[AMDGPU][GlobalISel] Fix waterfall loops

[sebastian-ne]

Summary

• Move the s_and exec to its correct position before the content of the waterfall loop
• Use the SI_WATERFALL pseudo instruction, like for sdag, to benefit from optimizations
• Add support for indirect function calls

To support indirect calls, add a G_SI_CALL instruction without register
class restrictions and insert a waterfall loop when applying register
banks.

Adjust the handling of new basic blocks in RegBankSelect to cope with
the new basic blocks inserted for indirect calls.

[foad]

No objections from me.

[arsenm]

llvm/lib/CodeGen/GlobalISel/RegBankSelect.cpp

754

2 c's

llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp

906–908

I think in the absence of knowing if the call target is uniform in CallLowering, we can't do tail calls

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

I'm not sure you can guarantee this is the range that needs to be moved. Could other instructions have been moved across these between the IRTranslator and here?

[sebastian-ne]

Disallow indirect tail calls

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

The running passes seem to be

• IRTranslator
• AMDGPUPreLegalizerCombiner
• Localizer
• Legalizer
• AMDGPUPostLegalizerCombiner
• RegBankSelect

It’s fine if VALU instructions are moved inside the loop, but SALU instructions would be a problem.

Can we prevent instructions from being moved into the call code?

[arsenm]

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

I was thinking of doing this before by using virtual registers for the call pseudo and expanding to physical registers later, but it seems like overkill. Practically speaking we don't really have code reordering right now that would present an issue, and don't see why we would ever add it. The localizer does some level of reordering, but it's really just a simple sink to uses. In practice we'll only see copies inside this range. Overall I can't see a real issue doing it this way, but it still feels wrong since there's no formal guarantee

[sebastian-ne]

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

I tried out putting the call code inside adjcallstackup to -down into a bundle when it is generated, but I hit some problems.

1. The MachineVerifier complains about undefined physical registers when they are defined and used inside the bundle. It only considers def-ed registers live after a bundle, not inside it, which is more the style of vliw-bundles.
2. Bundles in SSA form (before register allocation) apparently do not have a BUNDLE instruction and no defs/uses on the bundle. Also, a bundle always needs an instruction before it, so having a call as the first instruction of a block does not work, because there is no instruction before the bundle, which is rather unfortunate.

[foad]

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

Also, a bundle always needs an instruction before it

Really? Why?

[sebastian-ne]

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

Hm, maybe it doesn’t and just the syntax looks weird. I thought the ADJCALLSTACKUP would be outside the bundle, but probably it is part of it. (I tried adding it, which failed with an assert that I interpreted as “there must be an instruction before”, at a second look it sounds more like “it’s already part of a bundle”).

ADJCALLSTACKUP 0, 0, implicit-def $scc {
  %16:_(p4) = COPY %2:sgpr_64
  ; …
  $vgpr31 = COPY %33:_(s32)
  $sgpr30_sgpr31 = SI_CALL %15:sreg_64(p0), 0, <regmask $sgpr32 $sgpr33 $sgpr34 $sgpr35 $sgpr36 $sgpr37 $sgpr38 $sgpr39 $sgpr40 $sgpr41 $sgpr42 $sgpr43 $sgpr44 $sgpr45 $sgpr46 $sgpr47 $sgpr48 $sgpr49 $sgpr50 $sgpr51 $sgpr52 $sgpr53 $sgpr54 $sgpr55 $sgpr56 $sgpr57 $sgpr58 $sgpr59 $sgpr60 $sgpr61 $sgpr62 $sgpr63 $sgpr64 and 1092 more...>, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
  ADJCALLSTACKDOWN 0, 0, implicit-def $scc
}

[foad]

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3142–3146

Right, a normal bundle looks like

BUNDLE {
  INSN1
  INSN2
}

but as you noticed the bundle's "header" instruction doesn't need to be a BUNDLE instruction, so you can have this instead:

INSN1 {
  INSN2
  INSN3
}

[sebastian-ne]

Use bundles for call instructions, so no other instructions can be moved in-between.

[foad]

llvm/lib/CodeGen/GlobalISel/Localizer.cpp

172 ↗

(On Diff #377518)

Why is this change needed? Normally a BUNDLE instruction has use operands representing all the uses of the instructions inside the bundle, so I think this code should just work without any changes? See comment in lowerCall...

llvm/lib/CodeGen/MachineVerifier.cpp

2273 ↗

(On Diff #377518)

Why is this change needed?

Anyway I think you can write !is_contained(regsDefined, Reg).

llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp

1391

Use finalizeBundle instead of MIBundleBuilder, to add all the required operands to the BUNDLE instruction?

[sebastian-ne]

Use is_contained.

llvm/lib/CodeGen/GlobalISel/Localizer.cpp

172 ↗

(On Diff #377518)

As far as I understand, it only has that after register allocation (or somewhere in these passes, like phi elimination. The pass that adds the BUNDLE instruction also adds the uses and defs).

llvm/lib/CodeGen/MachineVerifier.cpp

2273 ↗

(On Diff #377518)

The MachineVerifier sees the whole bundle as a single instruction. regsLive will only be updated after the whole bundle is worked on. This makes sense for VLIW architectures like Hexagon but doesn’t make sense for grouping instructions in amdgpu.
So, making this target specific is probably the ideal way, maybe I should do this? This currently relaxes the check to allow uses to refer to defines that happen in the same bundle.

Thanks, is_contained looks useful.

llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp

1391

As far as I understood, finalizeBundle should only be used after register allocation. This will be done in some pass there, insert the BUNDLE instruction and add the uses and defs. Adding the uses and defs here will fail the MachineVerifier, because you can only have a single def in SSA mode.

The bundle will be dissolved before it gets to the finalizeBundle point though.

[foad]

llvm/lib/CodeGen/GlobalISel/Localizer.cpp

172 ↗

(On Diff #377518)

AMDGPU creates some bundles at instruction selection time, and it does use finalizeBundle, but they do not have any defs, so maybe that does not violate SSA? Example:

BUNDLE implicit killed %6:vgpr_32, implicit $m0, implicit $exec {
  DS_GWS_BARRIER killed %6:vgpr_32, 0, implicit $m0, implicit $exec :: (load (s32) from custom "GWSResource")
  S_WAITCNT 0
}

[foad]

I have just found this thread about bundling before register allocation: https://lists.llvm.org/pipermail/llvm-dev/2017-February/110467.html

[sebastian-ne]

Thanks, the Worklist.insert does read better than std::copy.

I guess we need to get some agreement about the high level design: is this a good way of handling newly created instructions and BBs? In particular I wonder if there is some way of handling newly created instructions automatically, without having to call setNextInstruction, but I can't quite see how it would work.

Yes, it gets quite complicated when introducing loops for call instructions because it moves multiple instructions and now also unbundles them (so the I++ used in RegBankSelect points to the wrong instruction – after the bundle).

[sebastian-ne]

That comment went to the wrong review :)

[sebastian-ne]

llvm/lib/CodeGen/GlobalISel/Localizer.cpp

172 ↗

(On Diff #377518)

It definitely complains if I try to finalize the bundle :)

*** Bad machine code: Multiple virtual register defs in SSA form ***
- function:    test_indirect_call_sgpr_ptr
- basic block: %bb.1  (0x6d87e8)
- instruction: BUNDLE implicit-def $scc, implicit-def %16:_(p4), implicit-def %17:_(p4), implicit-def %19:_(p4), implicit-def %20:_(s64), implicit-def %18:_(p4), implicit-def %21:_(s64), implicit-def %22:_(s32), implicit-def %23:_(s32), implicit-def %24:_(s32), implicit-def %25:_(s32), implicit-def %26:_(s32), implicit-def %27:_(s32), implicit-def %28:_(s32), implicit-def %29:_(s32), implicit-def %30:_(s32), implicit-def %31:_(s32), implicit-def %32:_(s32), implicit-def %33:_(s32), implicit-def %34:_(<4 x s32>), implicit-def $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $sgpr0, implicit-def $sgpr0_lo16, implicit-def $sgpr0_hi16, implicit-def $sgpr1, implicit-def $sgpr1_lo16, implicit-def $sgpr1_hi16, implicit-def $sgpr2, implicit-def $sgpr2_lo16, implicit-def $sgpr2_hi16, implicit-def $sgpr3, implicit-def $sgpr3_lo16, implicit-def $sgpr3_hi16, implicit-def $sgpr0_sgpr1, implicit-def $sgpr0_sgpr1_sgpr2, implicit-def $sgpr2_sgpr3, implicit-def $sgpr4_sgpr5, implicit-def $sgpr4, implicit-def $sgpr4_lo16, implicit-def $sgpr4_hi16, implicit-def $sgpr5, implicit-def $sgpr5_lo16, implicit-def $sgpr5_hi16, implicit-def $sgpr6_sgpr7, implicit-def $sgpr6, implicit-def $sgpr6_lo16, implicit-def $sgpr6_hi16, implicit-def $sgpr7, implicit-def $sgpr7_lo16, implicit-def $sgpr7_hi16, implicit-def $sgpr8_sgpr9, implicit-def $sgpr8, implicit-def $sgpr8_lo16, implicit-def $sgpr8_hi16, implicit-def $sgpr9, implicit-def $sgpr9_lo16, implicit-def $sgpr9_hi16, implicit-def $sgpr10_sgpr11, …

[sebastian-ne]

Do not use bundles anymore because they create more problems than they solve (e.g. the legalizer tries and fails to legalize instructions in a bundle).
Instead, select copies from and to physical registers that are used/defined by the call instruction and move them into the waterfall loop.

[arsenm]

llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp

1038

Add a todo to check divergence info?

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

3137–3138

It's OK to just hardcode the opcodes here

3169

This makes me nervous about overlapping register tuples. Exact match on physical registers is rarely a good idea

[sebastian-ne]

Fix review comments: Use readsRegister and modifiesRegister instead of exact register matches.

[arsenm]

Having to shuffle instructions around feels wrong but I don't have a better idea

llvm/lib/CodeGen/GlobalISel/RegBankSelect.cpp

 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
 #include <algorithm>
 #include <cassert>
 #include <cstdint>
+#include <deque>
 #include <limits>
 #include <memory>
 #include <utility>
 
 #define DEBUG_TYPE "regbankselect"
 
 using namespace llvm;
 
#ifndef NDEBUG
   if (!DisableGISelLegalityCheck)
     if (const MachineInstr *MI = machineFunctionIsIllegal(MF)) {
       reportGISelFailure(MF, *TPC, *MORE, "gisel-regbankselect",
                          "instruction is not legal", *MI);
       return false;
     }
 #endif
 
+  SmallSet<MachineBasicBlock *, 4> OrigBlocks;
+  for (auto &MBB : MF)
+    OrigBlocks.insert(&MBB);
+
   // Walk the function and assign register banks to all operands.
   // Use a RPOT to make sure all registers are assigned before we choose
   // the best mapping of the current instruction.
   ReversePostOrderTraversal<MachineFunction*> RPOT(&MF);
-  for (MachineBasicBlock *MBB : RPOT) {
+  for (MachineBasicBlock *OrigMBB : RPOT) {
+    std::deque<MachineBasicBlock *> Worklist;
+    Worklist.push_back(OrigMBB);
+    MachineBasicBlock *MBB;
+    while (!Worklist.empty()) {
+      MBB = Worklist.front();
+      Worklist.pop_front();
+
       // Set a sensible insertion point so that subsequent calls to
       // MIRBuilder.
       MIRBuilder.setMBB(*MBB);
       for (MachineBasicBlock::iterator MII = MBB->begin(), End = MBB->end();
            MII != End;) {
         // MI might be invalidated by the assignment, so move the
         // iterator before hand.
         MachineInstr &MI = *MII++;
 
         // Ignore target-specific post-isel instructions: they should use proper
         // regclasses.
         if (isTargetSpecificOpcode(MI.getOpcode()) && !MI.isPreISelOpcode())
           continue;
 
         // Ignore inline asm instructions: they should use physical
         // registers/regclasses
         if (MI.isInlineAsm())
           continue;
 
         // Ignore debug info.
         if (MI.isDebugInstr())
           continue;
 
         // Ignore IMPLICIT_DEF which must have a regclass.
         if (MI.isImplicitDef())
           continue;
 
         if (!assignInstr(MI)) {
           reportGISelFailure(MF, *TPC, *MORE, "gisel-regbankselect",
                              "unable to map instruction", MI);
           return false;
         }
 
-      // It's possible the mapping changed control flow, and moved the following
-      // instruction to a new block, so figure out the new parent.
+        // It's possible the mapping changed control flow, and moved the
+        // following instruction to a new block. Handle these cases below.
         if (MII != End) {
           MachineBasicBlock *NextInstBB = MII->getParent();
-        if (NextInstBB != MBB) {
-          LLVM_DEBUG(dbgs() << "Instruction mapping changed control flow\n");
-          MBB = NextInstBB;
-          MIRBuilder.setMBB(*MBB);
-          End = MBB->end();
+          if (NextInstBB != MBB)
+            break;
+        }
+      }
+
+      // If the mapping changed the control flow, look at the new successors.
+      for (auto *Succ : MBB->successors()) {
+        if (!OrigBlocks.count(Succ)) {
+          Worklist.push_back(Succ);
+          OrigBlocks.insert(Succ);
         }
       }
     }
   }
 
