
-; CGP-NEXT:    v_addc_u32_e32 v9, vcc, 0, v5, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, -1, s[4:5]
+; CGP-NEXT:    v_mov_b32_e32 v10, s6
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v9
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, 1, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v8, v10, v8, s[4:5]
+; CGP-NEXT:    v_addc_u32_e32 v10, vcc, 0, v6, vcc
 ; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v2
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v5
 ; CGP-NEXT:    v_cndmask_b32_e64 v2, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v10, s4
+; CGP-NEXT:    v_mov_b32_e32 v5, s4
 ; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v10, v2, vcc
-; CGP-NEXT:    v_add_i32_e32 v3, vcc, 1, v8
-; CGP-NEXT:    v_addc_u32_e32 v10, vcc, 0, v9, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v2, v5, v2, vcc
+; CGP-NEXT:    v_add_i32_e32 v3, vcc, 1, v9
+; CGP-NEXT:    v_addc_u32_e32 v5, vcc, 0, v10, vcc
 ; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v8, v3, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v9, v10, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v7
+; CGP-NEXT:    v_cndmask_b32_e32 v2, v9, v3, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v3, v10, v5, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v8
 ; CGP-NEXT:    v_cndmask_b32_e32 v2, v4, v2, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v5, v3, vcc
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v6, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v3, v6, v3, vcc
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
 ; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = sdiv <2 x i64> %num, <i64 1235195, i64 1235195>
   ret <2 x i64> %result
 }
 
 define i64 @v_sdiv_i64_pow2_shl_denom(i64 %x, i64 %y) {
 ; CHECK-LABEL: v_sdiv_i64_pow2_shl_denom:
 ; CHECK:       ; %bb.0:

llvm/test/CodeGen/AMDGPU/GlobalISel/srem.i64.ll

   %result = srem <2 x i64> %num, %den
   ret <2 x i64> %result
 }
 
 define i64 @v_srem_i64_pow2k_denom(i64 %num) {
 ; CHECK-LABEL: v_srem_i64_pow2k_denom:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, 0x1000
+; CHECK-NEXT:    s_movk_i32 s4, 0x1000
+; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, s4
 ; CHECK-NEXT:    v_cvt_f32_ubyte0_e32 v3, 0
-; CHECK-NEXT:    s_movk_i32 s4, 0xf000
-; CHECK-NEXT:    s_movk_i32 s6, 0x1000
+; CHECK-NEXT:    s_movk_i32 s5, 0xf000
+; CHECK-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0x4f800000, v3
 ; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; CHECK-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
 ; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v3
 ; CHECK-NEXT:    v_mul_f32_e32 v2, 0x5f7ffffc, v2
 ; CHECK-NEXT:    v_mul_f32_e32 v4, 0x2f800000, v2
 ; CHECK-NEXT:    v_trunc_f32_e32 v4, v4
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v3

 ; CHECK-NEXT:    v_mul_hi_u32 v5, v4, v5
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
-; CHECK-NEXT:    s_bfe_i32 s7, -1, 0x10000
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc

 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, v1, v2
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v0, v4
-; CHECK-NEXT:    v_mul_hi_u32 v7, v0, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; CHECK-NEXT:    v_mul_hi_u32 v2, v1, v2
+; CHECK-NEXT:    v_mov_b32_e32 v7, 0x1000
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
+; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v7, v1, v4
+; CHECK-NEXT:    v_mul_lo_u32 v8, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v6, v0, v4
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v7, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v8, v2
+; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
+; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
 ; CHECK-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v5, 0, v2
-; CHECK-NEXT:    v_mul_lo_u32 v4, s6, v4
-; CHECK-NEXT:    v_mul_lo_u32 v6, s6, v2
-; CHECK-NEXT:    v_mul_hi_u32 v2, s6, v2
+; CHECK-NEXT:    v_mul_lo_u32 v4, s4, v4
+; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v2
+; CHECK-NEXT:    v_mul_hi_u32 v2, s4, v2
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
 ; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v6
 ; CHECK-NEXT:    v_subb_u32_e64 v4, s[4:5], v1, v2, vcc
 ; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v2
-; CHECK-NEXT:    v_cmp_le_u32_e64 s[4:5], s6, v0
+; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v2, 0, -1, s[4:5]
-; CHECK-NEXT:    v_mov_b32_e32 v5, s7
+; CHECK-NEXT:    v_mov_b32_e32 v5, s6
 ; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v4
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e64 v2, v5, v2, s[4:5]
-; CHECK-NEXT:    v_subrev_i32_e32 v5, vcc, s6, v0
+; CHECK-NEXT:    v_sub_i32_e32 v5, vcc, v0, v7
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CHECK-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CHECK-NEXT:    v_cmp_le_u32_e32 vcc, s6, v5
+; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, -1, vcc
-; CHECK-NEXT:    v_mov_b32_e32 v7, s4
+; CHECK-NEXT:    v_mov_b32_e32 v8, s4
 ; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CHECK-NEXT:    v_cndmask_b32_e32 v6, v7, v6, vcc
-; CHECK-NEXT:    v_subrev_i32_e32 v7, vcc, s6, v5
+; CHECK-NEXT:    v_cndmask_b32_e32 v6, v8, v6, vcc
+; CHECK-NEXT:    v_sub_i32_e32 v7, vcc, v5, v7
 ; CHECK-NEXT:    v_subbrev_u32_e32 v8, vcc, 0, v1, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v6
 ; CHECK-NEXT:    v_cndmask_b32_e32 v5, v5, v7, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v1, v8, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
 ; CHECK-NEXT:    v_cndmask_b32_e32 v0, v0, v5, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v4, v1, vcc
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3

 ; GISEL-NEXT:    v_xor_b32_e32 v3, v3, v4
 ; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v2, v4
 ; GISEL-NEXT:    v_subb_u32_e32 v3, vcc, v3, v4, vcc
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; CGP-LABEL: v_srem_v2i64_pow2k_denom:
 ; CGP:       ; %bb.0:
 ; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_cvt_f32_u32_e32 v5, 0x1000
-; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
-; CGP-NEXT:    s_movk_i32 s6, 0xf000
-; CGP-NEXT:    s_movk_i32 s7, 0x1000
-; CGP-NEXT:    v_mov_b32_e32 v4, v5
-; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v6
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v7, v4
+; CGP-NEXT:    s_movk_i32 s6, 0x1000
+; CGP-NEXT:    v_cvt_f32_u32_e32 v4, s6
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
+; CGP-NEXT:    s_movk_i32 s7, 0xf000
+; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v4
 ; CGP-NEXT:    v_ashrrev_i32_e32 v4, 31, v1
 ; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x5f7ffffc, v7
-; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v7
-; CGP-NEXT:    v_trunc_f32_e32 v8, v8
-; CGP-NEXT:    v_mac_f32_e32 v7, 0xcf800000, v8
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x2f800000, v5
+; CGP-NEXT:    v_trunc_f32_e32 v6, v6
+; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v6
+; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
 ; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
 ; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, v1, v8
+; CGP-NEXT:    v_mul_lo_u32 v9, v0, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
+; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
+; CGP-NEXT:    v_mov_b32_e32 v5, 0x1000
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v10, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v9, v7
+; CGP-NEXT:    v_mul_hi_u32 v9, v0, v6
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_mul_hi_u32 v6, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v7
+; CGP-NEXT:    v_mul_lo_u32 v6, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
+; CGP-NEXT:    v_mul_hi_u32 v7, s6, v7
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v7
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v9
+; CGP-NEXT:    v_subb_u32_e64 v7, s[4:5], v1, v6, vcc
+; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v6
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[4:5]
+; CGP-NEXT:    v_mov_b32_e32 v8, s8
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v7
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v6, v8, v6, s[4:5]
+; CGP-NEXT:    v_sub_i32_e32 v8, vcc, v0, v5
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
+; CGP-NEXT:    v_mov_b32_e32 v10, s4
+; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
+; CGP-NEXT:    v_cndmask_b32_e32 v9, v10, v9, vcc
+; CGP-NEXT:    v_sub_i32_e32 v10, vcc, v8, v5
+; CGP-NEXT:    v_subbrev_u32_e32 v11, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
+; CGP-NEXT:    v_cvt_f32_u32_e32 v9, v5
+; CGP-NEXT:    v_cndmask_b32_e32 v8, v8, v10, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v1, v11, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v6
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
+; CGP-NEXT:    v_mac_f32_e32 v9, 0x4f800000, v6
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v6, v9
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v0, v8, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v7, v1, vcc
+; CGP-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x5f7ffffc, v6
+; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v6
+; CGP-NEXT:    v_trunc_f32_e32 v8, v8
+; CGP-NEXT:    v_mac_f32_e32 v6, 0xcf800000, v8
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
+; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
+; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v7, vcc
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
-; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_mac_f32_e32 v5, 0x4f800000, v6
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, v1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, v0, v8
-; CGP-NEXT:    v_mul_hi_u32 v11, v0, v7
-; CGP-NEXT:    v_mul_hi_u32 v7, v1, v7
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v5
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v11, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_lo_u32 v9, 0, v7
-; CGP-NEXT:    v_mul_lo_u32 v8, s7, v8
-; CGP-NEXT:    v_mul_lo_u32 v10, s7, v7
-; CGP-NEXT:    v_mul_hi_u32 v7, s7, v7
-; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v10
-; CGP-NEXT:    v_subb_u32_e64 v8, s[4:5], v1, v7, vcc
-; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v7
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v9, s8
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v8
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v7, v9, v7, s[4:5]
-; CGP-NEXT:    v_subrev_i32_e32 v9, vcc, s7, v0
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v11, s4
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CGP-NEXT:    v_cndmask_b32_e32 v10, v11, v10, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v11, vcc, s7, v9
-; CGP-NEXT:    v_subbrev_u32_e32 v12, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v10
-; CGP-NEXT:    v_cndmask_b32_e32 v9, v9, v11, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v1, v12, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v7
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x2f800000, v5
-; CGP-NEXT:    v_trunc_f32_e32 v7, v7
-; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v0, v9, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v8, v1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_ashrrev_i32_e32 v6, 31, v3
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v6, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
 ; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, v2, v7
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, v2, v8
 ; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
 ; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_hi_u32 v4, v2, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v8, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v7
-; CGP-NEXT:    v_mul_hi_u32 v5, v3, v5
+; CGP-NEXT:    v_mul_hi_u32 v4, v2, v6
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
-; CGP-NEXT:    v_mul_hi_u32 v9, v2, v7
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v8
+; CGP-NEXT:    v_mul_hi_u32 v6, v3, v6
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v10, v4
+; CGP-NEXT:    v_mul_hi_u32 v10, v2, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
 ; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_hi_u32 v7, v3, v7
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
-; CGP-NEXT:    v_mul_lo_u32 v7, 0, v4
-; CGP-NEXT:    v_mul_lo_u32 v5, s7, v5
-; CGP-NEXT:    v_mul_lo_u32 v8, s7, v4
-; CGP-NEXT:    v_mul_hi_u32 v4, s7, v4
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_mul_hi_u32 v8, v3, v8
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v4
+; CGP-NEXT:    v_mul_lo_u32 v6, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v4
+; CGP-NEXT:    v_mul_hi_u32 v4, s6, v4
 ; CGP-NEXT:    s_bfe_i32 s6, -1, 0x10000
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v8
-; CGP-NEXT:    v_subb_u32_e64 v5, s[4:5], v3, v4, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v9
+; CGP-NEXT:    v_subb_u32_e64 v6, s[4:5], v3, v4, vcc
 ; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v4
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v2
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v5
 ; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v7, s6
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v5
+; CGP-NEXT:    v_mov_b32_e32 v8, s6
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v6
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v4, v7, v4, s[4:5]
-; CGP-NEXT:    v_subrev_i32_e32 v7, vcc, s7, v2
+; CGP-NEXT:    v_cndmask_b32_e64 v4, v8, v4, s[4:5]
+; CGP-NEXT:    v_sub_i32_e32 v8, vcc, v2, v5
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
 ; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v9, s4
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
+; CGP-NEXT:    v_mov_b32_e32 v10, s4
 ; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; CGP-NEXT:    v_cndmask_b32_e32 v8, v9, v8, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v9, vcc, s7, v7
+; CGP-NEXT:    v_cndmask_b32_e32 v9, v10, v9, vcc
+; CGP-NEXT:    v_sub_i32_e32 v5, vcc, v8, v5
 ; CGP-NEXT:    v_subbrev_u32_e32 v10, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v8
-; CGP-NEXT:    v_cndmask_b32_e32 v7, v7, v9, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
+; CGP-NEXT:    v_cndmask_b32_e32 v5, v8, v5, vcc
 ; CGP-NEXT:    v_cndmask_b32_e32 v3, v3, v10, vcc
 ; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v4
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v2, v7, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v5, v3, vcc
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v6, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v2, v2, v5, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v3, v6, v3, vcc
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
 ; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = srem <2 x i64> %num, <i64 4096, i64 4096>
   ret <2 x i64> %result
 }
 
 define i64 @v_srem_i64_oddk_denom(i64 %num) {
 ; CHECK-LABEL: v_srem_i64_oddk_denom:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, 0x12d8fb
+; CHECK-NEXT:    s_mov_b32 s4, 0x12d8fb
+; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, s4
 ; CHECK-NEXT:    v_cvt_f32_ubyte0_e32 v3, 0
-; CHECK-NEXT:    s_mov_b32 s4, 0xffed2705
-; CHECK-NEXT:    s_mov_b32 s6, 0x12d8fb
+; CHECK-NEXT:    s_mov_b32 s5, 0xffed2705
+; CHECK-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0x4f800000, v3
 ; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; CHECK-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
 ; CHECK-NEXT:    v_add_i32_e32 v0, vcc, v0, v3
 ; CHECK-NEXT:    v_mul_f32_e32 v2, 0x5f7ffffc, v2
 ; CHECK-NEXT:    v_mul_f32_e32 v4, 0x2f800000, v2
 ; CHECK-NEXT:    v_trunc_f32_e32 v4, v4
 ; CHECK-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; CHECK-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; CHECK-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_xor_b32_e32 v1, v1, v3

 ; CHECK-NEXT:    v_mul_hi_u32 v5, v4, v5
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, -1, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v4
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
-; CHECK-NEXT:    s_bfe_i32 s7, -1, 0x10000
+; CHECK-NEXT:    v_mul_lo_u32 v6, s5, v4
+; CHECK-NEXT:    v_mul_hi_u32 v8, s5, v2
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v2
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v4, v7
 ; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v9, v2, v7
 ; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc

 ; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v5, vcc
 ; CHECK-NEXT:    v_mul_lo_u32 v5, v1, v2
 ; CHECK-NEXT:    v_mul_lo_u32 v6, v0, v4
-; CHECK-NEXT:    v_mul_hi_u32 v7, v0, v2
+; CHECK-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; CHECK-NEXT:    v_mul_hi_u32 v2, v1, v2
+; CHECK-NEXT:    v_mov_b32_e32 v7, 0x12d8fb
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
+; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v7, v1, v4
+; CHECK-NEXT:    v_mul_lo_u32 v8, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_mul_hi_u32 v6, v0, v4
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v7, v2
-; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v8, v2
+; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
+; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
 ; CHECK-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; CHECK-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; CHECK-NEXT:    v_mul_lo_u32 v5, 0, v2
-; CHECK-NEXT:    v_mul_lo_u32 v4, s6, v4
-; CHECK-NEXT:    v_mul_lo_u32 v6, s6, v2
-; CHECK-NEXT:    v_mul_hi_u32 v2, s6, v2
+; CHECK-NEXT:    v_mul_lo_u32 v4, s4, v4
+; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v2
+; CHECK-NEXT:    v_mul_hi_u32 v2, s4, v2
 ; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
 ; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v6
 ; CHECK-NEXT:    v_subb_u32_e64 v4, s[4:5], v1, v2, vcc
 ; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v2
-; CHECK-NEXT:    v_cmp_le_u32_e64 s[4:5], s6, v0
+; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v2, 0, -1, s[4:5]
-; CHECK-NEXT:    v_mov_b32_e32 v5, s7
+; CHECK-NEXT:    v_mov_b32_e32 v5, s6
 ; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v4
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e64 v2, v5, v2, s[4:5]
-; CHECK-NEXT:    v_subrev_i32_e32 v5, vcc, s6, v0
+; CHECK-NEXT:    v_sub_i32_e32 v5, vcc, v0, v7
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CHECK-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CHECK-NEXT:    v_cmp_le_u32_e32 vcc, s6, v5
+; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, -1, vcc
-; CHECK-NEXT:    v_mov_b32_e32 v7, s4
+; CHECK-NEXT:    v_mov_b32_e32 v8, s4
 ; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CHECK-NEXT:    v_cndmask_b32_e32 v6, v7, v6, vcc
-; CHECK-NEXT:    v_subrev_i32_e32 v7, vcc, s6, v5
+; CHECK-NEXT:    v_cndmask_b32_e32 v6, v8, v6, vcc
+; CHECK-NEXT:    v_sub_i32_e32 v7, vcc, v5, v7
 ; CHECK-NEXT:    v_subbrev_u32_e32 v8, vcc, 0, v1, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v6
 ; CHECK-NEXT:    v_cndmask_b32_e32 v5, v5, v7, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v1, v8, vcc
 ; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
 ; CHECK-NEXT:    v_cndmask_b32_e32 v0, v0, v5, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v4, v1, vcc
 ; CHECK-NEXT:    v_xor_b32_e32 v0, v0, v3

