[AMDGPU] Use instruction order in machine function to process workList of moveToVALU

[skc7]

Summary

Update moveToVALU method to accept worklist of instructions.
Top/first element of worklist will be picked up to process.

Instruction will be moved to VALU or legalizeOperands is called on it before all the instruction prior to it in machine function are processed. legalizeOperands will be able to process the operands together with this change.

This patch is required to legalize rsrc and soffset of MBUF instruction together and create single waterfall loop. Prerequisite for D141030

[foad]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

6205

Sorting the whole worklist every time round this loop is not acceptable. You need to find a more efficient way of doing this.

6210

Do you really need to guarantee the order between different basic blocks? If so, surely you need some kind of topological ordering, not just based on BB number. (And what if there are cycles in the CFG?)

[skc7]

Introduce SIInstrWorklist. It has a std::set with comparison operator to store instructions as per order in machine function. This tries to solve the previous issue with sorting the vector in every iteration.

[skc7]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

6205

Made changes in the current patch to use stl set with comparator for inserting and popping from worklist. This would not require to sort the worklist in every iteration. But the insertion and deletion from set have O(logn) time complexity.

6210

I'm finding it difficult to compare two instructions from different basic blocks other than using BB number. Can we use dominator tree information? May be DFS numbers be used to compare them?

[arsenm]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

6210

We used to do this in RPO. Not sure why this was changed.

Would it be easier to check all the operands of the one buffer instruction the first time you encounter it?

[skc7]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

6210

The first time a buffer instruction is processed in the worklist for legalizeOperands, both rsrc and soffset operands wouldn't have vgprs assigned by the pass. Only one of them would have vgpr. V2S copy result used by the other operand still need to be processed in the worklist.

So, if legalizeOperands has to legalize soffset and rsrc together, V2S copies that occur prior to buffer instruction have to be processed in the worklist first.

[arsenm]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

6210

Do they really need to be processed first? If you processed them at once, when those instructions are later processed they should see a copy with no uses which can be deleted

[skc7]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

6210

In the example below, si-fix-sgpr-copies pass calls moveToVALU for V2S copies in random order. So, BUFFER_LOAD instruction will have VGPRs for its soffset after processing %8:sreg_32 = COPY %5:vgpr_32 in the worklist. Legalize Operands would be called on the BUFFER_LOAD instruction which would only add a waterfall loop for soffset.

In further iterations, when rsrc would be converted to VGPR, then again LegalizeOperands would be called which would then add waterfall loop for it.

LegalizeOperands is called on the BUFFER_LOAD two times which is the cause of two waterfall loops in D141030.

So, the solution I think would work is to process the worklist instructions in order and each instruction will be called with legalizeOperands only once. This would make sure both rsrc and soffset are vgprs together once legalizeOperands is called. Then, D141030 logic can add single waterfall loop for them.

define float @llvm_amdgcn_raw_buffer_load_f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
  %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
  ret float %val
}

Input to si-fix-sgpr-copies pass:

Function Live Ins: $vgpr0 in %0, $vgpr1 in %1, $vgpr2 in %2, $vgpr3 in %3, $vgpr4 in %4, $vgpr5 in %5
bb.0 (%ir-block.0):
liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5
%5:vgpr_32 = COPY $vgpr5
%4:vgpr_32 = COPY $vgpr4
%3:vgpr_32 = COPY $vgpr3
%2:vgpr_32 = COPY $vgpr2
%1:vgpr_32 = COPY $vgpr1
%0:vgpr_32 = COPY $vgpr0
%9:sgpr_32 = COPY %0:vgpr_32
%10:sgpr_32 = COPY %1:vgpr_32
%11:sgpr_32 = COPY %2:vgpr_32
%12:sgpr_32 = COPY %3:vgpr_32
%6:sgpr_128 = REG_SEQUENCE %9:sgpr_32, %subreg.sub0, %10:sgpr_32, %subreg.sub1, %11:sgpr_32, %subreg.sub2, %12:sgpr_32, %subreg.sub3
%8:sreg_32 = COPY %5:vgpr_32
%7:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN %4:vgpr_32, killed %6:sgpr_128, %8:sreg_32, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 7)
$vgpr0 = COPY %7:vgpr_32
SI_RETURN implicit $vgpr0

++++++++++++++++++++++++++++++++++++++++++++++++

1st iteration
Top of worklist -> %8:sreg_32 = COPY %5:vgpr_32

%6:sgpr_128 = REG_SEQUENCE %9:sgpr_32, %subreg.sub0, %10:sgpr_32, %subreg.sub1, %11:sgpr_32, 
%subreg.sub2, %12:sgpr_32, %subreg.sub3 
%8:sreg_32 = COPY %5:vgpr_32 
%7:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN %4:vgpr_32, killed %6:sgpr_128, %8:sreg_32, 0, 0, 0, implicit $exec

++++++++++++++++++++++++++++++++++++++++++++++++

2nd iteration 
Top of worklist -> %11:sgpr_32 = COPY %2:vgpr_32

%6:sgpr_128 = REG_SEQUENCE %9:sgpr_32, %subreg.sub0, %10:sgpr_32, %subreg.sub1, %11:sgpr_32, 
%subreg.sub2, %12:sgpr_32, %subreg.sub3 
%8:sreg_32 = IMPLICIT_DEF 
%7:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN %4:vgpr_32, killed %6:sgpr_128, %5:vgpr_32, 0, 0, 0, implicit $exec

[skc7]

Use ReversePostOrderTraversal list to compare machine instructions from different basic blocks.

[arsenm]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

69

It doesn't make sense to me to construct RPO to create a list like this. The whole iteration would just work in RPO

75–77

This should be implied by the iteration order, shouldn't need to sort

[skc7]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

69

AFAIU, MBB order is created using RPOT list. New MBB might get created by legalizeOperands method once an instruction is processed in the worklist and the MBB order might change. So this list needs to be re-generated.
This list will be used by Cmp::Operator() to compare instructions from two different basic blocks.

75–77

New MachineInstructions might get added to worklist after it gets processed by legalizeOperands. So we need to maintain a sorted order of worklist. So, we need a Cmp::operator() for sort.

[arsenm]

Can this be abandoned now?

llvm/lib/Target/AMDGPU/SIFixSGPRCopies.cpp

if (DstReg == AMDGPU::M0 &&
             DefMI->eraseFromParent();
           MI.eraseFromParent();
         }
       }
     }
     return true;
   }
   if (!SrcReg.isVirtual() || TRI->isAGPR(*MRI, SrcReg)) {
-    TII->moveToVALU(MI, MDT);
+    SIInstrWorklist worklist;
+    worklist.insert(&MI);
+    TII->moveToVALU(worklist, MDT);
     return true;
   }
 
   unsigned SMovOp;
   int64_t Imm;
   // If we are just copying an immediate, we can replace the copy with
   // s_mov_b32.
   if (isSafeToFoldImmIntoCopy(&MI, MRI->getVRegDef(SrcReg), TII, SMovOp, Imm)) {

 void SIFixSGPRCopies::lowerVGPR2SGPRCopies(MachineFunction &MF) {
 
   SmallVector<unsigned, 8> LoweringWorklist;
   for (auto &C : V2SCopies) {
     if (needToBeConvertedToVALU(&C.second))
       LoweringWorklist.push_back(C.second.ID);
   }
 
+  // Store all the V2S copy instructions that need to be moved to VALU
+  // in the Copies worklist.
+  SIInstrWorklist Copies;
+
   while (!LoweringWorklist.empty()) {
     unsigned CurID = LoweringWorklist.pop_back_val();
     auto CurInfoIt = V2SCopies.find(CurID);
     if (CurInfoIt != V2SCopies.end()) {
       V2SCopyInfo C = CurInfoIt->getSecond();
       LLVM_DEBUG(dbgs() << "Processing ...\n"; C.dump());
       for (auto S : C.Siblings) {
         auto SibInfoIt = V2SCopies.find(S);
         if (SibInfoIt != V2SCopies.end()) {
           V2SCopyInfo &SI = SibInfoIt->getSecond();
           LLVM_DEBUG(dbgs() << "Sibling:\n"; SI.dump());
           if (!SI.NeedToBeConvertedToVALU) {
             set_subtract(SI.SChain, C.SChain);
             if (needToBeConvertedToVALU(&SI))
               LoweringWorklist.push_back(SI.ID);
           }
           SI.Siblings.remove_if([&](unsigned ID) { return ID == C.ID; });
         }
       }
       LLVM_DEBUG(dbgs() << "V2S copy " << *C.Copy
                         << " is being turned to VALU\n");
       V2SCopies.erase(C.ID);
-      TII->moveToVALU(*C.Copy, MDT);
+      Copies.insert(C.Copy);
     }
   }
 
+  TII->moveToVALU(Copies, MDT);
+  Copies.clear();
+
   // Now do actual lowering
   for (auto C : V2SCopies) {
     MachineInstr *MI = C.second.Copy;
     MachineBasicBlock *MBB = MI->getParent();
     // We decide to turn V2S copy to v_readfirstlane_b32
     // remove it from the V2SCopies and remove it from all its siblings
     LLVM_DEBUG(dbgs() << "V2S copy " << *MI
                       << " is being turned to v_readfirstlane_b32"

llvm/lib/Target/AMDGPU/SIInstrInfo.h

 class TargetRegisterClass;
 class ScheduleHazardRecognizer;
 
 /// Mark the MMO of a uniform load if there are no potentially clobbering stores
 /// on any path from the start of an entry function to this load.
 static const MachineMemOperand::Flags MONoClobber =
     MachineMemOperand::MOTargetFlag1;
 
+// Utility to store machine instructions worklist in order
+// as defined by comparator Cmp.
+struct SIInstrWorklist {
+  SIInstrWorklist() : InstrSet() {}
+
+  void insert(MachineInstr *MI) { InstrSet.insert(MI); }
+
+  MachineInstr *top() const {
+    auto iter = InstrSet.begin();
+    return *iter;
+  }
+
+  void erase_top() {
+    auto iter = InstrSet.begin();
+    InstrSet.erase(iter);
+  }
+
+  bool empty() const { return InstrSet.empty(); }
+
+  void clear() { InstrSet.clear(); }
+
+private:
+  struct Cmp {
+    Cmp() {}
+    bool operator()(const MachineInstr *InstA, const MachineInstr *InstB) const;
+  };
+  std::set<MachineInstr *, Cmp> InstrSet;
+};
+
 class SIInstrInfo final : public AMDGPUGenInstrInfo {
 private:
   const SIRegisterInfo RI;
   const GCNSubtarget &ST;
   TargetSchedModel SchedModel;
   mutable std::unique_ptr<AMDGPUMIRFormatter> Formatter;
 
   // The inverse predicate should have the negative value.
MachineOperand buildExtractSubRegOrImm(MachineBasicBlock::iterator MI,
                                          MachineOperand &SuperReg,
                                          const TargetRegisterClass *SuperRC,
                                          unsigned SubIdx,
                                          const TargetRegisterClass *SubRC) const;
 private:
   void swapOperands(MachineInstr &Inst) const;
 
   std::pair<bool, MachineBasicBlock *>
-  moveScalarAddSub(SetVectorType &Worklist, MachineInstr &Inst,
+  moveScalarAddSub(SIInstrWorklist &Worklist, MachineInstr &Inst,
                    MachineDominatorTree *MDT = nullptr) const;
 
-  void lowerSelect(SetVectorType &Worklist, MachineInstr &Inst,
+  void lowerSelect(SIInstrWorklist &Worklist, MachineInstr &Inst,
                    MachineDominatorTree *MDT = nullptr) const;
 
-  void lowerScalarAbs(SetVectorType &Worklist,
-                      MachineInstr &Inst) const;
+  void lowerScalarAbs(SIInstrWorklist &Worklist, MachineInstr &Inst) const;
 
-  void lowerScalarXnor(SetVectorType &Worklist,
-                       MachineInstr &Inst) const;
+  void lowerScalarXnor(SIInstrWorklist &Worklist, MachineInstr &Inst) const;
 
-  void splitScalarNotBinop(SetVectorType &Worklist,
-                           MachineInstr &Inst,
+  void splitScalarNotBinop(SIInstrWorklist &Worklist, MachineInstr &Inst,
                            unsigned Opcode) const;
 
-  void splitScalarBinOpN2(SetVectorType &Worklist,
-                          MachineInstr &Inst,
+  void splitScalarBinOpN2(SIInstrWorklist &Worklist, MachineInstr &Inst,
                           unsigned Opcode) const;
 
-  void splitScalar64BitUnaryOp(SetVectorType &Worklist,
-                               MachineInstr &Inst, unsigned Opcode,
-                               bool Swap = false) const;
+  void splitScalar64BitUnaryOp(SIInstrWorklist &Worklist, MachineInstr &Inst,
+                               unsigned Opcode, bool Swap = false) const;
 
-  void splitScalar64BitAddSub(SetVectorType &Worklist, MachineInstr &Inst,
+  void splitScalar64BitAddSub(SIInstrWorklist &Worklist, MachineInstr &Inst,
                               MachineDominatorTree *MDT = nullptr) const;
 
-  void splitScalar64BitBinaryOp(SetVectorType &Worklist, MachineInstr &Inst,
+  void splitScalar64BitBinaryOp(SIInstrWorklist &Worklist, MachineInstr &Inst,
                                 unsigned Opcode,
                                 MachineDominatorTree *MDT = nullptr) const;
 
-  void splitScalar64BitXnor(SetVectorType &Worklist, MachineInstr &Inst,
-                                MachineDominatorTree *MDT = nullptr) const;
+  void splitScalar64BitXnor(SIInstrWorklist &Worklist, MachineInstr &Inst,
 
-  void splitScalar64BitBCNT(SetVectorType &Worklist,
+  void splitScalar64BitBCNT(SIInstrWorklist &Worklist,
                             MachineInstr &Inst) const;
-  void splitScalar64BitBFE(SetVectorType &Worklist,
-                           MachineInstr &Inst) const;
-  void movePackToVALU(SetVectorType &Worklist,
-                      MachineRegisterInfo &MRI,
+  void splitScalar64BitBFE(SIInstrWorklist &Worklist, MachineInstr &Inst) const;
+  void movePackToVALU(SIInstrWorklist &Worklist, MachineRegisterInfo &MRI,
                       MachineInstr &Inst) const;
 
   void addUsersToMoveToVALUWorklist(Register Reg, MachineRegisterInfo &MRI,
-                                    SetVectorType &Worklist) const;
+                                    SIInstrWorklist &Worklist) const;
 
   void addSCCDefUsersToVALUWorklist(MachineOperand &Op,
                                     MachineInstr &SCCDefInst,
-                                    SetVectorType &Worklist,
+                                    SIInstrWorklist &Worklist,
                                     Register NewCond = Register()) const;
   void addSCCDefsToVALUWorklist(MachineInstr *SCCUseInst,
-                                SetVectorType &Worklist) const;
+                                SIInstrWorklist &Worklist) const;
 
   const TargetRegisterClass *
   getDestEquivalentVGPRClass(const MachineInstr &Inst) const;
 
   bool checkInstOffsetsDoNotOverlap(const MachineInstr &MIa,
                                     const MachineInstr &MIb) const;
 
   Register findUsedSGPR(const MachineInstr &MI, int OpIndices[3]) const;
public:
   /// \returns A new basic block that contains \p MI if new blocks were created.
   MachineBasicBlock *
   legalizeOperands(MachineInstr &MI, MachineDominatorTree *MDT = nullptr) const;
 
   /// Change SADDR form of a FLAT \p Inst to its VADDR form if saddr operand
   /// was moved to VGPR. \returns true if succeeded.
   bool moveFlatAddrToVGPR(MachineInstr &Inst) const;
 
-  /// Replace this instruction's opcode with the equivalent VALU
-  /// opcode.  This function will also move the users of \p MI to the
-  /// VALU if necessary. If present, \p MDT is updated.
-  MachineBasicBlock *moveToVALU(MachineInstr &MI,
-                                MachineDominatorTree *MDT = nullptr) const;
+  /// Replace the instructions opcode with the equivalent VALU
+  /// opcode.  This function will also move the users of MachineInstruntions
+  /// in the \p WorkList to the VALU if necessary. If present, \p MDT is
+  /// updated.
+  void moveToVALU(SIInstrWorklist &Worklist, MachineDominatorTree *MDT) const;
 
   void insertNoop(MachineBasicBlock &MBB,
                   MachineBasicBlock::iterator MI) const override;
 
   void insertNoops(MachineBasicBlock &MBB, MachineBasicBlock::iterator MI,
                    unsigned Quantity) const override;
 
   void insertReturn(MachineBasicBlock &MBB) const;

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

 //===----------------------------------------------------------------------===//
 
 #include "SIInstrInfo.h"
 #include "AMDGPU.h"
 #include "AMDGPUInstrInfo.h"
 #include "GCNHazardRecognizer.h"
 #include "GCNSubtarget.h"
 #include "SIMachineFunctionInfo.h"
+#include "llvm/ADT/PostOrderIterator.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/CodeGen/LiveIntervals.h"
 #include "llvm/CodeGen/LiveVariables.h"
 #include "llvm/CodeGen/MachineDominators.h"
 #include "llvm/CodeGen/MachineFrameInfo.h"
 #include "llvm/CodeGen/MachineScheduler.h"
 #include "llvm/CodeGen/RegisterScavenging.h"
 #include "llvm/CodeGen/ScheduleDAG.h"
BranchOffsetBits("amdgpu-s-branch-bits", cl::ReallyHidden, cl::init(16),
                  cl::desc("Restrict range of branch instructions (DEBUG)"));
 
 static cl::opt<bool> Fix16BitCopies(
   "amdgpu-fix-16-bit-physreg-copies",
   cl::desc("Fix copies between 32 and 16 bit registers by extending to 32 bit"),
   cl::init(true),
   cl::ReallyHidden);
 
+static std::vector<const MachineBasicBlock *>
+GetRPOList(const MachineFunction *MF) {
+  if (MF->empty())
+    return {};
+  ReversePostOrderTraversal<const MachineBasicBlock *> RPOT(&*MF->begin());
+  std::vector<const MachineBasicBlock *> RPOList;
+  append_range(RPOList, RPOT);
+  return RPOList;
+}
+
+// Compare machine instructions based on their order
+// in the machine function. Returns true if first instruction
+// occurs before second instruction in the machine function.
+bool SIInstrWorklist::Cmp::operator()(const MachineInstr *InstA,
+                                      const MachineInstr *InstB) const {
+  if (!InstA || !InstB)
+    return false;
+  if (InstA == InstB)
+    return false;
+  const MachineBasicBlock *BA = InstA->getParent();
+  const MachineBasicBlock *BB = InstB->getParent();
+  if (!BA || !BB)
+    return false;
+  if (BA != BB) {
+    const MachineFunction *MF = BA->getParent();
+    auto RPOList = GetRPOList(MF);
+    auto itA = std::find(RPOList.begin(), RPOList.end(), BA);
+    auto itB = std::find(RPOList.begin(), RPOList.end(), BB);
+    return std::distance(RPOList.begin(), itA) <
+           std::distance(RPOList.begin(), itB);
+  }
+  auto getDist = [](const MachineInstr *MI) {
+    MachineBasicBlock::const_iterator I = MI->getParent()->begin();
+    MachineBasicBlock::const_iterator E = MI->getIterator();
+    unsigned D = std::distance(I, E);
+    return D;
+  };
+  return getDist(InstA) < getDist(InstB);
+}
+
 SIInstrInfo::SIInstrInfo(const GCNSubtarget &ST)
   : AMDGPUGenInstrInfo(AMDGPU::ADJCALLSTACKUP, AMDGPU::ADJCALLSTACKDOWN),
     RI(ST), ST(ST) {
   SchedModel.init(&ST);
 }
 
 //===----------------------------------------------------------------------===//
 // TargetInstrInfo callbacks
if (VAddr && AMDGPU::getIfAddr64Inst(MI.getOpcode()) != -1) {
       // to SGPRs.
       CreatedBB = loadSRsrcFromVGPR(*this, MI, *Rsrc, MDT);
       return CreatedBB;
     }
   }
   return CreatedBB;
 }
 
-MachineBasicBlock *SIInstrInfo::moveToVALU(MachineInstr &TopInst,
-                                           MachineDominatorTree *MDT) const {
-  SetVectorType Worklist;
-  Worklist.insert(&TopInst);
-  MachineBasicBlock *CreatedBB = nullptr;
-  MachineBasicBlock *CreatedBBTmp = nullptr;
+void SIInstrInfo::moveToVALU(SIInstrWorklist &Worklist,
 
   while (!Worklist.empty()) {
-    MachineInstr &Inst = *Worklist.pop_back_val();
+    MachineInstr &Inst = *Worklist.top();
+    Worklist.erase_top();
     MachineBasicBlock *MBB = Inst.getParent();
+    if (!MBB)
+      continue;
     MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo();
-
     unsigned Opcode = Inst.getOpcode();
     unsigned NewOpcode = getVALUOp(Inst);
-
     // Handle some special cases
     switch (Opcode) {
     default:
       break;
     case AMDGPU::S_ADD_U64_PSEUDO:
     case AMDGPU::S_SUB_U64_PSEUDO:
       splitScalar64BitAddSub(Worklist, Inst, MDT);
       Inst.eraseFromParent();
       continue;
     case AMDGPU::S_ADD_I32:
     case AMDGPU::S_SUB_I32: {
       // FIXME: The u32 versions currently selected use the carry.
       bool Changed;
+      MachineBasicBlock *CreatedBBTmp = nullptr;
       std::tie(Changed, CreatedBBTmp) = moveScalarAddSub(Worklist, Inst, MDT);
-      if (CreatedBBTmp && TopInst.getParent() == CreatedBBTmp)
-        CreatedBB = CreatedBBTmp;
       if (Changed)
         continue;
 
       // Default handling
       break;
     }
     case AMDGPU::S_AND_B64:
       splitScalar64BitBinaryOp(Worklist, Inst, AMDGPU::S_AND_B32, MDT);
case AMDGPU::S_SUB_CO_PSEUDO: {
       unsigned Opc = (Inst.getOpcode() == AMDGPU::S_ADD_CO_PSEUDO)
                          ? AMDGPU::V_ADDC_U32_e64
                          : AMDGPU::V_SUBB_U32_e64;
       const auto *CarryRC = RI.getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
 
       Register CarryInReg = Inst.getOperand(4).getReg();
       if (!MRI.constrainRegClass(CarryInReg, CarryRC)) {
         Register NewCarryReg = MRI.createVirtualRegister(CarryRC);
-        BuildMI(*MBB, &Inst, Inst.getDebugLoc(), get(AMDGPU::COPY), NewCarryReg)
+        BuildMI(*MBB, Inst, Inst.getDebugLoc(), get(AMDGPU::COPY), NewCarryReg)
             .addReg(CarryInReg);
       }
 
       Register CarryOutReg = Inst.getOperand(1).getReg();
 
       Register DestReg = MRI.createVirtualRegister(RI.getEquivalentVGPRClass(
           MRI.getRegClass(Inst.getOperand(0).getReg())));
       MachineInstr *CarryOp =
           BuildMI(*MBB, &Inst, Inst.getDebugLoc(), get(Opc), DestReg)
               .addReg(CarryOutReg, RegState::Define)
               .add(Inst.getOperand(2))
               .add(Inst.getOperand(3))
               .addReg(CarryInReg)
               .addImm(0);
-      CreatedBBTmp = legalizeOperands(*CarryOp);
-      if (CreatedBBTmp && TopInst.getParent() == CreatedBBTmp)
-        CreatedBB = CreatedBBTmp;
+      legalizeOperands(*CarryOp);
       MRI.replaceRegWith(Inst.getOperand(0).getReg(), DestReg);
       addUsersToMoveToVALUWorklist(DestReg, MRI, Worklist);
       Inst.eraseFromParent();
     }
       continue;
     case AMDGPU::S_UADDO_PSEUDO:
     case AMDGPU::S_USUBO_PSEUDO: {
       const DebugLoc &DL = Inst.getDebugLoc();
case AMDGPU::S_USUBO_PSEUDO: {
           RI.getEquivalentVGPRClass(MRI.getRegClass(Dest0.getReg()));
       Register DestReg = MRI.createVirtualRegister(NewRC);
       MachineInstr *NewInstr = BuildMI(*MBB, &Inst, DL, get(Opc), DestReg)
                                    .addReg(Dest1.getReg(), RegState::Define)
                                    .add(Src0)
                                    .add(Src1)
                                    .addImm(0); // clamp bit
 
-      CreatedBBTmp = legalizeOperands(*NewInstr, MDT);
-      if (CreatedBBTmp && TopInst.getParent() == CreatedBBTmp)
-        CreatedBB = CreatedBBTmp;
-
+      legalizeOperands(*NewInstr, MDT);
       MRI.replaceRegWith(Dest0.getReg(), DestReg);
       addUsersToMoveToVALUWorklist(NewInstr->getOperand(0).getReg(), MRI,
                                    Worklist);
       Inst.eraseFromParent();
     }
       continue;
 
     case AMDGPU::S_CSELECT_B32:
case AMDGPU::S_CMP_LG_U64: {
         Inst.eraseFromParent();
       }
       continue;
     }
 
     if (NewOpcode == AMDGPU::INSTRUCTION_LIST_END) {
       // We cannot move this instruction to the VALU, so we should try to
       // legalize its operands instead.
-      CreatedBBTmp = legalizeOperands(Inst, MDT);
-      if (CreatedBBTmp && TopInst.getParent() == CreatedBBTmp)
-        CreatedBB = CreatedBBTmp;
+      legalizeOperands(Inst, MDT);
       continue;
     }
-
     // Handle converting generic instructions like COPY-to-SGPR into
     // COPY-to-VGPR.
     if (NewOpcode == Opcode) {
       Register DstReg = Inst.getOperand(0).getReg();
       const TargetRegisterClass *NewDstRC = getDestEquivalentVGPRClass(Inst);
+      if (!NewDstRC)
+        continue;
 
       if (Inst.isCopy() && Inst.getOperand(1).getReg().isVirtual() &&
           NewDstRC == RI.getRegClassForReg(MRI, Inst.getOperand(1).getReg())) {
         // Instead of creating a copy where src and dst are the same register
         // class, we just replace all uses of dst with src.  These kinds of
         // copies interfere with the heuristics MachineSink uses to decide
         // whether or not to split a critical edge.  Since the pass assumes
         // that copies will end up as machine instructions and not be
         // eliminated.
         addUsersToMoveToVALUWorklist(DstReg, MRI, Worklist);
         MRI.replaceRegWith(DstReg, Inst.getOperand(1).getReg());
         MRI.clearKillFlags(Inst.getOperand(1).getReg());
         Inst.getOperand(0).setReg(DstReg);
 
         // Make sure we don't leave around a dead VGPR->SGPR copy. Normally
         // these are deleted later, but at -O0 it would leave a suspicious
         // looking illegal copy of an undef register.
         for (unsigned I = Inst.getNumOperands() - 1; I != 0; --I)
           Inst.removeOperand(I);
         Inst.setDesc(get(AMDGPU::IMPLICIT_DEF));
         continue;
       }
-
       Register NewDstReg = MRI.createVirtualRegister(NewDstRC);
       MRI.replaceRegWith(DstReg, NewDstReg);
       legalizeOperands(Inst, MDT);
       addUsersToMoveToVALUWorklist(NewDstReg, MRI, Worklist);
       continue;
     }
 
     // Use the new VALU Opcode.
for (MachineOperand &Op : Inst.implicit_operands()) {
       if (Op.getReg() == AMDGPU::SCC) {
         // Only propagate through live-def of SCC.
         if (Op.isDef() && !Op.isDead())
           addSCCDefUsersToVALUWorklist(Op, Inst, Worklist);
         if (Op.isUse())
           addSCCDefsToVALUWorklist(NewInstr, Worklist);
       }
     }
-
     Inst.eraseFromParent();
-
     Register NewDstReg;
     if (NewInstr->getOperand(0).isReg() && NewInstr->getOperand(0).isDef()) {
       Register DstReg = NewInstr->getOperand(0).getReg();
       assert(DstReg.isVirtual());
-
       // Update the destination register class.
       const TargetRegisterClass *NewDstRC =
           getDestEquivalentVGPRClass(*NewInstr);
       assert(NewDstRC);
-
       NewDstReg = MRI.createVirtualRegister(NewDstRC);
       MRI.replaceRegWith(DstReg, NewDstReg);
     }
-
     if (Opcode == AMDGPU::S_SEXT_I32_I8 || Opcode == AMDGPU::S_SEXT_I32_I16) {
       // We are converting these to a BFE, so we need to add the missing
       // operands for the size and offset.
       unsigned Size = (Opcode == AMDGPU::S_SEXT_I32_I8) ? 8 : 16;
       NewInstr.addImm(0);
       NewInstr.addImm(Size);
     } else if (Opcode == AMDGPU::S_BCNT1_I32_B32) {
       // The VALU version adds the second operand to the result, so insert an
       // extra 0 operand.
       NewInstr.addImm(0);
     }
-
     if (Opcode == AMDGPU::S_BFE_I32 || Opcode == AMDGPU::S_BFE_U32) {
       const MachineOperand &OffsetWidthOp = NewInstr->getOperand(2);
       // If we need to move this to VGPRs, we need to unpack the second operand
       // back into the 2 separate ones for bit offset and width.
       assert(OffsetWidthOp.isImm() &&
              "Scalar BFE is only implemented for constant width and offset");
       uint32_t Imm = OffsetWidthOp.getImm();
-
       uint32_t Offset = Imm & 0x3f; // Extract bits [5:0].
       uint32_t BitWidth = (Imm & 0x7f0000) >> 16; // Extract bits [22:16].
       NewInstr->removeOperand(2);
       NewInstr.addImm(Offset);
       NewInstr.addImm(BitWidth);
     }
-
     fixImplicitOperands(*NewInstr);
-
     // Legalize the operands
-    CreatedBBTmp = legalizeOperands(*NewInstr, MDT);
-    if (CreatedBBTmp && TopInst.getParent() == CreatedBBTmp)
-      CreatedBB = CreatedBBTmp;
-
+    legalizeOperands(*NewInstr, MDT);
     if (NewDstReg)
       addUsersToMoveToVALUWorklist(NewDstReg, MRI, Worklist);
   }
-  return CreatedBB;
+  return;
 }
 
 // Add/sub require special handling to deal with carry outs.
 std::pair<bool, MachineBasicBlock *>
-SIInstrInfo::moveScalarAddSub(SetVectorType &Worklist, MachineInstr &Inst,
+SIInstrInfo::moveScalarAddSub(SIInstrWorklist &Worklist, MachineInstr &Inst,
                               MachineDominatorTree *MDT) const {
   if (ST.hasAddNoCarry()) {
     // Assume there is no user of scc since we don't select this in that case.
     // Since scc isn't used, it doesn't really matter if the i32 or u32 variant
     // is used.
 
     MachineBasicBlock &MBB = *Inst.getParent();
     MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
if (ST.hasAddNoCarry()) {
 
     addUsersToMoveToVALUWorklist(ResultReg, MRI, Worklist);
     return std::pair(true, NewBB);
   }
 
   return std::pair(false, nullptr);
 }
 
-void SIInstrInfo::lowerSelect(SetVectorType &Worklist, MachineInstr &Inst,
+void SIInstrInfo::lowerSelect(SIInstrWorklist &Worklist, MachineInstr &Inst,
                               MachineDominatorTree *MDT) const {
 
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   MachineBasicBlock::iterator MII = Inst;
   DebugLoc DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
auto UpdatedInst =
           .add(Src0) // True
           .addReg(IsSCC ? CopySCC : SCCSource);
 
   MRI.replaceRegWith(Dest.getReg(), ResultReg);
   legalizeOperands(*UpdatedInst, MDT);
   addUsersToMoveToVALUWorklist(ResultReg, MRI, Worklist);
 }
 
-void SIInstrInfo::lowerScalarAbs(SetVectorType &Worklist,
+void SIInstrInfo::lowerScalarAbs(SIInstrWorklist &Worklist,
                                  MachineInstr &Inst) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   MachineBasicBlock::iterator MII = Inst;
   DebugLoc DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src = Inst.getOperand(1);
void SIInstrInfo::lowerScalarAbs(SIInstrWorklist &Worklist,
   BuildMI(MBB, MII, DL, get(AMDGPU::V_MAX_I32_e64), ResultReg)
     .addReg(Src.getReg())
     .addReg(TmpReg);
 
   MRI.replaceRegWith(Dest.getReg(), ResultReg);
   addUsersToMoveToVALUWorklist(ResultReg, MRI, Worklist);
 }
 
-void SIInstrInfo::lowerScalarXnor(SetVectorType &Worklist,
+void SIInstrInfo::lowerScalarXnor(SIInstrWorklist &Worklist,
                                   MachineInstr &Inst) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   MachineBasicBlock::iterator MII = Inst;
   const DebugLoc &DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src0 = Inst.getOperand(1);
if (ST.hasDLInsts()) {
     MRI.replaceRegWith(Dest.getReg(), NewDest);
 
     Worklist.insert(Xor);
 
     addUsersToMoveToVALUWorklist(NewDest, MRI, Worklist);
   }
 }
 
-void SIInstrInfo::splitScalarNotBinop(SetVectorType &Worklist,
+void SIInstrInfo::splitScalarNotBinop(SIInstrWorklist &Worklist,
                                       MachineInstr &Inst,
                                       unsigned Opcode) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   MachineBasicBlock::iterator MII = Inst;
   const DebugLoc &DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
void SIInstrInfo::splitScalarNotBinop(SIInstrWorklist &Worklist,
 
   Worklist.insert(&Op);
   Worklist.insert(&Not);
 
   MRI.replaceRegWith(Dest.getReg(), NewDest);
   addUsersToMoveToVALUWorklist(NewDest, MRI, Worklist);
 }
 
-void SIInstrInfo::splitScalarBinOpN2(SetVectorType& Worklist,
+void SIInstrInfo::splitScalarBinOpN2(SIInstrWorklist &Worklist,
                                      MachineInstr &Inst,
                                      unsigned Opcode) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   MachineBasicBlock::iterator MII = Inst;
   const DebugLoc &DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
void SIInstrInfo::splitScalarBinOpN2(SIInstrWorklist &Worklist,
 
   Worklist.insert(&Not);
   Worklist.insert(&Op);
 
   MRI.replaceRegWith(Dest.getReg(), NewDest);
   addUsersToMoveToVALUWorklist(NewDest, MRI, Worklist);
 }
 
-void SIInstrInfo::splitScalar64BitUnaryOp(
-    SetVectorType &Worklist, MachineInstr &Inst,
-    unsigned Opcode, bool Swap) const {
+void SIInstrInfo::splitScalar64BitUnaryOp(SIInstrWorklist &Worklist,
+                                          MachineInstr &Inst, unsigned Opcode,
+                                          bool Swap) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src0 = Inst.getOperand(1);
   DebugLoc DL = Inst.getDebugLoc();
 
   MachineBasicBlock::iterator MII = Inst;
void SIInstrInfo::splitScalar64BitUnaryOp(SIInstrWorklist &Worklist,
 
   // We don't need to legalizeOperands here because for a single operand, src0
   // will support any kind of input.
 
   // Move all users of this moved value.
   addUsersToMoveToVALUWorklist(FullDestReg, MRI, Worklist);
 }
 
-void SIInstrInfo::splitScalar64BitAddSub(SetVectorType &Worklist,
+void SIInstrInfo::splitScalar64BitAddSub(SIInstrWorklist &Worklist,
                                          MachineInstr &Inst,
                                          MachineDominatorTree *MDT) const {
   bool IsAdd = (Inst.getOpcode() == AMDGPU::S_ADD_U64_PSEUDO);
 
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   const auto *CarryRC = RI.getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
 
void SIInstrInfo::splitScalar64BitAddSub(SIInstrWorklist &Worklist,
   // valid.
   legalizeOperands(*LoHalf, MDT);
   legalizeOperands(*HiHalf, MDT);
 
   // Move all users of this moved value.
   addUsersToMoveToVALUWorklist(FullDestReg, MRI, Worklist);
 }
 
-void SIInstrInfo::splitScalar64BitBinaryOp(SetVectorType &Worklist,
+void SIInstrInfo::splitScalar64BitBinaryOp(SIInstrWorklist &Worklist,
                                            MachineInstr &Inst, unsigned Opcode,
                                            MachineDominatorTree *MDT) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src0 = Inst.getOperand(1);
   MachineOperand &Src1 = Inst.getOperand(2);
void SIInstrInfo::splitScalar64BitBinaryOp(SIInstrWorklist &Worklist,
 
   Worklist.insert(&LoHalf);
   Worklist.insert(&HiHalf);
 
   // Move all users of this moved value.
   addUsersToMoveToVALUWorklist(FullDestReg, MRI, Worklist);
 }
 
-void SIInstrInfo::splitScalar64BitXnor(SetVectorType &Worklist,
+void SIInstrInfo::splitScalar64BitXnor(SIInstrWorklist &Worklist,
                                        MachineInstr &Inst,
                                        MachineDominatorTree *MDT) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src0 = Inst.getOperand(1);
   MachineOperand &Src1 = Inst.getOperand(2);
MachineInstr &Xor = *BuildMI(MBB, MII, DL, get(AMDGPU::S_XOR_B64), NewDest)
     .addReg(Interm)
     .add(*Op1);
 
   MRI.replaceRegWith(Dest.getReg(), NewDest);
 
   Worklist.insert(&Xor);
 }
 
-void SIInstrInfo::splitScalar64BitBCNT(
-    SetVectorType &Worklist, MachineInstr &Inst) const {
+void SIInstrInfo::splitScalar64BitBCNT(SIInstrWorklist &Worklist,
+                                       MachineInstr &Inst) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
 
   MachineBasicBlock::iterator MII = Inst;
   const DebugLoc &DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src = Inst.getOperand(1);
void SIInstrInfo::splitScalar64BitBCNT(SIInstrWorklist &Worklist,
 
   MRI.replaceRegWith(Dest.getReg(), ResultReg);
 
   // We don't need to legalize operands here. src0 for either instruction can be
   // an SGPR, and the second input is unused or determined here.
   addUsersToMoveToVALUWorklist(ResultReg, MRI, Worklist);
 }
 
-void SIInstrInfo::splitScalar64BitBFE(SetVectorType &Worklist,
+void SIInstrInfo::splitScalar64BitBFE(SIInstrWorklist &Worklist,
                                       MachineInstr &Inst) const {
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
   MachineBasicBlock::iterator MII = Inst;
   const DebugLoc &DL = Inst.getDebugLoc();
 
   MachineOperand &Dest = Inst.getOperand(0);
   uint32_t Imm = Inst.getOperand(2).getImm();
BuildMI(MBB, MII, DL, get(TargetOpcode::REG_SEQUENCE), ResultReg)
     .addReg(TmpReg)
     .addImm(AMDGPU::sub1);
 
   MRI.replaceRegWith(Dest.getReg(), ResultReg);
   addUsersToMoveToVALUWorklist(ResultReg, MRI, Worklist);
 }
 
 void SIInstrInfo::addUsersToMoveToVALUWorklist(
-  Register DstReg,
-  MachineRegisterInfo &MRI,
-  SetVectorType &Worklist) const {
+    Register DstReg, MachineRegisterInfo &MRI,
+    SIInstrWorklist &Worklist) const {
   for (MachineRegisterInfo::use_iterator I = MRI.use_begin(DstReg),
          E = MRI.use_end(); I != E;) {
     MachineInstr &UseMI = *I->getParent();
 
     unsigned OpNo = 0;
 
     switch (UseMI.getOpcode()) {
     case AMDGPU::COPY:
if (!RI.hasVectorRegisters(getOpRegClass(UseMI, OpNo))) {
         ++I;
       } while (I != E && I->getParent() == &UseMI);
     } else {
       ++I;
     }
   }
 }
 
-void SIInstrInfo::movePackToVALU(SetVectorType &Worklist,
+void SIInstrInfo::movePackToVALU(SIInstrWorklist &Worklist,
                                  MachineRegisterInfo &MRI,
                                  MachineInstr &Inst) const {
   Register ResultReg = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
   MachineBasicBlock *MBB = Inst.getParent();
   MachineOperand &Src0 = Inst.getOperand(1);
   MachineOperand &Src1 = Inst.getOperand(2);
   const DebugLoc &DL = Inst.getDebugLoc();
 
void SIInstrInfo::movePackToVALU(SIInstrWorklist &Worklist,
 
   MachineOperand &Dest = Inst.getOperand(0);
   MRI.replaceRegWith(Dest.getReg(), ResultReg);
   addUsersToMoveToVALUWorklist(ResultReg, MRI, Worklist);
 }
 
 void SIInstrInfo::addSCCDefUsersToVALUWorklist(MachineOperand &Op,
                                                MachineInstr &SCCDefInst,
-                                               SetVectorType &Worklist,
+                                               SIInstrWorklist &Worklist,
                                                Register NewCond) const {
 
   // Ensure that def inst defines SCC, which is still live.
   assert(Op.isReg() && Op.getReg() == AMDGPU::SCC && Op.isDef() &&
          !Op.isDead() && Op.getParent() == &SCCDefInst);
   SmallVector<MachineInstr *, 4> CopyToDelete;
   // This assumes that all the users of SCC are in the same block
   // as the SCC def.

 }
 
 // Instructions that use SCC may be converted to VALU instructions. When that
 // happens, the SCC register is changed to VCC_LO. The instruction that defines
 // SCC must be changed to an instruction that defines VCC. This function makes
 // sure that the instruction that defines SCC is added to the moveToVALU
 // worklist.
 void SIInstrInfo::addSCCDefsToVALUWorklist(MachineInstr *SCCUseInst,
-                                           SetVectorType &Worklist) const {
+                                           SIInstrWorklist &Worklist) const {
   // Look for a preceding instruction that either defines VCC or SCC. If VCC
   // then there is nothing to do because the defining instruction has been
   // converted to a VALU already. If SCC then that instruction needs to be
   // converted to a VALU.
   for (MachineInstr &MI :
        make_range(std::next(MachineBasicBlock::reverse_iterator(SCCUseInst)),
                   SCCUseInst->getParent()->rend())) {
     if (MI.modifiesRegister(AMDGPU::VCC, &RI))

llvm/test/CodeGen/AMDGPU/add3.ll

 ; but we don't.
 define amdgpu_ps float @add3_uniform_vgpr(float inreg %a, float inreg %b, float inreg %c) {
 ; VI-LABEL: add3_uniform_vgpr:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    v_add_f32_e64 v0, s2, 1.0
 ; VI-NEXT:    v_add_f32_e64 v1, s3, 2.0
 ; VI-NEXT:    v_mov_b32_e32 v2, 0x40400000
 ; VI-NEXT:    v_add_f32_e32 v2, s4, v2
-; VI-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; VI-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; VI-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; VI-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: add3_uniform_vgpr:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_add_f32_e64 v0, s2, 1.0
 ; GFX9-NEXT:    v_add_f32_e64 v1, s3, 2.0
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0x40400000

llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll

-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --force-update
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: opt -S -mtriple=amdgcn-- -mcpu=tahiti -amdgpu-codegenprepare -amdgpu-bypass-slow-div=0 %s | FileCheck %s
 ; RUN: llc -mtriple=amdgcn-- -mcpu=tahiti -amdgpu-bypass-slow-div=0 < %s | FileCheck -check-prefix=GFX6 %s
 ; RUN: llc -mtriple=amdgcn-- -mcpu=gfx900 -amdgpu-bypass-slow-div=0 < %s | FileCheck -check-prefix=GFX9 %s
 
 define amdgpu_kernel void @udiv_i32(ptr addrspace(1) %out, i32 %x, i32 %y) {
 ; CHECK-LABEL: @udiv_i32(
 ; CHECK-NEXT:    [[TMP1:%.*]] = uitofp i32 [[Y:%.*]] to float
 ; CHECK-NEXT:    [[TMP2:%.*]] = call fast float @llvm.amdgcn.rcp.f32(float [[TMP1]])

 ; GFX6-NEXT:    s_sub_i32 s4, 0, s3
 ; GFX6-NEXT:    s_mov_b32 s5, s1
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; GFX6-NEXT:    s_mov_b32 s4, s0
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX6-NEXT:    s_mul_i32 s0, s0, s3
 ; GFX6-NEXT:    s_sub_i32 s0, s2, s0
 ; GFX6-NEXT:    s_sub_i32 s1, s0, s3
 ; GFX6-NEXT:    s_cmp_ge_u32 s0, s3
 ; GFX6-NEXT:    s_cselect_b32 s0, s1, s0
 ; GFX6-NEXT:    s_sub_i32 s1, s0, s3

 ; GFX6-NEXT:    s_xor_b32 s2, s2, s9
 ; GFX6-NEXT:    s_mov_b32 s5, s1
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; GFX6-NEXT:    s_mov_b32 s4, s0
 ; GFX6-NEXT:    s_xor_b32 s0, s9, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s1, v0
 ; GFX6-NEXT:    s_mul_i32 s1, s1, s3
 ; GFX6-NEXT:    s_sub_i32 s1, s2, s1
 ; GFX6-NEXT:    s_sub_i32 s2, s1, s3
 ; GFX6-NEXT:    v_add_i32_e32 v1, vcc, 1, v0
 ; GFX6-NEXT:    s_cmp_ge_u32 s1, s3
 ; GFX6-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s12
 ; GFX6-NEXT:    s_cselect_b64 s[2:3], -1, 0
 ; GFX6-NEXT:    s_sub_i32 s4, 0, s13
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s4, v1
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v1, v3
 ; GFX6-NEXT:    v_cndmask_b32_e64 v0, v0, v2, s[2:3]
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v1, s9, v1
 ; GFX6-NEXT:    v_mul_f32_e32 v3, 0x4f7ffffe, v4
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v3, v3
 ; GFX6-NEXT:    v_readfirstlane_b32 s4, v1
 ; GFX6-NEXT:    s_mul_i32 s4, s4, s13
 ; GFX6-NEXT:    s_sub_i32 s4, s9, s4
 ; GFX6-NEXT:    s_sub_i32 s5, s4, s13
 ; GFX6-NEXT:    s_cmp_ge_u32 s4, s13

 ; GFX6-NEXT:    s_cmp_ge_u32 s6, s14
 ; GFX6-NEXT:    s_cselect_b64 s[6:7], -1, 0
 ; GFX6-NEXT:    s_sub_i32 s8, 0, s15
 ; GFX6-NEXT:    v_mul_lo_u32 v7, s8, v5
 ; GFX6-NEXT:    v_cndmask_b32_e32 v3, v3, v6, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, 1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v5, v7
 ; GFX6-NEXT:    v_cndmask_b32_e64 v2, v3, v6, s[6:7]
-; GFX6-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
+; GFX6-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; GFX6-NEXT:    v_mul_hi_u32 v5, s11, v5
 ; GFX6-NEXT:    v_readfirstlane_b32 s0, v5
 ; GFX6-NEXT:    s_mul_i32 s0, s0, s15
 ; GFX6-NEXT:    s_sub_i32 s0, s11, s0
 ; GFX6-NEXT:    s_sub_i32 s1, s0, s15
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, 1, v5
 ; GFX6-NEXT:    s_cmp_ge_u32 s0, s15
 ; GFX6-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GFX6-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s8
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s8
 ; GFX6-NEXT:    s_cselect_b32 s4, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s2, 0, s9
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s2, v1
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_cvt_f32_u32_e32 v2, s11
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s9
 ; GFX6-NEXT:    s_sub_i32 s2, s5, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s9
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s9
 ; GFX6-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s9
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s9
 ; GFX6-NEXT:    s_cselect_b32 s5, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s2, 0, s10
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s2, v1
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s6, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s10
 ; GFX6-NEXT:    s_sub_i32 s2, s6, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s10
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s10

 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; GFX6-NEXT:    s_ashr_i32 s4, s8, 31
 ; GFX6-NEXT:    s_add_i32 s5, s8, s4
 ; GFX6-NEXT:    s_xor_b32 s5, s5, s4
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
 ; GFX6-NEXT:    s_xor_b32 s8, s4, s2
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s3
 ; GFX6-NEXT:    s_sub_i32 s2, s5, s2
 ; GFX6-NEXT:    s_sub_i32 s4, s2, s3
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s3
 ; GFX6-NEXT:    v_add_i32_e32 v1, vcc, 1, v0
 ; GFX6-NEXT:    s_cselect_b32 s2, s4, s2

 ; GFX6-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s6, v2
 ; GFX6-NEXT:    s_ashr_i32 s6, s9, 31
 ; GFX6-NEXT:    s_add_i32 s7, s9, s6
 ; GFX6-NEXT:    s_xor_b32 s7, s7, s6
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v2, v3
 ; GFX6-NEXT:    s_xor_b32 s9, s6, s4
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v2, s7, v2
 ; GFX6-NEXT:    v_readfirstlane_b32 s4, v2
 ; GFX6-NEXT:    s_mul_i32 s4, s4, s5
 ; GFX6-NEXT:    s_sub_i32 s4, s7, s4
 ; GFX6-NEXT:    s_sub_i32 s6, s4, s5
 ; GFX6-NEXT:    s_cmp_ge_u32 s4, s5
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, 1, v2
 ; GFX6-NEXT:    s_cselect_b32 s4, s6, s4

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v3, v1
 ; GFX6-NEXT:    v_subrev_i32_e32 v1, vcc, s9, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s0, v3
 ; GFX6-NEXT:    s_ashr_i32 s0, s11, 31
 ; GFX6-NEXT:    s_add_i32 s1, s11, s0
 ; GFX6-NEXT:    s_xor_b32 s1, s1, s0
 ; GFX6-NEXT:    v_mul_hi_u32 v2, v3, v2
 ; GFX6-NEXT:    s_xor_b32 s0, s0, s10
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s1, v2
 ; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, s12, v4
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v3
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s13
 ; GFX6-NEXT:    s_sub_i32 s1, s1, s2
 ; GFX6-NEXT:    s_sub_i32 s2, s1, s13
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, 1, v3
 ; GFX6-NEXT:    s_cmp_ge_u32 s1, s13

 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s8, v0
 ; GFX6-NEXT:    s_ashr_i32 s8, s5, 31
 ; GFX6-NEXT:    s_add_i32 s5, s5, s8
 ; GFX6-NEXT:    s_xor_b32 s5, s5, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s4
 ; GFX6-NEXT:    s_sub_i32 s2, s5, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s4
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s4
 ; GFX6-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s4

 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; GFX6-NEXT:    s_ashr_i32 s4, s6, 31
 ; GFX6-NEXT:    s_add_i32 s5, s6, s4
 ; GFX6-NEXT:    s_xor_b32 s5, s5, s4
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
 ; GFX6-NEXT:    s_sub_i32 s6, s2, s8
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s3
 ; GFX6-NEXT:    s_sub_i32 s2, s5, s2
 ; GFX6-NEXT:    s_sub_i32 s5, s2, s3
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s3
 ; GFX6-NEXT:    s_cselect_b32 s2, s5, s2
 ; GFX6-NEXT:    s_sub_i32 s5, s2, s3

 ; GFX6-NEXT:    v_or_b32_e32 v0, 1, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v1, v5, v6
 ; GFX6-NEXT:    v_trunc_f32_e32 v1, v1
 ; GFX6-NEXT:    v_mad_f32 v5, -v1, v2, v5
 ; GFX6-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_cmp_ge_f32_e64 vcc, |v5|, |v2|
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0x7fff, v3
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_and_b32_e32 v3, 0x7fff, v4
 ; GFX6-NEXT:    v_lshl_b64 v[0:1], v[0:1], 30
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 15, v3
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_waitcnt expcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0x1fff, v1

 ; GFX6-NEXT:    s_cmp_ge_u32 s3, s2
 ; GFX6-NEXT:    s_cselect_b64 s[2:3], -1, 0
 ; GFX6-NEXT:    s_sub_i32 s4, 0, s6
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s4, v1
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v1, v3
 ; GFX6-NEXT:    v_cndmask_b32_e64 v0, v0, v2, s[2:3]
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX6-NEXT:    v_readfirstlane_b32 s0, v1
 ; GFX6-NEXT:    s_mul_i32 s0, s0, s6
 ; GFX6-NEXT:    s_sub_i32 s0, s5, s0
 ; GFX6-NEXT:    s_sub_i32 s1, s0, s6
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v1
 ; GFX6-NEXT:    s_cmp_ge_u32 s0, s6
 ; GFX6-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GFX6-NEXT:    s_sub_i32 s4, s3, s2
 ; GFX6-NEXT:    s_cmp_ge_u32 s3, s2
 ; GFX6-NEXT:    s_cselect_b32 s4, s4, s3
 ; GFX6-NEXT:    s_sub_i32 s2, 0, s6
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s2, v1
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s7, v0
 ; GFX6-NEXT:    s_mul_i32 s7, s7, s6
 ; GFX6-NEXT:    s_sub_i32 s5, s5, s7
 ; GFX6-NEXT:    s_sub_i32 s7, s5, s6
 ; GFX6-NEXT:    s_cmp_ge_u32 s5, s6
 ; GFX6-NEXT:    s_cselect_b32 s5, s7, s5
 ; GFX6-NEXT:    s_sub_i32 s7, s5, s6

 ; GFX6-NEXT:    v_mov_b32_e32 v0, 0xd9528441
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    v_mul_hi_i32 v0, s4, v0
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 31, v0
 ; GFX6-NEXT:    v_ashrrev_i32_e32 v0, 20, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_i32_oddk_denom:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x2c
 ; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0

 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX6-NEXT:    s_xor_b32 s2, s2, s9
 ; GFX6-NEXT:    s_mov_b32 s5, s1
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; GFX6-NEXT:    s_mov_b32 s4, s0
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX6-NEXT:    s_mul_i32 s0, s0, s3
 ; GFX6-NEXT:    s_sub_i32 s0, s2, s0
 ; GFX6-NEXT:    s_sub_i32 s1, s0, s3
 ; GFX6-NEXT:    v_add_i32_e32 v1, vcc, 1, v0
 ; GFX6-NEXT:    s_cmp_ge_u32 s0, s3
 ; GFX6-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GFX6-NEXT:    s_ashr_i32 s0, s2, 31
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, 20
 ; GFX6-NEXT:    s_add_i32 s0, s2, s0
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, s3, v0
 ; GFX6-NEXT:    s_ashr_i32 s0, s0, 12
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 31, v0
 ; GFX6-NEXT:    v_ashrrev_i32_e32 v0, 11, v0
 ; GFX6-NEXT:    s_mov_b32 s5, s1
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v0, v1
 ; GFX6-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX6-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GFX6-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: ssdiv_v2i32_mixed_pow2k_denom:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0

 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s6, v0
 ; GFX6-NEXT:    s_ashr_i32 s6, s4, 31
 ; GFX6-NEXT:    s_add_i32 s4, s4, s6
 ; GFX6-NEXT:    s_xor_b32 s4, s4, s6
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
 ; GFX6-NEXT:    s_xor_b32 s6, s6, s3
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX6-NEXT:    s_mul_i32 s3, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s4, s3
 ; GFX6-NEXT:    s_sub_i32 s4, s3, s2
 ; GFX6-NEXT:    v_add_i32_e32 v1, vcc, 1, v0
 ; GFX6-NEXT:    s_cmp_ge_u32 s3, s2
 ; GFX6-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s8, v2
 ; GFX6-NEXT:    s_ashr_i32 s8, s5, 31
 ; GFX6-NEXT:    s_add_i32 s5, s5, s8
 ; GFX6-NEXT:    s_xor_b32 s5, s5, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v2, v3
 ; GFX6-NEXT:    s_xor_b32 s4, s8, s4
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v2
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX6-NEXT:    v_readfirstlane_b32 s6, v1
 ; GFX6-NEXT:    s_mul_i32 s6, s6, s7
 ; GFX6-NEXT:    s_sub_i32 s5, s5, s6
 ; GFX6-NEXT:    s_sub_i32 s6, s5, s7
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v1
 ; GFX6-NEXT:    s_cmp_ge_u32 s5, s7
 ; GFX6-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX6-NEXT:    s_xor_b32 s6, s2, s5
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s3, v0
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s6, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s7, v0
 ; GFX6-NEXT:    s_mul_i32 s7, s7, s4
 ; GFX6-NEXT:    s_sub_i32 s6, s6, s7
 ; GFX6-NEXT:    s_sub_i32 s7, s6, s4
 ; GFX6-NEXT:    s_cmp_ge_u32 s6, s4
 ; GFX6-NEXT:    s_cselect_b32 s6, s7, s6
 ; GFX6-NEXT:    s_sub_i32 s7, s6, s4

 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s2, v0
 ; GFX6-NEXT:    s_add_i32 s2, s5, s8
 ; GFX6-NEXT:    s_xor_b32 s5, s2, s8
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s7, v0
 ; GFX6-NEXT:    s_mul_i32 s7, s7, s4
 ; GFX6-NEXT:    s_sub_i32 s5, s5, s7
 ; GFX6-NEXT:    s_sub_i32 s7, s5, s4
 ; GFX6-NEXT:    s_cmp_ge_u32 s5, s4
 ; GFX6-NEXT:    s_cselect_b32 s5, s7, s5
 ; GFX6-NEXT:    s_sub_i32 s7, s5, s4

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    s_movk_i32 s8, 0x11f
 ; GFX6-NEXT:    s_mov_b32 s9, 0x976a7377
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v0, s4
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s5
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v1, s5
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, s5
 ; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v0, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v5
 ; GFX6-NEXT:    s_mov_b32 s6, -1
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v4, v5, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v0, s4
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s5
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v1, s5
 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_mov_b32 s4, s0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, s5
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v3

 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, 0, v2, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v1, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v0, s8
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, 1, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v8, v0, s8
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v1, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, 2, v0
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v1, vcc
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX6-NEXT:    v_sub_i32_e32 v8, vcc, s2, v8
 ; GFX6-NEXT:    v_subb_u32_e32 v4, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_subrev_i32_e32 v5, vcc, s8, v8
 ; GFX6-NEXT:    v_subbrev_u32_e32 v9, vcc, 0, v4, vcc
 ; GFX6-NEXT:    s_movk_i32 s2, 0xffe
 ; GFX6-NEXT:    v_cmp_lt_u32_e32 vcc, s2, v5
 ; GFX6-NEXT:    v_cndmask_b32_e64 v5, 0, -1, vcc

 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_mov_b32 s8, s4
 ; GFX6-NEXT:    s_movk_i32 s4, 0x11f
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v0, s2
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s3
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v1, s3
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, s3
 ; GFX6-NEXT:    s_mov_b32 s12, 0x9761f7c9
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v0, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v5
 ; GFX6-NEXT:    s_mov_b32 s9, s5
 ; GFX6-NEXT:    s_movk_i32 s5, 0x11e
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v4, v5, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v0, s2
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s3
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v1, s3
 ; GFX6-NEXT:    s_mov_b32 s11, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s10, -1
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, s3
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v3

 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_mov_b32 s0, s4
 ; GFX6-NEXT:    s_mov_b32 s4, 0x12d8fb
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s8
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v0, s8
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    s_mov_b32 s1, s5
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s8
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, s8
-; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_add_u32 s2, s2, s12
 ; GFX6-NEXT:    s_mov_b32 s13, s12
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s5, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s4, v0
 ; GFX6-NEXT:    s_addc_u32 s3, s3, s12
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GFX6-NEXT:    s_xor_b64 s[2:3], s[2:3], s[12:13]
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s5, v0
 ; GFX6-NEXT:    s_mov_b32 s5, s1
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s4, v0
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GFX6-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX6-NEXT:    s_ashr_i32 s8, s1, 31
 ; GFX6-NEXT:    s_lshr_b32 s8, s8, 20
 ; GFX6-NEXT:    v_mul_hi_u32 v2, v0, s6
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, s6
 ; GFX6-NEXT:    s_add_u32 s0, s0, s8
 ; GFX6-NEXT:    s_addc_u32 s1, s1, 0
 ; GFX6-NEXT:    s_ashr_i64 s[0:1], s[0:1], 12
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, s6
-; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v3

 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, s6
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s6
 ; GFX6-NEXT:    s_mov_b32 s9, s8
 ; GFX6-NEXT:    s_addc_u32 s3, s3, s8
 ; GFX6-NEXT:    s_xor_b64 s[2:3], s[2:3], s[8:9]
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, s6
-; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_addc_u32 s1, s5, s16
 ; GFX6-NEXT:    s_xor_b64 s[4:5], s[0:1], s[16:17]
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s10, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s10, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s11, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s10, v0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s10, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s10, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s11, v0
 ; GFX6-NEXT:    s_mov_b32 s11, 0xf000
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s10, v0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GFX6-NEXT:    v_addc_u32_e32 v0, vcc, v3, v0, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v2, vcc, 0, v5, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, 0, v2, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s12, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s12, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s13, v0
 ; GFX6-NEXT:    v_mov_b32_e32 v5, s13
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s12, v0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_sub_i32_e32 v4, vcc, s5, v2
 ; GFX6-NEXT:    v_sub_i32_e32 v3, vcc, s4, v3
 ; GFX6-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v5, vcc
 ; GFX6-NEXT:    v_subrev_i32_e64 v5, s[0:1], s12, v3
 ; GFX6-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
 ; GFX6-NEXT:    v_cmp_le_u32_e64 s[0:1], s13, v4
 ; GFX6-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[0:1]
 ; GFX6-NEXT:    v_cmp_le_u32_e64 s[0:1], s12, v5

 ; GFX6-NEXT:    s_sub_u32 s12, 0, s2
 ; GFX6-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v4, vcc
 ; GFX6-NEXT:    v_mul_hi_u32 v4, s12, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s12, v3
 ; GFX6-NEXT:    s_subb_u32 s13, 0, s3
 ; GFX6-NEXT:    v_mul_lo_u32 v6, s13, v2
 ; GFX6-NEXT:    v_xor_b32_e32 v0, s0, v0
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s12, v2
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v2, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v2, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v9, v3, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v3, v4
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v8, v3, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v3, v5
 ; GFX6-NEXT:    v_xor_b32_e32 v1, s1, v1
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, v7, v5, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v6, vcc, 0, v9, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v3, v5, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s12, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v5, s12, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v6, s13, v2
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s12, v2
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
 ; GFX6-NEXT:    v_mul_lo_u32 v8, v2, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v9, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v10, v2, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v3, v5
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v3, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v3, v4
 ; GFX6-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
 ; GFX6-NEXT:    v_addc_u32_e32 v9, vcc, 0, v10, vcc

 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s2, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v5, s2, v2
 ; GFX6-NEXT:    v_subrev_i32_e32 v0, vcc, s0, v0
 ; GFX6-NEXT:    v_subb_u32_e32 v1, vcc, v1, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, s3, v2
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s2, v2
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
 ; GFX6-NEXT:    v_sub_i32_e32 v6, vcc, s7, v4
 ; GFX6-NEXT:    v_mov_b32_e32 v7, s3
 ; GFX6-NEXT:    v_sub_i32_e32 v5, vcc, s6, v5
 ; GFX6-NEXT:    v_subb_u32_e64 v6, s[0:1], v6, v7, vcc
 ; GFX6-NEXT:    v_subrev_i32_e64 v7, s[0:1], s2, v5
 ; GFX6-NEXT:    v_subbrev_u32_e64 v6, s[0:1], 0, v6, s[0:1]
 ; GFX6-NEXT:    v_cmp_le_u32_e64 s[0:1], s3, v6
 ; GFX6-NEXT:    v_cndmask_b32_e64 v8, 0, -1, s[0:1]

 ; GFX6-NEXT:    v_trunc_f32_e32 v1, v1
 ; GFX6-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s8
 ; GFX6-NEXT:    v_mul_lo_u32 v4, v0, s8
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, s8
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v0, s8
-; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
+; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_add_u32 s2, s2, s10
 ; GFX6-NEXT:    s_mov_b32 s11, s10
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s5, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s4, v0
 ; GFX6-NEXT:    s_addc_u32 s3, s3, s10
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GFX6-NEXT:    s_xor_b64 s[12:13], s[2:3], s[10:11]
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s5, v0
 ; GFX6-NEXT:    s_mov_b32 s5, s1
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s4, v0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GFX6-NEXT:    v_addc_u32_e32 v0, vcc, v3, v0, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v2, vcc, 0, v5, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, 0, v2, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s8, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v2, s8, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s9, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s8, v0
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v2
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
 ; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, s13, v1
 ; GFX6-NEXT:    v_mov_b32_e32 v3, s9
 ; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s12, v0
 ; GFX6-NEXT:    v_subb_u32_e64 v2, s[0:1], v2, v3, vcc
 ; GFX6-NEXT:    v_subrev_i32_e64 v4, s[0:1], s8, v0
 ; GFX6-NEXT:    v_subbrev_u32_e64 v5, s[2:3], 0, v2, s[0:1]
 ; GFX6-NEXT:    v_cmp_le_u32_e64 s[2:3], s9, v5
 ; GFX6-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[2:3]

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_addc_u32 s1, s5, s12
 ; GFX6-NEXT:    s_xor_b64 s[4:5], s[0:1], s[12:13]
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s2, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s2, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s3, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s2, v0
 ; GFX6-NEXT:    s_mov_b32 s10, -1
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s2, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s2, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s3, v0
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s2, v0
-; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GFX6-NEXT:    v_addc_u32_e32 v0, vcc, v3, v0, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v2, vcc, 0, v5, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_addc_u32_e32 v1, vcc, 0, v2, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s16, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v2, s16, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s17, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s16, v0
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v2
-; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
+; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
 ; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, s5, v1
 ; GFX6-NEXT:    v_mov_b32_e32 v3, s17
 ; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    v_subb_u32_e64 v2, s[0:1], v2, v3, vcc
 ; GFX6-NEXT:    v_subrev_i32_e64 v4, s[0:1], s16, v0
 ; GFX6-NEXT:    v_subbrev_u32_e64 v5, s[2:3], 0, v2, s[0:1]
 ; GFX6-NEXT:    v_cmp_le_u32_e64 s[2:3], s17, v5
 ; GFX6-NEXT:    v_cndmask_b32_e64 v6, 0, -1, s[2:3]

 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; GFX6-NEXT:    s_sub_u32 s0, 0, s4
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX6-NEXT:    v_mul_hi_u32 v3, s0, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s0, v4
 ; GFX6-NEXT:    s_subb_u32 s1, 0, s5
 ; GFX6-NEXT:    v_mul_lo_u32 v6, s1, v2
 ; GFX6-NEXT:    s_ashr_i32 s14, s7, 31
-; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
+; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s0, v2
-; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v6, v3
+; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GFX6-NEXT:    v_mul_lo_u32 v6, v2, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v8, v2, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v9, v4, v3
 ; GFX6-NEXT:    v_mul_lo_u32 v3, v4, v3
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GFX6-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v8, v4, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v5, v4, v5
 ; GFX6-NEXT:    s_mov_b32 s15, s14
 ; GFX6-NEXT:    v_xor_b32_e32 v0, s12, v0
 ; GFX6-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, v7, v5, vcc
 ; GFX6-NEXT:    v_addc_u32_e32 v6, vcc, 0, v9, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
 ; GFX6-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, v4, v5, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v4, s0, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v5, s0, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v6, s1, v2
 ; GFX6-NEXT:    v_xor_b32_e32 v1, s12, v1
 ; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s0, v2
-; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; GFX6-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
 ; GFX6-NEXT:    v_mul_lo_u32 v8, v2, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v9, v2, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v10, v2, v4
 ; GFX6-NEXT:    v_mul_hi_u32 v7, v3, v5
 ; GFX6-NEXT:    v_mul_lo_u32 v5, v3, v5
 ; GFX6-NEXT:    v_mul_hi_u32 v6, v3, v4
 ; GFX6-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
 ; GFX6-NEXT:    v_addc_u32_e32 v9, vcc, 0, v10, vcc

 ; GFX6-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GFX6-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GFX6-NEXT:    v_mul_lo_u32 v3, s4, v3
 ; GFX6-NEXT:    v_mul_hi_u32 v4, s4, v2
 ; GFX6-NEXT:    v_mul_lo_u32 v5, s5, v2
 ; GFX6-NEXT:    v_subrev_i32_e32 v0, vcc, s12, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v2, s4, v2
 ; GFX6-NEXT:    v_subb_u32_e32 v1, vcc, v1, v6, vcc
-; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
-; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
+; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
+; GFX6-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GFX6-NEXT:    v_sub_i32_e32 v4, vcc, s7, v3
 ; GFX6-NEXT:    v_mov_b32_e32 v5, s5
 ; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, s6, v2
 ; GFX6-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v5, vcc
 ; GFX6-NEXT:    v_subrev_i32_e64 v6, s[0:1], s4, v2
 ; GFX6-NEXT:    v_subbrev_u32_e64 v7, s[2:3], 0, v4, s[0:1]
 ; GFX6-NEXT:    v_cmp_le_u32_e64 s[2:3], s5, v7
 ; GFX6-NEXT:    v_cndmask_b32_e64 v8, 0, -1, s[2:3]

llvm/test/CodeGen/AMDGPU/carryout-selection.ll

 ; CISI-NEXT:    v_trunc_f32_e32 v1, v1
 ; CISI-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; CISI-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; CISI-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; CISI-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; CISI-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; CISI-NEXT:    v_mul_lo_u32 v5, s1, v0
 ; CISI-NEXT:    v_mul_lo_u32 v4, s0, v0
-; CISI-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; CISI-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; CISI-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; CISI-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; CISI-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; CISI-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; CISI-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; CISI-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; CISI-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; CISI-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; CISI-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; CISI-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; CISI-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; CISI-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; CISI-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; CISI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; CISI-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; CISI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; CISI-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; CISI-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; CISI-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; CISI-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; CISI-NEXT:    v_mul_lo_u32 v4, s1, v0
-; CISI-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; CISI-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; CISI-NEXT:    v_mul_lo_u32 v3, s0, v0
 ; CISI-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; CISI-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; CISI-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; CISI-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; CISI-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; CISI-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; CISI-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; VI-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
 ; VI-NEXT:    v_trunc_f32_e32 v1, v1
 ; VI-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; VI-NEXT:    v_cvt_u32_f32_e32 v4, v1
 ; VI-NEXT:    v_cvt_u32_f32_e32 v5, v0
 ; VI-NEXT:    v_mul_lo_u32 v2, s8, v4
 ; VI-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], s8, v5, 0
 ; VI-NEXT:    v_mul_lo_u32 v3, s9, v5
-; VI-NEXT:    v_add_u32_e32 v1, vcc, v2, v1
-; VI-NEXT:    v_add_u32_e32 v3, vcc, v3, v1
+; VI-NEXT:    v_add_u32_e32 v1, vcc, v1, v2
+; VI-NEXT:    v_add_u32_e32 v3, vcc, v1, v3
 ; VI-NEXT:    v_mul_hi_u32 v6, v5, v0
 ; VI-NEXT:    v_mad_u64_u32 v[1:2], s[0:1], v5, v3, 0
 ; VI-NEXT:    v_add_u32_e32 v6, vcc, v6, v1
 ; VI-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v4, v0, 0
 ; VI-NEXT:    v_addc_u32_e32 v7, vcc, 0, v2, vcc
 ; VI-NEXT:    v_mad_u64_u32 v[2:3], s[0:1], v4, v3, 0
 ; VI-NEXT:    v_add_u32_e32 v0, vcc, v6, v0
 ; VI-NEXT:    v_addc_u32_e32 v0, vcc, v7, v1, vcc
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v3, vcc
 ; VI-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    v_add_u32_e32 v6, vcc, v5, v0
 ; VI-NEXT:    v_addc_u32_e32 v7, vcc, v4, v1, vcc
 ; VI-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], s8, v6, 0
 ; VI-NEXT:    v_mul_lo_u32 v4, s8, v7
 ; VI-NEXT:    v_mul_lo_u32 v5, s9, v6
 ; VI-NEXT:    v_mul_hi_u32 v8, v6, v0
 ; VI-NEXT:    v_mad_u64_u32 v[2:3], s[0:1], v7, v0, 0
 ; VI-NEXT:    v_add_u32_e32 v1, vcc, v1, v4
-; VI-NEXT:    v_add_u32_e32 v1, vcc, v5, v1
+; VI-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
 ; VI-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v6, v1, 0
 ; VI-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v7, v1, 0
 ; VI-NEXT:    v_add_u32_e32 v4, vcc, v8, v4
 ; VI-NEXT:    v_addc_u32_e32 v5, vcc, 0, v5, vcc
 ; VI-NEXT:    v_add_u32_e32 v2, vcc, v4, v2
 ; VI-NEXT:    v_addc_u32_e32 v2, vcc, v5, v3, vcc
 ; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
 ; VI-NEXT:    v_add_u32_e32 v0, vcc, v2, v0

llvm/test/CodeGen/AMDGPU/control-flow-fastregalloc.ll

-; RUN: llc -O0 -mtriple=amdgcn--amdhsa -march=amdgcn -amdgpu-spill-sgpr-to-vgpr=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=VMEM -check-prefix=GCN %s
-; RUN: llc -O0 -mtriple=amdgcn--amdhsa -march=amdgcn -amdgpu-spill-sgpr-to-vgpr=1 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=VGPR -check-prefix=GCN %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -O0 -mtriple=amdgcn--amdhsa -march=amdgcn --amdhsa-code-object-version=2 -amdgpu-spill-sgpr-to-vgpr=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=VMEM %s
+; RUN: llc -O0 -mtriple=amdgcn--amdhsa -march=amdgcn --amdhsa-code-object-version=2 -amdgpu-spill-sgpr-to-vgpr=1 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=VGPR %s
 
 ; Verify registers used for tracking exec mask changes when all
 ; registers are spilled at the end of the block. The SGPR spill
 ; placement relative to the exec modifications are important.
 