   OptMode = SaveOptMode;
   return false;
 }

llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp

 /// CC.
 static std::pair<CCAssignFn *, CCAssignFn *>
 getAssignFnsForCC(CallingConv::ID CC, const SITargetLowering &TLI) {
   return {TLI.CCAssignFnForCall(CC, false), TLI.CCAssignFnForCall(CC, true)};
 }
 
 static unsigned getCallOpcode(const MachineFunction &CallerF, bool IsIndirect,
                               bool IsTailCall) {
-  return IsTailCall ? AMDGPU::SI_TCRETURN : AMDGPU::SI_CALL;
+  assert(!(IsIndirect && IsTailCall) && "Indirect calls can't be tail calls, "
+                                        "because the address can be divergent");
+  return IsTailCall ? AMDGPU::SI_TCRETURN : AMDGPU::G_SI_CALL;
 }
 
 // Add operands to call instruction to track the callee.
 static bool addCallTargetOperands(MachineInstrBuilder &CallInst,
                                   MachineIRBuilder &MIRBuilder,
                                   AMDGPUCallLowering::CallLoweringInfo &Info) {
   if (Info.Callee.isReg()) {
     CallInst.addReg(Info.Callee.getReg());

 
 bool AMDGPUCallLowering::isEligibleForTailCallOptimization(
     MachineIRBuilder &B, CallLoweringInfo &Info,
     SmallVectorImpl<ArgInfo> &InArgs, SmallVectorImpl<ArgInfo> &OutArgs) const {
   // Must pass all target-independent checks in order to tail call optimize.
   if (!Info.IsTailCall)
     return false;
 
+  // Indirect calls can't be tail calls, because the address can be divergent.
+  if (Info.Callee.isReg())
+    return false;
+
   MachineFunction &MF = B.getMF();
   const Function &CallerF = MF.getFunction();
   CallingConv::ID CalleeCC = Info.CallConv;
   CallingConv::ID CallerCC = CallerF.getCallingConv();
 
   const SIRegisterInfo *TRI = MF.getSubtarget<GCNSubtarget>().getRegisterInfo();
   const uint32_t *CallerPreserved = TRI->getCallPreservedMask(MF, CallerCC);
   // Kernels aren't callable, and don't have a live in return address so it
MIRBuilder.buildInstr(AMDGPU::ADJCALLSTACKDOWN)
             .addImm(CalleePopBytes);
 
   if (!Info.CanLowerReturn) {
     insertSRetLoads(MIRBuilder, Info.OrigRet.Ty, Info.OrigRet.Regs,
                     Info.DemoteRegister, Info.DemoteStackIndex);
   }
 
   return true;
 }

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp

bool AMDGPUInstructionSelector::select(MachineInstr &I) {
   }
   case AMDGPU::G_AMDGPU_INTRIN_BVH_INTERSECT_RAY:
     return selectBVHIntrinsic(I);
   case AMDGPU::G_AMDGPU_BUFFER_ATOMIC_FADD:
     return selectAMDGPU_BUFFER_ATOMIC_FADD(I);
   case AMDGPU::G_SBFX:
   case AMDGPU::G_UBFX:
     return selectG_SBFX_UBFX(I);
+  case AMDGPU::G_SI_CALL:
+    I.setDesc(TII.get(AMDGPU::SI_CALL));
+    return true;
   default:
     return selectImpl(I, *CoverageInfo);
   }
   return false;
 }
 
 InstructionSelector::ComplexRendererFns
 AMDGPUInstructionSelector::selectVCSRC(MachineOperand &Root) const {

llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp

const unsigned AndSaveExecOpc =  Subtarget.isWave32() ?
     AMDGPU::S_AND_SAVEEXEC_B32 : AMDGPU::S_AND_SAVEEXEC_B64;
   const unsigned ExecReg =  Subtarget.isWave32() ?
     AMDGPU::EXEC_LO : AMDGPU::EXEC;
 
 #ifndef NDEBUG
   const int OrigRangeSize = std::distance(Range.begin(), Range.end());
 #endif
 
-  for (MachineInstr &MI : Range) {
-    for (MachineOperand &Def : MI.defs()) {
-      if (MRI.use_nodbg_empty(Def.getReg()))
-        continue;
-
-      LLT ResTy = MRI.getType(Def.getReg());
-      const RegisterBank *DefBank = getRegBank(Def.getReg(), MRI, *TRI);
-      ResultRegs.push_back(Def.getReg());
-      Register InitReg = B.buildUndef(ResTy).getReg(0);
-      Register PhiReg = MRI.createGenericVirtualRegister(ResTy);
-      InitResultRegs.push_back(InitReg);
-      PhiRegs.push_back(PhiReg);
-      MRI.setRegBank(PhiReg, *DefBank);
-      MRI.setRegBank(InitReg, *DefBank);
-    }
-  }
-
   Register SaveExecReg = MRI.createVirtualRegister(WaveRC);
   Register InitSaveExecReg = MRI.createVirtualRegister(WaveRC);
 
   // Don't bother using generic instructions/registers for the exec mask.
   B.buildInstr(TargetOpcode::IMPLICIT_DEF)
     .addDef(InitSaveExecReg);
 
   Register PhiExec = MRI.createVirtualRegister(WaveRC);
for (MachineOperand &Op : MI.uses()) {
         LLT S32 = LLT::scalar(32);
         SmallVector<Register, 8> ReadlanePieces;
 
         // The compares can be done as 64-bit, but the extract needs to be done
         // in 32-bit pieces.
 
         bool Is64 = OpSize % 64 == 0;
 
-        LLT UnmergeTy = OpSize % 64 == 0 ? LLT::scalar(64) : LLT::scalar(32);
-        unsigned CmpOp = OpSize % 64 == 0 ? AMDGPU::V_CMP_EQ_U64_e64
-          : AMDGPU::V_CMP_EQ_U32_e64;
-
-        // The compares can be done as 64-bit, but the extract needs to be done
-        // in 32-bit pieces.
+        unsigned UnmergeTySize = Is64 ? 64 : 32;
+        unsigned CmpOp =
+            Is64 ? AMDGPU::V_CMP_EQ_U64_e64 : AMDGPU::V_CMP_EQ_U32_e64;
 
         // Insert the unmerge before the loop.
 
         B.setMBB(MBB);
-        auto Unmerge = B.buildUnmerge(UnmergeTy, OpReg);
+        unsigned NumPieces = OpSize / UnmergeTySize;
+        SmallVector<Register, 8> UnmergePieces;
+        if (NumPieces == 1) {
+          UnmergePieces.push_back(OpReg);
+        } else {
+          LLT UnmergeTy = LLT::scalar(UnmergeTySize);
+          MachineInstrBuilder Unmerge = B.buildUnmerge(UnmergeTy, OpReg);
+          for (unsigned PieceIdx = 0; PieceIdx != NumPieces; ++PieceIdx)
+            UnmergePieces.push_back(Unmerge.getReg(PieceIdx));
+        }
         B.setInstr(*I);
 
-        unsigned NumPieces = Unmerge->getNumOperands() - 1;
-        for (unsigned PieceIdx = 0; PieceIdx != NumPieces; ++PieceIdx) {
-          Register UnmergePiece = Unmerge.getReg(PieceIdx);
-
+        for (Register UnmergePiece : UnmergePieces) {
           Register CurrentLaneOpReg;
           if (Is64) {
             Register CurrentLaneOpRegLo = MRI.createGenericVirtualRegister(S32);
             Register CurrentLaneOpRegHi = MRI.createGenericVirtualRegister(S32);
 
             MRI.setRegClass(UnmergePiece, &AMDGPU::VReg_64RegClass);
             MRI.setRegClass(CurrentLaneOpRegLo, &AMDGPU::SReg_32_XM0RegClass);
             MRI.setRegClass(CurrentLaneOpRegHi, &AMDGPU::SReg_32_XM0RegClass);
for (MachineOperand &Op : MI.uses()) {
           }
         }
 
         // FIXME: Build merge seems to switch to CONCAT_VECTORS but not
         // BUILD_VECTOR
         if (OpTy.isVector()) {
           auto Merge = B.buildBuildVector(OpTy, ReadlanePieces);
           Op.setReg(Merge.getReg(0));
-        } else {
+          MRI.setRegBank(Op.getReg(), AMDGPU::SGPRRegBank);
+        } else if (ReadlanePieces.size() > 1) {
           auto Merge = B.buildMerge(OpTy, ReadlanePieces);
           Op.setReg(Merge.getReg(0));
-        }
-
           MRI.setRegBank(Op.getReg(), AMDGPU::SGPRRegBank);
+        } else {
+          Op.setReg(ReadlanePieces[0]);
+        }
       }
 
       // Make sure we don't re-process this register again.
       WaterfalledRegMap.insert(std::make_pair(OldReg, Op.getReg()));
     }
   }
 
-  B.setInsertPt(*LoopBB, LoopBB->end());
-
   // Update EXEC, save the original EXEC value to VCC.
   B.buildInstr(AndSaveExecOpc)
     .addDef(NewExec)
     .addReg(CondReg, RegState::Kill);
 
   MRI.setSimpleHint(NewExec, CondReg);
 
+  B.setInsertPt(*LoopBB, LoopBB->end());
+
   // Update EXEC, switch all done bits to 0 and all todo bits to 1.
   B.buildInstr(XorTermOpc)
     .addDef(ExecReg)
     .addReg(ExecReg)
     .addReg(NewExec);
 
   // XXX - s_xor_b64 sets scc to 1 if the result is nonzero, so can we use
   // s_cbranch_scc0?
 
   // Loop back to V_READFIRSTLANE_B32 if there are still variants to cover.
-  B.buildInstr(AMDGPU::S_CBRANCH_EXECNZ)
-    .addMBB(LoopBB);
+  B.buildInstr(AMDGPU::SI_WATERFALL_LOOP).addMBB(LoopBB);
 
   // Save the EXEC mask before the loop.
   BuildMI(MBB, MBB.end(), DL, TII->get(MovTermOpc), SaveExecReg)
     .addReg(ExecReg);
 
   // Restore the EXEC mask after the loop.
   B.setMBB(*RestoreExecBB);
   B.buildInstr(MovTermOpc)

 static unsigned extractSWZ(unsigned CachePolicy) {
   return (CachePolicy >> 3) & 1;
 }
 
 
 MachineInstr *
 AMDGPURegisterBankInfo::selectStoreIntrinsic(MachineIRBuilder &B,
                                              MachineInstr &MI) const {
-   MachineRegisterInfo &MRI = *B.getMRI();
   executeInWaterfallLoop(B, MI, MRI, {2, 4});
 
   // FIXME: DAG lowering brokenly changes opcode based on FP vs. integer.
 
   Register VData = MI.getOperand(1).getReg();
   LLT Ty = MRI.getType(VData);
 
   int EltSize = Ty.getScalarSizeInBits();
default: {
         }
       }
 
       break;
     }
     }
     break;
   }
+  case AMDGPU::G_SI_CALL: {
+    // Use a set to avoid extra readfirstlanes in the case where multiple
+    // operands are the same register.
+    SmallSet<Register, 4> SGPROperandRegs;
+
+    if (!collectWaterfallOperands(SGPROperandRegs, MI, MRI, {1}))
+      break;
+
+    // Move everything between ADJCALLSTACKUP and ADJCALLSTACKDOWN and
+    // following copies, we also need to move copies from and to physical
+    // registers into the loop block.
+    unsigned FrameSetupOpcode = TII->getCallFrameSetupOpcode();
+    unsigned FrameDestroyOpcode = TII->getCallFrameDestroyOpcode();
+
+    MachineBasicBlock &MBB = *MI.getParent();
+    MachineBasicBlock::iterator Start(&MI);
+    while (Start->getOpcode() != FrameSetupOpcode)
+      --Start;
+    MachineBasicBlock::iterator End(&MI);
+    while (End->getOpcode() != FrameDestroyOpcode)
+      ++End;
+    // Also include following copies of the return value
+    ++End;
+    while (End != MBB.end() && End->isCopy() && End->getOperand(1).isReg() &&
+           MI.definesRegister(End->getOperand(1).getReg()))
+      ++End;
+
+    MachineIRBuilder B(*Start);
+    executeInWaterfallLoop(B, make_range(Start, End), SGPROperandRegs, MRI);
+    break;
+  }
   case AMDGPU::G_LOAD:
   case AMDGPU::G_ZEXTLOAD:
   case AMDGPU::G_SEXTLOAD: {
     if (applyMappingLoad(MI, OpdMapper, MRI))
       return;
     break;
   }
   case AMDGPU::G_DYN_STACKALLOC:
     applyMappingDynStackAlloc(MI, OpdMapper, MRI);
     return;
   case AMDGPU::G_SBFX:
     applyMappingBFE(OpdMapper, /*Signed*/ true);
     return;
   case AMDGPU::G_UBFX:
     applyMappingBFE(OpdMapper, /*Signed*/ false);
     return;
   default:
     break;
   }
 
   return applyDefaultMapping(OpdMapper);
if (Size == 64) {
       OpdsMapping[1] = AMDGPU::getValueMapping(CondBank, 1);
       OpdsMapping[2] = AMDGPU::getValueMapping(Bank, Size);
       OpdsMapping[3] = AMDGPU::getValueMapping(Bank, Size);
     }
 
