Index: llvm/trunk/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
@@ -1014,6 +1014,9 @@
 }
 
 void AMDGPUDAGToDAGISel::SelectDIV_FMAS(SDNode *N) {
+  const GCNSubtarget *ST = static_cast<const GCNSubtarget *>(Subtarget);
+  const SIRegisterInfo *TRI = ST->getRegisterInfo();
+
   SDLoc SL(N);
   EVT VT = N->getValueType(0);
 
@@ -1025,7 +1028,7 @@
   SDValue CarryIn = N->getOperand(3);
   // V_DIV_FMAS implicitly reads VCC.
   SDValue VCC = CurDAG->getCopyToReg(CurDAG->getEntryNode(), SL,
-                                     AMDGPU::VCC, CarryIn, SDValue());
+                                     TRI->getVCC(), CarryIn, SDValue());
 
   SDValue Ops[10];
 
@@ -1842,9 +1845,12 @@
     return;
   }
 
+  const GCNSubtarget *ST = static_cast<const GCNSubtarget *>(Subtarget);
+  const SIRegisterInfo *TRI = ST->getRegisterInfo();
+
   bool UseSCCBr = isCBranchSCC(N) && isUniformBr(N);
   unsigned BrOp = UseSCCBr ? AMDGPU::S_CBRANCH_SCC1 : AMDGPU::S_CBRANCH_VCCNZ;
-  unsigned CondReg = UseSCCBr ? AMDGPU::SCC : AMDGPU::VCC;
+  unsigned CondReg = UseSCCBr ? (unsigned)AMDGPU::SCC : TRI->getVCC();
   SDLoc SL(N);
 
   if (!UseSCCBr) {
@@ -1861,9 +1867,13 @@
     // the S_AND when is unnecessary. But it would be better to add a separate
     // pass after SIFixSGPRCopies to do the unnecessary S_AND removal, so it
     // catches both cases.
-    Cond = SDValue(CurDAG->getMachineNode(AMDGPU::S_AND_B64, SL, MVT::i1,
-                               CurDAG->getRegister(AMDGPU::EXEC, MVT::i1),
-                               Cond),
+    Cond = SDValue(CurDAG->getMachineNode(ST->isWave32() ? AMDGPU::S_AND_B32
+                                                         : AMDGPU::S_AND_B64,
+                     SL, MVT::i1,
+                     CurDAG->getRegister(ST->isWave32() ? AMDGPU::EXEC_LO
+                                                        : AMDGPU::EXEC,
+                                         MVT::i1),
+                    Cond),
                    0);
   }
 
Index: llvm/trunk/lib/Target/AMDGPU/AMDGPUSubtarget.h
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -1042,6 +1042,10 @@
     return WavefrontSize == 32;
   }
 
+  const TargetRegisterClass *getBoolRC() const {
+    return getRegisterInfo()->getBoolRC();
+  }
+
   /// \returns Maximum number of work groups per compute unit supported by the
   /// subtarget and limited by given \p FlatWorkGroupSize.
   unsigned getMaxWorkGroupsPerCU(unsigned FlatWorkGroupSize) const override {
Index: llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIFrameLowering.cpp
@@ -627,9 +627,11 @@
 
       ScratchExecCopy
         = findScratchNonCalleeSaveRegister(MF, LiveRegs,
-                                           AMDGPU::SReg_64_XEXECRegClass);
+                                           *TRI.getWaveMaskRegClass());
 
-      BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::S_OR_SAVEEXEC_B64),
+      const unsigned OrSaveExec = ST.isWave32() ?
+        AMDGPU::S_OR_SAVEEXEC_B32 : AMDGPU::S_OR_SAVEEXEC_B64;
+      BuildMI(MBB, MBBI, DL, TII->get(OrSaveExec),
               ScratchExecCopy)
         .addImm(-1);
     }
@@ -641,7 +643,9 @@
 
   if (ScratchExecCopy != AMDGPU::NoRegister) {
     // FIXME: Split block and make terminator.
-    BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::S_MOV_B64), AMDGPU::EXEC)
+    unsigned ExecMov = ST.isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
+    unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+    BuildMI(MBB, MBBI, DL, TII->get(ExecMov), Exec)
       .addReg(ScratchExecCopy);
   }
 }
@@ -663,6 +667,7 @@
     if (!Reg.FI.hasValue())
       continue;
 
+    const SIRegisterInfo &TRI = TII->getRegisterInfo();
     if (ScratchExecCopy == AMDGPU::NoRegister) {
       // See emitPrologue
       LivePhysRegs LiveRegs(*ST.getRegisterInfo());
@@ -670,9 +675,12 @@
 
       ScratchExecCopy
         = findScratchNonCalleeSaveRegister(MF, LiveRegs,
-                                           AMDGPU::SReg_64_XEXECRegClass);
+                                           *TRI.getWaveMaskRegClass());
 
-      BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::S_OR_SAVEEXEC_B64), ScratchExecCopy)
+      const unsigned OrSaveExec = ST.isWave32() ?
+      AMDGPU::S_OR_SAVEEXEC_B32 : AMDGPU::S_OR_SAVEEXEC_B64;
+
+      BuildMI(MBB, MBBI, DL, TII->get(OrSaveExec), ScratchExecCopy)
         .addImm(-1);
     }
 
@@ -683,7 +691,9 @@
 
   if (ScratchExecCopy != AMDGPU::NoRegister) {
     // FIXME: Split block and make terminator.
-    BuildMI(MBB, MBBI, DL, TII->get(AMDGPU::S_MOV_B64), AMDGPU::EXEC)
+    unsigned ExecMov = ST.isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
+    unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+    BuildMI(MBB, MBBI, DL, TII->get(ExecMov), Exec)
       .addReg(ScratchExecCopy);
   }
 
Index: llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2924,12 +2924,16 @@
   int Offset,
   bool UseGPRIdxMode,
   bool IsIndirectSrc) {
+  MachineFunction *MF = OrigBB.getParent();
+  const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
   MachineBasicBlock::iterator I = LoopBB.begin();
 
-  unsigned PhiExec = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-  unsigned NewExec = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  const TargetRegisterClass *BoolRC = TRI->getBoolRC();
+  unsigned PhiExec = MRI.createVirtualRegister(BoolRC);
+  unsigned NewExec = MRI.createVirtualRegister(BoolRC);
   unsigned CurrentIdxReg = MRI.createVirtualRegister(&AMDGPU::SGPR_32RegClass);
-  unsigned CondReg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  unsigned CondReg = MRI.createVirtualRegister(BoolRC);
 
   BuildMI(LoopBB, I, DL, TII->get(TargetOpcode::PHI), PhiReg)
     .addReg(InitReg)
@@ -2953,7 +2957,9 @@
     .addReg(IdxReg.getReg(), 0, IdxReg.getSubReg());
 
   // Update EXEC, save the original EXEC value to VCC.
-  BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_AND_SAVEEXEC_B64), NewExec)
+  BuildMI(LoopBB, I, DL, TII->get(ST.isWave32() ? AMDGPU::S_AND_SAVEEXEC_B32
+                                                : AMDGPU::S_AND_SAVEEXEC_B64),
+          NewExec)
     .addReg(CondReg, RegState::Kill);
 
   MRI.setSimpleHint(NewExec, CondReg);
@@ -2988,10 +2994,12 @@
   }
 
   // Update EXEC, switch all done bits to 0 and all todo bits to 1.
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
   MachineInstr *InsertPt =
-    BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_XOR_B64_term), AMDGPU::EXEC)
-    .addReg(AMDGPU::EXEC)
-    .addReg(NewExec);
+    BuildMI(LoopBB, I, DL, TII->get(ST.isWave32() ? AMDGPU::S_XOR_B32_term
+                                                  : AMDGPU::S_XOR_B64_term), Exec)
+      .addReg(Exec)
+      .addReg(NewExec);
 
   // XXX - s_xor_b64 sets scc to 1 if the result is nonzero, so can we use
   // s_cbranch_scc0?
@@ -3017,19 +3025,24 @@
                                                   bool UseGPRIdxMode,
                                                   bool IsIndirectSrc) {
   MachineFunction *MF = MBB.getParent();
+  const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
   MachineRegisterInfo &MRI = MF->getRegInfo();
   const DebugLoc &DL = MI.getDebugLoc();
   MachineBasicBlock::iterator I(&MI);
 
+  const auto *BoolXExecRC = TRI->getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
   unsigned DstReg = MI.getOperand(0).getReg();
-  unsigned SaveExec = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-  unsigned TmpExec = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+  unsigned SaveExec = MRI.createVirtualRegister(BoolXExecRC);
+  unsigned TmpExec = MRI.createVirtualRegister(BoolXExecRC);
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+  unsigned MovExecOpc = ST.isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
 
   BuildMI(MBB, I, DL, TII->get(TargetOpcode::IMPLICIT_DEF), TmpExec);
 
   // Save the EXEC mask
-  BuildMI(MBB, I, DL, TII->get(AMDGPU::S_MOV_B64), SaveExec)
-    .addReg(AMDGPU::EXEC);
+  BuildMI(MBB, I, DL, TII->get(MovExecOpc), SaveExec)
+    .addReg(Exec);
 
   // To insert the loop we need to split the block. Move everything after this
   // point to a new block, and insert a new empty block between the two.
@@ -3057,7 +3070,7 @@
                                       Offset, UseGPRIdxMode, IsIndirectSrc);
 
   MachineBasicBlock::iterator First = RemainderBB->begin();
-  BuildMI(*RemainderBB, First, DL, TII->get(AMDGPU::S_MOV_B64), AMDGPU::EXEC)
+  BuildMI(*RemainderBB, First, DL, TII->get(MovExecOpc), Exec)
     .addReg(SaveExec);
 
   return InsPt;
@@ -3349,6 +3362,9 @@
   case AMDGPU::S_ADD_U64_PSEUDO:
   case AMDGPU::S_SUB_U64_PSEUDO: {
     MachineRegisterInfo &MRI = BB->getParent()->getRegInfo();
+    const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+    const SIRegisterInfo *TRI = ST.getRegisterInfo();
+    const TargetRegisterClass *BoolRC = TRI->getBoolRC();
     const DebugLoc &DL = MI.getDebugLoc();
 
     MachineOperand &Dest = MI.getOperand(0);
@@ -3359,17 +3375,17 @@
     unsigned DestSub1 = MRI.createVirtualRegister(&AMDGPU::SReg_32_XM0RegClass);
 
     MachineOperand Src0Sub0 = TII->buildExtractSubRegOrImm(MI, MRI,
-     Src0, &AMDGPU::SReg_64RegClass, AMDGPU::sub0,
+     Src0, BoolRC, AMDGPU::sub0,
      &AMDGPU::SReg_32_XM0RegClass);
     MachineOperand Src0Sub1 = TII->buildExtractSubRegOrImm(MI, MRI,
-      Src0, &AMDGPU::SReg_64RegClass, AMDGPU::sub1,
+      Src0, BoolRC, AMDGPU::sub1,
       &AMDGPU::SReg_32_XM0RegClass);
 
     MachineOperand Src1Sub0 = TII->buildExtractSubRegOrImm(MI, MRI,
-      Src1, &AMDGPU::SReg_64RegClass, AMDGPU::sub0,
+      Src1, BoolRC, AMDGPU::sub0,
       &AMDGPU::SReg_32_XM0RegClass);
     MachineOperand Src1Sub1 = TII->buildExtractSubRegOrImm(MI, MRI,
-      Src1, &AMDGPU::SReg_64RegClass, AMDGPU::sub1,
+      Src1, BoolRC, AMDGPU::sub1,
       &AMDGPU::SReg_32_XM0RegClass);
 
     bool IsAdd = (MI.getOpcode() == AMDGPU::S_ADD_U64_PSEUDO);
@@ -3405,6 +3421,14 @@
     MI.eraseFromParent();
     return BB;
 
+  case AMDGPU::SI_INIT_EXEC_LO:
+    // This should be before all vector instructions.
+    BuildMI(*BB, &*BB->begin(), MI.getDebugLoc(), TII->get(AMDGPU::S_MOV_B32),
+            AMDGPU::EXEC_LO)
+        .addImm(MI.getOperand(0).getImm());
+    MI.eraseFromParent();
+    return BB;
+
   case AMDGPU::SI_INIT_EXEC_FROM_INPUT: {
     // Extract the thread count from an SGPR input and set EXEC accordingly.
     // Since BFM can't shift by 64, handle that case with CMP + CMOV.
@@ -3438,18 +3462,23 @@
     (void)Found;
 