 ; GISEL-NEXT:    v_xor_b32_e32 v3, v3, v4
 ; GISEL-NEXT:    v_sub_i32_e32 v2, vcc, v2, v4
 ; GISEL-NEXT:    v_subb_u32_e32 v3, vcc, v3, v4, vcc
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; CGP-LABEL: v_srem_v2i64_oddk_denom:
 ; CGP:       ; %bb.0:
 ; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_cvt_f32_u32_e32 v5, 0x12d8fb
-; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
-; CGP-NEXT:    s_mov_b32 s6, 0xffed2705
-; CGP-NEXT:    s_mov_b32 s7, 0x12d8fb
-; CGP-NEXT:    v_mov_b32_e32 v4, v5
-; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v6
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v7, v4
+; CGP-NEXT:    s_mov_b32 s6, 0x12d8fb
+; CGP-NEXT:    v_cvt_f32_u32_e32 v4, s6
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
+; CGP-NEXT:    s_mov_b32 s7, 0xffed2705
+; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v4
 ; CGP-NEXT:    v_ashrrev_i32_e32 v4, 31, v1
 ; CGP-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x5f7ffffc, v7
-; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v7
-; CGP-NEXT:    v_trunc_f32_e32 v8, v8
-; CGP-NEXT:    v_mac_f32_e32 v7, 0xcf800000, v8
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x2f800000, v5
+; CGP-NEXT:    v_trunc_f32_e32 v6, v6
+; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v6
+; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
 ; CGP-NEXT:    v_addc_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
 ; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, -1, v5
+; CGP-NEXT:    v_mul_lo_u32 v8, s7, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, s7, v5
+; CGP-NEXT:    v_mul_lo_u32 v9, s7, v5
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_mul_lo_u32 v8, v6, v9
+; CGP-NEXT:    v_mul_lo_u32 v10, v5, v7
+; CGP-NEXT:    v_mul_hi_u32 v11, v5, v9
+; CGP-NEXT:    v_mul_hi_u32 v9, v6, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_mul_hi_u32 v10, v5, v7
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; CGP-NEXT:    v_mul_hi_u32 v7, v6, v7
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v5, v8
+; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v7, vcc
+; CGP-NEXT:    v_mul_lo_u32 v7, v1, v8
+; CGP-NEXT:    v_mul_lo_u32 v9, v0, v6
+; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
+; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
+; CGP-NEXT:    v_mov_b32_e32 v5, 0x12d8fb
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v10, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v9, v7
+; CGP-NEXT:    v_mul_hi_u32 v9, v0, v6
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CGP-NEXT:    v_mul_hi_u32 v6, v1, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v7
+; CGP-NEXT:    v_mul_lo_u32 v6, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
+; CGP-NEXT:    v_mul_hi_u32 v7, s6, v7
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v7
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v9
+; CGP-NEXT:    v_subb_u32_e64 v7, s[4:5], v1, v6, vcc
+; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v6
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[4:5]
+; CGP-NEXT:    v_mov_b32_e32 v8, s8
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v7
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v6, v8, v6, s[4:5]
+; CGP-NEXT:    v_sub_i32_e32 v8, vcc, v0, v5
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
+; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
+; CGP-NEXT:    v_mov_b32_e32 v10, s4
+; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
+; CGP-NEXT:    v_cndmask_b32_e32 v9, v10, v9, vcc
+; CGP-NEXT:    v_sub_i32_e32 v10, vcc, v8, v5
+; CGP-NEXT:    v_subbrev_u32_e32 v11, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
+; CGP-NEXT:    v_cvt_f32_u32_e32 v9, v5
+; CGP-NEXT:    v_cndmask_b32_e32 v8, v8, v10, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v1, v11, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v6
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
+; CGP-NEXT:    v_mac_f32_e32 v9, 0x4f800000, v6
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v6, v9
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v0, v8, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v7, v1, vcc
+; CGP-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x5f7ffffc, v6
+; CGP-NEXT:    v_mul_f32_e32 v8, 0x2f800000, v6
+; CGP-NEXT:    v_trunc_f32_e32 v8, v8
+; CGP-NEXT:    v_mac_f32_e32 v6, 0xcf800000, v8
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
+; CGP-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
+; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v7, vcc
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
+; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, -1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, s6, v7
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
-; CGP-NEXT:    s_bfe_i32 s8, -1, 0x10000
+; CGP-NEXT:    v_mul_lo_u32 v9, -1, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, s7, v8
+; CGP-NEXT:    v_mul_hi_u32 v12, s7, v6
+; CGP-NEXT:    v_mul_lo_u32 v11, s7, v6
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
 ; CGP-NEXT:    v_mul_lo_u32 v10, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v9
-; CGP-NEXT:    v_mul_hi_u32 v13, v7, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, v6, v9
+; CGP-NEXT:    v_mul_hi_u32 v13, v6, v11
 ; CGP-NEXT:    v_mul_hi_u32 v11, v8, v11
-; CGP-NEXT:    v_mac_f32_e32 v5, 0x4f800000, v6
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_mul_lo_u32 v13, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_mul_hi_u32 v12, v7, v9
+; CGP-NEXT:    v_mul_hi_u32 v12, v6, v9
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
 ; CGP-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
 ; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
 ; CGP-NEXT:    v_addc_u32_e32 v8, vcc, v8, v9, vcc
-; CGP-NEXT:    v_mul_lo_u32 v9, v1, v7
-; CGP-NEXT:    v_mul_lo_u32 v10, v0, v8
-; CGP-NEXT:    v_mul_hi_u32 v11, v0, v7
-; CGP-NEXT:    v_mul_hi_u32 v7, v1, v7
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v5
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_mul_hi_u32 v10, v0, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v11, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_mul_hi_u32 v8, v1, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_lo_u32 v9, 0, v7
-; CGP-NEXT:    v_mul_lo_u32 v8, s7, v8
-; CGP-NEXT:    v_mul_lo_u32 v10, s7, v7
-; CGP-NEXT:    v_mul_hi_u32 v7, s7, v7
-; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; CGP-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
-; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v10
-; CGP-NEXT:    v_subb_u32_e64 v8, s[4:5], v1, v7, vcc
-; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v7
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v7, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v9, s8
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v8
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v7, v9, v7, s[4:5]
-; CGP-NEXT:    v_subrev_i32_e32 v9, vcc, s7, v0
-; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v11, s4
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CGP-NEXT:    v_cndmask_b32_e32 v10, v11, v10, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v11, vcc, s7, v9
-; CGP-NEXT:    v_subbrev_u32_e32 v12, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v10
-; CGP-NEXT:    v_cndmask_b32_e32 v9, v9, v11, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v1, v12, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v7
-; CGP-NEXT:    v_mul_f32_e32 v7, 0x2f800000, v5
-; CGP-NEXT:    v_trunc_f32_e32 v7, v7
-; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v7
-; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
-; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; CGP-NEXT:    v_cndmask_b32_e32 v0, v0, v9, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v1, v8, v1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_ashrrev_i32_e32 v6, 31, v3
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_addc_u32_e32 v3, vcc, v3, v6, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v0, v0, v4
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, -1, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, s6, v7
-; CGP-NEXT:    v_mul_hi_u32 v11, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v10, s6, v5
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CGP-NEXT:    v_mul_lo_u32 v9, v7, v10
-; CGP-NEXT:    v_mul_lo_u32 v11, v5, v8
-; CGP-NEXT:    v_mul_hi_u32 v12, v5, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v7, v10
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
-; CGP-NEXT:    v_mul_hi_u32 v11, v5, v8
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v12, v10
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
-; CGP-NEXT:    v_mul_hi_u32 v8, v7, v8
-; CGP-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
-; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v8, vcc
 ; CGP-NEXT:    v_xor_b32_e32 v1, v1, v4
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v5
-; CGP-NEXT:    v_mul_lo_u32 v9, v2, v7
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v6
+; CGP-NEXT:    v_mul_lo_u32 v10, v2, v8
 ; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
 ; CGP-NEXT:    v_subb_u32_e32 v1, vcc, v1, v4, vcc
-; CGP-NEXT:    v_mul_hi_u32 v4, v2, v5
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v8, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
-; CGP-NEXT:    v_mul_lo_u32 v8, v3, v7
-; CGP-NEXT:    v_mul_hi_u32 v5, v3, v5
+; CGP-NEXT:    v_mul_hi_u32 v4, v2, v6
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
-; CGP-NEXT:    v_mul_hi_u32 v9, v2, v7
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v9
+; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, 1, vcc
+; CGP-NEXT:    v_mul_lo_u32 v9, v3, v8
+; CGP-NEXT:    v_mul_hi_u32 v6, v3, v6
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v10, v4
+; CGP-NEXT:    v_mul_hi_u32 v10, v2, v8
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
 ; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; CGP-NEXT:    v_mul_hi_u32 v7, v3, v7
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
-; CGP-NEXT:    v_mul_lo_u32 v7, 0, v4
-; CGP-NEXT:    v_mul_lo_u32 v5, s7, v5
-; CGP-NEXT:    v_mul_lo_u32 v8, s7, v4
-; CGP-NEXT:    v_mul_hi_u32 v4, s7, v4
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v10
+; CGP-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CGP-NEXT:    v_mul_hi_u32 v8, v3, v8
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_mul_lo_u32 v8, 0, v4
+; CGP-NEXT:    v_mul_lo_u32 v6, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v9, s6, v4
+; CGP-NEXT:    v_mul_hi_u32 v4, s6, v4
 ; CGP-NEXT:    s_bfe_i32 s6, -1, 0x10000
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v8
-; CGP-NEXT:    v_subb_u32_e64 v5, s[4:5], v3, v4, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v9
+; CGP-NEXT:    v_subb_u32_e64 v6, s[4:5], v3, v4, vcc
 ; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v4
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s7, v2
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v5
 ; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, -1, s[4:5]
-; CGP-NEXT:    v_mov_b32_e32 v7, s6
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v5
+; CGP-NEXT:    v_mov_b32_e32 v8, s6
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v6
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v4, v7, v4, s[4:5]
-; CGP-NEXT:    v_subrev_i32_e32 v7, vcc, s7, v2
+; CGP-NEXT:    v_cndmask_b32_e64 v4, v8, v4, s[4:5]
+; CGP-NEXT:    v_sub_i32_e32 v8, vcc, v2, v5
 ; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
 ; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s7, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v8, 0, -1, vcc
-; CGP-NEXT:    v_mov_b32_e32 v9, s4
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
+; CGP-NEXT:    v_mov_b32_e32 v10, s4
 ; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; CGP-NEXT:    v_cndmask_b32_e32 v8, v9, v8, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v9, vcc, s7, v7
+; CGP-NEXT:    v_cndmask_b32_e32 v9, v10, v9, vcc
+; CGP-NEXT:    v_sub_i32_e32 v5, vcc, v8, v5
 ; CGP-NEXT:    v_subbrev_u32_e32 v10, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v8
-; CGP-NEXT:    v_cndmask_b32_e32 v7, v7, v9, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
+; CGP-NEXT:    v_cndmask_b32_e32 v5, v8, v5, vcc
 ; CGP-NEXT:    v_cndmask_b32_e32 v3, v3, v10, vcc
 ; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v4
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v2, v7, vcc
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v5, v3, vcc
-; CGP-NEXT:    v_xor_b32_e32 v2, v2, v6
-; CGP-NEXT:    v_xor_b32_e32 v3, v3, v6
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
-; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v6, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v2, v2, v5, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v3, v6, v3, vcc
+; CGP-NEXT:    v_xor_b32_e32 v2, v2, v7
+; CGP-NEXT:    v_xor_b32_e32 v3, v3, v7
+; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v7
+; CGP-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
 ; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = srem <2 x i64> %num, <i64 1235195, i64 1235195>
   ret <2 x i64> %result
 }
 
 define i64 @v_srem_i64_pow2_shl_denom(i64 %x, i64 %y) {
 ; CHECK-LABEL: v_srem_i64_pow2_shl_denom:
 ; CHECK:       ; %bb.0:

llvm/test/CodeGen/AMDGPU/GlobalISel/urem.i64.ll

   %result = urem <2 x i64> %num, <i64 4096, i64 4096>
   ret <2 x i64> %result
 }
 
 define i64 @v_urem_i64_oddk_denom(i64 %num) {
 ; CHECK-LABEL: v_urem_i64_oddk_denom:
 ; CHECK:       ; %bb.0:
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-NEXT:    v_cvt_f32_u32_e32 v2, 0x12d8fb
+; CHECK-NEXT:    s_mov_b32 s4, 0x12d8fb
+; CHECK-NEXT:    v_mov_b32_e32 v2, 0x12d8fb
 ; CHECK-NEXT:    v_cvt_f32_ubyte0_e32 v3, 0
-; CHECK-NEXT:    s_mov_b32 s4, 0xffed2705
-; CHECK-NEXT:    s_mov_b32 s6, 0x12d8fb
-; CHECK-NEXT:    s_bfe_i32 s5, -1, 0x10000
+; CHECK-NEXT:    s_mov_b32 s5, 0xffed2705
+; CHECK-NEXT:    s_bfe_i32 s6, -1, 0x10000
 ; CHECK-NEXT:    s_bfe_i32 s7, -1, 0x10000
-; CHECK-NEXT:    v_mac_f32_e32 v2, 0x4f800000, v3
-; CHECK-NEXT:    v_mov_b32_e32 v3, s5
-; CHECK-NEXT:    v_mov_b32_e32 v4, s7
-; CHECK-NEXT:    v_rcp_iflag_f32_e32 v2, v2
-; CHECK-NEXT:    v_mul_f32_e32 v2, 0x5f7ffffc, v2
-; CHECK-NEXT:    v_mul_f32_e32 v5, 0x2f800000, v2
-; CHECK-NEXT:    v_trunc_f32_e32 v5, v5
-; CHECK-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v5
-; CHECK-NEXT:    v_cvt_u32_f32_e32 v5, v5
-; CHECK-NEXT:    v_cvt_u32_f32_e32 v2, v2
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v5
-; CHECK-NEXT:    v_mul_lo_u32 v7, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v8, -1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v9, s4, v2
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
-; CHECK-NEXT:    v_mul_lo_u32 v8, v5, v7
-; CHECK-NEXT:    v_mul_hi_u32 v10, v2, v7
-; CHECK-NEXT:    v_mul_hi_u32 v7, v5, v7
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v9
-; CHECK-NEXT:    v_mul_lo_u32 v9, v2, v6
-; CHECK-NEXT:    v_mul_lo_u32 v11, v5, v6
-; CHECK-NEXT:    v_mul_hi_u32 v12, v2, v6
-; CHECK-NEXT:    v_mul_hi_u32 v6, v5, v6
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CHECK-NEXT:    v_cvt_f32_u32_e32 v4, s4
+; CHECK-NEXT:    v_mov_b32_e32 v5, s6
+; CHECK-NEXT:    v_mov_b32_e32 v6, s7
+; CHECK-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v3
+; CHECK-NEXT:    v_rcp_iflag_f32_e32 v3, v4
+; CHECK-NEXT:    v_mul_f32_e32 v3, 0x5f7ffffc, v3
+; CHECK-NEXT:    v_mul_f32_e32 v4, 0x2f800000, v3
+; CHECK-NEXT:    v_trunc_f32_e32 v4, v4
+; CHECK-NEXT:    v_mac_f32_e32 v3, 0xcf800000, v4
+; CHECK-NEXT:    v_cvt_u32_f32_e32 v4, v4
+; CHECK-NEXT:    v_cvt_u32_f32_e32 v3, v3
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v4
+; CHECK-NEXT:    v_mul_lo_u32 v8, s5, v3
+; CHECK-NEXT:    v_mul_lo_u32 v9, -1, v3
+; CHECK-NEXT:    v_mul_hi_u32 v10, s5, v3
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v9, v7
+; CHECK-NEXT:    v_mul_lo_u32 v9, v4, v8
+; CHECK-NEXT:    v_mul_hi_u32 v11, v3, v8
+; CHECK-NEXT:    v_mul_hi_u32 v8, v4, v8
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v10
+; CHECK-NEXT:    v_mul_lo_u32 v10, v3, v7
+; CHECK-NEXT:    v_mul_lo_u32 v12, v4, v7
+; CHECK-NEXT:    v_mul_hi_u32 v13, v3, v7
+; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v9, v10
+; CHECK-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v12, v8
+; CHECK-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
 ; CHECK-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v11, v7
+; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v13
 ; CHECK-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CHECK-NEXT:    v_add_i32_e32 v10, vcc, v12, v11
+; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CHECK-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
+; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v3, v8
+; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v7, vcc
+; CHECK-NEXT:    v_mul_lo_u32 v7, s5, v3
+; CHECK-NEXT:    v_mul_lo_u32 v8, -1, v3
+; CHECK-NEXT:    v_mul_hi_u32 v9, s5, v3
+; CHECK-NEXT:    v_mul_lo_u32 v10, s5, v4
+; CHECK-NEXT:    v_mul_lo_u32 v11, v4, v7
+; CHECK-NEXT:    v_mul_hi_u32 v12, v3, v7
+; CHECK-NEXT:    v_mul_hi_u32 v7, v4, v7
 ; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v12
+; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CHECK-NEXT:    v_mul_lo_u32 v9, v3, v8
+; CHECK-NEXT:    v_mul_lo_u32 v10, v4, v8
+; CHECK-NEXT:    v_mul_hi_u32 v13, v3, v8
+; CHECK-NEXT:    v_mul_hi_u32 v8, v4, v8
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CHECK-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v10, v7
 ; CHECK-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v11, v10
-; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v7
-; CHECK-NEXT:    v_addc_u32_e32 v5, vcc, v5, v6, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v6, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v7, -1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v8, s4, v2
-; CHECK-NEXT:    v_mul_lo_u32 v9, s4, v5
-; CHECK-NEXT:    v_mul_lo_u32 v10, v5, v6
-; CHECK-NEXT:    v_mul_hi_u32 v11, v2, v6
-; CHECK-NEXT:    v_mul_hi_u32 v6, v5, v6
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
+; CHECK-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v13
+; CHECK-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v11, v9
+; CHECK-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
-; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
-; CHECK-NEXT:    v_mul_lo_u32 v8, v2, v7
-; CHECK-NEXT:    v_mul_lo_u32 v9, v5, v7
-; CHECK-NEXT:    v_mul_hi_u32 v12, v2, v7
-; CHECK-NEXT:    v_mul_hi_u32 v7, v5, v7
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v9, v6
 ; CHECK-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v12
-; CHECK-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
-; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
-; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CHECK-NEXT:    v_addc_u32_e32 v5, vcc, v5, v7, vcc
-; CHECK-NEXT:    v_mul_lo_u32 v6, v1, v2
-; CHECK-NEXT:    v_mul_hi_u32 v7, v0, v2
-; CHECK-NEXT:    v_mul_hi_u32 v2, v1, v2
-; CHECK-NEXT:    v_mul_lo_u32 v8, v0, v5
-; CHECK-NEXT:    v_mul_lo_u32 v9, v1, v5
-; CHECK-NEXT:    v_mul_hi_u32 v10, v0, v5
-; CHECK-NEXT:    v_mul_hi_u32 v5, v1, v5
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
-; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v9, v2
+; CHECK-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
+; CHECK-NEXT:    v_addc_u32_e32 v4, vcc, v4, v8, vcc
+; CHECK-NEXT:    v_mul_lo_u32 v7, v1, v3
+; CHECK-NEXT:    v_mul_hi_u32 v8, v0, v3
+; CHECK-NEXT:    v_mul_hi_u32 v3, v1, v3
+; CHECK-NEXT:    v_mul_lo_u32 v9, v0, v4
+; CHECK-NEXT:    v_mul_lo_u32 v10, v1, v4
+; CHECK-NEXT:    v_mul_hi_u32 v11, v0, v4
+; CHECK-NEXT:    v_mul_hi_u32 v4, v1, v4
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v9
 ; CHECK-NEXT:    v_cndmask_b32_e64 v9, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v6, v7
-; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v10
+; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v10, v3
+; CHECK-NEXT:    v_cndmask_b32_e64 v10, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v7, v8
 ; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v8, v6
+; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v3, v11
+; CHECK-NEXT:    v_cndmask_b32_e64 v8, 0, 1, vcc
 ; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v9, v7
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
-; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, 1, vcc
-; CHECK-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
-; CHECK-NEXT:    v_mul_lo_u32 v7, s6, v2
-; CHECK-NEXT:    v_mul_lo_u32 v8, 0, v2
-; CHECK-NEXT:    v_mul_hi_u32 v2, s6, v2
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v5, v6
-; CHECK-NEXT:    v_mul_lo_u32 v5, s6, v5
-; CHECK-NEXT:    v_add_i32_e32 v5, vcc, v8, v5
-; CHECK-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
-; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v7
-; CHECK-NEXT:    v_subb_u32_e64 v5, s[4:5], v1, v2, vcc
-; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v2
-; CHECK-NEXT:    v_cmp_le_u32_e64 s[4:5], s6, v0
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, 0, -1, s[4:5]
-; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v5
-; CHECK-NEXT:    v_cndmask_b32_e64 v2, v3, v2, s[4:5]
+; CHECK-NEXT:    v_add_i32_e32 v8, vcc, v10, v8
+; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
+; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, 1, vcc
+; CHECK-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
+; CHECK-NEXT:    v_mul_lo_u32 v8, s4, v3
+; CHECK-NEXT:    v_mul_lo_u32 v9, 0, v3
+; CHECK-NEXT:    v_mul_hi_u32 v3, s4, v3
+; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v4, v7
+; CHECK-NEXT:    v_mul_lo_u32 v4, s4, v4
+; CHECK-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
+; CHECK-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
+; CHECK-NEXT:    v_sub_i32_e32 v0, vcc, v0, v8
+; CHECK-NEXT:    v_subb_u32_e64 v4, s[4:5], v1, v3, vcc
+; CHECK-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v3
+; CHECK-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v2
+; CHECK-NEXT:    v_cndmask_b32_e64 v3, 0, -1, s[4:5]
+; CHECK-NEXT:    v_cmp_eq_u32_e64 s[4:5], 0, v4
+; CHECK-NEXT:    v_cndmask_b32_e64 v3, v5, v3, s[4:5]
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_subrev_i32_e32 v3, vcc, s6, v0
+; CHECK-NEXT:    v_sub_i32_e32 v5, vcc, v0, v2
 ; CHECK-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_cmp_le_u32_e32 vcc, s6, v3
-; CHECK-NEXT:    v_cndmask_b32_e64 v6, 0, -1, vcc
+; CHECK-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v2
+; CHECK-NEXT:    v_cndmask_b32_e64 v7, 0, -1, vcc
 ; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CHECK-NEXT:    v_cndmask_b32_e32 v4, v4, v6, vcc
-; CHECK-NEXT:    v_subrev_i32_e32 v6, vcc, s6, v3
+; CHECK-NEXT:    v_cndmask_b32_e32 v6, v6, v7, vcc
+; CHECK-NEXT:    v_sub_i32_e32 v2, vcc, v5, v2
 ; CHECK-NEXT:    v_subbrev_u32_e32 v7, vcc, 0, v1, vcc
-; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v4
-; CHECK-NEXT:    v_cndmask_b32_e32 v3, v3, v6, vcc
+; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v6
+; CHECK-NEXT:    v_cndmask_b32_e32 v2, v5, v2, vcc
 ; CHECK-NEXT:    v_cndmask_b32_e32 v1, v1, v7, vcc
-; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
-; CHECK-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
-; CHECK-NEXT:    v_cndmask_b32_e32 v1, v5, v1, vcc
+; CHECK-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; CHECK-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
+; CHECK-NEXT:    v_cndmask_b32_e32 v1, v4, v1, vcc
 ; CHECK-NEXT:    s_setpc_b64 s[30:31]
   %result = urem i64 %num, 1235195
   ret i64 %result
 }
 
 define <2 x i64> @v_urem_v2i64_oddk_denom(<2 x i64> %num) {
 ; GISEL-LABEL: v_urem_v2i64_oddk_denom:
 ; GISEL:       ; %bb.0:

 ; GISEL-NEXT:    v_cndmask_b32_e64 v0, v0, v4, s[4:5]
 ; GISEL-NEXT:    v_cndmask_b32_e64 v1, v9, v1, s[4:5]
 ; GISEL-NEXT:    v_cndmask_b32_e32 v3, v8, v3, vcc
 ; GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; CGP-LABEL: v_urem_v2i64_oddk_denom:
 ; CGP:       ; %bb.0:
 ; CGP-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CGP-NEXT:    v_cvt_f32_u32_e32 v4, 0x12d8fb
+; CGP-NEXT:    s_mov_b32 s8, 0x12d8fb
+; CGP-NEXT:    v_mov_b32_e32 v4, 0x12d8fb
 ; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v5, 0
 ; CGP-NEXT:    s_mov_b32 s6, 0xffed2705
-; CGP-NEXT:    s_mov_b32 s8, 0x12d8fb
 ; CGP-NEXT:    s_bfe_i32 s4, -1, 0x10000
 ; CGP-NEXT:    s_bfe_i32 s5, -1, 0x10000
+; CGP-NEXT:    v_cvt_f32_ubyte0_e32 v6, 0
 ; CGP-NEXT:    s_bfe_i32 s7, -1, 0x10000
 ; CGP-NEXT:    s_bfe_i32 s9, -1, 0x10000
-; CGP-NEXT:    v_mov_b32_e32 v6, v4
-; CGP-NEXT:    v_mov_b32_e32 v7, s4
-; CGP-NEXT:    v_mov_b32_e32 v8, s5
-; CGP-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
-; CGP-NEXT:    v_mov_b32_e32 v9, s7
-; CGP-NEXT:    v_mac_f32_e32 v6, 0x4f800000, v5
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v4, v4
-; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v6
-; CGP-NEXT:    v_mul_f32_e32 v4, 0x5f7ffffc, v4
+; CGP-NEXT:    v_cvt_f32_u32_e32 v7, s8
+; CGP-NEXT:    v_mov_b32_e32 v8, s4
+; CGP-NEXT:    v_mov_b32_e32 v9, s5
+; CGP-NEXT:    v_cvt_f32_u32_e32 v10, v4
+; CGP-NEXT:    v_mac_f32_e32 v7, 0x4f800000, v5
+; CGP-NEXT:    v_mac_f32_e32 v10, 0x4f800000, v6
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v5, v7
+; CGP-NEXT:    v_rcp_iflag_f32_e32 v6, v10
 ; CGP-NEXT:    v_mul_f32_e32 v5, 0x5f7ffffc, v5
-; CGP-NEXT:    v_mul_f32_e32 v6, 0x2f800000, v4
-; CGP-NEXT:    v_mul_f32_e32 v10, 0x2f800000, v5
-; CGP-NEXT:    v_trunc_f32_e32 v6, v6
+; CGP-NEXT:    v_mul_f32_e32 v6, 0x5f7ffffc, v6
+; CGP-NEXT:    v_mul_f32_e32 v7, 0x2f800000, v5
+; CGP-NEXT:    v_mul_f32_e32 v10, 0x2f800000, v6
+; CGP-NEXT:    v_trunc_f32_e32 v7, v7
 ; CGP-NEXT:    v_trunc_f32_e32 v10, v10
-; CGP-NEXT:    v_mac_f32_e32 v4, 0xcf800000, v6
-; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
-; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v10
+; CGP-NEXT:    v_mac_f32_e32 v5, 0xcf800000, v7
+; CGP-NEXT:    v_cvt_u32_f32_e32 v7, v7
+; CGP-NEXT:    v_mac_f32_e32 v6, 0xcf800000, v10
 ; CGP-NEXT:    v_cvt_u32_f32_e32 v10, v10
-; CGP-NEXT:    v_cvt_u32_f32_e32 v4, v4
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v6
 ; CGP-NEXT:    v_cvt_u32_f32_e32 v5, v5
+; CGP-NEXT:    v_mul_lo_u32 v11, s6, v7
+; CGP-NEXT:    v_cvt_u32_f32_e32 v6, v6
 ; CGP-NEXT:    v_mul_lo_u32 v12, s6, v10
-; CGP-NEXT:    v_mul_lo_u32 v13, s6, v4
-; CGP-NEXT:    v_mul_lo_u32 v14, -1, v4
-; CGP-NEXT:    v_mul_hi_u32 v15, s6, v4
-; CGP-NEXT:    v_mul_lo_u32 v16, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v17, -1, v5
-; CGP-NEXT:    v_mul_hi_u32 v18, s6, v5
+; CGP-NEXT:    v_mul_lo_u32 v13, s6, v5
+; CGP-NEXT:    v_mul_lo_u32 v14, -1, v5
+; CGP-NEXT:    v_mul_hi_u32 v15, s6, v5
+; CGP-NEXT:    v_mul_lo_u32 v16, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v17, -1, v6
+; CGP-NEXT:    v_mul_hi_u32 v18, s6, v6
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v14, v11
-; CGP-NEXT:    v_mul_lo_u32 v14, v6, v13
-; CGP-NEXT:    v_mul_hi_u32 v19, v4, v13
-; CGP-NEXT:    v_mul_hi_u32 v13, v6, v13
+; CGP-NEXT:    v_mul_lo_u32 v14, v7, v13
+; CGP-NEXT:    v_mul_hi_u32 v19, v5, v13
+; CGP-NEXT:    v_mul_hi_u32 v13, v7, v13
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v17, v12
 ; CGP-NEXT:    v_mul_lo_u32 v17, v10, v16
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v15
-; CGP-NEXT:    v_mul_hi_u32 v15, v5, v16
+; CGP-NEXT:    v_mul_hi_u32 v15, v6, v16
 ; CGP-NEXT:    v_mul_hi_u32 v16, v10, v16
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v18
-; CGP-NEXT:    v_mul_lo_u32 v18, v5, v12
+; CGP-NEXT:    v_mul_lo_u32 v18, v6, v12
 ; CGP-NEXT:    v_add_i32_e32 v17, vcc, v17, v18
 ; CGP-NEXT:    v_cndmask_b32_e64 v18, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v15, vcc, v17, v15
-; CGP-NEXT:    v_mul_lo_u32 v15, v4, v11
-; CGP-NEXT:    v_mul_lo_u32 v17, v6, v11
+; CGP-NEXT:    v_mul_lo_u32 v15, v5, v11
+; CGP-NEXT:    v_mul_lo_u32 v17, v7, v11
 ; CGP-NEXT:    v_add_i32_e64 v14, s[4:5], v14, v15
 ; CGP-NEXT:    v_cndmask_b32_e64 v15, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v14, s[4:5], v14, v19
-; CGP-NEXT:    v_mul_hi_u32 v14, v4, v11
-; CGP-NEXT:    v_mul_hi_u32 v11, v6, v11
+; CGP-NEXT:    v_mul_hi_u32 v14, v5, v11
+; CGP-NEXT:    v_mul_hi_u32 v11, v7, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v19, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v15, s[4:5], v15, v19
 ; CGP-NEXT:    v_mul_lo_u32 v19, v10, v12
 ; CGP-NEXT:    v_add_i32_e64 v13, s[4:5], v17, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v17, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v13, s[4:5], v13, v14
 ; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v14, s[4:5], v17, v14
 ; CGP-NEXT:    v_cndmask_b32_e64 v17, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v17, vcc, v18, v17
-; CGP-NEXT:    v_mul_hi_u32 v18, v5, v12
+; CGP-NEXT:    v_mul_hi_u32 v18, v6, v12
 ; CGP-NEXT:    v_mul_hi_u32 v12, v10, v12
 ; CGP-NEXT:    v_add_i32_e32 v16, vcc, v19, v16
 ; CGP-NEXT:    v_cndmask_b32_e64 v19, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v16, vcc, v16, v18
 ; CGP-NEXT:    v_cndmask_b32_e64 v18, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v18, vcc, v19, v18
 ; CGP-NEXT:    v_add_i32_e32 v13, vcc, v13, v15
 ; CGP-NEXT:    v_cndmask_b32_e64 v15, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v16, vcc, v16, v17
 ; CGP-NEXT:    v_cndmask_b32_e64 v17, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v14, vcc, v14, v15
 ; CGP-NEXT:    v_add_i32_e32 v15, vcc, v18, v17
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v14
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v15
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v4, v13
-; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v11, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, s6, v4
-; CGP-NEXT:    v_mul_lo_u32 v13, -1, v4
-; CGP-NEXT:    v_mul_hi_u32 v14, s6, v4
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v16
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v13
+; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v11, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, s6, v5
+; CGP-NEXT:    v_mul_lo_u32 v13, -1, v5
+; CGP-NEXT:    v_mul_hi_u32 v14, s6, v5
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v16
 ; CGP-NEXT:    v_addc_u32_e32 v10, vcc, v10, v12, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v15, -1, v5
-; CGP-NEXT:    v_mul_hi_u32 v16, s6, v5
-; CGP-NEXT:    v_mul_lo_u32 v17, s6, v6
-; CGP-NEXT:    v_mul_lo_u32 v18, v6, v11
-; CGP-NEXT:    v_mul_hi_u32 v19, v4, v11
-; CGP-NEXT:    v_mul_hi_u32 v11, v6, v11
+; CGP-NEXT:    v_mul_lo_u32 v12, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v15, -1, v6
+; CGP-NEXT:    v_mul_hi_u32 v16, s6, v6
+; CGP-NEXT:    v_mul_lo_u32 v17, s6, v7
+; CGP-NEXT:    v_mul_lo_u32 v18, v7, v11
+; CGP-NEXT:    v_mul_hi_u32 v19, v5, v11
+; CGP-NEXT:    v_mul_hi_u32 v11, v7, v11
 ; CGP-NEXT:    v_add_i32_e32 v13, vcc, v13, v17
 ; CGP-NEXT:    v_mul_lo_u32 v17, s6, v10
 ; CGP-NEXT:    v_add_i32_e32 v15, vcc, v15, v17
 ; CGP-NEXT:    v_mul_lo_u32 v17, v10, v12
 ; CGP-NEXT:    v_add_i32_e32 v13, vcc, v13, v14
-; CGP-NEXT:    v_mul_hi_u32 v14, v5, v12
+; CGP-NEXT:    v_mul_hi_u32 v14, v6, v12
 ; CGP-NEXT:    v_mul_hi_u32 v12, v10, v12
 ; CGP-NEXT:    v_add_i32_e32 v15, vcc, v15, v16
-; CGP-NEXT:    v_mul_lo_u32 v16, v5, v15
+; CGP-NEXT:    v_mul_lo_u32 v16, v6, v15
 ; CGP-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
 ; CGP-NEXT:    v_cndmask_b32_e64 v17, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v14, vcc, v16, v14
-; CGP-NEXT:    v_mul_lo_u32 v14, v4, v13
-; CGP-NEXT:    v_mul_lo_u32 v16, v6, v13
+; CGP-NEXT:    v_mul_lo_u32 v14, v5, v13
+; CGP-NEXT:    v_mul_lo_u32 v16, v7, v13
 ; CGP-NEXT:    v_add_i32_e64 v14, s[4:5], v18, v14
 ; CGP-NEXT:    v_cndmask_b32_e64 v18, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v14, s[4:5], v14, v19
-; CGP-NEXT:    v_mul_hi_u32 v14, v4, v13
+; CGP-NEXT:    v_mul_hi_u32 v14, v5, v13
 ; CGP-NEXT:    v_cndmask_b32_e64 v19, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v18, s[4:5], v18, v19
 ; CGP-NEXT:    v_mul_lo_u32 v19, v10, v15
 ; CGP-NEXT:    v_add_i32_e64 v11, s[4:5], v16, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v11, s[4:5], v11, v14
 ; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, s[4:5]
 ; CGP-NEXT:    v_add_i32_e64 v14, s[4:5], v16, v14
 ; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
-; CGP-NEXT:    v_mul_hi_u32 v17, v5, v15
+; CGP-NEXT:    v_mul_hi_u32 v17, v6, v15
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v19, v12
 ; CGP-NEXT:    v_cndmask_b32_e64 v19, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v17
 ; CGP-NEXT:    v_cndmask_b32_e64 v17, 0, 1, vcc
 ; CGP-NEXT:    v_add_i32_e32 v17, vcc, v19, v17
-; CGP-NEXT:    v_mov_b32_e32 v19, s9
-; CGP-NEXT:    v_mul_hi_u32 v13, v6, v13
-; CGP-NEXT:    v_mul_hi_u32 v15, v10, v15
+; CGP-NEXT:    v_mov_b32_e32 v19, s7
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v18
 ; CGP-NEXT:    v_cndmask_b32_e64 v18, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v14, vcc, v14, v18
+; CGP-NEXT:    v_mov_b32_e32 v18, s9
+; CGP-NEXT:    v_mul_hi_u32 v13, v7, v13
+; CGP-NEXT:    v_mul_hi_u32 v15, v10, v15
 ; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v16
 ; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v14, vcc, v14, v18
 ; CGP-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
 ; CGP-NEXT:    v_add_i32_e32 v13, vcc, v13, v14
 ; CGP-NEXT:    v_add_i32_e32 v14, vcc, v15, v16
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v4, v11
-; CGP-NEXT:    v_addc_u32_e32 v6, vcc, v6, v13, vcc
-; CGP-NEXT:    v_mul_lo_u32 v11, v3, v4
-; CGP-NEXT:    v_mul_hi_u32 v13, v2, v4
-; CGP-NEXT:    v_mul_hi_u32 v4, v3, v4
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v12
-; CGP-NEXT:    v_addc_u32_e32 v10, vcc, v10, v14, vcc
-; CGP-NEXT:    v_mul_lo_u32 v12, v1, v5
-; CGP-NEXT:    v_mul_hi_u32 v14, v0, v5
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v11
+; CGP-NEXT:    v_addc_u32_e32 v7, vcc, v7, v13, vcc
+; CGP-NEXT:    v_mul_lo_u32 v11, v1, v5
+; CGP-NEXT:    v_mul_hi_u32 v13, v0, v5
 ; CGP-NEXT:    v_mul_hi_u32 v5, v1, v5
-; CGP-NEXT:    v_mul_lo_u32 v15, v2, v6
-; CGP-NEXT:    v_mul_lo_u32 v16, v3, v6
-; CGP-NEXT:    v_mul_hi_u32 v17, v2, v6
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v12
+; CGP-NEXT:    v_addc_u32_e32 v10, vcc, v10, v14, vcc
+; CGP-NEXT:    v_mul_lo_u32 v12, v3, v6
+; CGP-NEXT:    v_mul_hi_u32 v14, v2, v6
 ; CGP-NEXT:    v_mul_hi_u32 v6, v3, v6
-; CGP-NEXT:    v_mul_lo_u32 v18, v0, v10
-; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v18
-; CGP-NEXT:    v_cndmask_b32_e64 v18, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v14
-; CGP-NEXT:    v_mul_lo_u32 v12, v1, v10
-; CGP-NEXT:    v_mul_hi_u32 v14, v0, v10
-; CGP-NEXT:    v_mul_hi_u32 v10, v1, v10
-; CGP-NEXT:    v_add_i32_e64 v11, s[4:5], v11, v15
-; CGP-NEXT:    v_cndmask_b32_e64 v15, 0, 1, s[4:5]
-; CGP-NEXT:    v_add_i32_e64 v4, s[4:5], v16, v4
-; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, s[4:5]
-; CGP-NEXT:    v_add_i32_e64 v5, s[4:5], v12, v5
+; CGP-NEXT:    v_mul_lo_u32 v15, v0, v7
+; CGP-NEXT:    v_mul_lo_u32 v16, v1, v7
+; CGP-NEXT:    v_mul_hi_u32 v17, v0, v7
+; CGP-NEXT:    v_mul_hi_u32 v7, v1, v7
+; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v15
+; CGP-NEXT:    v_cndmask_b32_e64 v15, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v13
+; CGP-NEXT:    v_mul_lo_u32 v11, v2, v10
+; CGP-NEXT:    v_mul_lo_u32 v13, v3, v10
+; CGP-NEXT:    v_add_i32_e64 v11, s[4:5], v12, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, s[4:5]
-; CGP-NEXT:    v_add_i32_e64 v11, s[4:5], v11, v13
-; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, s[4:5]
-; CGP-NEXT:    v_add_i32_e64 v4, s[4:5], v4, v17
-; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, s[4:5]
+; CGP-NEXT:    v_add_i32_e64 v11, s[4:5], v11, v14
+; CGP-NEXT:    v_mul_hi_u32 v11, v2, v10
+; CGP-NEXT:    v_mul_hi_u32 v10, v3, v10
+; CGP-NEXT:    v_add_i32_e64 v5, s[6:7], v16, v5
+; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, s[6:7]
+; CGP-NEXT:    v_add_i32_e64 v6, s[6:7], v13, v6
+; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, s[6:7]
+; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v17
 ; CGP-NEXT:    v_cndmask_b32_e64 v17, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v14
-; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v11, vcc, v15, v11
-; CGP-NEXT:    v_add_i32_e32 v13, vcc, v16, v13
-; CGP-NEXT:    v_add_i32_e32 v15, vcc, v18, v17
-; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v14
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v4, v11
+; CGP-NEXT:    v_add_i32_e32 v15, vcc, v15, v16
+; CGP-NEXT:    v_cndmask_b32_e64 v16, 0, 1, s[4:5]
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v11
 ; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, 1, vcc
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v15
-; CGP-NEXT:    v_cndmask_b32_e64 v14, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v14, vcc, v14, v17
+; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v16
 ; CGP-NEXT:    v_add_i32_e32 v11, vcc, v13, v11
-; CGP-NEXT:    v_mul_lo_u32 v13, s8, v4
-; CGP-NEXT:    v_mul_lo_u32 v15, 0, v4
-; CGP-NEXT:    v_mul_hi_u32 v4, s8, v4
-; CGP-NEXT:    v_add_i32_e32 v12, vcc, v12, v14
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v5, v15
+; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v12
+; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, 1, vcc
+; CGP-NEXT:    v_add_i32_e32 v13, vcc, v14, v13
 ; CGP-NEXT:    v_mul_lo_u32 v14, s8, v5
-; CGP-NEXT:    v_mul_lo_u32 v16, 0, v5
+; CGP-NEXT:    v_mul_lo_u32 v15, 0, v5
 ; CGP-NEXT:    v_mul_hi_u32 v5, s8, v5
-; CGP-NEXT:    v_add_i32_e32 v6, vcc, v6, v11
-; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
-; CGP-NEXT:    v_mul_lo_u32 v6, s8, v6
+; CGP-NEXT:    v_add_i32_e32 v11, vcc, v11, v12
+; CGP-NEXT:    v_mul_lo_u32 v12, s8, v6
+; CGP-NEXT:    v_mul_lo_u32 v16, 0, v6
+; CGP-NEXT:    v_mul_hi_u32 v6, s8, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v7, v13
+; CGP-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
+; CGP-NEXT:    v_mul_lo_u32 v7, s8, v7
 ; CGP-NEXT:    v_mul_lo_u32 v10, s8, v10
-; CGP-NEXT:    v_add_i32_e32 v6, vcc, v15, v6
+; CGP-NEXT:    v_add_i32_e32 v7, vcc, v15, v7
 ; CGP-NEXT:    v_add_i32_e32 v10, vcc, v16, v10
-; CGP-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
-; CGP-NEXT:    v_add_i32_e32 v5, vcc, v10, v5
-; CGP-NEXT:    v_sub_i32_e32 v2, vcc, v2, v13
-; CGP-NEXT:    v_subb_u32_e64 v6, s[4:5], v3, v4, vcc
-; CGP-NEXT:    v_sub_i32_e64 v3, s[4:5], v3, v4
-; CGP-NEXT:    v_cmp_le_u32_e64 s[4:5], s8, v2
-; CGP-NEXT:    v_cndmask_b32_e64 v4, 0, -1, s[4:5]
-; CGP-NEXT:    v_sub_i32_e64 v0, s[4:5], v0, v14
-; CGP-NEXT:    v_subb_u32_e64 v10, s[6:7], v1, v5, s[4:5]
-; CGP-NEXT:    v_sub_i32_e64 v1, s[6:7], v1, v5
-; CGP-NEXT:    v_cmp_le_u32_e64 s[6:7], s8, v0
-; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, -1, s[6:7]
-; CGP-NEXT:    v_cmp_eq_u32_e64 s[6:7], 0, v6
-; CGP-NEXT:    v_cndmask_b32_e64 v4, v9, v4, s[6:7]
-; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
+; CGP-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
+; CGP-NEXT:    v_add_i32_e32 v6, vcc, v10, v6
+; CGP-NEXT:    v_sub_i32_e32 v0, vcc, v0, v14
+; CGP-NEXT:    v_subb_u32_e64 v7, s[4:5], v1, v5, vcc
+; CGP-NEXT:    v_sub_i32_e64 v1, s[4:5], v1, v5
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[4:5], v0, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v5, 0, -1, s[4:5]
+; CGP-NEXT:    v_sub_i32_e64 v2, s[4:5], v2, v12
+; CGP-NEXT:    v_subb_u32_e64 v10, s[6:7], v3, v6, s[4:5]
+; CGP-NEXT:    v_sub_i32_e64 v3, s[6:7], v3, v6
+; CGP-NEXT:    v_cmp_ge_u32_e64 s[6:7], v2, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[6:7]
+; CGP-NEXT:    v_cmp_eq_u32_e64 s[6:7], 0, v7
+; CGP-NEXT:    v_cndmask_b32_e64 v5, v8, v5, s[6:7]
+; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
 ; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v10
-; CGP-NEXT:    v_cndmask_b32_e32 v5, v7, v5, vcc
-; CGP-NEXT:    v_subbrev_u32_e64 v1, vcc, 0, v1, s[4:5]
-; CGP-NEXT:    v_subrev_i32_e32 v7, vcc, s8, v2
-; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s8, v7
-; CGP-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v11, vcc, s8, v0
+; CGP-NEXT:    v_cndmask_b32_e32 v6, v19, v6, vcc
+; CGP-NEXT:    v_subbrev_u32_e64 v3, vcc, 0, v3, s[4:5]
+; CGP-NEXT:    v_sub_i32_e32 v8, vcc, v0, v4
 ; CGP-NEXT:    v_subbrev_u32_e32 v1, vcc, 0, v1, vcc
-; CGP-NEXT:    v_cmp_le_u32_e32 vcc, s8, v11
-; CGP-NEXT:    v_cndmask_b32_e64 v12, 0, -1, vcc
-; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; CGP-NEXT:    v_cndmask_b32_e32 v9, v19, v9, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v13, vcc, s8, v7
-; CGP-NEXT:    v_subbrev_u32_e32 v14, vcc, 0, v3, vcc
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v11, 0, -1, vcc
+; CGP-NEXT:    v_sub_i32_e32 v12, vcc, v2, v4
+; CGP-NEXT:    v_subbrev_u32_e32 v3, vcc, 0, v3, vcc
+; CGP-NEXT:    v_cmp_ge_u32_e32 vcc, v12, v4
+; CGP-NEXT:    v_cndmask_b32_e64 v13, 0, -1, vcc
 ; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v1
-; CGP-NEXT:    v_cndmask_b32_e32 v8, v8, v12, vcc
-; CGP-NEXT:    v_subrev_i32_e32 v12, vcc, s8, v11
-; CGP-NEXT:    v_subbrev_u32_e32 v15, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v9, v9, v11, vcc
+; CGP-NEXT:    v_sub_i32_e32 v11, vcc, v8, v4
+; CGP-NEXT:    v_subbrev_u32_e32 v14, vcc, 0, v1, vcc
+; CGP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
+; CGP-NEXT:    v_cndmask_b32_e32 v13, v18, v13, vcc
+; CGP-NEXT:    v_sub_i32_e32 v4, vcc, v12, v4
+; CGP-NEXT:    v_subbrev_u32_e32 v15, vcc, 0, v3, vcc
 ; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v9
-; CGP-NEXT:    v_cndmask_b32_e32 v7, v7, v13, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v8
-; CGP-NEXT:    v_cndmask_b32_e64 v8, v11, v12, s[4:5]
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v3, v14, vcc
-; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v4
-; CGP-NEXT:    v_cndmask_b32_e32 v2, v2, v7, vcc
-; CGP-NEXT:    v_cndmask_b32_e64 v1, v1, v15, s[4:5]
-; CGP-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v5
-; CGP-NEXT:    v_cndmask_b32_e64 v0, v0, v8, s[4:5]
-; CGP-NEXT:    v_cndmask_b32_e64 v1, v10, v1, s[4:5]
-; CGP-NEXT:    v_cndmask_b32_e32 v3, v6, v3, vcc
+; CGP-NEXT:    v_cndmask_b32_e32 v8, v8, v11, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v13
+; CGP-NEXT:    v_cndmask_b32_e64 v4, v12, v4, s[4:5]
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v1, v14, vcc
+; CGP-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v5
+; CGP-NEXT:    v_cndmask_b32_e32 v0, v0, v8, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v3, v3, v15, s[4:5]
+; CGP-NEXT:    v_cmp_ne_u32_e64 s[4:5], 0, v6
+; CGP-NEXT:    v_cndmask_b32_e64 v2, v2, v4, s[4:5]
+; CGP-NEXT:    v_cndmask_b32_e32 v1, v7, v1, vcc
+; CGP-NEXT:    v_cndmask_b32_e64 v3, v10, v3, s[4:5]
 ; CGP-NEXT:    s_setpc_b64 s[30:31]
   %result = urem <2 x i64> %num, <i64 1235195, i64 1235195>
   ret <2 x i64> %result
 }
 