     break;
   }
 
+  case AMDGPU::G_SI_CALL: {
+    OpdsMapping[0] = AMDGPU::getValueMapping(AMDGPU::SGPRRegBankID, 64);
+    // Lie and claim everything is legal, even though some need to be
+    // SGPRs. applyMapping will have to deal with it as a waterfall loop.
+    OpdsMapping[1] = getSGPROpMapping(MI.getOperand(1).getReg(), MRI, *TRI);
+
+    // Allow anything for implicit arguments
+    for (unsigned I = 4; I < MI.getNumOperands(); ++I) {
+      if (MI.getOperand(I).isReg()) {
+        Register Reg = MI.getOperand(I).getReg();
+        auto OpBank = getRegBankID(Reg, MRI);
+        unsigned Size = getSizeInBits(Reg, MRI, *TRI);
+        OpdsMapping[I] = AMDGPU::getValueMapping(OpBank, Size);
+      }
+    }
+    break;
+  }
   case AMDGPU::G_LOAD:
   case AMDGPU::G_ZEXTLOAD:
   case AMDGPU::G_SEXTLOAD:
     return getInstrMappingForLoad(MI);
 
   case AMDGPU::G_ATOMICRMW_XCHG:
   case AMDGPU::G_ATOMICRMW_ADD:
   case AMDGPU::G_ATOMICRMW_SUB:

llvm/lib/Target/AMDGPU/SIInstructions.td

 
 def G_AMDGPU_INTRIN_BVH_INTERSECT_RAY : AMDGPUGenericInstruction {
   let OutOperandList = (outs type0:$dst);
   let InOperandList = (ins unknown:$intrin, variable_ops);
   let hasSideEffects = 0;
   let mayLoad = 1;
   let mayStore = 0;
 }
+
+// Generic instruction for SI_CALL, so we can select the register bank and insert a waterfall loop
+// if necessary.
+def G_SI_CALL : AMDGPUGenericInstruction {
+  let OutOperandList = (outs SReg_64:$dst);
+  let InOperandList = (ins type0:$src0, unknown:$callee);
+  let Size = 4;
+  let isCall = 1;
+  let UseNamedOperandTable = 1;
+  let SchedRW = [WriteBranch];
+  // TODO: Should really base this on the call target
+  let isConvergent = 1;
+}

llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-abi-attribute-hints.ll

define amdgpu_kernel void @kernel_call_no_workitem_ids() {
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @extern
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; CHECK:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @kernel_call_no_workitem_ids() {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY12]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY14]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @extern() "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z"
   ret void
 }
 
 define amdgpu_kernel void @kernel_call_no_workgroup_ids() {
   ; CHECK-LABEL: name: kernel_call_no_workgroup_ids
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @extern
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; CHECK:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @kernel_call_no_workgroup_ids() {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY12]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY14]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @extern() "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }
 
 define amdgpu_kernel void @kernel_call_no_other_sgprs() {
   ; CHECK-LABEL: name: kernel_call_no_other_sgprs
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @extern
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; CHECK:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @kernel_call_no_other_sgprs() {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY12]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY14]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @extern() "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }
 
 define void @func_call_no_workitem_ids() {
   ; CHECK-LABEL: name: func_call_no_workitem_ids
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @extern
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK:   [[COPY17:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY17]](<4 x s32>)
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY18:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY18]]
   call void @extern() "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z"
   ret void
 }
 
 define void @func_call_no_workgroup_ids() {
   ; CHECK-LABEL: name: func_call_no_workgroup_ids
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @extern
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK:   [[COPY17:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY17]](<4 x s32>)
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY18:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY18]]
   call void @extern() "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }
 
 define void @func_call_no_other_sgprs() {
   ; CHECK-LABEL: name: func_call_no_other_sgprs
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @extern
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @extern
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK:   [[COPY17:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY17]](<4 x s32>)
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @extern, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY18:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY18]]
   call void @extern() "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z"
   ret void
 }

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

define amdgpu_kernel void @test_call_external_void_func_i32([17 x i8]) #0 {
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 20
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i32([17 x i8]) #0 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_call_external_void_func_i32
   ; GFX908: bb.1 (%ir-block.1):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 20
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i32([17 x i8]) #0 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define void @test_func_call_external_void_func_i32() #0 {
   ; GFX900-LABEL: name: test_func_call_external_void_func_i32
   ; GFX900: bb.1 (%ir-block.0):
   ; GFX900:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; GFX900:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; GFX900:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GFX900:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 99
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GFX900:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; GFX900:   $vgpr0 = COPY [[C]](s32)
   ; GFX900:   [[COPY17:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GFX900:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY17]](<4 x s32>)
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX900:   $vgpr31 = COPY [[COPY16]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   [[COPY18:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; GFX900:   S_SETPC_B64_return [[COPY18]]
   ; GFX908-LABEL: name: test_func_call_external_void_func_i32
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; GFX908:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 99
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GFX908:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; GFX908:   $vgpr0 = COPY [[C]](s32)
   ; GFX908:   [[COPY17:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GFX908:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY17]](<4 x s32>)
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY13]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY15]](s32)
   ; GFX908:   $vgpr31 = COPY [[COPY16]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   [[COPY18:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; GFX908:   S_SETPC_B64_return [[COPY18]]
   call void @external_void_func_i32(i32 99)
   ret void
 }
 
 ; Explicit argument is split between registers ad the stack due to v31
define amdgpu_kernel void @test_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GFX900:   [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32)
   ; GFX900:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 20
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_call_external_void_func_v32i32
   ; GFX908: bb.1 (%ir-block.1):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GFX908:   [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32)
   ; GFX908:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 20
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_v32i32(<32 x i32> zeroinitializer)
   ret void
 }
 
 define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900-LABEL: name: test_func_call_external_void_func_v32i32
define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900:   [[TRUNC31:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC30]](s16)
   ; GFX900:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr16
   ; GFX900:   [[TRUNC32:%[0-9]+]]:_(s16) = G_TRUNC [[COPY24]](s32)
   ; GFX900:   [[TRUNC33:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC32]](s16)
   ; GFX900:   [[COPY25:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GFX900:   [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32)
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
   ; GFX900:   [[COPY26:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY27:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GFX900:   [[COPY28:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GFX900:   [[COPY29:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GFX900:   [[COPY30:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GFX900:   [[COPY31:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GFX900:   [[COPY32:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GFX900:   [[COPY33:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY26]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY27]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[COPY28]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY29]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY30]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY31]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY32]](s32)
   ; GFX900:   $vgpr31 = COPY [[COPY33]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX900:   [[COPY36:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY25]]
   ; GFX900:   S_SETPC_B64_return [[COPY36]]
   ; GFX908-LABEL: name: test_func_call_external_void_func_v32i32
   ; GFX908: bb.1 (%ir-block.1):
   ; GFX908:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8, $vgpr9, $vgpr10, $vgpr11, $vgpr12, $vgpr13, $vgpr14, $vgpr15, $vgpr16, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; GFX908:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX908:   [[TRUNC31:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC30]](s16)
   ; GFX908:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr16
   ; GFX908:   [[TRUNC32:%[0-9]+]]:_(s16) = G_TRUNC [[COPY24]](s32)
   ; GFX908:   [[TRUNC33:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC32]](s16)
   ; GFX908:   [[COPY25:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GFX908:   [[BUILD_VECTOR:%[0-9]+]]:_(<32 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32), [[C]](s32)
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
   ; GFX908:   [[COPY26:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY27:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GFX908:   [[COPY28:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GFX908:   [[COPY29:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GFX908:   [[COPY30:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GFX908:   [[COPY31:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GFX908:   [[COPY32:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GFX908:   [[COPY33:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @test_func_call_external_void_func_v32i32([17 x i8]) #0 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY26]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY27]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[COPY28]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY29]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY30]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY31]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY32]](s32)
   ; GFX908:   $vgpr31 = COPY [[COPY33]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GFX908:   [[COPY36:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY25]]
   ; GFX908:   S_SETPC_B64_return [[COPY36]]
   call void @external_void_func_v32i32(<32 x i32> zeroinitializer)
   ret void
 }
 
 ; FIXME: Should fold out parts with known 0 id.
define amdgpu_kernel void @test_only_workitem_id_x() #0 !reqd_work_group_size !0 {
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_x() #0 !reqd_work_group_size !0 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_only_workitem_id_x
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_x() #0 !reqd_work_group_size !0 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_y() #0 !reqd_work_group_size !1 {
   ; GFX900-LABEL: name: test_only_workitem_id_y
   ; GFX900: bb.1 (%ir-block.0):
   ; GFX900:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX900:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX900:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX900:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX900:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_y() #0 !reqd_work_group_size !1 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_only_workitem_id_y
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_y() #0 !reqd_work_group_size !1 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_z() #0 !reqd_work_group_size !2 {
   ; GFX900-LABEL: name: test_only_workitem_id_z
   ; GFX900: bb.1 (%ir-block.0):
   ; GFX900:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX900:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX900:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX900:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX900:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_z() #0 !reqd_work_group_size !2 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_only_workitem_id_z
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_z() #0 !reqd_work_group_size !2 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_xy() #0 !reqd_work_group_size !3 {
   ; GFX900-LABEL: name: test_only_workitem_id_xy
   ; GFX900: bb.1 (%ir-block.0):
   ; GFX900:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX900:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX900:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX900:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX900:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_xy() #0 !reqd_work_group_size !3 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_only_workitem_id_xy
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_xy() #0 !reqd_work_group_size !3 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_yz() #0 !reqd_work_group_size !4 {
   ; GFX900-LABEL: name: test_only_workitem_id_yz
   ; GFX900: bb.1 (%ir-block.0):
   ; GFX900:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX900:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX900:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX900:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX900:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_yz() #0 !reqd_work_group_size !4 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_only_workitem_id_yz
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_yz() #0 !reqd_work_group_size !4 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_kernel void @test_only_workitem_id_xz() #0 !reqd_work_group_size !5 {
   ; GFX900-LABEL: name: test_only_workitem_id_xz
   ; GFX900: bb.1 (%ir-block.0):
   ; GFX900:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX900:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX900:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX900:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX900:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX900:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX900:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX900:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX900:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX900:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX900:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX900:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX900:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX900:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX900:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX900:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX900:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX900:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX900:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX900:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX900:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX900:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX900:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_xz() #0 !reqd_work_group_size !5 {
   ; GFX900:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX900:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX900:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX900:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX900:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX900:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX900:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX900:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX900:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX900:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX900:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX900:   S_ENDPGM 0
   ; GFX908-LABEL: name: test_only_workitem_id_xz
   ; GFX908: bb.1 (%ir-block.0):
   ; GFX908:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; GFX908:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GFX908:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GFX908:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GFX908:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; GFX908:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; GFX908:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GFX908:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GFX908:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GFX908:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GFX908:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GFX908:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GFX908:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GFX908:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; GFX908:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; GFX908:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GFX908:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GFX908:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GFX908:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GFX908:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GFX908:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GFX908:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GFX908:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_only_workitem_id_xz() #0 !reqd_work_group_size !5 {
   ; GFX908:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GFX908:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GFX908:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GFX908:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GFX908:   $sgpr12 = COPY [[COPY14]](s32)
   ; GFX908:   $sgpr13 = COPY [[COPY15]](s32)
   ; GFX908:   $sgpr14 = COPY [[COPY16]](s32)
   ; GFX908:   $vgpr31 = COPY [[OR1]](s32)
-  ; GFX908:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GFX908:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GFX908:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GFX908:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 declare i32 @llvm.amdgcn.workitem.id.x() #1
 declare i32 @llvm.amdgcn.workitem.id.y() #1

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

 ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
 ; RUN: llc -global-isel -stop-after=irtranslator -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -verify-machineinstrs -o - %s | FileCheck -enable-var-scope %s
 
 ; amdgpu_gfx calling convention
 declare hidden amdgpu_gfx void @external_gfx_void_func_void() #0
 declare hidden amdgpu_gfx void @external_gfx_void_func_i32(i32) #0
 declare hidden amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg) #0
 declare hidden amdgpu_gfx void @external_gfx_void_func_struct_i8_i32({ i8, i32 }) #0
 declare hidden amdgpu_gfx void @external_gfx_void_func_struct_i8_i32_inreg({ i8, i32 } inreg) #0
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_void() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_void
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_void
   ; CHECK:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK:   S_SETPC_B64_return [[COPY2]]
   call amdgpu_gfx void @external_gfx_void_func_void()
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $vgpr0, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32
   ; CHECK:   $vgpr0 = COPY [[C]](s32)
   ; CHECK:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
   ; CHECK:   S_SETPC_B64_return [[COPY3]]
   call amdgpu_gfx void @external_gfx_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm_inreg(i32 inreg) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm_inreg
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr4, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
   ; CHECK:   $sgpr4 = COPY [[C]](s32)
   ; CHECK:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
   ; CHECK:   S_SETPC_B64_return [[COPY3]]
   call amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_struct_i8_i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `{ i8, i32 } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[LOAD]](p1) :: (load (s8) from %ir.ptr0, align 4, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from %ir.ptr0 + 4, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32
   ; CHECK:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK:   $vgpr0 = COPY [[ANYEXT1]](s32)
   ; CHECK:   $vgpr1 = COPY [[LOAD2]](s32)
   ; CHECK:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr0 = load { i8, i32 } addrspace(1)*, { i8, i32 } addrspace(1)* addrspace(4)* undef
   %val = load { i8, i32 }, { i8, i32 } addrspace(1)* %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32({ i8, i32 } %val)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32_inreg() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_struct_i8_i32_inreg
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `{ i8, i32 } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[LOAD]](p1) :: (load (s8) from %ir.ptr0, align 4, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from %ir.ptr0 + 4, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
   ; CHECK:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK:   $sgpr4 = COPY [[ANYEXT1]](s32)
   ; CHECK:   $sgpr5 = COPY [[LOAD2]](s32)
   ; CHECK:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr0 = load { i8, i32 } addrspace(1)*, { i8, i32 } addrspace(1)* addrspace(4)* undef
   %val = load { i8, i32 }, { i8, i32 } addrspace(1)* %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32_inreg({ i8, i32 } inreg %val)
   ret void
 }
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind noinline }