     // This should be before all vector instructions.
+    unsigned Mask = (getSubtarget()->getWavefrontSize() << 1) - 1;
+    bool isWave32 = getSubtarget()->isWave32();
+    unsigned Exec = isWave32 ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
     BuildMI(*BB, FirstMI, DebugLoc(), TII->get(AMDGPU::S_BFE_U32), CountReg)
         .addReg(InputReg)
-        .addImm((MI.getOperand(1).getImm() & 0x7f) | 0x70000);
-    BuildMI(*BB, FirstMI, DebugLoc(), TII->get(AMDGPU::S_BFM_B64),
-            AMDGPU::EXEC)
+        .addImm((MI.getOperand(1).getImm() & Mask) | 0x70000);
+    BuildMI(*BB, FirstMI, DebugLoc(),
+            TII->get(isWave32 ? AMDGPU::S_BFM_B32 : AMDGPU::S_BFM_B64),
+            Exec)
         .addReg(CountReg)
         .addImm(0);
     BuildMI(*BB, FirstMI, DebugLoc(), TII->get(AMDGPU::S_CMP_EQ_U32))
         .addReg(CountReg, RegState::Kill)
-        .addImm(64);
-    BuildMI(*BB, FirstMI, DebugLoc(), TII->get(AMDGPU::S_CMOV_B64),
-            AMDGPU::EXEC)
+        .addImm(getSubtarget()->getWavefrontSize());
+    BuildMI(*BB, FirstMI, DebugLoc(),
+            TII->get(isWave32 ? AMDGPU::S_CMOV_B32 : AMDGPU::S_CMOV_B64),
+            Exec)
         .addImm(-1);
     MI.eraseFromParent();
     return BB;
@@ -3480,6 +3509,8 @@
     return splitKillBlock(MI, BB);
   case AMDGPU::V_CNDMASK_B64_PSEUDO: {
     MachineRegisterInfo &MRI = BB->getParent()->getRegInfo();
+    const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+    const SIRegisterInfo *TRI = ST.getRegisterInfo();
 
     unsigned Dst = MI.getOperand(0).getReg();
     unsigned Src0 = MI.getOperand(1).getReg();
@@ -3489,7 +3520,8 @@
 
     unsigned DstLo = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
     unsigned DstHi = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
-    unsigned SrcCondCopy = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+    const auto *CondRC = TRI->getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
+    unsigned SrcCondCopy = MRI.createVirtualRegister(CondRC);
 
     BuildMI(*BB, MI, DL, TII->get(AMDGPU::COPY), SrcCondCopy)
       .addReg(SrcCond);
@@ -3567,7 +3599,9 @@
 
     auto I = BuildMI(*BB, MI, DL, TII->get(Opc), MI.getOperand(0).getReg());
     if (TII->isVOP3(*I)) {
-      I.addReg(AMDGPU::VCC, RegState::Define);
+      const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+      const SIRegisterInfo *TRI = ST.getRegisterInfo();
+      I.addReg(TRI->getVCC(), RegState::Define);
     }
     I.add(MI.getOperand(1))
      .add(MI.getOperand(2));
@@ -10003,6 +10037,7 @@
 void SITargetLowering::finalizeLowering(MachineFunction &MF) const {
   MachineRegisterInfo &MRI = MF.getRegInfo();
   SIMachineFunctionInfo *Info = MF.getInfo<SIMachineFunctionInfo>();
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   const SIRegisterInfo *TRI = Subtarget->getRegisterInfo();
 
   if (Info->isEntryFunction()) {
@@ -10030,6 +10065,25 @@
 
   Info->limitOccupancy(MF);
 
+  if (ST.isWave32() && !MF.empty()) {
+    // Add VCC_HI def because many instructions marked as imp-use VCC where
+    // we may only define VCC_LO. If nothing defines VCC_HI we may end up
+    // having a use of undef.
+
+    const SIInstrInfo *TII = ST.getInstrInfo();
+    DebugLoc DL;
+
+    MachineBasicBlock &MBB = MF.front();
+    MachineBasicBlock::iterator I = MBB.getFirstNonDebugInstr();
+    BuildMI(MBB, I, DL, TII->get(TargetOpcode::IMPLICIT_DEF), AMDGPU::VCC_HI);
+
+    for (auto &MBB : MF) {
+      for (auto &MI : MBB) {
+        TII->fixImplicitOperands(MI);
+      }
+    }
+  }
+
   TargetLoweringBase::finalizeLowering(MF);
 }
 
Index: llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIInsertSkips.cpp
@@ -271,6 +271,9 @@
     break;
   }
   case AMDGPU::SI_KILL_I1_TERMINATOR: {
+    const MachineFunction *MF = MI.getParent()->getParent();
+    const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+    unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
     const MachineOperand &Op = MI.getOperand(0);
     int64_t KillVal = MI.getOperand(1).getImm();
     assert(KillVal == 0 || KillVal == -1);
@@ -281,14 +284,17 @@
       assert(Imm == 0 || Imm == -1);
 
       if (Imm == KillVal)
-        BuildMI(MBB, &MI, DL, TII->get(AMDGPU::S_MOV_B64), AMDGPU::EXEC)
+        BuildMI(MBB, &MI, DL, TII->get(ST.isWave32() ? AMDGPU::S_MOV_B32
+                                                     : AMDGPU::S_MOV_B64), Exec)
           .addImm(0);
       break;
     }
 
     unsigned Opcode = KillVal ? AMDGPU::S_ANDN2_B64 : AMDGPU::S_AND_B64;
-    BuildMI(MBB, &MI, DL, TII->get(Opcode), AMDGPU::EXEC)
-        .addReg(AMDGPU::EXEC)
+    if (ST.isWave32())
+      Opcode = KillVal ? AMDGPU::S_ANDN2_B32 : AMDGPU::S_AND_B32;
+    BuildMI(MBB, &MI, DL, TII->get(Opcode), Exec)
+        .addReg(Exec)
         .add(Op);
     break;
   }
@@ -337,9 +343,11 @@
   // S_CBRANCH_EXEC[N]Z
   bool Changed = false;
   MachineBasicBlock &MBB = *MI.getParent();
-  const unsigned CondReg = AMDGPU::VCC;
-  const unsigned ExecReg = AMDGPU::EXEC;
-  const unsigned And = AMDGPU::S_AND_B64;
+  const GCNSubtarget &ST = MBB.getParent()->getSubtarget<GCNSubtarget>();
+  const bool IsWave32 = ST.isWave32();
+  const unsigned CondReg = TRI->getVCC();
+  const unsigned ExecReg = IsWave32 ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+  const unsigned And = IsWave32 ? AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64;
 
   MachineBasicBlock::reverse_iterator A = MI.getReverseIterator(),
                                       E = MBB.rend();
Index: llvm/trunk/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -1426,9 +1426,9 @@
       // bit is updated, so we can restore the bit by reading the value of
       // vcc and then writing it back to the register.
       BuildMI(Block, Inst, Inst.getDebugLoc(),
-              TII->get(AMDGPU::S_MOV_B64),
-              AMDGPU::VCC)
-          .addReg(AMDGPU::VCC);
+              TII->get(ST->isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64),
+              TRI->getVCC())
+          .addReg(TRI->getVCC());
       VCCZBugHandledSet.insert(&Inst);
       Modified = true;
     }
Index: llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.h
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.h
+++ llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.h
@@ -942,6 +942,8 @@
   /// Return -1 if the target-specific opcode for the pseudo instruction does
   /// not exist. If Opcode is not a pseudo instruction, this is identity.
   int pseudoToMCOpcode(int Opcode) const;
+
+  void fixImplicitOperands(MachineInstr &MI) const;
 };
 
 /// \brief Returns true if a reg:subreg pair P has a TRC class
Index: llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -527,6 +527,21 @@
       return;
     }
 
+    if (DestReg == AMDGPU::VCC_LO) {
+      if (AMDGPU::SReg_32RegClass.contains(SrcReg)) {
+        BuildMI(MBB, MI, DL, get(AMDGPU::S_MOV_B32), AMDGPU::VCC_LO)
+          .addReg(SrcReg, getKillRegState(KillSrc));
+      } else {
+        // FIXME: Hack until VReg_1 removed.
+        assert(AMDGPU::VGPR_32RegClass.contains(SrcReg));
+        BuildMI(MBB, MI, DL, get(AMDGPU::V_CMP_NE_U32_e32))
+          .addImm(0)
+          .addReg(SrcReg, getKillRegState(KillSrc));
+      }
+
+      return;
+    }
+
     if (!AMDGPU::SReg_32RegClass.contains(SrcReg)) {
       reportIllegalCopy(this, MBB, MI, DL, DestReg, SrcReg, KillSrc);
       return;
@@ -698,11 +713,15 @@
                                      unsigned TrueReg,
                                      unsigned FalseReg) const {
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
+  MachineFunction *MF = MBB.getParent();
+  const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+  const TargetRegisterClass *BoolXExecRC =
+    RI.getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
   assert(MRI.getRegClass(DstReg) == &AMDGPU::VGPR_32RegClass &&
          "Not a VGPR32 reg");
 
   if (Cond.size() == 1) {
-    unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+    unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
     BuildMI(MBB, I, DL, get(AMDGPU::COPY), SReg)
       .add(Cond[0]);
     BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -715,8 +734,9 @@
     assert(Cond[0].isImm() && "Cond[0] is not an immediate");
     switch (Cond[0].getImm()) {
     case SIInstrInfo::SCC_TRUE: {
-      unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-      BuildMI(MBB, I, DL, get(AMDGPU::S_CSELECT_B64), SReg)
+      unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
+      BuildMI(MBB, I, DL, get(ST.isWave32() ? AMDGPU::S_CSELECT_B32
+                                            : AMDGPU::S_CSELECT_B64), SReg)
         .addImm(-1)
         .addImm(0);
       BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -728,8 +748,9 @@
       break;
     }
     case SIInstrInfo::SCC_FALSE: {
-      unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-      BuildMI(MBB, I, DL, get(AMDGPU::S_CSELECT_B64), SReg)
+      unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
+      BuildMI(MBB, I, DL, get(ST.isWave32() ? AMDGPU::S_CSELECT_B32
+                                            : AMDGPU::S_CSELECT_B64), SReg)
         .addImm(0)
         .addImm(-1);
       BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -743,7 +764,7 @@
     case SIInstrInfo::VCCNZ: {
       MachineOperand RegOp = Cond[1];
       RegOp.setImplicit(false);
-      unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+      unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
       BuildMI(MBB, I, DL, get(AMDGPU::COPY), SReg)
         .add(RegOp);
       BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -757,7 +778,7 @@
     case SIInstrInfo::VCCZ: {
       MachineOperand RegOp = Cond[1];
       RegOp.setImplicit(false);
-      unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+      unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
       BuildMI(MBB, I, DL, get(AMDGPU::COPY), SReg)
         .add(RegOp);
       BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -769,11 +790,13 @@
       break;
     }
     case SIInstrInfo::EXECNZ: {
-      unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-      unsigned SReg2 = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-      BuildMI(MBB, I, DL, get(AMDGPU::S_OR_SAVEEXEC_B64), SReg2)
+      unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
+      unsigned SReg2 = MRI.createVirtualRegister(RI.getBoolRC());
+      BuildMI(MBB, I, DL, get(ST.isWave32() ? AMDGPU::S_OR_SAVEEXEC_B32
+                                            : AMDGPU::S_OR_SAVEEXEC_B64), SReg2)
         .addImm(0);
-      BuildMI(MBB, I, DL, get(AMDGPU::S_CSELECT_B64), SReg)
+      BuildMI(MBB, I, DL, get(ST.isWave32() ? AMDGPU::S_CSELECT_B32
+                                            : AMDGPU::S_CSELECT_B64), SReg)
         .addImm(-1)
         .addImm(0);
       BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -785,11 +808,13 @@
       break;
     }
     case SIInstrInfo::EXECZ: {
-      unsigned SReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-      unsigned SReg2 = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-      BuildMI(MBB, I, DL, get(AMDGPU::S_OR_SAVEEXEC_B64), SReg2)
+      unsigned SReg = MRI.createVirtualRegister(BoolXExecRC);
+      unsigned SReg2 = MRI.createVirtualRegister(RI.getBoolRC());
+      BuildMI(MBB, I, DL, get(ST.isWave32() ? AMDGPU::S_OR_SAVEEXEC_B32
+                                            : AMDGPU::S_OR_SAVEEXEC_B64), SReg2)
         .addImm(0);
-      BuildMI(MBB, I, DL, get(AMDGPU::S_CSELECT_B64), SReg)
+      BuildMI(MBB, I, DL, get(ST.isWave32() ? AMDGPU::S_CSELECT_B32
+                                            : AMDGPU::S_CSELECT_B64), SReg)
         .addImm(0)
         .addImm(-1);
       BuildMI(MBB, I, DL, get(AMDGPU::V_CNDMASK_B32_e64), DstReg)
@@ -814,7 +839,7 @@
                                const DebugLoc &DL,
                                unsigned SrcReg, int Value) const {
   MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo();
-  unsigned Reg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  unsigned Reg = MRI.createVirtualRegister(RI.getBoolRC());
   BuildMI(*MBB, I, DL, get(AMDGPU::V_CMP_EQ_I32_e64), Reg)
     .addImm(Value)
     .addReg(SrcReg);
@@ -827,7 +852,7 @@
                                const DebugLoc &DL,
                                unsigned SrcReg, int Value) const {
   MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo();
-  unsigned Reg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  unsigned Reg = MRI.createVirtualRegister(RI.getBoolRC());
   BuildMI(*MBB, I, DL, get(AMDGPU::V_CMP_NE_I32_e64), Reg)
     .addImm(Value)
     .addReg(SrcReg);
@@ -1208,18 +1233,42 @@
     MI.setDesc(get(AMDGPU::S_MOV_B64));
     break;
 