 ; FIXME: This checks with SGPR to VGPR spilling disabled, but this may
 ; not work correctly in cases where no workitems take a branch.
 
-
-; GCN-LABEL: {{^}}divergent_if_endif:
-; VGPR: workitem_private_segment_byte_size = 12{{$}}
-
-
-; GCN: {{^}}; %bb.0:
-; GCN: s_mov_b32 m0, -1
-; GCN: ds_read_b32 [[LOAD0:v[0-9]+]]
-
-; Spill load
-; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], 0 offset:[[LOAD0_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-; GCN: v_cmp_eq_u32_e64 [[CMP0:s\[[0-9]+:[0-9]\]]], v0, s{{[0-9]+}}
-
-; Spill saved exec
-; GCN: s_mov_b64 s[[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]], exec
-; VGPR: v_writelane_b32 [[SPILL_VGPR:v[0-9]+]], s[[SAVEEXEC_LO]], [[SAVEEXEC_LO_LANE:[0-9]+]]
-; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[SAVEEXEC_HI]], [[SAVEEXEC_HI_LANE:[0-9]+]]
-
-; VMEM: v_writelane_b32 v[[V_SAVEEXEC:[0-9]+]], s[[SAVEEXEC_LO]], 0
-; VMEM: v_writelane_b32 v[[V_SAVEEXEC]], s[[SAVEEXEC_HI]], 1
-; VMEM: buffer_store_dword v[[V_SAVEEXEC]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; GCN: s_and_b64 s[[[ANDEXEC_LO:[0-9]+]]:[[ANDEXEC_HI:[0-9]+]]], s[[[SAVEEXEC_LO]]:[[SAVEEXEC_HI]]], [[CMP0]]
-; GCN: s_mov_b64 exec, s[[[ANDEXEC_LO]]:[[ANDEXEC_HI]]]
-
-; GCN: s_cbranch_execz [[ENDIF:.LBB[0-9]+_[0-9]+]]
-
-; GCN: ; %bb.{{[0-9]+}}: ; %if
-; GCN: buffer_load_dword [[RELOAD_LOAD0:v[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
-; GCN: s_mov_b32 m0, -1
-; GCN: ds_read_b32 [[LOAD1:v[0-9]+]]
-; GCN: s_waitcnt vmcnt(0) lgkmcnt(0)
-
-
-; Spill val register
-; GCN: v_add_i32_e32 [[VAL:v[0-9]+]], vcc, [[LOAD1]], [[RELOAD_LOAD0]]
-; GCN: buffer_store_dword [[VAL]], off, s[0:3], 0 offset:[[VAL_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; VMEM: [[ENDIF]]:
-
-; Restore val
-; GCN: buffer_load_dword [[RELOAD_VAL:v[0-9]+]], off, s[0:3], 0 offset:[[VAL_OFFSET]] ; 4-byte Folded Reload
-
-; Reload and restore exec mask
-; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_LO_LANE]]
-; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_HI_LANE]]
-
-; VMEM: buffer_load_dword v[[V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET]] ; 4-byte Folded Reload
-; VMEM: s_waitcnt vmcnt(0)
-; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 0
-; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 1
-
-; GCN: s_or_b64 exec, exec, s[[[S_RELOAD_SAVEEXEC_LO]]:[[S_RELOAD_SAVEEXEC_HI]]]
-
-; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[RELOAD_VAL]]
 define amdgpu_kernel void @divergent_if_endif(ptr addrspace(1) %out) #0 {
+; VMEM-LABEL: divergent_if_endif:
+; VMEM:         .amd_kernel_code_t
+; VMEM-NEXT:     amd_code_version_major = 1
+; VMEM-NEXT:     amd_code_version_minor = 2
+; VMEM-NEXT:     amd_machine_kind = 1
+; VMEM-NEXT:     amd_machine_version_major = 7
+; VMEM-NEXT:     amd_machine_version_minor = 0
+; VMEM-NEXT:     amd_machine_version_stepping = 0
+; VMEM-NEXT:     kernel_code_entry_byte_offset = 256
+; VMEM-NEXT:     kernel_code_prefetch_byte_size = 0
+; VMEM-NEXT:     granulated_workitem_vgpr_count = 1
+; VMEM-NEXT:     granulated_wavefront_sgpr_count = 1
+; VMEM-NEXT:     priority = 0
+; VMEM-NEXT:     float_mode = 240
+; VMEM-NEXT:     priv = 0
+; VMEM-NEXT:     enable_dx10_clamp = 1
+; VMEM-NEXT:     debug_mode = 0
+; VMEM-NEXT:     enable_ieee_mode = 1
+; VMEM-NEXT:     enable_wgp_mode = 0
+; VMEM-NEXT:     enable_mem_ordered = 0
+; VMEM-NEXT:     enable_fwd_progress = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; VMEM-NEXT:     user_sgpr_count = 6
+; VMEM-NEXT:     enable_trap_handler = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_id_x = 1
+; VMEM-NEXT:     enable_sgpr_workgroup_id_y = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_id_z = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_info = 0
+; VMEM-NEXT:     enable_vgpr_workitem_id = 0
+; VMEM-NEXT:     enable_exception_msb = 0
+; VMEM-NEXT:     granulated_lds_size = 0
+; VMEM-NEXT:     enable_exception = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_buffer = 1
+; VMEM-NEXT:     enable_sgpr_dispatch_ptr = 0
+; VMEM-NEXT:     enable_sgpr_queue_ptr = 0
+; VMEM-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; VMEM-NEXT:     enable_sgpr_dispatch_id = 0
+; VMEM-NEXT:     enable_sgpr_flat_scratch_init = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_size = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; VMEM-NEXT:     enable_wavefront_size32 = 0
+; VMEM-NEXT:     enable_ordered_append_gds = 0
+; VMEM-NEXT:     private_element_size = 1
+; VMEM-NEXT:     is_ptr64 = 1
+; VMEM-NEXT:     is_dynamic_callstack = 0
+; VMEM-NEXT:     is_debug_enabled = 0
+; VMEM-NEXT:     is_xnack_enabled = 0
+; VMEM-NEXT:     workitem_private_segment_byte_size = 28
+; VMEM-NEXT:     workgroup_group_segment_byte_size = 0
+; VMEM-NEXT:     gds_segment_byte_size = 0
+; VMEM-NEXT:     kernarg_segment_byte_size = 8
+; VMEM-NEXT:     workgroup_fbarrier_count = 0
+; VMEM-NEXT:     wavefront_sgpr_count = 14
+; VMEM-NEXT:     workitem_vgpr_count = 5
+; VMEM-NEXT:     reserved_vgpr_first = 0
+; VMEM-NEXT:     reserved_vgpr_count = 0
+; VMEM-NEXT:     reserved_sgpr_first = 0
+; VMEM-NEXT:     reserved_sgpr_count = 0
+; VMEM-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; VMEM-NEXT:     debug_private_segment_buffer_sgpr = 0
+; VMEM-NEXT:     kernarg_segment_alignment = 4
+; VMEM-NEXT:     group_segment_alignment = 4
+; VMEM-NEXT:     private_segment_alignment = 4
+; VMEM-NEXT:     wavefront_size = 6
+; VMEM-NEXT:     call_convention = -1
+; VMEM-NEXT:     runtime_loader_kernel_symbol = 0
+; VMEM-NEXT:    .end_amd_kernel_code_t
+; VMEM-NEXT:  ; %bb.0: ; %entry
+; VMEM-NEXT:    s_add_u32 s0, s0, s7
+; VMEM-NEXT:    s_addc_u32 s1, s1, 0
+; VMEM-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VMEM-NEXT:    s_mov_b64 s[10:11], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    v_writelane_b32 v2, s4, 0
+; VMEM-NEXT:    v_writelane_b32 v2, s5, 1
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[10:11]
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s4
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    buffer_store_dword v1, off, s[0:3], 0 offset:24 ; 4-byte Folded Spill
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    s_mov_b32 s4, 0
+; VMEM-NEXT:    v_cmp_eq_u32_e64 s[6:7], v0, s4
+; VMEM-NEXT:    v_mov_b32_e32 v0, s4
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:20 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_mov_b64 s[4:5], exec
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    v_writelane_b32 v0, s4, 0
+; VMEM-NEXT:    v_writelane_b32 v0, s5, 1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_and_b64 s[4:5], s[4:5], s[6:7]
+; VMEM-NEXT:    s_mov_b64 exec, s[4:5]
+; VMEM-NEXT:    s_cbranch_execz .LBB0_2
+; VMEM-NEXT:  ; %bb.1: ; %if
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:24 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s4
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VMEM-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:20 ; 4-byte Folded Spill
+; VMEM-NEXT:  .LBB0_2: ; %endif
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0 offset:20 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v4, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v4, off, s[0:3], 0 offset:4 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s6, v4, 0
+; VMEM-NEXT:    v_readlane_b32 s7, v4, 1
+; VMEM-NEXT:    buffer_load_dword v4, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_or_b64 exec, exec, s[6:7]
+; VMEM-NEXT:    s_mov_b64 s[6:7], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0 offset:12 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s4, v3, 0
+; VMEM-NEXT:    v_readlane_b32 s5, v3, 1
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[6:7]
+; VMEM-NEXT:    v_mov_b32_e32 v0, s4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s5
+; VMEM-NEXT:    flat_store_dword v[0:1], v2
+; VMEM-NEXT:    s_endpgm
+;
+; VGPR-LABEL: divergent_if_endif:
+; VGPR:         .amd_kernel_code_t
+; VGPR-NEXT:     amd_code_version_major = 1
+; VGPR-NEXT:     amd_code_version_minor = 2
+; VGPR-NEXT:     amd_machine_kind = 1
+; VGPR-NEXT:     amd_machine_version_major = 7
+; VGPR-NEXT:     amd_machine_version_minor = 0
+; VGPR-NEXT:     amd_machine_version_stepping = 0
+; VGPR-NEXT:     kernel_code_entry_byte_offset = 256
+; VGPR-NEXT:     kernel_code_prefetch_byte_size = 0
+; VGPR-NEXT:     granulated_workitem_vgpr_count = 0
+; VGPR-NEXT:     granulated_wavefront_sgpr_count = 1
+; VGPR-NEXT:     priority = 0
+; VGPR-NEXT:     float_mode = 240
+; VGPR-NEXT:     priv = 0
+; VGPR-NEXT:     enable_dx10_clamp = 1
+; VGPR-NEXT:     debug_mode = 0
+; VGPR-NEXT:     enable_ieee_mode = 1
+; VGPR-NEXT:     enable_wgp_mode = 0
+; VGPR-NEXT:     enable_mem_ordered = 0
+; VGPR-NEXT:     enable_fwd_progress = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; VGPR-NEXT:     user_sgpr_count = 6
+; VGPR-NEXT:     enable_trap_handler = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_id_x = 1
+; VGPR-NEXT:     enable_sgpr_workgroup_id_y = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_id_z = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_info = 0
+; VGPR-NEXT:     enable_vgpr_workitem_id = 0
+; VGPR-NEXT:     enable_exception_msb = 0
+; VGPR-NEXT:     granulated_lds_size = 0
+; VGPR-NEXT:     enable_exception = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_buffer = 1
+; VGPR-NEXT:     enable_sgpr_dispatch_ptr = 0
+; VGPR-NEXT:     enable_sgpr_queue_ptr = 0
+; VGPR-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; VGPR-NEXT:     enable_sgpr_dispatch_id = 0
+; VGPR-NEXT:     enable_sgpr_flat_scratch_init = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_size = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; VGPR-NEXT:     enable_wavefront_size32 = 0
+; VGPR-NEXT:     enable_ordered_append_gds = 0
+; VGPR-NEXT:     private_element_size = 1
+; VGPR-NEXT:     is_ptr64 = 1
+; VGPR-NEXT:     is_dynamic_callstack = 0
+; VGPR-NEXT:     is_debug_enabled = 0
+; VGPR-NEXT:     is_xnack_enabled = 0
+; VGPR-NEXT:     workitem_private_segment_byte_size = 12
+; VGPR-NEXT:     workgroup_group_segment_byte_size = 0
+; VGPR-NEXT:     gds_segment_byte_size = 0
+; VGPR-NEXT:     kernarg_segment_byte_size = 8
+; VGPR-NEXT:     workgroup_fbarrier_count = 0
+; VGPR-NEXT:     wavefront_sgpr_count = 10
+; VGPR-NEXT:     workitem_vgpr_count = 4
+; VGPR-NEXT:     reserved_vgpr_first = 0
+; VGPR-NEXT:     reserved_vgpr_count = 0
+; VGPR-NEXT:     reserved_sgpr_first = 0
+; VGPR-NEXT:     reserved_sgpr_count = 0
+; VGPR-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; VGPR-NEXT:     debug_private_segment_buffer_sgpr = 0
+; VGPR-NEXT:     kernarg_segment_alignment = 4
+; VGPR-NEXT:     group_segment_alignment = 4
+; VGPR-NEXT:     private_segment_alignment = 4
+; VGPR-NEXT:     wavefront_size = 6
+; VGPR-NEXT:     call_convention = -1
+; VGPR-NEXT:     runtime_loader_kernel_symbol = 0
+; VGPR-NEXT:    .end_amd_kernel_code_t
+; VGPR-NEXT:  ; %bb.0: ; %entry
+; VGPR-NEXT:    s_add_u32 s0, s0, s7
+; VGPR-NEXT:    s_addc_u32 s1, s1, 0
+; VGPR-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 0
+; VGPR-NEXT:    v_writelane_b32 v1, s5, 1
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; VGPR-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:8 ; 4-byte Folded Spill
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    s_mov_b32 s4, 0
+; VGPR-NEXT:    v_cmp_eq_u32_e64 s[6:7], v0, s4
+; VGPR-NEXT:    v_mov_b32_e32 v0, s4
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_mov_b64 s[4:5], exec
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 2
+; VGPR-NEXT:    v_writelane_b32 v1, s5, 3
+; VGPR-NEXT:    s_and_b64 s[4:5], s[4:5], s[6:7]
+; VGPR-NEXT:    s_mov_b64 exec, s[4:5]
+; VGPR-NEXT:    s_cbranch_execz .LBB0_2
+; VGPR-NEXT:  ; %bb.1: ; %if
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:8 ; 4-byte Folded Reload
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VGPR-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VGPR-NEXT:  .LBB0_2: ; %endif
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Reload
+; VGPR-NEXT:    v_readlane_b32 s6, v1, 2
+; VGPR-NEXT:    v_readlane_b32 s7, v1, 3
+; VGPR-NEXT:    s_or_b64 exec, exec, s[6:7]
+; VGPR-NEXT:    v_readlane_b32 s4, v1, 0
+; VGPR-NEXT:    v_readlane_b32 s5, v1, 1
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    v_mov_b32_e32 v3, s5
+; VGPR-NEXT:    s_waitcnt vmcnt(0)
+; VGPR-NEXT:    flat_store_dword v[2:3], v0
+; VGPR-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %load0 = load volatile i32, ptr addrspace(3) undef
   %cmp0 = icmp eq i32 %tid, 0
   br i1 %cmp0, label %if, label %endif
 
 if:
   %load1 = load volatile i32, ptr addrspace(3) undef
   %val = add i32 %load0, %load1
   br label %endif
 
 endif:
   %tmp4 = phi i32 [ %val, %if ], [ 0, %entry ]
   store i32 %tmp4, ptr addrspace(1) %out
   ret void
 }
 
-; GCN-LABEL: {{^}}divergent_loop:
-; VGPR: workitem_private_segment_byte_size = 16{{$}}
-
-; GCN: {{^}}; %bb.0:
-; GCN-DAG: s_mov_b32 m0, -1
-; GCN-DAG: v_mov_b32_e32 [[PTR0:v[0-9]+]], 0{{$}}
-; GCN: ds_read_b32 [[LOAD0:v[0-9]+]], [[PTR0]]
-; GCN: v_cmp_eq_u32_e64 [[CMP0:s\[[0-9]+:[0-9]+\]]], v0, s{{[0-9]+}}
-
-; Spill load
-; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], 0 offset:[[LOAD0_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-; GCN: s_mov_b64 s[[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]], exec
-
-; Spill saved exec
-; VGPR: v_writelane_b32 [[SPILL_VGPR:v[0-9]+]], s[[SAVEEXEC_LO]], [[SAVEEXEC_LO_LANE:[0-9]+]]
-; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[SAVEEXEC_HI]], [[SAVEEXEC_HI_LANE:[0-9]+]]
-
-; VMEM: v_writelane_b32 v[[V_SAVEEXEC:[0-9]+]], s[[SAVEEXEC_LO]], 0
-; VMEM: v_writelane_b32 v[[V_SAVEEXEC]], s[[SAVEEXEC_HI]], 1
-; VMEM: buffer_store_dword v[[V_SAVEEXEC]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-
-; GCN: s_and_b64 s[[[ANDEXEC_LO:[0-9]+]]:[[ANDEXEC_HI:[0-9]+]]], s[[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]], [[CMP0]]
-; GCN: s_mov_b64 exec, s[[[ANDEXEC_LO]]:[[ANDEXEC_HI]]]
-; GCN-NEXT: s_cbranch_execz [[END:.LBB[0-9]+_[0-9]+]]
-
-
-; GCN: [[LOOP:.LBB[0-9]+_[0-9]+]]:
-; GCN: buffer_load_dword v[[VAL_LOOP_RELOAD:[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
-; GCN: v_subrev_i32_e32 [[VAL_LOOP:v[0-9]+]], vcc, v{{[0-9]+}}, v[[VAL_LOOP_RELOAD]]
-; GCN: s_cmp_lg_u32
-; VMEM: buffer_store_dword
-; VMEM: buffer_store_dword
-; VMEM: buffer_store_dword
-; GCN: buffer_store_dword [[VAL_LOOP]], off, s[0:3], 0 offset:{{[0-9]+}} ; 4-byte Folded Spill
-; GCN-NEXT: s_cbranch_scc1 [[LOOP]]
-
-; GCN: buffer_store_dword [[VAL_LOOP]], off, s[0:3], 0 offset:[[VAL_SUB_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; GCN: [[END]]:
-; GCN: buffer_load_dword v[[VAL_END:[0-9]+]], off, s[0:3], 0 offset:[[VAL_SUB_OFFSET]] ; 4-byte Folded Reload
-; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_LO_LANE]]
-; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_HI_LANE]]
-
-; VMEM: buffer_load_dword v[[V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[V_EXEC_SPILL_OFFSET]] ; 4-byte Folded Reload
-; VMEM: s_waitcnt vmcnt(0)
-; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 0
-; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 1
-
-; GCN: s_or_b64 exec, exec, s[[[S_RELOAD_SAVEEXEC_LO]]:[[S_RELOAD_SAVEEXEC_HI]]]
-
-; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, v[[VAL_END]]
 define amdgpu_kernel void @divergent_loop(ptr addrspace(1) %out) #0 {
+; VMEM-LABEL: divergent_loop:
+; VMEM:         .amd_kernel_code_t
+; VMEM-NEXT:     amd_code_version_major = 1
+; VMEM-NEXT:     amd_code_version_minor = 2
+; VMEM-NEXT:     amd_machine_kind = 1
+; VMEM-NEXT:     amd_machine_version_major = 7
+; VMEM-NEXT:     amd_machine_version_minor = 0
+; VMEM-NEXT:     amd_machine_version_stepping = 0
+; VMEM-NEXT:     kernel_code_entry_byte_offset = 256
+; VMEM-NEXT:     kernel_code_prefetch_byte_size = 0
+; VMEM-NEXT:     granulated_workitem_vgpr_count = 1
+; VMEM-NEXT:     granulated_wavefront_sgpr_count = 1
+; VMEM-NEXT:     priority = 0
+; VMEM-NEXT:     float_mode = 240
+; VMEM-NEXT:     priv = 0
+; VMEM-NEXT:     enable_dx10_clamp = 1
+; VMEM-NEXT:     debug_mode = 0
+; VMEM-NEXT:     enable_ieee_mode = 1
+; VMEM-NEXT:     enable_wgp_mode = 0
+; VMEM-NEXT:     enable_mem_ordered = 0
+; VMEM-NEXT:     enable_fwd_progress = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; VMEM-NEXT:     user_sgpr_count = 6
+; VMEM-NEXT:     enable_trap_handler = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_id_x = 1
+; VMEM-NEXT:     enable_sgpr_workgroup_id_y = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_id_z = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_info = 0
+; VMEM-NEXT:     enable_vgpr_workitem_id = 0
+; VMEM-NEXT:     enable_exception_msb = 0
+; VMEM-NEXT:     granulated_lds_size = 0
+; VMEM-NEXT:     enable_exception = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_buffer = 1
+; VMEM-NEXT:     enable_sgpr_dispatch_ptr = 0
+; VMEM-NEXT:     enable_sgpr_queue_ptr = 0
+; VMEM-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; VMEM-NEXT:     enable_sgpr_dispatch_id = 0
+; VMEM-NEXT:     enable_sgpr_flat_scratch_init = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_size = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; VMEM-NEXT:     enable_wavefront_size32 = 0
+; VMEM-NEXT:     enable_ordered_append_gds = 0
+; VMEM-NEXT:     private_element_size = 1
+; VMEM-NEXT:     is_ptr64 = 1
+; VMEM-NEXT:     is_dynamic_callstack = 0
+; VMEM-NEXT:     is_debug_enabled = 0
+; VMEM-NEXT:     is_xnack_enabled = 0
+; VMEM-NEXT:     workitem_private_segment_byte_size = 36
+; VMEM-NEXT:     workgroup_group_segment_byte_size = 0
+; VMEM-NEXT:     gds_segment_byte_size = 0
+; VMEM-NEXT:     kernarg_segment_byte_size = 8
+; VMEM-NEXT:     workgroup_fbarrier_count = 0
+; VMEM-NEXT:     wavefront_sgpr_count = 16
+; VMEM-NEXT:     workitem_vgpr_count = 5
+; VMEM-NEXT:     reserved_vgpr_first = 0
+; VMEM-NEXT:     reserved_vgpr_count = 0
+; VMEM-NEXT:     reserved_sgpr_first = 0
+; VMEM-NEXT:     reserved_sgpr_count = 0
+; VMEM-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; VMEM-NEXT:     debug_private_segment_buffer_sgpr = 0
+; VMEM-NEXT:     kernarg_segment_alignment = 4
+; VMEM-NEXT:     group_segment_alignment = 4
+; VMEM-NEXT:     private_segment_alignment = 4
+; VMEM-NEXT:     wavefront_size = 6
+; VMEM-NEXT:     call_convention = -1
+; VMEM-NEXT:     runtime_loader_kernel_symbol = 0
+; VMEM-NEXT:    .end_amd_kernel_code_t
+; VMEM-NEXT:  ; %bb.0: ; %entry
+; VMEM-NEXT:    s_add_u32 s0, s0, s7
+; VMEM-NEXT:    s_addc_u32 s1, s1, 0
+; VMEM-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VMEM-NEXT:    s_mov_b64 s[12:13], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    v_writelane_b32 v3, s4, 0
+; VMEM-NEXT:    v_writelane_b32 v3, s5, 1
+; VMEM-NEXT:    buffer_store_dword v3, off, s[0:3], 0 offset:16 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[12:13]
+; VMEM-NEXT:    v_mov_b32_e32 v1, 0
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    s_mov_b32 s4, 0
+; VMEM-NEXT:    v_cmp_eq_u32_e64 s[6:7], v0, s4
+; VMEM-NEXT:    v_mov_b32_e32 v0, s4
+; VMEM-NEXT:    s_mov_b64 s[10:11], exec
+; VMEM-NEXT:    s_mov_b64 exec, 1
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    v_writelane_b32 v2, s4, 0
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[10:11]
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    buffer_store_dword v1, off, s[0:3], 0 offset:28 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:24 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_mov_b64 s[4:5], exec
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    v_writelane_b32 v0, s4, 0
+; VMEM-NEXT:    v_writelane_b32 v0, s5, 1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_and_b64 s[4:5], s[4:5], s[6:7]
+; VMEM-NEXT:    s_mov_b64 exec, s[4:5]
+; VMEM-NEXT:    s_cbranch_execz .LBB1_3
+; VMEM-NEXT:  .LBB1_1: ; %loop
+; VMEM-NEXT:    ; =>This Inner Loop Header: Depth=1
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:28 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 1
+; VMEM-NEXT:    buffer_store_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0 offset:12 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s5, v3, 0
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s4
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    s_mov_b32 s4, 1
+; VMEM-NEXT:    s_add_i32 s4, s5, s4
+; VMEM-NEXT:    ; implicit-def: $sgpr6
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
+; VMEM-NEXT:    s_mov_b32 s6, 0x100
+; VMEM-NEXT:    s_cmp_lg_u32 s5, s6
+; VMEM-NEXT:    v_mov_b32_e32 v1, v0
+; VMEM-NEXT:    s_mov_b64 s[6:7], exec
+; VMEM-NEXT:    s_mov_b64 exec, 1
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    v_writelane_b32 v2, s4, 0
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[6:7]
+; VMEM-NEXT:    buffer_store_dword v1, off, s[0:3], 0 offset:28 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:32 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_cbranch_scc1 .LBB1_1
+; VMEM-NEXT:  ; %bb.2: ; %Flow
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:32 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:24 ; 4-byte Folded Spill
+; VMEM-NEXT:  .LBB1_3: ; %end
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0 offset:24 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v4, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v4, off, s[0:3], 0 offset:4 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s6, v4, 0
+; VMEM-NEXT:    v_readlane_b32 s7, v4, 1
+; VMEM-NEXT:    buffer_load_dword v4, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_or_b64 exec, exec, s[6:7]
+; VMEM-NEXT:    s_mov_b64 s[6:7], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0 offset:16 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s4, v3, 0
+; VMEM-NEXT:    v_readlane_b32 s5, v3, 1
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[6:7]
+; VMEM-NEXT:    v_mov_b32_e32 v0, s4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s5
+; VMEM-NEXT:    flat_store_dword v[0:1], v2
+; VMEM-NEXT:    s_endpgm
+;
+; VGPR-LABEL: divergent_loop:
+; VGPR:         .amd_kernel_code_t
+; VGPR-NEXT:     amd_code_version_major = 1
+; VGPR-NEXT:     amd_code_version_minor = 2
+; VGPR-NEXT:     amd_machine_kind = 1
+; VGPR-NEXT:     amd_machine_version_major = 7
+; VGPR-NEXT:     amd_machine_version_minor = 0
+; VGPR-NEXT:     amd_machine_version_stepping = 0
+; VGPR-NEXT:     kernel_code_entry_byte_offset = 256
+; VGPR-NEXT:     kernel_code_prefetch_byte_size = 0
+; VGPR-NEXT:     granulated_workitem_vgpr_count = 0
+; VGPR-NEXT:     granulated_wavefront_sgpr_count = 1
+; VGPR-NEXT:     priority = 0
+; VGPR-NEXT:     float_mode = 240
+; VGPR-NEXT:     priv = 0
+; VGPR-NEXT:     enable_dx10_clamp = 1
+; VGPR-NEXT:     debug_mode = 0
+; VGPR-NEXT:     enable_ieee_mode = 1
+; VGPR-NEXT:     enable_wgp_mode = 0
+; VGPR-NEXT:     enable_mem_ordered = 0
+; VGPR-NEXT:     enable_fwd_progress = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; VGPR-NEXT:     user_sgpr_count = 6
+; VGPR-NEXT:     enable_trap_handler = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_id_x = 1
+; VGPR-NEXT:     enable_sgpr_workgroup_id_y = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_id_z = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_info = 0
+; VGPR-NEXT:     enable_vgpr_workitem_id = 0
+; VGPR-NEXT:     enable_exception_msb = 0
+; VGPR-NEXT:     granulated_lds_size = 0
+; VGPR-NEXT:     enable_exception = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_buffer = 1
+; VGPR-NEXT:     enable_sgpr_dispatch_ptr = 0
+; VGPR-NEXT:     enable_sgpr_queue_ptr = 0
+; VGPR-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; VGPR-NEXT:     enable_sgpr_dispatch_id = 0
+; VGPR-NEXT:     enable_sgpr_flat_scratch_init = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_size = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; VGPR-NEXT:     enable_wavefront_size32 = 0
+; VGPR-NEXT:     enable_ordered_append_gds = 0
+; VGPR-NEXT:     private_element_size = 1
+; VGPR-NEXT:     is_ptr64 = 1
+; VGPR-NEXT:     is_dynamic_callstack = 0
+; VGPR-NEXT:     is_debug_enabled = 0
+; VGPR-NEXT:     is_xnack_enabled = 0
+; VGPR-NEXT:     workitem_private_segment_byte_size = 16
+; VGPR-NEXT:     workgroup_group_segment_byte_size = 0
+; VGPR-NEXT:     gds_segment_byte_size = 0
+; VGPR-NEXT:     kernarg_segment_byte_size = 8
+; VGPR-NEXT:     workgroup_fbarrier_count = 0
+; VGPR-NEXT:     wavefront_sgpr_count = 10
+; VGPR-NEXT:     workitem_vgpr_count = 4
+; VGPR-NEXT:     reserved_vgpr_first = 0
+; VGPR-NEXT:     reserved_vgpr_count = 0
+; VGPR-NEXT:     reserved_sgpr_first = 0
+; VGPR-NEXT:     reserved_sgpr_count = 0
+; VGPR-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; VGPR-NEXT:     debug_private_segment_buffer_sgpr = 0
+; VGPR-NEXT:     kernarg_segment_alignment = 4
+; VGPR-NEXT:     group_segment_alignment = 4
+; VGPR-NEXT:     private_segment_alignment = 4
+; VGPR-NEXT:     wavefront_size = 6
+; VGPR-NEXT:     call_convention = -1
+; VGPR-NEXT:     runtime_loader_kernel_symbol = 0
+; VGPR-NEXT:    .end_amd_kernel_code_t
+; VGPR-NEXT:  ; %bb.0: ; %entry
+; VGPR-NEXT:    s_add_u32 s0, s0, s7
+; VGPR-NEXT:    s_addc_u32 s1, s1, 0
+; VGPR-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 0
+; VGPR-NEXT:    v_writelane_b32 v1, s5, 1
+; VGPR-NEXT:    v_mov_b32_e32 v2, 0
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    s_mov_b32 s4, 0
+; VGPR-NEXT:    v_cmp_eq_u32_e64 s[6:7], v0, s4
+; VGPR-NEXT:    v_mov_b32_e32 v0, s4
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 2
+; VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; VGPR-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:8 ; 4-byte Folded Spill
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_mov_b64 s[4:5], exec
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 3
+; VGPR-NEXT:    v_writelane_b32 v1, s5, 4
+; VGPR-NEXT:    s_and_b64 s[4:5], s[4:5], s[6:7]
+; VGPR-NEXT:    s_mov_b64 exec, s[4:5]
+; VGPR-NEXT:    s_cbranch_execz .LBB1_3
+; VGPR-NEXT:  .LBB1_1: ; %loop
+; VGPR-NEXT:    ; =>This Inner Loop Header: Depth=1
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:8 ; 4-byte Folded Reload
+; VGPR-NEXT:    v_readlane_b32 s5, v1, 2
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    s_mov_b32 s4, 1
+; VGPR-NEXT:    s_add_i32 s4, s5, s4
+; VGPR-NEXT:    ; implicit-def: $sgpr6
+; VGPR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VGPR-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
+; VGPR-NEXT:    s_mov_b32 s6, 0x100
+; VGPR-NEXT:    s_cmp_lg_u32 s5, s6
+; VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 2
+; VGPR-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:8 ; 4-byte Folded Spill
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_cbranch_scc1 .LBB1_1
+; VGPR-NEXT:  ; %bb.2: ; %Flow
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:12 ; 4-byte Folded Reload
+; VGPR-NEXT:    s_waitcnt vmcnt(0)
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VGPR-NEXT:  .LBB1_3: ; %end
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Reload
+; VGPR-NEXT:    v_readlane_b32 s6, v1, 3
+; VGPR-NEXT:    v_readlane_b32 s7, v1, 4
+; VGPR-NEXT:    s_or_b64 exec, exec, s[6:7]
+; VGPR-NEXT:    v_readlane_b32 s4, v1, 0
+; VGPR-NEXT:    v_readlane_b32 s5, v1, 1
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    v_mov_b32_e32 v3, s5
+; VGPR-NEXT:    s_waitcnt vmcnt(0)
+; VGPR-NEXT:    flat_store_dword v[2:3], v0
+; VGPR-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %load0 = load volatile i32, ptr addrspace(3) null
   %cmp0 = icmp eq i32 %tid, 0
   br i1 %cmp0, label %loop, label %end
 
 loop:
   %i = phi i32 [ %i.inc, %loop ], [ 0, %entry ]
   %val = phi i32 [ %val.sub, %loop ], [ %load0, %entry ]
   %load1 = load volatile i32, ptr addrspace(3) undef
   %i.inc = add i32 %i, 1
   %val.sub = sub i32 %val, %load1
   %cmp1 = icmp ne i32 %i, 256
   br i1 %cmp1, label %loop, label %end
 
 end:
   %tmp4 = phi i32 [ %val.sub, %loop ], [ 0, %entry ]
   store i32 %tmp4, ptr addrspace(1) %out
   ret void
 }
 
-; GCN-LABEL: {{^}}divergent_if_else_endif:
-; GCN: {{^}}; %bb.0:
-
-; GCN-DAG: s_mov_b32 m0, -1
-; GCN-DAG: v_mov_b32_e32 [[PTR0:v[0-9]+]], 0{{$}}
-; GCN: ds_read_b32 [[LOAD0:v[0-9]+]], [[PTR0]]
-
-; Spill load
-; GCN: buffer_store_dword [[LOAD0]], off, s[0:3], 0 offset:[[LOAD0_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; GCN: s_mov_b32 [[ZERO:s[0-9]+]], 0
-; GCN: v_cmp_ne_u32_e64 [[CMP0:s\[[0-9]+:[0-9]\]]], v0, [[ZERO]]
-
-; GCN: s_mov_b64 s[[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]], exec
-; GCN: s_and_b64 s[[[ANDEXEC_LO:[0-9]+]]:[[ANDEXEC_HI:[0-9]+]]], s[[[SAVEEXEC_LO:[0-9]+]]:[[SAVEEXEC_HI:[0-9]+]]], [[CMP0]]
-; GCN: s_xor_b64 s[[[SAVEEXEC_LO]]:[[SAVEEXEC_HI]]], s[[[ANDEXEC_LO]]:[[ANDEXEC_HI]]], s[[[SAVEEXEC_LO]]:[[SAVEEXEC_HI]]]
-
-; Spill saved exec
-; VGPR: v_writelane_b32 [[SPILL_VGPR:v[0-9]+]], s[[SAVEEXEC_LO]], [[SAVEEXEC_LO_LANE:[0-9]+]]
-; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[SAVEEXEC_HI]], [[SAVEEXEC_HI_LANE:[0-9]+]]
-
-; VMEM: v_writelane_b32 v[[V_SAVEEXEC:[0-9]+]], s[[SAVEEXEC_LO]], 0
-; VMEM: v_writelane_b32 v[[V_SAVEEXEC]], s[[SAVEEXEC_HI]], 1
-; VMEM: buffer_store_dword v[[V_SAVEEXEC]], off, s[0:3], 0 offset:[[SAVEEXEC_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; GCN: s_mov_b64 exec, [[CMP0]]
-
-; FIXME: It makes no sense to put this skip here
-; GCN: s_cbranch_execz [[FLOW:.LBB[0-9]+_[0-9]+]]
-; GCN-NEXT: s_branch [[ELSE:.LBB[0-9]+_[0-9]+]]
-
-; GCN: [[FLOW]]: ; %Flow
-; GCN: buffer_load_dword [[FLOW_VAL:v[0-9]+]], off, s[0:3], 0 offset:[[FLOW_VAL_OFFSET:[0-9]+]] ; 4-byte Folded Reload
-; VGPR: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_LO_LANE]]
-; VGPR: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[SAVEEXEC_HI_LANE]]
-
-; VMEM: buffer_load_dword v[[FLOW_V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[SAVEEXEC_OFFSET]]
-; VMEM: s_waitcnt vmcnt(0)
-; VMEM: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[FLOW_V_RELOAD_SAVEEXEC]], 0
-; VMEM: v_readlane_b32 s[[FLOW_S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[FLOW_V_RELOAD_SAVEEXEC]], 1
-
-; GCN: s_or_saveexec_b64 s[[[FLOW_S_RELOAD_SAVEEXEC_LO_SAVEEXEC:[0-9]+]]:[[FLOW_S_RELOAD_SAVEEXEC_HI_SAVEEXEC:[0-9]+]]], s[[[FLOW_S_RELOAD_SAVEEXEC_LO]]:[[FLOW_S_RELOAD_SAVEEXEC_HI]]]
-
-; Regular spill value restored after exec modification
-; Followed by spill
-; GCN: buffer_store_dword [[FLOW_VAL]], off, s[0:3], 0 offset:[[RESULT_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; GCN: s_and_b64 s[[[FLOW_AND_EXEC_LO:[0-9]+]]:[[FLOW_AND_EXEC_HI:[0-9]+]]], exec, s[[[FLOW_S_RELOAD_SAVEEXEC_LO_SAVEEXEC]]:[[FLOW_S_RELOAD_SAVEEXEC_HI_SAVEEXEC]]]
-
-; Spill saved exec
-; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[FLOW_AND_EXEC_LO]], [[FLOW_SAVEEXEC_LO_LANE:[0-9]+]]
-; VGPR: v_writelane_b32 [[SPILL_VGPR]], s[[FLOW_AND_EXEC_HI]], [[FLOW_SAVEEXEC_HI_LANE:[0-9]+]]
-
-; VMEM: v_writelane_b32 v[[FLOW_V_SAVEEXEC:[0-9]+]], s[[FLOW_AND_EXEC_LO]], 0
-; VMEM: v_writelane_b32 v[[FLOW_V_SAVEEXEC]], s[[FLOW_AND_EXEC_HI]], 1
-; VMEM: buffer_store_dword v[[FLOW_V_SAVEEXEC]], off, s[0:3], 0 offset:[[FLOW_SAVEEXEC_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-
-; GCN: s_xor_b64 exec, exec, s[[[FLOW_AND_EXEC_LO]]:[[FLOW_AND_EXEC_HI]]]
-; GCN-NEXT: s_cbranch_execz [[ENDIF:.LBB[0-9]+_[0-9]+]]
-
-
-; GCN: ; %bb.{{[0-9]+}}: ; %if
-; GCN: buffer_load_dword v[[LOAD0_RELOAD:[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
-; GCN: ds_read_b32
-; GCN: v_add_i32_e32 [[ADD:v[0-9]+]], vcc, v{{[0-9]+}}, v[[LOAD0_RELOAD]]
-; GCN: buffer_store_dword [[ADD]], off, s[0:3], 0 offset:[[RESULT_OFFSET]] ; 4-byte Folded Spill
-; GCN-NEXT: s_branch [[ENDIF:.LBB[0-9]+_[0-9]+]]
-
-; GCN: [[ELSE]]: ; %else
-; GCN: buffer_load_dword v[[LOAD0_RELOAD:[0-9]+]], off, s[0:3], 0 offset:[[LOAD0_OFFSET]] ; 4-byte Folded Reload
-; GCN: v_subrev_i32_e32 [[SUB:v[0-9]+]], vcc, v{{[0-9]+}}, v[[LOAD0_RELOAD]]
-; GCN: buffer_store_dword [[ADD]], off, s[0:3], 0 offset:[[FLOW_RESULT_OFFSET:[0-9]+]] ; 4-byte Folded Spill
-; GCN-NEXT: s_branch [[FLOW]]
-
-; GCN: [[ENDIF]]:
-; GCN: buffer_load_dword v[[RESULT:[0-9]+]], off, s[0:3], 0 offset:[[RESULT_OFFSET]] ; 4-byte Folded Reload
-; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], [[SPILL_VGPR]], [[FLOW_SAVEEXEC_LO_LANE]]
-; VGPR: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], [[SPILL_VGPR]], [[FLOW_SAVEEXEC_HI_LANE]]
-
-
-; VMEM: buffer_load_dword v[[V_RELOAD_SAVEEXEC:[0-9]+]], off, s[0:3], 0 offset:[[FLOW_SAVEEXEC_OFFSET]] ; 4-byte Folded Reload
-; VMEM: s_waitcnt vmcnt(0)
-; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_LO:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 0
-; VMEM: v_readlane_b32 s[[S_RELOAD_SAVEEXEC_HI:[0-9]+]], v[[V_RELOAD_SAVEEXEC]], 1
-
-; GCN: s_or_b64 exec, exec, s[[[S_RELOAD_SAVEEXEC_LO]]:[[S_RELOAD_SAVEEXEC_HI]]]
-
-; GCN: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, v[[RESULT]]
 define amdgpu_kernel void @divergent_if_else_endif(ptr addrspace(1) %out) #0 {
+; VMEM-LABEL: divergent_if_else_endif:
+; VMEM:         .amd_kernel_code_t
+; VMEM-NEXT:     amd_code_version_major = 1
+; VMEM-NEXT:     amd_code_version_minor = 2
+; VMEM-NEXT:     amd_machine_kind = 1
+; VMEM-NEXT:     amd_machine_version_major = 7
+; VMEM-NEXT:     amd_machine_version_minor = 0
+; VMEM-NEXT:     amd_machine_version_stepping = 0
+; VMEM-NEXT:     kernel_code_entry_byte_offset = 256
+; VMEM-NEXT:     kernel_code_prefetch_byte_size = 0
+; VMEM-NEXT:     granulated_workitem_vgpr_count = 1
+; VMEM-NEXT:     granulated_wavefront_sgpr_count = 1
+; VMEM-NEXT:     priority = 0
+; VMEM-NEXT:     float_mode = 240
+; VMEM-NEXT:     priv = 0
+; VMEM-NEXT:     enable_dx10_clamp = 1
+; VMEM-NEXT:     debug_mode = 0
+; VMEM-NEXT:     enable_ieee_mode = 1
+; VMEM-NEXT:     enable_wgp_mode = 0
+; VMEM-NEXT:     enable_mem_ordered = 0
+; VMEM-NEXT:     enable_fwd_progress = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; VMEM-NEXT:     user_sgpr_count = 6
+; VMEM-NEXT:     enable_trap_handler = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_id_x = 1
+; VMEM-NEXT:     enable_sgpr_workgroup_id_y = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_id_z = 0
+; VMEM-NEXT:     enable_sgpr_workgroup_info = 0
+; VMEM-NEXT:     enable_vgpr_workitem_id = 0
+; VMEM-NEXT:     enable_exception_msb = 0
+; VMEM-NEXT:     granulated_lds_size = 0
+; VMEM-NEXT:     enable_exception = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_buffer = 1
+; VMEM-NEXT:     enable_sgpr_dispatch_ptr = 0
+; VMEM-NEXT:     enable_sgpr_queue_ptr = 0
+; VMEM-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; VMEM-NEXT:     enable_sgpr_dispatch_id = 0
+; VMEM-NEXT:     enable_sgpr_flat_scratch_init = 0
+; VMEM-NEXT:     enable_sgpr_private_segment_size = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; VMEM-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; VMEM-NEXT:     enable_wavefront_size32 = 0
+; VMEM-NEXT:     enable_ordered_append_gds = 0
+; VMEM-NEXT:     private_element_size = 1
+; VMEM-NEXT:     is_ptr64 = 1
+; VMEM-NEXT:     is_dynamic_callstack = 0
+; VMEM-NEXT:     is_debug_enabled = 0
+; VMEM-NEXT:     is_xnack_enabled = 0
+; VMEM-NEXT:     workitem_private_segment_byte_size = 40
+; VMEM-NEXT:     workgroup_group_segment_byte_size = 0
+; VMEM-NEXT:     gds_segment_byte_size = 0
+; VMEM-NEXT:     kernarg_segment_byte_size = 8
+; VMEM-NEXT:     workgroup_fbarrier_count = 0
+; VMEM-NEXT:     wavefront_sgpr_count = 14
+; VMEM-NEXT:     workitem_vgpr_count = 5
+; VMEM-NEXT:     reserved_vgpr_first = 0
+; VMEM-NEXT:     reserved_vgpr_count = 0
+; VMEM-NEXT:     reserved_sgpr_first = 0
+; VMEM-NEXT:     reserved_sgpr_count = 0
+; VMEM-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; VMEM-NEXT:     debug_private_segment_buffer_sgpr = 0
+; VMEM-NEXT:     kernarg_segment_alignment = 4
+; VMEM-NEXT:     group_segment_alignment = 4
+; VMEM-NEXT:     private_segment_alignment = 4
+; VMEM-NEXT:     wavefront_size = 6
+; VMEM-NEXT:     call_convention = -1
+; VMEM-NEXT:     runtime_loader_kernel_symbol = 0
+; VMEM-NEXT:    .end_amd_kernel_code_t
+; VMEM-NEXT:  ; %bb.0: ; %entry
+; VMEM-NEXT:    s_add_u32 s0, s0, s7
+; VMEM-NEXT:    s_addc_u32 s1, s1, 0
+; VMEM-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VMEM-NEXT:    s_mov_b64 s[10:11], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    v_writelane_b32 v2, s4, 0
+; VMEM-NEXT:    v_writelane_b32 v2, s5, 1
+; VMEM-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[10:11]
+; VMEM-NEXT:    v_mov_b32_e32 v1, 0
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    s_waitcnt lgkmcnt(0)
+; VMEM-NEXT:    buffer_store_dword v1, off, s[0:3], 0 offset:32 ; 4-byte Folded Spill
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    s_mov_b32 s4, 0
+; VMEM-NEXT:    v_cmp_ne_u32_e64 s[4:5], v0, s4
+; VMEM-NEXT:    ; implicit-def: $sgpr6
+; VMEM-NEXT:    v_mov_b32_e32 v0, s6
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:28 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_mov_b64 s[6:7], exec
+; VMEM-NEXT:    s_and_b64 s[4:5], s[6:7], s[4:5]
+; VMEM-NEXT:    s_xor_b64 s[6:7], s[4:5], s[6:7]
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    v_writelane_b32 v0, s6, 0
+; VMEM-NEXT:    v_writelane_b32 v0, s7, 1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_mov_b64 exec, s[4:5]
+; VMEM-NEXT:    s_cbranch_execz .LBB2_1
+; VMEM-NEXT:    s_branch .LBB2_3
+; VMEM-NEXT:  .LBB2_1: ; %Flow
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:28 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v1, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v1, off, s[0:3], 0 offset:4 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s4, v1, 0
+; VMEM-NEXT:    v_readlane_b32 s5, v1, 1
+; VMEM-NEXT:    buffer_load_dword v1, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_or_saveexec_b64 s[4:5], s[4:5]
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:36 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_and_b64 s[4:5], exec, s[4:5]
+; VMEM-NEXT:    s_mov_b64 s[6:7], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    v_writelane_b32 v0, s4, 0
+; VMEM-NEXT:    v_writelane_b32 v0, s5, 1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:20 ; 4-byte Folded Spill
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[6:7]
+; VMEM-NEXT:    s_xor_b64 exec, exec, s[4:5]
+; VMEM-NEXT:    s_cbranch_execz .LBB2_4
+; VMEM-NEXT:  ; %bb.2: ; %if
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:32 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s4
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VMEM-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:36 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_branch .LBB2_4
+; VMEM-NEXT:  .LBB2_3: ; %else
+; VMEM-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:32 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b32 m0, -1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s4
+; VMEM-NEXT:    ds_read_b32 v1, v1
+; VMEM-NEXT:    ; implicit-def: $sgpr4
+; VMEM-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VMEM-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
+; VMEM-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:28 ; 4-byte Folded Spill
+; VMEM-NEXT:    s_branch .LBB2_1
+; VMEM-NEXT:  .LBB2_4: ; %endif
+; VMEM-NEXT:    buffer_load_dword v2, off, s[0:3], 0 offset:36 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_mov_b64 s[8:9], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v4, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v4, off, s[0:3], 0 offset:20 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s6, v4, 0
+; VMEM-NEXT:    v_readlane_b32 s7, v4, 1
+; VMEM-NEXT:    buffer_load_dword v4, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[8:9]
+; VMEM-NEXT:    s_or_b64 exec, exec, s[6:7]
+; VMEM-NEXT:    s_mov_b64 s[6:7], exec
+; VMEM-NEXT:    s_mov_b64 exec, 3
+; VMEM-NEXT:    buffer_store_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0 offset:12 ; 4-byte Folded Reload
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    v_readlane_b32 s4, v3, 0
+; VMEM-NEXT:    v_readlane_b32 s5, v3, 1
+; VMEM-NEXT:    buffer_load_dword v3, off, s[0:3], 0
+; VMEM-NEXT:    s_waitcnt vmcnt(0)
+; VMEM-NEXT:    s_mov_b64 exec, s[6:7]
+; VMEM-NEXT:    v_mov_b32_e32 v0, s4
+; VMEM-NEXT:    v_mov_b32_e32 v1, s5
+; VMEM-NEXT:    flat_store_dword v[0:1], v2
+; VMEM-NEXT:    s_endpgm
+;
+; VGPR-LABEL: divergent_if_else_endif:
+; VGPR:         .amd_kernel_code_t
+; VGPR-NEXT:     amd_code_version_major = 1
+; VGPR-NEXT:     amd_code_version_minor = 2
+; VGPR-NEXT:     amd_machine_kind = 1
+; VGPR-NEXT:     amd_machine_version_major = 7
+; VGPR-NEXT:     amd_machine_version_minor = 0
+; VGPR-NEXT:     amd_machine_version_stepping = 0
+; VGPR-NEXT:     kernel_code_entry_byte_offset = 256
+; VGPR-NEXT:     kernel_code_prefetch_byte_size = 0
+; VGPR-NEXT:     granulated_workitem_vgpr_count = 0
+; VGPR-NEXT:     granulated_wavefront_sgpr_count = 1
+; VGPR-NEXT:     priority = 0
+; VGPR-NEXT:     float_mode = 240
+; VGPR-NEXT:     priv = 0
+; VGPR-NEXT:     enable_dx10_clamp = 1
+; VGPR-NEXT:     debug_mode = 0
+; VGPR-NEXT:     enable_ieee_mode = 1
+; VGPR-NEXT:     enable_wgp_mode = 0
+; VGPR-NEXT:     enable_mem_ordered = 0
+; VGPR-NEXT:     enable_fwd_progress = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; VGPR-NEXT:     user_sgpr_count = 6
+; VGPR-NEXT:     enable_trap_handler = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_id_x = 1
+; VGPR-NEXT:     enable_sgpr_workgroup_id_y = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_id_z = 0
+; VGPR-NEXT:     enable_sgpr_workgroup_info = 0
+; VGPR-NEXT:     enable_vgpr_workitem_id = 0
+; VGPR-NEXT:     enable_exception_msb = 0
+; VGPR-NEXT:     granulated_lds_size = 0
+; VGPR-NEXT:     enable_exception = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_buffer = 1
+; VGPR-NEXT:     enable_sgpr_dispatch_ptr = 0
+; VGPR-NEXT:     enable_sgpr_queue_ptr = 0
+; VGPR-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; VGPR-NEXT:     enable_sgpr_dispatch_id = 0
+; VGPR-NEXT:     enable_sgpr_flat_scratch_init = 0
+; VGPR-NEXT:     enable_sgpr_private_segment_size = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; VGPR-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; VGPR-NEXT:     enable_wavefront_size32 = 0
+; VGPR-NEXT:     enable_ordered_append_gds = 0
+; VGPR-NEXT:     private_element_size = 1
+; VGPR-NEXT:     is_ptr64 = 1
+; VGPR-NEXT:     is_dynamic_callstack = 0
+; VGPR-NEXT:     is_debug_enabled = 0
+; VGPR-NEXT:     is_xnack_enabled = 0
+; VGPR-NEXT:     workitem_private_segment_byte_size = 16
+; VGPR-NEXT:     workgroup_group_segment_byte_size = 0
+; VGPR-NEXT:     gds_segment_byte_size = 0
+; VGPR-NEXT:     kernarg_segment_byte_size = 8
+; VGPR-NEXT:     workgroup_fbarrier_count = 0
+; VGPR-NEXT:     wavefront_sgpr_count = 10
+; VGPR-NEXT:     workitem_vgpr_count = 4
+; VGPR-NEXT:     reserved_vgpr_first = 0
+; VGPR-NEXT:     reserved_vgpr_count = 0
+; VGPR-NEXT:     reserved_sgpr_first = 0
+; VGPR-NEXT:     reserved_sgpr_count = 0
+; VGPR-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; VGPR-NEXT:     debug_private_segment_buffer_sgpr = 0
+; VGPR-NEXT:     kernarg_segment_alignment = 4
+; VGPR-NEXT:     group_segment_alignment = 4
+; VGPR-NEXT:     private_segment_alignment = 4
+; VGPR-NEXT:     wavefront_size = 6
+; VGPR-NEXT:     call_convention = -1
+; VGPR-NEXT:     runtime_loader_kernel_symbol = 0
+; VGPR-NEXT:    .end_amd_kernel_code_t
+; VGPR-NEXT:  ; %bb.0: ; %entry
+; VGPR-NEXT:    s_add_u32 s0, s0, s7
+; VGPR-NEXT:    s_addc_u32 s1, s1, 0
+; VGPR-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x0
+; VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 0
+; VGPR-NEXT:    v_writelane_b32 v1, s5, 1
+; VGPR-NEXT:    v_mov_b32_e32 v2, 0
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; VGPR-NEXT:    buffer_store_dword v2, off, s[0:3], 0 offset:8 ; 4-byte Folded Spill
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    s_mov_b32 s4, 0
+; VGPR-NEXT:    v_cmp_ne_u32_e64 s[4:5], v0, s4
+; VGPR-NEXT:    ; implicit-def: $sgpr6
+; VGPR-NEXT:    v_mov_b32_e32 v0, s6
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_mov_b64 s[6:7], exec
+; VGPR-NEXT:    s_and_b64 s[4:5], s[6:7], s[4:5]
+; VGPR-NEXT:    s_xor_b64 s[6:7], s[4:5], s[6:7]
+; VGPR-NEXT:    v_writelane_b32 v1, s6, 2
+; VGPR-NEXT:    v_writelane_b32 v1, s7, 3
+; VGPR-NEXT:    s_mov_b64 exec, s[4:5]
+; VGPR-NEXT:    s_cbranch_execz .LBB2_1
+; VGPR-NEXT:    s_branch .LBB2_3
+; VGPR-NEXT:  .LBB2_1: ; %Flow
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Reload
+; VGPR-NEXT:    v_readlane_b32 s4, v1, 2
+; VGPR-NEXT:    v_readlane_b32 s5, v1, 3
+; VGPR-NEXT:    s_or_saveexec_b64 s[4:5], s[4:5]
+; VGPR-NEXT:    s_waitcnt vmcnt(0)
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_and_b64 s[4:5], exec, s[4:5]
+; VGPR-NEXT:    v_writelane_b32 v1, s4, 4
+; VGPR-NEXT:    v_writelane_b32 v1, s5, 5
+; VGPR-NEXT:    s_xor_b64 exec, exec, s[4:5]
+; VGPR-NEXT:    s_cbranch_execz .LBB2_4
+; VGPR-NEXT:  ; %bb.2: ; %if
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:8 ; 4-byte Folded Reload
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VGPR-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:12 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_branch .LBB2_4
+; VGPR-NEXT:  .LBB2_3: ; %else
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:8 ; 4-byte Folded Reload
+; VGPR-NEXT:    s_mov_b32 m0, -1
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    ds_read_b32 v2, v2
+; VGPR-NEXT:    ; implicit-def: $sgpr4
+; VGPR-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; VGPR-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
+; VGPR-NEXT:    buffer_store_dword v0, off, s[0:3], 0 offset:4 ; 4-byte Folded Spill
+; VGPR-NEXT:    s_branch .LBB2_1
+; VGPR-NEXT:  .LBB2_4: ; %endif
+; VGPR-NEXT:    buffer_load_dword v0, off, s[0:3], 0 offset:12 ; 4-byte Folded Reload
+; VGPR-NEXT:    v_readlane_b32 s6, v1, 4
+; VGPR-NEXT:    v_readlane_b32 s7, v1, 5
+; VGPR-NEXT:    s_or_b64 exec, exec, s[6:7]
+; VGPR-NEXT:    v_readlane_b32 s4, v1, 0
+; VGPR-NEXT:    v_readlane_b32 s5, v1, 1
+; VGPR-NEXT:    v_mov_b32_e32 v2, s4
+; VGPR-NEXT:    v_mov_b32_e32 v3, s5
+; VGPR-NEXT:    s_waitcnt vmcnt(0)
+; VGPR-NEXT:    flat_store_dword v[2:3], v0
+; VGPR-NEXT:    s_endpgm
 entry:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %load0 = load volatile i32, ptr addrspace(3) null
   %cmp0 = icmp eq i32 %tid, 0
   br i1 %cmp0, label %if, label %else
 
 if:
   %load1 = load volatile i32, ptr addrspace(3) undef

llvm/test/CodeGen/AMDGPU/ds_read2.ll

 ; CI:       ; %bb.0:
 ; CI-NEXT:    v_mov_b32_e32 v0, 0
 ; CI-NEXT:    s_mov_b32 m0, -1
 ; CI-NEXT:    ds_read_b64 v[0:1], v0
 ; CI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
 ; CI-NEXT:    s_mov_b32 s3, 0xf000
 ; CI-NEXT:    s_mov_b32 s2, -1
 ; CI-NEXT:    s_waitcnt lgkmcnt(0)
-; CI-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; CI-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; CI-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; CI-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: load_constant_adjacent_offsets:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX9-NEXT:    ds_read_b64 v[0:1], v2
 ; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0

llvm/test/CodeGen/AMDGPU/greedy-reverse-local-assignment.ll

 ; FORWARDXNACK-NEXT:    v_mov_b32_e32 v0, v6
 ; FORWARDXNACK-NEXT:    s_waitcnt vmcnt(0)
 ; FORWARDXNACK-NEXT:    v_mov_b32_e32 v1, v4
 ; FORWARDXNACK-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; REVERSEXNACK-LABEL: shuffle_v4f16_234u:
 ; REVERSEXNACK:       ; %bb.0:
 ; REVERSEXNACK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; REVERSEXNACK-NEXT:    v_mov_b32_e32 v6, v1
-; REVERSEXNACK-NEXT:    v_mov_b32_e32 v5, v0
-; REVERSEXNACK-NEXT:    v_mov_b32_e32 v4, v3
-; REVERSEXNACK-NEXT:    v_mov_b32_e32 v3, v2
-; REVERSEXNACK-NEXT:    global_load_dword v0, v[5:6], off offset:4
-; REVERSEXNACK-NEXT:    global_load_dwordx2 v[1:2], v[3:4], off
+; REVERSEXNACK-NEXT:    v_mov_b32_e32 v6, v3
+; REVERSEXNACK-NEXT:    v_mov_b32_e32 v4, v1
+; REVERSEXNACK-NEXT:    v_mov_b32_e32 v3, v0
+; REVERSEXNACK-NEXT:    v_mov_b32_e32 v5, v2
+; REVERSEXNACK-NEXT:    global_load_dword v0, v[3:4], off offset:4
+; REVERSEXNACK-NEXT:    global_load_dwordx2 v[1:2], v[5:6], off
 ; REVERSEXNACK-NEXT:    s_waitcnt vmcnt(0)
 ; REVERSEXNACK-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; NOXNACK-LABEL: shuffle_v4f16_234u:
 ; NOXNACK:       ; %bb.0:
 ; NOXNACK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; NOXNACK-NEXT:    global_load_dword v0, v[0:1], off offset:4
 ; NOXNACK-NEXT:    global_load_dwordx2 v[1:2], v[2:3], off
 ; NOXNACK-NEXT:    s_waitcnt vmcnt(0)
 ; NOXNACK-NEXT:    s_setpc_b64 s[30:31]
   %val0 = load <4 x half>, ptr addrspace(1) %arg0
   %val1 = load <4 x half>, ptr addrspace(1) %arg1
   %shuffle = shufflevector <4 x half> %val0, <4 x half> %val1, <4 x i32> <i32 2, i32 3, i32 4, i32 undef>
   ret <4 x half> %shuffle
 }

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sbfe.ll

-; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN1 %s
+; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GCN2s %s
 