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

define amdgpu_kernel void @test_call_external_i32_func_i32_imm(i32 addrspace(1)* %out) #0 {
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; GCN:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GCN:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GCN:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (p1) from %ir.out.kernarg.offset.cast, align 16, addrspace 4)
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i32_func_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i32_func_i32
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GCN:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i32_func_i32_imm(i32 addrspace(1)* %out) #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i32_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY21]](s32), [[LOAD]](p1) :: (volatile store (s32) into %ir.out.load, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i32 @external_i32_func_i32(i32 42)
   store volatile i32 %val, i32 addrspace(1)* %out
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i32_func_i32_imm(i32 addrspace(1)* %out) #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i32_func_i32_imm
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
   ; GCN:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
   ; GCN:   [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_i32_func_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i32_func_i32
   ; GCN:   $vgpr0 = COPY [[C]](s32)
   ; GCN:   [[COPY3:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY3]](<4 x s32>)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_i32_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i32_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN:   [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY4]](s32), [[MV]](p1) :: (volatile store (s32) into %ir.out, addrspace 1)
   ; GCN:   [[COPY5:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY2]]
   ; GCN:   S_SETPC_B64_return [[COPY5]]
   %val = call amdgpu_gfx i32 @external_gfx_i32_func_i32(i32 42)
   store volatile i32 %val, i32 addrspace(1)* %out
   ret void
define amdgpu_kernel void @test_call_external_i1_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i1_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i1_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i1_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC]](s1), [[DEF]](p1) :: (volatile store (s1) into `i1 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i1 @external_i1_func_void()
   store volatile i1 %val, i1 addrspace(1)* undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i1_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i1_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr30_sgpr31
   ; GCN:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_i1_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i1_func_void
   ; GCN:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_i1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[COPY2]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC]](s1), [[DEF]](p1) :: (volatile store (s1) into `i1 addrspace(1)* undef`, addrspace 1)
   ; GCN:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; GCN:   S_SETPC_B64_return [[COPY3]]
   %val = call amdgpu_gfx i1 @external_gfx_i1_func_void()
   store volatile i1 %val, i1 addrspace(1)* undef
define amdgpu_kernel void @test_call_external_i1_zeroext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i1_zeroext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i1_zeroext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i1_zeroext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i1_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i1_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[TRUNC]](s1)
   ; GCN:   G_STORE [[ZEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i1 @external_i1_zeroext_func_void()
   %val.ext = zext i1 %val to i32
define amdgpu_kernel void @test_call_external_i1_signext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i1_signext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i1_signext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i1_signext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i1_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i1_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s1) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[TRUNC]](s1)
   ; GCN:   G_STORE [[SEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i1 @external_i1_signext_func_void()
   %val.ext = sext i1 %val to i32
define amdgpu_kernel void @test_call_external_i8_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i8_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i8_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i8_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i8_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i8_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC1]](s8), [[DEF]](p1) :: (volatile store (s8) into `i8 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i8 @external_i8_func_void()
   store volatile i8 %val, i8 addrspace(1)* undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i8_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i8_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr30_sgpr31
   ; GCN:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_i8_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i8_func_void
   ; GCN:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_i8_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i8_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY2]](s32)
   ; GCN:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC1]](s8), [[DEF]](p1) :: (volatile store (s8) into `i8 addrspace(1)* undef`, addrspace 1)
   ; GCN:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; GCN:   S_SETPC_B64_return [[COPY3]]
   %val = call amdgpu_gfx i8 @external_gfx_i8_func_void()
define amdgpu_kernel void @test_call_external_i8_zeroext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i8_zeroext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i8_zeroext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i8_zeroext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i8_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i8_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[TRUNC1]](s8)
   ; GCN:   G_STORE [[ZEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i8 @external_i8_zeroext_func_void()
define amdgpu_kernel void @test_call_external_i8_signext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i8_signext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i8_signext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i8_signext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i8_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i8_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[TRUNC1]](s8)
   ; GCN:   G_STORE [[SEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i8 @external_i8_signext_func_void()
define amdgpu_kernel void @test_call_external_i16_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC]](s16), [[DEF]](p1) :: (volatile store (s16) into `i16 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i16 @external_i16_func_void()
   store volatile i16 %val, i16 addrspace(1)* undef
   ret void
define amdgpu_kernel void @test_call_external_i16_zeroext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i16_zeroext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i16_zeroext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i16_zeroext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i16_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i16_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[TRUNC]](s16)
   ; GCN:   G_STORE [[ZEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i16 @external_i16_zeroext_func_void()
   %val.ext = zext i16 %val to i32
define amdgpu_kernel void @test_call_external_i16_signext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i16_signext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i16_signext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i16_signext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i16_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i16_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[SEXT:%[0-9]+]]:_(s32) = G_SEXT [[TRUNC]](s16)
   ; GCN:   G_STORE [[SEXT]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i16 @external_i16_signext_func_void()
   %val.ext = sext i16 %val to i32
define amdgpu_kernel void @test_call_external_i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY19]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i32 @external_i32_func_void()
   store volatile i32 %val, i32 addrspace(1)* undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i32_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i32_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr30_sgpr31
   ; GCN:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i32_func_void
   ; GCN:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0
   ; GCN:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY2]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; GCN:   S_SETPC_B64_return [[COPY3]]
   %val = call amdgpu_gfx i32 @external_gfx_i32_func_void()
   store volatile i32 %val, i32 addrspace(1)* undef
   ret void
define amdgpu_kernel void @test_call_external_i48_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i48_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i48_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i48_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i48_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i48_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s48) = G_TRUNC [[MV]](s64)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC]](s48), [[DEF]](p1) :: (volatile store (s48) into `i48 addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i48 @external_i48_func_void()
define amdgpu_kernel void @test_call_external_i48_zeroext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i48_zeroext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i48_zeroext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i48_zeroext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i48_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i48_zeroext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s48) = G_TRUNC [[MV]](s64)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[ZEXT:%[0-9]+]]:_(s64) = G_ZEXT [[TRUNC]](s48)
   ; GCN:   G_STORE [[ZEXT]](s64), [[DEF]](p1) :: (volatile store (s64) into `i64 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_i48_signext_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i48_signext_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i48_signext_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i48_signext_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i48_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i48_signext_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s48) = G_TRUNC [[MV]](s64)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[TRUNC]](s48)
   ; GCN:   G_STORE [[SEXT]](s64), [[DEF]](p1) :: (volatile store (s64) into `i64 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_i64_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i64_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i64_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i64_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[MV]](s64), [[DEF]](p1) :: (volatile store (s64) into `i64 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i64 @external_i64_func_void()
   store volatile i64 %val, i64 addrspace(1)* undef
define amdgpu_kernel void @test_call_external_p1_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_p1_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_p1_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_p1_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_p1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_p1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[MV]](p1), [[DEF]](p1) :: (volatile store (p1) into `i8 addrspace(1)* addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call i8 addrspace(1)* @external_p1_func_void()
   store volatile i8 addrspace(1)* %val, i8 addrspace(1)* addrspace(1)* undef
define amdgpu_kernel void @test_call_external_v2p1_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v2p1_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v2p1_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v2p1_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v2p1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2p1_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   [[MV1:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p1>) = G_BUILD_VECTOR [[MV]](p1), [[MV1]](p1)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
define amdgpu_kernel void @test_call_external_p3_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p3) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_p3_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_p3_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_p3_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_p3_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_p3_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(p3) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY19]](p3), [[DEF]](p3) :: (volatile store (p3) into `i8 addrspace(3)* addrspace(3)* undef`, addrspace 3)
   ; GCN:   S_ENDPGM 0
   %val = call i8 addrspace(3)* @external_p3_func_void()
   store volatile i8 addrspace(3)* %val, i8 addrspace(3)* addrspace(3)* undef
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_v2p3_func_void() #0 {
   ; GCN-LABEL: name: test_call_external_v2p3_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9
   ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p3) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v2p3_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v2p3_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v2p3_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v2p3_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2p3_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(p3) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(p3) = COPY $vgpr1
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x p3>) = G_BUILD_VECTOR [[COPY19]](p3), [[COPY20]](p3)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<2 x p3>), [[DEF]](p3) :: (volatile store (<2 x p3>) into `<2 x i8 addrspace(3)*> addrspace(3)* undef`, addrspace 3)
   ; GCN:   S_ENDPGM 0
   %val = call <2 x i8 addrspace(3)*> @external_v2p3_func_void()
   store volatile <2 x i8 addrspace(3)*> %val, <2 x i8 addrspace(3)*> addrspace(3)* undef
define amdgpu_kernel void @test_call_external_f16_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_f16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_f16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_f16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[TRUNC]](s16), [[DEF]](p1) :: (volatile store (s16) into `half addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call half @external_f16_func_void()
   store volatile half %val, half addrspace(1)* undef
   ret void
define amdgpu_kernel void @test_call_external_f32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_f32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_f32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_f32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_f32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_f32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY19]](s32), [[DEF]](p1) :: (volatile store (s32) into `float addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call float @external_f32_func_void()
   store volatile float %val, float addrspace(1)* undef
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_f64_func_void() #0 {
   ; GCN-LABEL: name: test_call_external_f64_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9
   ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_f64_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_f64_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_f64_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_f64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_f64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[MV]](s64), [[DEF]](p1) :: (volatile store (s64) into `double addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call double @external_f64_func_void()
   store volatile double %val, double addrspace(1)* undef
define amdgpu_kernel void @test_call_external_v2f64_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v2f64_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v2f64_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v2f64_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v2f64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2f64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[MV]](s64), [[MV1]](s64)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
define amdgpu_kernel void @test_call_external_v2i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v2i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v2i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v2i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v2i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY19]](s32), [[COPY20]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<2 x s32>), [[DEF]](p1) :: (volatile store (<2 x s32>) into `<2 x i32> addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <2 x i32> @external_v2i32_func_void()
   store volatile <2 x i32> %val, <2 x i32> addrspace(1)* undef
define amdgpu_kernel void @test_call_external_v3i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v3i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v3i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v3i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v3i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<3 x s32>), [[DEF]](p1) :: (volatile store (<3 x s32>) into `<3 x i32> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <3 x i32> @external_v3i32_func_void()
define amdgpu_kernel void @test_call_external_v4i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v4i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v4i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v4i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v4i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v4i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<4 x s32>), [[DEF]](p1) :: (volatile store (<4 x s32>) into `<4 x i32> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v5i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v5i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v5i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v5i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v5i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v5i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<5 x s32>), [[DEF]](p1) :: (volatile store (<5 x s32>) into `<5 x i32> addrspace(1)* undef`, align 8, addrspace 1)
define amdgpu_kernel void @test_call_external_v8i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v8i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v8i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v8i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v8i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v8i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr5
   ; GCN:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr6
   ; GCN:   [[COPY26:%[0-9]+]]:_(s32) = COPY $vgpr7
define amdgpu_kernel void @test_call_external_v16i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v16i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v16i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v16i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v16i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v16i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr5
   ; GCN:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr6
   ; GCN:   [[COPY26:%[0-9]+]]:_(s32) = COPY $vgpr7
define amdgpu_kernel void @test_call_external_v32i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v32i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v32i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v32i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v32i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15, implicit-def $vgpr16, implicit-def $vgpr17, implicit-def $vgpr18, implicit-def $vgpr19, implicit-def $vgpr20, implicit-def $vgpr21, implicit-def $vgpr22, implicit-def $vgpr23, implicit-def $vgpr24, implicit-def $vgpr25, implicit-def $vgpr26, implicit-def $vgpr27, implicit-def $vgpr28, implicit-def $vgpr29, implicit-def $vgpr30, implicit-def $vgpr31
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v32i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4, implicit-def $vgpr5, implicit-def $vgpr6, implicit-def $vgpr7, implicit-def $vgpr8, implicit-def $vgpr9, implicit-def $vgpr10, implicit-def $vgpr11, implicit-def $vgpr12, implicit-def $vgpr13, implicit-def $vgpr14, implicit-def $vgpr15, implicit-def $vgpr16, implicit-def $vgpr17, implicit-def $vgpr18, implicit-def $vgpr19, implicit-def $vgpr20, implicit-def $vgpr21, implicit-def $vgpr22, implicit-def $vgpr23, implicit-def $vgpr24, implicit-def $vgpr25, implicit-def $vgpr26, implicit-def $vgpr27, implicit-def $vgpr28, implicit-def $vgpr29, implicit-def $vgpr30, implicit-def $vgpr31
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN:   [[COPY24:%[0-9]+]]:_(s32) = COPY $vgpr5
   ; GCN:   [[COPY25:%[0-9]+]]:_(s32) = COPY $vgpr6
   ; GCN:   [[COPY26:%[0-9]+]]:_(s32) = COPY $vgpr7
define amdgpu_kernel void @test_call_external_v2i16_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v2i16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v2i16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v2i16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v2i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY19]](<2 x s16>), [[DEF]](p1) :: (volatile store (<2 x s16>) into `<2 x i16> addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <2 x i16> @external_v2i16_func_void()
   store volatile <2 x i16> %val, <2 x i16> addrspace(1)* undef
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_v3i16_func_void() #0 {
   ; GCN-LABEL: name: test_call_external_v3i16_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9
   ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v3i16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v3i16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v3i16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v3i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
   ; GCN:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY19]](<2 x s16>), [[COPY20]](<2 x s16>), [[DEF1]](<2 x s16>)
   ; GCN:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[UV]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v4i16_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v4i16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v4i16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v4i16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v4i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v4i16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY19]](<2 x s16>), [[COPY20]](<2 x s16>)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](p1) :: (volatile store (<4 x s16>) into `<4 x i16> addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <4 x i16> @external_v4i16_func_void()
   store volatile <4 x i16> %val, <4 x i16> addrspace(1)* undef
define amdgpu_kernel void @test_call_external_v2f16_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v2f16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v2f16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v2f16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v2f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v2f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY19:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY19]](<2 x s16>), [[DEF]](p1) :: (volatile store (<2 x s16>) into `<2 x half> addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <2 x half> @external_v2f16_func_void()
   store volatile <2 x half> %val, <2 x half> addrspace(1)* undef
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_v3f16_func_void() #0 {
   ; GCN-LABEL: name: test_call_external_v3f16_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9
   ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v3f16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v3f16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v3f16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v3f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN:   [[DEF1:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
   ; GCN:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY19]](<2 x s16>), [[COPY20]](<2 x s16>), [[DEF1]](<2 x s16>)
   ; GCN:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[UV]](<3 x s16>), [[DEF]](p1) :: (volatile store (<3 x s16>) into `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v4f16_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v4f16_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v4f16_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v4f16_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v4f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v4f16_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr1
   ; GCN:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY19]](<2 x s16>), [[COPY20]](<2 x s16>)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[CONCAT_VECTORS]](<4 x s16>), [[DEF]](p1) :: (volatile store (<4 x s16>) into `<4 x half> addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <4 x half> @external_v4f16_func_void()
   store volatile <4 x half> %val, <4 x half> addrspace(1)* undef
define amdgpu_kernel void @test_call_external_v3f32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v3f32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v3f32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v3f32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v3f32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v3f32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<3 x s32>), [[DEF]](p1) :: (volatile store (<3 x s32>) into `<3 x float> addrspace(1)* undef`, align 16, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <3 x float> @external_v3f32_func_void()
define amdgpu_kernel void @test_call_external_v5f32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v5f32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v5f32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v5f32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v5f32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v5f32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr3
   ; GCN:   [[COPY23:%[0-9]+]]:_(s32) = COPY $vgpr4
   ; GCN:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[COPY19]](s32), [[COPY20]](s32), [[COPY21]](s32), [[COPY22]](s32), [[COPY23]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[BUILD_VECTOR]](<5 x s32>), [[DEF]](p1) :: (volatile store (<5 x s32>) into `<5 x float> addrspace(1)* undef`, align 32, addrspace 1)
define amdgpu_kernel void @test_call_external_i32_i64_func_void() #0 {
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   [[COPY9:%[0-9]+]]:_(p1) = COPY [[DEF]](p1)
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i32_i64_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i32_i64_func_void
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i32_i64_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY14]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY15]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i32_i64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_i64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY21]](s32), [[COPY22]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY20]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   G_STORE [[MV]](s64), [[COPY9]](p1) :: (volatile store (s64) into `i64 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call { i32, i64 } @external_i32_i64_func_void()
   %val.0 = extractvalue { i32, i64 } %val, 0
   %val.1 = extractvalue { i32, i64 } %val, 1
   store volatile i32 %val.0, i32 addrspace(1)* undef
   store volatile i64 %val.1, i64 addrspace(1)* undef
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_i32_i64_func_void() #0 {
   ; GCN-LABEL: name: test_gfx_call_external_i32_i64_func_void
   ; GCN: bb.1 (%ir-block.0):
   ; GCN:   liveins: $sgpr30_sgpr31
   ; GCN:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   [[COPY1:%[0-9]+]]:_(p1) = COPY [[DEF]](p1)
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_i32_i64_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_i32_i64_func_void
   ; GCN:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_i32_i64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_i32_i64_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2
   ; GCN:   [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY4:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY5:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[COPY4]](s32), [[COPY5]](s32)
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY3]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   G_STORE [[MV]](s64), [[COPY1]](p1) :: (volatile store (s64) into `i64 addrspace(1)* undef`, addrspace 1)
   ; GCN:   [[COPY6:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
define amdgpu_kernel void @test_call_external_a2i32_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_a2i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_a2i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_a2i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_a2i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_a2i32_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   G_STORE [[COPY19]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   G_STORE [[COPY20]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call [2 x i32] @external_a2i32_func_void()
   %val.0 = extractvalue [2 x i32] %val, 0
define amdgpu_kernel void @test_call_external_a5i8_func_void() #0 {
   ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_a5i8_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_a5i8_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_a5i8_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_a5i8_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_a5i8_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0, implicit-def $vgpr1, implicit-def $vgpr2, implicit-def $vgpr3, implicit-def $vgpr4
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY19]](s32)
   ; GCN:   [[TRUNC1:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC]](s16)
   ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[TRUNC2:%[0-9]+]]:_(s16) = G_TRUNC [[COPY20]](s32)
   ; GCN:   [[TRUNC3:%[0-9]+]]:_(s8) = G_TRUNC [[TRUNC2]](s16)
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[TRUNC4:%[0-9]+]]:_(s16) = G_TRUNC [[COPY21]](s32)
define amdgpu_kernel void @test_call_external_v32i32_i32_func_void() #0 {
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   [[COPY9:%[0-9]+]]:_(p1) = COPY [[DEF]](p1)
   ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v32i32_i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v32i32_i32_func_void
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v32i32_i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY14]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY15]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v32i32_i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v32i32_i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[LOAD:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load (<32 x s32>) from %stack.0, addrspace 5)
   ; GCN:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
   ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C4]](s32)
   ; GCN:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load (s32) from %stack.0, align 128, addrspace 5)
   ; GCN:   G_STORE [[LOAD]](<32 x s32>), [[DEF]](p1) :: (volatile store (<32 x s32>) into `<32 x i32> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   G_STORE [[LOAD1]](s32), [[COPY9]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_i32_v32i32_func_void() #0 {
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   [[COPY9:%[0-9]+]]:_(p1) = COPY [[DEF]](p1)
   ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i32_v32i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_i32_v32i32_func_void
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_i32_v32i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY14]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY15]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i32_v32i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_i32_v32i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[FRAME_INDEX]](p5) :: (load (s32) from %stack.0, align 128, addrspace 5)
   ; GCN:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
   ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C4]](s32)
   ; GCN:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[PTR_ADD1]](p5) :: (load (<32 x s32>) from %stack.0, addrspace 5)
   ; GCN:   G_STORE [[LOAD]](s32), [[DEF]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   G_STORE [[LOAD1]](<32 x s32>), [[COPY9]](p1) :: (volatile store (<32 x s32>) into `<32 x i32> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
define amdgpu_kernel void @test_call_external_v33i32_func_void() #0 {
   ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v33i32_func_void
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v33i32_func_void
   ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[C:%[0-9]+]]:_(p4) = G_CONSTANT i64 0
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[C]], [[C1]](s64)
   ; GCN:   [[COPY11:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v33i32_func_void() #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY11]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY12]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY13]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY14]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v33i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v33i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[LOAD:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load (<33 x s32>) from %stack.0, align 256, addrspace 5)
   ; GCN:   G_STORE [[LOAD]](<33 x s32>), [[DEF]](p1) :: (volatile store (<33 x s32>) into `<33 x i32> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <33 x i32> @external_v33i32_func_void()
   store volatile <33 x i32> %val, <33 x i32> addrspace(1)* undef, align 8
   ret void
 }
define amdgpu_kernel void @test_call_external_v33i32_func_v33i32_i32(<33 x i32> addrspace(1)* %p, i32 %idx) #0 {
   ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; GCN:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GCN:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (p1) from %ir.p.kernarg.offset.cast, align 16, addrspace 4)
   ; GCN:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[INT]], [[C]](s64)
   ; GCN:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (dereferenceable invariant load (s32) from %ir.idx.kernarg.offset.cast, align 8, addrspace 4)
   ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v33i32_func_v33i32_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_v33i32_func_v33i32_i32
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
   ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; GCN:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_v33i32_func_v33i32_i32(<33 x i32> addrspace(1)* %p, i32 %idx) #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v33i32_func_v33i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_v33i32_func_v33i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[LOAD2:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load (<33 x s32>) from %stack.0, align 256, addrspace 5)
   ; GCN:   G_STORE [[LOAD2]](<33 x s32>), [[DEF]](p1) :: (volatile store (<33 x s32>) into `<33 x i32> addrspace(1)* undef`, align 8, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %val = call <33 x i32> @external_v33i32_func_v33i32_i32(<33 x i32> addrspace(1)* %p, i32 %idx)
   store volatile <33 x i32> %val, <33 x i32> addrspace(1)* undef, align 8
   ret void
 }
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind noinline }