+  case AMDGPU::S_MOV_B32_term:
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(get(AMDGPU::S_MOV_B32));
+    break;
+
   case AMDGPU::S_XOR_B64_term:
     // This is only a terminator to get the correct spill code placement during
     // register allocation.
     MI.setDesc(get(AMDGPU::S_XOR_B64));
     break;
 
+  case AMDGPU::S_XOR_B32_term:
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(get(AMDGPU::S_XOR_B32));
+    break;
+
+  case AMDGPU::S_OR_B32_term:
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(get(AMDGPU::S_OR_B32));
+    break;
+
   case AMDGPU::S_ANDN2_B64_term:
     // This is only a terminator to get the correct spill code placement during
     // register allocation.
     MI.setDesc(get(AMDGPU::S_ANDN2_B64));
     break;
 
+  case AMDGPU::S_ANDN2_B32_term:
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(get(AMDGPU::S_ANDN2_B32));
+    break;
+
   case AMDGPU::V_MOV_B64_PSEUDO: {
     unsigned Dst = MI.getOperand(0).getReg();
     unsigned DstLo = RI.getSubReg(Dst, AMDGPU::sub0);
@@ -1249,24 +1298,28 @@
     break;
   }
   case AMDGPU::V_SET_INACTIVE_B32: {
-    BuildMI(MBB, MI, DL, get(AMDGPU::S_NOT_B64), AMDGPU::EXEC)
-      .addReg(AMDGPU::EXEC);
+    unsigned NotOpc = ST.isWave32() ? AMDGPU::S_NOT_B32 : AMDGPU::S_NOT_B64;
+    unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+    BuildMI(MBB, MI, DL, get(NotOpc), Exec)
+      .addReg(Exec);
     BuildMI(MBB, MI, DL, get(AMDGPU::V_MOV_B32_e32), MI.getOperand(0).getReg())
       .add(MI.getOperand(2));
-    BuildMI(MBB, MI, DL, get(AMDGPU::S_NOT_B64), AMDGPU::EXEC)
-      .addReg(AMDGPU::EXEC);
+    BuildMI(MBB, MI, DL, get(NotOpc), Exec)
+      .addReg(Exec);
     MI.eraseFromParent();
     break;
   }
   case AMDGPU::V_SET_INACTIVE_B64: {
-    BuildMI(MBB, MI, DL, get(AMDGPU::S_NOT_B64), AMDGPU::EXEC)
-      .addReg(AMDGPU::EXEC);
+    unsigned NotOpc = ST.isWave32() ? AMDGPU::S_NOT_B32 : AMDGPU::S_NOT_B64;
+    unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+    BuildMI(MBB, MI, DL, get(NotOpc), Exec)
+      .addReg(Exec);
     MachineInstr *Copy = BuildMI(MBB, MI, DL, get(AMDGPU::V_MOV_B64_PSEUDO),
                                  MI.getOperand(0).getReg())
       .add(MI.getOperand(2));
     expandPostRAPseudo(*Copy);
-    BuildMI(MBB, MI, DL, get(AMDGPU::S_NOT_B64), AMDGPU::EXEC)
-      .addReg(AMDGPU::EXEC);
+    BuildMI(MBB, MI, DL, get(NotOpc), Exec)
+      .addReg(Exec);
     MI.eraseFromParent();
     break;
   }
@@ -1330,13 +1383,14 @@
   case AMDGPU::ENTER_WWM: {
     // This only gets its own opcode so that SIPreAllocateWWMRegs can tell when
     // WWM is entered.
-    MI.setDesc(get(AMDGPU::S_OR_SAVEEXEC_B64));
+    MI.setDesc(get(ST.isWave32() ? AMDGPU::S_OR_SAVEEXEC_B32
+                                 : AMDGPU::S_OR_SAVEEXEC_B64));
     break;
   }
   case AMDGPU::EXIT_WWM: {
     // This only gets its own opcode so that SIPreAllocateWWMRegs can tell when
     // WWM is exited.
-    MI.setDesc(get(AMDGPU::S_MOV_B64));
+    MI.setDesc(get(ST.isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64));
     break;
   }
   case TargetOpcode::BUNDLE: {
@@ -1699,6 +1753,10 @@
     case AMDGPU::S_MOV_B64_term:
     case AMDGPU::S_XOR_B64_term:
     case AMDGPU::S_ANDN2_B64_term:
+    case AMDGPU::S_MOV_B32_term:
+    case AMDGPU::S_XOR_B32_term:
+    case AMDGPU::S_OR_B32_term:
+    case AMDGPU::S_ANDN2_B32_term:
       break;
     case AMDGPU::SI_IF:
     case AMDGPU::SI_ELSE:
@@ -1978,6 +2036,7 @@
       .addReg(FalseReg, 0, SubIdx)
       .addReg(TrueReg, 0, SubIdx);
     preserveCondRegFlags(Select->getOperand(3), Cond[1]);
+    fixImplicitOperands(*Select);
 
     MIB.addReg(DstElt)
        .addImm(SubIdx);
@@ -2782,7 +2841,8 @@
     // dst
     Inst32.add(MI.getOperand(0));
   } else {
-    assert(MI.getOperand(0).getReg() == AMDGPU::VCC &&
+    assert(((MI.getOperand(0).getReg() == AMDGPU::VCC) ||
+            (MI.getOperand(0).getReg() == AMDGPU::VCC_LO)) &&
            "Unexpected case");
   }
 
@@ -2850,6 +2910,8 @@
 
     switch (MO.getReg()) {
     case AMDGPU::VCC:
+    case AMDGPU::VCC_LO:
+    case AMDGPU::VCC_HI:
     case AMDGPU::M0:
     case AMDGPU::FLAT_SCR:
       return MO.getReg();
@@ -3795,6 +3857,7 @@
 
   Src1.ChangeToRegister(Src0Reg, false, false, Src0Kill);
   Src1.setSubReg(Src0SubReg);
+  fixImplicitOperands(MI);
 }
 
 // Legalize VOP3 operands. All operand types are supported for any operand
@@ -3971,15 +4034,27 @@
 emitLoadSRsrcFromVGPRLoop(const SIInstrInfo &TII, MachineRegisterInfo &MRI,
                           MachineBasicBlock &OrigBB, MachineBasicBlock &LoopBB,
                           const DebugLoc &DL, MachineOperand &Rsrc) {
+  MachineFunction &MF = *OrigBB.getParent();
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+  unsigned SaveExecOpc =
+      ST.isWave32() ? AMDGPU::S_AND_SAVEEXEC_B32 : AMDGPU::S_AND_SAVEEXEC_B64;
+  unsigned XorTermOpc =
+      ST.isWave32() ? AMDGPU::S_XOR_B32_term : AMDGPU::S_XOR_B64_term;
+  unsigned AndOpc =
+      ST.isWave32() ? AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64;
+  const auto *BoolXExecRC = TRI->getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
+
   MachineBasicBlock::iterator I = LoopBB.begin();
 
   unsigned VRsrc = Rsrc.getReg();
   unsigned VRsrcUndef = getUndefRegState(Rsrc.isUndef());
 
-  unsigned SaveExec = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-  unsigned CondReg0 = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-  unsigned CondReg1 = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-  unsigned AndCond = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  unsigned SaveExec = MRI.createVirtualRegister(BoolXExecRC);
+  unsigned CondReg0 = MRI.createVirtualRegister(BoolXExecRC);
+  unsigned CondReg1 = MRI.createVirtualRegister(BoolXExecRC);
+  unsigned AndCond = MRI.createVirtualRegister(BoolXExecRC);
   unsigned SRsrcSub0 = MRI.createVirtualRegister(&AMDGPU::SGPR_32RegClass);
   unsigned SRsrcSub1 = MRI.createVirtualRegister(&AMDGPU::SGPR_32RegClass);
   unsigned SRsrcSub2 = MRI.createVirtualRegister(&AMDGPU::SGPR_32RegClass);
@@ -4017,22 +4092,22 @@
   BuildMI(LoopBB, I, DL, TII.get(AMDGPU::V_CMP_EQ_U64_e64), CondReg1)
       .addReg(SRsrc, 0, AMDGPU::sub2_sub3)
       .addReg(VRsrc, 0, AMDGPU::sub2_sub3);
-  BuildMI(LoopBB, I, DL, TII.get(AMDGPU::S_AND_B64), AndCond)
+  BuildMI(LoopBB, I, DL, TII.get(AndOpc), AndCond)
       .addReg(CondReg0)
       .addReg(CondReg1);
 
   MRI.setSimpleHint(SaveExec, AndCond);
 
   // Update EXEC to matching lanes, saving original to SaveExec.
-  BuildMI(LoopBB, I, DL, TII.get(AMDGPU::S_AND_SAVEEXEC_B64), SaveExec)
+  BuildMI(LoopBB, I, DL, TII.get(SaveExecOpc), SaveExec)
       .addReg(AndCond, RegState::Kill);
 
   // The original instruction is here; we insert the terminators after it.
   I = LoopBB.end();
 
   // Update EXEC, switch all done bits to 0 and all todo bits to 1.
-  BuildMI(LoopBB, I, DL, TII.get(AMDGPU::S_XOR_B64_term), AMDGPU::EXEC)
-      .addReg(AMDGPU::EXEC)
+  BuildMI(LoopBB, I, DL, TII.get(XorTermOpc), Exec)
+      .addReg(Exec)
       .addReg(SaveExec);
   BuildMI(LoopBB, I, DL, TII.get(AMDGPU::S_CBRANCH_EXECNZ)).addMBB(&LoopBB);
 }
@@ -4043,15 +4118,19 @@
                               MachineOperand &Rsrc, MachineDominatorTree *MDT) {
   MachineBasicBlock &MBB = *MI.getParent();
   MachineFunction &MF = *MBB.getParent();
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  const SIRegisterInfo *TRI = ST.getRegisterInfo();
   MachineRegisterInfo &MRI = MF.getRegInfo();
   MachineBasicBlock::iterator I(&MI);
   const DebugLoc &DL = MI.getDebugLoc();
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+  unsigned MovExecOpc = ST.isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
+  const auto *BoolXExecRC = TRI->getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
 
-  unsigned SaveExec = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+  unsigned SaveExec = MRI.createVirtualRegister(BoolXExecRC);
 
   // Save the EXEC mask
-  BuildMI(MBB, I, DL, TII.get(AMDGPU::S_MOV_B64), SaveExec)
-      .addReg(AMDGPU::EXEC);
+  BuildMI(MBB, I, DL, TII.get(MovExecOpc), SaveExec).addReg(Exec);
 
   // Killed uses in the instruction we are waterfalling around will be
   // incorrect due to the added control-flow.
@@ -4100,8 +4179,7 @@
 
   // Restore the EXEC mask
   MachineBasicBlock::iterator First = RemainderBB->begin();
-  BuildMI(*RemainderBB, First, DL, TII.get(AMDGPU::S_MOV_B64), AMDGPU::EXEC)
-      .addReg(SaveExec);
+  BuildMI(*RemainderBB, First, DL, TII.get(MovExecOpc), Exec).addReg(SaveExec);
 }
 
 // Extract pointer from Rsrc and return a zero-value Rsrc replacement.
@@ -4332,14 +4410,16 @@
 
       // NewVaddrLo = RsrcPtr:sub0 + VAddr:sub0
       DebugLoc DL = MI.getDebugLoc();
-      BuildMI(MBB, MI, DL, get(AMDGPU::V_ADD_I32_e32), NewVAddrLo)
+      fixImplicitOperands(*
+        BuildMI(MBB, MI, DL, get(AMDGPU::V_ADD_I32_e32), NewVAddrLo)
           .addReg(RsrcPtr, 0, AMDGPU::sub0)
-          .addReg(VAddr->getReg(), 0, AMDGPU::sub0);
+          .addReg(VAddr->getReg(), 0, AMDGPU::sub0));
 
       // NewVaddrHi = RsrcPtr:sub1 + VAddr:sub1
-      BuildMI(MBB, MI, DL, get(AMDGPU::V_ADDC_U32_e32), NewVAddrHi)
+      fixImplicitOperands(*
+        BuildMI(MBB, MI, DL, get(AMDGPU::V_ADDC_U32_e32), NewVAddrHi)
           .addReg(RsrcPtr, 0, AMDGPU::sub1)
-          .addReg(VAddr->getReg(), 0, AMDGPU::sub1);
+          .addReg(VAddr->getReg(), 0, AMDGPU::sub1));
 
       // NewVaddr = {NewVaddrHi, NewVaddrLo}
       BuildMI(MBB, MI, MI.getDebugLoc(), get(AMDGPU::REG_SEQUENCE), NewVAddr)
@@ -4563,10 +4643,16 @@
     case AMDGPU::S_CBRANCH_SCC0:
     case AMDGPU::S_CBRANCH_SCC1:
       // Clear unused bits of vcc
-      BuildMI(*MBB, Inst, Inst.getDebugLoc(), get(AMDGPU::S_AND_B64),
-              AMDGPU::VCC)
-          .addReg(AMDGPU::EXEC)
-          .addReg(AMDGPU::VCC);
+      if (ST.isWave32())
+        BuildMI(*MBB, Inst, Inst.getDebugLoc(), get(AMDGPU::S_AND_B32),
+                AMDGPU::VCC_LO)
+            .addReg(AMDGPU::EXEC_LO)
+            .addReg(AMDGPU::VCC_LO);
+      else
+        BuildMI(*MBB, Inst, Inst.getDebugLoc(), get(AMDGPU::S_AND_B64),
+                AMDGPU::VCC)
+            .addReg(AMDGPU::EXEC)
+            .addReg(AMDGPU::VCC);
       break;
 
     case AMDGPU::S_BFE_U64:
@@ -4644,6 +4730,7 @@
     }
 
     Inst.addImplicitDefUseOperands(*Inst.getParent()->getParent());
+    fixImplicitOperands(Inst);
 
     if (Opcode == AMDGPU::S_BFE_I32 || Opcode == AMDGPU::S_BFE_U32) {
       const MachineOperand &OffsetWidthOp = Inst.getOperand(2);
@@ -4957,13 +5044,14 @@
 