 define i64 @v_urem_i64_pow2_shl_denom(i64 %x, i64 %y) {
 ; CHECK-LABEL: v_urem_i64_pow2_shl_denom:
 ; CHECK:       ; %bb.0:

llvm/test/CodeGen/AMDGPU/fp-min-max-atomics.ll

 ; G_SI-NEXT:    buffer_atomic_fmin v0, v1, s[4:7], 4 offen glc slc
 ; G_SI-NEXT:    v_mov_b32_e32 v1, s0
 ; G_SI-NEXT:    s_waitcnt vmcnt(0)
 ; G_SI-NEXT:    ds_write_b32 v1, v0
 ; G_SI-NEXT:    s_endpgm
 ;
 ; G_GFX7-LABEL: raw_buffer_atomic_min_rtn_f32_off4_slc:
 ; G_GFX7:       ; %bb.0: ; %main_body
-; G_GFX7-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0xd
+; G_GFX7-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0xd
 ; G_GFX7-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
-; G_GFX7-NEXT:    s_load_dword s2, s[0:1], 0xf
+; G_GFX7-NEXT:    s_load_dword s0, s[0:1], 0xf
 ; G_GFX7-NEXT:    s_mov_b32 m0, -1
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX7-NEXT:    v_mov_b32_e32 v0, s12
-; G_GFX7-NEXT:    v_mov_b32_e32 v1, s13
+; G_GFX7-NEXT:    v_mov_b32_e32 v0, s2
+; G_GFX7-NEXT:    v_mov_b32_e32 v1, s3
 ; G_GFX7-NEXT:    buffer_atomic_fmin v0, v1, s[4:7], 4 offen glc slc
-; G_GFX7-NEXT:    v_mov_b32_e32 v1, s2
+; G_GFX7-NEXT:    v_mov_b32_e32 v1, s0
 ; G_GFX7-NEXT:    s_waitcnt vmcnt(0)
 ; G_GFX7-NEXT:    ds_write_b32 v1, v0
 ; G_GFX7-NEXT:    s_endpgm
 ;
 ; G_GFX10-LABEL: raw_buffer_atomic_min_rtn_f32_off4_slc:
 ; G_GFX10:       ; %bb.0: ; %main_body
-; G_GFX10-NEXT:    s_clause 0x2
-; G_GFX10-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x34
+; G_GFX10-NEXT:    s_clause 0x1
+; G_GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
 ; G_GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
-; G_GFX10-NEXT:    s_load_dword s2, s[0:1], 0x3c
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX10-NEXT:    v_mov_b32_e32 v0, s12
-; G_GFX10-NEXT:    v_mov_b32_e32 v1, s13
+; G_GFX10-NEXT:    v_mov_b32_e32 v0, s2
+; G_GFX10-NEXT:    v_mov_b32_e32 v1, s3
+; G_GFX10-NEXT:    s_load_dword s0, s[0:1], 0x3c
 ; G_GFX10-NEXT:    buffer_atomic_fmin v0, v1, s[4:7], 4 offen glc slc
-; G_GFX10-NEXT:    v_mov_b32_e32 v1, s2
+; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; G_GFX10-NEXT:    v_mov_b32_e32 v1, s0
 ; G_GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; G_GFX10-NEXT:    ds_write_b32 v1, v0
 ; G_GFX10-NEXT:    s_endpgm
 ;
 ; G_GFX1030-LABEL: raw_buffer_atomic_min_rtn_f32_off4_slc:
 ; G_GFX1030:       ; %bb.0: ; %main_body
 ; G_GFX1030-NEXT:    s_clause 0x2
-; G_GFX1030-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x34
+; G_GFX1030-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
 ; G_GFX1030-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
-; G_GFX1030-NEXT:    s_load_dword s2, s[0:1], 0x3c
+; G_GFX1030-NEXT:    s_load_dword s0, s[0:1], 0x3c
 ; G_GFX1030-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX1030-NEXT:    v_mov_b32_e32 v0, s12
-; G_GFX1030-NEXT:    v_mov_b32_e32 v1, s13
+; G_GFX1030-NEXT:    v_mov_b32_e32 v0, s2
+; G_GFX1030-NEXT:    v_mov_b32_e32 v1, s3
 ; G_GFX1030-NEXT:    buffer_atomic_fmin v0, v1, s[4:7], 4 offen glc slc
-; G_GFX1030-NEXT:    v_mov_b32_e32 v1, s2
+; G_GFX1030-NEXT:    v_mov_b32_e32 v1, s0
 ; G_GFX1030-NEXT:    s_waitcnt vmcnt(0)
 ; G_GFX1030-NEXT:    ds_write_b32 v1, v0
 ; G_GFX1030-NEXT:    s_endpgm
 ; GFX1010-LABEL: raw_buffer_atomic_min_rtn_f32_off4_slc:
 main_body:
   %ret = call float @llvm.amdgcn.raw.buffer.atomic.fmin.f32(float %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
   store float %ret, float addrspace(3)* %out, align 8
   ret void

llvm/test/CodeGen/AMDGPU/lds-atomic-fmin-fmax.ll

 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_VI-NEXT:    buffer_store_dword v0, v1, s[88:91], 0 offen
 ; G_VI-NEXT:    s_endpgm
 ;
 ; G_GFX9-LABEL: lds_ds_fmin:
 ; G_GFX9:       ; %bb.0:
 ; G_GFX9-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX9-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
-; G_GFX9-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; G_GFX9-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x24
 ; G_GFX9-NEXT:    s_mov_b32 s10, -1
 ; G_GFX9-NEXT:    s_mov_b32 s11, 0xe00000
 ; G_GFX9-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX9-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; G_GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
 ; G_GFX9-NEXT:    s_addc_u32 s9, s9, 0
-; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX9-NEXT:    s_add_i32 s0, s2, 4
-; G_GFX9-NEXT:    s_lshl_b32 s1, s0, 3
-; G_GFX9-NEXT:    v_mov_b32_e32 v0, s1
 ; G_GFX9-NEXT:    v_mov_b32_e32 v1, 0x42280000
+; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; G_GFX9-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 3
+; G_GFX9-NEXT:    v_mov_b32_e32 v0, s0
 ; G_GFX9-NEXT:    ds_min_rtn_f32 v0, v0, v1
-; G_GFX9-NEXT:    s_lshl_b32 s0, s0, 4
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 4
 ; G_GFX9-NEXT:    v_mov_b32_e32 v2, s0
 ; G_GFX9-NEXT:    ds_min_rtn_f32 v1, v2, v1
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX9-NEXT:    v_mov_b32_e32 v1, s7
+; G_GFX9-NEXT:    v_mov_b32_e32 v1, s3
 ; G_GFX9-NEXT:    ds_min_rtn_f32 v0, v1, v0
-; G_GFX9-NEXT:    v_mov_b32_e32 v1, s6
+; G_GFX9-NEXT:    v_mov_b32_e32 v1, s2
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX9-NEXT:    buffer_store_dword v0, v1, s[8:11], 0 offen
 ; G_GFX9-NEXT:    s_endpgm
 ;
 ; G_GFX10-LABEL: lds_ds_fmin:
 ; G_GFX10:       ; %bb.0:
 ; G_GFX10-NEXT:    s_load_dword s2, s[0:1], 0x2c
 ; G_GFX10-NEXT:    s_mov_b32 s4, SCRATCH_RSRC_DWORD0

 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_VI-NEXT:    buffer_store_dword v0, v1, s[88:91], 0 offen
 ; G_VI-NEXT:    s_endpgm
 ;
 ; G_GFX9-LABEL: lds_ds_fmax:
 ; G_GFX9:       ; %bb.0:
 ; G_GFX9-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX9-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
-; G_GFX9-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; G_GFX9-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x24
 ; G_GFX9-NEXT:    s_mov_b32 s10, -1
 ; G_GFX9-NEXT:    s_mov_b32 s11, 0xe00000
 ; G_GFX9-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX9-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; G_GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
 ; G_GFX9-NEXT:    s_addc_u32 s9, s9, 0
-; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX9-NEXT:    s_add_i32 s0, s2, 4
-; G_GFX9-NEXT:    s_lshl_b32 s1, s0, 3
-; G_GFX9-NEXT:    v_mov_b32_e32 v0, s1
 ; G_GFX9-NEXT:    v_mov_b32_e32 v1, 0x42280000
+; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; G_GFX9-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 3
+; G_GFX9-NEXT:    v_mov_b32_e32 v0, s0
 ; G_GFX9-NEXT:    ds_max_rtn_f32 v0, v0, v1
-; G_GFX9-NEXT:    s_lshl_b32 s0, s0, 4
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 4
 ; G_GFX9-NEXT:    v_mov_b32_e32 v2, s0
 ; G_GFX9-NEXT:    ds_max_rtn_f32 v1, v2, v1
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX9-NEXT:    v_mov_b32_e32 v1, s7
+; G_GFX9-NEXT:    v_mov_b32_e32 v1, s3
 ; G_GFX9-NEXT:    ds_max_rtn_f32 v0, v1, v0
-; G_GFX9-NEXT:    v_mov_b32_e32 v1, s6
+; G_GFX9-NEXT:    v_mov_b32_e32 v1, s2
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX9-NEXT:    buffer_store_dword v0, v1, s[8:11], 0 offen
 ; G_GFX9-NEXT:    s_endpgm
 ;
 ; G_GFX10-LABEL: lds_ds_fmax:
 ; G_GFX10:       ; %bb.0:
 ; G_GFX10-NEXT:    s_load_dword s2, s[0:1], 0x2c
 ; G_GFX10-NEXT:    s_mov_b32 s4, SCRATCH_RSRC_DWORD0

 ; G_SI-NEXT:    v_mov_b32_e32 v0, s0
 ; G_SI-NEXT:    buffer_store_dword v1, v0, s[8:11], 0 offen
 ; G_SI-NEXT:    s_endpgm
 ;
 ; G_GFX7-LABEL: lds_ds_fmin_f64:
 ; G_GFX7:       ; %bb.0:
 ; G_GFX7-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX7-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
-; G_GFX7-NEXT:    s_load_dword s2, s[0:1], 0xb
+; G_GFX7-NEXT:    s_load_dword s4, s[0:1], 0xb
 ; G_GFX7-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; G_GFX7-NEXT:    s_mov_b32 s10, -1
 ; G_GFX7-NEXT:    s_mov_b32 s11, 0xe8f000
 ; G_GFX7-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX7-NEXT:    s_mov_b32 s2, 0
 ; G_GFX7-NEXT:    s_addc_u32 s9, s9, 0
-; G_GFX7-NEXT:    s_mov_b32 s4, 0
+; G_GFX7-NEXT:    s_mov_b32 s3, 0x40450000
+; G_GFX7-NEXT:    v_mov_b32_e32 v0, s2
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX7-NEXT:    s_add_i32 s2, s2, 4
-; G_GFX7-NEXT:    s_mov_b32 s5, 0x40450000
-; G_GFX7-NEXT:    v_mov_b32_e32 v0, s4
-; G_GFX7-NEXT:    s_lshl_b32 s3, s2, 3
-; G_GFX7-NEXT:    v_mov_b32_e32 v1, s5
-; G_GFX7-NEXT:    v_mov_b32_e32 v2, s3
+; G_GFX7-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX7-NEXT:    v_mov_b32_e32 v1, s3
+; G_GFX7-NEXT:    s_lshl_b32 s2, s4, 3
+; G_GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; G_GFX7-NEXT:    s_mov_b32 m0, -1
 ; G_GFX7-NEXT:    ds_min_rtn_f64 v[2:3], v2, v[0:1]
-; G_GFX7-NEXT:    s_lshl_b32 s2, s2, 4
+; G_GFX7-NEXT:    s_lshl_b32 s2, s4, 4
 ; G_GFX7-NEXT:    v_mov_b32_e32 v4, s2
 ; G_GFX7-NEXT:    ds_min_rtn_f64 v[0:1], v4, v[0:1]
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX7-NEXT:    v_mov_b32_e32 v0, s1
 ; G_GFX7-NEXT:    ds_min_rtn_f64 v[0:1], v0, v[2:3]
 ; G_GFX7-NEXT:    v_mov_b32_e32 v2, s0
 ; G_GFX7-NEXT:    s_add_u32 s0, s0, 4
 ; G_GFX7-NEXT:    v_mov_b32_e32 v3, s0
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX7-NEXT:    buffer_store_dword v0, v2, s[8:11], 0 offen
 ; G_GFX7-NEXT:    buffer_store_dword v1, v3, s[8:11], 0 offen
 ; G_GFX7-NEXT:    s_endpgm
 ;
 ; G_VI-LABEL: lds_ds_fmin_f64:
 ; G_VI:       ; %bb.0:
 ; G_VI-NEXT:    s_mov_b32 s88, SCRATCH_RSRC_DWORD0
 ; G_VI-NEXT:    s_mov_b32 s89, SCRATCH_RSRC_DWORD1
-; G_VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; G_VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
 ; G_VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; G_VI-NEXT:    s_mov_b32 s90, -1
 ; G_VI-NEXT:    s_mov_b32 s91, 0xe80000
 ; G_VI-NEXT:    s_add_u32 s88, s88, s3
+; G_VI-NEXT:    s_mov_b32 s2, 0
 ; G_VI-NEXT:    s_addc_u32 s89, s89, 0
-; G_VI-NEXT:    s_mov_b32 s4, 0
+; G_VI-NEXT:    s_mov_b32 s3, 0x40450000
+; G_VI-NEXT:    v_mov_b32_e32 v0, s2
 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
-; G_VI-NEXT:    s_add_i32 s2, s2, 4
-; G_VI-NEXT:    s_mov_b32 s5, 0x40450000
-; G_VI-NEXT:    v_mov_b32_e32 v0, s4
-; G_VI-NEXT:    s_lshl_b32 s3, s2, 3
-; G_VI-NEXT:    v_mov_b32_e32 v1, s5
-; G_VI-NEXT:    v_mov_b32_e32 v2, s3
+; G_VI-NEXT:    s_add_i32 s4, s4, 4
+; G_VI-NEXT:    v_mov_b32_e32 v1, s3
+; G_VI-NEXT:    s_lshl_b32 s2, s4, 3
+; G_VI-NEXT:    v_mov_b32_e32 v2, s2
 ; G_VI-NEXT:    s_mov_b32 m0, -1
 ; G_VI-NEXT:    ds_min_rtn_f64 v[2:3], v2, v[0:1]
-; G_VI-NEXT:    s_lshl_b32 s2, s2, 4
+; G_VI-NEXT:    s_lshl_b32 s2, s4, 4
 ; G_VI-NEXT:    v_mov_b32_e32 v4, s2
 ; G_VI-NEXT:    ds_min_rtn_f64 v[0:1], v4, v[0:1]
 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_VI-NEXT:    v_mov_b32_e32 v0, s1
 ; G_VI-NEXT:    ds_min_rtn_f64 v[0:1], v0, v[2:3]
 ; G_VI-NEXT:    v_mov_b32_e32 v2, s0
 ; G_VI-NEXT:    s_add_u32 s0, s0, 4
 ; G_VI-NEXT:    v_mov_b32_e32 v3, s0
 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_VI-NEXT:    buffer_store_dword v0, v2, s[88:91], 0 offen
 ; G_VI-NEXT:    buffer_store_dword v1, v3, s[88:91], 0 offen
 ; G_VI-NEXT:    s_endpgm
 ;
 ; G_GFX9-LABEL: lds_ds_fmin_f64:
 ; G_GFX9:       ; %bb.0:
 ; G_GFX9-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX9-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
-; G_GFX9-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; G_GFX9-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x24
 ; G_GFX9-NEXT:    s_mov_b32 s10, -1
 ; G_GFX9-NEXT:    s_mov_b32 s11, 0xe00000
-; G_GFX9-NEXT:    s_mov_b32 s0, 0
 ; G_GFX9-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX9-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; G_GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; G_GFX9-NEXT:    s_mov_b32 s0, 0
+; G_GFX9-NEXT:    s_addc_u32 s9, s9, 0
 ; G_GFX9-NEXT:    s_mov_b32 s1, 0x40450000
 ; G_GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; G_GFX9-NEXT:    s_addc_u32 s9, s9, 0
-; G_GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX9-NEXT:    s_add_i32 s0, s2, 4
-; G_GFX9-NEXT:    s_lshl_b32 s1, s0, 3
-; G_GFX9-NEXT:    v_mov_b32_e32 v2, s1
+; G_GFX9-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 3
+; G_GFX9-NEXT:    v_mov_b32_e32 v2, s0
 ; G_GFX9-NEXT:    ds_min_rtn_f64 v[2:3], v2, v[0:1]
-; G_GFX9-NEXT:    s_lshl_b32 s0, s0, 4
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 4
 ; G_GFX9-NEXT:    v_mov_b32_e32 v5, s0
-; G_GFX9-NEXT:    v_mov_b32_e32 v4, s7
+; G_GFX9-NEXT:    v_mov_b32_e32 v4, s3
 ; G_GFX9-NEXT:    ds_min_rtn_f64 v[0:1], v5, v[0:1]
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX9-NEXT:    ds_min_rtn_f64 v[0:1], v4, v[2:3]
-; G_GFX9-NEXT:    v_mov_b32_e32 v2, s6
+; G_GFX9-NEXT:    v_mov_b32_e32 v2, s2
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX9-NEXT:    buffer_store_dword v0, v2, s[8:11], 0 offen
 ; G_GFX9-NEXT:    buffer_store_dword v1, v2, s[8:11], 0 offen offset:4
 ; G_GFX9-NEXT:    s_endpgm
 ;
 ; G_GFX10-LABEL: lds_ds_fmin_f64:
 ; G_GFX10:       ; %bb.0:
-; G_GFX10-NEXT:    s_clause 0x1
-; G_GFX10-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; G_GFX10-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x24
 ; G_GFX10-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX10-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
 ; G_GFX10-NEXT:    s_mov_b32 s10, -1
 ; G_GFX10-NEXT:    s_mov_b32 s11, 0x31c16000
 ; G_GFX10-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX10-NEXT:    s_clause 0x1
+; G_GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; G_GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
 ; G_GFX10-NEXT:    s_addc_u32 s9, s9, 0
 ; G_GFX10-NEXT:    s_mov_b32 s0, 0
 ; G_GFX10-NEXT:    s_mov_b32 s1, 0x40450000
 ; G_GFX10-NEXT:    v_mov_b32_e32 v0, s0
 ; G_GFX10-NEXT:    v_mov_b32_e32 v1, s1
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX10-NEXT:    s_add_i32 s2, s2, 4
-; G_GFX10-NEXT:    v_mov_b32_e32 v5, s7
-; G_GFX10-NEXT:    s_lshl_b32 s3, s2, 3
-; G_GFX10-NEXT:    s_lshl_b32 s0, s2, 4
-; G_GFX10-NEXT:    v_mov_b32_e32 v2, s3
+; G_GFX10-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX10-NEXT:    v_mov_b32_e32 v5, s3
+; G_GFX10-NEXT:    s_lshl_b32 s5, s4, 3
+; G_GFX10-NEXT:    s_lshl_b32 s0, s4, 4
+; G_GFX10-NEXT:    v_mov_b32_e32 v2, s5
 ; G_GFX10-NEXT:    v_mov_b32_e32 v4, s0
 ; G_GFX10-NEXT:    ds_min_rtn_f64 v[2:3], v2, v[0:1]
 ; G_GFX10-NEXT:    ds_min_rtn_f64 v[0:1], v4, v[0:1]
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX10-NEXT:    ds_min_rtn_f64 v[0:1], v5, v[2:3]
-; G_GFX10-NEXT:    v_mov_b32_e32 v2, s6
+; G_GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX10-NEXT:    buffer_store_dword v0, v2, s[8:11], 0 offen
 ; G_GFX10-NEXT:    buffer_store_dword v1, v2, s[8:11], 0 offen offset:4
 ; G_GFX10-NEXT:    s_endpgm
   %idx.add = add nuw i32 %idx, 4
   %shl0 = shl i32 %idx.add, 3
   %shl1 = shl i32 %idx.add, 4
   %ptr0 = inttoptr i32 %shl0 to double addrspace(3)*