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

define amdgpu_kernel void @test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32(i32) #0 {
   ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0.in.val
   ; GCN:   [[FRAME_INDEX1:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.1.out.val
   ; GCN:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GCN:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C2]](s32)
   ; GCN:   G_STORE [[C]](s8), [[FRAME_INDEX]](p5) :: (store (s8) into %ir.in.gep01, addrspace 5)
   ; GCN:   G_STORE [[C1]](s32), [[PTR_ADD]](p5) :: (store (s32) into %ir.in.gep1, addrspace 5)
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GCN:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
   ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY13]], [[C3]](s64)
   ; GCN:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32(i32) #0 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY15]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY16]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY17]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_sret_struct_i8_i32_byval_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; GCN:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; GCN:   [[PTR_ADD3:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX1]], [[C2]](s32)
   ; GCN:   [[LOAD:%[0-9]+]]:_(s8) = G_LOAD [[FRAME_INDEX1]](p5) :: (dereferenceable load (s8) from %ir.out.gep02, addrspace 5)
   ; GCN:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD3]](p5) :: (dereferenceable load (s32) from %ir.out.gep1, addrspace 5)
   ; GCN:   G_STORE [[LOAD]](s8), [[DEF]](p1) :: (volatile store (s8) into `i8 addrspace(1)* undef`, addrspace 1)
   ; GCN:   G_STORE [[LOAD1]](s32), [[COPY10]](p1) :: (volatile store (s32) into `i32 addrspace(1)* undef`, addrspace 1)
   ; GCN:   S_ENDPGM 0
   %in.val = alloca { i8, i32 }, align 4, addrspace(5)

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

define amdgpu_kernel void @test_call_external_void_func_void() #0 {
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_void
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_void
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_void() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_void()
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_void() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_void
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_void
   ; CHECK:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK:   S_SETPC_B64_return [[COPY2]]
   call amdgpu_gfx void @external_gfx_void_func_void()
   ret void
 }
 