   MachineBasicBlock &MBB = *Inst.getParent();
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
+  const auto *CarryRC = RI.getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
 
   unsigned FullDestReg = MRI.createVirtualRegister(&AMDGPU::VReg_64RegClass);
   unsigned DestSub0 = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
   unsigned DestSub1 = MRI.createVirtualRegister(&AMDGPU::VGPR_32RegClass);
 
-  unsigned CarryReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-  unsigned DeadCarryReg = MRI.createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+  unsigned CarryReg = MRI.createVirtualRegister(CarryRC);
+  unsigned DeadCarryReg = MRI.createVirtualRegister(CarryRC);
 
   MachineOperand &Dest = Inst.getOperand(0);
   MachineOperand &Src0 = Inst.getOperand(1);
@@ -5661,7 +5749,7 @@
   MachineRegisterInfo &MRI = IfEntry->getParent()->getRegInfo();
 
   if (Branch->getOpcode() == AMDGPU::SI_NON_UNIFORM_BRCOND_PSEUDO) {
-    unsigned DstReg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+    unsigned DstReg = MRI.createVirtualRegister(RI.getBoolRC());
     MachineInstr *SIIF =
         BuildMI(*MF, Branch->getDebugLoc(), get(AMDGPU::SI_IF), DstReg)
             .add(Branch->getOperand(0))
@@ -5688,8 +5776,8 @@
 
   if (Branch->getOpcode() == AMDGPU::SI_NON_UNIFORM_BRCOND_PSEUDO) {
 
-    unsigned DstReg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-    unsigned BackEdgeReg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+    unsigned DstReg = MRI.createVirtualRegister(RI.getBoolRC());
+    unsigned BackEdgeReg = MRI.createVirtualRegister(RI.getBoolRC());
     MachineInstrBuilder HeaderPHIBuilder =
         BuildMI(*(MF), Branch->getDebugLoc(), get(TargetOpcode::PHI), DstReg);
     for (MachineBasicBlock::pred_iterator PI = LoopEntry->pred_begin(),
@@ -5699,7 +5787,7 @@
         HeaderPHIBuilder.addReg(BackEdgeReg);
       } else {
         MachineBasicBlock *PMBB = *PI;
-        unsigned ZeroReg = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+        unsigned ZeroReg = MRI.createVirtualRegister(RI.getBoolRC());
         materializeImmediate(*PMBB, PMBB->getFirstTerminator(), DebugLoc(),
                              ZeroReg, 0);
         HeaderPHIBuilder.addReg(ZeroReg);
@@ -5781,8 +5869,8 @@
     return BuildMI(MBB, I, DL, get(AMDGPU::V_ADD_U32_e64), DestReg);
 
   MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
-  unsigned UnusedCarry = MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
-  MRI.setRegAllocationHint(UnusedCarry, 0, AMDGPU::VCC);
+  unsigned UnusedCarry = MRI.createVirtualRegister(RI.getBoolRC());
+  MRI.setRegAllocationHint(UnusedCarry, 0, RI.getVCC());
 
   return BuildMI(MBB, I, DL, get(AMDGPU::V_ADD_I32_e64), DestReg)
            .addReg(UnusedCarry, RegState::Define | RegState::Dead);
@@ -5809,6 +5897,20 @@
   }
 }
 
+void SIInstrInfo::fixImplicitOperands(MachineInstr &MI) const {
+  MachineBasicBlock *MBB = MI.getParent();
+  MachineFunction *MF = MBB->getParent();
+  const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
+
+  if (!ST.isWave32())
+    return;
+
+  for (auto &Op : MI.implicit_operands()) {
+    if (Op.isReg() && Op.getReg() == AMDGPU::VCC)
+      Op.setReg(AMDGPU::VCC_LO);
+  }
+}
+
 bool SIInstrInfo::isBufferSMRD(const MachineInstr &MI) const {
   if (!isSMRD(MI))
     return false;
Index: llvm/trunk/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
@@ -1144,9 +1144,10 @@
   MachineOperand OffsetLo = createRegOrImm(static_cast<int32_t>(Addr.Offset), MI);
   MachineOperand OffsetHi =
     createRegOrImm(static_cast<int32_t>(Addr.Offset >> 32), MI);
-  unsigned CarryReg = MRI->createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
-  unsigned DeadCarryReg =
-    MRI->createVirtualRegister(&AMDGPU::SReg_64_XEXECRegClass);
+
+  const auto *CarryRC = TRI->getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
+  unsigned CarryReg = MRI->createVirtualRegister(CarryRC);
+  unsigned DeadCarryReg = MRI->createVirtualRegister(CarryRC);
 
   unsigned DestSub0 = MRI->createVirtualRegister(&AMDGPU::VGPR_32RegClass);
   unsigned DestSub1 = MRI->createVirtualRegister(&AMDGPU::VGPR_32RegClass);
Index: llvm/trunk/lib/Target/AMDGPU/SILowerControlFlow.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SILowerControlFlow.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SILowerControlFlow.cpp
@@ -82,6 +82,16 @@
   LiveIntervals *LIS = nullptr;
   MachineRegisterInfo *MRI = nullptr;
 
+  const TargetRegisterClass *BoolRC = nullptr;
+  unsigned AndOpc;
+  unsigned OrOpc;
+  unsigned XorOpc;
+  unsigned MovTermOpc;
+  unsigned Andn2TermOpc;
+  unsigned XorTermrOpc;
+  unsigned OrSaveExecOpc;
+  unsigned Exec;
+
   void emitIf(MachineInstr &MI);
   void emitElse(MachineInstr &MI);
   void emitIfBreak(MachineInstr &MI);
@@ -188,16 +198,16 @@
   // Add an implicit def of exec to discourage scheduling VALU after this which
   // will interfere with trying to form s_and_saveexec_b64 later.
   unsigned CopyReg = SimpleIf ? SaveExecReg
-                       : MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
+                       : MRI->createVirtualRegister(BoolRC);
   MachineInstr *CopyExec =
     BuildMI(MBB, I, DL, TII->get(AMDGPU::COPY), CopyReg)
-    .addReg(AMDGPU::EXEC)
-    .addReg(AMDGPU::EXEC, RegState::ImplicitDefine);
+    .addReg(Exec)
+    .addReg(Exec, RegState::ImplicitDefine);
 
-  unsigned Tmp = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  unsigned Tmp = MRI->createVirtualRegister(BoolRC);
 
   MachineInstr *And =
-    BuildMI(MBB, I, DL, TII->get(AMDGPU::S_AND_B64), Tmp)
+    BuildMI(MBB, I, DL, TII->get(AndOpc), Tmp)
     .addReg(CopyReg)
     .add(Cond);
 
@@ -206,7 +216,7 @@
   MachineInstr *Xor = nullptr;
   if (!SimpleIf) {
     Xor =
-      BuildMI(MBB, I, DL, TII->get(AMDGPU::S_XOR_B64), SaveExecReg)
+      BuildMI(MBB, I, DL, TII->get(XorOpc), SaveExecReg)
       .addReg(Tmp)
       .addReg(CopyReg);
     setImpSCCDefDead(*Xor, ImpDefSCC.isDead());
@@ -215,7 +225,7 @@
   // Use a copy that is a terminator to get correct spill code placement it with
   // fast regalloc.
   MachineInstr *SetExec =
-    BuildMI(MBB, I, DL, TII->get(AMDGPU::S_MOV_B64_term), AMDGPU::EXEC)
+    BuildMI(MBB, I, DL, TII->get(MovTermOpc), Exec)
     .addReg(Tmp, RegState::Kill);
 
   // Insert a pseudo terminator to help keep the verifier happy. This will also
@@ -265,7 +275,7 @@
   // We are running before TwoAddressInstructions, and si_else's operands are
   // tied. In order to correctly tie the registers, split this into a copy of
   // the src like it does.
-  unsigned CopyReg = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  unsigned CopyReg = MRI->createVirtualRegister(BoolRC);
   MachineInstr *CopyExec =
     BuildMI(MBB, Start, DL, TII->get(AMDGPU::COPY), CopyReg)
       .add(MI.getOperand(1)); // Saved EXEC
@@ -273,9 +283,9 @@
   // This must be inserted before phis and any spill code inserted before the
   // else.
   unsigned SaveReg = ExecModified ?
-    MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass) : DstReg;
+    MRI->createVirtualRegister(BoolRC) : DstReg;
   MachineInstr *OrSaveExec =
-    BuildMI(MBB, Start, DL, TII->get(AMDGPU::S_OR_SAVEEXEC_B64), SaveReg)
+    BuildMI(MBB, Start, DL, TII->get(OrSaveExecOpc), SaveReg)
     .addReg(CopyReg);
 
   MachineBasicBlock *DestBB = MI.getOperand(2).getMBB();
@@ -284,8 +294,8 @@
 
   if (ExecModified) {
     MachineInstr *And =
-      BuildMI(MBB, ElsePt, DL, TII->get(AMDGPU::S_AND_B64), DstReg)
-      .addReg(AMDGPU::EXEC)
+      BuildMI(MBB, ElsePt, DL, TII->get(AndOpc), DstReg)
+      .addReg(Exec)
       .addReg(SaveReg);
 
     if (LIS)
@@ -293,8 +303,8 @@
   }
 
   MachineInstr *Xor =
-    BuildMI(MBB, ElsePt, DL, TII->get(AMDGPU::S_XOR_B64_term), AMDGPU::EXEC)
-    .addReg(AMDGPU::EXEC)
+    BuildMI(MBB, ElsePt, DL, TII->get(XorTermrOpc), Exec)
+    .addReg(Exec)
     .addReg(DstReg);
 
   MachineInstr *Branch =
@@ -347,14 +357,14 @@
   // exit" mask.
   MachineInstr *And = nullptr, *Or = nullptr;
   if (!SkipAnding) {
-    And = BuildMI(MBB, &MI, DL, TII->get(AMDGPU::S_AND_B64), Dst)
-             .addReg(AMDGPU::EXEC)
+    And = BuildMI(MBB, &MI, DL, TII->get(AndOpc), Dst)
+             .addReg(Exec)
              .add(MI.getOperand(1));
-    Or = BuildMI(MBB, &MI, DL, TII->get(AMDGPU::S_OR_B64), Dst)
+    Or = BuildMI(MBB, &MI, DL, TII->get(OrOpc), Dst)
              .addReg(Dst)
              .add(MI.getOperand(2));
   } else
-    Or = BuildMI(MBB, &MI, DL, TII->get(AMDGPU::S_OR_B64), Dst)
+    Or = BuildMI(MBB, &MI, DL, TII->get(OrOpc), Dst)
              .add(MI.getOperand(1))
              .add(MI.getOperand(2));
 
@@ -372,8 +382,8 @@
   const DebugLoc &DL = MI.getDebugLoc();
 
   MachineInstr *AndN2 =
-      BuildMI(MBB, &MI, DL, TII->get(AMDGPU::S_ANDN2_B64_term), AMDGPU::EXEC)
-          .addReg(AMDGPU::EXEC)
+      BuildMI(MBB, &MI, DL, TII->get(Andn2TermOpc), Exec)
+          .addReg(Exec)
           .add(MI.getOperand(0));
 
   MachineInstr *Branch =
@@ -394,8 +404,8 @@
 
   MachineBasicBlock::iterator InsPt = MBB.begin();
   MachineInstr *NewMI =
-      BuildMI(MBB, InsPt, DL, TII->get(AMDGPU::S_OR_B64), AMDGPU::EXEC)
-          .addReg(AMDGPU::EXEC)
+      BuildMI(MBB, InsPt, DL, TII->get(OrOpc), Exec)
+          .addReg(Exec)
           .add(MI.getOperand(0));
 
   if (LIS)
@@ -427,13 +437,13 @@
   // does not really modify exec.
   for (auto I = Def->getIterator(); I != MI.getIterator(); ++I)
     if (I->modifiesRegister(AMDGPU::EXEC, TRI) &&
-        !(I->isCopy() && I->getOperand(0).getReg() != AMDGPU::EXEC))
+        !(I->isCopy() && I->getOperand(0).getReg() != Exec))
       return;
 
   for (const auto &SrcOp : Def->explicit_operands())
     if (SrcOp.isReg() && SrcOp.isUse() &&
         (TargetRegisterInfo::isVirtualRegister(SrcOp.getReg()) ||
-        SrcOp.getReg() == AMDGPU::EXEC))
+        SrcOp.getReg() == Exec))
       Src.push_back(SrcOp);
 }
 