 ; G_SI-NEXT:    v_mov_b32_e32 v0, s0
 ; G_SI-NEXT:    buffer_store_dword v1, v0, s[8:11], 0 offen
 ; G_SI-NEXT:    s_endpgm
 ;
 ; G_GFX7-LABEL: lds_ds_fmax_f64:
 ; G_GFX7:       ; %bb.0:
 ; G_GFX7-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX7-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
-; G_GFX7-NEXT:    s_load_dword s2, s[0:1], 0xb
+; G_GFX7-NEXT:    s_load_dword s4, s[0:1], 0xb
 ; G_GFX7-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
 ; G_GFX7-NEXT:    s_mov_b32 s10, -1
 ; G_GFX7-NEXT:    s_mov_b32 s11, 0xe8f000
 ; G_GFX7-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX7-NEXT:    s_mov_b32 s2, 0
 ; G_GFX7-NEXT:    s_addc_u32 s9, s9, 0
-; G_GFX7-NEXT:    s_mov_b32 s4, 0
+; G_GFX7-NEXT:    s_mov_b32 s3, 0x40450000
+; G_GFX7-NEXT:    v_mov_b32_e32 v0, s2
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX7-NEXT:    s_add_i32 s2, s2, 4
-; G_GFX7-NEXT:    s_mov_b32 s5, 0x40450000
-; G_GFX7-NEXT:    v_mov_b32_e32 v0, s4
-; G_GFX7-NEXT:    s_lshl_b32 s3, s2, 3
-; G_GFX7-NEXT:    v_mov_b32_e32 v1, s5
-; G_GFX7-NEXT:    v_mov_b32_e32 v2, s3
+; G_GFX7-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX7-NEXT:    v_mov_b32_e32 v1, s3
+; G_GFX7-NEXT:    s_lshl_b32 s2, s4, 3
+; G_GFX7-NEXT:    v_mov_b32_e32 v2, s2
 ; G_GFX7-NEXT:    s_mov_b32 m0, -1
 ; G_GFX7-NEXT:    ds_max_rtn_f64 v[2:3], v2, v[0:1]
-; G_GFX7-NEXT:    s_lshl_b32 s2, s2, 4
+; G_GFX7-NEXT:    s_lshl_b32 s2, s4, 4
 ; G_GFX7-NEXT:    v_mov_b32_e32 v4, s2
 ; G_GFX7-NEXT:    ds_max_rtn_f64 v[0:1], v4, v[0:1]
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX7-NEXT:    v_mov_b32_e32 v0, s1
 ; G_GFX7-NEXT:    ds_max_rtn_f64 v[0:1], v0, v[2:3]
 ; G_GFX7-NEXT:    v_mov_b32_e32 v2, s0
 ; G_GFX7-NEXT:    s_add_u32 s0, s0, 4
 ; G_GFX7-NEXT:    v_mov_b32_e32 v3, s0
 ; G_GFX7-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX7-NEXT:    buffer_store_dword v0, v2, s[8:11], 0 offen
 ; G_GFX7-NEXT:    buffer_store_dword v1, v3, s[8:11], 0 offen
 ; G_GFX7-NEXT:    s_endpgm
 ;
 ; G_VI-LABEL: lds_ds_fmax_f64:
 ; G_VI:       ; %bb.0:
 ; G_VI-NEXT:    s_mov_b32 s88, SCRATCH_RSRC_DWORD0
 ; G_VI-NEXT:    s_mov_b32 s89, SCRATCH_RSRC_DWORD1
-; G_VI-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; G_VI-NEXT:    s_load_dword s4, s[0:1], 0x2c
 ; G_VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
 ; G_VI-NEXT:    s_mov_b32 s90, -1
 ; G_VI-NEXT:    s_mov_b32 s91, 0xe80000
 ; G_VI-NEXT:    s_add_u32 s88, s88, s3
+; G_VI-NEXT:    s_mov_b32 s2, 0
 ; G_VI-NEXT:    s_addc_u32 s89, s89, 0
-; G_VI-NEXT:    s_mov_b32 s4, 0
+; G_VI-NEXT:    s_mov_b32 s3, 0x40450000
+; G_VI-NEXT:    v_mov_b32_e32 v0, s2
 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
-; G_VI-NEXT:    s_add_i32 s2, s2, 4
-; G_VI-NEXT:    s_mov_b32 s5, 0x40450000
-; G_VI-NEXT:    v_mov_b32_e32 v0, s4
-; G_VI-NEXT:    s_lshl_b32 s3, s2, 3
-; G_VI-NEXT:    v_mov_b32_e32 v1, s5
-; G_VI-NEXT:    v_mov_b32_e32 v2, s3
+; G_VI-NEXT:    s_add_i32 s4, s4, 4
+; G_VI-NEXT:    v_mov_b32_e32 v1, s3
+; G_VI-NEXT:    s_lshl_b32 s2, s4, 3
+; G_VI-NEXT:    v_mov_b32_e32 v2, s2
 ; G_VI-NEXT:    s_mov_b32 m0, -1
 ; G_VI-NEXT:    ds_max_rtn_f64 v[2:3], v2, v[0:1]
-; G_VI-NEXT:    s_lshl_b32 s2, s2, 4
+; G_VI-NEXT:    s_lshl_b32 s2, s4, 4
 ; G_VI-NEXT:    v_mov_b32_e32 v4, s2
 ; G_VI-NEXT:    ds_max_rtn_f64 v[0:1], v4, v[0:1]
 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_VI-NEXT:    v_mov_b32_e32 v0, s1
 ; G_VI-NEXT:    ds_max_rtn_f64 v[0:1], v0, v[2:3]
 ; G_VI-NEXT:    v_mov_b32_e32 v2, s0
 ; G_VI-NEXT:    s_add_u32 s0, s0, 4
 ; G_VI-NEXT:    v_mov_b32_e32 v3, s0
 ; G_VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_VI-NEXT:    buffer_store_dword v0, v2, s[88:91], 0 offen
 ; G_VI-NEXT:    buffer_store_dword v1, v3, s[88:91], 0 offen
 ; G_VI-NEXT:    s_endpgm
 ;
 ; G_GFX9-LABEL: lds_ds_fmax_f64:
 ; G_GFX9:       ; %bb.0:
 ; G_GFX9-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX9-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
-; G_GFX9-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; G_GFX9-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x24
 ; G_GFX9-NEXT:    s_mov_b32 s10, -1
 ; G_GFX9-NEXT:    s_mov_b32 s11, 0xe00000
-; G_GFX9-NEXT:    s_mov_b32 s0, 0
 ; G_GFX9-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX9-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; G_GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; G_GFX9-NEXT:    s_mov_b32 s0, 0
+; G_GFX9-NEXT:    s_addc_u32 s9, s9, 0
 ; G_GFX9-NEXT:    s_mov_b32 s1, 0x40450000
 ; G_GFX9-NEXT:    v_mov_b32_e32 v0, s0
-; G_GFX9-NEXT:    s_addc_u32 s9, s9, 0
-; G_GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX9-NEXT:    s_add_i32 s0, s2, 4
-; G_GFX9-NEXT:    s_lshl_b32 s1, s0, 3
-; G_GFX9-NEXT:    v_mov_b32_e32 v2, s1
+; G_GFX9-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX9-NEXT:    v_mov_b32_e32 v1, s1
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 3
+; G_GFX9-NEXT:    v_mov_b32_e32 v2, s0
 ; G_GFX9-NEXT:    ds_max_rtn_f64 v[2:3], v2, v[0:1]
-; G_GFX9-NEXT:    s_lshl_b32 s0, s0, 4
+; G_GFX9-NEXT:    s_lshl_b32 s0, s4, 4
 ; G_GFX9-NEXT:    v_mov_b32_e32 v5, s0
-; G_GFX9-NEXT:    v_mov_b32_e32 v4, s7
+; G_GFX9-NEXT:    v_mov_b32_e32 v4, s3
 ; G_GFX9-NEXT:    ds_max_rtn_f64 v[0:1], v5, v[0:1]
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX9-NEXT:    ds_max_rtn_f64 v[0:1], v4, v[2:3]
-; G_GFX9-NEXT:    v_mov_b32_e32 v2, s6
+; G_GFX9-NEXT:    v_mov_b32_e32 v2, s2
 ; G_GFX9-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX9-NEXT:    buffer_store_dword v0, v2, s[8:11], 0 offen
 ; G_GFX9-NEXT:    buffer_store_dword v1, v2, s[8:11], 0 offen offset:4
 ; G_GFX9-NEXT:    s_endpgm
 ;
 ; G_GFX10-LABEL: lds_ds_fmax_f64:
 ; G_GFX10:       ; %bb.0:
-; G_GFX10-NEXT:    s_clause 0x1
-; G_GFX10-NEXT:    s_load_dword s2, s[0:1], 0x2c
-; G_GFX10-NEXT:    s_load_dwordx2 s[6:7], s[0:1], 0x24
 ; G_GFX10-NEXT:    s_mov_b32 s8, SCRATCH_RSRC_DWORD0
 ; G_GFX10-NEXT:    s_mov_b32 s9, SCRATCH_RSRC_DWORD1
 ; G_GFX10-NEXT:    s_mov_b32 s10, -1
 ; G_GFX10-NEXT:    s_mov_b32 s11, 0x31c16000
 ; G_GFX10-NEXT:    s_add_u32 s8, s8, s3
+; G_GFX10-NEXT:    s_clause 0x1
+; G_GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; G_GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
 ; G_GFX10-NEXT:    s_addc_u32 s9, s9, 0
 ; G_GFX10-NEXT:    s_mov_b32 s0, 0
 ; G_GFX10-NEXT:    s_mov_b32 s1, 0x40450000
 ; G_GFX10-NEXT:    v_mov_b32_e32 v0, s0
 ; G_GFX10-NEXT:    v_mov_b32_e32 v1, s1
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; G_GFX10-NEXT:    s_add_i32 s2, s2, 4
-; G_GFX10-NEXT:    v_mov_b32_e32 v5, s7
-; G_GFX10-NEXT:    s_lshl_b32 s3, s2, 3
-; G_GFX10-NEXT:    s_lshl_b32 s0, s2, 4
-; G_GFX10-NEXT:    v_mov_b32_e32 v2, s3
+; G_GFX10-NEXT:    s_add_i32 s4, s4, 4
+; G_GFX10-NEXT:    v_mov_b32_e32 v5, s3
+; G_GFX10-NEXT:    s_lshl_b32 s5, s4, 3
+; G_GFX10-NEXT:    s_lshl_b32 s0, s4, 4
+; G_GFX10-NEXT:    v_mov_b32_e32 v2, s5
 ; G_GFX10-NEXT:    v_mov_b32_e32 v4, s0
 ; G_GFX10-NEXT:    ds_max_rtn_f64 v[2:3], v2, v[0:1]
 ; G_GFX10-NEXT:    ds_max_rtn_f64 v[0:1], v4, v[0:1]
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX10-NEXT:    ds_max_rtn_f64 v[0:1], v5, v[2:3]
-; G_GFX10-NEXT:    v_mov_b32_e32 v2, s6
+; G_GFX10-NEXT:    v_mov_b32_e32 v2, s2
 ; G_GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; G_GFX10-NEXT:    buffer_store_dword v0, v2, s[8:11], 0 offen
 ; G_GFX10-NEXT:    buffer_store_dword v1, v2, s[8:11], 0 offen offset:4
 ; G_GFX10-NEXT:    s_endpgm
   %idx.add = add nuw i32 %idx, 4
   %shl0 = shl i32 %idx.add, 3
   %shl1 = shl i32 %idx.add, 4
   %ptr0 = inttoptr i32 %shl0 to double addrspace(3)*
   %ptr1 = inttoptr i32 %shl1 to double addrspace(3)*
   %a1 = call double @llvm.amdgcn.ds.fmax.f64(double addrspace(3)* %ptr0, double 4.2e+1, i32 0, i32 0, i1 false)
   %a2 = call double @llvm.amdgcn.ds.fmax.f64(double addrspace(3)* %ptr1, double 4.2e+1, i32 0, i32 0, i1 false)
   %a3 = call double @llvm.amdgcn.ds.fmax.f64(double addrspace(3)* %ptrf, double %a1, i32 0, i32 0, i1 false)
   store double %a3, double addrspace(5)* %out
   ret void
 }

llvm/test/CodeGen/AMDGPU/twoaddr-constrain.ll

 ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 ; RUN: llc -global-isel -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -stop-after twoaddressinstruction < %s | FileCheck %s
 