 ; GCN-LABEL: {{^}}bfe_i32_arg_arg_arg:
 ; GCN: v_bfe_i32
 define amdgpu_kernel void @bfe_i32_arg_arg_arg(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
+; GCN1-LABEL: bfe_i32_arg_arg_arg:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_mov_b32_e32 v0, s2
+; GCN1-NEXT:    v_bfe_i32 v0, v0, s3, s3
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_arg_arg_arg:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    v_mov_b32_e32 v0, s2
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, s3, s3
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 %src1, i32 %src1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_arg_arg_imm:
 ; GCN: v_bfe_i32
 define amdgpu_kernel void @bfe_i32_arg_arg_imm(ptr addrspace(1) %out, i32 %src0, i32 %src1) #0 {
+; GCN1-LABEL: bfe_i32_arg_arg_imm:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_mov_b32_e32 v1, s3
+; GCN1-NEXT:    v_bfe_i32 v0, s2, v1, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_arg_arg_imm:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    v_mov_b32_e32 v1, 0x7b
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    v_mov_b32_e32 v0, s3
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    v_bfe_i32 v0, s2, v0, v1
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 %src1, i32 123)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_arg_imm_arg:
 ; GCN: v_bfe_i32
 define amdgpu_kernel void @bfe_i32_arg_imm_arg(ptr addrspace(1) %out, i32 %src0, i32 %src2) #0 {
+; GCN1-LABEL: bfe_i32_arg_imm_arg:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_mov_b32_e32 v1, s3
+; GCN1-NEXT:    v_bfe_i32 v0, s2, v0, v1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_arg_imm_arg:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    v_bfe_i32 v0, s2, v0, v1
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 123, i32 %src2)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_imm_arg_arg:
 ; GCN: v_bfe_i32
 define amdgpu_kernel void @bfe_i32_imm_arg_arg(ptr addrspace(1) %out, i32 %src1, i32 %src2) #0 {
+; GCN1-LABEL: bfe_i32_imm_arg_arg:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_movk_i32 s8, 0x7b
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_mov_b32_e32 v0, s2
+; GCN1-NEXT:    v_mov_b32_e32 v1, s3
+; GCN1-NEXT:    v_bfe_i32 v0, s8, v0, v1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_imm_arg_arg:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_movk_i32 s8, 0x7b
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    v_mov_b32_e32 v0, s2
+; GCN2s-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    v_bfe_i32 v0, s8, v0, v1
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 123, i32 %src1, i32 %src2)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}v_bfe_print_arg:
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 2, 8
 define amdgpu_kernel void @v_bfe_print_arg(ptr addrspace(1) %out, ptr addrspace(1) %src0) #0 {
+; GCN1-LABEL: v_bfe_print_arg:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 2, 8
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: v_bfe_print_arg:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 2, 8
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i32, ptr addrspace(1) %src0, align 4
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %load, i32 2, i32 8)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_arg_0_width_reg_offset:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_arg_0_width_reg_offset(ptr addrspace(1) %out, i32 %src0, i32 %src1) #0 {
+; GCN1-LABEL: bfe_i32_arg_0_width_reg_offset:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_arg_0_width_reg_offset:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_u32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 %src1, i32 0)
   store i32 %bfe_u32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_arg_0_width_imm_offset:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_arg_0_width_imm_offset(ptr addrspace(1) %out, i32 %src0, i32 %src1) #0 {
+; GCN1-LABEL: bfe_i32_arg_0_width_imm_offset:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_arg_0_width_imm_offset:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_u32 = call i32 @llvm.amdgcn.sbfe.i32(i32 %src0, i32 8, i32 0)
   store i32 %bfe_u32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_6:
 ; GCN: v_lshlrev_b32_e32 v{{[0-9]+}}, 31, v{{[0-9]+}}
 ; GCN: v_ashrrev_i32_e32 v{{[0-9]+}}, 1, v{{[0-9]+}}
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_6(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_6:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_lshlrev_b32_e32 v0, 31, v0
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 1, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_6:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_lshlrev_b32_e32 v0, 31, v0
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 1, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = shl i32 %x, 31
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shl, i32 1, i32 31)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_7:
 ; GCN-NOT: shl
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_7(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_7:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_7:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = shl i32 %x, 31
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shl, i32 0, i32 31)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_8:
 ; GCN: buffer_load_dword
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 1
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_8(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_8:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_8:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 1
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = shl i32 %x, 31
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shl, i32 31, i32 1)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_9:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_ashrrev_i32_e32 v{{[0-9]+}}, 31, v{{[0-9]+}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_9(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_9:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 31, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_9:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 31, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %x, i32 31, i32 1)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_10:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_ashrrev_i32_e32 v{{[0-9]+}}, 1, v{{[0-9]+}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_10(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_10:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 1, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_10:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 1, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %x, i32 1, i32 31)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_11:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_ashrrev_i32_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_11(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_11:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 8, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_11:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 8, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %x, i32 8, i32 24)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_12:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_ashrrev_i32_e32 v{{[0-9]+}}, 24, v{{[0-9]+}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_12(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_12:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 24, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_12:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 24, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %x, i32 24, i32 8)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_13:
 ; GCN: v_ashrrev_i32_e32 {{v[0-9]+}}, 31, {{v[0-9]+}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_13(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_13:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 31, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_13:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 31, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = ashr i32 %x, 31
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shl, i32 31, i32 1)
   store i32 %bfe, ptr addrspace(1) %out, align 4 ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_test_14:
 ; GCN-NOT: lshr
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_test_14(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_i32_test_14:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_test_14:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = lshr i32 %x, 31
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shl, i32 31, i32 1)
   store i32 %bfe, ptr addrspace(1) %out, align 4 ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_0:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_0(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_0:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_0:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 0, i32 0, i32 0)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_1:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_1(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_1:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_1:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 12334, i32 0, i32 0)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_2:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_2(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_2:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_2:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 0, i32 0, i32 1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_3:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], -1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_3(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_3:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_3:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 1, i32 0, i32 1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_4:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], -1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_4(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_4:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_4:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 4294967295, i32 0, i32 1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_5:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], -1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_5(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_5:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_5:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 128, i32 7, i32 1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_6:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0xffffff80
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_6(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_6:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0xffffff80
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_6:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0xffffff80
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 128, i32 0, i32 8)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_7:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0x7f
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_7(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_7:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0x7f
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_7:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0x7f
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 127, i32 0, i32 8)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_8:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_8(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_8:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_8:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 127, i32 6, i32 8)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_9:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_9(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_9:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_9:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 65536, i32 16, i32 8)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_10:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_10(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_10:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_10:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 65535, i32 16, i32 16)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_11:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], -6
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_11(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_11:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, -6
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_11:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, -6
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 160, i32 4, i32 4)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_12:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_12(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_12:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_12:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 160, i32 31, i32 1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_13:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_13(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_13:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_13:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 131070, i32 16, i32 16)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_14:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 40
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_14(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_14:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 40
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_14:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 40
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 160, i32 2, i32 30)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_15:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 10
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_15(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_15:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 10
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_15:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 10
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 160, i32 4, i32 28)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_16:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], -1
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_16(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_16:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_16:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 4294967295, i32 1, i32 7)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_17:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0x7f
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_17(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_17:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0x7f
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_17:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0x7f
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 255, i32 1, i32 31)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_i32_constant_fold_test_18:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: v_mov_b32_e32 [[VREG:v[0-9]+]], 0
 ; GCN: buffer_store_dword [[VREG]],
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_i32_constant_fold_test_18(ptr addrspace(1) %out) #0 {
+; GCN1-LABEL: bfe_i32_constant_fold_test_18:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_i32_constant_fold_test_18:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %bfe_i32 = call i32 @llvm.amdgcn.sbfe.i32(i32 255, i32 31, i32 1)
   store i32 %bfe_i32, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_sext_in_reg_i24:
 ; GCN: buffer_load_dword [[LOAD:v[0-9]+]],
 ; GCN-NOT: v_lshl
 ; GCN-NOT: v_ashr
 ; GCN: v_bfe_i32 [[BFE:v[0-9]+]], [[LOAD]], 0, 24
 ; GCN: buffer_store_dword [[BFE]],
 define amdgpu_kernel void @bfe_sext_in_reg_i24(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: bfe_sext_in_reg_i24:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 24
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_sext_in_reg_i24:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 24
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %x, i32 0, i32 24)
   %shl = shl i32 %bfe, 8
   %ashr = ashr i32 %shl, 8
   store i32 %ashr, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: @simplify_demanded_bfe_sdiv
 ; GCN: buffer_load_dword [[LOAD:v[0-9]+]]
 ; GCN: v_bfe_i32 [[BFE:v[0-9]+]], [[LOAD]], 1, 16
 ; GCN: v_lshrrev_b32_e32 [[TMP0:v[0-9]+]], 31, [[BFE]]
 ; GCN: v_add_{{[iu]}}32_e32 [[TMP1:v[0-9]+]], vcc, [[TMP0]], [[BFE]]
 ; GCN: v_ashrrev_i32_e32 [[TMP2:v[0-9]+]], 1, [[TMP1]]
 ; GCN: buffer_store_dword [[TMP2]]
 define amdgpu_kernel void @simplify_demanded_bfe_sdiv(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: simplify_demanded_bfe_sdiv:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 1, 16
+; GCN1-NEXT:    v_lshrrev_b32_e32 v1, 31, v0
+; GCN1-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GCN1-NEXT:    v_ashrrev_i32_e32 v0, 1, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: simplify_demanded_bfe_sdiv:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 1, 16
+; GCN2s-NEXT:    v_lshrrev_b32_e32 v1, 31, v0
+; GCN2s-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; GCN2s-NEXT:    v_ashrrev_i32_e32 v0, 1, v0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %src = load i32, ptr addrspace(1) %in, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %src, i32 1, i32 16)
   %div = sdiv i32 %bfe, 2
   store i32 %div, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_0_width:
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_0_width(ptr addrspace(1) %out, ptr addrspace(1) %ptr) #0 {
+; GCN1-LABEL: bfe_0_width:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_0_width:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i32, ptr addrspace(1) %ptr, align 4
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %load, i32 8, i32 0)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_8_bfe_8:
 ; GCN: v_bfe_i32
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_8_bfe_8(ptr addrspace(1) %out, ptr addrspace(1) %ptr) #0 {
+; GCN1-LABEL: bfe_8_bfe_8:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 8
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_8_bfe_8:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 8
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i32, ptr addrspace(1) %ptr, align 4
   %bfe0 = call i32 @llvm.amdgcn.sbfe.i32(i32 %load, i32 0, i32 8)
   %bfe1 = call i32 @llvm.amdgcn.sbfe.i32(i32 %bfe0, i32 0, i32 8)
   store i32 %bfe1, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}bfe_8_bfe_16:
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 8
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_8_bfe_16(ptr addrspace(1) %out, ptr addrspace(1) %ptr) #0 {
+; GCN1-LABEL: bfe_8_bfe_16:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 8
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_8_bfe_16:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 8
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i32, ptr addrspace(1) %ptr, align 4
   %bfe0 = call i32 @llvm.amdgcn.sbfe.i32(i32 %load, i32 0, i32 8)
   %bfe1 = call i32 @llvm.amdgcn.sbfe.i32(i32 %bfe0, i32 0, i32 16)
   store i32 %bfe1, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; This really should be folded into 1
 ; GCN-LABEL: {{^}}bfe_16_bfe_8:
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 8
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @bfe_16_bfe_8(ptr addrspace(1) %out, ptr addrspace(1) %ptr) #0 {
+; GCN1-LABEL: bfe_16_bfe_8:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 8
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: bfe_16_bfe_8:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 8
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i32, ptr addrspace(1) %ptr, align 4
   %bfe0 = call i32 @llvm.amdgcn.sbfe.i32(i32 %load, i32 0, i32 16)
   %bfe1 = call i32 @llvm.amdgcn.sbfe.i32(i32 %bfe0, i32 0, i32 8)
   store i32 %bfe1, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; Make sure there isn't a redundant BFE
 ; GCN-LABEL: {{^}}sext_in_reg_i8_to_i32_bfe:
 ; GCN: s_sext_i32_i8 s{{[0-9]+}}, s{{[0-9]+}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @sext_in_reg_i8_to_i32_bfe(ptr addrspace(1) %out, i32 %a, i32 %b) #0 {
+; GCN1-LABEL: sext_in_reg_i8_to_i32_bfe:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_add_i32 s2, s2, s3
+; GCN1-NEXT:    s_sext_i32_i8 s2, s2
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_mov_b32_e32 v0, s2
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sext_in_reg_i8_to_i32_bfe:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_add_i32 s0, s2, s3
+; GCN2s-NEXT:    s_sext_i32_i8 s0, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %c = add i32 %a, %b ; add to prevent folding into extload
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %c, i32 0, i32 8)
   %shl = shl i32 %bfe, 24
   %ashr = ashr i32 %shl, 24
   store i32 %ashr, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}sext_in_reg_i8_to_i32_bfe_wrong:
 define amdgpu_kernel void @sext_in_reg_i8_to_i32_bfe_wrong(ptr addrspace(1) %out, i32 %a, i32 %b) #0 {
+; GCN1-LABEL: sext_in_reg_i8_to_i32_bfe_wrong:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sext_in_reg_i8_to_i32_bfe_wrong:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %c = add i32 %a, %b ; add to prevent folding into extload
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %c, i32 8, i32 0)
   %shl = shl i32 %bfe, 24
   %ashr = ashr i32 %shl, 24
   store i32 %ashr, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}sextload_i8_to_i32_bfe:
 ; GCN: buffer_load_sbyte
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @sextload_i8_to_i32_bfe(ptr addrspace(1) %out, ptr addrspace(1) %ptr) #0 {
+; GCN1-LABEL: sextload_i8_to_i32_bfe:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_sbyte v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sextload_i8_to_i32_bfe:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_sbyte v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i8, ptr addrspace(1) %ptr, align 1
   %sext = sext i8 %load to i32
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %sext, i32 0, i32 8)
   %shl = shl i32 %bfe, 24
   %ashr = ashr i32 %shl, 24
   store i32 %ashr, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN: .text
 ; GCN-LABEL: {{^}}sextload_i8_to_i32_bfe_0:{{.*$}}
 ; GCN-NOT: {{[^@]}}bfe
 ; GCN: s_endpgm
 define amdgpu_kernel void @sextload_i8_to_i32_bfe_0(ptr addrspace(1) %out, ptr addrspace(1) %ptr) #0 {
+; GCN1-LABEL: sextload_i8_to_i32_bfe_0:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    v_mov_b32_e32 v0, 0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sextload_i8_to_i32_bfe_0:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s3, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s2, -1
+; GCN2s-NEXT:    v_mov_b32_e32 v0, 0
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN2s-NEXT:    s_endpgm
   %load = load i8, ptr addrspace(1) %ptr, align 1
   %sext = sext i8 %load to i32
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %sext, i32 8, i32 0)
   %shl = shl i32 %bfe, 24
   %ashr = ashr i32 %shl, 24
   store i32 %ashr, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}sext_in_reg_i1_bfe_offset_0:
 ; GCN-NOT: shr
 ; GCN-NOT: shl
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 1
 ; GCN: s_endpgm
 define amdgpu_kernel void @sext_in_reg_i1_bfe_offset_0(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: sext_in_reg_i1_bfe_offset_0:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sext_in_reg_i1_bfe_offset_0:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 1
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = shl i32 %x, 31
   %shr = ashr i32 %shl, 31
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shr, i32 0, i32 1)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}sext_in_reg_i1_bfe_offset_1:
 ; GCN: buffer_load_dword
 ; GCN-NOT: shl
 ; GCN-NOT: shr
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 1, 1
 ; GCN: s_endpgm
 define amdgpu_kernel void @sext_in_reg_i1_bfe_offset_1(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: sext_in_reg_i1_bfe_offset_1:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 1, 1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sext_in_reg_i1_bfe_offset_1:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 1, 1
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = shl i32 %x, 30
   %shr = ashr i32 %shl, 30
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shr, i32 1, i32 1)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 ; GCN-LABEL: {{^}}sext_in_reg_i2_bfe_offset_1:
 ; GCN: buffer_load_dword
 ; GCN-NOT: v_lshl
 ; GCN-NOT: v_ashr
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 0, 2
 ; GCN: v_bfe_i32 v{{[0-9]+}}, v{{[0-9]+}}, 1, 2
 ; GCN: s_endpgm
 define amdgpu_kernel void @sext_in_reg_i2_bfe_offset_1(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; GCN1-LABEL: sext_in_reg_i2_bfe_offset_1:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 0, 2
+; GCN1-NEXT:    v_bfe_i32 v0, v0, 1, 2
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2s-LABEL: sext_in_reg_i2_bfe_offset_1:
+; GCN2s:       ; %bb.0:
+; GCN2s-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2s-NEXT:    s_mov_b32 s7, 0xf000
+; GCN2s-NEXT:    s_mov_b32 s6, -1
+; GCN2s-NEXT:    s_mov_b32 s10, s6
+; GCN2s-NEXT:    s_mov_b32 s11, s7
+; GCN2s-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2s-NEXT:    s_mov_b32 s8, s2
+; GCN2s-NEXT:    s_mov_b32 s9, s3
+; GCN2s-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN2s-NEXT:    s_mov_b32 s4, s0
+; GCN2s-NEXT:    s_mov_b32 s5, s1
+; GCN2s-NEXT:    s_waitcnt vmcnt(0)
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 0, 2
+; GCN2s-NEXT:    v_bfe_i32 v0, v0, 1, 2
+; GCN2s-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN2s-NEXT:    s_endpgm
   %x = load i32, ptr addrspace(1) %in, align 4
   %shl = shl i32 %x, 30
   %shr = ashr i32 %shl, 30
   %bfe = call i32 @llvm.amdgcn.sbfe.i32(i32 %shr, i32 1, i32 2)
   store i32 %bfe, ptr addrspace(1) %out, align 4
   ret void
 }
 
 declare i32 @llvm.amdgcn.sbfe.i32(i32, i32, i32) #1
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }

llvm/test/CodeGen/AMDGPU/mubuf-legalize-operands.ll

 ; W64-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    v_mov_b32_e32 v4, v3
 ; W64-O0-NEXT:    v_mov_b32_e32 v6, v2
 ; W64-O0-NEXT:    v_mov_b32_e32 v7, v1
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr0 killed $vgpr0 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr4 killed $vgpr4 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1_vgpr2_vgpr3 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v1, v7
 ; W64-O0-NEXT:    v_mov_b32_e32 v2, v6
 ; W64-O0-NEXT:    v_mov_b32_e32 v3, v4
 ; W64-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill

 ; W64-O0-NEXT:    v_mov_b32_e32 v8, v1
 ; W64-O0-NEXT:    buffer_load_dword v1, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
 ; W64-O0-NEXT:    v_mov_b32_e32 v2, v0
 ; W64-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr5 killed $vgpr5 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr4 killed $vgpr4 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr3 killed $vgpr3 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr14 killed $vgpr14 def $vgpr14_vgpr15_vgpr16_vgpr17 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v15, v5
 ; W64-O0-NEXT:    s_waitcnt vmcnt(3)
 ; W64-O0-NEXT:    v_mov_b32_e32 v16, v4
 ; W64-O0-NEXT:    s_waitcnt vmcnt(2)
 ; W64-O0-NEXT:    v_mov_b32_e32 v17, v3
 ; W64-O0-NEXT:    buffer_store_dword v14, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v15, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v16, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v17, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr2 killed $vgpr2 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr8 killed $vgpr8 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3_vgpr4_vgpr5 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v3, v8
 ; W64-O0-NEXT:    v_mov_b32_e32 v4, v7
 ; W64-O0-NEXT:    v_mov_b32_e32 v5, v6
 ; W64-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:16 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v5, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr12 killed $vgpr12 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v2, v12
 ; W64-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:12 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr10 killed $vgpr10 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v1, v10
 ; W64-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4_sgpr5
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4_sgpr5
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4_sgpr5_sgpr6_sgpr7

 ; W64-O0-NEXT:    v_mov_b32_e32 v6, v2
 ; W64-O0-NEXT:    v_mov_b32_e32 v7, v1
 ; W64-O0-NEXT:    v_mov_b32_e32 v13, v0
 ; W64-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr0 killed $vgpr0 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr11 killed $vgpr11 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr9 killed $vgpr9 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1_vgpr2_vgpr3 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v1, v11
 ; W64-O0-NEXT:    v_mov_b32_e32 v2, v10
 ; W64-O0-NEXT:    v_mov_b32_e32 v3, v9
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr13 killed $vgpr13 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 killed $exec
-; W64-O0-NEXT:    ; kill: def $vgpr6 killed $vgpr6 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr13 killed $vgpr13 def $vgpr13_vgpr14_vgpr15_vgpr16 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v14, v7
 ; W64-O0-NEXT:    v_mov_b32_e32 v15, v6
 ; W64-O0-NEXT:    v_mov_b32_e32 v16, v5
 ; W64-O0-NEXT:    buffer_store_dword v13, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v14, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v15, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v16, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
 ; W64-O0-NEXT:    ; implicit-def: $sgpr4
-; W64-O0-NEXT:    ; kill: def $vgpr12 killed $vgpr12 killed $exec
 ; W64-O0-NEXT:    ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
 ; W64-O0-NEXT:    v_mov_b32_e32 v5, v12
 ; W64-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v5, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
 ; W64-O0-NEXT:    s_waitcnt vmcnt(0)
 ; W64-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:8 ; 4-byte Folded Spill

llvm/test/CodeGen/AMDGPU/mubuf-legalize-operands.mir

bb.0:
     ; W64-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W64-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W64-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W64-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W64-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W64-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W64-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W64-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
+    ; W64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W64-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
     ; W64-NEXT: {{  $}}
     ; W64-NEXT: .1:
     ; W64-NEXT: successors: %bb.2(0x80000000)
     ; W64-NEXT: {{  $}}
     ; W64-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W64-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W64-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; W32-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W32-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W32-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W32-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W32-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W32-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W32-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W32-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
+    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W32-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: .1:
     ; W32-NEXT: successors: %bb.2(0x80000000)
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W32-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W32-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; W64-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W64-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W64-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W64-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W64-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W64-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W64-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY3]], implicit $exec
-    ; W64-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, killed [[COPY6]], %subreg.sub1, killed [[COPY7]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
+    ; W64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W64-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
     ; W64-NEXT: {{  $}}
     ; W64-NEXT: .1:
     ; W64-NEXT: successors: %bb.2(0x80000000)
     ; W64-NEXT: {{  $}}
     ; W64-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W64-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W64-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; W32-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W32-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W32-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W32-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W32-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W32-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W32-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY3]], implicit $exec
-    ; W32-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, killed [[COPY6]], %subreg.sub1, killed [[COPY7]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
+    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W32-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: .1:
     ; W32-NEXT: successors: %bb.2(0x80000000)
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W32-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W32-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; W64-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W64-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W64-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W64-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W64-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W64-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W64-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W64-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY3]], implicit $exec
-    ; W64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, killed [[COPY8]], %subreg.sub2, [[COPY2]], %subreg.sub3
+    ; W64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W64-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
     ; W64-NEXT: {{  $}}
     ; W64-NEXT: .1:
     ; W64-NEXT: successors: %bb.2(0x80000000)
     ; W64-NEXT: {{  $}}
     ; W64-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W64-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W64-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; W32-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W32-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W32-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W32-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W32-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W32-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W32-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W32-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY3]], implicit $exec
-    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, killed [[COPY8]], %subreg.sub2, [[COPY2]], %subreg.sub3
+    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W32-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: .1:
     ; W32-NEXT: successors: %bb.2(0x80000000)
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W32-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W32-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; ADDR64-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; ADDR64-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; ADDR64-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; ADDR64-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; ADDR64-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; ADDR64-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; ADDR64-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; ADDR64-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; ADDR64-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; ADDR64-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; ADDR64-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; ADDR64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
-    ; ADDR64-NEXT: [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
+    ; ADDR64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
+    ; ADDR64-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
     ; ADDR64-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
     ; ADDR64-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 0
     ; ADDR64-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sgpr_32 = S_MOV_B32 61440
     ; ADDR64-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[S_MOV_B32_]], %subreg.sub2, [[S_MOV_B32_1]], %subreg.sub3
-    ; ADDR64-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY9]].sub0, [[COPY1]].sub0, 0, implicit $exec
-    ; ADDR64-NEXT: %17:vgpr_32, dead %20:sreg_64_xexec = V_ADDC_U32_e64 [[COPY9]].sub1, [[COPY1]].sub1, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
-    ; ADDR64-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, %17, %subreg.sub1
+    ; ADDR64-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY6]].sub0, [[COPY1]].sub0, 0, implicit $exec
+    ; ADDR64-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_64_xexec = V_ADDC_U32_e64 [[COPY6]].sub1, [[COPY1]].sub1, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
+    ; ADDR64-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1
     ; ADDR64-NEXT: [[BUFFER_LOAD_FORMAT_X_ADDR64_:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_ADDR64 [[REG_SEQUENCE2]], killed [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec
     ; ADDR64-NEXT: $sgpr30_sgpr31 = COPY [[COPY]]
     ; ADDR64-NEXT: $vgpr0 = COPY [[BUFFER_LOAD_FORMAT_X_ADDR64_]]
     ; ADDR64-NEXT: S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
     ; W32-LABEL: name: addr64
     ; W32: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $sgpr30_sgpr31
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr30_sgpr31
     ; W32-NEXT: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr4_vgpr5
     ; W32-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W32-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W32-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W32-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W32-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W32-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W32-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W32-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
-    ; W32-NEXT: [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
+    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
+    ; W32-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
     ; W32-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
     ; W32-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 0
     ; W32-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sgpr_32 = S_MOV_B32 822173696
     ; W32-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[S_MOV_B32_]], %subreg.sub2, [[S_MOV_B32_1]], %subreg.sub3
-    ; W32-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY9]].sub0, [[COPY1]].sub0, 0, implicit $exec
-    ; W32-NEXT: %17:vgpr_32, dead %20:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY9]].sub1, [[COPY1]].sub1, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
-    ; W32-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, %17, %subreg.sub1
+    ; W32-NEXT: [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[COPY6]].sub0, [[COPY1]].sub0, 0, implicit $exec
+    ; W32-NEXT: [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 [[COPY6]].sub1, [[COPY1]].sub1, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
+    ; W32-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[V_ADD_CO_U32_e64_]], %subreg.sub0, [[V_ADDC_U32_e64_]], %subreg.sub1
     ; W32-NEXT: [[BUFFER_LOAD_FORMAT_X_ADDR64_:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_ADDR64 [[REG_SEQUENCE2]], killed [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec
     ; W32-NEXT: $sgpr30_sgpr31 = COPY [[COPY]]
     ; W32-NEXT: $vgpr0 = COPY [[BUFFER_LOAD_FORMAT_X_ADDR64_]]
     ; W32-NEXT: S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
     %5:sreg_64 = COPY $sgpr30_sgpr31
     %4:vreg_64 = COPY $vgpr4_vgpr5
     %3:vgpr_32 = COPY $vgpr3
     %2:vgpr_32 = COPY $vgpr2
bb.0:
     ; ADDR64-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; ADDR64-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; ADDR64-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; ADDR64-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; ADDR64-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; ADDR64-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; ADDR64-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; ADDR64-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; ADDR64-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; ADDR64-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; ADDR64-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; ADDR64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
-    ; ADDR64-NEXT: [[COPY9:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
+    ; ADDR64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
+    ; ADDR64-NEXT: [[COPY6:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
     ; ADDR64-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64 = S_MOV_B64 0
     ; ADDR64-NEXT: [[S_MOV_B32_:%[0-9]+]]:sgpr_32 = S_MOV_B32 0
     ; ADDR64-NEXT: [[S_MOV_B32_1:%[0-9]+]]:sgpr_32 = S_MOV_B32 61440
     ; ADDR64-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[S_MOV_B64_]], %subreg.sub0_sub1, [[S_MOV_B32_]], %subreg.sub2, [[S_MOV_B32_1]], %subreg.sub3
-    ; ADDR64-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]].sub0, %subreg.sub0, [[COPY9]].sub1, %subreg.sub1
+    ; ADDR64-NEXT: [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY6]].sub0, %subreg.sub0, [[COPY6]].sub1, %subreg.sub1
     ; ADDR64-NEXT: [[BUFFER_LOAD_FORMAT_X_ADDR64_:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_ADDR64 [[REG_SEQUENCE2]], [[REG_SEQUENCE1]], 0, 0, 0, 0, implicit $exec
     ; ADDR64-NEXT: $sgpr30_sgpr31 = COPY [[COPY]]
     ; ADDR64-NEXT: $vgpr0 = COPY [[BUFFER_LOAD_FORMAT_X_ADDR64_]]
     ; ADDR64-NEXT: S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
     ; W64-NO-ADDR64-LABEL: name: offset
     ; W64-NO-ADDR64: successors: %bb.1(0x80000000)
     ; W64-NO-ADDR64-NEXT: liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $sgpr30_sgpr31
     ; W64-NO-ADDR64-NEXT: {{  $}}
     ; W64-NO-ADDR64-NEXT: [[COPY:%[0-9]+]]:sreg_64 = COPY $sgpr30_sgpr31
     ; W64-NO-ADDR64-NEXT: [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr4_vgpr5
     ; W64-NO-ADDR64-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W64-NO-ADDR64-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W64-NO-ADDR64-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W64-NO-ADDR64-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W64-NO-ADDR64-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NO-ADDR64-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NO-ADDR64-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W64-NO-ADDR64-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W64-NO-ADDR64-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W64-NO-ADDR64-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W64-NO-ADDR64-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W64-NO-ADDR64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
+    ; W64-NO-ADDR64-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W64-NO-ADDR64-NEXT: [[S_MOV_B64_:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
     ; W64-NO-ADDR64-NEXT: {{  $}}
     ; W64-NO-ADDR64-NEXT: .1:
     ; W64-NO-ADDR64-NEXT: successors: %bb.2(0x80000000)
     ; W64-NO-ADDR64-NEXT: {{  $}}
     ; W64-NO-ADDR64-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W64-NO-ADDR64-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W64-NO-ADDR64-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
bb.0:
     ; W32-NEXT: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr3
     ; W32-NEXT: [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr2
     ; W32-NEXT: [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr1
     ; W32-NEXT: [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr0
     ; W32-NEXT: [[DEF:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF1:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF2:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
     ; W32-NEXT: [[DEF3:%[0-9]+]]:sgpr_32 = IMPLICIT_DEF
-    ; W32-NEXT: [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[COPY5]], implicit $exec
-    ; W32-NEXT: [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[COPY4]], implicit $exec
-    ; W32-NEXT: [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[COPY2]], implicit $exec
-    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE killed [[COPY6]], %subreg.sub0, killed [[COPY7]], %subreg.sub1, [[COPY3]], %subreg.sub2, killed [[COPY8]], %subreg.sub3
+    ; W32-NEXT: [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY5]], %subreg.sub0, [[COPY4]], %subreg.sub1, [[COPY3]], %subreg.sub2, [[COPY2]], %subreg.sub3
     ; W32-NEXT: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: .1:
     ; W32-NEXT: successors: %bb.2(0x80000000)
     ; W32-NEXT: {{  $}}
     ; W32-NEXT: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub0, implicit $exec
     ; W32-NEXT: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[REG_SEQUENCE]].sub1, implicit $exec
     ; W32-NEXT: [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1

llvm/test/CodeGen/AMDGPU/opt-sgpr-to-vgpr-copy.mir

body:             |
   ; GCN-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]], 9, 0
   ; GCN-NEXT:   [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]], 11, 0
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:sreg_64 = COPY [[S_LOAD_DWORDX2_IMM]]
   ; GCN-NEXT:   [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
   ; GCN-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, killed [[V_MOV_B32_e32_]], %subreg.sub1
   ; GCN-NEXT:   [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 [[S_LOAD_DWORDX2_IMM1]].sub0, [[REG_SEQUENCE]].sub0, implicit-def $vcc, implicit $exec
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[S_LOAD_DWORDX2_IMM1]].sub1
   ; GCN-NEXT:   [[V_ADDC_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADDC_U32_e32 0, [[COPY3]], implicit-def $vcc, implicit $vcc, implicit $exec
-  ; GCN-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY killed [[V_ADD_CO_U32_e32_]], implicit $exec
-  ; GCN-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[COPY4]], %subreg.sub0, killed [[V_ADDC_U32_e32_]], %subreg.sub1
+  ; GCN-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[V_ADD_CO_U32_e32_]], %subreg.sub0, killed [[V_ADDC_U32_e32_]], %subreg.sub1
   ; GCN-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 0
   ; GCN-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 1048576
   ; GCN-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64 = REG_SEQUENCE killed [[S_MOV_B32_1]], %subreg.sub0, killed [[S_MOV_B32_]], %subreg.sub1
   ; GCN-NEXT:   [[V_CMP_LT_U64_e64_:%[0-9]+]]:sreg_64 = V_CMP_LT_U64_e64 killed [[REG_SEQUENCE1]], [[REG_SEQUENCE2]], implicit $exec
   ; GCN-NEXT:   [[SI_IF:%[0-9]+]]:sreg_64 = SI_IF killed [[V_CMP_LT_U64_e64_]], %bb.2, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; GCN-NEXT:   S_BRANCH %bb.1
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.1.bb1:
body:             |
   ; GCN-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]], 9, 0
   ; GCN-NEXT:   [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[COPY]], 11, 0
   ; GCN-NEXT:   [[COPY2:%[0-9]+]]:sreg_64 = COPY [[S_LOAD_DWORDX2_IMM]]
   ; GCN-NEXT:   [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec
   ; GCN-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY1]], %subreg.sub0, killed [[V_MOV_B32_e32_]], %subreg.sub1
   ; GCN-NEXT:   [[V_ADD_CO_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADD_CO_U32_e32 [[S_LOAD_DWORDX2_IMM1]].sub0, [[REG_SEQUENCE]].sub0, implicit-def $vcc, implicit $exec
   ; GCN-NEXT:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[S_LOAD_DWORDX2_IMM1]].sub1
   ; GCN-NEXT:   [[V_ADDC_U32_e32_:%[0-9]+]]:vgpr_32 = V_ADDC_U32_e32 0, [[COPY3]], implicit-def $vcc, implicit $vcc, implicit $exec
-  ; GCN-NEXT:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY killed [[V_ADD_CO_U32_e32_]], implicit $exec
-  ; GCN-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[COPY4]], %subreg.sub0, killed [[V_ADDC_U32_e32_]], %subreg.sub1
+  ; GCN-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[V_ADD_CO_U32_e32_]], %subreg.sub0, killed [[V_ADDC_U32_e32_]], %subreg.sub1
   ; GCN-NEXT:   [[V_CMP_LT_U32_e64_:%[0-9]+]]:sreg_64 = V_CMP_LT_U32_e64 killed [[REG_SEQUENCE1]].sub0, 12, implicit $exec
   ; GCN-NEXT:   [[SI_IF:%[0-9]+]]:sreg_64 = SI_IF killed [[V_CMP_LT_U32_e64_]], %bb.2, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; GCN-NEXT:   S_BRANCH %bb.1
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.1.bb1:
   ; GCN-NEXT:   successors: %bb.2(0x80000000)
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   [[V_LSHL_B64_e64_:%[0-9]+]]:vreg_64 = V_LSHL_B64_e64 [[REG_SEQUENCE]], 2, implicit $exec

llvm/test/CodeGen/AMDGPU/sdiv.ll

 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_mov_b32 s8, s2
 ; GCN-NEXT:    s_mov_b32 s9, s3
 ; GCN-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
 ; GCN-NEXT:    s_mov_b32 s4, s0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_ashrrev_i32_e32 v2, 31, v1
-; GCN-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
+; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v2
 ; GCN-NEXT:    v_xor_b32_e32 v1, v1, v2
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v3, v1
 ; GCN-NEXT:    v_sub_i32_e32 v4, vcc, 0, v1
 ; GCN-NEXT:    v_ashrrev_i32_e32 v5, 31, v0
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v3, v3
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v5, v0
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v5
 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v5
 ; GCN-NEXT:    v_mul_f32_e32 v3, 0x4f7ffffe, v3
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v3, v3
 ; GCN-NEXT:    v_xor_b32_e32 v2, v5, v2
 ; GCN-NEXT:    v_mul_lo_u32 v4, v4, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v3, v4
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v3
 ; GCN-NEXT:    v_mul_lo_u32 v4, v3, v1
 ; GCN-NEXT:    v_add_i32_e32 v5, vcc, 1, v3
 ; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
-; GCN-NEXT:    v_subrev_i32_e32 v4, vcc, v1, v0
+; GCN-NEXT:    v_sub_i32_e32 v4, vcc, v0, v1
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, v3, v5, vcc
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, 1, v3
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v3, v4, vcc
 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v2
-; GCN-NEXT:    v_subrev_i32_e32 v0, vcc, v2, v0
+; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: sdiv_i32:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000
 ; TONGA-NEXT:    s_mov_b32 s6, -1
 ; TONGA-NEXT:    s_mov_b32 s10, s6
 ; TONGA-NEXT:    s_mov_b32 s11, s7
 ; TONGA-NEXT:    s_waitcnt lgkmcnt(0)
 ; TONGA-NEXT:    s_mov_b32 s8, s2
 ; TONGA-NEXT:    s_mov_b32 s9, s3
 ; TONGA-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
 ; TONGA-NEXT:    s_mov_b32 s4, s0
 ; TONGA-NEXT:    s_mov_b32 s5, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(0)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v2, 31, v1
-; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v2, v1
+; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v1, v2
 ; TONGA-NEXT:    v_xor_b32_e32 v1, v1, v2
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v3, v1
 ; TONGA-NEXT:    v_sub_u32_e32 v4, vcc, 0, v1
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v5, 31, v0
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v3, v3
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v5, v0
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v5
 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v5
 ; TONGA-NEXT:    v_mul_f32_e32 v3, 0x4f7ffffe, v3
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v3, v3
 ; TONGA-NEXT:    v_xor_b32_e32 v2, v5, v2
 ; TONGA-NEXT:    v_mul_lo_u32 v4, v4, v3
 ; TONGA-NEXT:    v_mul_hi_u32 v4, v3, v4
 ; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v3, v4
 ; TONGA-NEXT:    v_mul_hi_u32 v3, v0, v3
 ; TONGA-NEXT:    v_mul_lo_u32 v4, v3, v1
 ; TONGA-NEXT:    v_add_u32_e32 v5, vcc, 1, v3
 ; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v0, v4
-; TONGA-NEXT:    v_subrev_u32_e32 v4, vcc, v1, v0
+; TONGA-NEXT:    v_sub_u32_e32 v4, vcc, v0, v1
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; TONGA-NEXT:    v_cndmask_b32_e32 v3, v3, v5, vcc
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, v0, v4, vcc
 ; TONGA-NEXT:    v_add_u32_e32 v4, vcc, 1, v3
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, v3, v4, vcc
 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v2
-; TONGA-NEXT:    v_subrev_u32_e32 v0, vcc, v2, v0
+; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v0, v2
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s6, -1

 ; GCN-NEXT:    s_mov_b32 s8, s2
 ; GCN-NEXT:    s_mov_b32 s9, s3
 ; GCN-NEXT:    buffer_load_dword v0, off, s[8:11], 0
 ; GCN-NEXT:    s_mov_b32 s4, s0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    v_lshrrev_b32_e32 v1, 30, v1
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GCN-NEXT:    v_ashrrev_i32_e32 v0, 2, v0
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: sdiv_i32_4:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000
 ; TONGA-NEXT:    s_mov_b32 s6, -1
 ; TONGA-NEXT:    s_mov_b32 s10, s6
 ; TONGA-NEXT:    s_mov_b32 s11, s7
 ; TONGA-NEXT:    s_waitcnt lgkmcnt(0)
 ; TONGA-NEXT:    s_mov_b32 s8, s2
 ; TONGA-NEXT:    s_mov_b32 s9, s3
 ; TONGA-NEXT:    buffer_load_dword v0, off, s[8:11], 0
 ; TONGA-NEXT:    s_mov_b32 s4, s0
 ; TONGA-NEXT:    s_mov_b32 s5, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(0)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v1, 30, v1
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v0, 2, v0
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_i32_4:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000

 ; GCN-NEXT:    s_mov_b32 s2, 0x98a1930b
 ; GCN-NEXT:    s_mov_b32 s4, s0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_mul_hi_i32 v1, v0, s2
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GCN-NEXT:    v_lshrrev_b32_e32 v1, 31, v0
 ; GCN-NEXT:    v_ashrrev_i32_e32 v0, 11, v0
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: slow_sdiv_i32_3435:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000
 ; TONGA-NEXT:    s_mov_b32 s6, -1
 ; TONGA-NEXT:    s_mov_b32 s10, s6
 ; TONGA-NEXT:    s_mov_b32 s11, s7
 ; TONGA-NEXT:    s_waitcnt lgkmcnt(0)
 ; TONGA-NEXT:    s_mov_b32 s8, s2
 ; TONGA-NEXT:    s_mov_b32 s9, s3
 ; TONGA-NEXT:    buffer_load_dword v0, off, s[8:11], 0
 ; TONGA-NEXT:    s_mov_b32 s2, 0x98a1930b
 ; TONGA-NEXT:    s_mov_b32 s4, s0
 ; TONGA-NEXT:    s_mov_b32 s5, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(0)
 ; TONGA-NEXT:    v_mul_hi_i32 v1, v0, s2
 ; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v1, 31, v0
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v0, 11, v0
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: slow_sdiv_i32_3435:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s6, -1

 ; GCN-NEXT:    s_mov_b32 s8, s2
 ; GCN-NEXT:    s_mov_b32 s9, s3
 ; GCN-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; GCN-NEXT:    s_mov_b32 s4, s0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_ashrrev_i32_e32 v5, 31, v2
 ; GCN-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
-; GCN-NEXT:    v_add_i32_e32 v3, vcc, v7, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
+; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
 ; GCN-NEXT:    v_ashrrev_i32_e32 v4, 31, v0
 ; GCN-NEXT:    v_ashrrev_i32_e32 v6, 31, v1
 ; GCN-NEXT:    v_xor_b32_e32 v2, v2, v5
 ; GCN-NEXT:    v_xor_b32_e32 v3, v3, v7
 ; GCN-NEXT:    v_xor_b32_e32 v8, v4, v5
 ; GCN-NEXT:    v_xor_b32_e32 v9, v6, v7
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v5, v2
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v7, v3
 ; GCN-NEXT:    v_sub_i32_e32 v10, vcc, 0, v2
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v5, v5
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v7, v7
 ; GCN-NEXT:    v_sub_i32_e32 v11, vcc, 0, v3
 ; GCN-NEXT:    v_mul_f32_e32 v5, 0x4f7ffffe, v5
 ; GCN-NEXT:    v_mul_f32_e32 v7, 0x4f7ffffe, v7
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v5, v5
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v4, v0
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v10, v10, v5
 ; GCN-NEXT:    v_mul_lo_u32 v11, v11, v7
-; GCN-NEXT:    v_add_i32_e32 v1, vcc, v6, v1
+; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v6
 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v4
 ; GCN-NEXT:    v_mul_hi_u32 v4, v5, v10
 ; GCN-NEXT:    v_xor_b32_e32 v1, v1, v6
 ; GCN-NEXT:    v_mul_hi_u32 v6, v7, v11
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
-; GCN-NEXT:    v_add_i32_e32 v5, vcc, v6, v7
+; GCN-NEXT:    v_add_i32_e32 v5, vcc, v7, v6
 ; GCN-NEXT:    v_mul_hi_u32 v4, v0, v4
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v5
 ; GCN-NEXT:    v_mul_lo_u32 v6, v4, v2
 ; GCN-NEXT:    v_mul_lo_u32 v10, v5, v3
 ; GCN-NEXT:    v_add_i32_e32 v7, vcc, 1, v4
 ; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v0, v6
 ; GCN-NEXT:    v_sub_i32_e32 v1, vcc, v1, v10
 ; GCN-NEXT:    v_add_i32_e32 v11, vcc, 1, v5
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v2
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v3
-; GCN-NEXT:    v_subrev_i32_e32 v6, vcc, v2, v0
+; GCN-NEXT:    v_sub_i32_e32 v6, vcc, v0, v2
 ; GCN-NEXT:    v_cndmask_b32_e64 v4, v4, v7, s[0:1]
-; GCN-NEXT:    v_subrev_i32_e32 v7, vcc, v3, v1
+; GCN-NEXT:    v_sub_i32_e32 v7, vcc, v1, v3
 ; GCN-NEXT:    v_cndmask_b32_e64 v5, v5, v11, s[2:3]
 ; GCN-NEXT:    v_cndmask_b32_e64 v0, v0, v6, s[0:1]
 ; GCN-NEXT:    v_add_i32_e32 v6, vcc, 1, v4
 ; GCN-NEXT:    v_cndmask_b32_e64 v1, v1, v7, s[2:3]
 ; GCN-NEXT:    v_add_i32_e32 v7, vcc, 1, v5
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v4, v6, vcc
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, v5, v7, vcc
 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v8
 ; GCN-NEXT:    v_xor_b32_e32 v1, v1, v9
-; GCN-NEXT:    v_subrev_i32_e32 v0, vcc, v8, v0
-; GCN-NEXT:    v_subrev_i32_e32 v1, vcc, v9, v1
+; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v0, v8
+; GCN-NEXT:    v_sub_i32_e32 v1, vcc, v1, v9
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: sdiv_v2i32:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000
 ; TONGA-NEXT:    s_mov_b32 s6, -1
 ; TONGA-NEXT:    s_mov_b32 s10, s6
 ; TONGA-NEXT:    s_mov_b32 s11, s7
 ; TONGA-NEXT:    s_waitcnt lgkmcnt(0)
 ; TONGA-NEXT:    s_mov_b32 s8, s2
 ; TONGA-NEXT:    s_mov_b32 s9, s3
 ; TONGA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; TONGA-NEXT:    s_mov_b32 s4, s0
 ; TONGA-NEXT:    s_mov_b32 s5, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(0)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v5, 31, v2
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
-; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v5, v2
-; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v7, v3
+; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v2, v5
+; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v3, v7
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v4, 31, v0
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v6, 31, v1
 ; TONGA-NEXT:    v_xor_b32_e32 v2, v2, v5
 ; TONGA-NEXT:    v_xor_b32_e32 v3, v3, v7
 ; TONGA-NEXT:    v_xor_b32_e32 v8, v4, v5
 ; TONGA-NEXT:    v_xor_b32_e32 v9, v6, v7
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v5, v2
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v7, v3
 ; TONGA-NEXT:    v_sub_u32_e32 v10, vcc, 0, v2
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v5, v5
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v7, v7
 ; TONGA-NEXT:    v_sub_u32_e32 v11, vcc, 0, v3
 ; TONGA-NEXT:    v_mul_f32_e32 v5, 0x4f7ffffe, v5
 ; TONGA-NEXT:    v_mul_f32_e32 v7, 0x4f7ffffe, v7
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v5, v5
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v7, v7
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v4, v0
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v4
 ; TONGA-NEXT:    v_mul_lo_u32 v10, v10, v5
 ; TONGA-NEXT:    v_mul_lo_u32 v11, v11, v7
-; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v6, v1
+; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v1, v6
 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v4
 ; TONGA-NEXT:    v_mul_hi_u32 v4, v5, v10
 ; TONGA-NEXT:    v_xor_b32_e32 v1, v1, v6
 ; TONGA-NEXT:    v_mul_hi_u32 v6, v7, v11
 ; TONGA-NEXT:    v_add_u32_e32 v4, vcc, v5, v4
-; TONGA-NEXT:    v_add_u32_e32 v5, vcc, v6, v7
+; TONGA-NEXT:    v_add_u32_e32 v5, vcc, v7, v6
 ; TONGA-NEXT:    v_mul_hi_u32 v4, v0, v4
 ; TONGA-NEXT:    v_mul_hi_u32 v5, v1, v5
 ; TONGA-NEXT:    v_mul_lo_u32 v6, v4, v2
 ; TONGA-NEXT:    v_mul_lo_u32 v10, v5, v3
 ; TONGA-NEXT:    v_add_u32_e32 v7, vcc, 1, v4
 ; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v0, v6
 ; TONGA-NEXT:    v_sub_u32_e32 v1, vcc, v1, v10
 ; TONGA-NEXT:    v_add_u32_e32 v11, vcc, 1, v5
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v2
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v3
-; TONGA-NEXT:    v_subrev_u32_e32 v6, vcc, v2, v0
+; TONGA-NEXT:    v_sub_u32_e32 v6, vcc, v0, v2
 ; TONGA-NEXT:    v_cndmask_b32_e64 v4, v4, v7, s[0:1]
-; TONGA-NEXT:    v_subrev_u32_e32 v7, vcc, v3, v1
+; TONGA-NEXT:    v_sub_u32_e32 v7, vcc, v1, v3
 ; TONGA-NEXT:    v_cndmask_b32_e64 v5, v5, v11, s[2:3]
 ; TONGA-NEXT:    v_cndmask_b32_e64 v0, v0, v6, s[0:1]
 ; TONGA-NEXT:    v_add_u32_e32 v6, vcc, 1, v4
 ; TONGA-NEXT:    v_cndmask_b32_e64 v1, v1, v7, s[2:3]
 ; TONGA-NEXT:    v_add_u32_e32 v7, vcc, 1, v5
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, v4, v6, vcc
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
 ; TONGA-NEXT:    v_cndmask_b32_e32 v1, v5, v7, vcc
 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v8
 ; TONGA-NEXT:    v_xor_b32_e32 v1, v1, v9
-; TONGA-NEXT:    v_subrev_u32_e32 v0, vcc, v8, v0
-; TONGA-NEXT:    v_subrev_u32_e32 v1, vcc, v9, v1
+; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v0, v8
+; TONGA-NEXT:    v_sub_u32_e32 v1, vcc, v1, v9
 ; TONGA-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_v2i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1

 ; GCN-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
 ; GCN-NEXT:    s_mov_b32 s4, s0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_ashrrev_i32_e32 v2, 31, v0
 ; GCN-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
 ; GCN-NEXT:    v_lshrrev_b32_e32 v2, 30, v2
 ; GCN-NEXT:    v_lshrrev_b32_e32 v3, 30, v3
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v2, v0
-; GCN-NEXT:    v_add_i32_e32 v1, vcc, v3, v1
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
+; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
 ; GCN-NEXT:    v_ashrrev_i32_e32 v0, 2, v0
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 2, v1
 ; GCN-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: sdiv_v2i32_4:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000
 ; TONGA-NEXT:    s_mov_b32 s6, -1
 ; TONGA-NEXT:    s_mov_b32 s10, s6
 ; TONGA-NEXT:    s_mov_b32 s11, s7
 ; TONGA-NEXT:    s_waitcnt lgkmcnt(0)
 ; TONGA-NEXT:    s_mov_b32 s8, s2
 ; TONGA-NEXT:    s_mov_b32 s9, s3
 ; TONGA-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
 ; TONGA-NEXT:    s_mov_b32 s4, s0
 ; TONGA-NEXT:    s_mov_b32 s5, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(0)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v2, 31, v0
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v3, 31, v1
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v2, 30, v2
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v3, 30, v3
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v2, v0
-; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v3, v1
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v1, v3
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v0, 2, v0
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v1, 2, v1
 ; TONGA-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_v2i32_4:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24

 ; GCN-NEXT:    s_mov_b32 s8, s0
 ; GCN-NEXT:    s_mov_b32 s9, s1
 ; GCN-NEXT:    s_waitcnt vmcnt(1)
 ; GCN-NEXT:    v_ashrrev_i32_e32 v8, 31, v0
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_ashrrev_i32_e32 v9, 31, v4
 ; GCN-NEXT:    v_ashrrev_i32_e32 v11, 31, v5
 ; GCN-NEXT:    v_ashrrev_i32_e32 v10, 31, v1
-; GCN-NEXT:    v_add_i32_e32 v4, vcc, v9, v4
-; GCN-NEXT:    v_add_i32_e32 v5, vcc, v11, v5
+; GCN-NEXT:    v_add_i32_e32 v4, vcc, v4, v9
+; GCN-NEXT:    v_add_i32_e32 v5, vcc, v5, v11
 ; GCN-NEXT:    v_ashrrev_i32_e32 v13, 31, v6
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v8, v0
-; GCN-NEXT:    v_add_i32_e32 v1, vcc, v10, v1
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v8
+; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v10
 ; GCN-NEXT:    v_xor_b32_e32 v4, v4, v9
 ; GCN-NEXT:    v_xor_b32_e32 v5, v5, v11
 ; GCN-NEXT:    v_ashrrev_i32_e32 v12, 31, v2
 ; GCN-NEXT:    v_xor_b32_e32 v15, v8, v9
 ; GCN-NEXT:    v_xor_b32_e32 v16, v10, v11
-; GCN-NEXT:    v_add_i32_e32 v6, vcc, v13, v6
+; GCN-NEXT:    v_add_i32_e32 v6, vcc, v6, v13
 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v8
 ; GCN-NEXT:    v_xor_b32_e32 v1, v1, v10
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v8, v4
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v10, v5
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v12, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v12
 ; GCN-NEXT:    v_xor_b32_e32 v6, v6, v13
 ; GCN-NEXT:    v_xor_b32_e32 v17, v12, v13
 ; GCN-NEXT:    v_xor_b32_e32 v2, v2, v12
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v12, v6
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v8, v8
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v10, v10
 ; GCN-NEXT:    v_sub_i32_e32 v9, vcc, 0, v4
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v12, v12
 ; GCN-NEXT:    v_mul_f32_e32 v8, 0x4f7ffffe, v8
 ; GCN-NEXT:    v_mul_f32_e32 v10, 0x4f7ffffe, v10
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v8, v8
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v10, v10
 ; GCN-NEXT:    v_mul_f32_e32 v12, 0x4f7ffffe, v12
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v12, v12
 ; GCN-NEXT:    v_sub_i32_e32 v11, vcc, 0, v5
 ; GCN-NEXT:    v_mul_lo_u32 v9, v9, v8
 ; GCN-NEXT:    v_mul_lo_u32 v11, v11, v10
 ; GCN-NEXT:    v_sub_i32_e32 v13, vcc, 0, v6
 ; GCN-NEXT:    v_mul_lo_u32 v13, v13, v12
 ; GCN-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; GCN-NEXT:    v_mul_hi_u32 v11, v10, v11
 ; GCN-NEXT:    v_ashrrev_i32_e32 v14, 31, v7
-; GCN-NEXT:    v_add_i32_e32 v7, vcc, v14, v7
+; GCN-NEXT:    v_add_i32_e32 v7, vcc, v7, v14
 ; GCN-NEXT:    v_mul_hi_u32 v13, v12, v13
 ; GCN-NEXT:    v_xor_b32_e32 v7, v7, v14
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v18, v7
 ; GCN-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; GCN-NEXT:    v_add_i32_e32 v9, vcc, v11, v10
+; GCN-NEXT:    v_add_i32_e32 v9, vcc, v10, v11
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v8
 ; GCN-NEXT:    v_mul_hi_u32 v9, v1, v9
 ; GCN-NEXT:    v_add_i32_e32 v10, vcc, v12, v13
 ; GCN-NEXT:    v_mul_hi_u32 v10, v2, v10
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v18, v18
 ; GCN-NEXT:    v_mul_lo_u32 v11, v8, v4
 ; GCN-NEXT:    v_mul_lo_u32 v13, v9, v5
 ; GCN-NEXT:    v_mul_lo_u32 v21, v10, v6
 ; GCN-NEXT:    v_mul_f32_e32 v18, 0x4f7ffffe, v18
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v18, v18
 ; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v0, v11
 ; GCN-NEXT:    v_sub_i32_e32 v1, vcc, v1, v13
 ; GCN-NEXT:    v_add_i32_e32 v12, vcc, 1, v8
 ; GCN-NEXT:    v_add_i32_e32 v20, vcc, 1, v9
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v4
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v5
 ; GCN-NEXT:    v_sub_i32_e32 v2, vcc, v2, v21
-; GCN-NEXT:    v_subrev_i32_e32 v11, vcc, v4, v0
+; GCN-NEXT:    v_sub_i32_e32 v11, vcc, v0, v4
 ; GCN-NEXT:    v_cndmask_b32_e64 v8, v8, v12, s[0:1]
-; GCN-NEXT:    v_subrev_i32_e32 v12, vcc, v5, v1
+; GCN-NEXT:    v_sub_i32_e32 v12, vcc, v1, v5
 ; GCN-NEXT:    v_cndmask_b32_e64 v9, v9, v20, s[2:3]
 ; GCN-NEXT:    v_sub_i32_e32 v19, vcc, 0, v7
 ; GCN-NEXT:    v_add_i32_e32 v22, vcc, 1, v10
-; GCN-NEXT:    v_subrev_i32_e32 v13, vcc, v6, v2
+; GCN-NEXT:    v_sub_i32_e32 v13, vcc, v2, v6
 ; GCN-NEXT:    v_cndmask_b32_e64 v0, v0, v11, s[0:1]
 ; GCN-NEXT:    v_add_i32_e32 v11, vcc, 1, v8
 ; GCN-NEXT:    v_cndmask_b32_e64 v1, v1, v12, s[2:3]
 ; GCN-NEXT:    v_add_i32_e32 v12, vcc, 1, v9
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v19, v18
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v8, v11, vcc
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v5
 ; GCN-NEXT:    v_ashrrev_i32_e32 v8, 31, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v18, v4
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, v9, v12, vcc
-; GCN-NEXT:    v_add_i32_e32 v3, vcc, v8, v3
+; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v8
 ; GCN-NEXT:    v_xor_b32_e32 v3, v3, v8
-; GCN-NEXT:    v_add_i32_e32 v4, vcc, v4, v18
+; GCN-NEXT:    v_add_i32_e32 v4, vcc, v18, v4
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v6
 ; GCN-NEXT:    v_mul_hi_u32 v4, v3, v4
 ; GCN-NEXT:    v_cndmask_b32_e64 v10, v10, v22, s[4:5]
 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v15
 ; GCN-NEXT:    v_xor_b32_e32 v1, v1, v16
 ; GCN-NEXT:    v_cndmask_b32_e64 v2, v2, v13, s[4:5]
-; GCN-NEXT:    v_subrev_i32_e32 v0, vcc, v15, v0
-; GCN-NEXT:    v_subrev_i32_e32 v1, vcc, v16, v1
+; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v0, v15
+; GCN-NEXT:    v_sub_i32_e32 v1, vcc, v1, v16
 ; GCN-NEXT:    v_add_i32_e32 v5, vcc, 1, v10
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v6
 ; GCN-NEXT:    v_cndmask_b32_e32 v2, v10, v5, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v5, v4, v7
 ; GCN-NEXT:    v_xor_b32_e32 v2, v2, v17
-; GCN-NEXT:    v_subrev_i32_e32 v2, vcc, v17, v2
+; GCN-NEXT:    v_sub_i32_e32 v2, vcc, v2, v17
 ; GCN-NEXT:    v_sub_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_xor_b32_e32 v6, v8, v14
 ; GCN-NEXT:    v_add_i32_e32 v5, vcc, 1, v4
-; GCN-NEXT:    v_subrev_i32_e32 v8, vcc, v7, v3
+; GCN-NEXT:    v_sub_i32_e32 v8, vcc, v3, v7
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v3, v7
 ; GCN-NEXT:    v_cndmask_b32_e32 v4, v4, v5, vcc
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, v3, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v5, vcc, 1, v4
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v3, v7
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, v4, v5, vcc
 ; GCN-NEXT:    v_xor_b32_e32 v3, v3, v6
-; GCN-NEXT:    v_subrev_i32_e32 v3, vcc, v6, v3
+; GCN-NEXT:    v_sub_i32_e32 v3, vcc, v3, v6
 ; GCN-NEXT:    buffer_store_dwordx4 v[0:3], off, s[8:11], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: sdiv_v4i32:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s11, 0xf000
 ; TONGA-NEXT:    s_mov_b32 s10, -1
 ; TONGA-NEXT:    s_mov_b32 s6, s10
 ; TONGA-NEXT:    s_mov_b32 s7, s11
 ; TONGA-NEXT:    s_waitcnt lgkmcnt(0)
 ; TONGA-NEXT:    s_mov_b32 s4, s2
 ; TONGA-NEXT:    s_mov_b32 s5, s3
 ; TONGA-NEXT:    buffer_load_dwordx4 v[0:3], off, s[4:7], 0
 ; TONGA-NEXT:    buffer_load_dwordx4 v[4:7], off, s[4:7], 0 offset:16
 ; TONGA-NEXT:    s_mov_b32 s8, s0
 ; TONGA-NEXT:    s_mov_b32 s9, s1
 ; TONGA-NEXT:    s_waitcnt vmcnt(1)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v8, 31, v0
 ; TONGA-NEXT:    s_waitcnt vmcnt(0)
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v9, 31, v4
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v11, 31, v5
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v10, 31, v1
-; TONGA-NEXT:    v_add_u32_e32 v4, vcc, v9, v4
-; TONGA-NEXT:    v_add_u32_e32 v5, vcc, v11, v5
+; TONGA-NEXT:    v_add_u32_e32 v4, vcc, v4, v9
+; TONGA-NEXT:    v_add_u32_e32 v5, vcc, v5, v11
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v13, 31, v6
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v8, v0
-; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v10, v1
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v8
+; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v1, v10
 ; TONGA-NEXT:    v_xor_b32_e32 v4, v4, v9
 ; TONGA-NEXT:    v_xor_b32_e32 v5, v5, v11
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v12, 31, v2
 ; TONGA-NEXT:    v_xor_b32_e32 v15, v8, v9
 ; TONGA-NEXT:    v_xor_b32_e32 v16, v10, v11
-; TONGA-NEXT:    v_add_u32_e32 v6, vcc, v13, v6
+; TONGA-NEXT:    v_add_u32_e32 v6, vcc, v6, v13
 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v8
 ; TONGA-NEXT:    v_xor_b32_e32 v1, v1, v10
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v8, v4
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v10, v5
-; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v12, v2
+; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v2, v12
 ; TONGA-NEXT:    v_xor_b32_e32 v6, v6, v13
 ; TONGA-NEXT:    v_xor_b32_e32 v17, v12, v13
 ; TONGA-NEXT:    v_xor_b32_e32 v2, v2, v12
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v12, v6
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v8, v8
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v10, v10
 ; TONGA-NEXT:    v_sub_u32_e32 v9, vcc, 0, v4
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v12, v12
 ; TONGA-NEXT:    v_mul_f32_e32 v8, 0x4f7ffffe, v8
 ; TONGA-NEXT:    v_mul_f32_e32 v10, 0x4f7ffffe, v10
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v8, v8
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v10, v10
 ; TONGA-NEXT:    v_mul_f32_e32 v12, 0x4f7ffffe, v12
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v12, v12
 ; TONGA-NEXT:    v_sub_u32_e32 v11, vcc, 0, v5
 ; TONGA-NEXT:    v_mul_lo_u32 v9, v9, v8
 ; TONGA-NEXT:    v_mul_lo_u32 v11, v11, v10
 ; TONGA-NEXT:    v_sub_u32_e32 v13, vcc, 0, v6
 ; TONGA-NEXT:    v_mul_lo_u32 v13, v13, v12
 ; TONGA-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; TONGA-NEXT:    v_mul_hi_u32 v11, v10, v11
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v14, 31, v7
-; TONGA-NEXT:    v_add_u32_e32 v7, vcc, v14, v7
+; TONGA-NEXT:    v_add_u32_e32 v7, vcc, v7, v14
 ; TONGA-NEXT:    v_mul_hi_u32 v13, v12, v13
 ; TONGA-NEXT:    v_xor_b32_e32 v7, v7, v14
 ; TONGA-NEXT:    v_cvt_f32_u32_e32 v18, v7
 ; TONGA-NEXT:    v_add_u32_e32 v8, vcc, v8, v9
-; TONGA-NEXT:    v_add_u32_e32 v9, vcc, v11, v10
+; TONGA-NEXT:    v_add_u32_e32 v9, vcc, v10, v11
 ; TONGA-NEXT:    v_mul_hi_u32 v8, v0, v8
 ; TONGA-NEXT:    v_mul_hi_u32 v9, v1, v9
 ; TONGA-NEXT:    v_add_u32_e32 v10, vcc, v12, v13
 ; TONGA-NEXT:    v_mul_hi_u32 v10, v2, v10
 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v18, v18
 ; TONGA-NEXT:    v_mul_lo_u32 v11, v8, v4
 ; TONGA-NEXT:    v_mul_lo_u32 v13, v9, v5
 ; TONGA-NEXT:    v_mul_lo_u32 v21, v10, v6
 ; TONGA-NEXT:    v_mul_f32_e32 v18, 0x4f7ffffe, v18
 ; TONGA-NEXT:    v_cvt_u32_f32_e32 v18, v18
 ; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v0, v11
 ; TONGA-NEXT:    v_sub_u32_e32 v1, vcc, v1, v13
 ; TONGA-NEXT:    v_add_u32_e32 v12, vcc, 1, v8
 ; TONGA-NEXT:    v_add_u32_e32 v20, vcc, 1, v9
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v4
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v5
 ; TONGA-NEXT:    v_sub_u32_e32 v2, vcc, v2, v21
-; TONGA-NEXT:    v_subrev_u32_e32 v11, vcc, v4, v0
+; TONGA-NEXT:    v_sub_u32_e32 v11, vcc, v0, v4
 ; TONGA-NEXT:    v_cndmask_b32_e64 v8, v8, v12, s[0:1]
-; TONGA-NEXT:    v_subrev_u32_e32 v12, vcc, v5, v1
+; TONGA-NEXT:    v_sub_u32_e32 v12, vcc, v1, v5
 ; TONGA-NEXT:    v_cndmask_b32_e64 v9, v9, v20, s[2:3]
 ; TONGA-NEXT:    v_sub_u32_e32 v19, vcc, 0, v7
 ; TONGA-NEXT:    v_add_u32_e32 v22, vcc, 1, v10
-; TONGA-NEXT:    v_subrev_u32_e32 v13, vcc, v6, v2
+; TONGA-NEXT:    v_sub_u32_e32 v13, vcc, v2, v6
 ; TONGA-NEXT:    v_cndmask_b32_e64 v0, v0, v11, s[0:1]
 ; TONGA-NEXT:    v_add_u32_e32 v11, vcc, 1, v8
 ; TONGA-NEXT:    v_cndmask_b32_e64 v1, v1, v12, s[2:3]
 ; TONGA-NEXT:    v_add_u32_e32 v12, vcc, 1, v9
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v4
 ; TONGA-NEXT:    v_mul_lo_u32 v4, v19, v18
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, v8, v11, vcc
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v5
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v8, 31, v3
 ; TONGA-NEXT:    v_mul_hi_u32 v4, v18, v4
 ; TONGA-NEXT:    v_cndmask_b32_e32 v1, v9, v12, vcc
-; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v8, v3
+; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v3, v8
 ; TONGA-NEXT:    v_xor_b32_e32 v3, v3, v8
-; TONGA-NEXT:    v_add_u32_e32 v4, vcc, v4, v18
+; TONGA-NEXT:    v_add_u32_e32 v4, vcc, v18, v4
 ; TONGA-NEXT:    v_cmp_ge_u32_e64 s[4:5], v2, v6
 ; TONGA-NEXT:    v_mul_hi_u32 v4, v3, v4
 ; TONGA-NEXT:    v_cndmask_b32_e64 v10, v10, v22, s[4:5]
 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v15
 ; TONGA-NEXT:    v_xor_b32_e32 v1, v1, v16
 ; TONGA-NEXT:    v_cndmask_b32_e64 v2, v2, v13, s[4:5]
-; TONGA-NEXT:    v_subrev_u32_e32 v0, vcc, v15, v0
-; TONGA-NEXT:    v_subrev_u32_e32 v1, vcc, v16, v1
+; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v0, v15
+; TONGA-NEXT:    v_sub_u32_e32 v1, vcc, v1, v16
 ; TONGA-NEXT:    v_add_u32_e32 v5, vcc, 1, v10
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v6
 ; TONGA-NEXT:    v_cndmask_b32_e32 v2, v10, v5, vcc
 ; TONGA-NEXT:    v_mul_lo_u32 v5, v4, v7
 ; TONGA-NEXT:    v_xor_b32_e32 v2, v2, v17
-; TONGA-NEXT:    v_subrev_u32_e32 v2, vcc, v17, v2
+; TONGA-NEXT:    v_sub_u32_e32 v2, vcc, v2, v17
 ; TONGA-NEXT:    v_sub_u32_e32 v3, vcc, v3, v5
 ; TONGA-NEXT:    v_xor_b32_e32 v6, v8, v14
 ; TONGA-NEXT:    v_add_u32_e32 v5, vcc, 1, v4
-; TONGA-NEXT:    v_subrev_u32_e32 v8, vcc, v7, v3
+; TONGA-NEXT:    v_sub_u32_e32 v8, vcc, v3, v7
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v3, v7
 ; TONGA-NEXT:    v_cndmask_b32_e32 v4, v4, v5, vcc
 ; TONGA-NEXT:    v_cndmask_b32_e32 v3, v3, v8, vcc
 ; TONGA-NEXT:    v_add_u32_e32 v5, vcc, 1, v4
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v3, v7
 ; TONGA-NEXT:    v_cndmask_b32_e32 v3, v4, v5, vcc
 ; TONGA-NEXT:    v_xor_b32_e32 v3, v3, v6
-; TONGA-NEXT:    v_subrev_u32_e32 v3, vcc, v6, v3
+; TONGA-NEXT:    v_sub_u32_e32 v3, vcc, v3, v6
 ; TONGA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[8:11], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_v4i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX9-NEXT:    s_mov_b32 s2, -1

 ; GCN-NEXT:    v_ashrrev_i32_e32 v4, 31, v0
 ; GCN-NEXT:    v_ashrrev_i32_e32 v5, 31, v1
 ; GCN-NEXT:    v_ashrrev_i32_e32 v6, 31, v2
 ; GCN-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
 ; GCN-NEXT:    v_lshrrev_b32_e32 v4, 30, v4
 ; GCN-NEXT:    v_lshrrev_b32_e32 v5, 30, v5
 ; GCN-NEXT:    v_lshrrev_b32_e32 v6, 30, v6
 ; GCN-NEXT:    v_lshrrev_b32_e32 v7, 30, v7
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v4, v0
-; GCN-NEXT:    v_add_i32_e32 v1, vcc, v5, v1
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v6, v2
-; GCN-NEXT:    v_add_i32_e32 v3, vcc, v7, v3
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
+; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
+; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
 ; GCN-NEXT:    v_ashrrev_i32_e32 v0, 2, v0
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 2, v1
 ; GCN-NEXT:    v_ashrrev_i32_e32 v2, 2, v2
 ; GCN-NEXT:    v_ashrrev_i32_e32 v3, 2, v3
 ; GCN-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: sdiv_v4i32_4:

 ; TONGA-NEXT:    v_ashrrev_i32_e32 v4, 31, v0
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v5, 31, v1
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v6, 31, v2
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v7, 31, v3
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v4, 30, v4
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v5, 30, v5
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v6, 30, v6
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v7, 30, v7
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v4, v0
-; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v5, v1
-; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v6, v2
-; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v7, v3
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v4
+; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
+; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v2, v6
+; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v3, v7
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v0, 2, v0
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v1, 2, v1
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v2, 2, v2
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v3, 2, v3
 ; TONGA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: sdiv_v4i32_4:

 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v4, v2
 ; GCN-NEXT:    v_or_b32_e32 v0, 1, v0
 ; GCN-NEXT:    v_mul_f32_e32 v1, v3, v4
 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mad_f32 v3, -v1, v2, v3
 ; GCN-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v2|
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 8
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: v_sdiv_i8:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000

 ; TONGA-NEXT:    v_rcp_iflag_f32_e32 v4, v2
 ; TONGA-NEXT:    v_or_b32_e32 v0, 1, v0
 ; TONGA-NEXT:    v_mul_f32_e32 v1, v3, v4
 ; TONGA-NEXT:    v_trunc_f32_e32 v1, v1
 ; TONGA-NEXT:    v_mad_f32 v3, -v1, v2, v3
 ; TONGA-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; TONGA-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v2|
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
 ; TONGA-NEXT:    v_bfe_i32 v0, v0, 0, 8
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: v_sdiv_i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000

 ; GCN-NEXT:    v_ashrrev_i32_e32 v0, 30, v0
 ; GCN-NEXT:    v_or_b32_e32 v0, 1, v0
 ; GCN-NEXT:    v_mul_f32_e32 v1, v3, v4
 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mad_f32 v3, -v1, v2, v3
 ; GCN-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v2|
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 23
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: v_sdiv_i23:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000

 ; TONGA-NEXT:    v_ashrrev_i32_e32 v0, 30, v0
 ; TONGA-NEXT:    v_or_b32_e32 v0, 1, v0
 ; TONGA-NEXT:    v_mul_f32_e32 v1, v3, v4
 ; TONGA-NEXT:    v_trunc_f32_e32 v1, v1
 ; TONGA-NEXT:    v_mad_f32 v3, -v1, v2, v3
 ; TONGA-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; TONGA-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v2|
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
 ; TONGA-NEXT:    v_bfe_i32 v0, v0, 0, 23
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: v_sdiv_i23:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000

 ; GCN-NEXT:    v_ashrrev_i32_e32 v0, 30, v0
 ; GCN-NEXT:    v_or_b32_e32 v0, 1, v0
 ; GCN-NEXT:    v_mul_f32_e32 v2, v3, v4
 ; GCN-NEXT:    v_trunc_f32_e32 v2, v2
 ; GCN-NEXT:    v_mad_f32 v3, -v2, v1, v3
 ; GCN-NEXT:    v_cvt_i32_f32_e32 v2, v2
 ; GCN-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v1|
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
+; GCN-NEXT:    v_add_i32_e32 v0, vcc, v2, v0
 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: v_sdiv_i24:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000

 ; TONGA-NEXT:    v_ashrrev_i32_e32 v0, 30, v0
 ; TONGA-NEXT:    v_or_b32_e32 v0, 1, v0
 ; TONGA-NEXT:    v_mul_f32_e32 v2, v3, v4
 ; TONGA-NEXT:    v_trunc_f32_e32 v2, v2
 ; TONGA-NEXT:    v_mad_f32 v3, -v2, v1, v3
 ; TONGA-NEXT:    v_cvt_i32_f32_e32 v2, v2
 ; TONGA-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v1|
 ; TONGA-NEXT:    v_cndmask_b32_e32 v0, 0, v0, vcc
-; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v2, v0
 ; TONGA-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: v_sdiv_i24:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000

 ; GCN-NEXT:    v_xor_b32_e32 v0, v0, v1
 ; GCN-NEXT:    v_mul_lo_u32 v4, v4, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v3, v4
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_mul_hi_u32 v3, v5, v3
 ; GCN-NEXT:    v_mul_lo_u32 v1, v3, v2
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, 1, v3
 ; GCN-NEXT:    v_sub_i32_e32 v1, vcc, v5, v1
-; GCN-NEXT:    v_subrev_i32_e32 v5, vcc, v2, v1
+; GCN-NEXT:    v_sub_i32_e32 v5, vcc, v1, v2
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v2
 ; GCN-NEXT:    v_cndmask_b32_e32 v3, v3, v4, vcc
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, 1, v3
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v2
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, v3, v4, vcc
 ; GCN-NEXT:    v_xor_b32_e32 v1, v1, v0
-; GCN-NEXT:    v_subrev_i32_e32 v0, vcc, v0, v1
+; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v1, v0
 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 25
 ; GCN-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: v_sdiv_i25:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000

 ; TONGA-NEXT:    v_xor_b32_e32 v0, v0, v1
 ; TONGA-NEXT:    v_mul_lo_u32 v4, v4, v3
 ; TONGA-NEXT:    v_mul_hi_u32 v4, v3, v4
 ; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v3, v4
 ; TONGA-NEXT:    v_mul_hi_u32 v3, v5, v3
 ; TONGA-NEXT:    v_mul_lo_u32 v1, v3, v2
 ; TONGA-NEXT:    v_add_u32_e32 v4, vcc, 1, v3
 ; TONGA-NEXT:    v_sub_u32_e32 v1, vcc, v5, v1
-; TONGA-NEXT:    v_subrev_u32_e32 v5, vcc, v2, v1
+; TONGA-NEXT:    v_sub_u32_e32 v5, vcc, v1, v2
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v2
 ; TONGA-NEXT:    v_cndmask_b32_e32 v3, v3, v4, vcc
 ; TONGA-NEXT:    v_cndmask_b32_e32 v1, v1, v5, vcc
 ; TONGA-NEXT:    v_add_u32_e32 v4, vcc, 1, v3
 ; TONGA-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v2
 ; TONGA-NEXT:    v_cndmask_b32_e32 v1, v3, v4, vcc
 ; TONGA-NEXT:    v_xor_b32_e32 v1, v1, v0
-; TONGA-NEXT:    v_subrev_u32_e32 v0, vcc, v0, v1
+; TONGA-NEXT:    v_sub_u32_e32 v0, vcc, v1, v0
 ; TONGA-NEXT:    v_bfe_i32 v0, v0, 0, 25
 ; TONGA-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: v_sdiv_i25:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000

 ; GCN-NEXT:    v_lshrrev_b32_e32 v5, 31, v1
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 12, v1
 ; GCN-NEXT:    v_lshrrev_b32_e32 v6, 31, v2
 ; GCN-NEXT:    v_ashrrev_i32_e32 v2, 12, v2
 ; GCN-NEXT:    v_lshrrev_b32_e32 v7, 31, v3
 ; GCN-NEXT:    v_ashrrev_i32_e32 v3, 12, v3
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
 ; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v6, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v6
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v7
 ; GCN-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; GCN-NEXT:    s_endpgm
 ;
 ; TONGA-LABEL: scalarize_mulhs_4xi32:
 ; TONGA:       ; %bb.0:
 ; TONGA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; TONGA-NEXT:    s_mov_b32 s7, 0xf000

 ; TONGA-NEXT:    v_lshrrev_b32_e32 v5, 31, v1
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v1, 12, v1
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v6, 31, v2
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v2, 12, v2
 ; TONGA-NEXT:    v_lshrrev_b32_e32 v7, 31, v3
 ; TONGA-NEXT:    v_ashrrev_i32_e32 v3, 12, v3
 ; TONGA-NEXT:    v_add_u32_e32 v0, vcc, v0, v4
 ; TONGA-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
-; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v6, v2
+; TONGA-NEXT:    v_add_u32_e32 v2, vcc, v2, v6
 ; TONGA-NEXT:    v_add_u32_e32 v3, vcc, v3, v7
 ; TONGA-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; TONGA-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: scalarize_mulhs_4xi32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX9-NEXT:    s_mov_b32 s7, 0xf000

llvm/test/CodeGen/AMDGPU/sdiv64.ll

 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    s_add_u32 s2, s2, s12
 ; GCN-NEXT:    s_mov_b32 s13, s12
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s5, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s4, v0
 ; GCN-NEXT:    s_addc_u32 s3, s3, s12
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s5, v0
 ; GCN-NEXT:    s_xor_b64 s[2:3], s[2:3], s[12:13]
 ; GCN-NEXT:    s_mov_b32 s5, s1
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s4, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s5, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s4, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s5, v0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s4, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GCN-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

llvm/test/CodeGen/AMDGPU/sdwa-peephole.ll

-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=NOSDWA,GCN %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=VI,GFX89,SDWA,GCN %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -amdgpu-sdwa-peephole -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9,GFX9_10,SDWA,GCN %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx1010 -amdgpu-sdwa-peephole -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX10,GFX9_10,SDWA,GCN %s
-
-; GCN-LABEL: {{^}}add_shr_i32:
-; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_add_u32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v[[DST]]
-; NOSDWA-NOT: v_add_{{(_co)?}}_u32_sdwa
-
-; VI: v_add_u32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9: v_add_u32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX10: v_add_nc_u32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole=0 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=NOSDWA %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -amdgpu-sdwa-peephole -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=VI %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -amdgpu-sdwa-peephole -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx1010 -amdgpu-sdwa-peephole -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX10 %s
 
 define amdgpu_kernel void @add_shr_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; NOSDWA-LABEL: add_shr_i32:
+; NOSDWA:       ; %bb.0:
+; NOSDWA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s2
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s3
+; NOSDWA-NEXT:    flat_load_dword v2, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s1
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, v2, v3
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: add_shr_i32:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    flat_load_dword v2, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_add_u32_sdwa v2, vcc, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: add_shr_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_add_u32_sdwa v1, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: add_shr_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_add_nc_u32_sdwa v1, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX10-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT:    s_endpgm
   %a = load i32, ptr addrspace(1) %in, align 4
   %shr = lshr i32 %a, 16
   %add = add i32 %a, %shr
   store i32 %add, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}sub_shr_i32:
-; NOSDWA: v_lshrrev_b32_e32 v[[DST:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_subrev_u32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v[[DST]]
-; NOSDWA-NOT: v_subrev_{{(_co)?}}_u32_sdwa
-
-; VI: v_subrev_u32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX9: v_sub_u32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX10: v_sub_nc_u32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 define amdgpu_kernel void @sub_shr_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; NOSDWA-LABEL: sub_shr_i32:
+; NOSDWA:       ; %bb.0:
+; NOSDWA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s2
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s3
+; NOSDWA-NEXT:    flat_load_dword v2, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s1
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
+; NOSDWA-NEXT:    v_sub_u32_e32 v2, vcc, v3, v2
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: sub_shr_i32:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    flat_load_dword v2, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_sub_u32_sdwa v2, vcc, v2, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: sub_shr_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_sub_u32_sdwa v1, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: sub_shr_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_sub_nc_u32_sdwa v1, v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX10-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT:    s_endpgm
   %a = load i32, ptr addrspace(1) %in, align 4
   %shr = lshr i32 %a, 16
   %sub = sub i32 %shr, %a
   store i32 %sub, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_shr_i32:
-; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_mul_u32_u24_e32 v{{[0-9]+}}, v[[DST0]], v[[DST1]]
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; SDWA: v_mul_u32_u24_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-
 define amdgpu_kernel void @mul_shr_i32(ptr addrspace(1) %out, ptr addrspace(1) %in1, ptr addrspace(1) %in2) #0 {
+; NOSDWA-LABEL: mul_shr_i32:
+; NOSDWA:       ; %bb.0:
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dword v4, v[0:1]
+; NOSDWA-NEXT:    flat_load_dword v2, v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v4
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_mul_u32_u24_e32 v2, v3, v2
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_shr_i32:
+; VI:       ; %bb.0:
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_u32_u24_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_shr_i32:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_shr_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_mul_u32_u24_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX10-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gep1 = getelementptr i32, ptr addrspace(1) %in1, i32 %idx
   %gep2 = getelementptr i32, ptr addrspace(1) %in2, i32 %idx
   %a = load i32, ptr addrspace(1) %gep1, align 4
   %b = load i32, ptr addrspace(1) %gep2, align 4
   %shra = lshr i32 %a, 16
   %shrb = lshr i32 %b, 16
   %mul = mul i32 %shra, %shrb
   store i32 %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_i16:
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-; GFX89: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX10: v_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; SDWA-NOT: v_mul_u32_u24_sdwa
-
 define amdgpu_kernel void @mul_i16(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 1, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_ushort v4, v[0:1]
+; NOSDWA-NEXT:    flat_load_ushort v2, v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v4, v2
+; NOSDWA-NEXT:    flat_store_short v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 1, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_ushort v4, v[0:1]
+; VI-NEXT:    flat_load_ushort v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v2, v4, v2
+; VI-NEXT:    flat_store_short v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_ushort v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_ushort v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_lo_u16_e32 v1, v1, v2
+; GFX9-NEXT:    global_store_short v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_ushort v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_ushort v2, v0, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_mul_lo_u16 v1, v1, v2
+; GFX10-NEXT:    global_store_short v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr i16, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr i16, ptr addrspace(1) %inb, i32 %idx
   %a = load i16, ptr addrspace(1) %gepa, align 4
   %b = load i16, ptr addrspace(1) %gepb, align 4
   %mul = mul i16 %a, %b
   store i16 %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v2i16:
-; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v[[DST_MUL:[0-9]+]], v[[DST1]], v[[DST0]]
-; NOSDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MUL]]
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v[[DST_SHL]]
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL_LO:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL_HI:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL_LO]], v[[DST_MUL_HI]]
-
-; GFX9_10: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @mul_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v2i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dword v4, v[0:1]
+; NOSDWA-NEXT:    flat_load_dword v2, v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v4, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v4, 16, v4
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v4, v2
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v3, v2
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v2i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v3, v4, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v2, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_or_b32_e32 v2, v3, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v2i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_lo_u16 v1, v1, v2
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v2i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_lo_u16 v1, v1, v2
+; GFX10-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <2 x i16>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <2 x i16>, ptr addrspace(1) %inb, i32 %idx
   %a = load <2 x i16>, ptr addrspace(1) %gepa, align 4
   %b = load <2 x i16>, ptr addrspace(1) %gepb, align 4
   %mul = mul <2 x i16> %a, %b
   store <2 x i16> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v4i16:
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL0:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL1:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL2:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL3:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL2]], v[[DST_MUL3]]
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL0]], v[[DST_MUL1]]
-
-; GFX9_10-DAG: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @mul_v4i16(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v4i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; NOSDWA-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v4, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v5, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v6, v1, v3
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v1, 16, v1
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v7, v0, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v1, v1, v3
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v0, v0, v2
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v6, v1
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v7, v0
+; NOSDWA-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v4i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v6, v1, v3
+; VI-NEXT:    v_mul_lo_u16_sdwa v1, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_lo_u16_e32 v3, v0, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_or_b32_e32 v1, v6, v1
+; VI-NEXT:    v_or_b32_e32 v0, v3, v0
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v4i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GFX9-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_lo_u16 v1, v1, v3
+; GFX9-NEXT:    v_pk_mul_lo_u16 v0, v0, v2
+; GFX9-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v4i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GFX10-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_lo_u16 v1, v1, v3
+; GFX10-NEXT:    v_pk_mul_lo_u16 v0, v0, v2
+; GFX10-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <4 x i16>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <4 x i16>, ptr addrspace(1) %inb, i32 %idx
   %a = load <4 x i16>, ptr addrspace(1) %gepa, align 4
   %b = load <4 x i16>, ptr addrspace(1) %gepb, align 4
   %mul = mul <4 x i16> %a, %b
   store <4 x i16> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v8i16:
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL0:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL1:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL2:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL3:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL4:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL5:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_lo_u16_e32 v[[DST_MUL6:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v[[DST_MUL7:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL6]], v[[DST_MUL7]]
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL4]], v[[DST_MUL5]]
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL2]], v[[DST_MUL3]]
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL0]], v[[DST_MUL1]]
-
-; GFX9_10-DAG: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @mul_v8i16(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v8i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v4, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v5, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; NOSDWA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; NOSDWA-NEXT:    v_mov_b32_e32 v8, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v9, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v10, v3, v7
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v7, 16, v7
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v11, v2, v6
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v6, 16, v6
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v12, v1, v5
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v5, 16, v5
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v1, 16, v1
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v13, v0, v4
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v4, 16, v4
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v3, v7
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v2, v6
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v1, v1, v5
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v0, v0, v4
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; NOSDWA-NEXT:    v_or_b32_e32 v3, v10, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v11, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v12, v1
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v13, v0
+; NOSDWA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v8i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v4, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v5, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; VI-NEXT:    v_mov_b32_e32 v8, s4
+; VI-NEXT:    v_mov_b32_e32 v9, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v10, v3, v7
+; VI-NEXT:    v_mul_lo_u16_sdwa v3, v3, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_lo_u16_e32 v7, v2, v6
+; VI-NEXT:    v_mul_lo_u16_sdwa v2, v2, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_lo_u16_e32 v6, v1, v5
+; VI-NEXT:    v_mul_lo_u16_sdwa v1, v1, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_lo_u16_e32 v5, v0, v4
+; VI-NEXT:    v_mul_lo_u16_sdwa v0, v0, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_or_b32_e32 v3, v10, v3
+; VI-NEXT:    v_or_b32_e32 v2, v7, v2
+; VI-NEXT:    v_or_b32_e32 v1, v6, v1
+; VI-NEXT:    v_or_b32_e32 v0, v5, v0
+; VI-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v8i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v8, 4, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v8, s[6:7]
+; GFX9-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v8, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_lo_u16 v3, v3, v7
+; GFX9-NEXT:    v_pk_mul_lo_u16 v2, v2, v6
+; GFX9-NEXT:    v_pk_mul_lo_u16 v1, v1, v5
+; GFX9-NEXT:    v_pk_mul_lo_u16 v0, v0, v4
+; GFX9-NEXT:    global_store_dwordx4 v8, v[0:3], s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v8i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v8, 4, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v8, s[6:7]
+; GFX10-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v8, 0
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_lo_u16 v3, v3, v7
+; GFX10-NEXT:    v_pk_mul_lo_u16 v2, v2, v6
+; GFX10-NEXT:    v_pk_mul_lo_u16 v1, v1, v5
+; GFX10-NEXT:    v_pk_mul_lo_u16 v0, v0, v4
+; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <8 x i16>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <8 x i16>, ptr addrspace(1) %inb, i32 %idx
   %a = load <8 x i16>, ptr addrspace(1) %gepa, align 4
   %b = load <8 x i16>, ptr addrspace(1) %gepb, align 4
   %mul = mul <8 x i16> %a, %b
   store <8 x i16> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_half:
-; NOSDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_f16_sdwa
-; SDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; SDWA-NOT: v_mul_f16_sdwa
-
 define amdgpu_kernel void @mul_half(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_half:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s6
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    flat_load_ushort v4, v[0:1]
+; NOSDWA-NEXT:    flat_load_ushort v2, v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_f16_e32 v2, v4, v2
+; NOSDWA-NEXT:    flat_store_short v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_half:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    flat_load_ushort v4, v[0:1]
+; VI-NEXT:    flat_load_ushort v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_f16_e32 v2, v4, v2
+; VI-NEXT:    flat_store_short v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_half:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_ushort v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_ushort v2, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_f16_e32 v1, v1, v2
+; GFX9-NEXT:    global_store_short v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_half:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_ushort v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_ushort v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_mul_f16_e32 v1, v1, v2
+; GFX10-NEXT:    global_store_short v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %a = load half, ptr addrspace(1) %ina, align 4
   %b = load half, ptr addrspace(1) %inb, align 4
   %mul = fmul half %a, %b
   store half %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v2half:
-; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_mul_f16_e32 v[[DST_MUL:[0-9]+]], v[[DST0]], v[[DST1]]
-; NOSDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MUL]]
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v[[DST_SHL]]
-; NOSDWA-NOT: v_mul_f16_sdwa
-
-; VI-DAG: v_mul_f16_sdwa v[[DST_MUL_HI:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_f16_e32 v[[DST_MUL_LO:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; VI: v_or_b32_e32 v{{[0-9]+}}, v[[DST_MUL_LO]], v[[DST_MUL_HI]]
-
-; GFX9_10: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @mul_v2half(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v2half:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s6
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    flat_load_dword v2, v[2:3]
+; NOSDWA-NEXT:    flat_load_dword v3, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
+; NOSDWA-NEXT:    v_mul_f16_e32 v4, v5, v4
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; NOSDWA-NEXT:    v_mul_f16_e32 v2, v3, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v4
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v2half:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_f16_sdwa v3, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v2, v4, v2
+; VI-NEXT:    v_or_b32_e32 v2, v2, v3
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v2half:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_f16 v1, v1, v2
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v2half:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_f16 v1, v1, v2
+; GFX10-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %a = load <2 x half>, ptr addrspace(1) %ina, align 4
   %b = load <2 x half>, ptr addrspace(1) %inb, align 4
   %mul = fmul <2 x half> %a, %b
   store <2 x half> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v4half:
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_f16_sdwa
-
-; VI-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
-; GFX9_10-DAG: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @mul_v4half(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v4half:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s6
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; NOSDWA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v4, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v5, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v6, 16, v3
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v7, 16, v1
+; NOSDWA-NEXT:    v_mul_f16_e32 v1, v1, v3
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v8, 16, v0
+; NOSDWA-NEXT:    v_mul_f16_e32 v0, v0, v2
+; NOSDWA-NEXT:    v_mul_f16_e32 v2, v7, v6
+; NOSDWA-NEXT:    v_mul_f16_e32 v3, v8, v3
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v1, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v0, v3
+; NOSDWA-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v4half:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_f16_sdwa v6, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v1, v1, v3
+; VI-NEXT:    v_mul_f16_sdwa v3, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v0, v0, v2
+; VI-NEXT:    v_or_b32_e32 v1, v1, v6
+; VI-NEXT:    v_or_b32_e32 v0, v0, v3
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v4half:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GFX9-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_f16 v1, v1, v3
+; GFX9-NEXT:    v_pk_mul_f16 v0, v0, v2
+; GFX9-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v4half:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GFX10-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_f16 v1, v1, v3
+; GFX10-NEXT:    v_pk_mul_f16 v0, v0, v2
+; GFX10-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %a = load <4 x half>, ptr addrspace(1) %ina, align 4
   %b = load <4 x half>, ptr addrspace(1) %inb, align 4
   %mul = fmul <4 x half> %a, %b
   store <4 x half> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v8half:
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_mul_f16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_f16_sdwa
-
-; VI-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_mul_f16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; VI-DAG: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
-; GFX9_10-DAG: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10-DAG: v_pk_mul_f16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @mul_v8half(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v8half:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v4, s6
+; NOSDWA-NEXT:    v_mov_b32_e32 v5, s7
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s1
+; NOSDWA-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; NOSDWA-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; NOSDWA-NEXT:    v_mov_b32_e32 v8, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v9, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v10, 16, v3
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v11, 16, v7
+; NOSDWA-NEXT:    v_mul_f16_e32 v3, v7, v3
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v12, 16, v6
+; NOSDWA-NEXT:    v_mul_f16_e32 v2, v6, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v6, 16, v1
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v13, 16, v5
+; NOSDWA-NEXT:    v_mul_f16_e32 v1, v5, v1
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v5, 16, v0
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v14, 16, v4
+; NOSDWA-NEXT:    v_mul_f16_e32 v0, v4, v0
+; NOSDWA-NEXT:    v_mul_f16_e32 v4, v11, v10
+; NOSDWA-NEXT:    v_mul_f16_e32 v7, v12, v7
+; NOSDWA-NEXT:    v_mul_f16_e32 v6, v13, v6
+; NOSDWA-NEXT:    v_mul_f16_e32 v5, v14, v5
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v7, 16, v7
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v5, 16, v5
+; NOSDWA-NEXT:    v_or_b32_e32 v3, v3, v4
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v7
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v1, v6
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v0, v5
+; NOSDWA-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v8half:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_mov_b32_e32 v4, s0
+; VI-NEXT:    v_mov_b32_e32 v5, s1
+; VI-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; VI-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; VI-NEXT:    v_mov_b32_e32 v8, s4
+; VI-NEXT:    v_mov_b32_e32 v9, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_f16_sdwa v10, v3, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v3, v3, v7
+; VI-NEXT:    v_mul_f16_sdwa v7, v2, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v2, v2, v6
+; VI-NEXT:    v_mul_f16_sdwa v6, v1, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v1, v1, v5
+; VI-NEXT:    v_mul_f16_sdwa v5, v0, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_f16_e32 v0, v0, v4
+; VI-NEXT:    v_or_b32_e32 v3, v3, v10
+; VI-NEXT:    v_or_b32_e32 v2, v2, v7
+; VI-NEXT:    v_or_b32_e32 v1, v1, v6
+; VI-NEXT:    v_or_b32_e32 v0, v0, v5
+; VI-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v8half:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_mov_b32_e32 v8, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dwordx4 v[0:3], v8, s[6:7]
+; GFX9-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_f16 v3, v3, v7
+; GFX9-NEXT:    v_pk_mul_f16 v2, v2, v6
+; GFX9-NEXT:    v_pk_mul_f16 v1, v1, v5
+; GFX9-NEXT:    v_pk_mul_f16 v0, v0, v4
+; GFX9-NEXT:    global_store_dwordx4 v8, v[0:3], s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v8half:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v8, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dwordx4 v[0:3], v8, s[6:7]
+; GFX10-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_f16 v3, v3, v7
+; GFX10-NEXT:    v_pk_mul_f16 v2, v2, v6
+; GFX10-NEXT:    v_pk_mul_f16 v1, v1, v5
+; GFX10-NEXT:    v_pk_mul_f16 v0, v0, v4
+; GFX10-NEXT:    global_store_dwordx4 v8, v[0:3], s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %a = load <8 x half>, ptr addrspace(1) %ina, align 4
   %b = load <8 x half>, ptr addrspace(1) %inb, align 4
   %mul = fmul <8 x half> %a, %b
   store <8 x half> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_i8:
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-; GFX89: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; GFX10: v_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; SDWA-NOT: v_mul_u32_u24_sdwa
-
 define amdgpu_kernel void @mul_i8(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_i8:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v1, vcc, s6, v0
+; NOSDWA-NEXT:    v_addc_u32_e32 v2, vcc, 0, v2, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v4, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v3, vcc, s0, v0
+; NOSDWA-NEXT:    v_addc_u32_e32 v4, vcc, 0, v4, vcc
+; NOSDWA-NEXT:    flat_load_ubyte v2, v[1:2]
+; NOSDWA-NEXT:    flat_load_ubyte v3, v[3:4]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v2, v3
+; NOSDWA-NEXT:    flat_store_byte v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_i8:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v2, s7
+; VI-NEXT:    v_add_u32_e32 v1, vcc, s6, v0
+; VI-NEXT:    v_addc_u32_e32 v2, vcc, 0, v2, vcc
+; VI-NEXT:    v_mov_b32_e32 v4, s1
+; VI-NEXT:    v_add_u32_e32 v3, vcc, s0, v0
+; VI-NEXT:    v_addc_u32_e32 v4, vcc, 0, v4, vcc
+; VI-NEXT:    flat_load_ubyte v2, v[1:2]
+; VI-NEXT:    flat_load_ubyte v3, v[3:4]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v2, v2, v3
+; VI-NEXT:    flat_store_byte v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_i8:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_ubyte v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_ubyte v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_lo_u16_e32 v1, v1, v2
+; GFX9-NEXT:    global_store_byte v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_i8:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_ubyte v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_ubyte v2, v0, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_mul_lo_u16 v1, v1, v2
+; GFX10-NEXT:    global_store_byte v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr i8, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr i8, ptr addrspace(1) %inb, i32 %idx
   %a = load i8, ptr addrspace(1) %gepa, align 4
   %b = load i8, ptr addrspace(1) %gepb, align 4
   %mul = mul i8 %a, %b
   store i8 %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v2i8:
-; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI: v_mul_lo_u16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
-
-; GFX9-DAG: v_mul_lo_u16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
-; GFX9-DAG: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-
-; GFX9: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-
-; GFX10: v_lshlrev_b16 v{{[0-9]+}}, 8, v
-; GFX10: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 define amdgpu_kernel void @mul_v2i8(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v2i8:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 1, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_ushort v4, v[0:1]
+; NOSDWA-NEXT:    flat_load_ushort v2, v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v3, 8, v4
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v5, 8, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v4, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v3, v5
+; NOSDWA-NEXT:    v_and_b32_e32 v2, 0xff, v2
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v3, 8, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v3
+; NOSDWA-NEXT:    flat_store_short v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v2i8:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 1, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_ushort v4, v[0:1]
+; VI-NEXT:    flat_load_ushort v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v3, v4, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v2, v4, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    flat_store_short v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v2i8:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_ushort v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_ushort v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_lo_u16_e32 v3, v1, v2
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v1, v1, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; GFX9-NEXT:    v_or_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    global_store_short v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v2i8:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_ushort v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_ushort v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(1)
+; GFX10-NEXT:    v_lshrrev_b16 v0, 8, v1
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_lshrrev_b16 v3, 8, v2
+; GFX10-NEXT:    v_mul_lo_u16 v1, v1, v2
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_mul_lo_u16 v0, v0, v3
+; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
+; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    global_store_short v2, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <2 x i8>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <2 x i8>, ptr addrspace(1) %inb, i32 %idx
   %a = load <2 x i8>, ptr addrspace(1) %gepa, align 4
   %b = load <2 x i8>, ptr addrspace(1) %gepb, align 4
   %mul = mul <2 x i8> %a, %b
   store <2 x i8> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v4i8:
-; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-
 define amdgpu_kernel void @mul_v4i8(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v4i8:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    flat_load_dword v4, v[0:1]
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dword v2, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v4
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v5, 24, v4
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v6, 8, v4
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v8, 24, v2
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v9, 8, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v4, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v4, v6, v9
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v5, v5, v8
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v3, v7
+; NOSDWA-NEXT:    v_and_b32_e32 v2, 0xff, v2
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v4, 8, v4
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v5, 8, v5
+; NOSDWA-NEXT:    v_and_b32_e32 v3, 0xff, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v4
+; NOSDWA-NEXT:    v_or_b32_e32 v3, v3, v5
+; NOSDWA-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v3
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v4i8:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dword v4, v[0:1]
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v3, v4, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v5, v4, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT:    v_mul_lo_u16_sdwa v6, v4, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
+; VI-NEXT:    v_mul_lo_u16_sdwa v2, v4, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_or_b32_sdwa v3, v3, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v2, v2, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v4i8:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_lo_u16_e32 v3, v1, v2
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v4, v1, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v5, v1, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX9-NEXT:    v_or_b32_sdwa v2, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v4i8:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(1)
+; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 24, v1
+; GFX10-NEXT:    v_lshrrev_b16 v3, 8, v1
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_lshrrev_b16 v4, 8, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 24, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 16, v1
+; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
+; GFX10-NEXT:    v_mul_lo_u16 v1, v1, v2
+; GFX10-NEXT:    v_mul_lo_u16 v3, v3, v4
+; GFX10-NEXT:    v_mul_lo_u16 v0, v0, v5
+; GFX10-NEXT:    v_mul_lo_u16 v2, v6, v7
+; GFX10-NEXT:    v_lshlrev_b16 v3, 8, v3
+; GFX10-NEXT:    v_lshlrev_b16 v0, 8, v0
+; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT:    global_store_dword v2, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <4 x i8>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <4 x i8>, ptr addrspace(1) %inb, i32 %idx
   %a = load <4 x i8>, ptr addrspace(1) %gepa, align 4
   %b = load <4 x i8>, ptr addrspace(1) %gepb, align 4
   %mul = mul <4 x i8> %a, %b
   store <4 x i8> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}mul_v8i8:
-; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-; VI-DAG: v_mul_lo_u16_sdwa
-
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-; GFX9-DAG: v_mul_lo_u16_sdwa
-
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-; GFX10-DAG: v_mul_lo_u16
-
 define amdgpu_kernel void @mul_v8i8(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mul_v8i8:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; NOSDWA-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v4, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v5, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v6, 16, v0
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v7, 24, v0
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v8, 8, v0
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v9, 16, v1
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v10, 24, v1
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v11, 8, v1
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v12, 16, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v13, 24, v2
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v14, 8, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v15, 16, v3
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v16, 24, v3
+; NOSDWA-NEXT:    v_lshrrev_b16_e32 v17, 8, v3
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v1, v1, v3
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v0, v0, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v11, v17
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v10, v16
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v9, v9, v15
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v8, v8, v14
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v7, v7, v13
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v6, v6, v12
+; NOSDWA-NEXT:    v_and_b32_e32 v1, 0xff, v1
+; NOSDWA-NEXT:    v_and_b32_e32 v0, 0xff, v0
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v2, 8, v2
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v3, 8, v3
+; NOSDWA-NEXT:    v_and_b32_e32 v9, 0xff, v9
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v8, 8, v8
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v7, 8, v7
+; NOSDWA-NEXT:    v_and_b32_e32 v6, 0xff, v6
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v1, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v9, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v0, v8
+; NOSDWA-NEXT:    v_or_b32_e32 v3, v6, v7
+; NOSDWA-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v1, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v0, v3
+; NOSDWA-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mul_v8i8:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v4, s4
+; VI-NEXT:    v_mov_b32_e32 v5, s5
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v6, v1, v3
+; VI-NEXT:    v_mul_lo_u16_sdwa v7, v1, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT:    v_mul_lo_u16_sdwa v8, v1, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
+; VI-NEXT:    v_mul_lo_u16_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_mul_lo_u16_e32 v3, v0, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v9, v0, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; VI-NEXT:    v_mul_lo_u16_sdwa v10, v0, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
+; VI-NEXT:    v_mul_lo_u16_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_or_b32_sdwa v2, v6, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v1, v1, v8 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v3, v3, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v0, v0, v10 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mul_v8i8:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GFX9-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_mul_lo_u16_e32 v5, v1, v3
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v6, v1, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v7, v1, v3 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX9-NEXT:    v_mul_lo_u16_e32 v3, v0, v2
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v8, v0, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_1 src1_sel:BYTE_1
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v9, v0, v2 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:BYTE_3 src1_sel:BYTE_3
+; GFX9-NEXT:    v_mul_lo_u16_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GFX9-NEXT:    v_or_b32_sdwa v2, v5, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v3, v3, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v9 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mul_v8i8:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GFX10-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(1)
+; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
+; GFX10-NEXT:    v_lshrrev_b16 v6, 8, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v8, 24, v1
+; GFX10-NEXT:    v_lshrrev_b16 v9, 8, v1
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-NEXT:    v_lshrrev_b16 v12, 8, v2
+; GFX10-NEXT:    v_lshrrev_b16 v13, 8, v3
+; GFX10-NEXT:    v_lshrrev_b32_e32 v14, 24, v3
+; GFX10-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 16, v1
+; GFX10-NEXT:    v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v15, 16, v3
+; GFX10-NEXT:    v_mul_lo_u16 v1, v1, v3
+; GFX10-NEXT:    v_mul_lo_u16 v0, v0, v2
+; GFX10-NEXT:    v_mul_lo_u16 v2, v9, v13
+; GFX10-NEXT:    v_mul_lo_u16 v3, v8, v14
+; GFX10-NEXT:    v_mul_lo_u16 v6, v6, v12
+; GFX10-NEXT:    v_mul_lo_u16 v5, v5, v11
+; GFX10-NEXT:    v_mul_lo_u16 v7, v7, v15
+; GFX10-NEXT:    v_mul_lo_u16 v4, v4, v10
+; GFX10-NEXT:    v_lshlrev_b16 v2, 8, v2
+; GFX10-NEXT:    v_lshlrev_b16 v3, 8, v3
+; GFX10-NEXT:    v_lshlrev_b16 v6, 8, v6
+; GFX10-NEXT:    v_lshlrev_b16 v5, 8, v5
+; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v2, v7, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v3, v4, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <8 x i8>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <8 x i8>, ptr addrspace(1) %inb, i32 %idx
   %a = load <8 x i8>, ptr addrspace(1) %gepa, align 4
   %b = load <8 x i8>, ptr addrspace(1) %gepb, align 4
   %mul = mul <8 x i8> %a, %b
   store <8 x i8> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}sitofp_v2i16_to_v2f16:
-; NOSDWA-DAG: v_cvt_f16_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-DAG: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA-DAG: v_cvt_f16_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_cvt_f16_i16_sdwa
-
-; SDWA-DAG: v_cvt_f16_i16_e32 v{{[0-9]+}}, v{{[0-9]+}}
-; SDWA-DAG: v_cvt_f16_i16_sdwa v{{[0-9]+}}, v{{[0-9]+}} dst_sel:{{(WORD_1|DWORD)?}} dst_unused:UNUSED_PAD src0_sel:WORD_1
-
-; FIXME: Should be able to avoid or
 define amdgpu_kernel void @sitofp_v2i16_to_v2f16(
+; NOSDWA-LABEL: sitofp_v2i16_to_v2f16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s2
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s3
+; NOSDWA-NEXT:    flat_load_dword v2, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s1
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
+; NOSDWA-NEXT:    v_cvt_f16_i16_e32 v3, v3
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_cvt_f16_i16_e32 v2, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v3
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: sitofp_v2i16_to_v2f16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s2
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    flat_load_dword v2, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_cvt_f16_i16_sdwa v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
+; VI-NEXT:    v_cvt_f16_i16_e32 v2, v2
+; VI-NEXT:    v_or_b32_e32 v2, v2, v3
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: sitofp_v2i16_to_v2f16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_cvt_f16_i16_e32 v2, v1
+; GFX9-NEXT:    v_cvt_f16_i16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
+; GFX9-NEXT:    v_lshl_or_b32 v1, v1, 16, v2
+; GFX9-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: sitofp_v2i16_to_v2f16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    global_load_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_cvt_f16_i16_e32 v2, v1
+; GFX10-NEXT:    v_cvt_f16_i16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
+; GFX10-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10-NEXT:    v_lshl_or_b32 v1, v1, 16, v2
+; GFX10-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-NEXT:    s_endpgm
     ptr addrspace(1) %r,
     ptr addrspace(1) %a) #0 {
 entry:
   %a.val = load <2 x i16>, ptr addrspace(1) %a
   %r.val = sitofp <2 x i16> %a.val to <2 x half>
   store <2 x half> %r.val, ptr addrspace(1) %r
   ret void
 }
 
-
-; GCN-LABEL: {{^}}mac_v2half:
-; NOSDWA: v_lshrrev_b32_e32 v[[DST1:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v[[DST0:[0-9]+]], 16, v{{[0-9]+}}
-; NOSDWA: v_mac_f16_e32 v[[DST_MAC:[0-9]+]], v[[DST0]], v[[DST1]]
-; NOSDWA: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MAC]]
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v[[DST_SHL]]
-; NOSDWA-NOT: v_mac_f16_sdwa
-
-; VI: v_mac_f16_sdwa v[[DST_MAC:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI: v_lshlrev_b32_e32 v[[DST_SHL:[0-9]+]], 16, v[[DST_MAC]]
-
-; GFX9_10: v_pk_mul_f16 v[[DST_MUL:[0-9]+]], v{{[0-9]+}}, v[[SRC:[0-9]+]]
-; GFX9_10: v_pk_add_f16 v{{[0-9]+}}, v[[DST_MUL]], v[[SRC]]
-
 define amdgpu_kernel void @mac_v2half(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mac_v2half:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s6
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    flat_load_dword v2, v[2:3]
+; NOSDWA-NEXT:    flat_load_dword v3, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
+; NOSDWA-NEXT:    v_mac_f16_e32 v4, v5, v4
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; NOSDWA-NEXT:    v_mac_f16_e32 v2, v3, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v4
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mac_v2half:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mac_f16_sdwa v4, v3, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_lshlrev_b32_e32 v4, 16, v4
+; VI-NEXT:    v_mac_f16_e32 v2, v3, v2
+; VI-NEXT:    v_or_b32_e32 v2, v2, v4
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mac_v2half:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_f16 v1, v1, v2
+; GFX9-NEXT:    v_pk_add_f16 v1, v1, v2
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mac_v2half:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_f16 v1, v1, v2
+; GFX10-NEXT:    v_pk_add_f16 v1, v1, v2
+; GFX10-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %a = load <2 x half>, ptr addrspace(1) %ina, align 4
   %b = load <2 x half>, ptr addrspace(1) %inb, align 4
   %mul = fmul <2 x half> %a, %b
   %mac = fadd <2 x half> %mul, %b
   store <2 x half> %mac, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}immediate_mul_v2i16:
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-; VI-DAG: v_mov_b32_e32 v[[M321:[0-9]+]], 0x141
-; VI-DAG: v_mul_lo_u16_e32 v{{[0-9]+}}, 0x7b, v{{[0-9]+}}
-; VI-DAG: v_mul_lo_u16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v[[M321]] dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-
-; GFX9: s_mov_b32 s[[IMM:[0-9]+]], 0x141007b
-; GFX9: v_pk_mul_lo_u16 v{{[0-9]+}}, v{{[0-9]+}}, s[[IMM]]
-
-; GFX10: v_pk_mul_lo_u16 v{{[0-9]+}}, 0x141007b, v{{[0-9]+}}
-
 define amdgpu_kernel void @immediate_mul_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %in) #0 {
+; NOSDWA-LABEL: immediate_mul_v2i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s3
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    flat_load_dword v2, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s1
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, 0x7b, v2
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, 0x141, v2
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v3, v2
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: immediate_mul_v2i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; VI-NEXT:    v_mov_b32_e32 v3, 0x141
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s3
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    flat_load_dword v2, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v4, 0x7b, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v2, v2, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_e32 v2, v4, v2
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: immediate_mul_v2i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT:    v_mov_b32_e32 v1, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v0, v0, s[2:3]
+; GFX9-NEXT:    s_mov_b32 s2, 0x141007b
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_lo_u16 v0, v0, s2
+; GFX9-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: immediate_mul_v2i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    global_load_dword v0, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_lo_u16 v0, 0x141007b, v0
+; GFX10-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr <2 x i16>, ptr addrspace(1) %in, i32 %idx
   %a = load <2 x i16>, ptr addrspace(1) %gep, align 4
   %mul = mul <2 x i16> %a, <i16 123, i16 321>
   store <2 x i16> %mul, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; Double use of same src - should not convert it
-; GCN-LABEL: {{^}}mulmul_v2i16:
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_lshrrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_mul_lo_u16_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA: v_lshlrev_b32_e32 v{{[0-9]+}}, 16, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_mul_u32_u24_sdwa
-
-; VI: v_mul_lo_u16_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-
-; GFX9_10: v_pk_mul_lo_u16 v[[DST1:[0-9]+]], v{{[0-9]+}}, v{{[0-9]+}}
-; GFX9_10: v_pk_mul_lo_u16 v{{[0-9]+}}, v[[DST1]], v{{[0-9]+}}
-
 define amdgpu_kernel void @mulmul_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: mulmul_v2i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; NOSDWA-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; NOSDWA-NEXT:    flat_load_dword v2, v[2:3]
+; NOSDWA-NEXT:    flat_load_dword v3, v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v5, 16, v3
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v3, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v5, v5, v4
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v2, v3, v2
+; NOSDWA-NEXT:    v_mul_lo_u16_e32 v3, v5, v4
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v2, v3
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: mulmul_v2i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    v_lshlrev_b32_e32 v2, 2, v0
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; VI-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; VI-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    flat_load_dword v3, v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    s_waitcnt vmcnt(1)
+; VI-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_mul_lo_u16_e32 v5, v3, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v3, v3, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT:    v_mul_lo_u16_e32 v2, v5, v2
+; VI-NEXT:    v_mul_lo_u16_sdwa v3, v3, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; VI-NEXT:    v_or_b32_e32 v2, v2, v3
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: mulmul_v2i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_mul_lo_u16 v1, v1, v2
+; GFX9-NEXT:    v_pk_mul_lo_u16 v1, v1, v2
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: mulmul_v2i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_mul_lo_u16 v0, v1, v2
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    v_pk_mul_lo_u16 v0, v0, v2
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
   %gepa = getelementptr <2 x i16>, ptr addrspace(1) %ina, i32 %idx
   %gepb = getelementptr <2 x i16>, ptr addrspace(1) %inb, i32 %idx
   %a = load <2 x i16>, ptr addrspace(1) %gepa, align 4
   %b = load <2 x i16>, ptr addrspace(1) %gepb, align 4
   %mul = mul <2 x i16> %a, %b
   %mul2 = mul <2 x i16> %mul, %b
   store <2 x i16> %mul2, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; GCN-LABEL: {{^}}add_bb_v2i16:
-; NOSDWA-NOT: v_add_{{(_co)?}}_u32_sdwa
-
-; VI:      v_add_u32_sdwa v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
-; VI-NEXT: v_add_u32_e32 v{{[0-9]+}}, vcc, v{{[0-9]+}}, v{{[0-9]+}}
-; VI-NEXT: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-
-; GFX9_10: v_pk_add_u16 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-
 define amdgpu_kernel void @add_bb_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %ina, ptr addrspace(1) %inb) #0 {
+; NOSDWA-LABEL: add_bb_v2i16:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; NOSDWA-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s6
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s7
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s1
+; NOSDWA-NEXT:    flat_load_dword v1, v[0:1]
+; NOSDWA-NEXT:    flat_load_dword v2, v[2:3]
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s4
+; NOSDWA-NEXT:    s_waitcnt vmcnt(1)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v4, 16, v2
+; NOSDWA-NEXT:    v_add_u32_e32 v1, vcc, v1, v2
+; NOSDWA-NEXT:    v_add_u32_e32 v2, vcc, v3, v4
+; NOSDWA-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; NOSDWA-NEXT:    v_or_b32_e32 v2, v1, v2
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s5
+; NOSDWA-NEXT:    flat_store_dword v[0:1], v2
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: add_bb_v2i16:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s6
+; VI-NEXT:    v_mov_b32_e32 v1, s7
+; VI-NEXT:    v_mov_b32_e32 v2, s0
+; VI-NEXT:    v_mov_b32_e32 v3, s1
+; VI-NEXT:    flat_load_dword v1, v[0:1]
+; VI-NEXT:    flat_load_dword v2, v[2:3]
+; VI-NEXT:    v_mov_b32_e32 v0, s4
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_add_u32_sdwa v3, vcc, v1, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; VI-NEXT:    v_add_u32_e32 v1, vcc, v1, v2
+; VI-NEXT:    v_or_b32_sdwa v2, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    v_mov_b32_e32 v1, s5
+; VI-NEXT:    flat_store_dword v[0:1], v2
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: add_bb_v2i16:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX9-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX9-NEXT:    v_mov_b32_e32 v0, 0
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX9-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_pk_add_u16 v1, v1, v2
+; GFX9-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: add_bb_v2i16:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    global_load_dword v1, v0, s[6:7]
+; GFX10-NEXT:    global_load_dword v2, v0, s[2:3]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_pk_add_u16 v1, v1, v2
+; GFX10-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-NEXT:    s_endpgm
 entry:
   %a = load <2 x i16>, ptr addrspace(1) %ina, align 4
   %b = load <2 x i16>, ptr addrspace(1) %inb, align 4
   br label %add_label
 add_label:
   %add = add <2 x i16> %a, %b
   br label %store_label
 store_label:
   store <2 x i16> %add, ptr addrspace(1) %out, align 4
   ret void
 }
 
-
-; Check that "pulling out" SDWA operands works correctly.
-; GCN-LABEL: {{^}}pulled_out_test:
-; NOSDWA-DAG: v_and_b32_e32 v{{[0-9]+}}, 0xff, v{{[0-9]+}}
-; NOSDWA-DAG: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA-DAG: v_and_b32_e32 v{{[0-9]+}}, 0xff, v{{[0-9]+}}
-; NOSDWA-DAG: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}}
-; NOSDWA-NOT: v_and_b32_sdwa
-; NOSDWA-NOT: v_or_b32_sdwa
-
-; VI-DAG: v_and_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX9-DAG: v_and_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, s{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX10-DAG: v_and_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX89-DAG: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-;
-; GFX10-DAG: v_lshrrev_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-;
-; VI-DAG: v_and_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX9-DAG: v_and_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, s{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX10-DAG: v_and_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX89-DAG: v_lshlrev_b16_e32 v{{[0-9]+}}, 8, v{{[0-9]+}}
-;
-; GFX10-DAG: v_lshrrev_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-;
-; GFX89: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-;
-; GFX10: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX10: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
-; GFX10: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-; GFX10: v_or_b32_sdwa v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
-
 define amdgpu_kernel void @pulled_out_test(ptr addrspace(1) %sourceA, ptr addrspace(1) %destValues) #0 {
+; NOSDWA-LABEL: pulled_out_test:
+; NOSDWA:       ; %bb.0: ; %entry
+; NOSDWA-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; NOSDWA-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSDWA-NEXT:    v_mov_b32_e32 v0, s0
+; NOSDWA-NEXT:    v_mov_b32_e32 v1, s1
+; NOSDWA-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; NOSDWA-NEXT:    v_mov_b32_e32 v2, s2
+; NOSDWA-NEXT:    v_mov_b32_e32 v3, s3
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    v_lshrrev_b64 v[4:5], 24, v[0:1]
+; NOSDWA-NEXT:    v_and_b32_e32 v6, 0xff, v0
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v7, 8, v0
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
+; NOSDWA-NEXT:    v_and_b32_e32 v5, 0xff, v1
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v8, 8, v1
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v9, 24, v1
+; NOSDWA-NEXT:    v_lshrrev_b32_e32 v1, 16, v1
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v7, 8, v7
+; NOSDWA-NEXT:    v_and_b32_e32 v0, 0xff, v0
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v8, 8, v8
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v9, 8, v9
+; NOSDWA-NEXT:    v_and_b32_e32 v1, 0xff, v1
+; NOSDWA-NEXT:    v_lshlrev_b16_e32 v4, 8, v4
+; NOSDWA-NEXT:    v_or_b32_e32 v6, v6, v7
+; NOSDWA-NEXT:    v_or_b32_e32 v5, v5, v8
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v1, v9
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v0, v4
+; NOSDWA-NEXT:    v_and_b32_e32 v6, 0xffff, v6
+; NOSDWA-NEXT:    v_and_b32_e32 v4, 0xffff, v5
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; NOSDWA-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; NOSDWA-NEXT:    v_or_b32_e32 v0, v6, v0
+; NOSDWA-NEXT:    v_or_b32_e32 v1, v4, v1
+; NOSDWA-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; NOSDWA-NEXT:    s_endpgm
+;
+; VI-LABEL: pulled_out_test:
+; VI:       ; %bb.0: ; %entry
+; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; VI-NEXT:    v_mov_b32_e32 v6, 8
+; VI-NEXT:    v_mov_b32_e32 v7, 0xff
+; VI-NEXT:    s_waitcnt lgkmcnt(0)
+; VI-NEXT:    v_mov_b32_e32 v0, s0
+; VI-NEXT:    v_mov_b32_e32 v1, s1
+; VI-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    v_mov_b32_e32 v2, s2
+; VI-NEXT:    v_mov_b32_e32 v3, s3
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    v_lshrrev_b64 v[4:5], 24, v[0:1]
+; VI-NEXT:    v_lshrrev_b32_sdwa v8, v6, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; VI-NEXT:    v_lshrrev_b32_sdwa v6, v6, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; VI-NEXT:    v_lshrrev_b32_e32 v9, 24, v1
+; VI-NEXT:    v_and_b32_sdwa v5, v0, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT:    v_and_b32_sdwa v7, v1, v7 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v1, v1, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_lshlrev_b16_e32 v6, 8, v9
+; VI-NEXT:    v_lshlrev_b16_e32 v4, 8, v4
+; VI-NEXT:    v_or_b32_sdwa v0, v0, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v6, v7, v6 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v4, v5, v4 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v0, v0, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    v_or_b32_sdwa v1, v1, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; VI-NEXT:    s_endpgm
+;
+; GFX9-LABEL: pulled_out_test:
+; GFX9:       ; %bb.0: ; %entry
+; GFX9-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX9-NEXT:    v_mov_b32_e32 v4, 0
+; GFX9-NEXT:    v_mov_b32_e32 v5, 8
+; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX9-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
+; GFX9-NEXT:    s_movk_i32 s0, 0xff
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    v_lshrrev_b64 v[2:3], 24, v[0:1]
+; GFX9-NEXT:    v_lshrrev_b32_sdwa v6, v5, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX9-NEXT:    v_lshrrev_b32_sdwa v5, v5, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX9-NEXT:    v_lshrrev_b32_e32 v7, 24, v1
+; GFX9-NEXT:    v_and_b32_sdwa v3, v0, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX9-NEXT:    v_and_b32_sdwa v8, v1, s0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_lshlrev_b16_e32 v5, 8, v7
+; GFX9-NEXT:    v_lshlrev_b16_e32 v2, 8, v2
+; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v5, v8, v5 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT:    v_or_b32_sdwa v1, v1, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX9-NEXT:    global_store_dwordx2 v4, v[0:1], s[2:3]
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: pulled_out_test:
+; GFX10:       ; %bb.0: ; %entry
+; GFX10-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-NEXT:    v_mov_b32_e32 v5, 8
+; GFX10-NEXT:    v_mov_b32_e32 v6, 0xff
+; GFX10-NEXT:    v_mov_b32_e32 v7, 24
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    global_load_dwordx2 v[0:1], v4, s[0:1]
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    v_lshrrev_b64 v[2:3], 24, v[0:1]
+; GFX10-NEXT:    v_lshrrev_b32_sdwa v3, v5, v0 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT:    v_and_b32_sdwa v8, v0, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX10-NEXT:    v_lshrrev_b32_sdwa v5, v5, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT:    v_lshrrev_b32_sdwa v7, v7, v1 dst_sel:BYTE_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT:    v_and_b32_sdwa v6, v1, v6 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX10-NEXT:    v_lshlrev_b16 v2, 8, v2
+; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v5 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v3, v6, v7 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v2, v8, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT:    v_or_b32_sdwa v0, v0, v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX10-NEXT:    global_store_dwordx2 v4, v[0:1], s[2:3]
+; GFX10-NEXT:    s_endpgm
 entry:
   %idxprom = ashr exact i64 15, 32
   %arrayidx = getelementptr inbounds <8 x i8>, ptr addrspace(1) %sourceA, i64 %idxprom
   %tmp = load <8 x i8>, ptr addrspace(1) %arrayidx, align 8
 
   %tmp1 = extractelement <8 x i8> %tmp, i32 0
   %tmp2 = extractelement <8 x i8> %tmp, i32 1
   %tmp3 = extractelement <8 x i8> %tmp, i32 2
entry:
   %tmp18 = shufflevector <2 x i8> %tmp14, <2 x i8> %tmp16, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %tmp19 = shufflevector <4 x i8> %tmp17, <4 x i8> %tmp18, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 
   %arrayidx5 = getelementptr inbounds <8 x i8>, ptr addrspace(1) %destValues, i64 %idxprom
   store <8 x i8> %tmp19, ptr addrspace(1) %arrayidx5, align 8
   ret void
 }
 
-; GCN-LABEL: {{^}}sdwa_crash_inlineasm_def:
-; GCN: s_mov_b32 s{{[0-9]+}}, 0xffff
-; GCN: v_and_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}
+define amdgpu_kernel void @sdwa_crash_inlineasm_def() #0 {
+; NOSDWA-LABEL: sdwa_crash_inlineasm_def:
+; NOSDWA:       ; %bb.0: ; %bb
+; NOSDWA-NEXT:    s_mov_b32 s0, 0xffff
+; NOSDWA-NEXT:    ;;#ASMSTART
+; NOSDWA-NEXT:    v_and_b32_e32 v0, s0, v0
+; NOSDWA-NEXT:    ;;#ASMEND
+; NOSDWA-NEXT:    v_or_b32_e32 v0, 0x10000, v0
+; NOSDWA-NEXT:    s_and_b64 vcc, exec, -1
+; NOSDWA-NEXT:  .LBB21_1: ; %bb1
+; NOSDWA-NEXT:    ; =>This Inner Loop Header: Depth=1
+; NOSDWA-NEXT:    flat_store_dwordx2 v[0:1], v[0:1]
+; NOSDWA-NEXT:    s_waitcnt vmcnt(0)
+; NOSDWA-NEXT:    s_mov_b64 vcc, vcc
+; NOSDWA-NEXT:    s_cbranch_vccnz .LBB21_1
+; NOSDWA-NEXT:  ; %bb.2: ; %DummyReturnBlock
+; NOSDWA-NEXT:    s_endpgm
 ;
-; TODO: Why is the constant not peepholed into the v_or_b32_e32?
+; VI-LABEL: sdwa_crash_inlineasm_def:
+; VI:       ; %bb.0: ; %bb
+; VI-NEXT:    s_mov_b32 s0, 0xffff
+; VI-NEXT:    ;;#ASMSTART
+; VI-NEXT:    v_and_b32_e32 v0, s0, v0
+; VI-NEXT:    ;;#ASMEND
+; VI-NEXT:    v_or_b32_e32 v0, 0x10000, v0
+; VI-NEXT:    s_and_b64 vcc, exec, -1
+; VI-NEXT:  .LBB21_1: ; %bb1
+; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
+; VI-NEXT:    flat_store_dwordx2 v[0:1], v[0:1]
+; VI-NEXT:    s_waitcnt vmcnt(0)
+; VI-NEXT:    s_mov_b64 vcc, vcc
+; VI-NEXT:    s_cbranch_vccnz .LBB21_1
+; VI-NEXT:  ; %bb.2: ; %DummyReturnBlock
+; VI-NEXT:    s_endpgm
 ;
-; NOSDWA: v_or_b32_e32 v{{[0-9]+}}, 0x10000,
-; SDWA: v_or_b32_e32 v{{[0-9]+}}, 0x10000,
-define amdgpu_kernel void @sdwa_crash_inlineasm_def() #0 {
+; GFX9-LABEL: sdwa_crash_inlineasm_def:
+; GFX9:       ; %bb.0: ; %bb
+; GFX9-NEXT:    s_mov_b32 s0, 0xffff
+; GFX9-NEXT:    ;;#ASMSTART
+; GFX9-NEXT:    v_and_b32_e32 v0, s0, v0
+; GFX9-NEXT:    ;;#ASMEND
+; GFX9-NEXT:    v_or_b32_e32 v0, 0x10000, v0
+; GFX9-NEXT:    s_and_b64 vcc, exec, -1
+; GFX9-NEXT:  .LBB21_1: ; %bb1
+; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX9-NEXT:    global_store_dwordx2 v[0:1], v[0:1], off
+; GFX9-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-NEXT:    s_mov_b64 vcc, vcc
+; GFX9-NEXT:    s_cbranch_vccnz .LBB21_1
+; GFX9-NEXT:  ; %bb.2: ; %DummyReturnBlock
+; GFX9-NEXT:    s_endpgm
+;
+; GFX10-LABEL: sdwa_crash_inlineasm_def:
+; GFX10:       ; %bb.0: ; %bb
+; GFX10-NEXT:    s_mov_b32 s0, 0xffff
+; GFX10-NEXT:    ;;#ASMSTART
+; GFX10-NEXT:    v_and_b32_e32 v0, s0, v0
+; GFX10-NEXT:    ;;#ASMEND
+; GFX10-NEXT:    v_or_b32_e32 v0, 0x10000, v0
+; GFX10-NEXT:    s_mov_b32 vcc_lo, exec_lo
+; GFX10-NEXT:  .LBB21_1: ; %bb1
+; GFX10-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX10-NEXT:    global_store_dwordx2 v[0:1], v[0:1], off
+; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT:    s_cbranch_vccnz .LBB21_1
+; GFX10-NEXT:  ; %bb.2: ; %DummyReturnBlock
+; GFX10-NEXT:    s_endpgm
 bb:
   br label %bb1
 
 bb1:                                              ; preds = %bb11, %bb
   %tmp = phi <2 x i32> [ %tmp12, %bb11 ], [ undef, %bb ]
   br i1 true, label %bb2, label %bb11
 
 bb2:                                              ; preds = %bb1

llvm/test/CodeGen/AMDGPU/sgpr-spills-split-regalloc.ll

 ; GCN-NEXT:    v_writelane_b32 v4, s34, 0
 ; GCN-NEXT:    v_writelane_b32 v4, s35, 1
 ; GCN-NEXT:    v_writelane_b32 v4, s36, 2
 ; GCN-NEXT:    v_writelane_b32 v4, s37, 3
 ; GCN-NEXT:    v_mov_b32_e32 v5, v3
 ; GCN-NEXT:    v_mov_b32_e32 v3, v1
 ; GCN-NEXT:    ; implicit-def: $sgpr4
 ; GCN-NEXT:    ; implicit-def: $sgpr4
-; GCN-NEXT:    ; kill: def $vgpr3 killed $vgpr3 killed $exec
 ; GCN-NEXT:    ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1 killed $exec
 ; GCN-NEXT:    v_mov_b32_e32 v1, v3
 ; GCN-NEXT:    ; implicit-def: $sgpr4
 ; GCN-NEXT:    ; implicit-def: $sgpr4
-; GCN-NEXT:    ; kill: def $vgpr5 killed $vgpr5 killed $exec
 ; GCN-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
 ; GCN-NEXT:    v_mov_b32_e32 v3, v5
 ; GCN-NEXT:    ; implicit-def: $sgpr4_sgpr5
 ; GCN-NEXT:    ; implicit-def: $sgpr4_sgpr5
 ; GCN-NEXT:    flat_load_dwordx4 v[5:8], v[2:3]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    buffer_store_dword v5, off, s[0:3], s32 offset:448 ; 4-byte Folded Spill
 ; GCN-NEXT:    s_waitcnt vmcnt(0)

llvm/test/CodeGen/AMDGPU/shl.ll

 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    s_mov_b32 s8, s2
 ; SI-NEXT:    s_mov_b32 s9, s3
 ; SI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; SI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
 ; SI-NEXT:    s_mov_b32 s4, s0
 ; SI-NEXT:    s_mov_b32 s5, s1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
-; SI-NEXT:    v_lshlrev_b32_e32 v3, v7, v3
-; SI-NEXT:    v_lshlrev_b32_e32 v2, v6, v2
-; SI-NEXT:    v_lshlrev_b32_e32 v1, v5, v1
+; SI-NEXT:    v_lshl_b32_e32 v3, v3, v7
+; SI-NEXT:    v_lshl_b32_e32 v2, v2, v6
+; SI-NEXT:    v_lshl_b32_e32 v1, v1, v5
 ; SI-NEXT:    v_lshl_b32_e32 v0, v0, v4
 ; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: shl_v4i32:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)

 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    s_mov_b32 s8, s2
 ; SI-NEXT:    s_mov_b32 s9, s3
 ; SI-NEXT:    buffer_load_ushort v0, off, s[8:11], 0
 ; SI-NEXT:    buffer_load_ushort v1, off, s[8:11], 0 offset:2
 ; SI-NEXT:    s_mov_b32 s4, s0
 ; SI-NEXT:    s_mov_b32 s5, s1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
-; SI-NEXT:    v_lshlrev_b32_e32 v0, v1, v0
+; SI-NEXT:    v_lshl_b32_e32 v0, v0, v1
 ; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
 ; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: shl_i16:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; VI-NEXT:    s_mov_b32 s7, 0xf000
 ; VI-NEXT:    s_mov_b32 s6, -1

llvm/test/CodeGen/AMDGPU/sra.ll

 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    s_mov_b32 s8, s2
 ; SI-NEXT:    s_mov_b32 s9, s3
 ; SI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; SI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
 ; SI-NEXT:    s_mov_b32 s4, s0
 ; SI-NEXT:    s_mov_b32 s5, s1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
-; SI-NEXT:    v_ashrrev_i32_e32 v3, v7, v3
-; SI-NEXT:    v_ashrrev_i32_e32 v2, v6, v2
-; SI-NEXT:    v_ashrrev_i32_e32 v1, v5, v1
+; SI-NEXT:    v_ashr_i32_e32 v3, v3, v7
+; SI-NEXT:    v_ashr_i32_e32 v2, v2, v6
+; SI-NEXT:    v_ashr_i32_e32 v1, v1, v5
 ; SI-NEXT:    v_ashr_i32_e32 v0, v0, v4
 ; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: ashr_v4i32:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; VI-NEXT:    s_mov_b32 s7, 0xf000

llvm/test/CodeGen/AMDGPU/srem64.ll

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s1, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s0, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s1, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s0, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s1, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s0, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v6
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v4, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s1, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s0, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GCN-NEXT:    v_mad_f32 v3, -v4, v1, v3
 ; GCN-NEXT:    v_cvt_i32_f32_e32 v4, v4
 ; GCN-NEXT:    v_cmp_ge_f32_e64 vcc, |v3|, |v1|
 ; GCN-NEXT:    v_cndmask_b32_e32 v1, 0, v5, vcc
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    v_add_i32_e32 v1, vcc, v4, v1
 ; GCN-NEXT:    v_mul_lo_u32 v0, v1, v0
 ; GCN-NEXT:    s_mov_b32 s1, s5
-; GCN-NEXT:    v_subrev_i32_e32 v0, vcc, v0, v2
+; GCN-NEXT:    v_sub_i32_e32 v0, vcc, v2, v0
 ; GCN-NEXT:    v_bfe_i32 v0, v0, 0, 24
 ; GCN-NEXT:    v_ashrrev_i32_e32 v1, 31, v0
 ; GCN-NEXT:    buffer_store_dword v0, off, s[0:3], 0
 ; GCN-NEXT:    buffer_store_short v1, off, s[0:3], 0 offset:4
 ; GCN-NEXT:    s_endpgm
 ;
 ; GCN-IR-LABEL: s_test_srem24_48:
 ; GCN-IR:       ; %bb.0: ; %_udiv-special-cases

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s2, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s2, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s3, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s2, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s2, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s2, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s3, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s2, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

llvm/test/CodeGen/AMDGPU/srl.ll

 ; SI-NEXT:    s_mov_b32 s11, s7
 ; SI-NEXT:    s_waitcnt lgkmcnt(0)
 ; SI-NEXT:    s_mov_b32 s8, s2
 ; SI-NEXT:    s_mov_b32 s9, s3
 ; SI-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
 ; SI-NEXT:    s_mov_b32 s4, s0
 ; SI-NEXT:    s_mov_b32 s5, s1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
-; SI-NEXT:    v_lshrrev_b32_e32 v0, v1, v0
+; SI-NEXT:    v_lshr_b32_e32 v0, v0, v1
 ; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: lshr_i32:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_load_dwordx2 s[4:5], s[2:3], 0x0