@@ -471,6 +481,27 @@
   // This doesn't actually need LiveIntervals, but we can preserve them.
   LIS = getAnalysisIfAvailable<LiveIntervals>();
   MRI = &MF.getRegInfo();
+  BoolRC = TRI->getBoolRC();
+
+  if (ST.isWave32()) {
+    AndOpc = AMDGPU::S_AND_B32;
+    OrOpc = AMDGPU::S_OR_B32;
+    XorOpc = AMDGPU::S_XOR_B32;
+    MovTermOpc = AMDGPU::S_MOV_B32_term;
+    Andn2TermOpc = AMDGPU::S_ANDN2_B32_term;
+    XorTermrOpc = AMDGPU::S_XOR_B32_term;
+    OrSaveExecOpc = AMDGPU::S_OR_SAVEEXEC_B32;
+    Exec = AMDGPU::EXEC_LO;
+  } else {
+    AndOpc = AMDGPU::S_AND_B64;
+    OrOpc = AMDGPU::S_OR_B64;
+    XorOpc = AMDGPU::S_XOR_B64;
+    MovTermOpc = AMDGPU::S_MOV_B64_term;
+    Andn2TermOpc = AMDGPU::S_ANDN2_B64_term;
+    XorTermrOpc = AMDGPU::S_XOR_B64_term;
+    OrSaveExecOpc = AMDGPU::S_OR_SAVEEXEC_B64;
+    Exec = AMDGPU::EXEC;
+  }
 
   MachineFunction::iterator NextBB;
   for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
@@ -507,6 +538,8 @@
 
       case AMDGPU::S_AND_B64:
       case AMDGPU::S_OR_B64:
+      case AMDGPU::S_AND_B32:
+      case AMDGPU::S_OR_B32:
         // Cleanup bit manipulations on exec mask
         combineMasks(MI);
         Last = I;
Index: llvm/trunk/lib/Target/AMDGPU/SILowerI1Copies.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SILowerI1Copies.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SILowerI1Copies.cpp
@@ -7,8 +7,8 @@
 //===----------------------------------------------------------------------===//
 //
 // This pass lowers all occurrences of i1 values (with a vreg_1 register class)
-// to lane masks (64-bit scalar registers). The pass assumes machine SSA form
-// and a wave-level control flow graph.
+// to lane masks (32 / 64-bit scalar registers). The pass assumes machine SSA
+// form and a wave-level control flow graph.
 //
 // Before this pass, values that are semantically i1 and are defined and used
 // within the same basic block are already represented as lane masks in scalar
@@ -50,6 +50,7 @@
   static char ID;
 
 private:
+  bool IsWave32 = false;
   MachineFunction *MF = nullptr;
   MachineDominatorTree *DT = nullptr;
   MachinePostDominatorTree *PDT = nullptr;
@@ -57,6 +58,14 @@
   const GCNSubtarget *ST = nullptr;
   const SIInstrInfo *TII = nullptr;
 
+  unsigned ExecReg;
+  unsigned MovOp;
+  unsigned AndOp;
+  unsigned OrOp;
+  unsigned XorOp;
+  unsigned AndN2Op;
+  unsigned OrN2Op;
+
   DenseSet<unsigned> ConstrainRegs;
 
 public:
@@ -411,8 +420,10 @@
 }
 
 static unsigned createLaneMaskReg(MachineFunction &MF) {
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   MachineRegisterInfo &MRI = MF.getRegInfo();
-  return MRI.createVirtualRegister(&AMDGPU::SReg_64RegClass);
+  return MRI.createVirtualRegister(ST.isWave32() ? &AMDGPU::SReg_32RegClass
+                                                 : &AMDGPU::SReg_64RegClass);
 }
 
 static unsigned insertUndefLaneMask(MachineBasicBlock &MBB) {
@@ -442,13 +453,32 @@
 
   ST = &MF->getSubtarget<GCNSubtarget>();
   TII = ST->getInstrInfo();
+  IsWave32 = ST->isWave32();
+
+  if (IsWave32) {
+    ExecReg = AMDGPU::EXEC_LO;
+    MovOp = AMDGPU::S_MOV_B32;
+    AndOp = AMDGPU::S_AND_B32;
+    OrOp = AMDGPU::S_OR_B32;
+    XorOp = AMDGPU::S_XOR_B32;
+    AndN2Op = AMDGPU::S_ANDN2_B32;
+    OrN2Op = AMDGPU::S_ORN2_B32;
+  } else {
+    ExecReg = AMDGPU::EXEC;
+    MovOp = AMDGPU::S_MOV_B64;
+    AndOp = AMDGPU::S_AND_B64;
+    OrOp = AMDGPU::S_OR_B64;
+    XorOp = AMDGPU::S_XOR_B64;
+    AndN2Op = AMDGPU::S_ANDN2_B64;
+    OrN2Op = AMDGPU::S_ORN2_B64;
+  }
 
   lowerCopiesFromI1();
   lowerPhis();
   lowerCopiesToI1();
 
   for (unsigned Reg : ConstrainRegs)
-    MRI->constrainRegClass(Reg, &AMDGPU::SReg_64_XEXECRegClass);
+    MRI->constrainRegClass(Reg, &AMDGPU::SReg_1_XEXECRegClass);
   ConstrainRegs.clear();
 
   return true;
@@ -518,7 +548,8 @@
 
       LLVM_DEBUG(dbgs() << "Lower PHI: " << MI);
 
-      MRI->setRegClass(DstReg, &AMDGPU::SReg_64RegClass);
+      MRI->setRegClass(DstReg, IsWave32 ? &AMDGPU::SReg_32RegClass
+                                        : &AMDGPU::SReg_64RegClass);
 
       // Collect incoming values.
       for (unsigned i = 1; i < MI.getNumOperands(); i += 2) {
@@ -648,7 +679,8 @@
 
       LLVM_DEBUG(dbgs() << "Lower Other: " << MI);
 
-      MRI->setRegClass(DstReg, &AMDGPU::SReg_64RegClass);
+      MRI->setRegClass(DstReg, IsWave32 ? &AMDGPU::SReg_32RegClass
+                                        : &AMDGPU::SReg_64RegClass);
       if (MI.getOpcode() == AMDGPU::IMPLICIT_DEF)
         continue;
 
@@ -707,7 +739,7 @@
       return false;
   }
 
-  if (MI->getOpcode() != AMDGPU::S_MOV_B64)
+  if (MI->getOpcode() != MovOp)
     return false;
 
   if (!MI->getOperand(1).isImm())
@@ -782,10 +814,10 @@
     if (PrevVal == CurVal) {
       BuildMI(MBB, I, DL, TII->get(AMDGPU::COPY), DstReg).addReg(CurReg);
     } else if (CurVal) {
-      BuildMI(MBB, I, DL, TII->get(AMDGPU::COPY), DstReg).addReg(AMDGPU::EXEC);
+      BuildMI(MBB, I, DL, TII->get(AMDGPU::COPY), DstReg).addReg(ExecReg);
     } else {
-      BuildMI(MBB, I, DL, TII->get(AMDGPU::S_XOR_B64), DstReg)
-          .addReg(AMDGPU::EXEC)
+      BuildMI(MBB, I, DL, TII->get(XorOp), DstReg)
+          .addReg(ExecReg)
           .addImm(-1);
     }
     return;
@@ -798,9 +830,9 @@
       PrevMaskedReg = PrevReg;
     } else {
       PrevMaskedReg = createLaneMaskReg(*MF);
-      BuildMI(MBB, I, DL, TII->get(AMDGPU::S_ANDN2_B64), PrevMaskedReg)
+      BuildMI(MBB, I, DL, TII->get(AndN2Op), PrevMaskedReg)
           .addReg(PrevReg)
-          .addReg(AMDGPU::EXEC);
+          .addReg(ExecReg);
     }
   }
   if (!CurConstant) {
@@ -809,9 +841,9 @@
       CurMaskedReg = CurReg;
     } else {
       CurMaskedReg = createLaneMaskReg(*MF);
-      BuildMI(MBB, I, DL, TII->get(AMDGPU::S_AND_B64), CurMaskedReg)
+      BuildMI(MBB, I, DL, TII->get(AndOp), CurMaskedReg)
           .addReg(CurReg)
-          .addReg(AMDGPU::EXEC);
+          .addReg(ExecReg);
     }
   }
 
@@ -822,12 +854,12 @@
     BuildMI(MBB, I, DL, TII->get(AMDGPU::COPY), DstReg)
         .addReg(PrevMaskedReg);
   } else if (PrevConstant && PrevVal) {
-    BuildMI(MBB, I, DL, TII->get(AMDGPU::S_ORN2_B64), DstReg)
+    BuildMI(MBB, I, DL, TII->get(OrN2Op), DstReg)
         .addReg(CurMaskedReg)
-        .addReg(AMDGPU::EXEC);
+        .addReg(ExecReg);
   } else {
-    BuildMI(MBB, I, DL, TII->get(AMDGPU::S_OR_B64), DstReg)
+    BuildMI(MBB, I, DL, TII->get(OrOp), DstReg)
         .addReg(PrevMaskedReg)
-        .addReg(CurMaskedReg ? CurMaskedReg : (unsigned)AMDGPU::EXEC);
+        .addReg(CurMaskedReg ? CurMaskedReg : ExecReg);
   }
 }
Index: llvm/trunk/lib/Target/AMDGPU/SIOptimizeExecMasking.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIOptimizeExecMasking.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIOptimizeExecMasking.cpp
@@ -56,13 +56,16 @@
 char &llvm::SIOptimizeExecMaskingID = SIOptimizeExecMasking::ID;
 
 /// If \p MI is a copy from exec, return the register copied to.
-static unsigned isCopyFromExec(const MachineInstr &MI) {
+static unsigned isCopyFromExec(const MachineInstr &MI, const GCNSubtarget &ST) {
   switch (MI.getOpcode()) {
   case AMDGPU::COPY:
   case AMDGPU::S_MOV_B64:
-  case AMDGPU::S_MOV_B64_term: {
+  case AMDGPU::S_MOV_B64_term:
+  case AMDGPU::S_MOV_B32:
+  case AMDGPU::S_MOV_B32_term: {
     const MachineOperand &Src = MI.getOperand(1);
-    if (Src.isReg() && Src.getReg() == AMDGPU::EXEC)
+    if (Src.isReg() &&
+        Src.getReg() == (ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC))
       return MI.getOperand(0).getReg();
   }
   }
@@ -71,16 +74,20 @@
 }
 
 /// If \p MI is a copy to exec, return the register copied from.
-static unsigned isCopyToExec(const MachineInstr &MI) {
+static unsigned isCopyToExec(const MachineInstr &MI, const GCNSubtarget &ST) {
   switch (MI.getOpcode()) {
   case AMDGPU::COPY:
-  case AMDGPU::S_MOV_B64: {
+  case AMDGPU::S_MOV_B64:
+  case AMDGPU::S_MOV_B32: {
     const MachineOperand &Dst = MI.getOperand(0);
-    if (Dst.isReg() && Dst.getReg() == AMDGPU::EXEC && MI.getOperand(1).isReg())
+    if (Dst.isReg() &&
+        Dst.getReg() == (ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC) &&
+        MI.getOperand(1).isReg())
       return MI.getOperand(1).getReg();
     break;
   }
   case AMDGPU::S_MOV_B64_term:
+  case AMDGPU::S_MOV_B32_term:
     llvm_unreachable("should have been replaced");
   }
 
@@ -105,6 +112,23 @@
     const MachineOperand &Src2 = MI.getOperand(2);
     if (Src2.isReg() && Src2.getReg() == AMDGPU::EXEC)
       return MI.getOperand(0).getReg();
+    break;
+  }
+  case AMDGPU::S_AND_B32:
+  case AMDGPU::S_OR_B32:
+  case AMDGPU::S_XOR_B32:
+  case AMDGPU::S_ANDN2_B32:
+  case AMDGPU::S_ORN2_B32:
+  case AMDGPU::S_NAND_B32:
+  case AMDGPU::S_NOR_B32:
+  case AMDGPU::S_XNOR_B32: {
+    const MachineOperand &Src1 = MI.getOperand(1);
+    if (Src1.isReg() && Src1.getReg() == AMDGPU::EXEC_LO)
+      return MI.getOperand(0).getReg();
+    const MachineOperand &Src2 = MI.getOperand(2);
+    if (Src2.isReg() && Src2.getReg() == AMDGPU::EXEC_LO)
+      return MI.getOperand(0).getReg();
+    break;
   }
   }
 