 define void @test_func_call_external_void_func_void() #0 {
   ; CHECK-LABEL: name: test_func_call_external_void_func_void
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr12, $sgpr13, $sgpr14, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr31
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr_32 = COPY $sgpr13
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_void
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_void
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK:   [[COPY17:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY17]](<4 x s32>)
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_void, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY18:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY18]]
   call void @external_void_func_void()
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_empty_struct() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_empty_struct
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_empty_struct
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_empty_struct
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_empty_struct() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_empty_struct, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_empty_struct, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_empty_struct({} zeroinitializer, i32 23)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_empty_array() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_empty_array
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 23
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_empty_array
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_empty_array
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_empty_array() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_empty_array, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_empty_array, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_empty_array([0 x i8] zeroinitializer, i32 23)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i1_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s1) = G_CONSTANT i1 true
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i1
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i1
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i1, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i1, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_i1(i1 true)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i1_signext(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i1_signext
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s1) = G_LOAD [[DEF]](p1) :: (volatile load (s1) from `i1 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i1_signext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i1_signext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i1_signext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i1_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i1_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i1, i1 addrspace(1)* undef
   call void @external_void_func_i1_signext(i1 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s1) = G_LOAD [[DEF]](p1) :: (volatile load (s1) from `i1 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i1_zeroext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i1_zeroext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i1_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i1_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i1, i1 addrspace(1)* undef
   call void @external_void_func_i1_zeroext(i1 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i8_imm(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i8_imm
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s8) = G_CONSTANT i8 123
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i8
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i8
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i8_imm(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_i8(i8 123)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i8_signext(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i8_signext
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s8) = G_LOAD [[DEF]](p1) :: (volatile load (s8) from `i8 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i8_signext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i8_signext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i8_signext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i8_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i8_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i8, i8 addrspace(1)* undef
   call void @external_void_func_i8_signext(i8 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s8) = G_LOAD [[DEF]](p1) :: (volatile load (s8) from `i8 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i8_zeroext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i8_zeroext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i8_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i8_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i8, i8 addrspace(1)* undef
   call void @external_void_func_i8_zeroext(i8 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i16_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i16_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 123
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i16_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_i16(i16 123)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i16_signext(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i16_signext
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[DEF]](p1) :: (volatile load (s16) from `i16 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i16_signext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i16_signext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i16_signext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i16_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i16_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i16, i16 addrspace(1)* undef
   call void @external_void_func_i16_signext(i16 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s16) = G_LOAD [[DEF]](p1) :: (volatile load (s16) from `i16 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i16_zeroext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i16_zeroext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i16_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i16_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i16, i16 addrspace(1)* undef
   call void @external_void_func_i16_zeroext(i16 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i32_imm
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $vgpr0, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32
   ; CHECK:   $vgpr0 = COPY [[C]](s32)
   ; CHECK:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
   ; CHECK:   S_SETPC_B64_return [[COPY3]]
   call amdgpu_gfx void @external_gfx_void_func_i32(i32 42)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_i32_imm_inreg(i32 inreg) #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_i32_imm_inreg
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr4, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr4
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 42
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_i32_inreg
   ; CHECK:   $sgpr4 = COPY [[C]](s32)
   ; CHECK:   [[COPY2:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY3:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
   ; CHECK:   S_SETPC_B64_return [[COPY3]]
   call amdgpu_gfx void @external_gfx_void_func_i32_inreg(i32 inreg 42)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i64_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 123
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i64_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_i64(i64 123)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i64() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2i64
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(p1) = G_CONSTANT i64 0
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[C]](p1) :: (load (<2 x s64>) from `<2 x i64> addrspace(1)* null`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2i64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2i64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2i64() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <2 x i64>, <2 x i64> addrspace(1)* null
   call void @external_void_func_v2i64(<2 x i64> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8589934593
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 17179869187
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C1]](s64)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2i64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2i64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C2]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v2i64(<2 x i64> <i64 8589934593, i64 17179869187>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i48(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_i48
   ; CHECK: bb.1 (%ir-block.1):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s48) = G_LOAD [[DEF]](p1) :: (volatile load (s48) from `i48 addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i48
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i48
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i48(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i48, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i48, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i48, i48 addrspace(1)* undef
   call void @external_void_func_i48(i48 %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i48_signext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i48_signext(i32) #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s48) = G_LOAD [[DEF]](p1) :: (volatile load (s48) from `i48 addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i48_signext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i48_signext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i48_signext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i48_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i48_signext, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i48, i48 addrspace(1)* undef
   call void @external_void_func_i48_signext(i48 signext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_i48_zeroext(i32) #0 {
define amdgpu_kernel void @test_call_external_void_func_i48_zeroext(i32) #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(s48) = G_LOAD [[DEF]](p1) :: (volatile load (s48) from `i48 addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_i48_zeroext
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_i48_zeroext
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_i48_zeroext(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_i48_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_i48_zeroext, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %var = load volatile i48, i48 addrspace(1)* undef
   call void @external_void_func_i48_zeroext(i48 zeroext %var)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_p0_imm(i8* %arg) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_p0_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p0) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (p0) from %ir.arg.kernarg.offset.cast, align 16, addrspace 4)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_p0
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_p0
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_p0_imm(i8* %arg) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_p0, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_p0, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_p0(i8* %arg)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2p0() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2p0
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(p1) = G_CONSTANT i64 0
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x p0>) = G_LOAD [[C]](p1) :: (load (<2 x p0>) from `<2 x i8*> addrspace(1)* null`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2p0
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2p0
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2p0() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2p0, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2p0, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <2 x i8*>, <2 x i8*> addrspace(1)* null
   call void @external_void_func_v2p0(<2 x i8*> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 {
define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(p1) = G_CONSTANT i64 0
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8589934593
   ; CHECK:   [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C1]](s64), [[DEF]](s64)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[C]](p1) :: (load (<2 x s64>) from `<2 x i64> addrspace(1)* null`, addrspace 1)
   ; CHECK:   [[SHUF:%[0-9]+]]:_(<3 x s64>) = G_SHUFFLE_VECTOR [[LOAD]](<2 x s64>), [[BUILD_VECTOR]], shufflemask(0, 1, 2)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3i64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3i64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C2]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %load = load <2 x i64>, <2 x i64> addrspace(1)* null
   %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 undef>, <3 x i32> <i32 0, i32 1, i32 2>
 
   call void @external_void_func_v3i64(<3 x i64> %val)
   ret void
 }
define amdgpu_kernel void @test_call_external_void_func_v4i64() #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(p1) = G_CONSTANT i64 0
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 8589934593
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 17179869187
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C1]](s64), [[C2]](s64)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x s64>) = G_LOAD [[C]](p1) :: (load (<2 x s64>) from `<2 x i64> addrspace(1)* null`, addrspace 1)
   ; CHECK:   [[SHUF:%[0-9]+]]:_(<4 x s64>) = G_SHUFFLE_VECTOR [[LOAD]](<2 x s64>), [[BUILD_VECTOR]], shufflemask(0, 1, 2, 3)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v4i64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v4i64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C3]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v4i64() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v4i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %load = load <2 x i64>, <2 x i64> addrspace(1)* null
   %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 17179869187>, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   call void @external_void_func_v4i64(<4 x i64> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_f16_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_f16_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s16) = G_FCONSTANT half 0xH4400
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_f16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_f16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_f16_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_f16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_f16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_f16(half 4.0)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_f32_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 4.000000e+00
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_f32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_f32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_f32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_f32(float 4.0)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2f32_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 2.000000e+00
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2f32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2f32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C2]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2f32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v2f32(<2 x float> <float 1.0, float 2.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3f32_imm
define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 {
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 2.000000e+00
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_FCONSTANT float 4.000000e+00
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3f32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3f32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C3]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v3f32(<3 x float> <float 1.0, float 2.0, float 4.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v5f32_imm
define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.000000e+00
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_FCONSTANT float 2.000000e+00
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_FCONSTANT float 4.000000e+00
   ; CHECK:   [[C3:%[0-9]+]]:_(s32) = G_FCONSTANT float -1.000000e+00
   ; CHECK:   [[C4:%[0-9]+]]:_(s32) = G_FCONSTANT float 5.000000e-01
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32), [[C3]](s32), [[C4]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v5f32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v5f32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C5]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v5f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v5f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v5f32(<5 x float> <float 1.0, float 2.0, float 4.0, float -1.0, float 0.5>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_f64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_f64_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 4.000000e+00
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_f64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_f64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_f64_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_f64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_f64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_f64(double 4.0)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2f64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2f64_imm
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 2.000000e+00
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 4.000000e+00
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C1]](s64)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2f64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2f64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C2]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2f64_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2f64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2f64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v2f64(<2 x double> <double 2.0, double 4.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3f64_imm
define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 {
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_FCONSTANT double 2.000000e+00
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_FCONSTANT double 4.000000e+00
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_FCONSTANT double 8.000000e+00
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s64>) = G_BUILD_VECTOR [[C]](s64), [[C1]](s64), [[C2]](s64)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3f64
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3f64
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C3]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3f64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3f64, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v3f64(<3 x double> <double 2.0, double 4.0, double 8.0>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[DEF]](p1) :: (load (<2 x s16>) from `<2 x i16> addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <2 x i16>, <2 x i16> addrspace(1)* undef
   call void @external_void_func_v2i16(<2 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<3 x s16>) = G_LOAD [[DEF]](p1) :: (load (<3 x s16>) from `<3 x i16> addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <3 x i16>, <3 x i16> addrspace(1)* undef
   call void @external_void_func_v3i16(<3 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3f16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<3 x s16>) = G_LOAD [[DEF]](p1) :: (load (<3 x s16>) from `<3 x half> addrspace(1)* undef`, align 8, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3f16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3f16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3f16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3f16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <3 x half>, <3 x half> addrspace(1)* undef
   call void @external_void_func_v3f16(<3 x half> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v4i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<4 x s16>) = G_LOAD [[DEF]](p1) :: (load (<4 x s16>) from `<4 x i16> addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v4i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v4i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v4i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <4 x i16>, <4 x i16> addrspace(1)* undef
   call void @external_void_func_v4i16(<4 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 {
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s16) = G_CONSTANT i16 1
   ; CHECK:   [[C1:%[0-9]+]]:_(s16) = G_CONSTANT i16 2
   ; CHECK:   [[C2:%[0-9]+]]:_(s16) = G_CONSTANT i16 3
   ; CHECK:   [[C3:%[0-9]+]]:_(s16) = G_CONSTANT i16 4
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s16>) = G_BUILD_VECTOR [[C]](s16), [[C1]](s16), [[C2]](s16), [[C3]](s16)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v4i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v4i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C4]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v4i16(<4 x i16> <i16 1, i16 2, i16 3, i16 4>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v5i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v5i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<5 x s16>) = G_LOAD [[DEF]](p1) :: (load (<5 x s16>) from `<5 x i16> addrspace(1)* undef`, align 16, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v5i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v5i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v5i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v5i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v5i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <5 x i16>, <5 x i16> addrspace(1)* undef
   call void @external_void_func_v5i16(<5 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v7i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v7i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<7 x s16>) = G_LOAD [[DEF]](p1) :: (load (<7 x s16>) from `<7 x i16> addrspace(1)* undef`, align 16, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v7i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v7i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v7i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v7i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v7i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <7 x i16>, <7 x i16> addrspace(1)* undef
   call void @external_void_func_v7i16(<7 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v63i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v63i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<63 x s16>) = G_LOAD [[DEF]](p1) :: (load (<63 x s16>) from `<63 x i16> addrspace(1)* undef`, align 128, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v63i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v63i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v63i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v63i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v63i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <63 x i16>, <63 x i16> addrspace(1)* undef
   call void @external_void_func_v63i16(<63 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v65i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v65i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<65 x s16>) = G_LOAD [[DEF]](p1) :: (load (<65 x s16>) from `<65 x i16> addrspace(1)* undef`, align 256, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v65i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v65i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v65i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v65i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v65i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <65 x i16>, <65 x i16> addrspace(1)* undef
   call void @external_void_func_v65i16(<65 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v66i16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v66i16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<66 x s16>) = G_LOAD [[DEF]](p1) :: (load (<66 x s16>) from `<66 x i16> addrspace(1)* undef`, align 256, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v66i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v66i16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v66i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v66i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v66i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <66 x i16>, <66 x i16> addrspace(1)* undef
   call void @external_void_func_v66i16(<66 x i16> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2f16() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2f16
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x s16>) = G_LOAD [[DEF]](p1) :: (load (<2 x s16>) from `<2 x half> addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2f16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2f16
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2f16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2f16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2f16, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <2 x half>, <2 x half> addrspace(1)* undef
   call void @external_void_func_v2f16(<2 x half> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v2i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[DEF]](p1) :: (load (<2 x s32>) from `<2 x i32> addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <2 x i32>, <2 x i32> addrspace(1)* undef
   call void @external_void_func_v2i32(<2 x i32> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C2]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v2i32(<2 x i32> <i32 1, i32 2>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3i32_imm
define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 {
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32)
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C3]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v3i32(<3 x i32> <i32 3, i32 4, i32 5>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v3i32_i32
define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32)
   ; CHECK:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3i32_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3i32_i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C4]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v3i32_i32(<3 x i32> <i32 3, i32 4, i32 5>, i32 6)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v4i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<4 x s32>) = G_LOAD [[DEF]](p1) :: (load (<4 x s32>) from `<4 x i32> addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v4i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v4i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v4i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = load <4 x i32>, <4 x i32> addrspace(1)* undef
   call void @external_void_func_v4i32(<4 x i32> %val)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 {
define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 {
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
   ; CHECK:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32), [[C3]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v4i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v4i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C4:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C4]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v4i32(<4 x i32> <i32 1, i32 2, i32 3, i32 4>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v5i32_imm
define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 2
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
   ; CHECK:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32), [[C3]](s32), [[C4]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v5i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v5i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C5:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C5]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v5i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v5i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v5i32(<5 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v8i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v8i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<8 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<8 x s32>) = G_LOAD [[LOAD]](p1) :: (load (<8 x s32>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v8i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v8i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v8i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <8 x i32> addrspace(1)*, <8 x i32> addrspace(1)* addrspace(4)* undef
   %val = load <8 x i32>, <8 x i32> addrspace(1)* %ptr
   call void @external_void_func_v8i32(<8 x i32> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v8i32_imm() #0 {
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 3
   ; CHECK:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK:   [[C4:%[0-9]+]]:_(s32) = G_CONSTANT i32 5
   ; CHECK:   [[C5:%[0-9]+]]:_(s32) = G_CONSTANT i32 6
   ; CHECK:   [[C6:%[0-9]+]]:_(s32) = G_CONSTANT i32 7
   ; CHECK:   [[C7:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C1]](s32), [[C2]](s32), [[C3]](s32), [[C4]](s32), [[C5]](s32), [[C6]](s32), [[C7]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v8i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v8i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C8:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C8]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v8i32_imm() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v8i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void @external_void_func_v8i32(<8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v16i32() #0 {
   ; CHECK-LABEL: name: test_call_external_void_func_v16i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<16 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<16 x s32>) = G_LOAD [[LOAD]](p1) :: (load (<16 x s32>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v16i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v16i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v16i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v16i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v16i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <16 x i32> addrspace(1)*, <16 x i32> addrspace(1)* addrspace(4)* undef
   %val = load <16 x i32>, <16 x i32> addrspace(1)* %ptr
   call void @external_void_func_v16i32(<16 x i32> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v32i32() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<32 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[LOAD]](p1) :: (load (<32 x s32>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v32i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <32 x i32> addrspace(1)*, <32 x i32> addrspace(1)* addrspace(4)* undef
   %val = load <32 x i32>, <32 x i32> addrspace(1)* %ptr
   call void @external_void_func_v32i32(<32 x i32> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v32i32_i32(i32) #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<32 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[LOAD]](p1) :: (load (<32 x s32>) from %ir.ptr0, addrspace 1)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[DEF1]](p1) :: (load (s32) from `i32 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32_i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v32i32_i32(i32) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr0 = load <32 x i32> addrspace(1)*, <32 x i32> addrspace(1)* addrspace(4)* undef
   %val0 = load <32 x i32>, <32 x i32> addrspace(1)* %ptr0
   %val1 = load i32, i32 addrspace(1)* undef
   call void @external_void_func_v32i32_i32(<32 x i32> %val0, i32 %val1)
   ret void
 }
define amdgpu_kernel void @test_call_external_void_func_v32i32_i8_i8_i16() #0 {
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p1) = COPY [[DEF1]](p1)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<32 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[LOAD]](p1) :: (load (<32 x s32>) from %ir.ptr0, addrspace 1)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s8) = G_LOAD [[DEF1]](p1) :: (load (s8) from `i8 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   [[LOAD3:%[0-9]+]]:_(s16) = G_LOAD [[COPY10]](p1) :: (load (s16) from `i16 addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32_i8_i8_i16
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32_i8_i8_i16
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY13]], [[C]](s64)
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v32i32_i8_i8_i16() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY16]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY17]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32_i8_i8_i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32_i8_i8_i16, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 16, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr0 = load <32 x i32> addrspace(1)*, <32 x i32> addrspace(1)* addrspace(4)* undef
   %val0 = load <32 x i32>, <32 x i32> addrspace(1)* %ptr0
   %val1 = load i8, i8 addrspace(1)* undef
   %val2 = load i8, i8 addrspace(1)* undef
   %val3 = load i16, i16 addrspace(1)* undef
   call void @external_void_func_v32i32_i8_i8_i16(<32 x i32> %val0, i8 %val1, i8 %val2, i16 %val3)
define amdgpu_kernel void @test_call_external_void_func_v32i32_p3_p5() #0 {
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[DEF1:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p1) = COPY [[DEF1]](p1)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<32 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[LOAD]](p1) :: (load (<32 x s32>) from %ir.ptr0, addrspace 1)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(p3) = G_LOAD [[DEF1]](p1) :: (load (p3) from `i8 addrspace(3)* addrspace(1)* undef`, addrspace 1)
   ; CHECK:   [[LOAD3:%[0-9]+]]:_(p5) = G_LOAD [[COPY10]](p1) :: (load (p5) from `i8 addrspace(5)* addrspace(1)* undef`, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v32i32_p3_p5
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v32i32_p3_p5
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY13]], [[C]](s64)
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v32i32_p3_p5() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY16]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY17]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v32i32_p3_p5, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v32i32_p3_p5, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 12, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr0 = load <32 x i32> addrspace(1)*, <32 x i32> addrspace(1)* addrspace(4)* undef
   %val0 = load <32 x i32>, <32 x i32> addrspace(1)* %ptr0
   %val1 = load i8 addrspace(3)*, i8 addrspace(3)* addrspace(1)* undef
   %val2 = load i8 addrspace(5)*, i8 addrspace(5)* addrspace(1)* undef
   call void @external_void_func_v32i32_p3_p5(<32 x i32> %val0, i8 addrspace(3)* %val1, i8 addrspace(5)* %val2)
   ret void
define amdgpu_kernel void @test_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `{ i8, i32 } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[LOAD]](p1) :: (load (s8) from %ir.ptr0, align 4, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from %ir.ptr0 + 4, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_struct_i8_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_struct_i8_i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr0 = load { i8, i32 } addrspace(1)*, { i8, i32 } addrspace(1)* addrspace(4)* undef
   %val = load { i8, i32 }, { i8, i32 } addrspace(1)* %ptr0
   call void @external_void_func_struct_i8_i32({ i8, i32 } %val)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_struct_i8_i32
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `{ i8, i32 } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[LOAD]](p1) :: (load (s8) from %ir.ptr0, align 4, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from %ir.ptr0 + 4, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32
   ; CHECK:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK:   $vgpr0 = COPY [[ANYEXT1]](s32)
   ; CHECK:   $vgpr1 = COPY [[LOAD2]](s32)
   ; CHECK:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr0 = load { i8, i32 } addrspace(1)*, { i8, i32 } addrspace(1)* addrspace(4)* undef
   %val = load { i8, i32 }, { i8, i32 } addrspace(1)* %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32({ i8, i32 } %val)
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_call_external_void_func_struct_i8_i32_inreg() #0 {
   ; CHECK-LABEL: name: test_gfx_call_external_void_func_struct_i8_i32_inreg
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `{ i8, i32 } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s8) = G_LOAD [[LOAD]](p1) :: (load (s8) from %ir.ptr0, align 4, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load (s32) from %ir.ptr0 + 4, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_gfx_void_func_struct_i8_i32_inreg
   ; CHECK:   [[ANYEXT:%[0-9]+]]:_(s16) = G_ANYEXT [[LOAD1]](s8)
   ; CHECK:   [[ANYEXT1:%[0-9]+]]:_(s32) = G_ANYEXT [[ANYEXT]](s16)
   ; CHECK:   $sgpr4 = COPY [[ANYEXT1]](s32)
   ; CHECK:   $sgpr5 = COPY [[LOAD2]](s32)
   ; CHECK:   [[COPY1:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY1]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_gfx_void_func_struct_i8_i32_inreg, csr_amdgpu_highregs, implicit $sgpr4, implicit $sgpr5, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY]]
   ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr0 = load { i8, i32 } addrspace(1)*, { i8, i32 } addrspace(1)* addrspace(4)* undef
   %val = load { i8, i32 }, { i8, i32 } addrspace(1)* %ptr0
   call amdgpu_gfx void @external_gfx_void_func_struct_i8_i32_inreg({ i8, i32 } inreg %val)
   ret void
 }
define amdgpu_kernel void @test_call_external_void_func_byval_struct_i8_i32() #0 {
   ; CHECK:   [[C:%[0-9]+]]:_(s8) = G_CONSTANT i8 3
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 8
   ; CHECK:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0.val
   ; CHECK:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C2]](s32)
   ; CHECK:   G_STORE [[C]](s8), [[FRAME_INDEX]](p5) :: (store (s8) into %ir.gep01, addrspace 5)
   ; CHECK:   G_STORE [[C1]](s32), [[PTR_ADD]](p5) :: (store (s32) into %ir.gep1, addrspace 5)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_byval_struct_i8_i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_byval_struct_i8_i32
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C3]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_byval_struct_i8_i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_byval_struct_i8_i32, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_byval_struct_i8_i32, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 8, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %val = alloca { i8, i32 }, align 4, addrspace(5)
   %gep0 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %val, i32 0, i32 0
   %gep1 = getelementptr inbounds { i8, i32 }, { i8, i32 } addrspace(5)* %val, i32 0, i32 1
   store i8 3, i8 addrspace(5)* %gep0
   store i32 8, i32 addrspace(5)* %gep1
   call void @external_void_func_byval_struct_i8_i32({ i8, i32 } addrspace(5)* byval({ i8, i32 }) %val)
define void @call_byval_3ai32_byval_i8_align32([3 x i32] addrspace(5)* %incoming0, i8 addrspace(5)* align 32 %incoming1) #0 {
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY8:%[0-9]+]]:_(p5) = COPY $vgpr0
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p5) = COPY $vgpr1
   ; CHECK:   [[COPY10:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 999
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @void_func_byval_a3i32_byval_i8_align32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @void_func_byval_a3i32_byval_i8_align32
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @call_byval_3ai32_byval_i8_align32([3 x i32] addrspace(5)* %incoming0, i8 addrspace(5)* align 32 %incoming1) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY13]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY16]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY17]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY18]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @void_func_byval_a3i32_byval_i8_align32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @void_func_byval_a3i32_byval_i8_align32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 36, implicit-def $scc
   ; CHECK:   [[COPY21:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY10]]
   ; CHECK:   S_SETPC_B64_return [[COPY21]]
   call void @void_func_byval_a3i32_byval_i8_align32([3 x i32] addrspace(5)* byval([3 x i32]) %incoming0, i8 addrspace(5)* align 32 %incoming1, i32 999)
   ret void
 }
 