 ; SI-NEXT:    s_mov_b32 s8, s2
 ; SI-NEXT:    s_mov_b32 s9, s3
 ; SI-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
 ; SI-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
 ; SI-NEXT:    s_mov_b32 s4, s0
 ; SI-NEXT:    s_mov_b32 s5, s1
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_lshr_b32_e32 v3, v3, v7
-; SI-NEXT:    v_lshrrev_b32_e32 v2, v6, v2
-; SI-NEXT:    v_lshrrev_b32_e32 v1, v5, v1
-; SI-NEXT:    v_lshrrev_b32_e32 v0, v4, v0
+; SI-NEXT:    v_lshr_b32_e32 v2, v2, v6
+; SI-NEXT:    v_lshr_b32_e32 v1, v1, v5
+; SI-NEXT:    v_lshr_b32_e32 v0, v0, v4
 ; SI-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
 ; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: lshr_v4i32:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x24
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    s_load_dwordx8 s[0:7], s[10:11], 0x0

llvm/test/CodeGen/AMDGPU/sub.ll

-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SI %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,VI,GFX89 %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX9,GFX89 %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN1 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN2 %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GCN3 %s
 
 declare i32 @llvm.amdgcn.workitem.id.x() nounwind readnone speculatable
 
 ; GCN-LABEL: {{^}}s_sub_i32:
 ; GCN: s_load_dwordx4 s[[[#LOAD:]]:{{[0-9]+}}]
 ; GCN: s_sub_i32 s{{[0-9]+}}, s[[#LOAD + 2]], s[[#LOAD + 3]]
 define amdgpu_kernel void @s_sub_i32(ptr addrspace(1) %out, i32 %a, i32 %b) {
+; GCN1-LABEL: s_sub_i32:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_sub_i32 s0, s2, s3
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_mov_b32_e32 v0, s0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: s_sub_i32:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    s_sub_i32 s2, s2, s3
+; GCN2-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    v_mov_b32_e32 v2, s2
+; GCN2-NEXT:    flat_store_dword v[0:1], v2
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: s_sub_i32:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v0, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    s_sub_i32 s2, s2, s3
+; GCN3-NEXT:    v_mov_b32_e32 v1, s2
+; GCN3-NEXT:    global_store_dword v0, v1, s[0:1]
+; GCN3-NEXT:    s_endpgm
   %result = sub i32 %a, %b
   store i32 %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}s_sub_imm_i32:
 ; GCN: s_load_dword [[A:s[0-9]+]]
 ; GCN: s_sub_i32 s{{[0-9]+}}, 0x4d2, [[A]]
 define amdgpu_kernel void @s_sub_imm_i32(ptr addrspace(1) %out, i32 %a) {
+; GCN1-LABEL: s_sub_imm_i32:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dword s4, s[0:1], 0xb
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_sub_i32 s4, 0x4d2, s4
+; GCN1-NEXT:    v_mov_b32_e32 v0, s4
+; GCN1-NEXT:    buffer_store_dword v0, off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: s_sub_imm_i32:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dword s2, s[0:1], 0x2c
+; GCN2-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    s_sub_i32 s2, 0x4d2, s2
+; GCN2-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    v_mov_b32_e32 v2, s2
+; GCN2-NEXT:    flat_store_dword v[0:1], v2
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: s_sub_imm_i32:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GCN3-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v0, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    s_sub_i32 s0, 0x4d2, s4
+; GCN3-NEXT:    v_mov_b32_e32 v1, s0
+; GCN3-NEXT:    global_store_dword v0, v1, s[2:3]
+; GCN3-NEXT:    s_endpgm
   %result = sub i32 1234, %a
   store i32 %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_i32:
 ; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_i32:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_i32:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v0, s2
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v1
+; GCN2-NEXT:    flat_store_dword v[2:3], v0
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_i32:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v2, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx2 v[0:1], v2, s[2:3]
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_u32_e32 v0, v0, v1
+; GCN3-NEXT:    global_store_dword v2, v0, s[0:1]
+; GCN3-NEXT:    s_endpgm
   %b_ptr = getelementptr i32, ptr addrspace(1) %in, i32 1
   %a = load i32, ptr addrspace(1) %in
   %b = load i32, ptr addrspace(1) %b_ptr
   %result = sub i32 %a, %b
   store i32 %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_imm_i32:
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc, 0x7b, v{{[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+}}, 0x7b, v{{[0-9]+}}
 define amdgpu_kernel void @test_sub_imm_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_imm_i32:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, 0x7b, v0
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_imm_i32:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v0, s2
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    flat_load_dword v2, v[0:1]
+; GCN2-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u32_e32 v2, vcc, 0x7b, v2
+; GCN2-NEXT:    flat_store_dword v[0:1], v2
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_imm_i32:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v0, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dword v1, v0, s[2:3]
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_u32_e32 v1, 0x7b, v1
+; GCN3-NEXT:    global_store_dword v0, v1, s[0:1]
+; GCN3-NEXT:    s_endpgm
   %a = load i32, ptr addrspace(1) %in
   %result = sub i32 123, %a
   store i32 %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_v2i32:
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_v2i32:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v1, vcc, v1, v3
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
+; GCN1-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_v2i32:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v0, s2
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCN2-NEXT:    v_mov_b32_e32 v4, s0
+; GCN2-NEXT:    v_mov_b32_e32 v5, s1
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u32_e32 v1, vcc, v1, v3
+; GCN2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v2
+; GCN2-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_v2i32:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v4, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_u32_e32 v1, v1, v3
+; GCN3-NEXT:    v_sub_u32_e32 v0, v0, v2
+; GCN3-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
+; GCN3-NEXT:    s_endpgm
   %b_ptr = getelementptr <2 x i32>, ptr addrspace(1) %in, i32 1
   %a = load <2 x i32>, ptr addrspace(1) %in
   %b = load <2 x i32>, ptr addrspace(1) %b_ptr
   %result = sub <2 x i32> %a, %b
   store <2 x i32> %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_v4i32:
 ; SI: v_subrev_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 ; SI: v_sub_i32_e32 v{{[0-9]+, vcc, v[0-9]+, v[0-9]+}}
 
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; GFX9: v_sub_u32_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_v4i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_v4i32:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s10, s6
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s8, s2
+; GCN1-NEXT:    s_mov_b32 s9, s3
+; GCN1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; GCN1-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0 offset:16
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v3, vcc, v3, v7
+; GCN1-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
+; GCN1-NEXT:    v_sub_i32_e32 v1, vcc, v1, v5
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
+; GCN1-NEXT:    buffer_store_dwordx4 v[0:3], off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_v4i32:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v0, s2
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    s_add_u32 s2, s2, 16
+; GCN2-NEXT:    s_addc_u32 s3, s3, 0
+; GCN2-NEXT:    v_mov_b32_e32 v5, s3
+; GCN2-NEXT:    v_mov_b32_e32 v4, s2
+; GCN2-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCN2-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; GCN2-NEXT:    v_mov_b32_e32 v8, s0
+; GCN2-NEXT:    v_mov_b32_e32 v9, s1
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u32_e32 v3, vcc, v3, v7
+; GCN2-NEXT:    v_sub_u32_e32 v2, vcc, v2, v6
+; GCN2-NEXT:    v_sub_u32_e32 v1, vcc, v1, v5
+; GCN2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v4
+; GCN2-NEXT:    flat_store_dwordx4 v[8:9], v[0:3]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_v4i32:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v8, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx4 v[0:3], v8, s[2:3] offset:16
+; GCN3-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_u32_e32 v3, v7, v3
+; GCN3-NEXT:    v_sub_u32_e32 v2, v6, v2
+; GCN3-NEXT:    v_sub_u32_e32 v1, v5, v1
+; GCN3-NEXT:    v_sub_u32_e32 v0, v4, v0
+; GCN3-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GCN3-NEXT:    s_endpgm
   %b_ptr = getelementptr <4 x i32>, ptr addrspace(1) %in, i32 1
   %a = load <4 x i32>, ptr addrspace(1) %in
   %b = load <4 x i32>, ptr addrspace(1) %b_ptr
   %result = sub <4 x i32> %a, %b
   store <4 x i32> %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_i16:
 ; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc,
 ; GFX89: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 define amdgpu_kernel void @test_sub_i16(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_i16:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s10, 0
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; GCN1-NEXT:    v_mov_b32_e32 v1, 0
+; GCN1-NEXT:    buffer_load_ushort v2, v[0:1], s[8:11], 0 addr64 glc
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    buffer_load_ushort v0, v[0:1], s[8:11], 0 addr64 offset:2 glc
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v2, v0
+; GCN1-NEXT:    buffer_store_short v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_i16:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; GCN2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GCN2-NEXT:    v_add_u32_e32 v2, vcc, 2, v0
+; GCN2-NEXT:    v_addc_u32_e32 v3, vcc, 0, v1, vcc
+; GCN2-NEXT:    flat_load_ushort v4, v[0:1] glc
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    flat_load_ushort v2, v[2:3] glc
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    v_sub_u16_e32 v2, v4, v2
+; GCN2-NEXT:    flat_store_short v[0:1], v2
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_i16:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_ushort v1, v0, s[2:3] glc
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    global_load_ushort v2, v0, s[2:3] offset:2 glc
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_mov_b32_e32 v0, 0
+; GCN3-NEXT:    v_sub_u16_e32 v1, v1, v2
+; GCN3-NEXT:    global_store_short v0, v1, s[0:1]
+; GCN3-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr i16, ptr addrspace(1) %in, i32 %tid
   %b_ptr = getelementptr i16, ptr addrspace(1) %gep, i32 1
   %a = load volatile i16, ptr addrspace(1) %gep
   %b = load volatile i16, ptr addrspace(1) %b_ptr
   %result = sub i16 %a, %b
   store i16 %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_v2i16:
 ; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
 
 ; GFX9: v_pk_sub_i16
 define amdgpu_kernel void @test_sub_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_v2i16:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s10, 0
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; GCN1-NEXT:    v_mov_b32_e32 v1, 0
+; GCN1-NEXT:    buffer_load_dwordx2 v[0:1], v[0:1], s[8:11], 0 addr64
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
+; GCN1-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
+; GCN1-NEXT:    v_sub_i32_e32 v1, vcc, v2, v3
+; GCN1-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GCN1-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GCN1-NEXT:    v_or_b32_e32 v0, v0, v1
+; GCN1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_v2i16:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; GCN2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GCN2-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; GCN2-NEXT:    v_mov_b32_e32 v2, s0
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u16_e32 v4, v0, v1
+; GCN2-NEXT:    v_sub_u16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GCN2-NEXT:    v_or_b32_e32 v0, v4, v0
+; GCN2-NEXT:    flat_store_dword v[2:3], v0
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_v2i16:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
+; GCN3-NEXT:    v_mov_b32_e32 v2, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx2 v[0:1], v0, s[2:3]
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_pk_sub_i16 v0, v0, v1
+; GCN3-NEXT:    global_store_dword v2, v0, s[0:1]
+; GCN3-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr <2 x i16>, ptr addrspace(1) %in, i32 %tid
   %b_ptr = getelementptr <2 x i16>, ptr addrspace(1) %gep, i16 1
   %a = load <2 x i16>, ptr addrspace(1) %gep
   %b = load <2 x i16>, ptr addrspace(1) %b_ptr
   %result = sub <2 x i16> %a, %b
   store <2 x i16> %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}test_sub_v4i16:
 ; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_e32 v{{[0-9]+, v[0-9]+, v[0-9]+}}
 ; VI: v_sub_u16_sdwa v{{[0-9]+, v[0-9]+, v[0-9]+}}
 
 ; GFX9: v_pk_sub_i16
 ; GFX9: v_pk_sub_i16
 define amdgpu_kernel void @test_sub_v4i16(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; GCN1-LABEL: test_sub_v4i16:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s7, 0xf000
+; GCN1-NEXT:    s_mov_b32 s10, 0
+; GCN1-NEXT:    s_mov_b32 s11, s7
+; GCN1-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; GCN1-NEXT:    v_mov_b32_e32 v1, 0
+; GCN1-NEXT:    buffer_load_dwordx4 v[0:3], v[0:1], s[8:11], 0 addr64
+; GCN1-NEXT:    s_mov_b32 s6, -1
+; GCN1-NEXT:    s_mov_b32 s4, s0
+; GCN1-NEXT:    s_mov_b32 s5, s1
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
+; GCN1-NEXT:    v_lshrrev_b32_e32 v5, 16, v1
+; GCN1-NEXT:    v_lshrrev_b32_e32 v6, 16, v2
+; GCN1-NEXT:    v_lshrrev_b32_e32 v7, 16, v3
+; GCN1-NEXT:    v_sub_i32_e32 v1, vcc, v1, v3
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
+; GCN1-NEXT:    v_sub_i32_e32 v2, vcc, v5, v7
+; GCN1-NEXT:    v_sub_i32_e32 v3, vcc, v4, v6
+; GCN1-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GCN1-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GCN1-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
+; GCN1-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
+; GCN1-NEXT:    v_or_b32_e32 v1, v1, v2
+; GCN1-NEXT:    v_or_b32_e32 v0, v0, v3
+; GCN1-NEXT:    buffer_store_dwordx2 v[0:1], off, s[4:7], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: test_sub_v4i16:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN2-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v1, s3
+; GCN2-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; GCN2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GCN2-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCN2-NEXT:    v_mov_b32_e32 v4, s0
+; GCN2-NEXT:    v_mov_b32_e32 v5, s1
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u16_e32 v6, v1, v3
+; GCN2-NEXT:    v_sub_u16_sdwa v1, v1, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GCN2-NEXT:    v_sub_u16_e32 v3, v0, v2
+; GCN2-NEXT:    v_sub_u16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1
+; GCN2-NEXT:    v_or_b32_e32 v1, v6, v1
+; GCN2-NEXT:    v_or_b32_e32 v0, v3, v0
+; GCN2-NEXT:    flat_store_dwordx2 v[4:5], v[0:1]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: test_sub_v4i16:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GCN3-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GCN3-NEXT:    v_mov_b32_e32 v4, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx4 v[0:3], v0, s[2:3]
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_pk_sub_i16 v1, v1, v3
+; GCN3-NEXT:    v_pk_sub_i16 v0, v0, v2
+; GCN3-NEXT:    global_store_dwordx2 v4, v[0:1], s[0:1]
+; GCN3-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr <4 x i16>, ptr addrspace(1) %in, i32 %tid
   %b_ptr = getelementptr <4 x i16>, ptr addrspace(1) %gep, i16 1
   %a = load <4 x i16>, ptr addrspace(1) %gep
   %b = load <4 x i16>, ptr addrspace(1) %b_ptr
   %result = sub <4 x i16> %a, %b
   store <4 x i16> %result, ptr addrspace(1) %out
   ret void
 }
 
 ; GCN-LABEL: {{^}}s_sub_i64:
 ; GCN: s_sub_u32
 ; GCN: s_subb_u32
 define amdgpu_kernel void @s_sub_i64(ptr addrspace(1) noalias %out, i64 %a, i64 %b) nounwind {
+; GCN1-LABEL: s_sub_i64:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0xb
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
+; GCN1-NEXT:    s_mov_b32 s3, 0xf000
+; GCN1-NEXT:    s_mov_b32 s2, -1
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_sub_u32 s4, s4, s6
+; GCN1-NEXT:    s_subb_u32 s5, s5, s7
+; GCN1-NEXT:    v_mov_b32_e32 v0, s4
+; GCN1-NEXT:    v_mov_b32_e32 v1, s5
+; GCN1-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: s_sub_i64:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x2c
+; GCN2-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    s_sub_u32 s2, s4, s6
+; GCN2-NEXT:    s_subb_u32 s3, s5, s7
+; GCN2-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2-NEXT:    v_mov_b32_e32 v2, s2
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    v_mov_b32_e32 v3, s3
+; GCN2-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: s_sub_i64:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x2c
+; GCN3-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GCN3-NEXT:    v_mov_b32_e32 v2, 0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    s_sub_u32 s0, s4, s6
+; GCN3-NEXT:    s_subb_u32 s1, s5, s7
+; GCN3-NEXT:    v_mov_b32_e32 v0, s0
+; GCN3-NEXT:    v_mov_b32_e32 v1, s1
+; GCN3-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
+; GCN3-NEXT:    s_endpgm
   %result = sub i64 %a, %b
   store i64 %result, ptr addrspace(1) %out, align 8
   ret void
 }
 
 ; GCN-LABEL: {{^}}v_sub_i64:
 ; SI: v_sub_i32_e32
 ; SI: v_subb_u32_e32
 
 ; VI: v_sub_u32_e32
 ; VI: v_subb_u32_e32
 
 ; GFX9: v_sub_co_u32_e32
 ; GFX9: v_subb_co_u32_e32
 define amdgpu_kernel void @v_sub_i64(ptr addrspace(1) noalias %out, ptr addrspace(1) noalias %inA, ptr addrspace(1) noalias %inB) nounwind {
+; GCN1-LABEL: v_sub_i64:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; GCN1-NEXT:    s_mov_b32 s11, 0xf000
+; GCN1-NEXT:    s_mov_b32 s14, 0
+; GCN1-NEXT:    s_mov_b32 s15, s11
+; GCN1-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GCN1-NEXT:    v_mov_b32_e32 v1, 0
+; GCN1-NEXT:    s_mov_b64 s[2:3], s[14:15]
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b64 s[12:13], s[6:7]
+; GCN1-NEXT:    buffer_load_dwordx2 v[2:3], v[0:1], s[0:3], 0 addr64
+; GCN1-NEXT:    buffer_load_dwordx2 v[0:1], v[0:1], s[12:15], 0 addr64
+; GCN1-NEXT:    s_mov_b32 s10, -1
+; GCN1-NEXT:    s_mov_b32 s8, s4
+; GCN1-NEXT:    s_mov_b32 s9, s5
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v2
+; GCN1-NEXT:    v_subb_u32_e32 v1, vcc, v1, v3, vcc
+; GCN1-NEXT:    buffer_store_dwordx2 v[0:1], off, s[8:11], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: v_sub_i64:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GCN2-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; GCN2-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v1, s7
+; GCN2-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; GCN2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
+; GCN2-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
+; GCN2-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
+; GCN2-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; GCN2-NEXT:    flat_load_dwordx2 v[2:3], v[2:3]
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v2
+; GCN2-NEXT:    v_subb_u32_e32 v1, vcc, v1, v3, vcc
+; GCN2-NEXT:    v_mov_b32_e32 v2, s4
+; GCN2-NEXT:    v_mov_b32_e32 v3, s5
+; GCN2-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: v_sub_i64:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GCN3-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GCN3-NEXT:    v_lshlrev_b32_e32 v4, 3, v0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx2 v[0:1], v4, s[6:7]
+; GCN3-NEXT:    global_load_dwordx2 v[2:3], v4, s[2:3]
+; GCN3-NEXT:    v_mov_b32_e32 v4, 0
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_co_u32_e32 v0, vcc, v0, v2
+; GCN3-NEXT:    v_subb_co_u32_e32 v1, vcc, v1, v3, vcc
+; GCN3-NEXT:    global_store_dwordx2 v4, v[0:1], s[4:5]
+; GCN3-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
   %a_ptr = getelementptr i64, ptr addrspace(1) %inA, i32 %tid
   %b_ptr = getelementptr i64, ptr addrspace(1) %inB, i32 %tid
   %a = load i64, ptr addrspace(1) %a_ptr
   %b = load i64, ptr addrspace(1) %b_ptr
   %result = sub i64 %a, %b
   store i64 %result, ptr addrspace(1) %out, align 8
   ret void

 ; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc,
 ; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc,
 
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 define amdgpu_kernel void @v_test_sub_v2i64(ptr addrspace(1) %out, ptr addrspace(1) noalias %inA, ptr addrspace(1) noalias %inB) {
+; GCN1-LABEL: v_test_sub_v2i64:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; GCN1-NEXT:    s_mov_b32 s11, 0xf000
+; GCN1-NEXT:    s_mov_b32 s14, 0
+; GCN1-NEXT:    s_mov_b32 s15, s11
+; GCN1-NEXT:    v_lshlrev_b32_e32 v4, 4, v0
+; GCN1-NEXT:    v_mov_b32_e32 v5, 0
+; GCN1-NEXT:    s_mov_b64 s[2:3], s[14:15]
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b64 s[12:13], s[6:7]
+; GCN1-NEXT:    buffer_load_dwordx4 v[0:3], v[4:5], s[0:3], 0 addr64
+; GCN1-NEXT:    buffer_load_dwordx4 v[4:7], v[4:5], s[12:15], 0 addr64
+; GCN1-NEXT:    s_mov_b32 s10, -1
+; GCN1-NEXT:    s_mov_b32 s8, s4
+; GCN1-NEXT:    s_mov_b32 s9, s5
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v2, vcc, v6, v2
+; GCN1-NEXT:    v_subb_u32_e32 v3, vcc, v7, v3, vcc
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v4, v0
+; GCN1-NEXT:    v_subb_u32_e32 v1, vcc, v5, v1, vcc
+; GCN1-NEXT:    buffer_store_dwordx4 v[0:3], off, s[8:11], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: v_test_sub_v2i64:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GCN2-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; GCN2-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v1, s7
+; GCN2-NEXT:    v_add_u32_e32 v0, vcc, s6, v2
+; GCN2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GCN2-NEXT:    v_mov_b32_e32 v3, s1
+; GCN2-NEXT:    v_add_u32_e32 v4, vcc, s0, v2
+; GCN2-NEXT:    v_addc_u32_e32 v5, vcc, 0, v3, vcc
+; GCN2-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; GCN2-NEXT:    flat_load_dwordx4 v[4:7], v[4:5]
+; GCN2-NEXT:    s_waitcnt vmcnt(0)
+; GCN2-NEXT:    v_sub_u32_e32 v2, vcc, v2, v6
+; GCN2-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
+; GCN2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v4
+; GCN2-NEXT:    v_subb_u32_e32 v1, vcc, v1, v5, vcc
+; GCN2-NEXT:    v_mov_b32_e32 v4, s4
+; GCN2-NEXT:    v_mov_b32_e32 v5, s5
+; GCN2-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: v_test_sub_v2i64:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GCN3-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GCN3-NEXT:    v_lshlrev_b32_e32 v8, 4, v0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx4 v[0:3], v8, s[6:7]
+; GCN3-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3]
+; GCN3-NEXT:    v_mov_b32_e32 v8, 0
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_co_u32_e32 v2, vcc, v2, v6
+; GCN3-NEXT:    v_subb_co_u32_e32 v3, vcc, v3, v7, vcc
+; GCN3-NEXT:    v_sub_co_u32_e32 v0, vcc, v0, v4
+; GCN3-NEXT:    v_subb_co_u32_e32 v1, vcc, v1, v5, vcc
+; GCN3-NEXT:    global_store_dwordx4 v8, v[0:3], s[4:5]
+; GCN3-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
   %a_ptr = getelementptr <2 x i64>, ptr addrspace(1) %inA, i32 %tid
   %b_ptr = getelementptr <2 x i64>, ptr addrspace(1) %inB, i32 %tid
   %a = load <2 x i64>, ptr addrspace(1) %a_ptr
   %b = load <2 x i64>, ptr addrspace(1) %b_ptr
   %result = sub <2 x i64> %a, %b
   store <2 x i64> %result, ptr addrspace(1) %out
   ret void

 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc,
 ; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc,
 define amdgpu_kernel void @v_test_sub_v4i64(ptr addrspace(1) %out, ptr addrspace(1) noalias %inA, ptr addrspace(1) noalias %inB) {
+; GCN1-LABEL: v_test_sub_v4i64:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; GCN1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0xd
+; GCN1-NEXT:    s_mov_b32 s11, 0xf000
+; GCN1-NEXT:    s_mov_b32 s14, 0
+; GCN1-NEXT:    s_mov_b32 s15, s11
+; GCN1-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN1-NEXT:    s_mov_b64 s[12:13], s[6:7]
+; GCN1-NEXT:    v_lshlrev_b32_e32 v12, 5, v0
+; GCN1-NEXT:    v_mov_b32_e32 v13, 0
+; GCN1-NEXT:    s_mov_b64 s[2:3], s[14:15]
+; GCN1-NEXT:    buffer_load_dwordx4 v[0:3], v[12:13], s[12:15], 0 addr64
+; GCN1-NEXT:    buffer_load_dwordx4 v[4:7], v[12:13], s[0:3], 0 addr64
+; GCN1-NEXT:    buffer_load_dwordx4 v[8:11], v[12:13], s[0:3], 0 addr64 offset:16
+; GCN1-NEXT:    buffer_load_dwordx4 v[12:15], v[12:13], s[12:15], 0 addr64 offset:16
+; GCN1-NEXT:    s_mov_b32 s10, -1
+; GCN1-NEXT:    s_mov_b32 s8, s4
+; GCN1-NEXT:    s_mov_b32 s9, s5
+; GCN1-NEXT:    s_waitcnt vmcnt(2)
+; GCN1-NEXT:    v_sub_i32_e32 v2, vcc, v2, v6
+; GCN1-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
+; GCN1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v4
+; GCN1-NEXT:    v_subb_u32_e32 v1, vcc, v1, v5, vcc
+; GCN1-NEXT:    s_waitcnt vmcnt(0)
+; GCN1-NEXT:    v_sub_i32_e32 v6, vcc, v14, v10
+; GCN1-NEXT:    v_subb_u32_e32 v7, vcc, v15, v11, vcc
+; GCN1-NEXT:    v_sub_i32_e32 v4, vcc, v12, v8
+; GCN1-NEXT:    v_subb_u32_e32 v5, vcc, v13, v9, vcc
+; GCN1-NEXT:    buffer_store_dwordx4 v[4:7], off, s[8:11], 0 offset:16
+; GCN1-NEXT:    buffer_store_dwordx4 v[0:3], off, s[8:11], 0
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: v_test_sub_v4i64:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GCN2-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x34
+; GCN2-NEXT:    v_lshlrev_b32_e32 v0, 5, v0
+; GCN2-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN2-NEXT:    v_mov_b32_e32 v1, s7
+; GCN2-NEXT:    v_add_u32_e32 v8, vcc, s6, v0
+; GCN2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v1, vcc
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    v_add_u32_e32 v12, vcc, s0, v0
+; GCN2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v1, vcc
+; GCN2-NEXT:    flat_load_dwordx4 v[0:3], v[8:9]
+; GCN2-NEXT:    flat_load_dwordx4 v[4:7], v[12:13]
+; GCN2-NEXT:    v_add_u32_e32 v8, vcc, 16, v8
+; GCN2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; GCN2-NEXT:    v_add_u32_e32 v12, vcc, 16, v12
+; GCN2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v13, vcc
+; GCN2-NEXT:    flat_load_dwordx4 v[8:11], v[8:9]
+; GCN2-NEXT:    flat_load_dwordx4 v[12:15], v[12:13]
+; GCN2-NEXT:    v_mov_b32_e32 v17, s5
+; GCN2-NEXT:    v_mov_b32_e32 v16, s4
+; GCN2-NEXT:    s_add_u32 s0, s4, 16
+; GCN2-NEXT:    s_addc_u32 s1, s5, 0
+; GCN2-NEXT:    s_waitcnt vmcnt(2)
+; GCN2-NEXT:    v_sub_u32_e32 v2, vcc, v2, v6
+; GCN2-NEXT:    v_subb_u32_e32 v3, vcc, v3, v7, vcc
+; GCN2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v4
+; GCN2-NEXT:    v_subb_u32_e32 v1, vcc, v1, v5, vcc
+; GCN2-NEXT:    flat_store_dwordx4 v[16:17], v[0:3]
+; GCN2-NEXT:    s_waitcnt vmcnt(1)
+; GCN2-NEXT:    v_sub_u32_e32 v6, vcc, v10, v14
+; GCN2-NEXT:    v_subb_u32_e32 v7, vcc, v11, v15, vcc
+; GCN2-NEXT:    v_sub_u32_e32 v4, vcc, v8, v12
+; GCN2-NEXT:    v_mov_b32_e32 v0, s0
+; GCN2-NEXT:    v_subb_u32_e32 v5, vcc, v9, v13, vcc
+; GCN2-NEXT:    v_mov_b32_e32 v1, s1
+; GCN2-NEXT:    flat_store_dwordx4 v[0:1], v[4:7]
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: v_test_sub_v4i64:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GCN3-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GCN3-NEXT:    v_lshlrev_b32_e32 v16, 5, v0
+; GCN3-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN3-NEXT:    global_load_dwordx4 v[0:3], v16, s[6:7]
+; GCN3-NEXT:    global_load_dwordx4 v[4:7], v16, s[2:3]
+; GCN3-NEXT:    global_load_dwordx4 v[8:11], v16, s[6:7] offset:16
+; GCN3-NEXT:    global_load_dwordx4 v[12:15], v16, s[2:3] offset:16
+; GCN3-NEXT:    v_mov_b32_e32 v16, 0
+; GCN3-NEXT:    s_waitcnt vmcnt(2)
+; GCN3-NEXT:    v_sub_co_u32_e32 v2, vcc, v2, v6
+; GCN3-NEXT:    v_subb_co_u32_e32 v3, vcc, v3, v7, vcc
+; GCN3-NEXT:    v_sub_co_u32_e32 v0, vcc, v0, v4
+; GCN3-NEXT:    v_subb_co_u32_e32 v1, vcc, v1, v5, vcc
+; GCN3-NEXT:    s_waitcnt vmcnt(0)
+; GCN3-NEXT:    v_sub_co_u32_e32 v6, vcc, v10, v14
+; GCN3-NEXT:    v_subb_co_u32_e32 v7, vcc, v11, v15, vcc
+; GCN3-NEXT:    v_sub_co_u32_e32 v4, vcc, v8, v12
+; GCN3-NEXT:    v_subb_co_u32_e32 v5, vcc, v9, v13, vcc
+; GCN3-NEXT:    global_store_dwordx4 v16, v[4:7], s[4:5] offset:16
+; GCN3-NEXT:    global_store_dwordx4 v16, v[0:3], s[4:5]
+; GCN3-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() readnone
   %a_ptr = getelementptr <4 x i64>, ptr addrspace(1) %inA, i32 %tid
   %b_ptr = getelementptr <4 x i64>, ptr addrspace(1) %inB, i32 %tid
   %a = load <4 x i64>, ptr addrspace(1) %a_ptr
   %b = load <4 x i64>, ptr addrspace(1) %b_ptr
   %result = sub <4 x i64> %a, %b
   store <4 x i64> %result, ptr addrspace(1) %out
   ret void
 }
 
 ; Make sure the VOP3 form of sub is initially selected. Otherwise pair
 ; of opies from/to VCC would be necessary
 
 ; GCN-LABEL: {{^}}sub_select_vop3:
 ; SI: v_subrev_i32_e64 v0, s[0:1], s0, v0
 ; VI: v_subrev_u32_e64 v0, s[0:1], s0, v0
 ; GFX9: v_subrev_u32_e32 v0, s0, v0
 
 ; GCN: ; def vcc
 ; GCN: ds_write_b32
 ; GCN: ; use vcc
 define amdgpu_ps void @sub_select_vop3(i32 inreg %s, i32 %v) {
+; GCN1-LABEL: sub_select_vop3:
+; GCN1:       ; %bb.0:
+; GCN1-NEXT:    v_subrev_i32_e64 v0, s[0:1], s0, v0
+; GCN1-NEXT:    s_mov_b32 m0, -1
+; GCN1-NEXT:    ;;#ASMSTART
+; GCN1-NEXT:    ; def vcc
+; GCN1-NEXT:    ;;#ASMEND
+; GCN1-NEXT:    ds_write_b32 v0, v0
+; GCN1-NEXT:    ;;#ASMSTART
+; GCN1-NEXT:    ; use vcc
+; GCN1-NEXT:    ;;#ASMEND
+; GCN1-NEXT:    s_endpgm
+;
+; GCN2-LABEL: sub_select_vop3:
+; GCN2:       ; %bb.0:
+; GCN2-NEXT:    v_subrev_u32_e64 v0, s[0:1], s0, v0
+; GCN2-NEXT:    s_mov_b32 m0, -1
+; GCN2-NEXT:    ;;#ASMSTART
+; GCN2-NEXT:    ; def vcc
+; GCN2-NEXT:    ;;#ASMEND
+; GCN2-NEXT:    ds_write_b32 v0, v0
+; GCN2-NEXT:    ;;#ASMSTART
+; GCN2-NEXT:    ; use vcc
+; GCN2-NEXT:    ;;#ASMEND
+; GCN2-NEXT:    s_endpgm
+;
+; GCN3-LABEL: sub_select_vop3:
+; GCN3:       ; %bb.0:
+; GCN3-NEXT:    v_subrev_u32_e32 v0, s0, v0
+; GCN3-NEXT:    ;;#ASMSTART
+; GCN3-NEXT:    ; def vcc
+; GCN3-NEXT:    ;;#ASMEND
+; GCN3-NEXT:    ds_write_b32 v0, v0
+; GCN3-NEXT:    ;;#ASMSTART
+; GCN3-NEXT:    ; use vcc
+; GCN3-NEXT:    ;;#ASMEND
+; GCN3-NEXT:    s_endpgm
   %vcc = call i64 asm sideeffect "; def vcc", "={vcc}"()
   %sub = sub i32 %v, %s
   store i32 %sub, ptr addrspace(3) undef
   call void asm sideeffect "; use vcc", "{vcc}"(i64 %vcc)
   ret void
 }

llvm/test/CodeGen/AMDGPU/udiv.ll

 ; SI-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; SI-NEXT:    v_mul_lo_u32 v3, v3, v2
 ; SI-NEXT:    v_mul_hi_u32 v3, v2, v3
 ; SI-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; SI-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; SI-NEXT:    v_mul_lo_u32 v3, v2, v1
 ; SI-NEXT:    v_add_i32_e32 v4, vcc, 1, v2
 ; SI-NEXT:    v_sub_i32_e32 v0, vcc, v0, v3
-; SI-NEXT:    v_subrev_i32_e32 v3, vcc, v1, v0
+; SI-NEXT:    v_sub_i32_e32 v3, vcc, v0, v1
 ; SI-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; SI-NEXT:    v_cndmask_b32_e32 v2, v2, v4, vcc
 ; SI-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; SI-NEXT:    v_add_i32_e32 v3, vcc, 1, v2
 ; SI-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; SI-NEXT:    v_cndmask_b32_e32 v0, v2, v3, vcc
 ; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; SI-NEXT:    s_endpgm

 ; VI-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; VI-NEXT:    v_mul_lo_u32 v3, v3, v2
 ; VI-NEXT:    v_mul_hi_u32 v3, v2, v3
 ; VI-NEXT:    v_add_u32_e32 v2, vcc, v2, v3
 ; VI-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; VI-NEXT:    v_mul_lo_u32 v3, v2, v1
 ; VI-NEXT:    v_add_u32_e32 v4, vcc, 1, v2
 ; VI-NEXT:    v_sub_u32_e32 v0, vcc, v0, v3
-; VI-NEXT:    v_subrev_u32_e32 v3, vcc, v1, v0
+; VI-NEXT:    v_sub_u32_e32 v3, vcc, v0, v1
 ; VI-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; VI-NEXT:    v_cndmask_b32_e32 v2, v2, v4, vcc
 ; VI-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; VI-NEXT:    v_add_u32_e32 v3, vcc, 1, v2
 ; VI-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; VI-NEXT:    v_cndmask_b32_e32 v0, v2, v3, vcc
 ; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
 ; VI-NEXT:    s_endpgm

 ; GCN-NEXT:    v_mul_hi_u32 v3, v2, v3
 ; GCN-NEXT:    v_add_u32_e32 v2, vcc, v2, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v0, v2
 ; GCN-NEXT:    v_mov_b32_e32 v2, s0
 ; GCN-NEXT:    v_mov_b32_e32 v3, s1
 ; GCN-NEXT:    v_mul_lo_u32 v5, v4, v1
 ; GCN-NEXT:    v_add_u32_e32 v6, vcc, 1, v4
 ; GCN-NEXT:    v_sub_u32_e32 v0, vcc, v0, v5
-; GCN-NEXT:    v_subrev_u32_e32 v5, vcc, v1, v0
+; GCN-NEXT:    v_sub_u32_e32 v5, vcc, v0, v1
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; GCN-NEXT:    v_cndmask_b32_e32 v4, v4, v6, vcc
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v0, v5, vcc
 ; GCN-NEXT:    v_add_u32_e32 v5, vcc, 1, v4
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v1
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v4, v5, vcc
 ; GCN-NEXT:    flat_store_dword v[2:3], v0
 ; GCN-NEXT:    s_endpgm

 ; SI-NEXT:    s_sub_i32 s4, 0, s3
 ; SI-NEXT:    s_mov_b32 s5, s1
 ; SI-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; SI-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; SI-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; SI-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; SI-NEXT:    s_mov_b32 s4, s0
 ; SI-NEXT:    v_mul_hi_u32 v1, v0, v1
-; SI-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; SI-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; SI-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; SI-NEXT:    v_readfirstlane_b32 s0, v0
 ; SI-NEXT:    s_mul_i32 s0, s0, s3
 ; SI-NEXT:    s_sub_i32 s0, s2, s0
 ; SI-NEXT:    s_sub_i32 s1, s0, s3
 ; SI-NEXT:    v_add_i32_e32 v1, vcc, 1, v0
 ; SI-NEXT:    s_cmp_ge_u32 s0, s3
 ; SI-NEXT:    s_cselect_b64 vcc, -1, 0

 ; VI-NEXT:    s_sub_i32 s4, 0, s3
 ; VI-NEXT:    s_mov_b32 s5, s1
 ; VI-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; VI-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; VI-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; VI-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; VI-NEXT:    s_mov_b32 s4, s0
 ; VI-NEXT:    v_mul_hi_u32 v1, v0, v1
-; VI-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; VI-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; VI-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; VI-NEXT:    v_readfirstlane_b32 s0, v0
 ; VI-NEXT:    s_mul_i32 s0, s0, s3
 ; VI-NEXT:    s_sub_i32 s0, s2, s0
 ; VI-NEXT:    s_sub_i32 s1, s0, s3
 ; VI-NEXT:    v_add_u32_e32 v1, vcc, 1, v0
 ; VI-NEXT:    s_cmp_ge_u32 s0, s3
 ; VI-NEXT:    s_cselect_b64 vcc, -1, 0

 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_cvt_f32_u32_e32 v0, s3
 ; GCN-NEXT:    s_sub_i32 s4, 0, s3
 ; GCN-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v1, s4, v0
 ; GCN-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GCN-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; GCN-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; GCN-NEXT:    v_readfirstlane_b32 s4, v0
 ; GCN-NEXT:    s_mul_i32 s4, s4, s3
 ; GCN-NEXT:    s_sub_i32 s2, s2, s4
 ; GCN-NEXT:    s_sub_i32 s4, s2, s3
 ; GCN-NEXT:    v_add_u32_e32 v1, vcc, 1, v0
 ; GCN-NEXT:    s_cmp_ge_u32 s2, s3
 ; GCN-NEXT:    s_cselect_b64 vcc, -1, 0

 ; SI-NEXT:    v_mul_f32_e32 v4, 0x4f7ffffe, v4
 ; SI-NEXT:    v_mul_f32_e32 v5, 0x4f7ffffe, v5
 ; SI-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; SI-NEXT:    v_cvt_u32_f32_e32 v5, v5
 ; SI-NEXT:    v_mul_lo_u32 v6, v6, v4
 ; SI-NEXT:    v_mul_lo_u32 v7, v7, v5
 ; SI-NEXT:    v_mul_hi_u32 v6, v4, v6
 ; SI-NEXT:    v_mul_hi_u32 v7, v5, v7
-; SI-NEXT:    v_add_i32_e32 v4, vcc, v6, v4
+; SI-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
 ; SI-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
 ; SI-NEXT:    v_mul_hi_u32 v4, v0, v4
 ; SI-NEXT:    v_mul_hi_u32 v5, v1, v5
 ; SI-NEXT:    v_mul_lo_u32 v6, v4, v2
 ; SI-NEXT:    v_mul_lo_u32 v8, v5, v3
 ; SI-NEXT:    v_add_i32_e32 v7, vcc, 1, v4
 ; SI-NEXT:    v_sub_i32_e32 v0, vcc, v0, v6
 ; SI-NEXT:    v_sub_i32_e32 v1, vcc, v1, v8
 ; SI-NEXT:    v_add_i32_e32 v9, vcc, 1, v5
 ; SI-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v2
 ; SI-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v3
-; SI-NEXT:    v_subrev_i32_e32 v6, vcc, v2, v0
+; SI-NEXT:    v_sub_i32_e32 v6, vcc, v0, v2
 ; SI-NEXT:    v_cndmask_b32_e64 v4, v4, v7, s[0:1]
-; SI-NEXT:    v_subrev_i32_e32 v7, vcc, v3, v1
+; SI-NEXT:    v_sub_i32_e32 v7, vcc, v1, v3
 ; SI-NEXT:    v_cndmask_b32_e64 v5, v5, v9, s[2:3]
 ; SI-NEXT:    v_cndmask_b32_e64 v0, v0, v6, s[0:1]
 ; SI-NEXT:    v_add_i32_e32 v6, vcc, 1, v4
 ; SI-NEXT:    v_cndmask_b32_e64 v1, v1, v7, s[2:3]
 ; SI-NEXT:    v_add_i32_e32 v7, vcc, 1, v5
 ; SI-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
 ; SI-NEXT:    v_cndmask_b32_e32 v0, v4, v6, vcc
 ; SI-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3

 ; VI-NEXT:    v_mul_f32_e32 v4, 0x4f7ffffe, v4
 ; VI-NEXT:    v_mul_f32_e32 v5, 0x4f7ffffe, v5
 ; VI-NEXT:    v_cvt_u32_f32_e32 v4, v4
 ; VI-NEXT:    v_cvt_u32_f32_e32 v5, v5
 ; VI-NEXT:    v_mul_lo_u32 v6, v6, v4
 ; VI-NEXT:    v_mul_lo_u32 v7, v7, v5
 ; VI-NEXT:    v_mul_hi_u32 v6, v4, v6
 ; VI-NEXT:    v_mul_hi_u32 v7, v5, v7
-; VI-NEXT:    v_add_u32_e32 v4, vcc, v6, v4
+; VI-NEXT:    v_add_u32_e32 v4, vcc, v4, v6
 ; VI-NEXT:    v_add_u32_e32 v5, vcc, v5, v7
 ; VI-NEXT:    v_mul_hi_u32 v4, v0, v4
 ; VI-NEXT:    v_mul_hi_u32 v5, v1, v5
 ; VI-NEXT:    v_mul_lo_u32 v6, v4, v2
 ; VI-NEXT:    v_mul_lo_u32 v8, v5, v3
 ; VI-NEXT:    v_add_u32_e32 v7, vcc, 1, v4
 ; VI-NEXT:    v_sub_u32_e32 v0, vcc, v0, v6
 ; VI-NEXT:    v_sub_u32_e32 v1, vcc, v1, v8
 ; VI-NEXT:    v_add_u32_e32 v9, vcc, 1, v5
 ; VI-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v2
 ; VI-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v3
-; VI-NEXT:    v_subrev_u32_e32 v6, vcc, v2, v0
+; VI-NEXT:    v_sub_u32_e32 v6, vcc, v0, v2
 ; VI-NEXT:    v_cndmask_b32_e64 v4, v4, v7, s[0:1]
-; VI-NEXT:    v_subrev_u32_e32 v7, vcc, v3, v1
+; VI-NEXT:    v_sub_u32_e32 v7, vcc, v1, v3
 ; VI-NEXT:    v_cndmask_b32_e64 v5, v5, v9, s[2:3]
 ; VI-NEXT:    v_cndmask_b32_e64 v0, v0, v6, s[0:1]
 ; VI-NEXT:    v_add_u32_e32 v6, vcc, 1, v4
 ; VI-NEXT:    v_cndmask_b32_e64 v1, v1, v7, s[2:3]
 ; VI-NEXT:    v_add_u32_e32 v7, vcc, 1, v5
 ; VI-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
 ; VI-NEXT:    v_cndmask_b32_e32 v0, v4, v6, vcc
 ; VI-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3

 ; GCN-NEXT:    v_sub_u32_e32 v4, vcc, 0, v2
 ; GCN-NEXT:    v_mul_lo_u32 v5, v4, v6
 ; GCN-NEXT:    v_sub_u32_e32 v4, vcc, 0, v3
 ; GCN-NEXT:    v_mul_lo_u32 v8, v4, v7
 ; GCN-NEXT:    v_mul_hi_u32 v9, v6, v5
 ; GCN-NEXT:    v_mov_b32_e32 v4, s0
 ; GCN-NEXT:    v_mov_b32_e32 v5, s1
 ; GCN-NEXT:    v_mul_hi_u32 v8, v7, v8
-; GCN-NEXT:    v_add_u32_e32 v6, vcc, v9, v6
+; GCN-NEXT:    v_add_u32_e32 v6, vcc, v6, v9
 ; GCN-NEXT:    v_mul_hi_u32 v6, v0, v6
 ; GCN-NEXT:    v_add_u32_e32 v7, vcc, v7, v8
 ; GCN-NEXT:    v_mul_hi_u32 v7, v1, v7
 ; GCN-NEXT:    v_mul_lo_u32 v8, v6, v2
 ; GCN-NEXT:    v_add_u32_e32 v9, vcc, 1, v6
 ; GCN-NEXT:    v_mul_lo_u32 v10, v7, v3
 ; GCN-NEXT:    v_sub_u32_e32 v0, vcc, v0, v8
 ; GCN-NEXT:    v_add_u32_e32 v11, vcc, 1, v7
 ; GCN-NEXT:    v_sub_u32_e32 v1, vcc, v1, v10
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[0:1], v0, v2
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[2:3], v1, v3
-; GCN-NEXT:    v_subrev_u32_e32 v8, vcc, v2, v0
+; GCN-NEXT:    v_sub_u32_e32 v8, vcc, v0, v2
 ; GCN-NEXT:    v_cndmask_b32_e64 v6, v6, v9, s[0:1]
-; GCN-NEXT:    v_subrev_u32_e32 v9, vcc, v3, v1
+; GCN-NEXT:    v_sub_u32_e32 v9, vcc, v1, v3
 ; GCN-NEXT:    v_cndmask_b32_e64 v7, v7, v11, s[2:3]
 ; GCN-NEXT:    v_cndmask_b32_e64 v0, v0, v8, s[0:1]
 ; GCN-NEXT:    v_add_u32_e32 v8, vcc, 1, v6
 ; GCN-NEXT:    v_cndmask_b32_e64 v1, v1, v9, s[2:3]
 ; GCN-NEXT:    v_add_u32_e32 v9, vcc, 1, v7
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
 ; GCN-NEXT:    v_cndmask_b32_e32 v0, v6, v8, vcc
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3

 ; SI-NEXT:    v_mul_lo_u32 v11, v11, v10
 ; SI-NEXT:    v_mul_lo_u32 v13, v13, v12
 ; SI-NEXT:    v_mul_lo_u32 v15, v15, v14
 ; SI-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; SI-NEXT:    v_mul_hi_u32 v11, v10, v11
 ; SI-NEXT:    v_mul_hi_u32 v13, v12, v13
 ; SI-NEXT:    v_mul_hi_u32 v15, v14, v15
 ; SI-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
-; SI-NEXT:    v_add_i32_e32 v9, vcc, v11, v10
+; SI-NEXT:    v_add_i32_e32 v9, vcc, v10, v11
 ; SI-NEXT:    v_add_i32_e32 v10, vcc, v12, v13
-; SI-NEXT:    v_add_i32_e32 v11, vcc, v15, v14
+; SI-NEXT:    v_add_i32_e32 v11, vcc, v14, v15
 ; SI-NEXT:    s_waitcnt vmcnt(0)
 ; SI-NEXT:    v_mul_hi_u32 v8, v4, v8
 ; SI-NEXT:    v_mul_hi_u32 v9, v5, v9
 ; SI-NEXT:    v_mul_hi_u32 v10, v6, v10
 ; SI-NEXT:    v_mul_hi_u32 v11, v7, v11
 ; SI-NEXT:    v_mul_lo_u32 v12, v8, v0
 ; SI-NEXT:    v_mul_lo_u32 v14, v9, v1
 ; SI-NEXT:    v_mul_lo_u32 v16, v10, v2
 ; SI-NEXT:    v_mul_lo_u32 v18, v11, v3
 ; SI-NEXT:    v_sub_i32_e32 v4, vcc, v4, v12
 ; SI-NEXT:    v_sub_i32_e32 v5, vcc, v5, v14
 ; SI-NEXT:    v_sub_i32_e32 v6, vcc, v6, v16
 ; SI-NEXT:    v_sub_i32_e32 v7, vcc, v7, v18
 ; SI-NEXT:    v_add_i32_e32 v13, vcc, 1, v8
 ; SI-NEXT:    v_add_i32_e32 v15, vcc, 1, v9
 ; SI-NEXT:    v_add_i32_e32 v17, vcc, 1, v10
 ; SI-NEXT:    v_add_i32_e32 v19, vcc, 1, v11
 ; SI-NEXT:    v_cmp_ge_u32_e64 s[0:1], v4, v0
 ; SI-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v1
 ; SI-NEXT:    v_cmp_ge_u32_e64 s[4:5], v6, v2
 ; SI-NEXT:    v_cmp_ge_u32_e64 s[6:7], v7, v3
-; SI-NEXT:    v_subrev_i32_e32 v12, vcc, v0, v4
+; SI-NEXT:    v_sub_i32_e32 v12, vcc, v4, v0
 ; SI-NEXT:    v_cndmask_b32_e64 v8, v8, v13, s[0:1]
-; SI-NEXT:    v_subrev_i32_e32 v13, vcc, v1, v5
+; SI-NEXT:    v_sub_i32_e32 v13, vcc, v5, v1
 ; SI-NEXT:    v_cndmask_b32_e64 v9, v9, v15, s[2:3]
-; SI-NEXT:    v_subrev_i32_e32 v14, vcc, v2, v6
+; SI-NEXT:    v_sub_i32_e32 v14, vcc, v6, v2
 ; SI-NEXT:    v_cndmask_b32_e64 v10, v10, v17, s[4:5]
-; SI-NEXT:    v_subrev_i32_e32 v15, vcc, v3, v7
+; SI-NEXT:    v_sub_i32_e32 v15, vcc, v7, v3
 ; SI-NEXT:    v_cndmask_b32_e64 v11, v11, v19, s[6:7]
 ; SI-NEXT:    v_cndmask_b32_e64 v4, v4, v12, s[0:1]
 ; SI-NEXT:    v_add_i32_e32 v12, vcc, 1, v8
 ; SI-NEXT:    v_cndmask_b32_e64 v5, v5, v13, s[2:3]
 ; SI-NEXT:    v_add_i32_e32 v13, vcc, 1, v9
 ; SI-NEXT:    v_cndmask_b32_e64 v6, v6, v14, s[4:5]
 ; SI-NEXT:    v_add_i32_e32 v14, vcc, 1, v10
 ; SI-NEXT:    v_cndmask_b32_e64 v7, v7, v15, s[6:7]

 ; VI-NEXT:    v_mul_lo_u32 v11, v11, v10
 ; VI-NEXT:    v_mul_lo_u32 v13, v13, v12
 ; VI-NEXT:    v_mul_lo_u32 v15, v15, v14
 ; VI-NEXT:    v_mul_hi_u32 v9, v8, v9
 ; VI-NEXT:    v_mul_hi_u32 v11, v10, v11
 ; VI-NEXT:    v_mul_hi_u32 v13, v12, v13
 ; VI-NEXT:    v_mul_hi_u32 v15, v14, v15
 ; VI-NEXT:    v_add_u32_e32 v8, vcc, v8, v9
-; VI-NEXT:    v_add_u32_e32 v9, vcc, v11, v10
+; VI-NEXT:    v_add_u32_e32 v9, vcc, v10, v11
 ; VI-NEXT:    v_add_u32_e32 v10, vcc, v12, v13
-; VI-NEXT:    v_add_u32_e32 v11, vcc, v15, v14
+; VI-NEXT:    v_add_u32_e32 v11, vcc, v14, v15
 ; VI-NEXT:    s_waitcnt vmcnt(0)
 ; VI-NEXT:    v_mul_hi_u32 v8, v4, v8
 ; VI-NEXT:    v_mul_hi_u32 v9, v5, v9
 ; VI-NEXT:    v_mul_hi_u32 v10, v6, v10
 ; VI-NEXT:    v_mul_hi_u32 v11, v7, v11
 ; VI-NEXT:    v_mul_lo_u32 v12, v8, v0
 ; VI-NEXT:    v_mul_lo_u32 v14, v9, v1
 ; VI-NEXT:    v_mul_lo_u32 v16, v10, v2
 ; VI-NEXT:    v_mul_lo_u32 v18, v11, v3
 ; VI-NEXT:    v_sub_u32_e32 v4, vcc, v4, v12
 ; VI-NEXT:    v_sub_u32_e32 v5, vcc, v5, v14
 ; VI-NEXT:    v_sub_u32_e32 v6, vcc, v6, v16
 ; VI-NEXT:    v_sub_u32_e32 v7, vcc, v7, v18
 ; VI-NEXT:    v_add_u32_e32 v13, vcc, 1, v8
 ; VI-NEXT:    v_add_u32_e32 v15, vcc, 1, v9
 ; VI-NEXT:    v_add_u32_e32 v17, vcc, 1, v10
 ; VI-NEXT:    v_add_u32_e32 v19, vcc, 1, v11
 ; VI-NEXT:    v_cmp_ge_u32_e64 s[0:1], v4, v0
 ; VI-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v1
 ; VI-NEXT:    v_cmp_ge_u32_e64 s[4:5], v6, v2
 ; VI-NEXT:    v_cmp_ge_u32_e64 s[6:7], v7, v3
-; VI-NEXT:    v_subrev_u32_e32 v12, vcc, v0, v4
+; VI-NEXT:    v_sub_u32_e32 v12, vcc, v4, v0
 ; VI-NEXT:    v_cndmask_b32_e64 v8, v8, v13, s[0:1]
-; VI-NEXT:    v_subrev_u32_e32 v13, vcc, v1, v5
+; VI-NEXT:    v_sub_u32_e32 v13, vcc, v5, v1
 ; VI-NEXT:    v_cndmask_b32_e64 v9, v9, v15, s[2:3]
-; VI-NEXT:    v_subrev_u32_e32 v14, vcc, v2, v6
+; VI-NEXT:    v_sub_u32_e32 v14, vcc, v6, v2
 ; VI-NEXT:    v_cndmask_b32_e64 v10, v10, v17, s[4:5]
-; VI-NEXT:    v_subrev_u32_e32 v15, vcc, v3, v7
+; VI-NEXT:    v_sub_u32_e32 v15, vcc, v7, v3
 ; VI-NEXT:    v_cndmask_b32_e64 v11, v11, v19, s[6:7]
 ; VI-NEXT:    v_cndmask_b32_e64 v4, v4, v12, s[0:1]
 ; VI-NEXT:    v_add_u32_e32 v12, vcc, 1, v8
 ; VI-NEXT:    v_cndmask_b32_e64 v5, v5, v13, s[2:3]
 ; VI-NEXT:    v_add_u32_e32 v13, vcc, 1, v9
 ; VI-NEXT:    v_cndmask_b32_e64 v6, v6, v14, s[4:5]
 ; VI-NEXT:    v_add_u32_e32 v14, vcc, 1, v10
 ; VI-NEXT:    v_cndmask_b32_e64 v7, v7, v15, s[6:7]

 ; GCN-NEXT:    v_mul_lo_u32 v13, v13, v12
 ; GCN-NEXT:    v_mul_lo_u32 v15, v15, v14
 ; GCN-NEXT:    v_mul_lo_u32 v17, v17, v16
 ; GCN-NEXT:    v_mul_hi_u32 v11, v10, v11
 ; GCN-NEXT:    v_mul_hi_u32 v13, v12, v13
 ; GCN-NEXT:    v_mul_hi_u32 v15, v14, v15
 ; GCN-NEXT:    v_mul_hi_u32 v17, v16, v17
 ; GCN-NEXT:    v_add_u32_e32 v10, vcc, v10, v11
-; GCN-NEXT:    v_add_u32_e32 v11, vcc, v13, v12
+; GCN-NEXT:    v_add_u32_e32 v11, vcc, v12, v13
 ; GCN-NEXT:    v_add_u32_e32 v12, vcc, v14, v15
-; GCN-NEXT:    v_add_u32_e32 v13, vcc, v17, v16
+; GCN-NEXT:    v_add_u32_e32 v13, vcc, v16, v17
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    v_mul_hi_u32 v10, v4, v10
 ; GCN-NEXT:    v_mul_hi_u32 v11, v5, v11
 ; GCN-NEXT:    v_mul_hi_u32 v12, v6, v12
 ; GCN-NEXT:    v_mul_hi_u32 v13, v7, v13
 ; GCN-NEXT:    v_mul_lo_u32 v14, v10, v0
 ; GCN-NEXT:    v_mul_lo_u32 v16, v11, v1
 ; GCN-NEXT:    v_mul_lo_u32 v18, v12, v2
 ; GCN-NEXT:    v_mul_lo_u32 v20, v13, v3
 ; GCN-NEXT:    v_sub_u32_e32 v4, vcc, v4, v14
 ; GCN-NEXT:    v_sub_u32_e32 v5, vcc, v5, v16
 ; GCN-NEXT:    v_sub_u32_e32 v6, vcc, v6, v18
 ; GCN-NEXT:    v_sub_u32_e32 v7, vcc, v7, v20
 ; GCN-NEXT:    v_add_u32_e32 v15, vcc, 1, v10
 ; GCN-NEXT:    v_add_u32_e32 v17, vcc, 1, v11
 ; GCN-NEXT:    v_add_u32_e32 v19, vcc, 1, v12
 ; GCN-NEXT:    v_add_u32_e32 v21, vcc, 1, v13
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[0:1], v4, v0
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v1
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[4:5], v6, v2
 ; GCN-NEXT:    v_cmp_ge_u32_e64 s[6:7], v7, v3
-; GCN-NEXT:    v_subrev_u32_e32 v14, vcc, v0, v4
+; GCN-NEXT:    v_sub_u32_e32 v14, vcc, v4, v0
 ; GCN-NEXT:    v_cndmask_b32_e64 v10, v10, v15, s[0:1]
-; GCN-NEXT:    v_subrev_u32_e32 v15, vcc, v1, v5
+; GCN-NEXT:    v_sub_u32_e32 v15, vcc, v5, v1
 ; GCN-NEXT:    v_cndmask_b32_e64 v11, v11, v17, s[2:3]
-; GCN-NEXT:    v_subrev_u32_e32 v16, vcc, v2, v6
+; GCN-NEXT:    v_sub_u32_e32 v16, vcc, v6, v2
 ; GCN-NEXT:    v_cndmask_b32_e64 v12, v12, v19, s[4:5]
-; GCN-NEXT:    v_subrev_u32_e32 v17, vcc, v3, v7
+; GCN-NEXT:    v_sub_u32_e32 v17, vcc, v7, v3
 ; GCN-NEXT:    v_cndmask_b32_e64 v13, v13, v21, s[6:7]
 ; GCN-NEXT:    v_cndmask_b32_e64 v4, v4, v14, s[0:1]
 ; GCN-NEXT:    v_add_u32_e32 v14, vcc, 1, v10
 ; GCN-NEXT:    v_cndmask_b32_e64 v5, v5, v15, s[2:3]
 ; GCN-NEXT:    v_add_u32_e32 v15, vcc, 1, v11
 ; GCN-NEXT:    v_cndmask_b32_e64 v6, v6, v16, s[4:5]
 ; GCN-NEXT:    v_add_u32_e32 v16, vcc, 1, v12
 ; GCN-NEXT:    v_cndmask_b32_e64 v7, v7, v17, s[6:7]

 ; SI-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; SI-NEXT:    v_mul_lo_u32 v4, v4, v1
 ; SI-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; SI-NEXT:    v_add_i32_e32 v1, vcc, v1, v4
 ; SI-NEXT:    v_mul_hi_u32 v1, v2, v1
 ; SI-NEXT:    v_mul_lo_u32 v3, v1, v0
 ; SI-NEXT:    v_add_i32_e32 v4, vcc, 1, v1
 ; SI-NEXT:    v_sub_i32_e32 v2, vcc, v2, v3
-; SI-NEXT:    v_subrev_i32_e32 v3, vcc, v0, v2
+; SI-NEXT:    v_sub_i32_e32 v3, vcc, v2, v0
 ; SI-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v0
 ; SI-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc
 ; SI-NEXT:    v_cndmask_b32_e32 v2, v2, v3, vcc
 ; SI-NEXT:    v_add_i32_e32 v3, vcc, 1, v1
 ; SI-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v0
 ; SI-NEXT:    v_cndmask_b32_e32 v0, v1, v3, vcc
 ; SI-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
 ; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0

 ; VI-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; VI-NEXT:    v_mul_lo_u32 v4, v4, v1
 ; VI-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; VI-NEXT:    v_add_u32_e32 v1, vcc, v1, v4
 ; VI-NEXT:    v_mul_hi_u32 v1, v2, v1
 ; VI-NEXT:    v_mul_lo_u32 v3, v1, v0
 ; VI-NEXT:    v_add_u32_e32 v4, vcc, 1, v1
 ; VI-NEXT:    v_sub_u32_e32 v2, vcc, v2, v3
-; VI-NEXT:    v_subrev_u32_e32 v3, vcc, v0, v2
+; VI-NEXT:    v_sub_u32_e32 v3, vcc, v2, v0
 ; VI-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v0
 ; VI-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc
 ; VI-NEXT:    v_cndmask_b32_e32 v2, v2, v3, vcc
 ; VI-NEXT:    v_add_u32_e32 v3, vcc, 1, v1
 ; VI-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v0
 ; VI-NEXT:    v_cndmask_b32_e32 v0, v1, v3, vcc
 ; VI-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
 ; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0

 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v4
 ; GCN-NEXT:    v_add_u32_e32 v0, vcc, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v4, v2, v0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s0
 ; GCN-NEXT:    v_mov_b32_e32 v1, s1
 ; GCN-NEXT:    v_mul_lo_u32 v5, v4, v3
 ; GCN-NEXT:    v_add_u32_e32 v6, vcc, 1, v4
 ; GCN-NEXT:    v_sub_u32_e32 v2, vcc, v2, v5
-; GCN-NEXT:    v_subrev_u32_e32 v5, vcc, v3, v2
+; GCN-NEXT:    v_sub_u32_e32 v5, vcc, v2, v3
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v3
 ; GCN-NEXT:    v_cndmask_b32_e32 v4, v4, v6, vcc
 ; GCN-NEXT:    v_cndmask_b32_e32 v2, v2, v5, vcc
 ; GCN-NEXT:    v_add_u32_e32 v5, vcc, 1, v4
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v2, v3
 ; GCN-NEXT:    v_cndmask_b32_e32 v2, v4, v5, vcc
 ; GCN-NEXT:    v_and_b32_e32 v2, 0xffffff, v2
 ; GCN-NEXT:    flat_store_dword v[0:1], v2

 ; VI-NEXT:    v_mul_f32_e32 v3, 0x2f800000, v2
 ; VI-NEXT:    v_trunc_f32_e32 v3, v3
 ; VI-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v3
 ; VI-NEXT:    v_cvt_u32_f32_e32 v6, v2
 ; VI-NEXT:    v_cvt_u32_f32_e32 v7, v3
 ; VI-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v6, s6, 0
 ; VI-NEXT:    v_mul_lo_u32 v4, v7, s6
 ; VI-NEXT:    v_sub_u32_e32 v3, vcc, v3, v6
-; VI-NEXT:    v_add_u32_e32 v8, vcc, v4, v3
+; VI-NEXT:    v_add_u32_e32 v8, vcc, v3, v4
 ; VI-NEXT:    v_mul_hi_u32 v5, v6, v2
 ; VI-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], v6, v8, 0
 ; VI-NEXT:    v_add_u32_e32 v9, vcc, v5, v3
 ; VI-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v7, v2, 0
 ; VI-NEXT:    v_addc_u32_e32 v10, vcc, 0, v4, vcc
 ; VI-NEXT:    v_mad_u64_u32 v[4:5], s[4:5], v7, v8, 0
 ; VI-NEXT:    v_add_u32_e32 v2, vcc, v9, v2
 ; VI-NEXT:    v_addc_u32_e32 v2, vcc, v10, v3, vcc

 ; GCN-NEXT:    v_mul_f32_e32 v3, 0x2f800000, v2
 ; GCN-NEXT:    v_trunc_f32_e32 v3, v3
 ; GCN-NEXT:    v_mac_f32_e32 v2, 0xcf800000, v3
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v6, v2
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v7, v3
 ; GCN-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v6, s6, 0
 ; GCN-NEXT:    v_mul_lo_u32 v4, v7, s6
 ; GCN-NEXT:    v_sub_u32_e32 v3, vcc, v3, v6
-; GCN-NEXT:    v_add_u32_e32 v8, vcc, v4, v3
+; GCN-NEXT:    v_add_u32_e32 v8, vcc, v3, v4
 ; GCN-NEXT:    v_mul_hi_u32 v5, v6, v2
 ; GCN-NEXT:    v_mad_u64_u32 v[3:4], s[4:5], v6, v8, 0
 ; GCN-NEXT:    v_add_u32_e32 v9, vcc, v5, v3
 ; GCN-NEXT:    v_mad_u64_u32 v[2:3], s[4:5], v7, v2, 0
 ; GCN-NEXT:    v_addc_u32_e32 v10, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_mad_u64_u32 v[4:5], s[4:5], v7, v8, 0
 ; GCN-NEXT:    v_add_u32_e32 v2, vcc, v9, v2
 ; GCN-NEXT:    v_addc_u32_e32 v2, vcc, v10, v3, vcc

llvm/test/CodeGen/AMDGPU/udiv64.ll

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s5, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s4, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s5, v0
 ; GCN-NEXT:    s_mov_b32 s5, s1
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s4, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    v_mul_lo_u32 v3, s8, v2
 ; GCN-NEXT:    v_mul_hi_u32 v4, s8, v1
 ; GCN-NEXT:    v_mul_lo_u32 v5, s9, v1
 ; GCN-NEXT:    v_mul_lo_u32 v6, s8, v1
 ; GCN-NEXT:    s_mov_b32 s1, s5
-; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
-; GCN-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
+; GCN-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
+; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v6
 ; GCN-NEXT:    v_mul_hi_u32 v7, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v2, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v2, v3
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v7, v2, v6
 ; GCN-NEXT:    v_mul_hi_u32 v6, v2, v6
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v4, v7
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v5, vcc
 ; GCN-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
 ; GCN-NEXT:    v_addc_u32_e32 v2, vcc, v2, v4, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v3, s8, v2
 ; GCN-NEXT:    v_mul_hi_u32 v4, s8, v1
 ; GCN-NEXT:    v_mul_lo_u32 v5, s9, v1
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v4, v3
 ; GCN-NEXT:    v_mul_lo_u32 v4, s8, v1
-; GCN-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
+; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_mul_lo_u32 v7, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v9, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v6, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v2, v4
 ; GCN-NEXT:    v_mul_hi_u32 v5, v2, v3
 ; GCN-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
 ; GCN-NEXT:    v_addc_u32_e32 v8, vcc, 0, v9, vcc

 ; GCN-NEXT:    v_addc_u32_e32 v2, vcc, 0, v2, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v1
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, 1, v1
 ; GCN-NEXT:    v_mul_lo_u32 v10, v0, v1
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v2, vcc
 ; GCN-NEXT:    v_add_i32_e32 v8, vcc, 2, v1
 ; GCN-NEXT:    v_addc_u32_e32 v9, vcc, 0, v2, vcc
-; GCN-NEXT:    v_add_i32_e32 v6, vcc, v6, v7
+; GCN-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GCN-NEXT:    v_sub_i32_e32 v3, vcc, v3, v10
 ; GCN-NEXT:    v_subb_u32_e32 v6, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_sub_i32_e32 v7, vcc, v3, v0
 ; GCN-NEXT:    v_subbrev_u32_e32 v10, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_cmp_ge_u32_e32 vcc, v7, v0
 ; GCN-NEXT:    v_cndmask_b32_e64 v7, 0, -1, vcc
 ; GCN-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v10
 ; GCN-NEXT:    v_cndmask_b32_e32 v7, -1, v7, vcc

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s5, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s4, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s4, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s4, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s5, v0
 ; GCN-NEXT:    s_mov_b32 s5, s1
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s4, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
 ; GCN-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_hi_u32 v2, v0, s8
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, s8
 ; GCN-NEXT:    v_mul_lo_u32 v4, v0, s8
-; GCN-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v0, v4
 ; GCN-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v6, v1, v4

 ; GCN-NEXT:    v_addc_u32_e32 v6, vcc, 0, v9, vcc
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v3, v5, vcc
 ; GCN-NEXT:    v_mul_hi_u32 v4, v2, s4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v3, s4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v2, s4
-; GCN-NEXT:    v_subrev_i32_e32 v4, vcc, v2, v4
+; GCN-NEXT:    v_sub_i32_e32 v4, vcc, v4, v2
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
 ; GCN-NEXT:    v_mul_lo_u32 v5, v2, v4
 ; GCN-NEXT:    v_mul_hi_u32 v7, v2, v6
 ; GCN-NEXT:    v_mul_hi_u32 v8, v2, v4
 ; GCN-NEXT:    v_mul_hi_u32 v9, v3, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v3, v4
 ; GCN-NEXT:    v_add_i32_e32 v5, vcc, v7, v5
 ; GCN-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc

llvm/test/CodeGen/AMDGPU/udivrem.ll

 ; GFX6-NEXT:    s_sub_i32 s2, 0, s6
 ; GFX6-NEXT:    v_cvt_f32_u32_e32 v2, s7
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    v_mul_lo_u32 v1, s2, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s6
 ; GFX6-NEXT:    s_sub_i32 s2, s4, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s6
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s6
 ; GFX6-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s6
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s6
 ; GFX6-NEXT:    s_cselect_b32 s4, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s2, 0, s7
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s2, v1
 ; GFX6-NEXT:    s_mov_b32 s3, 0xf000
 ; GFX6-NEXT:    s_mov_b32 s2, -1
 ; GFX6-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s6, v0
 ; GFX6-NEXT:    s_mul_i32 s6, s6, s7
 ; GFX6-NEXT:    s_sub_i32 s5, s5, s6
 ; GFX6-NEXT:    s_sub_i32 s6, s5, s7
 ; GFX6-NEXT:    s_cmp_ge_u32 s5, s7
 ; GFX6-NEXT:    s_cselect_b32 s5, s6, s5
 ; GFX6-NEXT:    s_sub_i32 s6, s5, s7