@@ -129,6 +153,22 @@
     return AMDGPU::S_NOR_SAVEEXEC_B64;
   case AMDGPU::S_XNOR_B64:
     return AMDGPU::S_XNOR_SAVEEXEC_B64;
+  case AMDGPU::S_AND_B32:
+    return AMDGPU::S_AND_SAVEEXEC_B32;
+  case AMDGPU::S_OR_B32:
+    return AMDGPU::S_OR_SAVEEXEC_B32;
+  case AMDGPU::S_XOR_B32:
+    return AMDGPU::S_XOR_SAVEEXEC_B32;
+  case AMDGPU::S_ANDN2_B32:
+    return AMDGPU::S_ANDN2_SAVEEXEC_B32;
+  case AMDGPU::S_ORN2_B32:
+    return AMDGPU::S_ORN2_SAVEEXEC_B32;
+  case AMDGPU::S_NAND_B32:
+    return AMDGPU::S_NAND_SAVEEXEC_B32;
+  case AMDGPU::S_NOR_B32:
+    return AMDGPU::S_NOR_SAVEEXEC_B32;
+  case AMDGPU::S_XNOR_B32:
+    return AMDGPU::S_XNOR_SAVEEXEC_B32;
   default:
     return AMDGPU::INSTRUCTION_LIST_END;
   }
@@ -139,7 +179,8 @@
 // these is expected per block.
 static bool removeTerminatorBit(const SIInstrInfo &TII, MachineInstr &MI) {
   switch (MI.getOpcode()) {
-  case AMDGPU::S_MOV_B64_term: {
+  case AMDGPU::S_MOV_B64_term:
+  case AMDGPU::S_MOV_B32_term: {
     MI.setDesc(TII.get(AMDGPU::COPY));
     return true;
   }
@@ -149,12 +190,30 @@
     MI.setDesc(TII.get(AMDGPU::S_XOR_B64));
     return true;
   }
+  case AMDGPU::S_XOR_B32_term: {
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(TII.get(AMDGPU::S_XOR_B32));
+    return true;
+  }
+  case AMDGPU::S_OR_B32_term: {
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(TII.get(AMDGPU::S_OR_B32));
+    return true;
+  }
   case AMDGPU::S_ANDN2_B64_term: {
     // This is only a terminator to get the correct spill code placement during
     // register allocation.
     MI.setDesc(TII.get(AMDGPU::S_ANDN2_B64));
     return true;
   }
+  case AMDGPU::S_ANDN2_B32_term: {
+    // This is only a terminator to get the correct spill code placement during
+    // register allocation.
+    MI.setDesc(TII.get(AMDGPU::S_ANDN2_B32));
+    return true;
+  }
   default:
     return false;
   }
@@ -177,6 +236,7 @@
 
 static MachineBasicBlock::reverse_iterator findExecCopy(
   const SIInstrInfo &TII,
+  const GCNSubtarget &ST,
   MachineBasicBlock &MBB,
   MachineBasicBlock::reverse_iterator I,
   unsigned CopyToExec) {
@@ -184,7 +244,7 @@
 
   auto E = MBB.rend();
   for (unsigned N = 0; N <= InstLimit && I != E; ++I, ++N) {
-    unsigned CopyFromExec = isCopyFromExec(*I);
+    unsigned CopyFromExec = isCopyFromExec(*I, ST);
     if (CopyFromExec != AMDGPU::NoRegister)
       return I;
   }
@@ -211,6 +271,7 @@
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   const SIRegisterInfo *TRI = ST.getRegisterInfo();
   const SIInstrInfo *TII = ST.getInstrInfo();
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
 
   // Optimize sequences emitted for control flow lowering. They are originally
   // emitted as the separate operations because spill code may need to be
@@ -229,13 +290,13 @@
     if (I == E)
       continue;
 
-    unsigned CopyToExec = isCopyToExec(*I);
+    unsigned CopyToExec = isCopyToExec(*I, ST);
     if (CopyToExec == AMDGPU::NoRegister)
       continue;
 
     // Scan backwards to find the def.
     auto CopyToExecInst = &*I;
-    auto CopyFromExecInst = findExecCopy(*TII, MBB, I, CopyToExec);
+    auto CopyFromExecInst = findExecCopy(*TII, ST, MBB, I, CopyToExec);
     if (CopyFromExecInst == E) {
       auto PrepareExecInst = std::next(I);
       if (PrepareExecInst == E)
@@ -245,7 +306,7 @@
           isLogicalOpOnExec(*PrepareExecInst) == CopyToExec) {
         LLVM_DEBUG(dbgs() << "Fold exec copy: " << *PrepareExecInst);
 
-        PrepareExecInst->getOperand(0).setReg(AMDGPU::EXEC);
+        PrepareExecInst->getOperand(0).setReg(Exec);
 
         LLVM_DEBUG(dbgs() << "into: " << *PrepareExecInst << '\n');
 
@@ -268,7 +329,7 @@
     for (MachineBasicBlock::iterator J
            = std::next(CopyFromExecInst->getIterator()), JE = I->getIterator();
          J != JE; ++J) {
-      if (SaveExecInst && J->readsRegister(AMDGPU::EXEC, TRI)) {
+      if (SaveExecInst && J->readsRegister(Exec, TRI)) {
         LLVM_DEBUG(dbgs() << "exec read prevents saveexec: " << *J << '\n');
         // Make sure this is inserted after any VALU ops that may have been
         // scheduled in between.
@@ -352,7 +413,7 @@
     CopyToExecInst->eraseFromParent();
 
     for (MachineInstr *OtherInst : OtherUseInsts) {
-      OtherInst->substituteRegister(CopyToExec, AMDGPU::EXEC,
+      OtherInst->substituteRegister(CopyToExec, Exec,
                                     AMDGPU::NoSubRegister, *TRI);
     }
   }
Index: llvm/trunk/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
@@ -82,13 +82,21 @@
   return new SIOptimizeExecMaskingPreRA();
 }
 
-static bool isEndCF(const MachineInstr& MI, const SIRegisterInfo* TRI) {
+static bool isEndCF(const MachineInstr &MI, const SIRegisterInfo *TRI,
+                    const GCNSubtarget &ST) {
+  if (ST.isWave32()) {
+    return MI.getOpcode() == AMDGPU::S_OR_B32 &&
+           MI.modifiesRegister(AMDGPU::EXEC_LO, TRI);
+  }
+
   return MI.getOpcode() == AMDGPU::S_OR_B64 &&
          MI.modifiesRegister(AMDGPU::EXEC, TRI);
 }
 
-static bool isFullExecCopy(const MachineInstr& MI) {
-  if (MI.isCopy() && MI.getOperand(1).getReg() == AMDGPU::EXEC) {
+static bool isFullExecCopy(const MachineInstr& MI, const GCNSubtarget& ST) {
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+
+  if (MI.isCopy() && MI.getOperand(1).getReg() == Exec) {
     assert(MI.isFullCopy());
     return true;
   }
@@ -97,24 +105,27 @@
 }
 
 static unsigned getOrNonExecReg(const MachineInstr &MI,
-                                const SIInstrInfo &TII) {
+                                const SIInstrInfo &TII,
+                                const GCNSubtarget& ST) {
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
   auto Op = TII.getNamedOperand(MI, AMDGPU::OpName::src1);
-  if (Op->isReg() && Op->getReg() != AMDGPU::EXEC)
+  if (Op->isReg() && Op->getReg() != Exec)
      return Op->getReg();
   Op = TII.getNamedOperand(MI, AMDGPU::OpName::src0);
-  if (Op->isReg() && Op->getReg() != AMDGPU::EXEC)
+  if (Op->isReg() && Op->getReg() != Exec)
      return Op->getReg();
   return AMDGPU::NoRegister;
 }
 
 static MachineInstr* getOrExecSource(const MachineInstr &MI,
                                      const SIInstrInfo &TII,
-                                     const MachineRegisterInfo &MRI) {
-  auto SavedExec = getOrNonExecReg(MI, TII);
+                                     const MachineRegisterInfo &MRI,
+                                     const GCNSubtarget& ST) {
+  auto SavedExec = getOrNonExecReg(MI, TII, ST);
   if (SavedExec == AMDGPU::NoRegister)
     return nullptr;
   auto SaveExecInst = MRI.getUniqueVRegDef(SavedExec);
-  if (!SaveExecInst || !isFullExecCopy(*SaveExecInst))
+  if (!SaveExecInst || !isFullExecCopy(*SaveExecInst, ST))
     return nullptr;
   return SaveExecInst;
 }
@@ -180,10 +191,11 @@
                                      LiveIntervals *LIS) {
   const SIRegisterInfo *TRI = ST.getRegisterInfo();
   const SIInstrInfo *TII = ST.getInstrInfo();
-  const unsigned AndOpc = AMDGPU::S_AND_B64;
-  const unsigned Andn2Opc = AMDGPU::S_ANDN2_B64;
-  const unsigned CondReg = AMDGPU::VCC;
-  const unsigned ExecReg = AMDGPU::EXEC;
+  bool Wave32 = ST.isWave32();
+  const unsigned AndOpc = Wave32 ? AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64;
+  const unsigned Andn2Opc = Wave32 ? AMDGPU::S_ANDN2_B32 : AMDGPU::S_ANDN2_B64;
+  const unsigned CondReg = Wave32 ? AMDGPU::VCC_LO : AMDGPU::VCC;
+  const unsigned ExecReg = Wave32 ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
 
   auto I = llvm::find_if(MBB.terminators(), [](const MachineInstr &MI) {
                            unsigned Opc = MI.getOpcode();
@@ -290,6 +302,7 @@
   MachineRegisterInfo &MRI = MF.getRegInfo();
   LiveIntervals *LIS = &getAnalysis<LiveIntervals>();
   DenseSet<unsigned> RecalcRegs({AMDGPU::EXEC_LO, AMDGPU::EXEC_HI});
+  unsigned Exec = ST.isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
   bool Changed = false;
 
   for (MachineBasicBlock &MBB : MF) {
@@ -368,19 +381,19 @@
     // Try to collapse adjacent endifs.
     auto E = MBB.end();
     auto Lead = skipDebugInstructionsForward(MBB.begin(), E);
-    if (MBB.succ_size() != 1 || Lead == E || !isEndCF(*Lead, TRI))
+    if (MBB.succ_size() != 1 || Lead == E || !isEndCF(*Lead, TRI, ST))
       continue;
 
     MachineBasicBlock *TmpMBB = &MBB;
     auto NextLead = skipIgnoreExecInstsTrivialSucc(TmpMBB, std::next(Lead));
-    if (NextLead == TmpMBB->end() || !isEndCF(*NextLead, TRI) ||
-        !getOrExecSource(*NextLead, *TII, MRI))
+    if (NextLead == TmpMBB->end() || !isEndCF(*NextLead, TRI, ST) ||
+        !getOrExecSource(*NextLead, *TII, MRI, ST))
       continue;
 
     LLVM_DEBUG(dbgs() << "Redundant EXEC = S_OR_B64 found: " << *Lead << '\n');
 
-    auto SaveExec = getOrExecSource(*Lead, *TII, MRI);
-    unsigned SaveExecReg = getOrNonExecReg(*Lead, *TII);
+    auto SaveExec = getOrExecSource(*Lead, *TII, MRI, ST);
+    unsigned SaveExecReg = getOrNonExecReg(*Lead, *TII, ST);
     for (auto &Op : Lead->operands()) {
       if (Op.isReg())
         RecalcRegs.insert(Op.getReg());
@@ -414,7 +427,7 @@
     if (SafeToReplace) {
       LIS->RemoveMachineInstrFromMaps(*SaveExec);
       SaveExec->eraseFromParent();
-      MRI.replaceRegWith(SavedExec, AMDGPU::EXEC);
+      MRI.replaceRegWith(SavedExec, Exec);
       LIS->removeInterval(SavedExec);
     }
   }
Index: llvm/trunk/lib/Target/AMDGPU/SIPeepholeSDWA.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIPeepholeSDWA.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIPeepholeSDWA.cpp
@@ -954,7 +954,8 @@
   if (TII->isVOPC(Opc)) {
     if (!ST.hasSDWASdst()) {
       const MachineOperand *SDst = TII->getNamedOperand(MI, AMDGPU::OpName::sdst);
-      if (SDst && SDst->getReg() != AMDGPU::VCC)
+      if (SDst && (SDst->getReg() != AMDGPU::VCC &&
+                   SDst->getReg() != AMDGPU::VCC_LO))
         return false;
     }
 
@@ -1019,7 +1020,7 @@
     SDWAInst.add(*Dst);
   } else {
     assert(AMDGPU::getNamedOperandIdx(SDWAOpcode, AMDGPU::OpName::sdst) != -1);
-    SDWAInst.addReg(AMDGPU::VCC, RegState::Define);
+    SDWAInst.addReg(TRI->getVCC(), RegState::Define);
   }
 
   // Copy src0, initialize src0_modifiers. All sdwa instructions has src0 and
Index: llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.h
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.h
+++ llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.h
@@ -33,6 +33,7 @@
   BitVector VGPRPressureSets;
   bool SpillSGPRToVGPR;
   bool SpillSGPRToSMEM;
+  bool isWave32;
 
   void classifyPressureSet(unsigned PSetID, unsigned Reg,
                            BitVector &PressureSets) const;
@@ -231,6 +232,20 @@
   getConstrainedRegClassForOperand(const MachineOperand &MO,
                                  const MachineRegisterInfo &MRI) const override;
 