 declare void @void_func_byval_a4i64_align4([4 x i64] addrspace(5)* byval([4 x i64]) align 4 %arg0) #0
define void @call_byval_a4i64_align4_higher_source_align([4 x i64] addrspace(5)* align 256 %incoming_high_align) #0 {
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr12
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY8:%[0-9]+]]:_(p5) = COPY $vgpr0
   ; CHECK:   [[COPY9:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @void_func_byval_a4i64_align4
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @void_func_byval_a4i64_align4
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; CHECK:   [[COPY18:%[0-9]+]]:_(p5) = COPY $sgpr32
   ; CHECK:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY18]], [[C]](s32)
   ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
   ; CHECK:   G_MEMCPY [[PTR_ADD]](p5), [[COPY8]](p5), [[C1]](s32), 0 :: (dereferenceable store (s256) into stack, align 4, addrspace 5), (dereferenceable load (s256) from %ir.incoming_high_align, align 256, addrspace 5)
   ; CHECK:   [[COPY19:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY19]](<4 x s32>)
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY12]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY17]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @void_func_byval_a4i64_align4, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @void_func_byval_a4i64_align4, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 32, implicit-def $scc
   ; CHECK:   [[COPY20:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY9]]
   ; CHECK:   S_SETPC_B64_return [[COPY20]]
   call void @void_func_byval_a4i64_align4([4 x i64] addrspace(5)* byval([4 x i64]) align 4 %incoming_high_align)
   ret void
 }
 
 define amdgpu_kernel void @test_call_external_void_func_v2i8() #0 {
define amdgpu_kernel void @test_call_external_void_func_v2i8() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<2 x i8> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<2 x s8>) = G_LOAD [[LOAD]](p1) :: (load (<2 x s8>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v2i8
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v2i8
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v2i8() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v2i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v2i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <2 x i8> addrspace(1)*, <2 x i8> addrspace(1)* addrspace(4)* undef
   %val = load <2 x i8>, <2 x i8> addrspace(1)* %ptr
   call void @external_void_func_v2i8(<2 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v3i8() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<3 x i8> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<3 x s8>) = G_LOAD [[LOAD]](p1) :: (load (<3 x s8>) from %ir.ptr, align 4, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v3i8
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v3i8
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v3i8() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v3i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v3i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <3 x i8> addrspace(1)*, <3 x i8> addrspace(1)* addrspace(4)* undef
   %val = load <3 x i8>, <3 x i8> addrspace(1)* %ptr
   call void @external_void_func_v3i8(<3 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v4i8() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<4 x i8> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<4 x s8>) = G_LOAD [[LOAD]](p1) :: (load (<4 x s8>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v4i8
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v4i8
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v4i8() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v4i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v4i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <4 x i8> addrspace(1)*, <4 x i8> addrspace(1)* addrspace(4)* undef
   %val = load <4 x i8>, <4 x i8> addrspace(1)* %ptr
   call void @external_void_func_v4i8(<4 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v8i8() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<8 x i8> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<8 x s8>) = G_LOAD [[LOAD]](p1) :: (load (<8 x s8>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v8i8
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v8i8
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v8i8() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v8i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v8i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <8 x i8> addrspace(1)*, <8 x i8> addrspace(1)* addrspace(4)* undef
   %val = load <8 x i8>, <8 x i8> addrspace(1)* %ptr
   call void @external_void_func_v8i8(<8 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @test_call_external_void_func_v16i8() #0 {
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (load (p1) from `<16 x i8> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<16 x s8>) = G_LOAD [[LOAD]](p1) :: (load (<16 x s8>) from %ir.ptr, addrspace 1)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_v16i8
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_v16i8
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @test_call_external_void_func_v16i8() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_v16i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_v16i8, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   %ptr = load <16 x i8> addrspace(1)*, <16 x i8> addrspace(1)* addrspace(4)* undef
   %val = load <16 x i8>, <16 x i8> addrspace(1)* %ptr
   call void @external_void_func_v16i8(<16 x i8> %val)
   ret void
 }
 
define amdgpu_kernel void @stack_passed_arg_alignment_v32i32_f64(<32 x i32> %val, double %tmp) #0 {
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; CHECK:   [[LOAD:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (<32 x s32>) from %ir.val.kernarg.offset.cast, align 16, addrspace 4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 128
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[INT]], [[C]](s64)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s64) = G_LOAD [[PTR_ADD]](p4) :: (dereferenceable invariant load (s64) from %ir.tmp.kernarg.offset.cast, align 16, addrspace 4)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @stack_passed_f64_arg
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @stack_passed_f64_arg
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 136
   ; CHECK:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @stack_passed_arg_alignment_v32i32_f64(<32 x i32> %val, double %tmp) #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @stack_passed_f64_arg, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @stack_passed_f64_arg, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 12, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
 entry:
   call void @stack_passed_f64_arg(<32 x i32> %val, double %tmp)
   ret void
 }
 