 ; Check that %16 gets constrained to register class sgpr_96_with_sub0_sub1.
 define amdgpu_ps <3 x i32> @s_load_constant_v3i32_align4(<3 x i32> addrspace(4)* inreg %ptr) {
   ; CHECK-LABEL: name: s_load_constant_v3i32_align4
   ; CHECK: bb.0 (%ir-block.0):
   ; CHECK-NEXT:   liveins: $sgpr0, $sgpr1
   ; CHECK-NEXT: {{  $}}
   ; CHECK-NEXT:   [[COPY:%[0-9]+]]:sreg_32 = COPY killed $sgpr0
   ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:sreg_32 = COPY killed $sgpr1
   ; CHECK-NEXT:   undef %0.sub0:sreg_64 = COPY killed [[COPY]]
   ; CHECK-NEXT:   %0.sub1:sreg_64 = COPY killed [[COPY1]]
   ; CHECK-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 0, 0 :: (load (<2 x s32>) from %ir.ptr, align 4, addrspace 4)
   ; CHECK-NEXT:   [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM killed %0, 8, 0 :: (load (s32) from %ir.ptr + 8, addrspace 4)
-  ; CHECK-NEXT:   undef %16.sub0_sub1:sgpr_96_with_sub0_sub1 = COPY killed [[S_LOAD_DWORDX2_IMM]]
-  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sgpr_96_with_sub0_sub1 = COPY killed %16
-  ; CHECK-NEXT:   [[COPY2]].sub2:sgpr_96_with_sub0_sub1 = COPY undef [[S_LOAD_DWORD_IMM]]
-  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY [[COPY2]].sub0
-  ; CHECK-NEXT:   $sgpr0 = COPY killed [[COPY3]]
-  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY killed [[COPY2]].sub1
-  ; CHECK-NEXT:   $sgpr1 = COPY killed [[COPY4]]
-  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:sreg_32 = COPY killed [[S_LOAD_DWORD_IMM]]
-  ; CHECK-NEXT:   $sgpr2 = COPY killed [[COPY5]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:sreg_32 = COPY [[S_LOAD_DWORDX2_IMM]].sub0
+  ; CHECK-NEXT:   $sgpr0 = COPY killed [[COPY2]]
+  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:sreg_32 = COPY killed [[S_LOAD_DWORDX2_IMM]].sub1
+  ; CHECK-NEXT:   $sgpr1 = COPY killed [[COPY3]]
+  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:sreg_32 = COPY killed [[S_LOAD_DWORD_IMM]]
+  ; CHECK-NEXT:   $sgpr2 = COPY killed [[COPY4]]
   ; CHECK-NEXT:   SI_RETURN_TO_EPILOG implicit killed $sgpr0, implicit killed $sgpr1, implicit killed $sgpr2
   %load = load <3 x i32>, <3 x i32> addrspace(4)* %ptr, align 4
   ret <3 x i32> %load
 }

llvm/unittests/CodeGen/GlobalISel/LegalizerHelperTest.cpp

   B.setInstr(*And);
   EXPECT_TRUE(Helper.fewerElementsVector(*And, 0, V2S32) ==
               LegalizerHelper::LegalizeResult::Legalized);
 
   auto CheckStr = R"(
   CHECK: [[IMP_DEF0:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
   CHECK: [[IMP_DEF1:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
   CHECK: [[VALUE0:%[0-9]+]]:_(s32), [[VALUE1:%[0-9]+]]:_(s32), [[VALUE2:%[0-9]+]]:_(s32), [[VALUE3:%[0-9]+]]:_(s32), [[VALUE4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[IMP_DEF0]]:_(<5 x s32>)
-  CHECK: [[IMP_DEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
   CHECK: [[VECTOR0:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE0]]:_(s32), [[VALUE1]]:_(s32)
   CHECK: [[VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE2]]:_(s32), [[VALUE3]]:_(s32)
-  CHECK: [[VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE4]]:_(s32), [[IMP_DEF2]]:_(s32)
-  CHECK: [[IMP_DEF3:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
   CHECK: [[VALUE5:%[0-9]+]]:_(s32), [[VALUE6:%[0-9]+]]:_(s32), [[VALUE7:%[0-9]+]]:_(s32), [[VALUE8:%[0-9]+]]:_(s32), [[VALUE9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[IMP_DEF1]]:_(<5 x s32>)
-  CHECK: [[IMP_DEF4:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-  CHECK: [[VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE5]]:_(s32), [[VALUE6]]:_(s32)
-  CHECK: [[VECTOR4:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE7]]:_(s32), [[VALUE8]]:_(s32)
-  CHECK: [[VECTOR5:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE9]]:_(s32), [[IMP_DEF4]]:_(s32)
-  CHECK: [[IMP_DEF5:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
-
-  CHECK: [[AND0:%[0-9]+]]:_(<2 x s32>) = G_AND [[VECTOR0]]:_, [[VECTOR3]]:_
-  CHECK: [[AND1:%[0-9]+]]:_(<2 x s32>) = G_AND [[VECTOR1]]:_, [[VECTOR4]]:_
-  CHECK: [[AND2:%[0-9]+]]:_(<2 x s32>) = G_AND [[VECTOR2]]:_, [[VECTOR5]]:_
-  CHECK: [[IMP_DEF6:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
-
-  CHECK: [[VECTOR6:%[0-9]+]]:_(<10 x s32>) = G_CONCAT_VECTORS [[AND0]]:_(<2 x s32>), [[AND1]]:_(<2 x s32>), [[AND2]]:_(<2 x s32>), [[IMP_DEF6]]:_(<2 x s32>), [[IMP_DEF6]]:_(<2 x s32>)
-  CHECK: [[VECTOR7:%[0-9]+]]:_(<10 x s32>) = G_CONCAT_VECTORS [[AND0]]:_(<2 x s32>), [[AND1]]:_(<2 x s32>), [[AND2]]:_(<2 x s32>), [[IMP_DEF6]]:_(<2 x s32>), [[IMP_DEF6]]:_(<2 x s32>)
-  CHECK: [[VECTOR8:%[0-9]+]]:_(<5 x s32>), [[VECTOR9:%[0-9]+]]:_(<5 x s32>) = G_UNMERGE_VALUES [[VECTOR7]]:_(<10 x s32>)
+  CHECK: [[VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE5]]:_(s32), [[VALUE6]]:_(s32)
+  CHECK: [[VECTOR3:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[VALUE7]]:_(s32), [[VALUE8]]:_(s32)
+  CHECK: [[AND0:%[0-9]+]]:_(<2 x s32>) = G_AND [[VECTOR0]]:_, [[VECTOR2]]:_
+  CHECK: [[AND1:%[0-9]+]]:_(<2 x s32>) = G_AND [[VECTOR1]]:_, [[VECTOR3]]:_
+  CHECK: [[AND2:%[0-9]+]]:_(s32) = G_AND [[VALUE4]]:_, [[VALUE9]]:_
+  CHECK: [[AND0_E0:%[0-9]+]]:_(s32), [[AND0_E1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND0]]:_(<2 x s32>)
+  CHECK: [[AND1_E0:%[0-9]+]]:_(s32), [[AND1_E1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND1]]:_(<2 x s32>)
+  CHECK: [[RESULT:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[AND0_E0]]:_(s32), [[AND0_E1]]:_(s32), [[AND1_E0]]:_(s32), [[AND1_E1]]:_(s32), [[AND2]]:_(s32)
   )";
 
   // Check
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 TEST_F(AArch64GISelMITest, MoreElementsAnd) {
   setUp();

 
   B.setInstr(*And);
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
             Helper.moreElementsVector(*And, 0, v6s32));
 
   auto CheckStr = R"(
   CHECK: [[BITCAST0:%[0-9]+]]:_(<2 x s32>) = G_BITCAST
   CHECK: [[BITCAST1:%[0-9]+]]:_(<2 x s32>) = G_BITCAST
-  CHECK: [[IMP_DEF0:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
-  CHECK: [[CONCAT0:%[0-9]+]]:_(<6 x s32>) = G_CONCAT_VECTORS [[BITCAST0]]:_(<2 x s32>), [[IMP_DEF0]]:_(<2 x s32>), [[IMP_DEF0]]:_(<2 x s32>)
-  CHECK: [[IMP_DEF1:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
-  CHECK: [[CONCAT1:%[0-9]+]]:_(<6 x s32>) = G_CONCAT_VECTORS [[BITCAST1]]:_(<2 x s32>), [[IMP_DEF1]]:_(<2 x s32>), [[IMP_DEF1]]:_(<2 x s32>)
-  CHECK: [[AND:%[0-9]+]]:_(<6 x s32>) = G_AND [[CONCAT0]]:_, [[CONCAT1]]:_
-  CHECK: (<2 x s32>) = G_UNMERGE_VALUES [[AND]]:_(<6 x s32>)
+
+  CHECK: [[BITCAST0_E0:%[0-9]+]]:_(s32), [[BITCAST0_E1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST0]]:_(<2 x s32>)
+  CHECK: [[IMP_DEF0:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+  CHECK: [[BITCAST0_LARGE:%[0-9]+]]:_(<6 x s32>) = G_BUILD_VECTOR [[BITCAST0_E0]]:_(s32), [[BITCAST0_E1]]:_(s32), [[IMP_DEF0]]:_(s32), [[IMP_DEF0]]:_(s32), [[IMP_DEF0]]:_(s32), [[IMP_DEF0]]:_(s32)
+
+  CHECK: [[BITCAST1_E0:%[0-9]+]]:_(s32), [[BITCAST1_E1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]]:_(<2 x s32>)
+  CHECK: [[IMP_DEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+  CHECK: [[BITCAST1_LARGE:%[0-9]+]]:_(<6 x s32>) = G_BUILD_VECTOR [[BITCAST1_E0]]:_(s32), [[BITCAST1_E1]]:_(s32), [[IMP_DEF1]]:_(s32), [[IMP_DEF1]]:_(s32), [[IMP_DEF1]]:_(s32), [[IMP_DEF1]]:_(s32)
+
+  CHECK: [[AND:%[0-9]+]]:_(<6 x s32>) = G_AND [[BITCAST0_LARGE]]:_, [[BITCAST1_LARGE]]:_
+
+  CHECK: [[AND_E0:%[0-9]+]]:_(s32), [[AND_E1:%[0-9]+]]:_(s32), [[AND_E2:%[0-9]+]]:_(s32), [[AND_E3:%[0-9]+]]:_(s32), [[AND_E4:%[0-9]+]]:_(s32), [[AND_E5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[AND]]:_(<6 x s32>)
+  CHECK: (<2 x s32>) = G_BUILD_VECTOR [[AND_E0]]:_(s32), [[AND_E1]]:_(s32)
   )";
 
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 TEST_F(AArch64GISelMITest, FewerElementsPhi) {
   setUp();
   if (!TM)

   B.buildAnd(PhiTy, Phi.getReg(0), Phi.getReg(0));
 
   B.setInstr(*Phi);
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
             Helper.fewerElementsVector(*Phi, 0, v2s32));
 
   auto CheckStr = R"(
   CHECK: [[INITVAL:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
-  CHECK: [[EXTRACT0:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[INITVAL]]:_(<5 x s32>), 0
-  CHECK: [[EXTRACT1:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[INITVAL]]:_(<5 x s32>), 64
-  CHECK: [[EXTRACT2:%[0-9]+]]:_(s32) = G_EXTRACT [[INITVAL]]:_(<5 x s32>), 128
+  CHECK: [[INITVAL_E0:%[0-9]+]]:_(s32), [[INITVAL_E1:%[0-9]+]]:_(s32), [[INITVAL_E2:%[0-9]+]]:_(s32), [[INITVAL_E3:%[0-9]+]]:_(s32), [[INITVAL_E4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[INITVAL]]:_(<5 x s32>)
+  CHECK: [[INITVAL_E01:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INITVAL_E0]]:_(s32), [[INITVAL_E1]]:_(s32)
+  CHECK: [[INITVAL_E23:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[INITVAL_E2]]:_(s32), [[INITVAL_E3]]:_(s32)
   CHECK: G_BRCOND
 
   CHECK: [[MIDVAL:%[0-9]+]]:_(<5 x s32>) = G_IMPLICIT_DEF
-  CHECK: [[EXTRACT3:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[MIDVAL]]:_(<5 x s32>), 0
-  CHECK: [[EXTRACT4:%[0-9]+]]:_(<2 x s32>) = G_EXTRACT [[MIDVAL]]:_(<5 x s32>), 64
-  CHECK: [[EXTRACT5:%[0-9]+]]:_(s32) = G_EXTRACT [[MIDVAL]]:_(<5 x s32>), 128
+  CHECK: [[MIDVAL_E0:%[0-9]+]]:_(s32), [[MIDVAL_E1:%[0-9]+]]:_(s32), [[MIDVAL_E2:%[0-9]+]]:_(s32), [[MIDVAL_E3:%[0-9]+]]:_(s32), [[MIDVAL_E4:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[MIDVAL]]:_(<5 x s32>)
+  CHECK: [[MIDVAL_E01:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[MIDVAL_E0]]:_(s32), [[MIDVAL_E1]]:_(s32)
+  CHECK: [[MIDVAL_E23:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[MIDVAL_E2]]:_(s32), [[MIDVAL_E3]]:_(s32)
   CHECK: G_BR
 
-  CHECK: [[PHI0:%[0-9]+]]:_(<2 x s32>) = G_PHI [[EXTRACT0]]:_(<2 x s32>), %bb.0, [[EXTRACT3]]:_(<2 x s32>), %bb.1
-  CHECK: [[PHI1:%[0-9]+]]:_(<2 x s32>) = G_PHI [[EXTRACT1]]:_(<2 x s32>), %bb.0, [[EXTRACT4]]:_(<2 x s32>), %bb.1
-  CHECK: [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[EXTRACT2]]:_(s32), %bb.0, [[EXTRACT5]]:_(s32), %bb.1
-
-  CHECK: [[OTHER_PHI:%[0-9]+]]:_(s64) = G_PHI
-
+  CHECK: [[PHI0:%[0-9]+]]:_(<2 x s32>) = G_PHI [[INITVAL_E01]]:_(<2 x s32>), %bb.0, [[MIDVAL_E01]]:_(<2 x s32>), %bb.1
+  CHECK: [[PHI1:%[0-9]+]]:_(<2 x s32>) = G_PHI [[INITVAL_E23]]:_(<2 x s32>), %bb.0, [[MIDVAL_E23]]:_(<2 x s32>), %bb.1
+  CHECK: [[PHI2:%[0-9]+]]:_(s32) = G_PHI [[INITVAL_E4]]:_(s32), %bb.0, [[MIDVAL_E4]]:_(s32), %bb.1
   CHECK: [[UNMERGE0:%[0-9]+]]:_(s32), [[UNMERGE1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[PHI0]]:_(<2 x s32>)
   CHECK: [[UNMERGE2:%[0-9]+]]:_(s32), [[UNMERGE3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[PHI1]]:_(<2 x s32>)
   CHECK: [[BV:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[UNMERGE0]]:_(s32), [[UNMERGE1]]:_(s32), [[UNMERGE2]]:_(s32), [[UNMERGE3]]:_(s32), [[PHI2]]:_(s32)
+
+  CHECK: [[OTHER_PHI:%[0-9]+]]:_(s64) = G_PHI
+
   CHECK: [[USE_OP:%[0-9]+]]:_(<5 x s32>) = G_AND [[BV]]:_, [[BV]]:_
   )";
 
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 // FNEG expansion in terms of XOR
 TEST_F(AArch64GISelMITest, LowerFNEG) {

   CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[PTRTOINT1]]
   CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT
   CHECK: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[ZEXT]]:_, [[C]]:_(s64)
   CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT
   CHECK: [[AND:%[0-9]+]]:_(s64) = G_AND [[PTRTOINT]]:_, [[C]]:_
   CHECK: [[OR:%[0-9]+]]:_(s64) = G_OR [[AND]]:_, [[SHL]]:_
   CHECK: [[INTTOPTR:%[0-9]+]]:_(p0) = G_INTTOPTR [[OR]]
 
-  CHECK: [[BITCAST:%[0-9]+]]:_(s64) = G_BITCAST [[V2S32]]
-  CHECK: [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[S32]]
-  CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT
-  CHECK: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[ZEXT]]:_, [[C]]:_(s64)
-  CHECK: [[C:%[0-9]+]]:_(s64) = G_CONSTANT
-  CHECK: [[AND:%[0-9]+]]:_(s64) = G_AND [[BITCAST]]:_, [[C]]:_
-  CHECK: [[OR:%[0-9]+]]:_(s64) = G_OR [[AND]]:_, [[SHL]]:_
-  CHECK: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[OR]]
+  CHECK: [[V2S32_E0:%[0-9]+]]:_(s32), [[V2S32_E1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[V2S32]]
+  CHECK: [[BV:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[V2S32_E0]]:_(s32), [[S32]]:_(s32)
   )";
 