+  const TargetRegisterClass *getBoolRC() const {
+    return isWave32 ? &AMDGPU::SReg_32_XM0RegClass
+                    : &AMDGPU::SReg_64RegClass;
+  }
+
+  const TargetRegisterClass *getWaveMaskRegClass() const {
+    return isWave32 ? &AMDGPU::SReg_32_XM0_XEXECRegClass
+                    : &AMDGPU::SReg_64_XEXECRegClass;
+  }
+
+  unsigned getVCC() const;
+
+  const TargetRegisterClass *getRegClass(unsigned RCID) const;
+
   // Find reaching register definition
   MachineInstr *findReachingDef(unsigned Reg, unsigned SubReg,
                                 MachineInstr &Use,
Index: llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -63,7 +63,8 @@
   SGPRPressureSets(getNumRegPressureSets()),
   VGPRPressureSets(getNumRegPressureSets()),
   SpillSGPRToVGPR(false),
-  SpillSGPRToSMEM(false) {
+  SpillSGPRToSMEM(false),
+  isWave32(ST.isWave32()) {
   if (EnableSpillSGPRToSMEM && ST.hasScalarStores())
     SpillSGPRToSMEM = true;
   else if (EnableSpillSGPRToVGPR)
@@ -184,6 +185,13 @@
   // Reserve null register - it shall never be allocated
   reserveRegisterTuples(Reserved, AMDGPU::SGPR_NULL);
 
+  // Disallow vcc_hi allocation in wave32. It may be allocated but most likely
+  // will result in bugs.
+  if (isWave32) {
+    Reserved.set(AMDGPU::VCC);
+    Reserved.set(AMDGPU::VCC_HI);
+  }
+
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
 
   unsigned MaxNumSGPRs = ST.getMaxNumSGPRs(MF);
@@ -1706,6 +1714,25 @@
   }
 }
 
+unsigned SIRegisterInfo::getVCC() const {
+  return isWave32 ? AMDGPU::VCC_LO : AMDGPU::VCC;
+}
+
+const TargetRegisterClass *
+SIRegisterInfo::getRegClass(unsigned RCID) const {
+  switch ((int)RCID) {
+  case AMDGPU::SReg_1RegClassID:
+    return getBoolRC();
+  case AMDGPU::SReg_1_XEXECRegClassID:
+    return isWave32 ? &AMDGPU::SReg_32_XM0_XEXECRegClass
+      : &AMDGPU::SReg_64_XEXECRegClass;
+  case -1:
+    return nullptr;
+  default:
+    return AMDGPURegisterInfo::getRegClass(RCID);
+  }
+}
+
 // Find reaching register definition
 MachineInstr *SIRegisterInfo::findReachingDef(unsigned Reg, unsigned SubReg,
                                               MachineInstr &Use,
Index: llvm/trunk/lib/Target/AMDGPU/SIShrinkInstructions.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIShrinkInstructions.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIShrinkInstructions.cpp
@@ -551,6 +551,7 @@
   MachineRegisterInfo &MRI = MF.getRegInfo();
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   const SIInstrInfo *TII = ST.getInstrInfo();
+  unsigned VCCReg = ST.isWave32() ? AMDGPU::VCC_LO : AMDGPU::VCC;
 
   std::vector<unsigned> I1Defs;
 
@@ -726,10 +727,10 @@
           // So, instead of forcing the instruction to write to VCC, we provide
           // a hint to the register allocator to use VCC and then we will run
           // this pass again after RA and shrink it if it outputs to VCC.
-          MRI.setRegAllocationHint(MI.getOperand(0).getReg(), 0, AMDGPU::VCC);
+          MRI.setRegAllocationHint(MI.getOperand(0).getReg(), 0, VCCReg);
           continue;
         }
-        if (DstReg != AMDGPU::VCC)
+        if (DstReg != VCCReg)
           continue;
       }
 
@@ -742,10 +743,10 @@
           continue;
         unsigned SReg = Src2->getReg();
         if (TargetRegisterInfo::isVirtualRegister(SReg)) {
-          MRI.setRegAllocationHint(SReg, 0, AMDGPU::VCC);
+          MRI.setRegAllocationHint(SReg, 0, VCCReg);
           continue;
         }
-        if (SReg != AMDGPU::VCC)
+        if (SReg != VCCReg)
           continue;
       }
 
@@ -758,20 +759,24 @@
                                                         AMDGPU::OpName::src2);
 
       if (SDst) {
-        if (SDst->getReg() != AMDGPU::VCC) {
+        bool Next = false;
+
+        if (SDst->getReg() != VCCReg) {
           if (TargetRegisterInfo::isVirtualRegister(SDst->getReg()))
-            MRI.setRegAllocationHint(SDst->getReg(), 0, AMDGPU::VCC);
-          continue;
+            MRI.setRegAllocationHint(SDst->getReg(), 0, VCCReg);
+          Next = true;
         }
 
         // All of the instructions with carry outs also have an SGPR input in
         // src2.
-        if (Src2 && Src2->getReg() != AMDGPU::VCC) {
+        if (Src2 && Src2->getReg() != VCCReg) {
           if (TargetRegisterInfo::isVirtualRegister(Src2->getReg()))
-            MRI.setRegAllocationHint(Src2->getReg(), 0, AMDGPU::VCC);
+            MRI.setRegAllocationHint(Src2->getReg(), 0, VCCReg);
+          Next = true;
+        }
 
+        if (Next)
           continue;
-        }
       }
 
       // We can shrink this instruction
Index: llvm/trunk/lib/Target/AMDGPU/SIWholeQuadMode.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIWholeQuadMode.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIWholeQuadMode.cpp
@@ -148,6 +148,7 @@
   CallingConv::ID CallingConv;
   const SIInstrInfo *TII;
   const SIRegisterInfo *TRI;
+  const GCNSubtarget *ST;
   MachineRegisterInfo *MRI;
   LiveIntervals *LIS;
 
@@ -278,7 +279,7 @@
     // for VCC, which can appear as the (implicit) input of a uniform branch,
     // e.g. when a loop counter is stored in a VGPR.
     if (!TargetRegisterInfo::isVirtualRegister(Reg)) {
-      if (Reg == AMDGPU::EXEC)
+      if (Reg == AMDGPU::EXEC || Reg == AMDGPU::EXEC_LO)
         continue;
 
       for (MCRegUnitIterator RegUnit(Reg, TRI); RegUnit.isValid(); ++RegUnit) {
@@ -620,13 +621,16 @@
   MachineInstr *MI;
 
   if (SaveWQM) {
-    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::S_AND_SAVEEXEC_B64),
+    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(ST->isWave32() ?
+                   AMDGPU::S_AND_SAVEEXEC_B32 : AMDGPU::S_AND_SAVEEXEC_B64),
                  SaveWQM)
              .addReg(LiveMaskReg);
   } else {
-    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::S_AND_B64),
-                 AMDGPU::EXEC)
-             .addReg(AMDGPU::EXEC)
+    unsigned Exec = ST->isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(ST->isWave32() ?
+                   AMDGPU::S_AND_B32 : AMDGPU::S_AND_B64),
+                 Exec)
+             .addReg(Exec)
              .addReg(LiveMaskReg);
   }
 
@@ -638,13 +642,15 @@
                             unsigned SavedWQM) {
   MachineInstr *MI;
 
+  unsigned Exec = ST->isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
   if (SavedWQM) {
-    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::COPY), AMDGPU::EXEC)
+    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::COPY), Exec)
              .addReg(SavedWQM);
   } else {
-    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::S_WQM_B64),
-                 AMDGPU::EXEC)
-             .addReg(AMDGPU::EXEC);
+    MI = BuildMI(MBB, Before, DebugLoc(), TII->get(ST->isWave32() ?
+                   AMDGPU::S_WQM_B32 : AMDGPU::S_WQM_B64),
+                 Exec)
+             .addReg(Exec);
   }
 
   LIS->InsertMachineInstrInMaps(*MI);
@@ -667,7 +673,8 @@
   MachineInstr *MI;
 
   assert(SavedOrig);
-  MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::EXIT_WWM), AMDGPU::EXEC)
+  MI = BuildMI(MBB, Before, DebugLoc(), TII->get(AMDGPU::EXIT_WWM),
+               ST->isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC)
            .addReg(SavedOrig);
   LIS->InsertMachineInstrInMaps(*MI);
 }
@@ -693,6 +700,7 @@
   bool WQMFromExec = isEntry;
   char State = (isEntry || !(BI.InNeeds & StateWQM)) ? StateExact : StateWQM;
   char NonWWMState = 0;
+  const TargetRegisterClass *BoolRC = TRI->getBoolRC();
 
   auto II = MBB.getFirstNonPHI(), IE = MBB.end();
   if (isEntry)
@@ -780,13 +788,13 @@
 
       if (Needs == StateWWM) {
         NonWWMState = State;
-        SavedNonWWMReg = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
+        SavedNonWWMReg = MRI->createVirtualRegister(BoolRC);
         toWWM(MBB, Before, SavedNonWWMReg);
         State = StateWWM;
       } else {
         if (State == StateWQM && (Needs & StateExact) && !(Needs & StateWQM)) {
           if (!WQMFromExec && (OutNeeds & StateWQM))
-            SavedWQMReg = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
+            SavedWQMReg = MRI->createVirtualRegister(BoolRC);
 
           toExact(MBB, Before, SavedWQMReg, LiveMaskReg);
           State = StateExact;
@@ -865,17 +873,18 @@
   LowerToCopyInstrs.clear();
   CallingConv = MF.getFunction().getCallingConv();
 
-  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  ST = &MF.getSubtarget<GCNSubtarget>();
 
-  TII = ST.getInstrInfo();
+  TII = ST->getInstrInfo();
   TRI = &TII->getRegisterInfo();
   MRI = &MF.getRegInfo();
   LIS = &getAnalysis<LiveIntervals>();
 
   char GlobalFlags = analyzeFunction(MF);
   unsigned LiveMaskReg = 0;
+  unsigned Exec = ST->isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
   if (!(GlobalFlags & StateWQM)) {
-    lowerLiveMaskQueries(AMDGPU::EXEC);
+    lowerLiveMaskQueries(Exec);
     if (!(GlobalFlags & StateWWM))
       return !LiveMaskQueries.empty();
   } else {
@@ -884,10 +893,10 @@
     MachineBasicBlock::iterator EntryMI = Entry.getFirstNonPHI();
 
     if (GlobalFlags & StateExact || !LiveMaskQueries.empty()) {
-      LiveMaskReg = MRI->createVirtualRegister(&AMDGPU::SReg_64RegClass);
+      LiveMaskReg = MRI->createVirtualRegister(TRI->getBoolRC());
       MachineInstr *MI = BuildMI(Entry, EntryMI, DebugLoc(),
                                  TII->get(AMDGPU::COPY), LiveMaskReg)
-                             .addReg(AMDGPU::EXEC);
+                             .addReg(Exec);
       LIS->InsertMachineInstrInMaps(*MI);
     }
 
@@ -895,9 +904,10 @@
 
     if (GlobalFlags == StateWQM) {
       // For a shader that needs only WQM, we can just set it once.
-      BuildMI(Entry, EntryMI, DebugLoc(), TII->get(AMDGPU::S_WQM_B64),
-              AMDGPU::EXEC)
-          .addReg(AMDGPU::EXEC);
+      BuildMI(Entry, EntryMI, DebugLoc(), TII->get(ST->isWave32() ?
+                AMDGPU::S_WQM_B32 : AMDGPU::S_WQM_B64),
+              Exec)
+          .addReg(Exec);
 
       lowerCopyInstrs();
       // EntryMI may become invalid here
Index: llvm/trunk/test/CodeGen/AMDGPU/mubuf-legalize-operands.mir
===================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/mubuf-legalize-operands.mir
+++ llvm/trunk/test/CodeGen/AMDGPU/mubuf-legalize-operands.mir
@@ -1,5 +1,6 @@
-# RUN: llc -march=amdgcn -mcpu=gfx700 -verify-machineinstrs -verify-machine-dom-info --run-pass=si-fix-sgpr-copies -o - %s | FileCheck %s --check-prefixes=COMMON,ADDR64
-# RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -verify-machine-dom-info --run-pass=si-fix-sgpr-copies -o - %s | FileCheck %s --check-prefixes=COMMON,NO-ADDR64
+# RUN: llc -march=amdgcn -mcpu=gfx700 -verify-machineinstrs -verify-machine-dom-info --run-pass=si-fix-sgpr-copies -o - %s | FileCheck %s --check-prefixes=W64,ADDR64
+# RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -verify-machine-dom-info --run-pass=si-fix-sgpr-copies -o - %s | FileCheck %s --check-prefixes=W64,W64-NO-ADDR64
+# RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -verify-machineinstrs -verify-machine-dom-info --run-pass=si-fix-sgpr-copies -o - %s | FileCheck %s --check-prefixes=W64,W64-NO-ADDR64
 
 # Test that we correctly legalize VGPR Rsrc operands in MUBUF instructions.
 #
@@ -7,27 +8,50 @@
 # needing a waterfall. For all other instruction variants, and when we are
 # on non-ADDR64 hardware, we emit a waterfall loop.
 