 ; GFX8-NEXT:    s_sub_i32 s2, 0, s6
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v2, s7
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX8-NEXT:    v_mul_lo_u32 v1, s2, v0
 ; GFX8-NEXT:    v_mul_hi_u32 v1, v0, v1
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX8-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX8-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX8-NEXT:    s_mul_i32 s2, s2, s6
 ; GFX8-NEXT:    s_sub_i32 s2, s4, s2
 ; GFX8-NEXT:    s_sub_i32 s3, s2, s6
 ; GFX8-NEXT:    s_cmp_ge_u32 s2, s6
 ; GFX8-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX8-NEXT:    s_sub_i32 s3, s2, s6
 ; GFX8-NEXT:    s_cmp_ge_u32 s2, s6
 ; GFX8-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX8-NEXT:    s_sub_i32 s3, 0, s7
 ; GFX8-NEXT:    v_mul_lo_u32 v0, s3, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX8-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
 ; GFX8-NEXT:    v_mul_hi_u32 v1, s5, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX8-NEXT:    v_readfirstlane_b32 s2, v1
 ; GFX8-NEXT:    s_mul_i32 s2, s2, s7
 ; GFX8-NEXT:    s_sub_i32 s2, s5, s2
 ; GFX8-NEXT:    s_sub_i32 s3, s2, s7
 ; GFX8-NEXT:    s_cmp_ge_u32 s2, s7
 ; GFX8-NEXT:    s_cselect_b32 s2, s3, s2

 ; GFX6-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s8
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s8
 ; GFX6-NEXT:    s_cselect_b32 s4, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s2, 0, s9
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s2, v1
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_cvt_f32_u32_e32 v2, s11
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s9
 ; GFX6-NEXT:    s_sub_i32 s2, s5, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s9
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s9
 ; GFX6-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s9
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s9
 ; GFX6-NEXT:    s_cselect_b32 s5, s3, s2
 ; GFX6-NEXT:    s_sub_i32 s2, 0, s10
 ; GFX6-NEXT:    v_mul_lo_u32 v0, s2, v1
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX6-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s6, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX6-NEXT:    v_readfirstlane_b32 s2, v0
 ; GFX6-NEXT:    s_mul_i32 s2, s2, s10
 ; GFX6-NEXT:    s_sub_i32 s2, s6, s2
 ; GFX6-NEXT:    s_sub_i32 s3, s2, s10
 ; GFX6-NEXT:    s_cmp_ge_u32 s2, s10

 ; GFX8-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX8-NEXT:    s_sub_i32 s3, s2, s8
 ; GFX8-NEXT:    s_cmp_ge_u32 s2, s8
 ; GFX8-NEXT:    s_cselect_b32 s2, s3, s2
 ; GFX8-NEXT:    s_sub_i32 s3, 0, s9
 ; GFX8-NEXT:    v_mul_lo_u32 v0, s3, v1
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX8-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s5, v0
 ; GFX8-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX8-NEXT:    v_cvt_f32_u32_e32 v2, s11
 ; GFX8-NEXT:    v_readfirstlane_b32 s3, v0
 ; GFX8-NEXT:    s_mul_i32 s3, s3, s9
 ; GFX8-NEXT:    s_sub_i32 s3, s5, s3
 ; GFX8-NEXT:    s_sub_i32 s4, s3, s9
 ; GFX8-NEXT:    s_cmp_ge_u32 s3, s9
 ; GFX8-NEXT:    s_cselect_b32 s3, s4, s3
 ; GFX8-NEXT:    s_sub_i32 s4, s3, s9
 ; GFX8-NEXT:    s_cmp_ge_u32 s3, s9
 ; GFX8-NEXT:    s_cselect_b32 s3, s4, s3
 ; GFX8-NEXT:    s_sub_i32 s4, 0, s10
 ; GFX8-NEXT:    v_mul_lo_u32 v0, s4, v1
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX8-NEXT:    v_mul_hi_u32 v0, v1, v0
-; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s6, v0
 ; GFX8-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GFX8-NEXT:    v_readfirstlane_b32 s4, v0
 ; GFX8-NEXT:    s_mul_i32 s4, s4, s10
 ; GFX8-NEXT:    s_sub_i32 s4, s6, s4
 ; GFX8-NEXT:    s_sub_i32 s5, s4, s10
 ; GFX8-NEXT:    s_cmp_ge_u32 s4, s10

llvm/test/CodeGen/AMDGPU/urem.ll

-; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC1 %s
+; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=FUNC2 %s
+; RUN:  llc -amdgpu-scalarize-global-loads=false  -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG %s
 
 ; The code generated by urem is long and complex and may frequently
 ; change.  The goal of this test is to make sure the ISel doesn't fail
 ; when it gets a v2i32/v4i32 urem
 
-; FUNC-LABEL: {{^}}test_urem_i32:
-; SI: s_endpgm
-; EG: CF_END
 define amdgpu_kernel void @test_urem_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_i32:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s6, -1
+; FUNC1-NEXT:    s_mov_b32 s10, s6
+; FUNC1-NEXT:    s_mov_b32 s11, s7
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s8, s2
+; FUNC1-NEXT:    s_mov_b32 s9, s3
+; FUNC1-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; FUNC1-NEXT:    s_mov_b32 s4, s0
+; FUNC1-NEXT:    s_mov_b32 s5, s1
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_readfirstlane_b32 s2, v1
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, s2
+; FUNC1-NEXT:    s_sub_i32 s3, 0, s2
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v2, s3, v1
+; FUNC1-NEXT:    v_readfirstlane_b32 s3, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v2, v1, v2
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v1, v2
+; FUNC1-NEXT:    v_mul_hi_u32 v0, s3, v0
+; FUNC1-NEXT:    v_readfirstlane_b32 s0, v0
+; FUNC1-NEXT:    s_mul_i32 s0, s0, s2
+; FUNC1-NEXT:    s_sub_i32 s0, s3, s0
+; FUNC1-NEXT:    s_sub_i32 s1, s0, s2
+; FUNC1-NEXT:    s_cmp_ge_u32 s0, s2
+; FUNC1-NEXT:    s_cselect_b32 s0, s1, s0
+; FUNC1-NEXT:    s_sub_i32 s1, s0, s2
+; FUNC1-NEXT:    s_cmp_ge_u32 s0, s2
+; FUNC1-NEXT:    s_cselect_b32 s0, s1, s0
+; FUNC1-NEXT:    v_mov_b32_e32 v0, s0
+; FUNC1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; FUNC1-NEXT:    s_endpgm
+;
+; FUNC2-LABEL: test_urem_i32:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s6, -1
+; FUNC2-NEXT:    s_mov_b32 s10, s6
+; FUNC2-NEXT:    s_mov_b32 s11, s7
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s8, s2
+; FUNC2-NEXT:    s_mov_b32 s9, s3
+; FUNC2-NEXT:    buffer_load_dwordx2 v[0:1], off, s[8:11], 0
+; FUNC2-NEXT:    s_mov_b32 s4, s0
+; FUNC2-NEXT:    s_mov_b32 s5, s1
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_readfirstlane_b32 s2, v1
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, s2
+; FUNC2-NEXT:    s_sub_i32 s3, 0, s2
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v2, s3, v1
+; FUNC2-NEXT:    v_readfirstlane_b32 s3, v0
+; FUNC2-NEXT:    v_mul_hi_u32 v2, v1, v2
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v1, v2
+; FUNC2-NEXT:    v_mul_hi_u32 v0, s3, v0
+; FUNC2-NEXT:    v_readfirstlane_b32 s0, v0
+; FUNC2-NEXT:    s_mul_i32 s0, s0, s2
+; FUNC2-NEXT:    s_sub_i32 s0, s3, s0
+; FUNC2-NEXT:    s_sub_i32 s1, s0, s2
+; FUNC2-NEXT:    s_cmp_ge_u32 s0, s2
+; FUNC2-NEXT:    s_cselect_b32 s0, s1, s0
+; FUNC2-NEXT:    s_sub_i32 s1, s0, s2
+; FUNC2-NEXT:    s_cmp_ge_u32 s0, s2
+; FUNC2-NEXT:    s_cselect_b32 s0, s1, s0
+; FUNC2-NEXT:    v_mov_b32_e32 v0, s0
+; FUNC2-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; FUNC2-NEXT:    s_endpgm
+;
+; EG-LABEL: test_urem_i32:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 15, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_64 T0.XY, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     SUB_INT T0.W, 0.0, T0.Y,
+; EG-NEXT:     RECIP_UINT * T0.Z, T0.Y,
+; EG-NEXT:     MULLO_INT * T0.W, PV.W, PS,
+; EG-NEXT:     MULHI * T0.W, T0.Z, PS,
+; EG-NEXT:     ADD_INT * T0.W, T0.Z, PS,
+; EG-NEXT:     MULHI * T0.Z, T0.X, PV.W,
+; EG-NEXT:     MULLO_INT * T0.Z, PS, T0.Y,
+; EG-NEXT:     SUB_INT * T0.W, T0.X, PS,
+; EG-NEXT:     SETGE_UINT T1.W, PV.W, T0.Y,
+; EG-NEXT:     SUB_INT * T2.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT * T0.W, PV.W, T0.W, PS,
+; EG-NEXT:     SETGE_UINT T1.W, PV.W, T0.Y,
+; EG-NEXT:     SUB_INT * T2.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T0.X, PV.W, T0.W, PS,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %b_ptr = getelementptr i32, ptr addrspace(1) %in, i32 1
   %a = load i32, ptr addrspace(1) %in
   %b = load i32, ptr addrspace(1) %b_ptr
   %result = urem i32 %a, %b
   store i32 %result, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_urem_i32_7:
-; SI: s_mov_b32 [[MAGIC:s[0-9]+]], 0x24924925
-; SI: v_mul_hi_u32 {{v[0-9]+}}, {{v[0-9]+}}, [[MAGIC]]
-; SI: v_sub_{{[iu]}}32
-; SI: v_mul_lo_u32
-; SI: v_subrev_{{[iu]}}32
-; SI: buffer_store_dword
-; SI: s_endpgm
 define amdgpu_kernel void @test_urem_i32_7(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_i32_7:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s6, -1
+; FUNC1-NEXT:    s_mov_b32 s10, s6
+; FUNC1-NEXT:    s_mov_b32 s11, s7
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s8, s2
+; FUNC1-NEXT:    s_mov_b32 s9, s3
+; FUNC1-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; FUNC1-NEXT:    s_mov_b32 s2, 0x24924925
+; FUNC1-NEXT:    s_mov_b32 s4, s0
+; FUNC1-NEXT:    s_mov_b32 s5, s1
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_mul_hi_u32 v1, v0, s2
+; FUNC1-NEXT:    v_sub_i32_e32 v2, vcc, v0, v1
+; FUNC1-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
+; FUNC1-NEXT:    v_lshrrev_b32_e32 v1, 2, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v1, 7
+; FUNC1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
+; FUNC1-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; FUNC1-NEXT:    s_endpgm
+;
+; FUNC2-LABEL: test_urem_i32_7:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s6, -1
+; FUNC2-NEXT:    s_mov_b32 s10, s6
+; FUNC2-NEXT:    s_mov_b32 s11, s7
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s8, s2
+; FUNC2-NEXT:    s_mov_b32 s9, s3
+; FUNC2-NEXT:    buffer_load_dword v0, off, s[8:11], 0
+; FUNC2-NEXT:    s_mov_b32 s2, 0x24924925
+; FUNC2-NEXT:    s_mov_b32 s4, s0
+; FUNC2-NEXT:    s_mov_b32 s5, s1
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_mul_hi_u32 v1, v0, s2
+; FUNC2-NEXT:    v_sub_u32_e32 v2, vcc, v0, v1
+; FUNC2-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v2, v1
+; FUNC2-NEXT:    v_lshrrev_b32_e32 v1, 2, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v1, v1, 7
+; FUNC2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v1
+; FUNC2-NEXT:    buffer_store_dword v0, off, s[4:7], 0
+; FUNC2-NEXT:    s_endpgm
+;
+; EG-LABEL: test_urem_i32_7:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 11, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_32 T0.X, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     MULHI * T0.Y, T0.X, literal.x,
+; EG-NEXT:    613566757(6.344132e-17), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT * T0.W, T0.X, PS,
+; EG-NEXT:     LSHR * T0.W, PV.W, 1,
+; EG-NEXT:     ADD_INT * T0.W, PV.W, T0.Y,
+; EG-NEXT:     LSHR * T0.W, PV.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     MULLO_INT * T0.Y, PV.W, literal.x,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T0.X, T0.X, PS,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %num = load i32, ptr addrspace(1) %in
   %result = urem i32 %num, 7
   store i32 %result, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_urem_v2i32:
-; SI: s_endpgm
-; EG: CF_END
 define amdgpu_kernel void @test_urem_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_v2i32:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s2, -1
+; FUNC1-NEXT:    s_mov_b32 s10, s2
+; FUNC1-NEXT:    s_mov_b32 s11, s3
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s8, s6
+; FUNC1-NEXT:    s_mov_b32 s9, s7
+; FUNC1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_readfirstlane_b32 s0, v2
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v2, s0
+; FUNC1-NEXT:    s_sub_i32 s1, 0, s0
+; FUNC1-NEXT:    v_readfirstlane_b32 s6, v3
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v3, s6
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; FUNC1-NEXT:    v_readfirstlane_b32 s8, v1
+; FUNC1-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v2, v2
+; FUNC1-NEXT:    v_mul_lo_u32 v4, s1, v2
+; FUNC1-NEXT:    v_readfirstlane_b32 s1, v0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v0, v3
+; FUNC1-NEXT:    v_mul_hi_u32 v4, v2, v4
+; FUNC1-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC1-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v2, s1, v2
+; FUNC1-NEXT:    v_readfirstlane_b32 s7, v2
+; FUNC1-NEXT:    s_mul_i32 s7, s7, s0
+; FUNC1-NEXT:    s_sub_i32 s1, s1, s7
+; FUNC1-NEXT:    s_sub_i32 s7, s1, s0
+; FUNC1-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC1-NEXT:    s_cselect_b32 s1, s7, s1
+; FUNC1-NEXT:    s_sub_i32 s7, s1, s0
+; FUNC1-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC1-NEXT:    s_cselect_b32 s7, s7, s1
+; FUNC1-NEXT:    s_sub_i32 s0, 0, s6
+; FUNC1-NEXT:    v_mul_lo_u32 v2, s0, v0
+; FUNC1-NEXT:    s_mov_b32 s0, s4
+; FUNC1-NEXT:    s_mov_b32 s1, s5
+; FUNC1-NEXT:    v_mul_hi_u32 v2, v0, v2
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
+; FUNC1-NEXT:    v_mul_hi_u32 v0, s8, v0
+; FUNC1-NEXT:    v_readfirstlane_b32 s4, v0
+; FUNC1-NEXT:    s_mul_i32 s4, s4, s6
+; FUNC1-NEXT:    s_sub_i32 s4, s8, s4
+; FUNC1-NEXT:    s_sub_i32 s5, s4, s6
+; FUNC1-NEXT:    s_cmp_ge_u32 s4, s6
+; FUNC1-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC1-NEXT:    s_sub_i32 s5, s4, s6
+; FUNC1-NEXT:    s_cmp_ge_u32 s4, s6
+; FUNC1-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC1-NEXT:    v_mov_b32_e32 v0, s7
+; FUNC1-NEXT:    v_mov_b32_e32 v1, s4
+; FUNC1-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; FUNC1-NEXT:    s_endpgm
+;
+; FUNC2-LABEL: test_urem_v2i32:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s2, -1
+; FUNC2-NEXT:    s_mov_b32 s10, s2
+; FUNC2-NEXT:    s_mov_b32 s11, s3
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s8, s6
+; FUNC2-NEXT:    s_mov_b32 s9, s7
+; FUNC2-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_readfirstlane_b32 s0, v2
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v2, s0
+; FUNC2-NEXT:    s_sub_i32 s1, 0, s0
+; FUNC2-NEXT:    v_readfirstlane_b32 s6, v3
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v3, s6
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; FUNC2-NEXT:    v_readfirstlane_b32 s8, v1
+; FUNC2-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v2, v2
+; FUNC2-NEXT:    v_mul_lo_u32 v4, s1, v2
+; FUNC2-NEXT:    v_readfirstlane_b32 s1, v0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v0, v3
+; FUNC2-NEXT:    v_mul_hi_u32 v4, v2, v4
+; FUNC2-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC2-NEXT:    v_add_u32_e32 v2, vcc, v2, v4
+; FUNC2-NEXT:    v_mul_hi_u32 v2, s1, v2
+; FUNC2-NEXT:    v_readfirstlane_b32 s7, v2
+; FUNC2-NEXT:    s_mul_i32 s7, s7, s0
+; FUNC2-NEXT:    s_sub_i32 s1, s1, s7
+; FUNC2-NEXT:    s_sub_i32 s7, s1, s0
+; FUNC2-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC2-NEXT:    s_cselect_b32 s1, s7, s1
+; FUNC2-NEXT:    s_sub_i32 s7, s1, s0
+; FUNC2-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC2-NEXT:    s_cselect_b32 s7, s7, s1
+; FUNC2-NEXT:    s_sub_i32 s0, 0, s6
+; FUNC2-NEXT:    v_mul_lo_u32 v2, s0, v0
+; FUNC2-NEXT:    s_mov_b32 s0, s4
+; FUNC2-NEXT:    s_mov_b32 s1, s5
+; FUNC2-NEXT:    v_mul_hi_u32 v2, v0, v2
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; FUNC2-NEXT:    v_mul_hi_u32 v0, s8, v0
+; FUNC2-NEXT:    v_readfirstlane_b32 s4, v0
+; FUNC2-NEXT:    s_mul_i32 s4, s4, s6
+; FUNC2-NEXT:    s_sub_i32 s4, s8, s4
+; FUNC2-NEXT:    s_sub_i32 s5, s4, s6
+; FUNC2-NEXT:    s_cmp_ge_u32 s4, s6
+; FUNC2-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC2-NEXT:    s_sub_i32 s5, s4, s6
+; FUNC2-NEXT:    s_cmp_ge_u32 s4, s6
+; FUNC2-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC2-NEXT:    v_mov_b32_e32 v0, s7
+; FUNC2-NEXT:    v_mov_b32_e32 v1, s4
+; FUNC2-NEXT:    buffer_store_dwordx2 v[0:1], off, s[0:3], 0
+; FUNC2-NEXT:    s_endpgm
+;
+; EG-LABEL: test_urem_v2i32:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @8, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @6
+; EG-NEXT:    ALU 29, @9, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 8:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 9:
+; EG-NEXT:     SUB_INT T1.W, 0.0, T0.W,
+; EG-NEXT:     RECIP_UINT * T1.X, T0.W,
+; EG-NEXT:     MULLO_INT * T1.Y, PV.W, PS,
+; EG-NEXT:     SUB_INT T1.W, 0.0, T0.Z,
+; EG-NEXT:     RECIP_UINT * T1.Z, T0.Z,
+; EG-NEXT:     MULLO_INT * T1.W, PV.W, PS,
+; EG-NEXT:     MULHI * T1.W, T1.Z, PS,
+; EG-NEXT:     ADD_INT T1.W, T1.Z, PS,
+; EG-NEXT:     MULHI * T1.Y, T1.X, T1.Y,
+; EG-NEXT:     ADD_INT T2.W, T1.X, PS,
+; EG-NEXT:     MULHI * T1.X, T0.X, PV.W,
+; EG-NEXT:     MULHI * T1.Y, T0.Y, PV.W,
+; EG-NEXT:     MULLO_INT * T1.Y, PS, T0.W,
+; EG-NEXT:     SUB_INT T1.W, T0.Y, PS,
+; EG-NEXT:     MULLO_INT * T0.Y, T1.X, T0.Z,
+; EG-NEXT:     SUB_INT T1.Z, T0.X, PS,
+; EG-NEXT:     SETGE_UINT T2.W, PV.W, T0.W,
+; EG-NEXT:     SUB_INT * T3.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T2.Z, PV.W, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T1.W, PV.Z, T0.Z,
+; EG-NEXT:     SUB_INT * T2.W, PV.Z, T0.Z,
+; EG-NEXT:     CNDE_INT T1.Z, PV.W, T1.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.W, PV.Z, T0.W,
+; EG-NEXT:     SUB_INT * T0.W, PV.Z, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, T2.Z, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T0.W, PV.Z, T0.Z,
+; EG-NEXT:     SUB_INT * T1.W, PV.Z, T0.Z,
+; EG-NEXT:     CNDE_INT T0.X, PV.W, T1.Z, PS,
+; EG-NEXT:     LSHR * T1.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %b_ptr = getelementptr <2 x i32>, ptr addrspace(1) %in, i32 1
   %a = load <2 x i32>, ptr addrspace(1) %in
   %b = load <2 x i32>, ptr addrspace(1) %b_ptr
   %result = urem <2 x i32> %a, %b
   store <2 x i32> %result, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_urem_v4i32:
-; SI: s_endpgm
-; EG: CF_END
 define amdgpu_kernel void @test_urem_v4i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_v4i32:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s2, -1
+; FUNC1-NEXT:    s_mov_b32 s10, s2
+; FUNC1-NEXT:    s_mov_b32 s11, s3
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s8, s6
+; FUNC1-NEXT:    s_mov_b32 s9, s7
+; FUNC1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0 offset:16
+; FUNC1-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0
+; FUNC1-NEXT:    s_waitcnt vmcnt(1)
+; FUNC1-NEXT:    v_readfirstlane_b32 s0, v0
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v0, s0
+; FUNC1-NEXT:    s_sub_i32 s1, 0, s0
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_readfirstlane_b32 s6, v4
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v8, s1, v0
+; FUNC1-NEXT:    v_readfirstlane_b32 s1, v1
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, s1
+; FUNC1-NEXT:    v_mul_hi_u32 v8, v0, v8
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v0, s6, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_readfirstlane_b32 s7, v0
+; FUNC1-NEXT:    s_mul_i32 s7, s7, s0
+; FUNC1-NEXT:    s_sub_i32 s6, s6, s7
+; FUNC1-NEXT:    s_sub_i32 s7, s6, s0
+; FUNC1-NEXT:    s_cmp_ge_u32 s6, s0
+; FUNC1-NEXT:    s_cselect_b32 s6, s7, s6
+; FUNC1-NEXT:    s_sub_i32 s7, s6, s0
+; FUNC1-NEXT:    s_cmp_ge_u32 s6, s0
+; FUNC1-NEXT:    s_cselect_b32 s6, s7, s6
+; FUNC1-NEXT:    s_sub_i32 s0, 0, s1
+; FUNC1-NEXT:    v_mul_lo_u32 v0, s0, v1
+; FUNC1-NEXT:    v_readfirstlane_b32 s0, v2
+; FUNC1-NEXT:    v_readfirstlane_b32 s7, v5
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v2, s0
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v1, v0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v0, s7, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_readfirstlane_b32 s8, v0
+; FUNC1-NEXT:    s_mul_i32 s8, s8, s1
+; FUNC1-NEXT:    s_sub_i32 s7, s7, s8
+; FUNC1-NEXT:    s_sub_i32 s8, s7, s1
+; FUNC1-NEXT:    s_cmp_ge_u32 s7, s1
+; FUNC1-NEXT:    s_cselect_b32 s7, s8, s7
+; FUNC1-NEXT:    s_sub_i32 s8, s7, s1
+; FUNC1-NEXT:    s_cmp_ge_u32 s7, s1
+; FUNC1-NEXT:    s_cselect_b32 s7, s8, s7
+; FUNC1-NEXT:    s_sub_i32 s1, 0, s0
+; FUNC1-NEXT:    v_mul_lo_u32 v0, s1, v1
+; FUNC1-NEXT:    v_readfirstlane_b32 s8, v3
+; FUNC1-NEXT:    v_readfirstlane_b32 s1, v6
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v2, s8
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v1, v0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v0, s1, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_readfirstlane_b32 s9, v0
+; FUNC1-NEXT:    s_mul_i32 s9, s9, s0
+; FUNC1-NEXT:    s_sub_i32 s1, s1, s9
+; FUNC1-NEXT:    s_sub_i32 s9, s1, s0
+; FUNC1-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC1-NEXT:    s_cselect_b32 s1, s9, s1
+; FUNC1-NEXT:    s_sub_i32 s9, s1, s0
+; FUNC1-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC1-NEXT:    s_cselect_b32 s9, s9, s1
+; FUNC1-NEXT:    s_sub_i32 s0, 0, s8
+; FUNC1-NEXT:    v_mul_lo_u32 v0, s0, v1
+; FUNC1-NEXT:    s_mov_b32 s0, s4
+; FUNC1-NEXT:    v_readfirstlane_b32 s4, v7
+; FUNC1-NEXT:    s_mov_b32 s1, s5
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v1, v0
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v1, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v2, s4, v0
+; FUNC1-NEXT:    v_mov_b32_e32 v0, s6
+; FUNC1-NEXT:    v_mov_b32_e32 v1, s7
+; FUNC1-NEXT:    v_readfirstlane_b32 s5, v2
+; FUNC1-NEXT:    s_mul_i32 s5, s5, s8
+; FUNC1-NEXT:    s_sub_i32 s4, s4, s5
+; FUNC1-NEXT:    s_sub_i32 s5, s4, s8
+; FUNC1-NEXT:    s_cmp_ge_u32 s4, s8
+; FUNC1-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC1-NEXT:    s_sub_i32 s5, s4, s8
+; FUNC1-NEXT:    s_cmp_ge_u32 s4, s8
+; FUNC1-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC1-NEXT:    v_mov_b32_e32 v2, s9
+; FUNC1-NEXT:    v_mov_b32_e32 v3, s4
+; FUNC1-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; FUNC1-NEXT:    s_endpgm
+;
+; FUNC2-LABEL: test_urem_v4i32:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s2, -1
+; FUNC2-NEXT:    s_mov_b32 s10, s2
+; FUNC2-NEXT:    s_mov_b32 s11, s3
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s8, s6
+; FUNC2-NEXT:    s_mov_b32 s9, s7
+; FUNC2-NEXT:    buffer_load_dwordx4 v[0:3], off, s[8:11], 0 offset:16
+; FUNC2-NEXT:    buffer_load_dwordx4 v[4:7], off, s[8:11], 0
+; FUNC2-NEXT:    s_waitcnt vmcnt(1)
+; FUNC2-NEXT:    v_readfirstlane_b32 s0, v0
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v0, s0
+; FUNC2-NEXT:    s_sub_i32 s1, 0, s0
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_readfirstlane_b32 s6, v4
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v8, s1, v0
+; FUNC2-NEXT:    v_readfirstlane_b32 s1, v1
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, s1
+; FUNC2-NEXT:    v_mul_hi_u32 v8, v0, v8
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v0, v8
+; FUNC2-NEXT:    v_mul_hi_u32 v0, s6, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_readfirstlane_b32 s7, v0
+; FUNC2-NEXT:    s_mul_i32 s7, s7, s0
+; FUNC2-NEXT:    s_sub_i32 s6, s6, s7
+; FUNC2-NEXT:    s_sub_i32 s7, s6, s0
+; FUNC2-NEXT:    s_cmp_ge_u32 s6, s0
+; FUNC2-NEXT:    s_cselect_b32 s6, s7, s6
+; FUNC2-NEXT:    s_sub_i32 s7, s6, s0
+; FUNC2-NEXT:    s_cmp_ge_u32 s6, s0
+; FUNC2-NEXT:    s_cselect_b32 s6, s7, s6
+; FUNC2-NEXT:    s_sub_i32 s0, 0, s1
+; FUNC2-NEXT:    v_mul_lo_u32 v0, s0, v1
+; FUNC2-NEXT:    v_readfirstlane_b32 s0, v2
+; FUNC2-NEXT:    v_readfirstlane_b32 s7, v5
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v2, s0
+; FUNC2-NEXT:    v_mul_hi_u32 v0, v1, v0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; FUNC2-NEXT:    v_mul_hi_u32 v0, s7, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_readfirstlane_b32 s8, v0
+; FUNC2-NEXT:    s_mul_i32 s8, s8, s1
+; FUNC2-NEXT:    s_sub_i32 s7, s7, s8
+; FUNC2-NEXT:    s_sub_i32 s8, s7, s1
+; FUNC2-NEXT:    s_cmp_ge_u32 s7, s1
+; FUNC2-NEXT:    s_cselect_b32 s7, s8, s7
+; FUNC2-NEXT:    s_sub_i32 s8, s7, s1
+; FUNC2-NEXT:    s_cmp_ge_u32 s7, s1
+; FUNC2-NEXT:    s_cselect_b32 s7, s8, s7
+; FUNC2-NEXT:    s_sub_i32 s1, 0, s0
+; FUNC2-NEXT:    v_mul_lo_u32 v0, s1, v1
+; FUNC2-NEXT:    v_readfirstlane_b32 s8, v3
+; FUNC2-NEXT:    v_readfirstlane_b32 s1, v6
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v2, s8
+; FUNC2-NEXT:    v_mul_hi_u32 v0, v1, v0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v2, v2
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; FUNC2-NEXT:    v_mul_hi_u32 v0, s1, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v2
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_readfirstlane_b32 s9, v0
+; FUNC2-NEXT:    s_mul_i32 s9, s9, s0
+; FUNC2-NEXT:    s_sub_i32 s1, s1, s9
+; FUNC2-NEXT:    s_sub_i32 s9, s1, s0
+; FUNC2-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC2-NEXT:    s_cselect_b32 s1, s9, s1
+; FUNC2-NEXT:    s_sub_i32 s9, s1, s0
+; FUNC2-NEXT:    s_cmp_ge_u32 s1, s0
+; FUNC2-NEXT:    s_cselect_b32 s9, s9, s1
+; FUNC2-NEXT:    s_sub_i32 s0, 0, s8
+; FUNC2-NEXT:    v_mul_lo_u32 v0, s0, v1
+; FUNC2-NEXT:    s_mov_b32 s0, s4
+; FUNC2-NEXT:    v_readfirstlane_b32 s4, v7
+; FUNC2-NEXT:    s_mov_b32 s1, s5
+; FUNC2-NEXT:    v_mul_hi_u32 v0, v1, v0
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v1, v0
+; FUNC2-NEXT:    v_mul_hi_u32 v2, s4, v0
+; FUNC2-NEXT:    v_mov_b32_e32 v0, s6
+; FUNC2-NEXT:    v_mov_b32_e32 v1, s7
+; FUNC2-NEXT:    v_readfirstlane_b32 s5, v2
+; FUNC2-NEXT:    s_mul_i32 s5, s5, s8
+; FUNC2-NEXT:    s_sub_i32 s4, s4, s5
+; FUNC2-NEXT:    s_sub_i32 s5, s4, s8
+; FUNC2-NEXT:    s_cmp_ge_u32 s4, s8
+; FUNC2-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC2-NEXT:    s_sub_i32 s5, s4, s8
+; FUNC2-NEXT:    s_cmp_ge_u32 s4, s8
+; FUNC2-NEXT:    s_cselect_b32 s4, s5, s4
+; FUNC2-NEXT:    v_mov_b32_e32 v2, s9
+; FUNC2-NEXT:    v_mov_b32_e32 v3, s4
+; FUNC2-NEXT:    buffer_store_dwordx4 v[0:3], off, s[0:3], 0
+; FUNC2-NEXT:    s_endpgm
+;
+; EG-LABEL: test_urem_v4i32:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @10, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 1 @6
+; EG-NEXT:    ALU 57, @11, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T3.XYZW, T0.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 6:
+; EG-NEXT:     VTX_READ_128 T1.XYZW, T0.X, 16, #1
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 10:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 11:
+; EG-NEXT:     SUB_INT T2.W, 0.0, T1.Z,
+; EG-NEXT:     RECIP_UINT * T2.X, T1.Z,
+; EG-NEXT:     MULLO_INT * T2.Y, PV.W, PS,
+; EG-NEXT:     SUB_INT T2.W, 0.0, T1.Y,
+; EG-NEXT:     RECIP_UINT * T2.Z, T1.Y,
+; EG-NEXT:     MULLO_INT * T2.W, PV.W, PS,
+; EG-NEXT:     MULHI * T2.W, T2.Z, PS,
+; EG-NEXT:     ADD_INT T2.W, T2.Z, PS,
+; EG-NEXT:     MULHI * T2.Y, T2.X, T2.Y,
+; EG-NEXT:     ADD_INT T3.W, T2.X, PS,
+; EG-NEXT:     MULHI * T2.X, T0.Y, PV.W,
+; EG-NEXT:     MULHI * T2.Y, T0.Z, PV.W,
+; EG-NEXT:     SUB_INT T2.W, 0.0, T1.W,
+; EG-NEXT:     RECIP_UINT * T2.Z, T1.W,
+; EG-NEXT:     MULLO_INT * T2.W, PV.W, PS,
+; EG-NEXT:     MULHI * T2.W, T2.Z, PS,
+; EG-NEXT:     ADD_INT T2.W, T2.Z, PS,
+; EG-NEXT:     MULLO_INT * T2.Y, T2.Y, T1.Z,
+; EG-NEXT:     MULHI * T2.Z, T0.W, PV.W,
+; EG-NEXT:     MULLO_INT * T2.Z, PS, T1.W,
+; EG-NEXT:     SUB_INT T2.W, 0.0, T1.X,
+; EG-NEXT:     RECIP_UINT * T3.X, T1.X,
+; EG-NEXT:     MULLO_INT * T2.W, PV.W, PS,
+; EG-NEXT:     MULHI * T2.W, T3.X, PS,
+; EG-NEXT:     ADD_INT T3.Y, T3.X, PS,
+; EG-NEXT:     SUB_INT T2.Z, T0.W, T2.Z,
+; EG-NEXT:     SUB_INT T0.W, T0.Z, T2.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     MULLO_INT * T0.Z, T2.X, T1.Y,
+; EG-NEXT:     SETGE_UINT T2.X, PV.W, T1.Z,
+; EG-NEXT:     SUB_INT T0.Y, T0.Y, PS,
+; EG-NEXT:     SETGE_UINT T0.Z, PV.Z, T1.W,
+; EG-NEXT:     SUB_INT T2.W, PV.Z, T1.W,
+; EG-NEXT:     MULHI * T2.Y, T0.X, PV.Y,
+; EG-NEXT:     SUB_INT T3.X, T0.W, T1.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.Z, T2.Z, PV.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T0.Z, PV.Y, T1.Y,
+; EG-NEXT:     SUB_INT T2.W, PV.Y, T1.Y,
+; EG-NEXT:     MULLO_INT * T2.Y, PS, T1.X,
+; EG-NEXT:     CNDE_INT T4.X, PV.Z, T0.Y, PV.W,
+; EG-NEXT:     SETGE_UINT T0.Y, PV.Y, T1.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T0.Z, PV.Y, T1.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T1.W, T0.X, PS,
+; EG-NEXT:     CNDE_INT * T0.W, T2.X, T0.W, PV.X,
+; EG-NEXT:     SETGE_UINT T0.X, PS, T1.Z,
+; EG-NEXT:     SUB_INT T2.Y, PS, T1.Z,
+; EG-NEXT:     SETGE_UINT T1.Z, PV.W, T1.X,
+; EG-NEXT:     SUB_INT T2.W, PV.W, T1.X,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.Y, PV.Z,
+; EG-NEXT:     CNDE_INT T0.Y, PV.Z, T1.W, PV.W,
+; EG-NEXT:     CNDE_INT T3.Z, PV.X, T0.W, PV.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T0.W, T4.X, T1.Y,
+; EG-NEXT:     SUB_INT * T1.W, T4.X, T1.Y,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, T4.X, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T0.W, PV.Y, T1.X,
+; EG-NEXT:     SUB_INT * T1.W, PV.Y, T1.X,
+; EG-NEXT:     CNDE_INT T3.X, PV.W, T0.Y, PS,
+; EG-NEXT:     LSHR * T0.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %b_ptr = getelementptr <4 x i32>, ptr addrspace(1) %in, i32 1
   %a = load <4 x i32>, ptr addrspace(1) %in
   %b = load <4 x i32>, ptr addrspace(1) %b_ptr
   %result = urem <4 x i32> %a, %b
   store <4 x i32> %result, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_urem_i64:
-; SI: s_endpgm
-; EG: CF_END
 define amdgpu_kernel void @test_urem_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_i64:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s2, -1
+; FUNC1-NEXT:    v_mov_b32_e32 v4, 0
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s0, s6
+; FUNC1-NEXT:    s_mov_b32 s1, s7
+; FUNC1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[0:3], 0
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_or_b32_e32 v5, v1, v3
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[4:5]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB4_4
+; FUNC1-NEXT:  ; %bb.1:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v4, v2
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v5, v3
+; FUNC1-NEXT:    v_sub_i32_e32 v6, vcc, 0, v2
+; FUNC1-NEXT:    v_subb_u32_e32 v7, vcc, 0, v3, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
+; FUNC1-NEXT:    v_rcp_f32_e32 v4, v4
+; FUNC1-NEXT:    v_mul_f32_e32 v4, 0x5f7ffffc, v4
+; FUNC1-NEXT:    v_mul_f32_e32 v5, 0x2f800000, v4
+; FUNC1-NEXT:    v_trunc_f32_e32 v5, v5
+; FUNC1-NEXT:    v_mac_f32_e32 v4, 0xcf800000, v5
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v5, v5
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v4, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v6, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v9, v6, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v7, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v6, v4
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v9, v4, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v4, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v12, v4, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v5, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v5, v8
+; FUNC1-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, 0, v12, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v5, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v10, v5, v10
+; FUNC1-NEXT:    v_add_i32_e32 v9, vcc, v9, v12
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, v11, v10, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v10, vcc, 0, v13, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, 0, v10, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v8
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, v5, v9, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v6, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v9, v6, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v7, v7, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v6, v6, v4
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v9, v8
+; FUNC1-NEXT:    v_add_i32_e32 v7, vcc, v8, v7
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v4, v7
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v4, v6
+; FUNC1-NEXT:    v_mul_hi_u32 v12, v4, v7
+; FUNC1-NEXT:    v_mul_hi_u32 v9, v5, v6
+; FUNC1-NEXT:    v_mul_lo_u32 v6, v5, v6
+; FUNC1-NEXT:    v_mul_hi_u32 v8, v5, v7
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, 0, v12, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v7, v5, v7
+; FUNC1-NEXT:    v_add_i32_e32 v6, vcc, v10, v6
+; FUNC1-NEXT:    v_addc_u32_e32 v6, vcc, v11, v9, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v8, vcc, 0, v8, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v6, vcc, v6, v7
+; FUNC1-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, v5, v7, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v6, v0, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v7, v0, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v8, v0, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v9, v1, v5
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v6, vcc, v7, v6
+; FUNC1-NEXT:    v_addc_u32_e32 v7, vcc, 0, v8, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v1, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v4, v1, v4
+; FUNC1-NEXT:    v_add_i32_e32 v6, vcc, v6, v8
+; FUNC1-NEXT:    v_addc_u32_e32 v4, vcc, v7, v4, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v6, vcc, 0, v9, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v2, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v6, v2, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v7, v3, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v2, v4
+; FUNC1-NEXT:    v_add_i32_e32 v5, vcc, v6, v5
+; FUNC1-NEXT:    v_add_i32_e32 v5, vcc, v5, v7
+; FUNC1-NEXT:    v_sub_i32_e32 v6, vcc, v1, v5
+; FUNC1-NEXT:    v_sub_i32_e32 v7, vcc, v0, v4
+; FUNC1-NEXT:    v_subb_u32_e64 v4, s[0:1], v6, v3, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v6, s[0:1], v7, v2
+; FUNC1-NEXT:    v_subbrev_u32_e64 v8, s[2:3], 0, v4, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v8, v3
+; FUNC1-NEXT:    v_cndmask_b32_e64 v9, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v6, v2
+; FUNC1-NEXT:    v_cndmask_b32_e64 v10, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v8, v3
+; FUNC1-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v9, v9, v10, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v10, s[0:1], v6, v2
+; FUNC1-NEXT:    v_subb_u32_e32 v1, vcc, v1, v5, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v9
+; FUNC1-NEXT:    v_cndmask_b32_e64 v5, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v7, v2
+; FUNC1-NEXT:    v_cndmask_b32_e64 v4, v8, v4, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v8, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v1, v3
+; FUNC1-NEXT:    v_cndmask_b32_e32 v3, v5, v8, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v6, v6, v10, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; FUNC1-NEXT:    v_cndmask_b32_e32 v4, v1, v4, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v3, v7, v6, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB4_3
+; FUNC1-NEXT:  .LBB4_2:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, v2
+; FUNC1-NEXT:    v_sub_i32_e32 v3, vcc, 0, v2
+; FUNC1-NEXT:    v_mov_b32_e32 v4, 0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v3, v3, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v3, v1, v3
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v1, v3
+; FUNC1-NEXT:    v_mul_hi_u32 v1, v0, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v1, v2
+; FUNC1-NEXT:    v_sub_i32_e32 v0, vcc, v0, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v0, v2
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v0, v2
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e32 v3, v0, v1, vcc
+; FUNC1-NEXT:  .LBB4_3:
+; FUNC1-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s6, -1
+; FUNC1-NEXT:    buffer_store_dwordx2 v[3:4], off, s[4:7], 0
+; FUNC1-NEXT:    s_endpgm
+; FUNC1-NEXT:  .LBB4_4:
+; FUNC1-NEXT:    ; implicit-def: $vgpr3_vgpr4
+; FUNC1-NEXT:    s_branch .LBB4_2
+;
+; FUNC2-LABEL: test_urem_i64:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s2, -1
+; FUNC2-NEXT:    v_mov_b32_e32 v4, 0
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s0, s6
+; FUNC2-NEXT:    s_mov_b32 s1, s7
+; FUNC2-NEXT:    buffer_load_dwordx4 v[0:3], off, s[0:3], 0
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_or_b32_e32 v5, v1, v3
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[4:5]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB4_4
+; FUNC2-NEXT:  ; %bb.1:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v4, v2
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v5, v3
+; FUNC2-NEXT:    v_sub_u32_e32 v10, vcc, 0, v2
+; FUNC2-NEXT:    v_subb_u32_e32 v11, vcc, 0, v3, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v4, 0x4f800000, v5
+; FUNC2-NEXT:    v_rcp_f32_e32 v4, v4
+; FUNC2-NEXT:    v_mul_f32_e32 v4, 0x5f7ffffc, v4
+; FUNC2-NEXT:    v_mul_f32_e32 v5, 0x2f800000, v4
+; FUNC2-NEXT:    v_trunc_f32_e32 v5, v5
+; FUNC2-NEXT:    v_mac_f32_e32 v4, 0xcf800000, v5
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v8, v5
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v9, v4
+; FUNC2-NEXT:    v_mul_lo_u32 v6, v10, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v10, v9, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v7, v11, v9
+; FUNC2-NEXT:    v_add_u32_e32 v5, vcc, v5, v6
+; FUNC2-NEXT:    v_add_u32_e32 v7, vcc, v5, v7
+; FUNC2-NEXT:    v_mul_hi_u32 v12, v9, v4
+; FUNC2-NEXT:    v_mad_u64_u32 v[5:6], s[0:1], v9, v7, 0
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v12, v5
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v8, v4, 0
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v6, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[6:7], s[0:1], v8, v7, 0
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v12, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v4, vcc, v13, v5, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v4, v6
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v9, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, v8, v5, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v10, v12, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v8, v10, v13
+; FUNC2-NEXT:    v_mul_lo_u32 v9, v11, v12
+; FUNC2-NEXT:    v_mul_hi_u32 v10, v12, v4
+; FUNC2-NEXT:    v_mad_u64_u32 v[6:7], s[0:1], v13, v4, 0
+; FUNC2-NEXT:    v_add_u32_e32 v5, vcc, v5, v8
+; FUNC2-NEXT:    v_add_u32_e32 v5, vcc, v5, v9
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v12, v5, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v13, v5, 0
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v10, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v6, vcc, v8, v6
+; FUNC2-NEXT:    v_addc_u32_e32 v6, vcc, v9, v7, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v6, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v6, vcc, v12, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v7, vcc, v13, v5, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v0, v7, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v8, v0, v6
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v8, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v1, v6, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[6:7], s[0:1], v1, v7, 0
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v8, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v4, vcc, v9, v5, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v6, vcc, v4, v6
+; FUNC2-NEXT:    v_addc_u32_e32 v4, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v7, v2, v4
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v2, v6, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v6, v3, v6
+; FUNC2-NEXT:    v_add_u32_e32 v5, vcc, v5, v7
+; FUNC2-NEXT:    v_add_u32_e32 v5, vcc, v5, v6
+; FUNC2-NEXT:    v_sub_u32_e32 v6, vcc, v1, v5
+; FUNC2-NEXT:    v_sub_u32_e32 v7, vcc, v0, v4
+; FUNC2-NEXT:    v_subb_u32_e64 v4, s[0:1], v6, v3, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v6, s[0:1], v7, v2
+; FUNC2-NEXT:    v_subbrev_u32_e64 v8, s[2:3], 0, v4, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v8, v3
+; FUNC2-NEXT:    v_cndmask_b32_e64 v9, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v6, v2
+; FUNC2-NEXT:    v_cndmask_b32_e64 v10, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v8, v3
+; FUNC2-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v9, v9, v10, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v10, s[0:1], v6, v2
+; FUNC2-NEXT:    v_subb_u32_e32 v1, vcc, v1, v5, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v9
+; FUNC2-NEXT:    v_cndmask_b32_e64 v5, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v7, v2
+; FUNC2-NEXT:    v_cndmask_b32_e64 v4, v8, v4, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v8, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v1, v3
+; FUNC2-NEXT:    v_cndmask_b32_e32 v3, v5, v8, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v6, v6, v10, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; FUNC2-NEXT:    v_cndmask_b32_e32 v4, v1, v4, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v3, v7, v6, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB4_3
+; FUNC2-NEXT:  .LBB4_2:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, v2
+; FUNC2-NEXT:    v_sub_u32_e32 v3, vcc, 0, v2
+; FUNC2-NEXT:    v_mov_b32_e32 v4, 0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v3, v3, v1
+; FUNC2-NEXT:    v_mul_hi_u32 v3, v1, v3
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v3
+; FUNC2-NEXT:    v_mul_hi_u32 v1, v0, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v1, v1, v2
+; FUNC2-NEXT:    v_sub_u32_e32 v0, vcc, v0, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v0, v2
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v0, v2
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e32 v3, v0, v1, vcc
+; FUNC2-NEXT:  .LBB4_3:
+; FUNC2-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s6, -1
+; FUNC2-NEXT:    buffer_store_dwordx2 v[3:4], off, s[4:7], 0
+; FUNC2-NEXT:    s_endpgm
+; FUNC2-NEXT:  .LBB4_4:
+; FUNC2-NEXT:    ; implicit-def: $vgpr3_vgpr4
+; FUNC2-NEXT:    s_branch .LBB4_2
+;
+; EG-LABEL: test_urem_i64:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @14, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 0 @12
+; EG-NEXT:    ALU_PUSH_BEFORE 6, @15, KC0[], KC1[]
+; EG-NEXT:    JUMP @9 POP:1
+; EG-NEXT:    ALU 114, @22, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @137, KC0[], KC1[]
+; EG-NEXT:    ALU 116, @253, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @370, KC0[], KC1[]
+; EG-NEXT:    ALU_POP_AFTER 57, @486, KC0[], KC1[]
+; EG-NEXT:    ALU 18, @544, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.XY, T0.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 12:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T0.X, 0, #1
+; EG-NEXT:    ALU clause starting at 14:
+; EG-NEXT:     MOV * T0.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 15:
+; EG-NEXT:     OR_INT * T1.W, T0.Y, T0.W,
+; EG-NEXT:     SUB_INT T1.Y, 0.0, T0.Z,
+; EG-NEXT:     MOV T1.Z, literal.x,
+; EG-NEXT:     SETNE_INT T1.W, PV.W, 0.0,
+; EG-NEXT:     RECIP_UINT * T1.X, T0.Z,
+; EG-NEXT:    1(1.401298e-45), 0(0.000000e+00)
+; EG-NEXT:     PRED_SETNE_INT * ExecMask,PredicateBit (MASKED), PV.W, 0.0,
+; EG-NEXT:    ALU clause starting at 22:
+; EG-NEXT:     MULLO_INT * T1.Z, T1.Y, T1.X,
+; EG-NEXT:     MULHI * T1.Z, T1.X, PS,
+; EG-NEXT:     ADD_INT * T1.W, T1.X, PS,
+; EG-NEXT:     MULHI * T1.Z, T0.Y, PV.W,
+; EG-NEXT:     MULLO_INT * T1.Z, PS, T0.Z,
+; EG-NEXT:     SUB_INT * T1.W, T0.Y, PS,
+; EG-NEXT:     SETGE_UINT T2.W, PV.W, T0.Z,
+; EG-NEXT:     SUB_INT * T3.W, PV.W, T0.Z,
+; EG-NEXT:     CNDE_INT * T1.W, PV.W, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.W, PV.W, T0.Z,
+; EG-NEXT:     SUB_INT * T3.W, PV.W, T0.Z,
+; EG-NEXT:     CNDE_INT * T1.W, PV.W, T1.W, PS,
+; EG-NEXT:     CNDE_INT * T1.W, T0.W, PV.W, T0.Y,
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PV.W, T0.X, literal.x,
+; EG-NEXT:     LSHR * T1.W, PV.W, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T1.Z, PS, T0.W,
+; EG-NEXT:     SETGE_UINT T3.W, PS, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.Z,
+; EG-NEXT:     CNDE_INT T0.Y, PV.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z, BS:VEC_201
+; EG-NEXT:     SUB_INT T3.W, T1.W, T0.W,
+; EG-NEXT:     SUBB_UINT * T4.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.W, PV.W, PS,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    30(4.203895e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    29(4.063766e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    26(3.643376e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:    ALU clause starting at 137:
+; EG-NEXT:     CNDE_INT T0.Y, T3.W, T4.W, T1.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z, BS:VEC_201
+; EG-NEXT:     SUBB_UINT * T3.W, T2.W, T0.Z, BS:VEC_201
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PV.W, T3.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T2.W, T0.Y, T2.W, T1.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    21(2.942727e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    19(2.662467e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:    ALU clause starting at 253:
+; EG-NEXT:     SUB_INT T3.W, T4.W, T3.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T2.W, T0.Y, T2.W, T1.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    14(1.961818e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    11(1.541428e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT * T4.W, T0.X, literal.x, 1,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 370:
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, T3.W,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.W, T2.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T2.W, T1.Z, T4.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    9(1.261169e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 486:
+; EG-NEXT:     SETGE_UINT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, T1.W, T0.W, BS:VEC_210
+; EG-NEXT:     SETGE_UINT * T4.W, T1.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T4.W, T0.X, 1, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T0.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T1.Z, T2.W, T0.Z,
+; EG-NEXT:     SUBB_UINT T3.W, T2.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, T1.W, T0.W,
+; EG-NEXT:     SUB_INT T3.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Y, T2.W, PV.Z,
+; EG-NEXT:     LSHL T1.Z, PS, 1,
+; EG-NEXT:     AND_INT T4.W, T0.X, 1,
+; EG-NEXT:     CNDE_INT * T1.W, T0.Y, T1.W, PV.W,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PS, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.X, PS, T0.Z,
+; EG-NEXT:     SETE_INT T0.Y, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T1.Z, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.W, PS, T0.Z,
+; EG-NEXT:     SUB_INT * T0.W, PV.W, T0.W,
+; EG-NEXT:     SUB_INT T0.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, PV.Z, PV.X,
+; EG-NEXT:     CNDE_INT T2.Y, PS, T1.W, PV.W,
+; EG-NEXT:     SUB_INT * T0.W, T2.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.X, T3.W, T2.W, PV.W,
+; EG-NEXT:     MOV * T1.Z, literal.x,
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 544:
+; EG-NEXT:     MOV T0.W, KC0[2].Y,
+; EG-NEXT:     SETE_INT * T1.W, T1.Z, 0.0,
+; EG-NEXT:     PRED_SETNE_INT * Pred,PredicateBit (MASKED), PS, 0.0,
+; EG-NEXT:     MULLO_INT * T0.Y, T1.Y, T1.X, Pred_sel_zero
+; EG-NEXT:     MULHI * T0.Y, T1.X, T0.Y, Pred_sel_zero
+; EG-NEXT:     ADD_INT * T1.W, T1.X, T0.Y, Pred_sel_zero
+; EG-NEXT:     MULHI * T0.Y, T0.X, T1.W, Pred_sel_zero
+; EG-NEXT:     MULLO_INT * T0.Y, T0.Y, T0.Z, Pred_sel_zero
+; EG-NEXT:     SUB_INT * T1.W, T0.X, T0.Y, Pred_sel_zero
+; EG-NEXT:     SETGE_UINT T2.W, T1.W, T0.Z, Pred_sel_zero
+; EG-NEXT:     SUB_INT * T3.W, T1.W, T0.Z, Pred_sel_zero
+; EG-NEXT:     CNDE_INT * T1.W, T2.W, T1.W, T3.W, Pred_sel_zero
+; EG-NEXT:     SETGE_UINT T2.W, T1.W, T0.Z, Pred_sel_zero
+; EG-NEXT:     SUB_INT * T3.W, T1.W, T0.Z, Pred_sel_zero
+; EG-NEXT:     CNDE_INT T2.X, T2.W, T1.W, T3.W, Pred_sel_zero
+; EG-NEXT:     MOV * T2.Y, literal.x, Pred_sel_zero
+; EG-NEXT:    0(0.000000e+00), 0(0.000000e+00)
+; EG-NEXT:     LSHR * T0.X, T0.W, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %b_ptr = getelementptr i64, ptr addrspace(1) %in, i64 1
   %a = load i64, ptr addrspace(1) %in
   %b = load i64, ptr addrspace(1) %b_ptr
   %result = urem i64 %a, %b
   store i64 %result, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_urem_v2i64:
-; SI: s_endpgm
-; EG: CF_END
 define amdgpu_kernel void @test_urem_v2i64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_v2i64:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s2, -1
+; FUNC1-NEXT:    v_mov_b32_e32 v8, 0
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s0, s6
+; FUNC1-NEXT:    s_mov_b32 s1, s7
+; FUNC1-NEXT:    buffer_load_dwordx4 v[4:7], off, s[0:3], 0
+; FUNC1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_or_b32_e32 v9, v5, v1
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[8:9]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB5_7
+; FUNC1-NEXT:  ; %bb.1:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v8, v0
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v9, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v10, vcc, 0, v0
+; FUNC1-NEXT:    v_subb_u32_e32 v11, vcc, 0, v1, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v8, 0x4f800000, v9
+; FUNC1-NEXT:    v_rcp_f32_e32 v8, v8
+; FUNC1-NEXT:    v_mul_f32_e32 v8, 0x5f7ffffc, v8
+; FUNC1-NEXT:    v_mul_f32_e32 v9, 0x2f800000, v8
+; FUNC1-NEXT:    v_trunc_f32_e32 v9, v9
+; FUNC1-NEXT:    v_mac_f32_e32 v8, 0xcf800000, v9
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v9, v9
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v10, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v10, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v11, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v10, v8
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v12, v15
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v8, v14
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v8, v12
+; FUNC1-NEXT:    v_mul_hi_u32 v16, v8, v12
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v9, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v9, v12
+; FUNC1-NEXT:    v_add_i32_e32 v13, vcc, v13, v15
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v16, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v9, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v9, v14
+; FUNC1-NEXT:    v_add_i32_e32 v13, vcc, v13, v16
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, v15, v14, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v14, vcc, 0, v17, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v12
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, v9, v13, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v10, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v10, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v11, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v10, v8
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
+; FUNC1-NEXT:    v_add_i32_e32 v11, vcc, v12, v11
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v8, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v8, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v16, v8, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v9, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v9, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v12, v9, v11
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v16, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v9, v11
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v14, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v10, vcc, v15, v13, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v12, vcc, 0, v12, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, 0, v12, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, v9, v11, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v4, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v4, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v12, v4, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v5, v9
+; FUNC1-NEXT:    v_mul_lo_u32 v9, v5, v9
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, 0, v12, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v5, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v8, v5, v8
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v12
+; FUNC1-NEXT:    v_addc_u32_e32 v8, vcc, v11, v8, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v10, vcc, 0, v13, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, 0, v10, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v9, v0, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v10, v0, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v1, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v0, v8
+; FUNC1-NEXT:    v_add_i32_e32 v9, vcc, v10, v9
+; FUNC1-NEXT:    v_add_i32_e32 v9, vcc, v9, v11
+; FUNC1-NEXT:    v_sub_i32_e32 v10, vcc, v5, v9
+; FUNC1-NEXT:    v_sub_i32_e32 v8, vcc, v4, v8
+; FUNC1-NEXT:    v_subb_u32_e64 v10, s[0:1], v10, v1, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v11, s[0:1], v8, v0
+; FUNC1-NEXT:    v_subbrev_u32_e64 v12, s[2:3], 0, v10, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v12, v1
+; FUNC1-NEXT:    v_cndmask_b32_e64 v13, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v11, v0
+; FUNC1-NEXT:    v_cndmask_b32_e64 v14, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v12, v1
+; FUNC1-NEXT:    v_subb_u32_e64 v10, s[0:1], v10, v1, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v13, v13, v14, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v14, s[0:1], v11, v0
+; FUNC1-NEXT:    v_subb_u32_e32 v5, vcc, v5, v9, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v10, s[0:1], 0, v10, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v1
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v13
+; FUNC1-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v0
+; FUNC1-NEXT:    v_cndmask_b32_e64 v10, v12, v10, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v12, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v5, v1
+; FUNC1-NEXT:    v_cndmask_b32_e32 v1, v9, v12, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v11, v11, v14, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v1
+; FUNC1-NEXT:    v_cndmask_b32_e32 v9, v5, v10, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v8, v8, v11, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB5_3
+; FUNC1-NEXT:  .LBB5_2:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, v0
+; FUNC1-NEXT:    v_sub_i32_e32 v5, vcc, 0, v0
+; FUNC1-NEXT:    v_mov_b32_e32 v9, 0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v5, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v5, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v1, v4, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v4, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, v1, v0
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC1-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, v1, v0
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC1-NEXT:    v_cndmask_b32_e32 v8, v1, v4, vcc
+; FUNC1-NEXT:  .LBB5_3:
+; FUNC1-NEXT:    v_or_b32_e32 v1, v7, v3
+; FUNC1-NEXT:    v_mov_b32_e32 v0, 0
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB5_8
+; FUNC1-NEXT:  ; %bb.4:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, v3
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, 0, v2
+; FUNC1-NEXT:    v_subb_u32_e32 v5, vcc, 0, v3, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v0, 0x4f800000, v1
+; FUNC1-NEXT:    v_rcp_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
+; FUNC1-NEXT:    v_trunc_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v4, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v4, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v13, v5, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v4, v0
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v13
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v0, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v13, v0, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v0, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v1, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v1, v10
+; FUNC1-NEXT:    v_add_i32_e32 v11, vcc, v11, v13
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v1, v12
+; FUNC1-NEXT:    v_mul_hi_u32 v12, v1, v12
+; FUNC1-NEXT:    v_add_i32_e32 v11, vcc, v11, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, v13, v12, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v12, vcc, 0, v15, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, 0, v12, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v1, vcc, v1, v11, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v4, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v4, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v5, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v4, v0
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v11, v10
+; FUNC1-NEXT:    v_add_i32_e32 v5, vcc, v10, v5
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v0, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v13, v0, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v0, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v1, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v1, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v10, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v13, v12
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v12, v4
+; FUNC1-NEXT:    v_addc_u32_e32 v4, vcc, v13, v11, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v10, vcc, 0, v10, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, 0, v10, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
+; FUNC1-NEXT:    v_addc_u32_e32 v1, vcc, v1, v5, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v6, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v5, v6, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v10, v6, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v7, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v7, v1
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, 0, v10, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v7, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v7, v0
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v10
+; FUNC1-NEXT:    v_addc_u32_e32 v0, vcc, v5, v0, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v4, vcc, 0, v11, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; FUNC1-NEXT:    v_addc_u32_e32 v1, vcc, 0, v4, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v2, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v4, v2, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v3, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v0, v2, v0
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v4, v1
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, v7, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v0, vcc, v6, v0
+; FUNC1-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v5, s[0:1], v0, v2
+; FUNC1-NEXT:    v_subbrev_u32_e64 v10, s[2:3], 0, v4, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v10, v3
+; FUNC1-NEXT:    v_cndmask_b32_e64 v11, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v2
+; FUNC1-NEXT:    v_cndmask_b32_e64 v12, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v10, v3
+; FUNC1-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v11, v11, v12, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v12, s[0:1], v5, v2
+; FUNC1-NEXT:    v_subb_u32_e32 v1, vcc, v7, v1, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v11
+; FUNC1-NEXT:    v_cndmask_b32_e64 v7, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e64 v4, v10, v4, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v10, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v1, v3
+; FUNC1-NEXT:    v_cndmask_b32_e32 v3, v7, v10, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v5, v5, v12, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; FUNC1-NEXT:    v_cndmask_b32_e32 v11, v1, v4, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v10, v0, v5, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB5_6
+; FUNC1-NEXT:  .LBB5_5:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, 0, v2
+; FUNC1-NEXT:    v_mov_b32_e32 v11, 0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v1, v0, v1
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v6, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v0, v0, v2
+; FUNC1-NEXT:    v_sub_i32_e32 v0, vcc, v6, v0
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v0, v2
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v0, v2
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e32 v10, v0, v1, vcc
+; FUNC1-NEXT:  .LBB5_6:
+; FUNC1-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s6, -1
+; FUNC1-NEXT:    buffer_store_dwordx4 v[8:11], off, s[4:7], 0
+; FUNC1-NEXT:    s_endpgm
+; FUNC1-NEXT:  .LBB5_7:
+; FUNC1-NEXT:    ; implicit-def: $vgpr8_vgpr9
+; FUNC1-NEXT:    s_branch .LBB5_2
+; FUNC1-NEXT:  .LBB5_8:
+; FUNC1-NEXT:    s_branch .LBB5_5
+;
+; FUNC2-LABEL: test_urem_v2i64:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s2, -1
+; FUNC2-NEXT:    v_mov_b32_e32 v8, 0
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s0, s6
+; FUNC2-NEXT:    s_mov_b32 s1, s7
+; FUNC2-NEXT:    buffer_load_dwordx4 v[4:7], off, s[0:3], 0
+; FUNC2-NEXT:    buffer_load_dwordx4 v[0:3], off, s[0:3], 0 offset:16
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_or_b32_e32 v9, v5, v1
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[8:9]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB5_7
+; FUNC2-NEXT:  ; %bb.1:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v8, v0
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v9, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v14, vcc, 0, v0
+; FUNC2-NEXT:    v_subb_u32_e32 v15, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v8, 0x4f800000, v9
+; FUNC2-NEXT:    v_rcp_f32_e32 v8, v8
+; FUNC2-NEXT:    v_mul_f32_e32 v8, 0x5f7ffffc, v8
+; FUNC2-NEXT:    v_mul_f32_e32 v9, 0x2f800000, v8
+; FUNC2-NEXT:    v_trunc_f32_e32 v9, v9
+; FUNC2-NEXT:    v_mac_f32_e32 v8, 0xcf800000, v9
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v12, v9
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v13, v8
+; FUNC2-NEXT:    v_mul_lo_u32 v10, v14, v12
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v14, v13, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v11, v15, v13
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v10
+; FUNC2-NEXT:    v_add_u32_e32 v11, vcc, v9, v11
+; FUNC2-NEXT:    v_mul_hi_u32 v16, v13, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[9:10], s[0:1], v13, v11, 0
+; FUNC2-NEXT:    v_add_u32_e32 v16, vcc, v16, v9
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v12, v8, 0
+; FUNC2-NEXT:    v_addc_u32_e32 v17, vcc, 0, v10, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v12, v11, 0
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v16, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v8, vcc, v17, v9, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v8, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v16, vcc, v13, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v17, vcc, v12, v9, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v14, v16, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v12, v14, v17
+; FUNC2-NEXT:    v_mul_lo_u32 v13, v15, v16
+; FUNC2-NEXT:    v_mul_hi_u32 v14, v16, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v17, v8, 0
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v12
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v13
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v16, v9, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v17, v9, 0
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v14, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v12, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v10, vcc, v13, v11, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v10, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v16, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, v17, v9, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v4, v11, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v12, v4, v10
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v12, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v5, v10, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v5, v11, 0
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v12, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v8, vcc, v13, v9, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v8, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v8, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v11, v0, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v0, v10, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v10, v1, v10
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v11
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v10
+; FUNC2-NEXT:    v_sub_u32_e32 v10, vcc, v5, v9
+; FUNC2-NEXT:    v_sub_u32_e32 v8, vcc, v4, v8
+; FUNC2-NEXT:    v_subb_u32_e64 v10, s[0:1], v10, v1, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v11, s[0:1], v8, v0
+; FUNC2-NEXT:    v_subbrev_u32_e64 v12, s[2:3], 0, v10, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v12, v1
+; FUNC2-NEXT:    v_cndmask_b32_e64 v13, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v11, v0
+; FUNC2-NEXT:    v_cndmask_b32_e64 v14, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v12, v1
+; FUNC2-NEXT:    v_subb_u32_e64 v10, s[0:1], v10, v1, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v13, v13, v14, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v14, s[0:1], v11, v0
+; FUNC2-NEXT:    v_subb_u32_e32 v5, vcc, v5, v9, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v10, s[0:1], 0, v10, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v1
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v13
+; FUNC2-NEXT:    v_cndmask_b32_e64 v9, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v0
+; FUNC2-NEXT:    v_cndmask_b32_e64 v10, v12, v10, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v12, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v5, v1
+; FUNC2-NEXT:    v_cndmask_b32_e32 v1, v9, v12, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v11, v11, v14, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v1
+; FUNC2-NEXT:    v_cndmask_b32_e32 v9, v5, v10, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v8, v8, v11, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB5_3
+; FUNC2-NEXT:  .LBB5_2:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, v0
+; FUNC2-NEXT:    v_sub_u32_e32 v5, vcc, 0, v0
+; FUNC2-NEXT:    v_mov_b32_e32 v9, 0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v5, v5, v1
+; FUNC2-NEXT:    v_mul_hi_u32 v5, v1, v5
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
+; FUNC2-NEXT:    v_mul_hi_u32 v1, v4, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v4, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v4, vcc, v1, v0
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC2-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v4, vcc, v1, v0
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC2-NEXT:    v_cndmask_b32_e32 v8, v1, v4, vcc
+; FUNC2-NEXT:  .LBB5_3:
+; FUNC2-NEXT:    v_or_b32_e32 v1, v7, v3
+; FUNC2-NEXT:    v_mov_b32_e32 v0, 0
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB5_8
+; FUNC2-NEXT:  ; %bb.4:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, v3
+; FUNC2-NEXT:    v_sub_u32_e32 v12, vcc, 0, v2
+; FUNC2-NEXT:    v_subb_u32_e32 v13, vcc, 0, v3, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v0, 0x4f800000, v1
+; FUNC2-NEXT:    v_rcp_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
+; FUNC2-NEXT:    v_trunc_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v10, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v11, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v4, v12, v10
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v12, v11, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v5, v13, v11
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v4
+; FUNC2-NEXT:    v_add_u32_e32 v15, vcc, v1, v5
+; FUNC2-NEXT:    v_mul_hi_u32 v14, v11, v0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v11, v15, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v10, v0, 0
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v14, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v16, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v10, v15, 0
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v14, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v0, vcc, v16, v1, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v0, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v11, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v15, vcc, v10, v1, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v12, v14, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v10, v12, v15
+; FUNC2-NEXT:    v_mul_lo_u32 v11, v13, v14
+; FUNC2-NEXT:    v_mul_hi_u32 v12, v14, v0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v15, v0, 0
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v10
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v11
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v14, v1, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v15, v1, 0
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v12, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v10, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v4, vcc, v11, v5, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v4, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v14, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, v15, v1, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v6, v5, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v10, v6, v4
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v10, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v7, v4, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v7, v5, 0
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v10, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v0, vcc, v11, v1, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v0, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v0, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v5, v2, v0
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v2, v4, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v4, v3, v4
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v4
+; FUNC2-NEXT:    v_sub_u32_e32 v4, vcc, v7, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v0, vcc, v6, v0
+; FUNC2-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v5, s[0:1], v0, v2
+; FUNC2-NEXT:    v_subbrev_u32_e64 v10, s[2:3], 0, v4, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v10, v3
+; FUNC2-NEXT:    v_cndmask_b32_e64 v11, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v2
+; FUNC2-NEXT:    v_cndmask_b32_e64 v12, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v10, v3
+; FUNC2-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v11, v11, v12, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v12, s[0:1], v5, v2
+; FUNC2-NEXT:    v_subb_u32_e32 v1, vcc, v7, v1, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v11
+; FUNC2-NEXT:    v_cndmask_b32_e64 v7, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e64 v4, v10, v4, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v10, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v1, v3
+; FUNC2-NEXT:    v_cndmask_b32_e32 v3, v7, v10, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v5, v5, v12, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; FUNC2-NEXT:    v_cndmask_b32_e32 v11, v1, v4, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v10, v0, v5, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB5_6
+; FUNC2-NEXT:  .LBB5_5:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, 0, v2
+; FUNC2-NEXT:    v_mov_b32_e32 v11, 0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC2-NEXT:    v_mul_hi_u32 v1, v0, v1
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; FUNC2-NEXT:    v_mul_hi_u32 v0, v6, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v0, v0, v2
+; FUNC2-NEXT:    v_sub_u32_e32 v0, vcc, v6, v0
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v0, v2
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v0, v2
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e32 v10, v0, v1, vcc
+; FUNC2-NEXT:  .LBB5_6:
+; FUNC2-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s6, -1
+; FUNC2-NEXT:    buffer_store_dwordx4 v[8:11], off, s[4:7], 0
+; FUNC2-NEXT:    s_endpgm
+; FUNC2-NEXT:  .LBB5_7:
+; FUNC2-NEXT:    ; implicit-def: $vgpr8_vgpr9
+; FUNC2-NEXT:    s_branch .LBB5_2
+; FUNC2-NEXT:  .LBB5_8:
+; FUNC2-NEXT:    s_branch .LBB5_5
+;
+; EG-LABEL: test_urem_v2i64:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @18, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 1 @14
+; EG-NEXT:    ALU 113, @19, KC0[], KC1[]
+; EG-NEXT:    ALU 116, @133, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @250, KC0[], KC1[]
+; EG-NEXT:    ALU 116, @366, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @483, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @599, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @715, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @831, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @947, KC0[], KC1[]
+; EG-NEXT:    ALU 1, @1063, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T4.XYZW, T0.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    Fetch clause starting at 14:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T1.X, 16, #1
+; EG-NEXT:     VTX_READ_128 T1.XYZW, T1.X, 0, #1
+; EG-NEXT:    ALU clause starting at 18:
+; EG-NEXT:     MOV * T1.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 19:
+; EG-NEXT:     RECIP_UINT * T2.X, T0.X,
+; EG-NEXT:     SUB_INT T2.W, 0.0, T0.Z,
+; EG-NEXT:     RECIP_UINT * T2.Y, T0.Z,
+; EG-NEXT:     MULLO_INT * T2.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T2.W, 0.0, T0.X,
+; EG-NEXT:     MULHI * T2.Z, T2.Y, PS,
+; EG-NEXT:     ADD_INT T3.W, T2.Y, PS,
+; EG-NEXT:     MULLO_INT * T2.Y, PV.W, T2.X,
+; EG-NEXT:     MULHI * T2.Z, T1.W, PV.W,
+; EG-NEXT:     MULLO_INT * T2.Z, PS, T0.Z,
+; EG-NEXT:     SUB_INT T2.W, T1.W, PS,
+; EG-NEXT:     MULHI * T2.Y, T2.X, T2.Y,
+; EG-NEXT:     ADD_INT T2.Z, T2.X, PS,
+; EG-NEXT:     SETGE_UINT T3.W, PV.W, T0.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.W, PV.W, T2.W, PS,
+; EG-NEXT:     MULHI * T2.X, T1.Y, PV.Z,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.W, PV.W, T0.Z,
+; EG-NEXT:     MULLO_INT * T2.X, PS, T0.X,
+; EG-NEXT:     CNDE_INT T2.W, PV.Z, T2.W, PV.W,
+; EG-NEXT:     SUB_INT * T3.W, T1.Y, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SUB_INT T4.W, PS, T0.X,
+; EG-NEXT:     CNDE_INT * T1.W, T0.W, PV.W, T1.W,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PS, T1.Z, literal.x,
+; EG-NEXT:     LSHR T1.W, PS, literal.x,
+; EG-NEXT:     CNDE_INT * T2.W, PV.Z, T3.W, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.X, PS, T0.X,
+; EG-NEXT:     SUB_INT T2.Y, PS, T0.X,
+; EG-NEXT:     SETE_INT T2.Z, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T3.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Z, PV.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T3.W, T3.Z, T0.Z,
+; EG-NEXT:     CNDE_INT * T2.W, PV.X, T2.W, PV.Y,
+; EG-NEXT:     SUB_INT T2.Y, T1.W, T0.W,
+; EG-NEXT:     SUBB_UINT T4.Z, T3.Z, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T2.W, T0.Y, PS, T1.Y,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Z, T3.Z, PV.W,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T1.X, literal.x,
+; EG-NEXT:     LSHR T2.W, PV.W, literal.x,
+; EG-NEXT:     SUB_INT * T4.W, PV.Y, PV.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T2.Y, T2.Z, T1.W, PS, BS:VEC_120/SCL_212
+; EG-NEXT:     SETE_INT T2.Z, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT T1.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.X,
+; EG-NEXT:    30(4.203895e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T2.Z, T3.Z, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T2.X, T2.W, T0.Y,
+; EG-NEXT:     SUBB_UINT T1.Y, T3.Z, T0.X,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T4.W, PV.X, PV.Y,
+; EG-NEXT:     CNDE_INT * T5.W, T3.Y, T3.Z, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, PV.W, BS:VEC_021/SCL_122
+; EG-NEXT:    30(4.203895e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT * T3.W, T2.Y, T3.W, T4.Z,
+; EG-NEXT:     LSHL T3.Y, PV.W, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, T2.W, T5.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, T1.Y, T2.Z,
+; EG-NEXT:     SUB_INT * T4.W, T2.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    29(4.063766e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    29(4.063766e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT * T2.Z, PV.W, T0.X,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 133:
+; EG-NEXT:     SETE_INT T1.W, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, T3.Z, T0.Y,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, T2.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X, BS:VEC_021/SCL_122
+; EG-NEXT:     BIT_ALIGN_INT T1.W, T1.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, T2.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    26(3.643376e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    26(3.643376e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT * T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 250:
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, T4.W,
+; EG-NEXT:     CNDE_INT * T3.W, T3.Y, T3.W, T4.Z,
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, T1.Y, T2.Z,
+; EG-NEXT:     SUB_INT * T4.W, T2.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    21(2.942727e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT * T1.Y, T3.Z, T0.Y,
+; EG-NEXT:    ALU clause starting at 366:
+; EG-NEXT:     SETGE_UINT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, T1.W, T0.W, BS:VEC_210
+; EG-NEXT:     SETGE_UINT * T5.W, T1.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, T1.Y, T2.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    21(2.942727e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    19(2.662467e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    19(2.662467e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT * T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 483:
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, T4.W,
+; EG-NEXT:     SETGE_UINT T2.Z, T2.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, T3.Z, T0.Y,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X, BS:VEC_021/SCL_122
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, T2.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    14(1.961818e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT * T2.X, T1.W, T0.W,
+; EG-NEXT:    ALU clause starting at 599:
+; EG-NEXT:     LSHL T1.Y, T2.W, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, T4.W,
+; EG-NEXT:     CNDE_INT * T3.W, T2.Y, T3.W, T4.Z,
+; EG-NEXT:    14(1.961818e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, T2.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    11(1.541428e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    11(1.541428e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT * T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 715:
+; EG-NEXT:     OR_INT * T3.W, T2.Y, T2.X,
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, T1.W, T0.W, BS:VEC_210
+; EG-NEXT:     SETGE_UINT * T5.W, T1.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    9(1.261169e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    9(1.261169e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT * T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 831:
+; EG-NEXT:     OR_INT T2.W, T1.Y, T2.Z,
+; EG-NEXT:     SUB_INT * T4.W, T2.X, T3.X,
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, T3.Z, T0.Y, BS:VEC_120/SCL_212
+; EG-NEXT:     SETGE_UINT * T4.W, T3.Z, T0.Y,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT * T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:    ALU clause starting at 947:
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, T1.Y, T2.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, T3.Y, T3.W, PV.Z,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T2.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T2.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.X, T3.W, T0.Z,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T3.Y, T2.W, T2.Z,
+; EG-NEXT:     SUB_INT T2.X, T1.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T4.W, T3.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT T2.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T4.W, PV.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T2.X, T1.Z, 1, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T2.Y, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T2.Z, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T1.W, PV.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT * T4.W, PV.Z, T0.Y,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T2.W, T0.X,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T3.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T2.X, T2.W, T0.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.Z, PS, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.W,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Z, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT T4.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T2.W, T2.Y, T2.W, T2.Z,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T2.Z, T1.X, 1, 1,
+; EG-NEXT:     CNDE_INT T4.W, T2.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T5.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.X, PS, 1,
+; EG-NEXT:     SUBB_UINT T2.Y, T3.W, T0.Z,
+; EG-NEXT:     SUB_INT * T3.Z, T1.W, T0.W, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T2.W, T4.W, T2.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, T1.Y, T2.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     AND_INT T3.X, T1.Z, 1,
+; EG-NEXT:     SETGE_UINT T1.Y, PS, T0.X,
+; EG-NEXT:     SETE_INT T1.Z, PV.W, T0.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T4.W, PV.W, T0.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT * T6.W, T3.Z, T2.Y,
+; EG-NEXT:     CNDE_INT T4.X, T3.Y, T1.W, PS,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T1.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T1.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T4.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T2.Y, PS, PV.W,
+; EG-NEXT:     CNDE_INT T1.Z, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.X, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T3.W, T2.X, T3.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.X, PS, T0.Z,
+; EG-NEXT:     SETE_INT T3.Y, PV.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T2.Z, PV.Z, 1,
+; EG-NEXT:     AND_INT T4.W, T1.X, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.Y,
+; EG-NEXT:     SETGE_UINT T1.X, T1.W, T0.W,
+; EG-NEXT:     SUBB_UINT T1.Y, T3.W, T0.Z, BS:VEC_201
+; EG-NEXT:     SUB_INT T3.Z, T1.W, T0.W,
+; EG-NEXT:     BIT_ALIGN_INT T0.W, PS, T1.Z, literal.x,
+; EG-NEXT:     OR_INT * T2.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Y, PS, T0.X,
+; EG-NEXT:     SETE_INT T1.Z, PV.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PV.Z, PV.Y,
+; EG-NEXT:     CNDE_INT * T5.W, T3.Y, PV.X, T2.X,
+; EG-NEXT:     SETGE_UINT T1.X, T0.W, T0.Y,
+; EG-NEXT:     SUBB_UINT T1.Y, T2.W, T0.X, BS:VEC_120/SCL_212
+; EG-NEXT:     SUB_INT T2.Z, T0.W, T0.Y,
+; EG-NEXT:     SUB_INT * T6.W, T3.W, T0.Z, BS:VEC_201
+; EG-NEXT:     CNDE_INT * T4.W, T5.W, T1.W, T4.W,
+; EG-NEXT:     CNDE_INT T4.Z, T5.W, T3.W, T6.W,
+; EG-NEXT:     SUB_INT T1.W, T2.Z, T1.Y,
+; EG-NEXT:     CNDE_INT * T3.W, T1.Z, T1.X, T2.Y,
+; EG-NEXT:     CNDE_INT T4.Y, PS, T0.W, PV.W,
+; EG-NEXT:     SUB_INT * T0.W, T2.W, T0.X,
+; EG-NEXT:     CNDE_INT * T4.X, T3.W, T2.W, PV.W,
+; EG-NEXT:    ALU clause starting at 1063:
+; EG-NEXT:     LSHR * T0.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %b_ptr = getelementptr <2 x i64>, ptr addrspace(1) %in, i64 1
   %a = load <2 x i64>, ptr addrspace(1) %in
   %b = load <2 x i64>, ptr addrspace(1) %b_ptr
   %result = urem <2 x i64> %a, %b
   store <2 x i64> %result, ptr addrspace(1) %out
   ret void
 }
 