   // Check
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 // Test lowering of G_FFLOOR
 TEST_F(AArch64GISelMITest, LowerFFloor) {

 TEST_F(AArch64GISelMITest, NarrowFreeze) {
   setUp();
   if (!TM)
     return;
 
   DefineLegalizerInfo(A, {});
 
   // Make sure that G_FREEZE is narrowed using unmerge/extract
-  LLT S16{LLT::scalar(16)};
   LLT S32{LLT::scalar(32)};
   LLT S33{LLT::scalar(33)};
+  LLT S48{LLT::scalar(48)};
   LLT S64{LLT::scalar(64)};
   LLT V2S16{LLT::fixed_vector(2, 16)};
-  LLT V2S32{LLT::fixed_vector(2, 32)};
+  LLT V3S16{LLT::fixed_vector(3, 16)};
+  LLT V4S16{LLT::fixed_vector(4, 16)};
 
   auto Trunc = B.buildTrunc(S33, {Copies[0]});
-  auto Vector = B.buildBitcast(V2S32, Copies[0]);
+  auto Trunc1 = B.buildTrunc(S48, {Copies[0]});
+  auto Vector = B.buildBitcast(V3S16, Trunc1);
 
   auto FreezeScalar = B.buildInstr(TargetOpcode::G_FREEZE, {S64}, {Copies[0]});
   auto FreezeOdd = B.buildInstr(TargetOpcode::G_FREEZE, {S33}, {Trunc});
-  auto FreezeVector = B.buildInstr(TargetOpcode::G_FREEZE, {V2S32}, {Vector});
-  auto FreezeVector1 = B.buildInstr(TargetOpcode::G_FREEZE, {V2S32}, {Vector});
+  auto FreezeVector = B.buildInstr(TargetOpcode::G_FREEZE, {V3S16}, {Vector});
+  auto FreezeVector1 = B.buildInstr(TargetOpcode::G_FREEZE, {V3S16}, {Vector});
 
   AInfo Info(MF->getSubtarget());
   DummyGISelObserver Observer;
   LegalizerHelper Helper(*MF, Info, Observer, B);
 
   // Perform Legalization
 
   B.setInsertPt(*EntryMBB, FreezeScalar->getIterator());
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
             Helper.narrowScalar(*FreezeScalar, 0, S32));
 
+  // This should be followed by narrowScalar to S32.
   B.setInsertPt(*EntryMBB, FreezeOdd->getIterator());
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
-            Helper.narrowScalar(*FreezeOdd, 0, S32));
+            Helper.widenScalar(*FreezeOdd, 0, S64));
 
   B.setInsertPt(*EntryMBB, FreezeVector->getIterator());
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
-            Helper.narrowScalar(*FreezeVector, 0, V2S16));
+            Helper.fewerElementsVector(*FreezeVector, 0, V2S16));
 
+  // This should be followed by fewerElements to V2S16.
   B.setInsertPt(*EntryMBB, FreezeVector1->getIterator());
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
-            Helper.narrowScalar(*FreezeVector1, 0, S16));
+            Helper.moreElementsVector(*FreezeVector1, 0, V4S16));
 
   const auto *CheckStr = R"(
   CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY
   CHECK: [[TRUNC:%[0-9]+]]:_(s33) = G_TRUNC [[COPY]]
-  CHECK: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY]]
+  CHECK: [[TRUNC1:%[0-9]+]]:_(s48) = G_TRUNC [[COPY]]
+  CHECK: [[BITCAST:%[0-9]+]]:_(<3 x s16>) = G_BITCAST [[TRUNC1]]
 
   CHECK: [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]]
   CHECK: [[FREEZE:%[0-9]+]]:_(s32) = G_FREEZE [[UV]]
   CHECK: [[FREEZE1:%[0-9]+]]:_(s32) = G_FREEZE [[UV1]]
   CHECK: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[FREEZE]]:_(s32), [[FREEZE1]]
 
-  CHECK: (s1) = G_UNMERGE_VALUES [[TRUNC]]:_(s33)
-  CHECK: [[UNDEF:%[0-9]+]]:_(s1) = G_IMPLICIT_DEF
-  CHECK: [[MV1:%[0-9]+]]:_(s32) = G_MERGE_VALUES
-  CHECK: [[MV2:%[0-9]+]]:_(s32) = G_MERGE_VALUES
-  CHECK: [[UNDEF1:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-  CHECK: [[FREEZE2:%[0-9]+]]:_(s32) = G_FREEZE [[MV1]]
-  CHECK: [[FREEZE3:%[0-9]+]]:_(s32) = G_FREEZE [[MV2]]
-  CHECK: [[UNDEF2:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-  CHECK: [[MV3:%[0-9]+]]:_(s1056) = G_MERGE_VALUES [[FREEZE2]]:_(s32), [[FREEZE3]]:_(s32), [[UNDEF2]]
-  CHECK: [[TRUNC1:%[0-9]+]]:_(s33) = G_TRUNC [[MV3]]
-
-  CHECK: [[BITCAST1:%[0-9]+]]:_(s64) = G_BITCAST [[BITCAST]]
-  CHECK: [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST1]]
-  CHECK: [[FREEZE4:%[0-9]+]]:_(s32) = G_FREEZE [[UV2]]
-  CHECK: [[FREEZE5:%[0-9]+]]:_(s32) = G_FREEZE [[UV3]]
-  CHECK: [[MV4:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[FREEZE4]]:_(s32), [[FREEZE5]]:_(s32)
-  CHECK: [[BITCAST2:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[MV4]]
-
-  CHECK: [[BITCAST3:%[0-9]+]]:_(s64) = G_BITCAST [[BITCAST]]
-  CHECK: [[UV4:%[0-9]+]]:_(s16), [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST3]]
-  CHECK: [[FREEZE6:%[0-9]+]]:_(s16) = G_FREEZE [[UV4]]
-  CHECK: [[FREEZE7:%[0-9]+]]:_(s16) = G_FREEZE [[UV5]]
-  CHECK: [[FREEZE8:%[0-9]+]]:_(s16) = G_FREEZE [[UV6]]
-  CHECK: [[FREEZE9:%[0-9]+]]:_(s16) = G_FREEZE [[UV7]]
-  CHECK: [[MV5:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[FREEZE6]]:_(s16), [[FREEZE7]]:_(s16), [[FREEZE8]]:_(s16), [[FREEZE9]]
-  CHECK: [[BITCAST3:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[MV5]]
+  CHECK: [[ANYEXT:%[0-9]+]]:_(s64) = G_ANYEXT [[TRUNC]]
+  CHECK: [[FREEZE2:%[0-9]+]]:_(s64) = G_FREEZE [[ANYEXT]]
+  CHECK: [[TRUNC1:%[0-9]+]]:_(s33) = G_TRUNC [[FREEZE2]]
+
+  CHECK: [[UV2:%[0-9]+]]:_(s16), [[UV3:%[0-9]+]]:_(s16), [[UV4:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]]
+  CHECK: [[BV:%[0-9]+]]:_(<2 x s16>) = G_BUILD_VECTOR [[UV2]]:_(s16), [[UV3]]:_(s16)
+  CHECK: [[FREEZE3:%[0-9]+]]:_(<2 x s16>) = G_FREEZE [[BV]]
+  CHECK: [[FREEZE4:%[0-9]+]]:_(s16) = G_FREEZE [[UV4]]
+  CHECK: [[FREEZE3_E0:%[0-9]+]]:_(s16), [[FREEZE3_E1:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[FREEZE3]]
+  CHECK: [[BV1:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[FREEZE3_E0]]:_(s16), [[FREEZE3_E1]]:_(s16), [[FREEZE4]]:_(s16)
+
+  CHECK: [[UV5:%[0-9]+]]:_(s16), [[UV6:%[0-9]+]]:_(s16), [[UV7:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[BITCAST]]
+  CHECK: [[IMP_DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+  CHECK: [[BV1:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[UV5]]:_(s16), [[UV6]]:_(s16), [[UV7]]:_(s16), [[IMP_DEF]]:_(s16)
+  CHECK: [[FREEZE5:%[0-9]+]]:_(<4 x s16>) = G_FREEZE [[BV1]]
+  CHECK: [[FREEZE5_E0:%[0-9]+]]:_(s16), [[FREEZE5_E1:%[0-9]+]]:_(s16), [[FREEZE5_E2:%[0-9]+]]:_(s16), [[FREEZE5_E3:%[0-9]+]]:_(s16) = G_UNMERGE_VALUES [[FREEZE5]]
+  CHECK: [[BV2:%[0-9]+]]:_(<3 x s16>) = G_BUILD_VECTOR [[FREEZE5_E0]]:_(s16), [[FREEZE5_E1]]:_(s16), [[FREEZE5_E2]]:_(s16)
   )";
 
   // Check
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 // Test fewer elements of G_FREEZE
 TEST_F(AArch64GISelMITest, FewerElementsFreeze) {

   // Perform Legalization
   B.setInsertPt(*EntryMBB, FreezeVector1->getIterator());
   EXPECT_EQ(LegalizerHelper::LegalizeResult::Legalized,
             Helper.moreElementsVector(*FreezeVector1, 0, V4S32));
 
   const auto *CheckStr = R"(
   CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY
   CHECK: [[BITCAST:%[0-9]+]]:_(<2 x s32>) = G_BITCAST [[COPY]]
-  CHECK: [[UNDEF:%[0-9]+]]:_(<2 x s32>) = G_IMPLICIT_DEF
-  CHECK: [[CV:%[0-9]+]]:_(<4 x s32>) = G_CONCAT_VECTORS [[BITCAST]]:_(<2 x s32>), [[UNDEF]]
-  CHECK: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[CV]]
-  CHECK: [[EXTR0:%[0-9]+]]:_(<2 x s32>), [[EXTR1:%[0-9]+]]:_(<2 x s32>) = G_UNMERGE_VALUES [[FREEZE]]:_(<4 x s32>)
+  CHECK: [[BITCAST_E0:%[0-9]+]]:_(s32), [[BITCAST_E1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[BITCAST]]:_(<2 x s32>)
+  CHECK: [[IMP_DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
+  CHECK: [[BITCAST_LARGE:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[BITCAST_E0]]:_(s32), [[BITCAST_E1]]:_(s32), [[IMP_DEF]]:_(s32), [[IMP_DEF]]:_(s32)
+  CHECK: [[FREEZE:%[0-9]+]]:_(<4 x s32>) = G_FREEZE [[BITCAST_LARGE]]
+  CHECK: [[FREEZE_E0:%[0-9]+]]:_(s32), [[FREEZE_E1:%[0-9]+]]:_(s32), [[FREEZE_E2:%[0-9]+]]:_(s32), [[FREEZE_E3:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[FREEZE]]:_(<4 x s32>)
+  CHECK: (<2 x s32>) = G_BUILD_VECTOR [[FREEZE_E0]]:_(s32), [[FREEZE_E1]]:_(s32)
   )";
 
   // Check
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 // Test fewer elements of G_INSERT_VECTOR_ELEMENT
 TEST_F(AArch64GISelMITest, FewerElementsInsertVectorElt) {

             Helper.moreElementsVector(*Shuffle, 0, LLT::fixed_vector(8, S64)));
 
   const auto *CheckStr = R"(
   CHECK: [[COPY0:%[0-9]+]]:_(s64) = COPY
   CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY
   CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY
   CHECK: [[BV1:%[0-9]+]]:_(<6 x s64>) = G_BUILD_VECTOR
   CHECK: [[BV2:%[0-9]+]]:_(<6 x s64>) = G_BUILD_VECTOR
-  CHECK: [[IMPDEF1:%[0-9]+]]:_(<8 x s64>) = G_IMPLICIT_DEF
-  CHECK: [[INSERT1:%[0-9]+]]:_(<8 x s64>) = G_INSERT [[IMPDEF1]]:_, [[BV1]]:_(<6 x s64>), 0
-  CHECK: [[IMPDEF2:%[0-9]+]]:_(<8 x s64>) = G_IMPLICIT_DEF
-  CHECK: [[INSERT2:%[0-9]+]]:_(<8 x s64>) = G_INSERT [[IMPDEF2]]:_, [[BV2]]:_(<6 x s64>), 0
-  CHECK: [[SHUF:%[0-9]+]]:_(<8 x s64>) = G_SHUFFLE_VECTOR [[INSERT1]]:_(<8 x s64>), [[INSERT2]]:_, shufflemask(3, 4, 9, 0, 1, 13, undef, undef)
-  CHECK: [[IMPDEF3:%[0-9]+]]:_(<8 x s64>) = G_IMPLICIT_DEF
-  CHECK: [[CONCAT:%[0-9]+]]:_(<24 x s64>) = G_CONCAT_VECTORS [[SHUF]]:_(<8 x s64>), [[IMPDEF3]]:_(<8 x s64>), [[IMPDEF3]]:_(<8 x s64>)
-  CHECK: [[UNMERGE:%[0-9]+]]:_(<6 x s64>), [[UNMERGE2:%[0-9]+]]:_(<6 x s64>), [[UNMERGE3:%[0-9]+]]:_(<6 x s64>), [[UNMERGE4:%[0-9]+]]:_(<6 x s64>) = G_UNMERGE_VALUES [[CONCAT]]:_(<24 x s64>)
+
+  CHECK: [[BV1_E0:%[0-9]+]]:_(s64), [[BV1_E1:%[0-9]+]]:_(s64), [[BV1_E2:%[0-9]+]]:_(s64), [[BV1_E3:%[0-9]+]]:_(s64), [[BV1_E4:%[0-9]+]]:_(s64), [[BV1_E5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[BV1]]:_(<6 x s64>)
+  CHECK: [[IMP_DEF0:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+  CHECK: [[BV1_LARGE:%[0-9]+]]:_(<8 x s64>) = G_BUILD_VECTOR [[BV1_E0]]:_(s64), [[BV1_E1]]:_(s64), [[BV1_E2]]:_(s64), [[BV1_E3]]:_(s64), [[BV1_E4]]:_(s64), [[BV1_E5]]:_(s64), [[IMP_DEF0]]:_(s64), [[IMP_DEF0]]:_(s64)
+
+  CHECK: [[BV2_E0:%[0-9]+]]:_(s64), [[BV2_E1:%[0-9]+]]:_(s64), [[BV2_E2:%[0-9]+]]:_(s64), [[BV2_E3:%[0-9]+]]:_(s64), [[BV2_E4:%[0-9]+]]:_(s64), [[BV2_E5:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[BV2]]:_(<6 x s64>)
+  CHECK: [[IMP_DEF1:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+  CHECK: [[BV2_LARGE:%[0-9]+]]:_(<8 x s64>) = G_BUILD_VECTOR [[BV2_E0]]:_(s64), [[BV2_E1]]:_(s64), [[BV2_E2]]:_(s64), [[BV2_E3]]:_(s64), [[BV2_E4]]:_(s64), [[BV2_E5]]:_(s64), [[IMP_DEF1]]:_(s64), [[IMP_DEF1]]:_(s64)
+
+  CHECK: [[SHUF:%[0-9]+]]:_(<8 x s64>) = G_SHUFFLE_VECTOR [[BV1_LARGE]]:_(<8 x s64>), [[BV2_LARGE]]:_, shufflemask(3, 4, 9, 0, 1, 13, undef, undef)
+
+  CHECK: [[SHUF_E0:%[0-9]+]]:_(s64), [[SHUF_E1:%[0-9]+]]:_(s64), [[SHUF_E2:%[0-9]+]]:_(s64), [[SHUF_E3:%[0-9]+]]:_(s64), [[SHUF_E4:%[0-9]+]]:_(s64), [[SHUF_E5:%[0-9]+]]:_(s64), [[SHUF_E6:%[0-9]+]]:_(s64), [[SHUF_E7:%[0-9]+]]:_(s64) = G_UNMERGE_VALUES [[SHUF]]:_(<8 x s64>)
+  CHECK: (<6 x s64>) = G_BUILD_VECTOR [[SHUF_E0]]:_(s64), [[SHUF_E1]]:_(s64), [[SHUF_E2]]:_(s64), [[SHUF_E3]]:_(s64), [[SHUF_E4]]:_(s64), [[SHUF_E5]]:_(s64)
   )";
 
   // Check
   EXPECT_TRUE(CheckMachineFunction(*MF, CheckStr)) << *MF;
 }
 
 // Test narror scalar of G_SHL with constant shift amount
 TEST_F(AArch64GISelMITest, narrowScalarShiftByConstant) {