-# COMMON-LABEL: name: idxen
-# COMMON-LABEL:  bb.0:
-# COMMON-NEXT: successors: %bb.1({{.*}})
-# COMMON: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
-# COMMON: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
-# COMMON-LABEL: bb.1:
-# COMMON-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
-# COMMON: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
-# COMMON: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
-# COMMON: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
-# COMMON: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
-# COMMON: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
-# COMMON: [[CMP0:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
-# COMMON: [[CMP1:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
-# COMMON: [[CMP:%[0-9]+]]:sreg_64 = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
-# COMMON: [[TMPEXEC:%[0-9]+]]:sreg_64 = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
-# COMMON: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
-# COMMON: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
-# COMMON: S_CBRANCH_EXECNZ %bb.1, implicit $exec
-# COMMON-LABEL  bb.2:
-# COMMON: $exec = S_MOV_B64 [[SAVEEXEC]]
+# W64-LABEL: name: idxen
+# W64-LABEL:  bb.0:
+# W64-NEXT: successors: %bb.1({{.*}})
+# W64: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W64: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
+# W64-LABEL: bb.1:
+# W64-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W64: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W64: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W64: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W64: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W64: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W64: [[CMP0:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W64: [[CMP1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W64: [[CMP:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
+# W64: [[TMPEXEC:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W64: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# W64: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
+# W64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W64-LABEL  bb.2:
+# W64: $exec = S_MOV_B64 [[SAVEEXEC]]
+
+# W32-LABEL: name: idxen
+# W32-LABEL:  bb.0:
+# W32-NEXT: successors: %bb.1({{.*}})
+# W32: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W32: [[SAVEEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
+# W32-LABEL: bb.1:
+# W32-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W32: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W32: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W32: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W32: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W32: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W32: [[CMP0:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W32: [[CMP1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W32: [[CMP:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[CMP0]], [[CMP1]], implicit-def $scc
+# W32: [[TMPEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W32: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# TODO: S_XOR_B32_term should be `implicit-def $scc`
+# W32: $exec_lo = S_XOR_B32_term $exec_lo, [[TMPEXEC]]
+# W32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W32-LABEL  bb.2:
+# W32: $exec_lo = S_MOV_B32 [[SAVEEXEC]]
 ---
 name:            idxen
 liveins:
@@ -53,27 +77,50 @@
     S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
 ...
 
-# COMMON-LABEL: name: offen
-# COMMON-LABEL:  bb.0:
-# COMMON-NEXT: successors: %bb.1({{.*}})
-# COMMON: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
-# COMMON: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
-# COMMON-LABEL: bb.1:
-# COMMON-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
-# COMMON: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
-# COMMON: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
-# COMMON: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
-# COMMON: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
-# COMMON: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
-# COMMON: [[CMP0:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
-# COMMON: [[CMP1:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
-# COMMON: [[CMP:%[0-9]+]]:sreg_64 = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
-# COMMON: [[TMPEXEC:%[0-9]+]]:sreg_64 = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
-# COMMON: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
-# COMMON: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
-# COMMON: S_CBRANCH_EXECNZ %bb.1, implicit $exec
-# COMMON-LABEL  bb.2:
-# COMMON: $exec = S_MOV_B64 [[SAVEEXEC]]
+# W64-LABEL: name: offen
+# W64-LABEL:  bb.0:
+# W64-NEXT: successors: %bb.1({{.*}})
+# W64: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W64: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
+# W64-LABEL: bb.1:
+# W64-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W64: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W64: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W64: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W64: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W64: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W64: [[CMP0:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W64: [[CMP1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W64: [[CMP:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
+# W64: [[TMPEXEC:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W64: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# W64: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
+# W64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W64-LABEL  bb.2:
+# W64: $exec = S_MOV_B64 [[SAVEEXEC]]
+
+# W32-LABEL: name: offen
+# W32-LABEL:  bb.0:
+# W32-NEXT: successors: %bb.1({{.*}})
+# W32: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W32: [[SAVEEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
+# W32-LABEL: bb.1:
+# W32-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W32: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W32: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W32: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W32: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W32: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W32: [[CMP0:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W32: [[CMP1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W32: [[CMP:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[CMP0]], [[CMP1]], implicit-def $scc
+# W32: [[TMPEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W32: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# TODO: S_XOR_B32_term should be `implicit-def $scc`
+# W32: $exec_lo = S_XOR_B32_term $exec_lo, [[TMPEXEC]]
+# W32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W32-LABEL  bb.2:
+# W32: $exec_lo = S_MOV_B32 [[SAVEEXEC]]
 ---
 name:            offen
 liveins:
@@ -99,27 +146,50 @@
     S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
 ...
 
-# COMMON-LABEL: name: bothen
-# COMMON-LABEL:  bb.0:
-# COMMON-NEXT: successors: %bb.1({{.*}})
-# COMMON: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
-# COMMON: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
-# COMMON-LABEL: bb.1:
-# COMMON-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
-# COMMON: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
-# COMMON: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
-# COMMON: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
-# COMMON: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
-# COMMON: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
-# COMMON: [[CMP0:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
-# COMMON: [[CMP1:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
-# COMMON: [[CMP:%[0-9]+]]:sreg_64 = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
-# COMMON: [[TMPEXEC:%[0-9]+]]:sreg_64 = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
-# COMMON: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_BOTHEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
-# COMMON: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
-# COMMON: S_CBRANCH_EXECNZ %bb.1, implicit $exec
-# COMMON-LABEL  bb.2:
-# COMMON: $exec = S_MOV_B64 [[SAVEEXEC]]
+# W64-LABEL: name: bothen
+# W64-LABEL:  bb.0:
+# W64-NEXT: successors: %bb.1({{.*}})
+# W64: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W64: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
+# W64-LABEL: bb.1:
+# W64-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W64: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W64: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W64: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W64: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W64: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W64: [[CMP0:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W64: [[CMP1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W64: [[CMP:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
+# W64: [[TMPEXEC:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W64: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_BOTHEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# W64: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
+# W64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W64-LABEL  bb.2:
+# W64: $exec = S_MOV_B64 [[SAVEEXEC]]
+
+# W32-LABEL: name: bothen
+# W32-LABEL:  bb.0:
+# W32-NEXT: successors: %bb.1({{.*}})
+# W32: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W32: [[SAVEEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
+# W32-LABEL: bb.1:
+# W32-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W32: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W32: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W32: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W32: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W32: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W32: [[CMP0:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W32: [[CMP1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W32: [[CMP:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[CMP0]], [[CMP1]], implicit-def $scc
+# W32: [[TMPEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W32: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_BOTHEN %4, killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# TODO: S_XOR_B32_term should be `implicit-def $scc`
+# W32: $exec_lo = S_XOR_B32_term $exec_lo, [[TMPEXEC]]
+# W32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W32-LABEL  bb.2:
+# W32: $exec_lo = S_MOV_B32 [[SAVEEXEC]]
 ---
 name:            bothen
 liveins:
@@ -145,17 +215,17 @@
     S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
 ...
 
-# COMMON-LABEL: name: addr64
-# COMMON-LABEL:  bb.0:
-# COMMON: %12:vreg_64 = COPY %8.sub0_sub1
-# COMMON: %13:sreg_64 = S_MOV_B64 0
-# COMMON: %14:sgpr_32 = S_MOV_B32 0
-# COMMON: %15:sgpr_32 = S_MOV_B32 61440
-# COMMON: %16:sreg_128 = REG_SEQUENCE %13, %subreg.sub0_sub1, %14, %subreg.sub2, %15, %subreg.sub3
-# COMMON: %9:vgpr_32 = V_ADD_I32_e32 %12.sub0, %4.sub0, implicit-def $vcc, implicit $exec
-# COMMON: %10:vgpr_32 = V_ADDC_U32_e32 %12.sub1, %4.sub1, implicit-def $vcc, implicit $vcc, implicit $exec
-# COMMON: %11:vreg_64 = REG_SEQUENCE %9, %subreg.sub0, %10, %subreg.sub1
-# COMMON: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_ADDR64 %11, killed %16, 0, 0, 0, 0, 0, 0, implicit $exec
+# ADDR64-LABEL: name: addr64
+# ADDR64-LABEL:  bb.0:
+# ADDR64: %12:vreg_64 = COPY %8.sub0_sub1
+# ADDR64: %13:sreg_64 = S_MOV_B64 0
+# ADDR64: %14:sgpr_32 = S_MOV_B32 0
+# ADDR64: %15:sgpr_32 = S_MOV_B32 61440
+# ADDR64: %16:sreg_128 = REG_SEQUENCE %13, %subreg.sub0_sub1, %14, %subreg.sub2, %15, %subreg.sub3
+# ADDR64: %9:vgpr_32 = V_ADD_I32_e32 %12.sub0, %4.sub0, implicit-def $vcc, implicit $exec
+# ADDR64: %10:vgpr_32 = V_ADDC_U32_e32 %12.sub1, %4.sub1, implicit-def $vcc, implicit $vcc, implicit $exec
+# ADDR64: %11:vreg_64 = REG_SEQUENCE %9, %subreg.sub0, %10, %subreg.sub1
+# ADDR64: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_ADDR64 %11, killed %16, 0, 0, 0, 0, 0, 0, implicit $exec
 ---
 name:            addr64
 liveins:
@@ -181,28 +251,49 @@
     S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0
 ...
 
-# COMMON-LABEL: name: offset
-# COMMON-LABEL:  bb.0:
+# W64-LABEL: name: offset
+# W64-LABEL:  bb.0:
+
+# W64-NO-ADDR64: successors: %bb.1({{.*}})
+# W64-NO-ADDR64: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W64-NO-ADDR64: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
+# W64-NO-ADDR64-LABEL: bb.1:
+# W64-NO-ADDR64-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W64-NO-ADDR64: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W64-NO-ADDR64: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W64-NO-ADDR64: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W64-NO-ADDR64: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W64-NO-ADDR64: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W64-NO-ADDR64: [[CMP0:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W64-NO-ADDR64: [[CMP1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W64-NO-ADDR64: [[CMP:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
+# W64-NO-ADDR64: [[TMPEXEC:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W64-NO-ADDR64: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFSET killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# W64-NO-ADDR64: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
+# W64-NO-ADDR64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W64-NO-ADDR64-LABEL  bb.2:
+# W64-NO-ADDR64: $exec = S_MOV_B64 [[SAVEEXEC]]
 
-# NO-ADDR64-NEXT: successors: %bb.1({{.*}})
-# NO-ADDR64: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
-# NO-ADDR64: [[SAVEEXEC:%[0-9]+]]:sreg_64_xexec = S_MOV_B64 $exec
-# NO-ADDR64-LABEL: bb.1:
-# NO-ADDR64-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
-# NO-ADDR64: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
-# NO-ADDR64: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
-# NO-ADDR64: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
-# NO-ADDR64: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
-# NO-ADDR64: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
-# NO-ADDR64: [[CMP0:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
-# NO-ADDR64: [[CMP1:%[0-9]+]]:sreg_64 = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
-# NO-ADDR64: [[CMP:%[0-9]+]]:sreg_64 = S_AND_B64 [[CMP0]], [[CMP1]], implicit-def $scc
-# NO-ADDR64: [[TMPEXEC:%[0-9]+]]:sreg_64 = S_AND_SAVEEXEC_B64 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
-# NO-ADDR64: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFSET killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
-# NO-ADDR64: $exec = S_XOR_B64_term $exec, [[TMPEXEC]], implicit-def $scc
-# NO-ADDR64: S_CBRANCH_EXECNZ %bb.1, implicit $exec
-# NO-ADDR64-LABEL  bb.2:
-# NO-ADDR64: $exec = S_MOV_B64 [[SAVEEXEC]]
+# W32: successors: %bb.1({{.*}})
+# W32: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
+# W32: [[SAVEEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_MOV_B32 $exec_lo
+# W32-LABEL: bb.1:
+# W32-NEXT: successors: %bb.1({{.*}}), %bb.2({{.*}})
+# W32: [[SRSRC0:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub0, implicit $exec
+# W32: [[SRSRC1:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub1, implicit $exec
+# W32: [[SRSRC2:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub2, implicit $exec
+# W32: [[SRSRC3:%[0-9]+]]:sgpr_32 = V_READFIRSTLANE_B32 [[VRSRC]].sub3, implicit $exec
+# W32: [[SRSRC:%[0-9]+]]:sreg_128 = REG_SEQUENCE [[SRSRC0]], %subreg.sub0, [[SRSRC1]], %subreg.sub1, [[SRSRC2]], %subreg.sub2, [[SRSRC3]], %subreg.sub3
+# W32: [[CMP0:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub0_sub1, [[VRSRC]].sub0_sub1, implicit $exec
+# W32: [[CMP1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[SRSRC]].sub2_sub3, [[VRSRC]].sub2_sub3, implicit $exec
+# W32: [[CMP:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[CMP0]], [[CMP1]], implicit-def $scc
+# W32: [[TMPEXEC:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[CMP]], implicit-def $exec, implicit-def $scc, implicit $exec
+# W32: {{[0-9]+}}:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFSET killed [[SRSRC]], 0, 0, 0, 0, 0, 0, implicit $exec
+# TODO: S_XOR_B32_term should be `implicit-def $scc`
+# W32: $exec_lo = S_XOR_B32_term $exec_lo, [[TMPEXEC]]
+# W32: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+# W32-LABEL  bb.2:
+# W32: $exec_lo = S_MOV_B32 [[SAVEEXEC]]
 
 # ADDR64: [[VRSRC:%[0-9]+]]:vreg_128 = REG_SEQUENCE %0, %subreg.sub0, %1, %subreg.sub1, %2, %subreg.sub2, %3, %subreg.sub3
 # ADDR64: [[RSRCPTR:%[0-9]+]]:vreg_64 = COPY [[VRSRC]].sub0_sub1