-; FUNC-LABEL: {{^}}test_urem_v4i64:
-; SI: s_endpgm
-; EG: CF_END
 define amdgpu_kernel void @test_urem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
+; FUNC1-LABEL: test_urem_v4i64:
+; FUNC1:       ; %bb.0:
+; FUNC1-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x9
+; FUNC1-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s2, -1
+; FUNC1-NEXT:    v_mov_b32_e32 v8, 0
+; FUNC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC1-NEXT:    s_mov_b32 s0, s6
+; FUNC1-NEXT:    s_mov_b32 s1, s7
+; FUNC1-NEXT:    buffer_load_dwordx4 v[10:13], off, s[0:3], 0 offset:32
+; FUNC1-NEXT:    buffer_load_dwordx4 v[14:17], off, s[0:3], 0
+; FUNC1-NEXT:    buffer_load_dwordx4 v[4:7], off, s[0:3], 0 offset:16
+; FUNC1-NEXT:    buffer_load_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; FUNC1-NEXT:    s_waitcnt vmcnt(2)
+; FUNC1-NEXT:    v_or_b32_e32 v9, v15, v11
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[8:9]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB6_13
+; FUNC1-NEXT:  ; %bb.1:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v8, v10
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v9, v11
+; FUNC1-NEXT:    v_sub_i32_e32 v18, vcc, 0, v10
+; FUNC1-NEXT:    v_subb_u32_e32 v19, vcc, 0, v11, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v8, 0x4f800000, v9
+; FUNC1-NEXT:    v_rcp_f32_e32 v8, v8
+; FUNC1-NEXT:    v_mul_f32_e32 v8, 0x5f7ffffc, v8
+; FUNC1-NEXT:    v_mul_f32_e32 v9, 0x2f800000, v8
+; FUNC1-NEXT:    v_trunc_f32_e32 v9, v9
+; FUNC1-NEXT:    v_mac_f32_e32 v8, 0xcf800000, v9
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v9, v9
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v18, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v18, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v23, v19, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v22, v18, v8
+; FUNC1-NEXT:    v_add_i32_e32 v20, vcc, v21, v20
+; FUNC1-NEXT:    v_add_i32_e32 v20, vcc, v20, v23
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v8, v22
+; FUNC1-NEXT:    v_mul_lo_u32 v23, v8, v20
+; FUNC1-NEXT:    v_mul_hi_u32 v24, v8, v20
+; FUNC1-NEXT:    v_mul_hi_u32 v25, v9, v20
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v9, v20
+; FUNC1-NEXT:    v_add_i32_e32 v21, vcc, v21, v23
+; FUNC1-NEXT:    v_addc_u32_e32 v23, vcc, 0, v24, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v24, v9, v22
+; FUNC1-NEXT:    v_mul_hi_u32 v22, v9, v22
+; FUNC1-NEXT:    v_add_i32_e32 v21, vcc, v21, v24
+; FUNC1-NEXT:    v_addc_u32_e32 v21, vcc, v23, v22, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v22, vcc, 0, v25, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v20, vcc, v21, v20
+; FUNC1-NEXT:    v_addc_u32_e32 v21, vcc, 0, v22, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v20
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, v9, v21, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v18, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v18, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v19, v19, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v18, v8
+; FUNC1-NEXT:    v_add_i32_e32 v20, vcc, v21, v20
+; FUNC1-NEXT:    v_add_i32_e32 v19, vcc, v20, v19
+; FUNC1-NEXT:    v_mul_lo_u32 v22, v8, v19
+; FUNC1-NEXT:    v_mul_hi_u32 v23, v8, v18
+; FUNC1-NEXT:    v_mul_hi_u32 v24, v8, v19
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v9, v18
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v9, v18
+; FUNC1-NEXT:    v_mul_hi_u32 v20, v9, v19
+; FUNC1-NEXT:    v_add_i32_e32 v22, vcc, v23, v22
+; FUNC1-NEXT:    v_addc_u32_e32 v23, vcc, 0, v24, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v19, v9, v19
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v22, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v18, vcc, v23, v21, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v20, vcc, 0, v20, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v18, v19
+; FUNC1-NEXT:    v_addc_u32_e32 v19, vcc, 0, v20, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, v9, v19, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v14, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v14, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v20, v14, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v15, v9
+; FUNC1-NEXT:    v_mul_lo_u32 v9, v15, v9
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v19, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v19, vcc, 0, v20, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v15, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v8, v15, v8
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v18, v20
+; FUNC1-NEXT:    v_addc_u32_e32 v8, vcc, v19, v8, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v18, vcc, 0, v21, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; FUNC1-NEXT:    v_addc_u32_e32 v9, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v9, v10, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v18, v10, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v19, v11, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v10, v8
+; FUNC1-NEXT:    v_add_i32_e32 v9, vcc, v18, v9
+; FUNC1-NEXT:    v_add_i32_e32 v9, vcc, v9, v19
+; FUNC1-NEXT:    v_sub_i32_e32 v18, vcc, v15, v9
+; FUNC1-NEXT:    v_sub_i32_e32 v8, vcc, v14, v8
+; FUNC1-NEXT:    v_subb_u32_e64 v18, s[0:1], v18, v11, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v19, s[0:1], v8, v10
+; FUNC1-NEXT:    v_subbrev_u32_e64 v20, s[2:3], 0, v18, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v20, v11
+; FUNC1-NEXT:    v_cndmask_b32_e64 v21, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v19, v10
+; FUNC1-NEXT:    v_cndmask_b32_e64 v22, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v20, v11
+; FUNC1-NEXT:    v_subb_u32_e64 v18, s[0:1], v18, v11, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v21, v21, v22, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v22, s[0:1], v19, v10
+; FUNC1-NEXT:    v_subb_u32_e32 v9, vcc, v15, v9, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v18, s[0:1], 0, v18, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v9, v11
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v21
+; FUNC1-NEXT:    v_cndmask_b32_e64 v15, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v10
+; FUNC1-NEXT:    v_cndmask_b32_e64 v18, v20, v18, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v20, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v9, v11
+; FUNC1-NEXT:    v_cndmask_b32_e32 v11, v15, v20, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v19, v19, v22, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v11
+; FUNC1-NEXT:    v_cndmask_b32_e32 v9, v9, v18, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v8, v8, v19, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB6_3
+; FUNC1-NEXT:  .LBB6_2:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v8, v10
+; FUNC1-NEXT:    v_sub_i32_e32 v9, vcc, 0, v10
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v8, v8
+; FUNC1-NEXT:    v_mul_f32_e32 v8, 0x4f7ffffe, v8
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v9, v9, v8
+; FUNC1-NEXT:    v_mul_hi_u32 v9, v8, v9
+; FUNC1-NEXT:    v_add_i32_e32 v8, vcc, v8, v9
+; FUNC1-NEXT:    v_mul_hi_u32 v8, v14, v8
+; FUNC1-NEXT:    v_mul_lo_u32 v8, v8, v10
+; FUNC1-NEXT:    v_sub_i32_e32 v8, vcc, v14, v8
+; FUNC1-NEXT:    v_sub_i32_e32 v9, vcc, v8, v10
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v10
+; FUNC1-NEXT:    v_cndmask_b32_e32 v8, v8, v9, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v9, vcc, v8, v10
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v10
+; FUNC1-NEXT:    v_cndmask_b32_e32 v8, v8, v9, vcc
+; FUNC1-NEXT:    v_mov_b32_e32 v9, 0
+; FUNC1-NEXT:  .LBB6_3:
+; FUNC1-NEXT:    v_or_b32_e32 v11, v17, v13
+; FUNC1-NEXT:    v_mov_b32_e32 v10, 0
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[10:11]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB6_14
+; FUNC1-NEXT:  ; %bb.4:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v10, v12
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v11, v13
+; FUNC1-NEXT:    v_sub_i32_e32 v14, vcc, 0, v12
+; FUNC1-NEXT:    v_subb_u32_e32 v15, vcc, 0, v13, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v10, 0x4f800000, v11
+; FUNC1-NEXT:    v_rcp_f32_e32 v10, v10
+; FUNC1-NEXT:    v_mul_f32_e32 v10, 0x5f7ffffc, v10
+; FUNC1-NEXT:    v_mul_f32_e32 v11, 0x2f800000, v10
+; FUNC1-NEXT:    v_trunc_f32_e32 v11, v11
+; FUNC1-NEXT:    v_mac_f32_e32 v10, 0xcf800000, v11
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v11, v11
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v10, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v14, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v14, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v21, v15, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v14, v10
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v19, v18
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v18, v21
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v10, v20
+; FUNC1-NEXT:    v_mul_lo_u32 v21, v10, v18
+; FUNC1-NEXT:    v_mul_hi_u32 v22, v10, v18
+; FUNC1-NEXT:    v_mul_hi_u32 v23, v11, v18
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v11, v18
+; FUNC1-NEXT:    v_add_i32_e32 v19, vcc, v19, v21
+; FUNC1-NEXT:    v_addc_u32_e32 v21, vcc, 0, v22, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v22, v11, v20
+; FUNC1-NEXT:    v_mul_hi_u32 v20, v11, v20
+; FUNC1-NEXT:    v_add_i32_e32 v19, vcc, v19, v22
+; FUNC1-NEXT:    v_addc_u32_e32 v19, vcc, v21, v20, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v20, vcc, 0, v23, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v19, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v19, vcc, 0, v20, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, v11, v19, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v14, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v14, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v15, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v14, v10
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v19, v18
+; FUNC1-NEXT:    v_add_i32_e32 v15, vcc, v18, v15
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v10, v15
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v10, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v22, v10, v15
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v11, v14
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v11, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v18, v11, v15
+; FUNC1-NEXT:    v_add_i32_e32 v20, vcc, v21, v20
+; FUNC1-NEXT:    v_addc_u32_e32 v21, vcc, 0, v22, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v11, v15
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v20, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v14, vcc, v21, v19, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v18, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v14, v15
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, v11, v15, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v16, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v16, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v18, v16, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v17, v11
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v17, v11
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v17, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v10, v17, v10
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v14, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v10, vcc, v15, v10, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v14, vcc, 0, v19, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
+; FUNC1-NEXT:    v_addc_u32_e32 v11, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v12, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v12, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v13, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v12, v10
+; FUNC1-NEXT:    v_add_i32_e32 v11, vcc, v14, v11
+; FUNC1-NEXT:    v_add_i32_e32 v11, vcc, v11, v15
+; FUNC1-NEXT:    v_sub_i32_e32 v14, vcc, v17, v11
+; FUNC1-NEXT:    v_sub_i32_e32 v10, vcc, v16, v10
+; FUNC1-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v13, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v15, s[0:1], v10, v12
+; FUNC1-NEXT:    v_subbrev_u32_e64 v18, s[2:3], 0, v14, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v18, v13
+; FUNC1-NEXT:    v_cndmask_b32_e64 v19, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v15, v12
+; FUNC1-NEXT:    v_cndmask_b32_e64 v20, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v18, v13
+; FUNC1-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v13, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v19, v19, v20, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v20, s[0:1], v15, v12
+; FUNC1-NEXT:    v_subb_u32_e32 v11, vcc, v17, v11, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v14, s[0:1], 0, v14, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v11, v13
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v19
+; FUNC1-NEXT:    v_cndmask_b32_e64 v17, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; FUNC1-NEXT:    v_cndmask_b32_e64 v14, v18, v14, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v18, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v11, v13
+; FUNC1-NEXT:    v_cndmask_b32_e32 v13, v17, v18, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v15, v15, v20, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v13
+; FUNC1-NEXT:    v_cndmask_b32_e32 v11, v11, v14, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v10, v10, v15, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB6_6
+; FUNC1-NEXT:  .LBB6_5:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v10, v12
+; FUNC1-NEXT:    v_sub_i32_e32 v11, vcc, 0, v12
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v10, v10
+; FUNC1-NEXT:    v_mul_f32_e32 v10, 0x4f7ffffe, v10
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v10, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v11, v11, v10
+; FUNC1-NEXT:    v_mul_hi_u32 v11, v10, v11
+; FUNC1-NEXT:    v_add_i32_e32 v10, vcc, v10, v11
+; FUNC1-NEXT:    v_mul_hi_u32 v10, v16, v10
+; FUNC1-NEXT:    v_mul_lo_u32 v10, v10, v12
+; FUNC1-NEXT:    v_sub_i32_e32 v10, vcc, v16, v10
+; FUNC1-NEXT:    v_sub_i32_e32 v11, vcc, v10, v12
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; FUNC1-NEXT:    v_cndmask_b32_e32 v10, v10, v11, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v11, vcc, v10, v12
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; FUNC1-NEXT:    v_cndmask_b32_e32 v10, v10, v11, vcc
+; FUNC1-NEXT:    v_mov_b32_e32 v11, 0
+; FUNC1-NEXT:  .LBB6_6:
+; FUNC1-NEXT:    s_waitcnt vmcnt(0)
+; FUNC1-NEXT:    v_or_b32_e32 v13, v5, v1
+; FUNC1-NEXT:    v_mov_b32_e32 v12, 0
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[12:13]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB6_15
+; FUNC1-NEXT:  ; %bb.7:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v12, v0
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v13, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v14, vcc, 0, v0
+; FUNC1-NEXT:    v_subb_u32_e32 v15, vcc, 0, v1, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v12, 0x4f800000, v13
+; FUNC1-NEXT:    v_rcp_f32_e32 v12, v12
+; FUNC1-NEXT:    v_mul_f32_e32 v12, 0x5f7ffffc, v12
+; FUNC1-NEXT:    v_mul_f32_e32 v13, 0x2f800000, v12
+; FUNC1-NEXT:    v_trunc_f32_e32 v13, v13
+; FUNC1-NEXT:    v_mac_f32_e32 v12, 0xcf800000, v13
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v13, v13
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v12, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v14, v13
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v14, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v19, v15, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v14, v12
+; FUNC1-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
+; FUNC1-NEXT:    v_add_i32_e32 v16, vcc, v16, v19
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v12, v18
+; FUNC1-NEXT:    v_mul_lo_u32 v19, v12, v16
+; FUNC1-NEXT:    v_mul_hi_u32 v20, v12, v16
+; FUNC1-NEXT:    v_mul_hi_u32 v21, v13, v16
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v13, v16
+; FUNC1-NEXT:    v_add_i32_e32 v17, vcc, v17, v19
+; FUNC1-NEXT:    v_addc_u32_e32 v19, vcc, 0, v20, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v20, v13, v18
+; FUNC1-NEXT:    v_mul_hi_u32 v18, v13, v18
+; FUNC1-NEXT:    v_add_i32_e32 v17, vcc, v17, v20
+; FUNC1-NEXT:    v_addc_u32_e32 v17, vcc, v19, v18, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v18, vcc, 0, v21, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
+; FUNC1-NEXT:    v_addc_u32_e32 v17, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v12, v16
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, v13, v17, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v14, v13
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v14, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v15, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v14, v12
+; FUNC1-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
+; FUNC1-NEXT:    v_add_i32_e32 v15, vcc, v16, v15
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v12, v15
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v12, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v20, v12, v15
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v13, v14
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v13, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v16, v13, v15
+; FUNC1-NEXT:    v_add_i32_e32 v18, vcc, v19, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v19, vcc, 0, v20, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v13, v15
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v18, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v14, vcc, v19, v17, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v16, vcc, 0, v16, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v14, v15
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v16, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v12, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, v13, v15, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v4, v13
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v4, v12
+; FUNC1-NEXT:    v_mul_hi_u32 v16, v4, v13
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v5, v13
+; FUNC1-NEXT:    v_mul_lo_u32 v13, v5, v13
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v16, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v5, v12
+; FUNC1-NEXT:    v_mul_hi_u32 v12, v5, v12
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v14, v16
+; FUNC1-NEXT:    v_addc_u32_e32 v12, vcc, v15, v12, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v14, vcc, 0, v17, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v12, vcc, v12, v13
+; FUNC1-NEXT:    v_addc_u32_e32 v13, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v13, v0, v13
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v0, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v15, v1, v12
+; FUNC1-NEXT:    v_mul_lo_u32 v12, v0, v12
+; FUNC1-NEXT:    v_add_i32_e32 v13, vcc, v14, v13
+; FUNC1-NEXT:    v_add_i32_e32 v13, vcc, v13, v15
+; FUNC1-NEXT:    v_sub_i32_e32 v14, vcc, v5, v13
+; FUNC1-NEXT:    v_sub_i32_e32 v12, vcc, v4, v12
+; FUNC1-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v1, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v15, s[0:1], v12, v0
+; FUNC1-NEXT:    v_subbrev_u32_e64 v16, s[2:3], 0, v14, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v16, v1
+; FUNC1-NEXT:    v_cndmask_b32_e64 v17, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v15, v0
+; FUNC1-NEXT:    v_cndmask_b32_e64 v18, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v16, v1
+; FUNC1-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v1, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v17, v17, v18, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v18, s[0:1], v15, v0
+; FUNC1-NEXT:    v_subb_u32_e32 v5, vcc, v5, v13, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v14, s[0:1], 0, v14, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v1
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v17
+; FUNC1-NEXT:    v_cndmask_b32_e64 v13, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v12, v0
+; FUNC1-NEXT:    v_cndmask_b32_e64 v14, v16, v14, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v16, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v5, v1
+; FUNC1-NEXT:    v_cndmask_b32_e32 v1, v13, v16, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v15, v15, v18, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v1
+; FUNC1-NEXT:    v_cndmask_b32_e32 v13, v5, v14, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v12, v12, v15, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB6_9
+; FUNC1-NEXT:  .LBB6_8:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, v0
+; FUNC1-NEXT:    v_sub_i32_e32 v5, vcc, 0, v0
+; FUNC1-NEXT:    v_mov_b32_e32 v13, 0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v5, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v5, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v1, v4, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v4, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, v1, v0
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC1-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, v1, v0
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC1-NEXT:    v_cndmask_b32_e32 v12, v1, v4, vcc
+; FUNC1-NEXT:  .LBB6_9:
+; FUNC1-NEXT:    v_or_b32_e32 v1, v7, v3
+; FUNC1-NEXT:    v_mov_b32_e32 v0, 0
+; FUNC1-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
+; FUNC1-NEXT:    s_cbranch_vccz .LBB6_16
+; FUNC1-NEXT:  ; %bb.10:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v1, v3
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, 0, v2
+; FUNC1-NEXT:    v_subb_u32_e32 v5, vcc, 0, v3, vcc
+; FUNC1-NEXT:    v_mac_f32_e32 v0, 0x4f800000, v1
+; FUNC1-NEXT:    v_rcp_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
+; FUNC1-NEXT:    v_trunc_f32_e32 v1, v1
+; FUNC1-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v4, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v4, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v17, v5, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v4, v0
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v14, v17
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v0, v16
+; FUNC1-NEXT:    v_mul_lo_u32 v17, v0, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v18, v0, v14
+; FUNC1-NEXT:    v_mul_hi_u32 v19, v1, v14
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v1, v14
+; FUNC1-NEXT:    v_add_i32_e32 v15, vcc, v15, v17
+; FUNC1-NEXT:    v_addc_u32_e32 v17, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v18, v1, v16
+; FUNC1-NEXT:    v_mul_hi_u32 v16, v1, v16
+; FUNC1-NEXT:    v_add_i32_e32 v15, vcc, v15, v18
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, v17, v16, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v16, vcc, 0, v19, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v15, vcc, 0, v16, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v1, vcc, v1, v15, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v4, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v4, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v5, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v4, v0
+; FUNC1-NEXT:    v_add_i32_e32 v14, vcc, v15, v14
+; FUNC1-NEXT:    v_add_i32_e32 v5, vcc, v14, v5
+; FUNC1-NEXT:    v_mul_lo_u32 v16, v0, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v17, v0, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v18, v0, v5
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v1, v4
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v1, v4
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v16, vcc, v17, v16
+; FUNC1-NEXT:    v_addc_u32_e32 v17, vcc, 0, v18, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v1, v5
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v16, v4
+; FUNC1-NEXT:    v_addc_u32_e32 v4, vcc, v17, v15, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v14, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v5
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v4
+; FUNC1-NEXT:    v_addc_u32_e32 v1, vcc, v1, v5, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v4, v6, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v5, v6, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v14, v6, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v15, v7, v1
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v7, v1
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
+; FUNC1-NEXT:    v_addc_u32_e32 v5, vcc, 0, v14, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v14, v7, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v7, v0
+; FUNC1-NEXT:    v_add_i32_e32 v4, vcc, v4, v14
+; FUNC1-NEXT:    v_addc_u32_e32 v0, vcc, v5, v0, vcc
+; FUNC1-NEXT:    v_addc_u32_e32 v4, vcc, 0, v15, vcc
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; FUNC1-NEXT:    v_addc_u32_e32 v1, vcc, 0, v4, vcc
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v2, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v4, v2, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v5, v3, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v0, v2, v0
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v4, v1
+; FUNC1-NEXT:    v_add_i32_e32 v1, vcc, v1, v5
+; FUNC1-NEXT:    v_sub_i32_e32 v4, vcc, v7, v1
+; FUNC1-NEXT:    v_sub_i32_e32 v0, vcc, v6, v0
+; FUNC1-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, vcc
+; FUNC1-NEXT:    v_sub_i32_e64 v5, s[0:1], v0, v2
+; FUNC1-NEXT:    v_subbrev_u32_e64 v14, s[2:3], 0, v4, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v14, v3
+; FUNC1-NEXT:    v_cndmask_b32_e64 v15, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v2
+; FUNC1-NEXT:    v_cndmask_b32_e64 v16, 0, -1, s[2:3]
+; FUNC1-NEXT:    v_cmp_eq_u32_e64 s[2:3], v14, v3
+; FUNC1-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v15, v15, v16, s[2:3]
+; FUNC1-NEXT:    v_sub_i32_e64 v16, s[0:1], v5, v2
+; FUNC1-NEXT:    v_subb_u32_e32 v1, vcc, v7, v1, vcc
+; FUNC1-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
+; FUNC1-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v15
+; FUNC1-NEXT:    v_cndmask_b32_e64 v7, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e64 v4, v14, v4, s[0:1]
+; FUNC1-NEXT:    v_cndmask_b32_e64 v14, 0, -1, vcc
+; FUNC1-NEXT:    v_cmp_eq_u32_e32 vcc, v1, v3
+; FUNC1-NEXT:    v_cndmask_b32_e32 v3, v7, v14, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e64 v5, v5, v16, s[0:1]
+; FUNC1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; FUNC1-NEXT:    v_cndmask_b32_e32 v15, v1, v4, vcc
+; FUNC1-NEXT:    v_cndmask_b32_e32 v14, v0, v5, vcc
+; FUNC1-NEXT:    s_cbranch_execnz .LBB6_12
+; FUNC1-NEXT:  .LBB6_11:
+; FUNC1-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, 0, v2
+; FUNC1-NEXT:    v_mov_b32_e32 v15, 0
+; FUNC1-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC1-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC1-NEXT:    v_mul_hi_u32 v1, v0, v1
+; FUNC1-NEXT:    v_add_i32_e32 v0, vcc, v0, v1
+; FUNC1-NEXT:    v_mul_hi_u32 v0, v6, v0
+; FUNC1-NEXT:    v_mul_lo_u32 v0, v0, v2
+; FUNC1-NEXT:    v_sub_i32_e32 v0, vcc, v6, v0
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v0, v2
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
+; FUNC1-NEXT:    v_sub_i32_e32 v1, vcc, v0, v2
+; FUNC1-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC1-NEXT:    v_cndmask_b32_e32 v14, v0, v1, vcc
+; FUNC1-NEXT:  .LBB6_12:
+; FUNC1-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC1-NEXT:    s_mov_b32 s6, -1
+; FUNC1-NEXT:    buffer_store_dwordx4 v[12:15], off, s[4:7], 0 offset:16
+; FUNC1-NEXT:    buffer_store_dwordx4 v[8:11], off, s[4:7], 0
+; FUNC1-NEXT:    s_endpgm
+; FUNC1-NEXT:  .LBB6_13:
+; FUNC1-NEXT:    ; implicit-def: $vgpr8_vgpr9
+; FUNC1-NEXT:    s_branch .LBB6_2
+; FUNC1-NEXT:  .LBB6_14:
+; FUNC1-NEXT:    s_branch .LBB6_5
+; FUNC1-NEXT:  .LBB6_15:
+; FUNC1-NEXT:    ; implicit-def: $vgpr12_vgpr13
+; FUNC1-NEXT:    s_branch .LBB6_8
+; FUNC1-NEXT:  .LBB6_16:
+; FUNC1-NEXT:    s_branch .LBB6_11
+;
+; FUNC2-LABEL: test_urem_v4i64:
+; FUNC2:       ; %bb.0:
+; FUNC2-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; FUNC2-NEXT:    s_mov_b32 s3, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s2, -1
+; FUNC2-NEXT:    v_mov_b32_e32 v8, 0
+; FUNC2-NEXT:    s_waitcnt lgkmcnt(0)
+; FUNC2-NEXT:    s_mov_b32 s0, s6
+; FUNC2-NEXT:    s_mov_b32 s1, s7
+; FUNC2-NEXT:    buffer_load_dwordx4 v[10:13], off, s[0:3], 0 offset:32
+; FUNC2-NEXT:    buffer_load_dwordx4 v[14:17], off, s[0:3], 0
+; FUNC2-NEXT:    buffer_load_dwordx4 v[4:7], off, s[0:3], 0 offset:16
+; FUNC2-NEXT:    buffer_load_dwordx4 v[0:3], off, s[0:3], 0 offset:48
+; FUNC2-NEXT:    s_waitcnt vmcnt(2)
+; FUNC2-NEXT:    v_or_b32_e32 v9, v15, v11
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[8:9]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB6_13
+; FUNC2-NEXT:  ; %bb.1:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v8, v10
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v9, v11
+; FUNC2-NEXT:    v_sub_u32_e32 v22, vcc, 0, v10
+; FUNC2-NEXT:    v_subb_u32_e32 v23, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v8, 0x4f800000, v9
+; FUNC2-NEXT:    v_rcp_f32_e32 v8, v8
+; FUNC2-NEXT:    v_mul_f32_e32 v8, 0x5f7ffffc, v8
+; FUNC2-NEXT:    v_mul_f32_e32 v9, 0x2f800000, v8
+; FUNC2-NEXT:    v_trunc_f32_e32 v9, v9
+; FUNC2-NEXT:    v_mac_f32_e32 v8, 0xcf800000, v9
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v20, v9
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v21, v8
+; FUNC2-NEXT:    v_mul_lo_u32 v18, v22, v20
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v22, v21, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v19, v23, v21
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v18
+; FUNC2-NEXT:    v_add_u32_e32 v25, vcc, v9, v19
+; FUNC2-NEXT:    v_mul_hi_u32 v24, v21, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v21, v25, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v20, v8, 0
+; FUNC2-NEXT:    v_add_u32_e32 v24, vcc, v24, v18
+; FUNC2-NEXT:    v_addc_u32_e32 v26, vcc, 0, v19, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v20, v25, 0
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v24, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v8, vcc, v26, v9, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v19, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v8, v18
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v24, vcc, v21, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v25, vcc, v20, v9, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v22, v24, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v20, v22, v25
+; FUNC2-NEXT:    v_mul_lo_u32 v21, v23, v24
+; FUNC2-NEXT:    v_mul_hi_u32 v22, v24, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v25, v8, 0
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v20
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v21
+; FUNC2-NEXT:    v_mad_u64_u32 v[20:21], s[0:1], v24, v9, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v25, v9, 0
+; FUNC2-NEXT:    v_add_u32_e32 v20, vcc, v22, v20
+; FUNC2-NEXT:    v_addc_u32_e32 v21, vcc, 0, v21, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v20, v18
+; FUNC2-NEXT:    v_addc_u32_e32 v18, vcc, v21, v19, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v18, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v24, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v19, vcc, v25, v9, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v14, v19, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v20, v14, v18
+; FUNC2-NEXT:    v_add_u32_e32 v20, vcc, v20, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v21, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v15, v18, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v15, v19, 0
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v20, v8
+; FUNC2-NEXT:    v_addc_u32_e32 v8, vcc, v21, v9, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v9, vcc, 0, v19, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v8, v18
+; FUNC2-NEXT:    v_addc_u32_e32 v8, vcc, 0, v9, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v19, v10, v8
+; FUNC2-NEXT:    v_mad_u64_u32 v[8:9], s[0:1], v10, v18, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v18, v11, v18
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v19
+; FUNC2-NEXT:    v_add_u32_e32 v9, vcc, v9, v18
+; FUNC2-NEXT:    v_sub_u32_e32 v18, vcc, v15, v9
+; FUNC2-NEXT:    v_sub_u32_e32 v8, vcc, v14, v8
+; FUNC2-NEXT:    v_subb_u32_e64 v18, s[0:1], v18, v11, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v19, s[0:1], v8, v10
+; FUNC2-NEXT:    v_subbrev_u32_e64 v20, s[2:3], 0, v18, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v20, v11
+; FUNC2-NEXT:    v_cndmask_b32_e64 v21, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v19, v10
+; FUNC2-NEXT:    v_cndmask_b32_e64 v22, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v20, v11
+; FUNC2-NEXT:    v_subb_u32_e64 v18, s[0:1], v18, v11, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v21, v21, v22, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v22, s[0:1], v19, v10
+; FUNC2-NEXT:    v_subb_u32_e32 v9, vcc, v15, v9, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v18, s[0:1], 0, v18, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v9, v11
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v21
+; FUNC2-NEXT:    v_cndmask_b32_e64 v15, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v10
+; FUNC2-NEXT:    v_cndmask_b32_e64 v18, v20, v18, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v20, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v9, v11
+; FUNC2-NEXT:    v_cndmask_b32_e32 v11, v15, v20, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v19, v19, v22, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v11
+; FUNC2-NEXT:    v_cndmask_b32_e32 v9, v9, v18, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v8, v8, v19, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB6_3
+; FUNC2-NEXT:  .LBB6_2:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v8, v10
+; FUNC2-NEXT:    v_sub_u32_e32 v9, vcc, 0, v10
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v8, v8
+; FUNC2-NEXT:    v_mul_f32_e32 v8, 0x4f7ffffe, v8
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v8, v8
+; FUNC2-NEXT:    v_mul_lo_u32 v9, v9, v8
+; FUNC2-NEXT:    v_mul_hi_u32 v9, v8, v9
+; FUNC2-NEXT:    v_add_u32_e32 v8, vcc, v8, v9
+; FUNC2-NEXT:    v_mul_hi_u32 v8, v14, v8
+; FUNC2-NEXT:    v_mul_lo_u32 v8, v8, v10
+; FUNC2-NEXT:    v_sub_u32_e32 v8, vcc, v14, v8
+; FUNC2-NEXT:    v_sub_u32_e32 v9, vcc, v8, v10
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v10
+; FUNC2-NEXT:    v_cndmask_b32_e32 v8, v8, v9, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v9, vcc, v8, v10
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v8, v10
+; FUNC2-NEXT:    v_cndmask_b32_e32 v8, v8, v9, vcc
+; FUNC2-NEXT:    v_mov_b32_e32 v9, 0
+; FUNC2-NEXT:  .LBB6_3:
+; FUNC2-NEXT:    v_or_b32_e32 v11, v17, v13
+; FUNC2-NEXT:    v_mov_b32_e32 v10, 0
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[10:11]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB6_14
+; FUNC2-NEXT:  ; %bb.4:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v10, v12
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v11, v13
+; FUNC2-NEXT:    v_sub_u32_e32 v20, vcc, 0, v12
+; FUNC2-NEXT:    v_subb_u32_e32 v21, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v10, 0x4f800000, v11
+; FUNC2-NEXT:    v_rcp_f32_e32 v10, v10
+; FUNC2-NEXT:    v_mul_f32_e32 v10, 0x5f7ffffc, v10
+; FUNC2-NEXT:    v_mul_f32_e32 v11, 0x2f800000, v10
+; FUNC2-NEXT:    v_trunc_f32_e32 v11, v11
+; FUNC2-NEXT:    v_mac_f32_e32 v10, 0xcf800000, v11
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v18, v11
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v19, v10
+; FUNC2-NEXT:    v_mul_lo_u32 v14, v20, v18
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v20, v19, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v15, v21, v19
+; FUNC2-NEXT:    v_add_u32_e32 v11, vcc, v11, v14
+; FUNC2-NEXT:    v_add_u32_e32 v23, vcc, v11, v15
+; FUNC2-NEXT:    v_mul_hi_u32 v22, v19, v10
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v19, v23, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v18, v10, 0
+; FUNC2-NEXT:    v_add_u32_e32 v22, vcc, v22, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v24, vcc, 0, v15, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v18, v23, 0
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v22, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v10, vcc, v24, v11, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v15, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v10, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v22, vcc, v19, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v23, vcc, v18, v11, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v20, v22, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v18, v20, v23
+; FUNC2-NEXT:    v_mul_lo_u32 v19, v21, v22
+; FUNC2-NEXT:    v_mul_hi_u32 v20, v22, v10
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v23, v10, 0
+; FUNC2-NEXT:    v_add_u32_e32 v11, vcc, v11, v18
+; FUNC2-NEXT:    v_add_u32_e32 v11, vcc, v11, v19
+; FUNC2-NEXT:    v_mad_u64_u32 v[18:19], s[0:1], v22, v11, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v23, v11, 0
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v20, v18
+; FUNC2-NEXT:    v_addc_u32_e32 v19, vcc, 0, v19, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v18, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v14, vcc, v19, v15, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v14, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v22, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v15, vcc, v23, v11, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v16, v15, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v18, v16, v14
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v18, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v19, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v17, v14, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v17, v15, 0
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v18, v10
+; FUNC2-NEXT:    v_addc_u32_e32 v10, vcc, v19, v11, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v11, vcc, 0, v15, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v10, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v10, vcc, 0, v11, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v15, v12, v10
+; FUNC2-NEXT:    v_mad_u64_u32 v[10:11], s[0:1], v12, v14, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v14, v13, v14
+; FUNC2-NEXT:    v_add_u32_e32 v11, vcc, v11, v15
+; FUNC2-NEXT:    v_add_u32_e32 v11, vcc, v11, v14
+; FUNC2-NEXT:    v_sub_u32_e32 v14, vcc, v17, v11
+; FUNC2-NEXT:    v_sub_u32_e32 v10, vcc, v16, v10
+; FUNC2-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v13, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v15, s[0:1], v10, v12
+; FUNC2-NEXT:    v_subbrev_u32_e64 v18, s[2:3], 0, v14, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v18, v13
+; FUNC2-NEXT:    v_cndmask_b32_e64 v19, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v15, v12
+; FUNC2-NEXT:    v_cndmask_b32_e64 v20, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v18, v13
+; FUNC2-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v13, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v19, v19, v20, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v20, s[0:1], v15, v12
+; FUNC2-NEXT:    v_subb_u32_e32 v11, vcc, v17, v11, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v14, s[0:1], 0, v14, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v11, v13
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v19
+; FUNC2-NEXT:    v_cndmask_b32_e64 v17, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; FUNC2-NEXT:    v_cndmask_b32_e64 v14, v18, v14, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v18, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v11, v13
+; FUNC2-NEXT:    v_cndmask_b32_e32 v13, v17, v18, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v15, v15, v20, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v13
+; FUNC2-NEXT:    v_cndmask_b32_e32 v11, v11, v14, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v10, v10, v15, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB6_6
+; FUNC2-NEXT:  .LBB6_5:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v10, v12
+; FUNC2-NEXT:    v_sub_u32_e32 v11, vcc, 0, v12
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v10, v10
+; FUNC2-NEXT:    v_mul_f32_e32 v10, 0x4f7ffffe, v10
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v10, v10
+; FUNC2-NEXT:    v_mul_lo_u32 v11, v11, v10
+; FUNC2-NEXT:    v_mul_hi_u32 v11, v10, v11
+; FUNC2-NEXT:    v_add_u32_e32 v10, vcc, v10, v11
+; FUNC2-NEXT:    v_mul_hi_u32 v10, v16, v10
+; FUNC2-NEXT:    v_mul_lo_u32 v10, v10, v12
+; FUNC2-NEXT:    v_sub_u32_e32 v10, vcc, v16, v10
+; FUNC2-NEXT:    v_sub_u32_e32 v11, vcc, v10, v12
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; FUNC2-NEXT:    v_cndmask_b32_e32 v10, v10, v11, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v11, vcc, v10, v12
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v10, v12
+; FUNC2-NEXT:    v_cndmask_b32_e32 v10, v10, v11, vcc
+; FUNC2-NEXT:    v_mov_b32_e32 v11, 0
+; FUNC2-NEXT:  .LBB6_6:
+; FUNC2-NEXT:    s_waitcnt vmcnt(0)
+; FUNC2-NEXT:    v_or_b32_e32 v13, v5, v1
+; FUNC2-NEXT:    v_mov_b32_e32 v12, 0
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[12:13]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB6_15
+; FUNC2-NEXT:  ; %bb.7:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v12, v0
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v13, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v18, vcc, 0, v0
+; FUNC2-NEXT:    v_subb_u32_e32 v19, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v12, 0x4f800000, v13
+; FUNC2-NEXT:    v_rcp_f32_e32 v12, v12
+; FUNC2-NEXT:    v_mul_f32_e32 v12, 0x5f7ffffc, v12
+; FUNC2-NEXT:    v_mul_f32_e32 v13, 0x2f800000, v12
+; FUNC2-NEXT:    v_trunc_f32_e32 v13, v13
+; FUNC2-NEXT:    v_mac_f32_e32 v12, 0xcf800000, v13
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v16, v13
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v17, v12
+; FUNC2-NEXT:    v_mul_lo_u32 v14, v18, v16
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v18, v17, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v15, v19, v17
+; FUNC2-NEXT:    v_add_u32_e32 v13, vcc, v13, v14
+; FUNC2-NEXT:    v_add_u32_e32 v15, vcc, v13, v15
+; FUNC2-NEXT:    v_mul_hi_u32 v20, v17, v12
+; FUNC2-NEXT:    v_mad_u64_u32 v[13:14], s[0:1], v17, v15, 0
+; FUNC2-NEXT:    v_add_u32_e32 v20, vcc, v20, v13
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v16, v12, 0
+; FUNC2-NEXT:    v_addc_u32_e32 v21, vcc, 0, v14, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v16, v15, 0
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v20, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v12, vcc, v21, v13, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v15, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v12, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v20, vcc, v17, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v21, vcc, v16, v13, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v18, v20, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v16, v18, v21
+; FUNC2-NEXT:    v_mul_lo_u32 v17, v19, v20
+; FUNC2-NEXT:    v_mul_hi_u32 v18, v20, v12
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v21, v12, 0
+; FUNC2-NEXT:    v_add_u32_e32 v13, vcc, v13, v16
+; FUNC2-NEXT:    v_add_u32_e32 v13, vcc, v13, v17
+; FUNC2-NEXT:    v_mad_u64_u32 v[16:17], s[0:1], v20, v13, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v21, v13, 0
+; FUNC2-NEXT:    v_add_u32_e32 v16, vcc, v18, v16
+; FUNC2-NEXT:    v_addc_u32_e32 v17, vcc, 0, v17, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v16, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v14, vcc, v17, v15, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v14, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v20, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v15, vcc, v21, v13, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v4, v15, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v16, v4, v14
+; FUNC2-NEXT:    v_add_u32_e32 v16, vcc, v16, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v17, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v5, v14, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v5, v15, 0
+; FUNC2-NEXT:    v_add_u32_e32 v12, vcc, v16, v12
+; FUNC2-NEXT:    v_addc_u32_e32 v12, vcc, v17, v13, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v13, vcc, 0, v15, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v12, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v12, vcc, 0, v13, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v15, v0, v12
+; FUNC2-NEXT:    v_mad_u64_u32 v[12:13], s[0:1], v0, v14, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v14, v1, v14
+; FUNC2-NEXT:    v_add_u32_e32 v13, vcc, v13, v15
+; FUNC2-NEXT:    v_add_u32_e32 v13, vcc, v13, v14
+; FUNC2-NEXT:    v_sub_u32_e32 v14, vcc, v5, v13
+; FUNC2-NEXT:    v_sub_u32_e32 v12, vcc, v4, v12
+; FUNC2-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v1, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v15, s[0:1], v12, v0
+; FUNC2-NEXT:    v_subbrev_u32_e64 v16, s[2:3], 0, v14, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v16, v1
+; FUNC2-NEXT:    v_cndmask_b32_e64 v17, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v15, v0
+; FUNC2-NEXT:    v_cndmask_b32_e64 v18, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v16, v1
+; FUNC2-NEXT:    v_subb_u32_e64 v14, s[0:1], v14, v1, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v17, v17, v18, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v18, s[0:1], v15, v0
+; FUNC2-NEXT:    v_subb_u32_e32 v5, vcc, v5, v13, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v14, s[0:1], 0, v14, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v5, v1
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v17
+; FUNC2-NEXT:    v_cndmask_b32_e64 v13, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v12, v0
+; FUNC2-NEXT:    v_cndmask_b32_e64 v14, v16, v14, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v16, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v5, v1
+; FUNC2-NEXT:    v_cndmask_b32_e32 v1, v13, v16, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v15, v15, v18, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v1
+; FUNC2-NEXT:    v_cndmask_b32_e32 v13, v5, v14, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v12, v12, v15, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB6_9
+; FUNC2-NEXT:  .LBB6_8:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, v0
+; FUNC2-NEXT:    v_sub_u32_e32 v5, vcc, 0, v0
+; FUNC2-NEXT:    v_mov_b32_e32 v13, 0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v5, v5, v1
+; FUNC2-NEXT:    v_mul_hi_u32 v5, v1, v5
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
+; FUNC2-NEXT:    v_mul_hi_u32 v1, v4, v1
+; FUNC2-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v4, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v4, vcc, v1, v0
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC2-NEXT:    v_cndmask_b32_e32 v1, v1, v4, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v4, vcc, v1, v0
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v0
+; FUNC2-NEXT:    v_cndmask_b32_e32 v12, v1, v4, vcc
+; FUNC2-NEXT:  .LBB6_9:
+; FUNC2-NEXT:    v_or_b32_e32 v1, v7, v3
+; FUNC2-NEXT:    v_mov_b32_e32 v0, 0
+; FUNC2-NEXT:    v_cmp_ne_u64_e32 vcc, 0, v[0:1]
+; FUNC2-NEXT:    s_cbranch_vccz .LBB6_16
+; FUNC2-NEXT:  ; %bb.10:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v1, v3
+; FUNC2-NEXT:    v_sub_u32_e32 v16, vcc, 0, v2
+; FUNC2-NEXT:    v_subb_u32_e32 v17, vcc, 0, v3, vcc
+; FUNC2-NEXT:    v_mac_f32_e32 v0, 0x4f800000, v1
+; FUNC2-NEXT:    v_rcp_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v0, 0x5f7ffffc, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v1, 0x2f800000, v0
+; FUNC2-NEXT:    v_trunc_f32_e32 v1, v1
+; FUNC2-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v14, v1
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v15, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v4, v16, v14
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v16, v15, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v5, v17, v15
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v4
+; FUNC2-NEXT:    v_add_u32_e32 v19, vcc, v1, v5
+; FUNC2-NEXT:    v_mul_hi_u32 v18, v15, v0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v15, v19, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v14, v0, 0
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v18, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v20, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v14, v19, 0
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v18, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v0, vcc, v20, v1, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v0, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v18, vcc, v15, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v19, vcc, v14, v1, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v16, v18, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v14, v16, v19
+; FUNC2-NEXT:    v_mul_lo_u32 v15, v17, v18
+; FUNC2-NEXT:    v_mul_hi_u32 v16, v18, v0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v19, v0, 0
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v14
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v15
+; FUNC2-NEXT:    v_mad_u64_u32 v[14:15], s[0:1], v18, v1, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v19, v1, 0
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v16, v14
+; FUNC2-NEXT:    v_addc_u32_e32 v15, vcc, 0, v15, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v14, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v4, vcc, v15, v5, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v4, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v18, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v5, vcc, v19, v1, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v6, v5, 0
+; FUNC2-NEXT:    v_mul_hi_u32 v14, v6, v4
+; FUNC2-NEXT:    v_add_u32_e32 v14, vcc, v14, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v15, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v7, v4, 0
+; FUNC2-NEXT:    v_mad_u64_u32 v[4:5], s[0:1], v7, v5, 0
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v14, v0
+; FUNC2-NEXT:    v_addc_u32_e32 v0, vcc, v15, v1, vcc
+; FUNC2-NEXT:    v_addc_u32_e32 v1, vcc, 0, v5, vcc
+; FUNC2-NEXT:    v_add_u32_e32 v4, vcc, v0, v4
+; FUNC2-NEXT:    v_addc_u32_e32 v0, vcc, 0, v1, vcc
+; FUNC2-NEXT:    v_mul_lo_u32 v5, v2, v0
+; FUNC2-NEXT:    v_mad_u64_u32 v[0:1], s[0:1], v2, v4, 0
+; FUNC2-NEXT:    v_mul_lo_u32 v4, v3, v4
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v5
+; FUNC2-NEXT:    v_add_u32_e32 v1, vcc, v1, v4
+; FUNC2-NEXT:    v_sub_u32_e32 v4, vcc, v7, v1
+; FUNC2-NEXT:    v_sub_u32_e32 v0, vcc, v6, v0
+; FUNC2-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, vcc
+; FUNC2-NEXT:    v_sub_u32_e64 v5, s[0:1], v0, v2
+; FUNC2-NEXT:    v_subbrev_u32_e64 v14, s[2:3], 0, v4, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v14, v3
+; FUNC2-NEXT:    v_cndmask_b32_e64 v15, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_ge_u32_e64 s[2:3], v5, v2
+; FUNC2-NEXT:    v_cndmask_b32_e64 v16, 0, -1, s[2:3]
+; FUNC2-NEXT:    v_cmp_eq_u32_e64 s[2:3], v14, v3
+; FUNC2-NEXT:    v_subb_u32_e64 v4, s[0:1], v4, v3, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v15, v15, v16, s[2:3]
+; FUNC2-NEXT:    v_sub_u32_e64 v16, s[0:1], v5, v2
+; FUNC2-NEXT:    v_subb_u32_e32 v1, vcc, v7, v1, vcc
+; FUNC2-NEXT:    v_subbrev_u32_e64 v4, s[0:1], 0, v4, s[0:1]
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v1, v3
+; FUNC2-NEXT:    v_cmp_ne_u32_e64 s[0:1], 0, v15
+; FUNC2-NEXT:    v_cndmask_b32_e64 v7, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e64 v4, v14, v4, s[0:1]
+; FUNC2-NEXT:    v_cndmask_b32_e64 v14, 0, -1, vcc
+; FUNC2-NEXT:    v_cmp_eq_u32_e32 vcc, v1, v3
+; FUNC2-NEXT:    v_cndmask_b32_e32 v3, v7, v14, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e64 v5, v5, v16, s[0:1]
+; FUNC2-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v3
+; FUNC2-NEXT:    v_cndmask_b32_e32 v15, v1, v4, vcc
+; FUNC2-NEXT:    v_cndmask_b32_e32 v14, v0, v5, vcc
+; FUNC2-NEXT:    s_cbranch_execnz .LBB6_12
+; FUNC2-NEXT:  .LBB6_11:
+; FUNC2-NEXT:    v_cvt_f32_u32_e32 v0, v2
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, 0, v2
+; FUNC2-NEXT:    v_mov_b32_e32 v15, 0
+; FUNC2-NEXT:    v_rcp_iflag_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
+; FUNC2-NEXT:    v_cvt_u32_f32_e32 v0, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v1, v1, v0
+; FUNC2-NEXT:    v_mul_hi_u32 v1, v0, v1
+; FUNC2-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
+; FUNC2-NEXT:    v_mul_hi_u32 v0, v6, v0
+; FUNC2-NEXT:    v_mul_lo_u32 v0, v0, v2
+; FUNC2-NEXT:    v_sub_u32_e32 v0, vcc, v6, v0
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v0, v2
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e32 v0, v0, v1, vcc
+; FUNC2-NEXT:    v_sub_u32_e32 v1, vcc, v0, v2
+; FUNC2-NEXT:    v_cmp_ge_u32_e32 vcc, v0, v2
+; FUNC2-NEXT:    v_cndmask_b32_e32 v14, v0, v1, vcc
+; FUNC2-NEXT:  .LBB6_12:
+; FUNC2-NEXT:    s_mov_b32 s7, 0xf000
+; FUNC2-NEXT:    s_mov_b32 s6, -1
+; FUNC2-NEXT:    buffer_store_dwordx4 v[12:15], off, s[4:7], 0 offset:16
+; FUNC2-NEXT:    buffer_store_dwordx4 v[8:11], off, s[4:7], 0
+; FUNC2-NEXT:    s_endpgm
+; FUNC2-NEXT:  .LBB6_13:
+; FUNC2-NEXT:    ; implicit-def: $vgpr8_vgpr9
+; FUNC2-NEXT:    s_branch .LBB6_2
+; FUNC2-NEXT:  .LBB6_14:
+; FUNC2-NEXT:    s_branch .LBB6_5
+; FUNC2-NEXT:  .LBB6_15:
+; FUNC2-NEXT:    ; implicit-def: $vgpr12_vgpr13
+; FUNC2-NEXT:    s_branch .LBB6_8
+; FUNC2-NEXT:  .LBB6_16:
+; FUNC2-NEXT:    s_branch .LBB6_11
+;
+; EG-LABEL: test_urem_v4i64:
+; EG:       ; %bb.0:
+; EG-NEXT:    ALU 0, @34, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    TEX 1 @26
+; EG-NEXT:    ALU 114, @35, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @150, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @266, KC0[], KC1[]
+; EG-NEXT:    ALU 115, @382, KC0[], KC1[]
+; EG-NEXT:    ALU 22, @498, KC0[], KC1[]
+; EG-NEXT:    TEX 1 @30
+; EG-NEXT:    ALU 112, @521, KC0[], KC1[]
+; EG-NEXT:    ALU 114, @634, KC0[], KC1[]
+; EG-NEXT:    ALU 114, @749, KC0[], KC1[]
+; EG-NEXT:    ALU 113, @864, KC0[], KC1[]
+; EG-NEXT:    ALU 114, @978, KC0[], KC1[]
+; EG-NEXT:    ALU 113, @1093, KC0[], KC1[]
+; EG-NEXT:    ALU 114, @1207, KC0[], KC1[]
+; EG-NEXT:    ALU 113, @1322, KC0[], KC1[]
+; EG-NEXT:    ALU 114, @1436, KC0[], KC1[]
+; EG-NEXT:    ALU 113, @1551, KC0[], KC1[]
+; EG-NEXT:    ALU 114, @1665, KC0[], KC1[]
+; EG-NEXT:    ALU 113, @1780, KC0[], KC1[]
+; EG-NEXT:    ALU 112, @1894, KC0[], KC1[]
+; EG-NEXT:    ALU 101, @2007, KC0[CB0:0-32], KC1[]
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T2.XYZW, T0.X, 0
+; EG-NEXT:    MEM_RAT_CACHELESS STORE_RAW T6.XYZW, T1.X, 1
+; EG-NEXT:    CF_END
+; EG-NEXT:    PAD
+; EG-NEXT:    Fetch clause starting at 26:
+; EG-NEXT:     VTX_READ_128 T0.XYZW, T2.X, 32, #1
+; EG-NEXT:     VTX_READ_128 T1.XYZW, T2.X, 0, #1
+; EG-NEXT:    Fetch clause starting at 30:
+; EG-NEXT:     VTX_READ_128 T4.XYZW, T2.X, 16, #1
+; EG-NEXT:     VTX_READ_128 T2.XYZW, T2.X, 48, #1
+; EG-NEXT:    ALU clause starting at 34:
+; EG-NEXT:     MOV * T2.X, KC0[2].Z,
+; EG-NEXT:    ALU clause starting at 35:
+; EG-NEXT:     SUB_INT T2.W, 0.0, T0.X,
+; EG-NEXT:     RECIP_UINT * T2.Y, T0.X,
+; EG-NEXT:     MULLO_INT * T2.Z, PV.W, PS,
+; EG-NEXT:     MULHI * T2.Z, T2.Y, PS,
+; EG-NEXT:     ADD_INT * T2.W, T2.Y, PS,
+; EG-NEXT:     MULHI * T2.Y, T1.Y, PV.W,
+; EG-NEXT:     MULLO_INT * T2.Y, PS, T0.X,
+; EG-NEXT:     SUB_INT * T2.W, T1.Y, PS,
+; EG-NEXT:     SETGE_UINT T3.W, PV.W, T0.X,
+; EG-NEXT:     SUB_INT * T4.W, PV.W, T0.X,
+; EG-NEXT:     CNDE_INT * T2.W, PV.W, T2.W, PS,
+; EG-NEXT:     SETGE_UINT T3.W, PV.W, T0.X,
+; EG-NEXT:     SUB_INT * T4.W, PV.W, T0.X,
+; EG-NEXT:     CNDE_INT * T2.W, PV.W, T2.W, PS,
+; EG-NEXT:     CNDE_INT * T2.W, T0.Y, PV.W, T1.Y,
+; EG-NEXT:     BIT_ALIGN_INT T3.W, PV.W, T1.X, literal.x,
+; EG-NEXT:     LSHR * T2.W, PV.W, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T2.Z, PS, T0.Y,
+; EG-NEXT:     SETGE_UINT T4.W, PS, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.X,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUB_INT T4.W, T2.W, T0.Y, BS:VEC_120/SCL_212
+; EG-NEXT:     SUBB_UINT * T5.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT T4.W, PV.W, PS,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    30(4.203895e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    29(4.063766e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    26(3.643376e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT * T2.Z, PS, T0.X,
+; EG-NEXT:    ALU clause starting at 150:
+; EG-NEXT:     SETE_INT T4.W, T2.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, T2.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    21(2.942727e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    19(2.662467e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT * T2.Z, T3.W, T0.X,
+; EG-NEXT:    ALU clause starting at 266:
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, T1.Y, T3.W, T2.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    14(1.961818e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    11(1.541428e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:    ALU clause starting at 382:
+; EG-NEXT:     LSHL T2.Z, T3.W, 1, BS:VEC_201
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, T4.W,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    9(1.261169e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T2.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T3.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 498:
+; EG-NEXT:     BIT_ALIGN_INT T2.W, T2.W, T3.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T3.W, T2.Z, T5.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T2.Z, PS, T0.X,
+; EG-NEXT:     SETE_INT T4.W, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.W, T0.Y,
+; EG-NEXT:     CNDE_INT T1.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T2.Z, T3.W, T0.X,
+; EG-NEXT:     SUBB_UINT T4.W, T3.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T2.W, T0.Y,
+; EG-NEXT:     SUB_INT T4.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T3.W, PV.Y, T3.W, PV.Z,
+; EG-NEXT:     LSHL T2.Z, PS, 1,
+; EG-NEXT:     BFE_UINT T5.W, T1.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, PV.W,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     BIT_ALIGN_INT T3.W, PS, T3.W, literal.x,
+; EG-NEXT:     OR_INT * T5.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T1.Y, PS, T0.X,
+; EG-NEXT:     SETE_INT T3.Z, PV.W, T0.Y,
+; EG-NEXT:     SUB_INT T6.W, 0.0, T0.Z,
+; EG-NEXT:     RECIP_UINT * T3.X, T0.Z,
+; EG-NEXT:    ALU clause starting at 521:
+; EG-NEXT:     SETGE_UINT T7.W, T3.W, T0.Y,
+; EG-NEXT:     MULLO_INT * T3.Y, T6.W, T3.X,
+; EG-NEXT:     CNDE_INT T5.X, T3.Z, PV.W, T1.Y,
+; EG-NEXT:     SUB_INT T1.Y, T5.W, T0.X,
+; EG-NEXT:     SUBB_UINT T3.Z, T5.W, T0.X,
+; EG-NEXT:     SUB_INT T6.W, T3.W, T0.Y, BS:VEC_102/SCL_221
+; EG-NEXT:     MULHI * T3.Y, T3.X, PS,
+; EG-NEXT:     ADD_INT T3.Y, T3.X, PS,
+; EG-NEXT:     SUB_INT T3.Z, PV.W, PV.Z,
+; EG-NEXT:     CNDE_INT T5.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     RECIP_UINT * T1.Y, T2.Z,
+; EG-NEXT:     LSHL T5.Y, PV.W, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T1.X, 1, 1,
+; EG-NEXT:     CNDE_INT T3.W, T5.X, T3.W, PV.Z, BS:VEC_102/SCL_221
+; EG-NEXT:     MULHI * T3.X, T1.W, PV.Y,
+; EG-NEXT:     SUB_INT T3.Y, 0.0, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     MULLO_INT * T3.X, PS, T0.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T3.X, T1.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Y, PV.W, T0.X,
+; EG-NEXT:     SETE_INT T5.Z, PV.Z, T0.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.W, PV.Z, T0.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     MULLO_INT * T3.Y, PV.Y, T1.Y,
+; EG-NEXT:     CNDE_INT T5.X, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T5.Y, T5.W, T0.X,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.X, T0.Z,
+; EG-NEXT:     SUB_INT T3.W, PV.X, T0.Z,
+; EG-NEXT:     MULHI * T3.Y, T1.Y, PS,
+; EG-NEXT:     ADD_INT T6.Z, T1.Y, PS,
+; EG-NEXT:     CNDE_INT T3.W, PV.Z, T3.X, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T3.X, PS, 1,
+; EG-NEXT:     AND_INT T1.Y, T1.X, 1,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T0.Z,
+; EG-NEXT:     SUB_INT T7.W, PV.W, T0.Z,
+; EG-NEXT:     MULHI * T1.X, T4.W, PV.Z,
+; EG-NEXT:     CNDE_INT T5.Z, PV.Z, T3.W, PV.W,
+; EG-NEXT:     OR_INT T3.W, PV.X, PV.Y,
+; EG-NEXT:     MULLO_INT * T1.Y, PS, T2.Z,
+; EG-NEXT:     SETGE_UINT T1.X, PV.W, T0.X,
+; EG-NEXT:     SUB_INT T3.Y, 0.0, T2.X, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T5.Z, T0.W, PV.Z, T1.W,
+; EG-NEXT:     SUB_INT T1.W, T4.W, PS, BS:VEC_120/SCL_212
+; EG-NEXT:     RECIP_UINT * T1.Y, T2.X,
+; EG-NEXT:     SETGE_UINT T3.X, PV.W, T2.Z,
+; EG-NEXT:     SUB_INT T5.Y, PV.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.Z, T1.Z, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHR T7.W, PV.Z, literal.x,
+; EG-NEXT:     MULLO_INT * T3.Y, PV.Y, PS,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T6.X, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T6.Y, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.Z, T0.Z,
+; EG-NEXT:     CNDE_INT T8.W, PV.X, T1.W, PV.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     MULHI * T1.W, T1.Y, PS,
+; EG-NEXT:     SETGE_UINT T7.X, PV.W, T2.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.X, PV.Y, PV.Z,
+; EG-NEXT:     ADD_INT T5.Z, T1.Y, PS,
+; EG-NEXT:     SUBB_UINT T1.W, T5.W, T0.X,
+; EG-NEXT:     SUB_INT * T5.W, T3.Z, T0.Y,
+; EG-NEXT:     SUB_INT T3.X, T6.Z, T0.Z,
+; EG-NEXT:     SUB_INT T1.Y, T7.W, T0.W,
+; EG-NEXT:     SUBB_UINT T7.Z, T6.Z, T0.Z,
+; EG-NEXT:     SUB_INT T1.W, PS, PV.W,
+; EG-NEXT:     MULHI * T5.Y, T4.Y, PV.Z,
+; EG-NEXT:     SUB_INT T6.X, T8.W, T2.Z,
+; EG-NEXT:     CNDE_INT T6.Y, T5.X, T3.Z, PV.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T3.Z, PV.Y, PV.Z,
+; EG-NEXT:     CNDE_INT T5.W, T3.Y, T6.Z, PV.X, BS:VEC_102/SCL_221
+; EG-NEXT:     MULLO_INT * T1.Y, PS, T2.X,
+; EG-NEXT:     LSHL T5.X, PV.W, 1,
+; EG-NEXT:     BFE_UINT T5.Y, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Z, T3.Y, T7.W, PV.Z,
+; EG-NEXT:     BIT_ALIGN_INT T1.W, PV.Y, T6.W, literal.y, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT * T6.W, T4.Y, PS,
+; EG-NEXT:    30(4.203895e-44), 31(4.344025e-44)
+; EG-NEXT:     SETE_INT T3.X, PV.W, T0.Y,
+; EG-NEXT:     SETGE_UINT T1.Y, PS, T2.X,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.Z, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, PV.X, PV.Y,
+; EG-NEXT:     CNDE_INT * T7.W, T7.X, T8.W, T6.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T5.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T3.Y, T2.W, PS, T4.W, BS:VEC_102/SCL_221
+; EG-NEXT:     SETGE_UINT * T5.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T7.W, T3.Z, T0.W,
+; EG-NEXT:     CNDE_INT T6.X, PV.W, PS, T5.Z,
+; EG-NEXT:     SUB_INT T5.Y, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, T3.Y, T4.Z, literal.x, BS:VEC_102/SCL_221
+; EG-NEXT:     LSHR T4.W, T3.Y, literal.x,
+; EG-NEXT:     CNDE_INT * T6.W, T1.Y, T6.W, T5.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T7.X, PS, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, PS, T2.X,
+; EG-NEXT:     SETE_INT T6.Z, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT T7.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T2.Z,
+; EG-NEXT:     SETGE_UINT T5.X, T1.W, T0.Y,
+; EG-NEXT:     CNDE_INT T3.Y, PV.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T6.Z, T5.Z, T2.Z,
+; EG-NEXT:     CNDE_INT T6.W, PV.X, T6.W, PV.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, T6.X, T5.W, T5.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     SUB_INT T1.Y, T4.W, T2.W,
+; EG-NEXT:     SUBB_UINT T7.Z, T5.Z, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T6.W, T2.Y, PV.W, T4.Y,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T5.Z, PV.Z,
+; EG-NEXT:     BFE_UINT T8.X, T1.Z, literal.x, 1,
+; EG-NEXT:     LSHL * T4.Y, PS, 1,
+; EG-NEXT:    29(4.063766e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 634:
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, T6.W, T4.X, literal.x,
+; EG-NEXT:     LSHR T6.W, T6.W, literal.x,
+; EG-NEXT:     SUB_INT * T9.W, T1.Y, T7.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS,
+; EG-NEXT:     SETE_INT T6.Z, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT T4.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.X,
+; EG-NEXT:    30(4.203895e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T10.X, T5.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.Z, PV.W, PS,
+; EG-NEXT:     SUB_INT T6.Z, T5.Z, T2.X,
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T5.W, T4.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T9.X, T6.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T1.Y, T5.Z, T2.X,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T11.X, T5.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T5.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T8.W, PV.X, PV.Y,
+; EG-NEXT:     CNDE_INT * T9.W, T3.Y, T5.Z, T6.Z,
+; EG-NEXT:     SUB_INT T9.X, T4.W, T2.W,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T6.W, T3.Y, T6.W, PV.W, BS:VEC_021/SCL_122
+; EG-NEXT:    30(4.203895e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T7.Z,
+; EG-NEXT:     SUB_INT T12.X, T3.Z, T0.W,
+; EG-NEXT:     LSHL T3.Y, PV.W, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, T6.W, T9.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T6.W, T1.Y, T5.Z,
+; EG-NEXT:     SUB_INT * T8.W, T9.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T4.Y, T4.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T2.Y,
+; EG-NEXT:    29(4.063766e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T4.Y, T6.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.Y, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, T3.Y, PV.X,
+; EG-NEXT:     SUB_INT * T5.W, T12.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T6.X, T3.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T5.W, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T6.W, PV.Y,
+; EG-NEXT:     LSHL T9.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T3.Z, T4.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.X, T7.W, literal.x,
+; EG-NEXT:     OR_INT * T7.W, T7.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T9.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T0.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T8.X, T4.X, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    29(4.063766e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T5.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.W,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Z, T4.W, T2.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.Z,
+; EG-NEXT:     SUB_INT T1.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T7.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T0.Z,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T6.Z, T2.Y,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT * T5.Z, PV.W, T4.W, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 749:
+; EG-NEXT:     OR_INT T4.W, T1.Y, T3.Z,
+; EG-NEXT:     SUB_INT * T7.W, T6.X, T10.X,
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T5.W, T5.Z, T2.W, BS:VEC_201
+; EG-NEXT:     SETGE_UINT * T7.W, T5.Z, T2.W,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T6.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T6.Z, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T6.Z, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T8.W, T9.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T2.X,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.Y,
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    26(3.643376e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    26(3.643376e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T0.W,
+; EG-NEXT:    28(3.923636e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T6.Z, T5.W, T0.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T0.Z,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.X,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T2.Y,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T5.Z, T2.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, PV.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T0.W,
+; EG-NEXT:    27(3.783506e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T5.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.Z, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T5.Z, T4.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T3.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T2.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T5.Z,
+; EG-NEXT:     SUB_INT * T6.X, T6.W, T2.W,
+; EG-NEXT:    ALU clause starting at 864:
+; EG-NEXT:     LSHL T3.Y, T4.W, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T6.Z, T9.W,
+; EG-NEXT:     CNDE_INT * T7.W, T4.Y, T7.W, T7.Z,
+; EG-NEXT:    26(3.643376e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T1.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, T6.X, T7.X,
+; EG-NEXT:    24(3.363116e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.Y,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T5.Z, T4.W, T2.X, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T7.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T2.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.Z,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    25(3.503246e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T3.Z, T0.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T4.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T2.Y,
+; EG-NEXT:    24(3.363116e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.Y, T6.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T3.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T3.Z, T5.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T9.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T0.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.X, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    24(3.363116e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.W,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Z, T5.W, T2.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.Z,
+; EG-NEXT:     SUB_INT T1.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T8.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T0.Z,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T0.W,
+; EG-NEXT:     LSHL * T1.Y, PS, 1,
+; EG-NEXT:    ALU clause starting at 978:
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T5.Z, T8.W, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT * T6.W, T3.Y, T6.W, T7.Z,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T6.Z, T2.Y,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T5.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, T1.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, T6.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T2.W,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T6.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T6.Z, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T6.Z, T4.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T7.W, literal.x,
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T2.X,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.Y,
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    21(2.942727e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    21(2.942727e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T0.W,
+; EG-NEXT:    23(3.222986e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T6.Z, T4.W, T0.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T0.Z,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T7.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.X,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T2.Y,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T5.Z, T2.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T4.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T0.W,
+; EG-NEXT:    22(3.082857e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T5.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.Z, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T5.Z, T5.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.X,
+; EG-NEXT:     SUB_INT T3.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T7.Z, PS, T2.Z,
+; EG-NEXT:    ALU clause starting at 1093:
+; EG-NEXT:     SETE_INT T9.W, T6.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T10.W, T6.W, T2.W,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, T7.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T2.Z,
+; EG-NEXT:     SUB_INT T9.W, T3.Y, T6.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    21(2.942727e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T1.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    19(2.662467e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.Y,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T5.Z, T5.W, T2.X, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T2.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.Z,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    20(2.802597e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T3.Z, T0.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, PV.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T2.Y,
+; EG-NEXT:    19(2.662467e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.Y, T6.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T3.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T3.Z, T4.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T7.W, literal.x,
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T9.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T0.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.X, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    19(2.662467e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.W,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Z, T4.W, T2.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.Z,
+; EG-NEXT:     SUB_INT T1.Y, T5.Z, T2.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT * T7.W, PV.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:    ALU clause starting at 1207:
+; EG-NEXT:     SETGE_UINT * T9.W, T5.W, T0.W,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T3.Y, T7.W, PV.W, T7.Z, BS:VEC_201
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T7.W, T1.Y, T6.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T6.Z, T2.Y,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T4.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T2.W,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T6.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T6.Z, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T6.Z, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T2.X,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.Y,
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    16(2.242078e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T0.W,
+; EG-NEXT:    18(2.522337e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T6.Z, T5.W, T0.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T0.Z,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.X,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T2.Y,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T5.Z, T2.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, PV.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T0.W,
+; EG-NEXT:    17(2.382207e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T5.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.Z, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL * T8.X, PS, 1,
+; EG-NEXT:    ALU clause starting at 1322:
+; EG-NEXT:     CNDE_INT T1.Y, T5.Z, T6.W, T1.Y,
+; EG-NEXT:     SUB_INT T5.Z, T4.W, T2.X,
+; EG-NEXT:     BIT_ALIGN_INT * T6.W, T6.X, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T3.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T9.W, T6.W, T2.W, BS:VEC_210
+; EG-NEXT:     SETGE_UINT * T10.W, T6.W, T2.W,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T2.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T1.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    14(1.961818e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.Y,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T5.Z, T4.W, T2.X, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T7.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T2.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.Z,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    15(2.101948e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T3.Z, T0.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T4.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T2.Y,
+; EG-NEXT:    14(1.961818e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.Y, T6.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T3.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T3.Z, T5.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T9.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T0.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.X, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    14(1.961818e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.W,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T2.X,
+; EG-NEXT:     CNDE_INT * T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:    ALU clause starting at 1436:
+; EG-NEXT:     SUB_INT T3.Z, T5.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T4.W, T3.Y, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T6.W, T1.Y, T6.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.Z,
+; EG-NEXT:     SUB_INT T1.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T8.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T0.Z,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T6.Z, T2.Y,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T5.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, PV.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T2.W,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T6.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T6.Z, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T6.Z, T4.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T7.W, literal.x,
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T2.X,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.Y,
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    11(1.541428e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    11(1.541428e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T0.W,
+; EG-NEXT:    13(1.821688e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T6.Z, T4.W, T0.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T0.Z,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T7.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.X,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T2.Y,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T5.Z, T2.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T4.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T0.W,
+; EG-NEXT:    12(1.681558e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT * T6.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 1551:
+; EG-NEXT:     OR_INT T5.W, T4.Y, T6.X, BS:VEC_102/SCL_221
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T5.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.Z, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T6.W, T6.Z, T2.Y,
+; EG-NEXT:     CNDE_INT * T7.W, T10.X, T4.W, T3.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T5.Z, T5.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.X,
+; EG-NEXT:     SUB_INT T3.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T2.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    11(1.541428e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T1.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    9(1.261169e-44), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.Y,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T5.Z, T5.W, T2.X, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T2.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.Z,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    10(1.401298e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T3.Z, T0.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, PV.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T4.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T8.W, PV.Z, T2.Y,
+; EG-NEXT:    9(1.261169e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.Y, T6.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T3.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T3.Z, T4.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T7.W, literal.x,
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T9.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T0.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.X, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    9(1.261169e-44), 31(4.344025e-44)
+; EG-NEXT:    ALU clause starting at 1665:
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, T9.W,
+; EG-NEXT:     SETGE_UINT * T3.Z, T4.W, T2.Z,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T6.W, T5.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, T5.Z, T2.W,
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T2.X, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, T3.Z,
+; EG-NEXT:     SUB_INT T3.Z, T4.W, T2.Z, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.W, T3.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, T6.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.Z,
+; EG-NEXT:     SUB_INT T1.Y, T5.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T7.W, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T0.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T0.Z,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T3.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Z, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T5.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T6.Z, T2.Y,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.W, T4.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.Z,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T2.W,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T2.W,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T6.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T6.Z, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T6.Z, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.x,
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T0.Z,
+; EG-NEXT:     SUB_INT T3.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T8.X, T8.W, T2.X,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T8.W, T2.X,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T1.Y, T5.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.Y,
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T8.W, PV.Y, T8.W, PV.Z,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T4.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.y,
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T8.X,
+; EG-NEXT:    6(8.407791e-45), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.Z, PV.W, T0.Z,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T0.W,
+; EG-NEXT:    8(1.121039e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T4.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T6.Z, T5.W, T0.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Y, T8.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T5.W, T0.Z,
+; EG-NEXT:     SUB_INT T1.Y, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T11.X, T6.W, T2.X,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.X,
+; EG-NEXT:     SUB_INT T8.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, T6.Z,
+; EG-NEXT:     SUB_INT T6.X, T4.W, T2.Y,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T6.Z, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T8.W, T4.Y, T3.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T5.Z, T2.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.W, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T8.W, PV.X, T11.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT * T6.X, T4.X, literal.x, 1,
+; EG-NEXT:    7(9.809089e-45), 0(0.000000e+00)
+; EG-NEXT:    ALU clause starting at 1780:
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T4.W, T8.W,
+; EG-NEXT:     SETGE_UINT * T6.Z, T5.W, T0.Z,
+; EG-NEXT:     SETE_INT T4.W, T3.Z, T0.W,
+; EG-NEXT:     SETGE_UINT * T8.W, T3.Z, T0.W,
+; EG-NEXT:     CNDE_INT T11.X, PV.W, PS, T6.Z,
+; EG-NEXT:     SUB_INT T3.Y, T5.W, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, T1.Y, T6.W, literal.x,
+; EG-NEXT:     OR_INT T4.W, T4.Y, T6.X, BS:VEC_102/SCL_221
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T8.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T10.X, T5.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.Z, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     CNDE_INT * T8.W, PV.X, T5.W, PV.Y,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     SUB_INT T5.Z, T4.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T7.W, T7.X, T9.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T3.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T9.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T10.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T7.X, T7.W, T2.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T7.W, T2.Z,
+; EG-NEXT:     SUB_INT T9.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T1.Y, T4.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T6.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T9.W, T1.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T7.W, PV.Z,
+; EG-NEXT:    6(8.407791e-45), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T9.X, T1.Z, literal.x, 1,
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T4.W, literal.y,
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T9.W, PV.X, T7.X,
+; EG-NEXT:    4(5.605194e-45), 31(4.344025e-44)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Y, T4.Y, T6.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.Z, T2.Y,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T7.X, T5.W, T0.Z,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T5.Z, T4.W, T2.X, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Y, T7.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T6.W, T1.Y, PV.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, T6.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.Z, PS, T2.Z,
+; EG-NEXT:     SETE_INT T7.W, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.W,
+; EG-NEXT:     SUBB_UINT T10.X, T6.W, T2.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T7.Z, T6.W, T2.Z,
+; EG-NEXT:     SUB_INT T7.W, PV.Y, PV.X,
+; EG-NEXT:     CNDE_INT * T4.W, T4.Y, T4.W, T5.Z,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T1.Y, PS, 1,
+; EG-NEXT:     BFE_UINT T5.Z, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T7.W, T4.Y, T6.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.Y, T6.W, PV.Z,
+; EG-NEXT:    5(7.006492e-45), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T12.X, T3.Z, T0.W,
+; EG-NEXT:     LSHL T4.Y, PS, 1,
+; EG-NEXT:     BIT_ALIGN_INT T6.Z, PV.W, T4.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T4.W, PV.Y, PV.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.X, T10.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     BFE_UINT T6.X, T4.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T1.Y, T3.Y, T5.W, PS,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.W, T2.X,
+; EG-NEXT:     SETE_INT T5.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T2.Y,
+; EG-NEXT:    4(5.605194e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Y, T4.W, T2.X, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T5.Z, PV.Y, T6.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT T5.W, T4.Y, PV.X,
+; EG-NEXT:     SUB_INT * T6.W, T12.X, T7.X,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T6.X, T11.X, T3.Z, PS,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T6.W, PV.Z, T2.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T7.W, PV.X, T4.W, PV.Y,
+; EG-NEXT:     LSHL T7.X, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Y, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Z, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PV.X, T8.W, literal.y,
+; EG-NEXT:     OR_INT * T8.W, T8.X, T9.X,
+; EG-NEXT:    4(5.605194e-45), 31(4.344025e-44)
+; EG-NEXT:     SETGE_UINT T6.X, PS, T0.Z,
+; EG-NEXT:     SETE_INT T1.Y, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T7.Z, PV.W, T0.W,
+; EG-NEXT:     SUBB_UINT T9.W, PS, T0.Z,
+; EG-NEXT:     SUB_INT * T10.W, PV.W, T0.W,
+; EG-NEXT:     SUB_INT T8.X, T5.W, T2.Z,
+; EG-NEXT:     SUBB_UINT T4.Y, T5.W, T2.Z,
+; EG-NEXT:     SUB_INT T8.Z, T5.Z, T2.W,
+; EG-NEXT:     SUB_INT T9.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T10.W, PV.Y, PV.Z, PV.X,
+; EG-NEXT:     CNDE_INT T6.X, PS, T6.W, PV.W,
+; EG-NEXT:     SUBB_UINT T1.Y, T4.W, T2.X,
+; EG-NEXT:     SUB_INT T7.Z, PV.Z, PV.Y,
+; EG-NEXT:     CNDE_INT * T4.W, T3.Z, T5.W, PV.X, BS:VEC_021/SCL_122
+; EG-NEXT:    ALU clause starting at 1894:
+; EG-NEXT:     SUB_INT * T5.W, T8.W, T0.Z,
+; EG-NEXT:     SUB_INT T8.X, T6.Z, T2.Y,
+; EG-NEXT:     CNDE_INT T4.Y, T10.W, T8.W, PV.W,
+; EG-NEXT:     LSHL T8.Z, T4.W, 1, BS:VEC_201
+; EG-NEXT:     BFE_UINT * T5.W, T4.Z, literal.x, 1, BS:VEC_120/SCL_212
+; EG-NEXT:    3(4.203895e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT * T6.W, T3.Z, T5.Z, T7.Z,
+; EG-NEXT:     BIT_ALIGN_INT T9.X, PV.W, T4.W, literal.x,
+; EG-NEXT:     OR_INT T5.Y, T8.Z, T5.W, BS:VEC_021/SCL_122
+; EG-NEXT:     LSHL T3.Z, T4.Y, 1,
+; EG-NEXT:     BFE_UINT T4.W, T1.Z, literal.y, 1, BS:VEC_120/SCL_212
+; EG-NEXT:     SUB_INT * T5.W, T8.X, T1.Y,
+; EG-NEXT:    31(4.344025e-44), 3(4.203895e-45)
+; EG-NEXT:     CNDE_INT T8.X, T10.X, T6.Z, PS,
+; EG-NEXT:     OR_INT T1.Y, PV.Z, PV.W,
+; EG-NEXT:     SETGE_UINT T3.Z, PV.Y, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T4.W, PV.X, T2.W,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.X, T2.W,
+; EG-NEXT:     CNDE_INT T10.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SETGE_UINT T6.Y, PV.Y, T0.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Z, PV.X, T7.W, literal.x,
+; EG-NEXT:     BIT_ALIGN_INT T4.W, T6.X, T4.Y, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T5.W, T7.X, T3.Y,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T6.X, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T3.Y, PV.W, T0.W,
+; EG-NEXT:     SETGE_UINT T5.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T6.W, PV.Z, T2.Y,
+; EG-NEXT:     SETGE_UINT * T7.W, PV.Z, T2.Y,
+; EG-NEXT:     CNDE_INT T7.X, PV.W, PS, PV.Z,
+; EG-NEXT:     CNDE_INT T3.Y, PV.X, PV.Y, T6.Y,
+; EG-NEXT:     SUB_INT T5.Z, T1.Y, T0.Z,
+; EG-NEXT:     SUBB_UINT T6.W, T1.Y, T0.Z,
+; EG-NEXT:     SUB_INT * T7.W, T4.W, T0.W,
+; EG-NEXT:     SUB_INT T6.X, T5.W, T2.X,
+; EG-NEXT:     SUBB_UINT T4.Y, T5.W, T2.X,
+; EG-NEXT:     SUB_INT T6.Z, T3.Z, T2.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T6.W, PS, PV.W,
+; EG-NEXT:     CNDE_INT * T7.W, PV.Y, T1.Y, PV.Z,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     BFE_UINT T1.Y, T1.Z, literal.x, 1,
+; EG-NEXT:     CNDE_INT T5.Z, T3.Y, T4.W, PV.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT T4.W, PV.Z, PV.Y,
+; EG-NEXT:     CNDE_INT * T5.W, T7.X, T5.W, PV.X,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     LSHL T6.X, PS, 1,
+; EG-NEXT:     BFE_UINT T3.Y, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT T3.Z, T7.X, T3.Z, PV.W, BS:VEC_120/SCL_212
+; EG-NEXT:     BIT_ALIGN_INT T4.W, PV.Z, T7.W, literal.y,
+; EG-NEXT:     OR_INT * T6.W, PV.X, PV.Y,
+; EG-NEXT:    3(4.203895e-45), 31(4.344025e-44)
+; EG-NEXT:     SUB_INT T7.X, T5.Y, T2.Z,
+; EG-NEXT:     SETGE_UINT T1.Y, PS, T0.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T5.Z, PV.W, T0.W,
+; EG-NEXT:     BIT_ALIGN_INT T5.W, PV.Z, T5.W, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T7.W, PV.X, PV.Y,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T6.X, T4.W, T0.W,
+; EG-NEXT:     SUBB_UINT T3.Y, T6.W, T0.Z, BS:VEC_201
+; EG-NEXT:     SETGE_UINT T3.Z, PS, T2.X,
+; EG-NEXT:     SETE_INT T8.W, PV.W, T2.Y,
+; EG-NEXT:     SETGE_UINT * T9.W, PV.W, T2.Y,
+; EG-NEXT:     SUB_INT T8.X, T4.W, T0.W,
+; EG-NEXT:     CNDE_INT T4.Y, PV.W, PS, PV.Z,
+; EG-NEXT:     SUB_INT T3.Z, T7.W, T2.X, BS:VEC_201
+; EG-NEXT:     SUBB_UINT * T8.W, T7.W, T2.X, BS:VEC_201
+; EG-NEXT:     SUB_INT * T9.W, T5.W, T2.Y,
+; EG-NEXT:     SUB_INT T11.X, PV.W, T8.W,
+; EG-NEXT:     CNDE_INT T6.Y, T4.Y, T7.W, T3.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     SUB_INT T3.Z, T6.W, T0.Z,
+; EG-NEXT:     SUB_INT T7.W, T8.X, T3.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T8.W, T5.Z, T6.X, T1.Y,
+; EG-NEXT:     CNDE_INT T6.X, PS, T4.W, PV.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT T1.Y, PS, T6.W, PV.Z, BS:VEC_102/SCL_221
+; EG-NEXT:     LSHL T3.Z, PV.Y, 1,
+; EG-NEXT:     BFE_UINT T4.W, T4.X, literal.x, 1,
+; EG-NEXT:     CNDE_INT * T5.W, T4.Y, T5.W, PV.X,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T8.X, T5.Y, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T3.Y, PS, T6.Y, literal.x,
+; EG-NEXT:     OR_INT T3.Z, PV.Z, PV.W,
+; EG-NEXT:     BFE_UINT T4.W, T1.Z, 1, 1,
+; EG-NEXT:     LSHL * T5.W, PV.Y, 1,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T11.X, T9.X, T2.W,
+; EG-NEXT:     OR_INT T4.Y, PS, PV.W,
+; EG-NEXT:     SETGE_UINT T5.Z, PV.Z, T2.X,
+; EG-NEXT:     SETE_INT T4.W, PV.Y, T2.Y,
+; EG-NEXT:     SETGE_UINT * T5.W, PV.Y, T2.Y,
+; EG-NEXT:     CNDE_INT T12.X, PV.W, PS, PV.Z,
+; EG-NEXT:     SETGE_UINT T6.Y, PV.Y, T0.Z,
+; EG-NEXT:     SUB_INT T5.Z, PV.X, T8.X,
+; EG-NEXT:     BIT_ALIGN_INT * T4.W, T6.X, T1.Y, literal.x,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT * T5.W, T10.X, T5.Y, T7.X,
+; EG-NEXT:     SETE_INT T6.X, T4.W, T0.W,
+; EG-NEXT:     SETGE_UINT T1.Y, T4.W, T0.W,
+; EG-NEXT:     LSHL T6.Z, PV.W, 1,
+; EG-NEXT:     BFE_UINT T6.W, T4.Z, literal.x, 1, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT * T7.W, T10.X, T9.X, T5.Z,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT T7.X, T3.Z, T2.X,
+; EG-NEXT:     BIT_ALIGN_INT T5.Y, PS, T5.W, literal.x,
+; EG-NEXT:     OR_INT T5.Z, PV.Z, PV.W,
+; EG-NEXT:     CNDE_INT T5.W, PV.X, PV.Y, T6.Y, BS:VEC_021/SCL_122
+; EG-NEXT:     SUB_INT * T6.W, T4.Y, T0.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUBB_UINT T6.X, T3.Z, T2.X,
+; EG-NEXT:     SUB_INT T1.Y, T3.Y, T2.Y,
+; EG-NEXT:     CNDE_INT T6.Z, PV.W, T4.Y, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SUBB_UINT T6.W, PV.Z, T2.Z,
+; EG-NEXT:     SUB_INT * T7.W, PV.Y, T2.W,
+; EG-NEXT:     SUB_INT * T8.X, PS, PV.W,
+; EG-NEXT:    ALU clause starting at 2007:
+; EG-NEXT:     LSHL T6.Y, T6.Z, 1,
+; EG-NEXT:     SETGE_UINT T7.Z, T5.Z, T2.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     SUB_INT * T6.W, T1.Y, T6.X,
+; EG-NEXT:     CNDE_INT * T7.W, T12.X, T3.Z, T7.X,
+; EG-NEXT:     SETE_INT T6.X, T5.Y, T2.W,
+; EG-NEXT:     SETGE_UINT T1.Y, T5.Y, T2.W,
+; EG-NEXT:     LSHL T3.Z, PV.W, 1,
+; EG-NEXT:     BFE_UINT T8.W, T4.X, 1, 1,
+; EG-NEXT:     CNDE_INT * T6.W, T12.X, T3.Y, T6.W,
+; EG-NEXT:     AND_INT T7.X, T1.Z, 1,
+; EG-NEXT:     BIT_ALIGN_INT T3.Y, PS, T7.W, literal.x,
+; EG-NEXT:     OR_INT T1.Z, PV.Z, PV.W,
+; EG-NEXT:     CNDE_INT * T6.W, PV.X, PV.Y, T7.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SUB_INT * T7.W, T5.Z, T2.Z,
+; EG-NEXT:     CNDE_INT T6.X, T6.W, T5.Z, PV.W,
+; EG-NEXT:     SETGE_UINT T1.Y, T1.Z, T2.X, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Z, T3.Y, T2.Y,
+; EG-NEXT:     SETGE_UINT T7.W, T3.Y, T2.Y,
+; EG-NEXT:     OR_INT * T8.W, T6.Y, T7.X,
+; EG-NEXT:     SUB_INT T7.X, PS, T0.Z,
+; EG-NEXT:     CNDE_INT T1.Y, PV.Z, PV.W, PV.Y,
+; EG-NEXT:     LSHL T3.Z, PV.X, 1,
+; EG-NEXT:     BFE_UINT T7.W, T4.Z, 1, 1,
+; EG-NEXT:     CNDE_INT * T6.W, T6.W, T5.Y, T8.X,
+; EG-NEXT:     SUB_INT T8.X, T1.Z, T2.X,
+; EG-NEXT:     SUBB_UINT T5.Y, T1.Z, T2.X,
+; EG-NEXT:     SUB_INT T5.Z, T3.Y, T2.Y,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, PS, T6.X, literal.x, BS:VEC_021/SCL_122
+; EG-NEXT:     OR_INT * T7.W, PV.Z, PV.W,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T6.X, PS, T2.Z, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T6.Y, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT T3.Z, PV.W, T2.W,
+; EG-NEXT:     SUB_INT T9.W, PV.Z, PV.Y,
+; EG-NEXT:     CNDE_INT * T10.W, T1.Y, T1.Z, PV.X,
+; EG-NEXT:     LSHL T8.X, PS, 1,
+; EG-NEXT:     AND_INT T5.Y, T4.X, 1,
+; EG-NEXT:     CNDE_INT T1.Z, T1.Y, T3.Y, PV.W,
+; EG-NEXT:     CNDE_INT T9.W, PV.Y, PV.Z, PV.X,
+; EG-NEXT:     SUB_INT * T11.W, T7.W, T2.Z,
+; EG-NEXT:     SUBB_UINT T4.X, T7.W, T2.Z,
+; EG-NEXT:     SUB_INT T1.Y, T6.W, T2.W, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT * T3.Z, PV.W, T7.W, PS, BS:VEC_102/SCL_221
+; EG-NEXT:     BIT_ALIGN_INT T7.W, T1.Z, T10.W, literal.x,
+; EG-NEXT:     OR_INT * T10.W, T8.X, T5.Y,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETGE_UINT T6.X, PS, T2.X, BS:VEC_021/SCL_122
+; EG-NEXT:     SETE_INT T3.Y, PV.W, T2.Y,
+; EG-NEXT:     LSHL T1.Z, T3.Z, 1,
+; EG-NEXT:     AND_INT T11.W, T4.Z, 1, BS:VEC_120/SCL_212
+; EG-NEXT:     SUB_INT * T12.W, T1.Y, T4.X,
+; EG-NEXT:     SETGE_UINT T4.X, T7.W, T2.Y,
+; EG-NEXT:     SUBB_UINT T1.Y, T4.Y, T0.Z,
+; EG-NEXT:     SUB_INT * T4.Z, T4.W, T0.W, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT * T6.W, T9.W, T6.W, T12.W,
+; EG-NEXT:     OR_INT * T9.W, T1.Z, T11.W,
+; EG-NEXT:     SUBB_UINT T8.X, T10.W, T2.X,
+; EG-NEXT:     SUB_INT T2.Y, T7.W, T2.Y, BS:VEC_120/SCL_212
+; EG-NEXT:     SETGE_UINT T1.Z, PV.W, T2.Z,
+; EG-NEXT:     BIT_ALIGN_INT T6.W, T6.W, T3.Z, literal.x, BS:VEC_201
+; EG-NEXT:     SUB_INT * T11.W, T4.Z, T1.Y,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T9.X, PV.W, T2.W,
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T2.W,
+; EG-NEXT:     CNDE_INT * T3.Z, T5.W, T4.W, PS, BS:VEC_021/SCL_122
+; EG-NEXT:     SUBB_UINT T4.W, T9.W, T2.Z,
+; EG-NEXT:     SUB_INT * T2.W, T6.W, T2.W,
+; EG-NEXT:     SUB_INT T10.X, T9.W, T2.Z,
+; EG-NEXT:     SETGE_UINT T4.Y, T8.W, T0.Z, BS:VEC_120/SCL_212
+; EG-NEXT:     SUB_INT * T2.Z, PS, PV.W,
+; EG-NEXT:     BIT_ALIGN_INT T2.W, T3.Z, T6.Z, literal.x, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT * T4.W, T9.X, T1.Y, T1.Z,
+; EG-NEXT:    31(4.344025e-44), 0(0.000000e+00)
+; EG-NEXT:     SETE_INT T9.X, PV.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SETGE_UINT T1.Y, PV.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     SUBB_UINT T0.Z, T8.W, T0.Z,
+; EG-NEXT:     SUB_INT T0.W, PV.W, T0.W, BS:VEC_021/SCL_122
+; EG-NEXT:     CNDE_INT * T6.W, PS, T6.W, T2.Z,
+; EG-NEXT:     SUB_INT T11.X, PV.W, PV.Z,
+; EG-NEXT:     CNDE_INT T1.Y, PV.X, PV.Y, T4.Y,
+; EG-NEXT:     CNDE_INT T6.Z, T4.W, T9.W, T10.X,
+; EG-NEXT:     SUB_INT * T0.W, T2.Y, T8.X,
+; EG-NEXT:     CNDE_INT * T4.W, T3.Y, T4.X, T6.X,
+; EG-NEXT:     SUBB_UINT T4.X, T3.W, T0.X,
+; EG-NEXT:     CNDE_INT * T6.Y, PV.W, T7.W, T0.W,
+; EG-NEXT:     SUB_INT T0.Z, T1.W, T0.Y,
+; EG-NEXT:     SUB_INT T0.W, T10.W, T2.X, BS:VEC_210
+; EG-NEXT:     CNDE_INT * T2.W, T1.Y, T2.W, T11.X,
+; EG-NEXT:     CNDE_INT T6.X, T4.W, T10.W, PV.W,
+; EG-NEXT:     SUB_INT T0.Y, PV.Z, T4.X,
+; EG-NEXT:     CNDE_INT * T2.Z, T1.Y, T8.W, T7.X, BS:VEC_120/SCL_212
+; EG-NEXT:     CNDE_INT T0.W, T3.X, T5.X, T1.X,
+; EG-NEXT:     ADD_INT * T4.W, KC0[2].Y, literal.x,
+; EG-NEXT:    16(2.242078e-44), 0(0.000000e+00)
+; EG-NEXT:     LSHR T1.X, PS, literal.x,
+; EG-NEXT:     CNDE_INT T2.Y, PV.W, T1.W, T0.Y,
+; EG-NEXT:     SUB_INT * T1.W, T3.W, T0.X,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
+; EG-NEXT:     CNDE_INT T2.X, T0.W, T3.W, PV.W,
+; EG-NEXT:     LSHR * T0.X, KC0[2].Y, literal.x,
+; EG-NEXT:    2(2.802597e-45), 0(0.000000e+00)
   %b_ptr = getelementptr <4 x i64>, ptr addrspace(1) %in, i64 1
   %a = load <4 x i64>, ptr addrspace(1) %in
   %b = load <4 x i64>, ptr addrspace(1) %b_ptr
   %result = urem <4 x i64> %a, %b
   store <4 x i64> %result, ptr addrspace(1) %out
   ret void
 }

llvm/test/CodeGen/AMDGPU/urem64.ll

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s1, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s0, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s1, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s0, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GCN-NEXT:    v_trunc_f32_e32 v1, v1
 ; GCN-NEXT:    v_mac_f32_e32 v0, 0xcf800000, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v5, s1, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s0, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v5, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v5
 ; GCN-NEXT:    v_mul_hi_u32 v3, v0, v4
 ; GCN-NEXT:    v_mul_lo_u32 v5, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v1, v4
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, v4
 ; GCN-NEXT:    v_mul_hi_u32 v8, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v5
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v3, v4
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v6, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v8, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v2, s0, v1
 ; GCN-NEXT:    v_mul_hi_u32 v3, s0, v0
 ; GCN-NEXT:    v_mul_lo_u32 v4, s1, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_mul_lo_u32 v3, s0, v0
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_lo_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v0, v3
 ; GCN-NEXT:    v_mul_hi_u32 v8, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v1, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v4, v1, v2

 ; GCN-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GCN-NEXT:    v_cvt_u32_f32_e32 v1, v1
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_mov_b32 s0, s4
 ; GCN-NEXT:    s_mov_b32 s1, s5
 ; GCN-NEXT:    v_mul_hi_u32 v2, v0, s2
 ; GCN-NEXT:    v_mul_lo_u32 v4, v1, s2
 ; GCN-NEXT:    v_mul_lo_u32 v3, v0, s2
-; GCN-NEXT:    v_subrev_i32_e32 v2, vcc, v0, v2
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v4, v2
+; GCN-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v4
 ; GCN-NEXT:    v_mul_hi_u32 v5, v0, v3
 ; GCN-NEXT:    v_mul_lo_u32 v4, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v5, v4
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v6, v1, v3
 ; GCN-NEXT:    v_mul_hi_u32 v3, v1, v3
 ; GCN-NEXT:    v_add_i32_e32 v4, vcc, v4, v6
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, v5, v3, vcc
 ; GCN-NEXT:    v_addc_u32_e32 v4, vcc, 0, v7, vcc
 ; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
 ; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v4, vcc
 ; GCN-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GCN-NEXT:    v_addc_u32_e32 v1, vcc, v1, v3, vcc
 ; GCN-NEXT:    v_mul_hi_u32 v2, v0, s2
 ; GCN-NEXT:    v_mul_lo_u32 v3, v1, s2
 ; GCN-NEXT:    v_mul_lo_u32 v4, v0, s2
 ; GCN-NEXT:    s_mov_b32 s2, -1
 ; GCN-NEXT:    v_sub_i32_e32 v2, vcc, v2, v0
-; GCN-NEXT:    v_add_i32_e32 v2, vcc, v3, v2
+; GCN-NEXT:    v_add_i32_e32 v2, vcc, v2, v3
 ; GCN-NEXT:    v_mul_lo_u32 v3, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v5, v0, v4
 ; GCN-NEXT:    v_mul_hi_u32 v6, v0, v2
 ; GCN-NEXT:    v_mul_hi_u32 v7, v1, v2
 ; GCN-NEXT:    v_mul_lo_u32 v2, v1, v2
 ; GCN-NEXT:    v_add_i32_e32 v3, vcc, v5, v3
 ; GCN-NEXT:    v_addc_u32_e32 v5, vcc, 0, v6, vcc
 ; GCN-NEXT:    v_mul_lo_u32 v6, v1, v4

llvm/test/CodeGen/AMDGPU/vgpr-liverange-ir.ll

define amdgpu_ps float @loop(i32 %z, float %v, i32 inreg %bound, ptr %extern_func, ptr %extern_func2) #0 {
   ; SI-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed $vgpr0
   ; SI-NEXT:   [[V_CMP_GT_I32_e64_:%[0-9]+]]:sreg_32 = V_CMP_GT_I32_e64 6, killed [[COPY5]], implicit $exec
   ; SI-NEXT:   [[SI_IF:%[0-9]+]]:sreg_32 = SI_IF killed [[V_CMP_GT_I32_e64_]], %bb.1, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_BRANCH %bb.6
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.1.Flow:
   ; SI-NEXT:   successors: %bb.2(0x40000000), %bb.10(0x40000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI undef %49:vgpr_32, %bb.0, %4, %bb.9
-  ; SI-NEXT:   [[PHI1:%[0-9]+]]:vgpr_32 = PHI [[COPY4]], %bb.0, undef %51:vgpr_32, %bb.9
-  ; SI-NEXT:   [[PHI2:%[0-9]+]]:vgpr_32 = PHI [[COPY3]], %bb.0, undef %53:vgpr_32, %bb.9
-  ; SI-NEXT:   [[PHI3:%[0-9]+]]:vgpr_32 = PHI [[COPY2]], %bb.0, undef %55:vgpr_32, %bb.9
+  ; SI-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI undef %47:vgpr_32, %bb.0, %4, %bb.9
+  ; SI-NEXT:   [[PHI1:%[0-9]+]]:vgpr_32 = PHI [[COPY4]], %bb.0, undef %49:vgpr_32, %bb.9
+  ; SI-NEXT:   [[PHI2:%[0-9]+]]:vgpr_32 = PHI [[COPY3]], %bb.0, undef %51:vgpr_32, %bb.9
+  ; SI-NEXT:   [[PHI3:%[0-9]+]]:vgpr_32 = PHI [[COPY2]], %bb.0, undef %53:vgpr_32, %bb.9
   ; SI-NEXT:   [[SI_ELSE:%[0-9]+]]:sreg_32 = SI_ELSE killed [[SI_IF]], %bb.10, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_BRANCH %bb.2
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.2.if:
   ; SI-NEXT:   successors: %bb.3(0x80000000)
   ; SI-NEXT: {{  $}}
   ; SI-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[PHI2]], %subreg.sub0, killed [[PHI3]], %subreg.sub1
   ; SI-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.3:
   ; SI-NEXT:   successors: %bb.4(0x80000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI4:%[0-9]+]]:vreg_64 = PHI undef %57:vreg_64, %bb.4, [[REG_SEQUENCE]], %bb.2
-  ; SI-NEXT:   [[PHI5:%[0-9]+]]:vgpr_32 = PHI undef %59:vgpr_32, %bb.4, [[PHI1]], %bb.2
+  ; SI-NEXT:   [[PHI4:%[0-9]+]]:vreg_64 = PHI undef %55:vreg_64, %bb.4, [[REG_SEQUENCE]], %bb.2
+  ; SI-NEXT:   [[PHI5:%[0-9]+]]:vgpr_32 = PHI undef %57:vgpr_32, %bb.4, [[PHI1]], %bb.2
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI4]].sub0, implicit $exec
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI4]].sub1, implicit $exec
   ; SI-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE killed [[V_READFIRSTLANE_B32_]], %subreg.sub0, killed [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; SI-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], killed [[PHI4]], implicit $exec
   ; SI-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[V_CMP_EQ_U64_e64_]], implicit-def $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.4:
   ; SI-NEXT:   successors: %bb.3(0x40000000), %bb.5(0x40000000)
define amdgpu_ps float @loop(i32 %z, float %v, i32 inreg %bound, ptr %extern_func, ptr %extern_func2) #0 {
   ; SI-NEXT:   successors: %bb.7(0x80000000)
   ; SI-NEXT: {{  $}}
   ; SI-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[COPY1]], %subreg.sub0, killed [[COPY]], %subreg.sub1
   ; SI-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.7:
   ; SI-NEXT:   successors: %bb.8(0x80000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI6:%[0-9]+]]:vreg_64 = PHI undef %61:vreg_64, %bb.8, [[REG_SEQUENCE2]], %bb.6
-  ; SI-NEXT:   [[PHI7:%[0-9]+]]:vgpr_32 = PHI undef %63:vgpr_32, %bb.8, [[COPY4]], %bb.6
+  ; SI-NEXT:   [[PHI6:%[0-9]+]]:vreg_64 = PHI undef %59:vreg_64, %bb.8, [[REG_SEQUENCE2]], %bb.6
+  ; SI-NEXT:   [[PHI7:%[0-9]+]]:vgpr_32 = PHI undef %61:vgpr_32, %bb.8, [[COPY4]], %bb.6
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI6]].sub0, implicit $exec
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI6]].sub1, implicit $exec
   ; SI-NEXT:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_64 = REG_SEQUENCE killed [[V_READFIRSTLANE_B32_2]], %subreg.sub0, killed [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; SI-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], killed [[PHI6]], implicit $exec
   ; SI-NEXT:   [[S_AND_SAVEEXEC_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[V_CMP_EQ_U64_e64_1]], implicit-def $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.8:
   ; SI-NEXT:   successors: %bb.7(0x40000000), %bb.9(0x40000000)
define amdgpu_ps float @loop_with_use(i32 %z, float %v, i32 inreg %bound, ptr %extern_func, ptr %extern_func2) #0 {
   ; SI-NEXT:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY killed $vgpr0
   ; SI-NEXT:   [[V_CMP_GT_I32_e64_:%[0-9]+]]:sreg_32 = V_CMP_GT_I32_e64 6, killed [[COPY5]], implicit $exec
   ; SI-NEXT:   [[SI_IF:%[0-9]+]]:sreg_32 = SI_IF killed [[V_CMP_GT_I32_e64_]], %bb.1, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_BRANCH %bb.6
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.1.Flow:
   ; SI-NEXT:   successors: %bb.2(0x40000000), %bb.10(0x40000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI undef %50:vgpr_32, %bb.0, %4, %bb.9
-  ; SI-NEXT:   [[PHI1:%[0-9]+]]:vgpr_32 = PHI [[COPY3]], %bb.0, undef %52:vgpr_32, %bb.9
-  ; SI-NEXT:   [[PHI2:%[0-9]+]]:vgpr_32 = PHI [[COPY2]], %bb.0, undef %54:vgpr_32, %bb.9
+  ; SI-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI undef %48:vgpr_32, %bb.0, %4, %bb.9
+  ; SI-NEXT:   [[PHI1:%[0-9]+]]:vgpr_32 = PHI [[COPY3]], %bb.0, undef %50:vgpr_32, %bb.9
+  ; SI-NEXT:   [[PHI2:%[0-9]+]]:vgpr_32 = PHI [[COPY2]], %bb.0, undef %52:vgpr_32, %bb.9
   ; SI-NEXT:   [[SI_ELSE:%[0-9]+]]:sreg_32 = SI_ELSE killed [[SI_IF]], %bb.10, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_BRANCH %bb.2
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.2.if:
   ; SI-NEXT:   successors: %bb.3(0x80000000)
   ; SI-NEXT: {{  $}}
   ; SI-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[PHI1]], %subreg.sub0, killed [[PHI2]], %subreg.sub1
   ; SI-NEXT:   [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.3:
   ; SI-NEXT:   successors: %bb.4(0x80000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI3:%[0-9]+]]:vreg_64 = PHI undef %56:vreg_64, %bb.4, [[REG_SEQUENCE]], %bb.2
+  ; SI-NEXT:   [[PHI3:%[0-9]+]]:vreg_64 = PHI undef %54:vreg_64, %bb.4, [[REG_SEQUENCE]], %bb.2
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI3]].sub0, implicit $exec
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI3]].sub1, implicit $exec
   ; SI-NEXT:   [[REG_SEQUENCE1:%[0-9]+]]:sgpr_64 = REG_SEQUENCE killed [[V_READFIRSTLANE_B32_]], %subreg.sub0, killed [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; SI-NEXT:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], killed [[PHI3]], implicit $exec
   ; SI-NEXT:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[V_CMP_EQ_U64_e64_]], implicit-def $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.4:
   ; SI-NEXT:   successors: %bb.3(0x40000000), %bb.5(0x40000000)
define amdgpu_ps float @loop_with_use(i32 %z, float %v, i32 inreg %bound, ptr %extern_func, ptr %extern_func2) #0 {
   ; SI-NEXT:   successors: %bb.7(0x80000000)
   ; SI-NEXT: {{  $}}
   ; SI-NEXT:   [[REG_SEQUENCE2:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[COPY1]], %subreg.sub0, killed [[COPY]], %subreg.sub1
   ; SI-NEXT:   [[S_MOV_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.7:
   ; SI-NEXT:   successors: %bb.8(0x80000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI4:%[0-9]+]]:vreg_64 = PHI undef %58:vreg_64, %bb.8, [[REG_SEQUENCE2]], %bb.6
+  ; SI-NEXT:   [[PHI4:%[0-9]+]]:vreg_64 = PHI undef %56:vreg_64, %bb.8, [[REG_SEQUENCE2]], %bb.6
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI4]].sub0, implicit $exec
   ; SI-NEXT:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[PHI4]].sub1, implicit $exec
   ; SI-NEXT:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_64 = REG_SEQUENCE killed [[V_READFIRSTLANE_B32_2]], %subreg.sub0, killed [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; SI-NEXT:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], killed [[PHI4]], implicit $exec
   ; SI-NEXT:   [[S_AND_SAVEEXEC_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[V_CMP_EQ_U64_e64_1]], implicit-def $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.8:
   ; SI-NEXT:   successors: %bb.7(0x40000000), %bb.9(0x40000000)
define amdgpu_kernel void @livevariables_update_missed_block(ptr addrspace(1) %src1) {
   ; SI-NEXT:   [[SI_IF:%[0-9]+]]:sreg_32 = SI_IF killed [[V_CMP_NE_U32_e64_]], %bb.5, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_BRANCH %bb.2
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.1.if.then:
   ; SI-NEXT:   successors: %bb.7(0x80000000)
   ; SI-NEXT: {{  $}}
   ; SI-NEXT:   [[S_LOAD_DWORDX2_IMM:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM killed [[COPY]](p4), 36, 0 :: (dereferenceable invariant load (s64) from %ir.src1.kernarg.offset, align 4, addrspace 4)
   ; SI-NEXT:   [[V_ADD_CO_U32_e64_:%[0-9]+]]:vgpr_32, [[V_ADD_CO_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADD_CO_U32_e64 [[S_LOAD_DWORDX2_IMM]].sub0, killed %51, 0, implicit $exec
-  ; SI-NEXT:   %44:vgpr_32, dead %46:sreg_32_xm0_xexec = V_ADDC_U32_e64 0, killed [[S_LOAD_DWORDX2_IMM]].sub1, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
-  ; SI-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[V_ADD_CO_U32_e64_]], %subreg.sub0, killed %44, %subreg.sub1
+  ; SI-NEXT:   [[V_ADDC_U32_e64_:%[0-9]+]]:vgpr_32, dead [[V_ADDC_U32_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_ADDC_U32_e64 0, killed [[S_LOAD_DWORDX2_IMM]].sub1, killed [[V_ADD_CO_U32_e64_1]], 0, implicit $exec
+  ; SI-NEXT:   [[REG_SEQUENCE:%[0-9]+]]:vreg_64 = REG_SEQUENCE killed [[V_ADD_CO_U32_e64_]], %subreg.sub0, killed [[V_ADDC_U32_e64_]], %subreg.sub1
   ; SI-NEXT:   [[GLOBAL_LOAD_UBYTE:%[0-9]+]]:vgpr_32 = GLOBAL_LOAD_UBYTE killed [[REG_SEQUENCE]], 0, 0, implicit $exec :: (load (s8) from %ir.i10, addrspace 1)
   ; SI-NEXT:   [[V_MOV_B:%[0-9]+]]:vreg_64 = V_MOV_B64_PSEUDO 0, implicit $exec
   ; SI-NEXT:   GLOBAL_STORE_BYTE killed [[V_MOV_B]], killed [[GLOBAL_LOAD_UBYTE]], 0, 0, implicit $exec :: (store (s8) into `ptr addrspace(1) null`, addrspace 1)
   ; SI-NEXT:   S_BRANCH %bb.7
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.2.if.then9:
   ; SI-NEXT:   successors: %bb.4(0x40000000), %bb.3(0x40000000)
   ; SI-NEXT: {{  $}}
define amdgpu_kernel void @livevariables_update_missed_block(ptr addrspace(1) %src1) {
   ; SI-NEXT: {{  $}}
   ; SI-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI [[COPY1]](s32), %bb.0, undef %52:vgpr_32, %bb.6
   ; SI-NEXT:   [[SI_ELSE:%[0-9]+]]:sreg_32 = SI_ELSE killed [[SI_IF]], %bb.7, implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_BRANCH %bb.1
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.6.sw.bb18:
   ; SI-NEXT:   successors: %bb.5(0x80000000)
   ; SI-NEXT: {{  $}}
-  ; SI-NEXT:   [[PHI1:%[0-9]+]]:vgpr_32 = PHI undef %36:vgpr_32, %bb.3, [[GLOBAL_LOAD_UBYTE1]], %bb.4
+  ; SI-NEXT:   [[PHI1:%[0-9]+]]:vgpr_32 = PHI undef %40:vgpr_32, %bb.3, [[GLOBAL_LOAD_UBYTE1]], %bb.4
   ; SI-NEXT:   [[V_MOV_B2:%[0-9]+]]:vreg_64 = V_MOV_B64_PSEUDO 0, implicit $exec
   ; SI-NEXT:   GLOBAL_STORE_BYTE killed [[V_MOV_B2]], killed [[PHI1]], 0, 0, implicit $exec :: (store (s8) into `ptr addrspace(1) null`, addrspace 1)
   ; SI-NEXT:   S_BRANCH %bb.5
   ; SI-NEXT: {{  $}}
   ; SI-NEXT: bb.7.UnifiedReturnBlock:
   ; SI-NEXT:   SI_END_CF killed [[SI_ELSE]], implicit-def dead $exec, implicit-def dead $scc, implicit $exec
   ; SI-NEXT:   S_ENDPGM 0
 entry:

llvm/test/CodeGen/AMDGPU/wwm-reserved-spill.ll

 
 define amdgpu_gfx i64 @strict_wwm_called_i64(i64 %a) noinline {
 ; GFX9-O0-LABEL: strict_wwm_called_i64:
 ; GFX9-O0:       ; %bb.0:
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
-; GFX9-O0-NEXT:    ; kill: def $vgpr1 killed $vgpr1 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34_sgpr35
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v2
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v2
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v0, v3
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX9-O0-NEXT:    v_add_co_u32_e64 v4, s[34:35], v4, v5

 ; GFX9-O0-NEXT:    v_readlane_b32 s36, v10, 2
 ; GFX9-O0-NEXT:    v_readlane_b32 s37, v10, 3
 ; GFX9-O0-NEXT:    v_readlane_b32 s38, v10, 4
 ; GFX9-O0-NEXT:    v_readlane_b32 s39, v10, 5
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v1
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr40
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr40
-; GFX9-O0-NEXT:    ; kill: def $vgpr2 killed $vgpr2 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v8
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v9
 ; GFX9-O0-NEXT:    v_add_co_u32_e64 v2, s[40:41], v2, v4
 ; GFX9-O0-NEXT:    v_addc_co_u32_e64 v3, s[40:41], v3, v5, s[40:41]
 ; GFX9-O0-NEXT:    s_mov_b64 exec, s[34:35]
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v0, v2
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX9-O0-NEXT:    s_mov_b32 s34, 0

 ; GFX9-O0-NEXT:    s_mov_b32 s34, 0
 ; GFX9-O0-NEXT:    buffer_load_dwordx4 v[10:13], v0, s[36:39], s34 offen
 ; GFX9-O0-NEXT:    buffer_load_dwordx2 v[3:4], v0, s[36:39], s34 offen offset:16
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(1)
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v11
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v10
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
-; GFX9-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v7
 ; GFX9-O0-NEXT:    s_mov_b32 s35, 0x7fffffff
 ; GFX9-O0-NEXT:    s_mov_b32 s40, -1
 ; GFX9-O0-NEXT:    ; kill: def $sgpr40 killed $sgpr40 def $sgpr40_sgpr41
 ; GFX9-O0-NEXT:    s_mov_b32 s41, s35
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v6
 ; GFX9-O0-NEXT:    s_not_b64 exec, exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s40
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s41
 ; GFX9-O0-NEXT:    s_not_b64 exec, exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v1
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v2
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v13
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v12
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
-; GFX9-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v7
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v6
 ; GFX9-O0-NEXT:    s_not_b64 exec, exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s40
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s41
 ; GFX9-O0-NEXT:    s_not_b64 exec, exec

 ; GFX9-O0-NEXT:    v_mov_b32_e32 v11, v9
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v8
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v7
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v10, v6
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr35
-; GFX9-O0-NEXT:    ; kill: def $vgpr5 killed $vgpr5 killed $exec
-; GFX9-O0-NEXT:    ; kill: def $vgpr11 killed $vgpr11 killed $exec
-; GFX9-O0-NEXT:    ; kill: def $vgpr9 killed $vgpr9 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr5 killed $vgpr5 def $vgpr5_vgpr6_vgpr7_vgpr8 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v11
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v10
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v9
 ; GFX9-O0-NEXT:    buffer_store_dwordx4 v[5:8], v0, s[36:39], s34 offen
 ; GFX9-O0-NEXT:    buffer_store_dwordx2 v[3:4], v0, s[36:39], s34 offen offset:16
 ; GFX9-O0-NEXT:    s_xor_saveexec_b64 s[34:35], -1
 ; GFX9-O0-NEXT:    buffer_load_dword v1, off, s[0:3], s32 ; 4-byte Folded Reload

 }
 
 ; FIXME: This spills v40 and v41 twice, once in whole-wave-mode and once for the active lanes
 define amdgpu_gfx <32 x i32> @strict_wwm_callee_saves(<32 x i32> inreg %keep, ptr addrspace(5) %ptr, i64 %a, i64 %b, i64 %c, i64 %d, i64 %e) #0 {
 ; GFX9-O0-LABEL: strict_wwm_callee_saves:
 ; GFX9-O0:       ; %bb.0:
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-O0-NEXT:    s_xor_saveexec_b64 s[34:35], -1
-; GFX9-O0-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v32, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v33, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v34, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v33, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v34, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v35, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v36, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v35, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v36, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v37, off, s[0:3], s32 offset:244 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v38, off, s[0:3], s32 offset:248 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v37, off, s[0:3], s32 offset:224 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v38, off, s[0:3], s32 offset:228 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v39, off, s[0:3], s32 offset:252 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:256 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v39, off, s[0:3], s32 offset:232 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:236 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:260 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:240 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_mov_b64 exec, -1
-; GFX9-O0-NEXT:    buffer_store_dword v42, off, s[0:3], s32 offset:220 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v42, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_mov_b64 exec, s[34:35]
 ; GFX9-O0-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v41, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v43, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v44, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v45, off, s[0:3], s32 offset:32 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v46, off, s[0:3], s32 offset:28 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v47, off, s[0:3], s32 offset:24 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    v_writelane_b32 v42, s64, 0
 ; GFX9-O0-NEXT:    v_writelane_b32 v42, s65, 1
 ; GFX9-O0-NEXT:    v_writelane_b32 v42, s66, 2
 ; GFX9-O0-NEXT:    v_writelane_b32 v42, s67, 3
 ; GFX9-O0-NEXT:    buffer_store_dword v10, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v9, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v8, off, s[0:3], s32 offset:76 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v7, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v6, off, s[0:3], s32 offset:84 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v5, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:92 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:80 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:88 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_load_dword v1, off, s[0:3], s32 offset:20
+; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:20
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_load_dword v44, off, s[0:3], s32 offset:16
+; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:116 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:16
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:112 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:12
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:108 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_load_dword v46, off, s[0:3], s32 offset:8
+; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:8
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v46, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_load_dword v47, off, s[0:3], s32 offset:4
+; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:104 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:4
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v47, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_load_dword v45, off, s[0:3], s32
+; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:100 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v45, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v0, off, s[0:3], s32 offset:96 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v43, s5
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s6
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:216 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s7
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:212 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s8
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:208 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s9
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:204 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s10
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:200 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s11
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s12
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s13
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s14
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s15
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s16
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s17
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s18
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s19
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s20
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s21
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s22
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s23
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s24
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s25
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s26
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s27
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s28
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, s29
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    ; kill: def $vgpr45 killed $vgpr45 killed $exec
-; GFX9-O0-NEXT:    ; kill: def $vgpr47 killed $vgpr47 killed $exec
-; GFX9-O0-NEXT:    ; kill: def $vgpr46 killed $vgpr46 killed $exec
-; GFX9-O0-NEXT:    ; kill: def $vgpr44 killed $vgpr44 killed $exec
-; GFX9-O0-NEXT:    ; kill: def $vgpr1 killed $vgpr1 killed $exec
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s6
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:196 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s7
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:192 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s8
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:188 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s9
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:184 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s10
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:180 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s11
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:176 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s12
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:172 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s13
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:168 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s14
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:164 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s15
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:160 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s16
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:156 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s17
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:152 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s18
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:148 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s19
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:144 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s20
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:140 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s21
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:136 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s22
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:132 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s23
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:128 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s24
+; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:124 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v47, s25
+; GFX9-O0-NEXT:    v_mov_b32_e32 v46, s26
+; GFX9-O0-NEXT:    v_mov_b32_e32 v45, s27
+; GFX9-O0-NEXT:    v_mov_b32_e32 v44, s28
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, s29
 ; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:120 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    ; kill: def $vgpr0 killed $vgpr0 def $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17_vgpr18_vgpr19_vgpr20_vgpr21_vgpr22_vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v43
-; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v43
-; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v43
-; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v43
-; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v43
-; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:196 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:192 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:188 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:184 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v10, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:180 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v11, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:176 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v12, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:172 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v13, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:168 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v14, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:164 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v15, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v10, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:160 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v16, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v11, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:156 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v17, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v12, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:152 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v18, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v13, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:148 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v19, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v14, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:144 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v20, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v15, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:140 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v21, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v16, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:136 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v22, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v17, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:132 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v23, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v18, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:128 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v24, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v19, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:124 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    v_mov_b32_e32 v25, v43
+; GFX9-O0-NEXT:    v_mov_b32_e32 v20, v43
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:120 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    v_mov_b32_e32 v26, v45
+; GFX9-O0-NEXT:    v_mov_b32_e32 v21, v47
+; GFX9-O0-NEXT:    buffer_load_dword v47, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    v_mov_b32_e32 v22, v46
+; GFX9-O0-NEXT:    buffer_load_dword v46, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    v_mov_b32_e32 v23, v45
 ; GFX9-O0-NEXT:    buffer_load_dword v45, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    v_mov_b32_e32 v24, v44
+; GFX9-O0-NEXT:    buffer_load_dword v44, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    s_waitcnt vmcnt(4)
+; GFX9-O0-NEXT:    v_mov_b32_e32 v25, v43
+; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
+; GFX9-O0-NEXT:    v_mov_b32_e32 v26, v43
+; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v27, v47
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v28, v46
-; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v29, v45
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v30, v44
 ; GFX9-O0-NEXT:    ; kill: def $vgpr31 killed $vgpr43 killed $exec
 ; GFX9-O0-NEXT:    buffer_load_dword v31, off, s[0:3], s32 offset:116 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v30, off, s[0:3], s32 offset:112 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v29, off, s[0:3], s32 offset:108 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v28, off, s[0:3], s32 offset:104 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v27, off, s[0:3], s32 offset:100 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v26, off, s[0:3], s32 offset:96 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v11, off, s[0:3], s32 offset:92 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v10, off, s[0:3], s32 offset:88 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v9, off, s[0:3], s32 offset:84 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v8, off, s[0:3], s32 offset:80 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v7, off, s[0:3], s32 offset:76 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v6, off, s[0:3], s32 offset:72 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v5, off, s[0:3], s32 offset:68 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v4, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v2, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v1, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
-; GFX9-O0-NEXT:    ; kill: def $vgpr5 killed $vgpr5 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr2 killed $vgpr2 def $vgpr2_vgpr3 killed $exec
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(4)
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v5
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
-; GFX9-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr4 killed $vgpr4 def $vgpr4_vgpr5 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v7
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
-; GFX9-O0-NEXT:    ; kill: def $vgpr9 killed $vgpr9 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr6 killed $vgpr6 def $vgpr6_vgpr7 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v9
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
-; GFX9-O0-NEXT:    ; kill: def $vgpr8 killed $vgpr8 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v11
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34
-; GFX9-O0-NEXT:    ; kill: def $vgpr10 killed $vgpr10 killed $exec
 ; GFX9-O0-NEXT:    ; kill: def $vgpr10 killed $vgpr10 def $vgpr10_vgpr11 killed $exec
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v11, v0
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34_sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34_sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34_sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34_sgpr35
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr34_sgpr35

 ; GFX9-O0-NEXT:    buffer_load_dword v47, off, s[0:3], s32 offset:24 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v46, off, s[0:3], s32 offset:28 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v45, off, s[0:3], s32 offset:32 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v44, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v43, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_xor_saveexec_b64 s[34:35], -1
-; GFX9-O0-NEXT:    buffer_load_dword v32, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v32, off, s[0:3], s32 offset:204 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v33, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v33, off, s[0:3], s32 offset:208 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v34, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v34, off, s[0:3], s32 offset:212 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v35, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v35, off, s[0:3], s32 offset:216 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v36, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v36, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v37, off, s[0:3], s32 offset:244 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v37, off, s[0:3], s32 offset:224 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v38, off, s[0:3], s32 offset:248 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v38, off, s[0:3], s32 offset:228 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v39, off, s[0:3], s32 offset:252 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v39, off, s[0:3], s32 offset:232 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:256 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:236 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_nop 0
-; GFX9-O0-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v41, off, s[0:3], s32 offset:240 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_mov_b64 exec, -1
-; GFX9-O0-NEXT:    buffer_load_dword v42, off, s[0:3], s32 offset:220 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v42, off, s[0:3], s32 offset:200 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_mov_b64 exec, s[34:35]
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-O0-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-O3-LABEL: strict_wwm_callee_saves:
 ; GFX9-O3:       ; %bb.0:
 ; GFX9-O3-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-O3-NEXT:    s_xor_saveexec_b64 s[34:35], -1