 define void @stack_12xv3i32() #0 {
define void @stack_12xv3i32() #0 {
   ; CHECK:   [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
   ; CHECK:   [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
   ; CHECK:   [[BUILD_VECTOR10:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C10]](s32), [[C11]](s32), [[C12]](s32)
   ; CHECK:   [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
   ; CHECK:   [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
   ; CHECK:   [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
   ; CHECK:   [[BUILD_VECTOR11:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C13]](s32), [[C14]](s32), [[C15]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_12xv3i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_12xv3i32
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @stack_12xv3i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_12xv3i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_12xv3i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 20, implicit-def $scc
   ; CHECK:   [[COPY19:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY19]]
 entry:
   call void @external_void_func_12xv3i32(
       <3 x i32> <i32 0, i32 0, i32 0>,
       <3 x i32> <i32 1, i32 1, i32 1>,
       <3 x i32> <i32 2, i32 2, i32 2>,
define void @stack_12xv3f32() #0 {
   ; CHECK:   [[C11:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.100000e+01
   ; CHECK:   [[C12:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.200000e+01
   ; CHECK:   [[BUILD_VECTOR10:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C10]](s32), [[C11]](s32), [[C12]](s32)
   ; CHECK:   [[C13:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.300000e+01
   ; CHECK:   [[C14:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.400000e+01
   ; CHECK:   [[C15:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.500000e+01
   ; CHECK:   [[BUILD_VECTOR11:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[C13]](s32), [[C14]](s32), [[C15]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_12xv3f32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_12xv3f32
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @stack_12xv3f32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_12xv3f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_12xv3f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 20, implicit-def $scc
   ; CHECK:   [[COPY19:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY19]]
 entry:
   call void @external_void_func_12xv3f32(
       <3 x float> <float 0.0, float 0.0, float 0.0>,
       <3 x float> <float 1.0, float 1.0, float 1.0>,
       <3 x float> <float 2.0, float 2.0, float 2.0>,
define void @stack_8xv5i32() #0 {
   ; CHECK:   [[BUILD_VECTOR6:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[C6]](s32), [[C7]](s32), [[C8]](s32), [[C9]](s32), [[C10]](s32)
   ; CHECK:   [[C11:%[0-9]+]]:_(s32) = G_CONSTANT i32 11
   ; CHECK:   [[C12:%[0-9]+]]:_(s32) = G_CONSTANT i32 12
   ; CHECK:   [[C13:%[0-9]+]]:_(s32) = G_CONSTANT i32 13
   ; CHECK:   [[C14:%[0-9]+]]:_(s32) = G_CONSTANT i32 14
   ; CHECK:   [[C15:%[0-9]+]]:_(s32) = G_CONSTANT i32 15
   ; CHECK:   [[BUILD_VECTOR7:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[C11]](s32), [[C12]](s32), [[C13]](s32), [[C14]](s32), [[C15]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_8xv5i32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_8xv5i32
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @stack_8xv5i32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_8xv5i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_8xv5i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 36, implicit-def $scc
   ; CHECK:   [[COPY19:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY19]]
 entry:
   call void @external_void_func_8xv5i32(
       <5 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0>,
       <5 x i32> <i32 1, i32 1, i32 1, i32 1, i32 1>,
       <5 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2>,
define void @stack_8xv5f32() #0 {
   ; CHECK:   [[BUILD_VECTOR6:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[C6]](s32), [[C7]](s32), [[C8]](s32), [[C9]](s32), [[C10]](s32)
   ; CHECK:   [[C11:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.100000e+01
   ; CHECK:   [[C12:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.200000e+01
   ; CHECK:   [[C13:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.300000e+01
   ; CHECK:   [[C14:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.400000e+01
   ; CHECK:   [[C15:%[0-9]+]]:_(s32) = G_FCONSTANT float 1.500000e+01
   ; CHECK:   [[BUILD_VECTOR7:%[0-9]+]]:_(<5 x s32>) = G_BUILD_VECTOR [[C11]](s32), [[C12]](s32), [[C13]](s32), [[C14]](s32), [[C15]](s32)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; CHECK:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_void_func_8xv5f32
+  ; CHECK:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @external_void_func_8xv5f32
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; CHECK:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; CHECK:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define void @stack_8xv5f32() #0 {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY9]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY12]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY13]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY15]](s32)
   ; CHECK:   $vgpr31 = COPY [[COPY16]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_void_func_8xv5f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @external_void_func_8xv5f32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 36, implicit-def $scc
   ; CHECK:   [[COPY19:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY8]]
   ; CHECK:   S_SETPC_B64_return [[COPY19]]
 entry:
   call void @external_void_func_8xv5f32(
       <5 x float> <float 0.0, float 0.0, float 0.0, float 0.0, float 0.0>,
       <5 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>,
       <5 x float> <float 2.0, float 2.0, float 2.0, float 2.0, float 2.0>,

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

define amdgpu_kernel void @test_indirect_call_sgpr_ptr(void()* %fptr) {
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
   ; CHECK:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; CHECK:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; CHECK:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
   ; CHECK:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
-  ; CHECK:   [[LOAD:%[0-9]+]]:sreg_64(p0) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (p0) from %ir.fptr.kernarg.offset.cast, align 16, addrspace 4)
+  ; CHECK:   [[LOAD:%[0-9]+]]:_(p0) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (p0) from %ir.fptr.kernarg.offset.cast, align 16, addrspace 4)
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; CHECK:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; CHECK:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
   ; CHECK:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; CHECK:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
define amdgpu_kernel void @test_indirect_call_sgpr_ptr(void()* %fptr) {
   ; CHECK:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; CHECK:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; CHECK:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
   ; CHECK:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; CHECK:   $sgpr12 = COPY [[COPY14]](s32)
   ; CHECK:   $sgpr13 = COPY [[COPY15]](s32)
   ; CHECK:   $sgpr14 = COPY [[COPY16]](s32)
   ; CHECK:   $vgpr31 = COPY [[OR1]](s32)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[LOAD]](p0), 0, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[LOAD]](p0), 0, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   S_ENDPGM 0
   call void %fptr()
   ret void
 }
 
 define amdgpu_gfx void @test_gfx_indirect_call_sgpr_ptr(void()* %fptr) {
   ; CHECK-LABEL: name: test_gfx_indirect_call_sgpr_ptr
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
   ; CHECK:   [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
-  ; CHECK:   [[MV:%[0-9]+]]:sreg_64(p0) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
+  ; CHECK:   [[MV:%[0-9]+]]:_(p0) = G_MERGE_VALUES [[COPY]](s32), [[COPY1]](s32)
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; CHECK:   ADJCALLSTACKUP 0, 0, implicit-def $scc
   ; CHECK:   [[COPY3:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY3]](<4 x s32>)
-  ; CHECK:   $sgpr30_sgpr31 = SI_CALL [[MV]](p0), 0, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3
+  ; CHECK:   $sgpr30_sgpr31 = G_SI_CALL [[MV]](p0), 0, csr_amdgpu_highregs, implicit $sgpr0_sgpr1_sgpr2_sgpr3
   ; CHECK:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; CHECK:   [[COPY4:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY2]]
   ; CHECK:   S_SETPC_B64_return [[COPY4]]
   call amdgpu_gfx void %fptr()
   ret void
 }

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

define amdgpu_kernel void @kernel_call_i32_fastcc_i32_i32_unused_result(i32 %a, i32 %b, i32 %c) #1 {
   ; GCN:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 1
   ; GCN:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
   ; GCN:   [[LOAD:%[0-9]+]]:_(<2 x s32>) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load (<2 x s32>) from %ir.0, align 16, addrspace 4)
   ; GCN:   [[EVEC:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[LOAD]](<2 x s32>), [[C]](s32)
   ; GCN:   [[EVEC1:%[0-9]+]]:_(s32) = G_EXTRACT_VECTOR_ELT [[LOAD]](<2 x s32>), [[C1]](s32)
   ; GCN:   [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 4
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[INT]], [[C2]](s64)
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32
   ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
   ; GCN:   [[C3:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
   ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C3]](s64)
   ; GCN:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
define amdgpu_kernel void @kernel_call_i32_fastcc_i32_i32_unused_result(i32 %a, i32 %b, i32 %c) #1 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY14]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY15]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY16]](s32)
   ; GCN:   $vgpr31 = COPY [[OR1]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   S_ENDPGM 0
 entry:
   %ret = tail call fastcc i32 @i32_fastcc_i32_i32(i32 %a, i32 %b)
   ret void
 }
 
define fastcc i32 @sibling_call_i32_fastcc_i32_byval_i32_byval_parent(i32 %a, i32 addrspace(5)* byval(i32) %b.byval, i32 %c) #1 {
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %fixed-stack.0
   ; GCN:   [[COPY9:%[0-9]+]]:_(p5) = COPY [[FRAME_INDEX]](p5)
   ; GCN:   [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY11:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_byval_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_byval_i32
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; GCN:   $vgpr0 = COPY [[COPY8]](s32)
   ; GCN:   [[COPY20:%[0-9]+]]:_(p5) = COPY $sgpr32
   ; GCN:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY20]], [[C]](s32)
   ; GCN:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 4
   ; GCN:   G_MEMCPY [[PTR_ADD]](p5), [[COPY9]](p5), [[C1]](s32), 0 :: (dereferenceable store (s32) into stack, addrspace 5), (dereferenceable load (s32) from %ir.b.byval, addrspace 5)
   ; GCN:   [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY12]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY13]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[COPY14]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY15]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY16]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY17]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY18]](s32)
   ; GCN:   $vgpr31 = COPY [[COPY19]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @i32_fastcc_i32_byval_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_byval_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY22:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 4, implicit-def $scc
   ; GCN:   $vgpr0 = COPY [[COPY22]](s32)
   ; GCN:   [[COPY23:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY11]]
   ; GCN:   S_SETPC_B64_return [[COPY23]], implicit $vgpr0
 entry:
   %ret = tail call fastcc i32 @i32_fastcc_i32_byval_i32(i32 %a, i32 addrspace(5)* byval(i32) %b.byval)
   ret i32 %ret
define fastcc i32 @no_sibling_call_callee_more_stack_space(i32 %a, i32 %b) #1 {
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY10:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32_a32i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32_a32i32
   ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
define fastcc i32 @no_sibling_call_callee_more_stack_space(i32 %a, i32 %b) #1 {
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[COPY13]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY15]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY16]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY17]](s32)
   ; GCN:   $vgpr31 = COPY [[COPY18]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @i32_fastcc_i32_i32_a32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32_a32i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $vgpr4, implicit $vgpr5, implicit $vgpr6, implicit $vgpr7, implicit $vgpr8, implicit $vgpr9, implicit $vgpr10, implicit $vgpr11, implicit $vgpr12, implicit $vgpr13, implicit $vgpr14, implicit $vgpr15, implicit $vgpr16, implicit $vgpr17, implicit $vgpr18, implicit $vgpr19, implicit $vgpr20, implicit $vgpr21, implicit $vgpr22, implicit $vgpr23, implicit $vgpr24, implicit $vgpr25, implicit $vgpr26, implicit $vgpr27, implicit $vgpr28, implicit $vgpr29, implicit $vgpr30, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 12, implicit-def $scc
   ; GCN:   $vgpr0 = COPY [[COPY21]](s32)
   ; GCN:   [[COPY22:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY10]]
   ; GCN:   S_SETPC_B64_return [[COPY22]], implicit $vgpr0
 entry:
   %ret = tail call fastcc i32 @i32_fastcc_i32_i32_a32i32(i32 %a, i32 %b, [32 x i32] zeroinitializer)
   ret i32 %ret
define fastcc i32 @sibling_call_i32_fastcc_i32_i32_other_call(i32 %a, i32 %b, i32 %c) #1 {
   ; GCN:   [[COPY5:%[0-9]+]]:sgpr_64 = COPY $sgpr8_sgpr9
   ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
   ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
   ; GCN:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
   ; GCN:   [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
   ; GCN:   [[COPY11:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
   ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
-  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32
+  ; GCN:   [[GV:%[0-9]+]]:_(p0) = G_GLOBAL_VALUE @i32_fastcc_i32_i32
   ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GCN:   [[COPY14:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GCN:   [[COPY15:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY3]]
   ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY2]]
   ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]]
   ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
   ; GCN:   $vgpr0 = COPY [[COPY8]](s32)
   ; GCN:   $vgpr1 = COPY [[COPY9]](s32)
   ; GCN:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
   ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
   ; GCN:   $sgpr4_sgpr5 = COPY [[COPY12]](p4)
   ; GCN:   $sgpr6_sgpr7 = COPY [[COPY13]](p4)
   ; GCN:   $sgpr8_sgpr9 = COPY [[COPY14]](p4)
   ; GCN:   $sgpr10_sgpr11 = COPY [[COPY15]](s64)
   ; GCN:   $sgpr12 = COPY [[COPY16]](s32)
   ; GCN:   $sgpr13 = COPY [[COPY17]](s32)
   ; GCN:   $sgpr14 = COPY [[COPY18]](s32)
   ; GCN:   $vgpr31 = COPY [[COPY19]](s32)
-  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
+  ; GCN:   $sgpr30_sgpr31 = G_SI_CALL [[GV]](p0), @i32_fastcc_i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31, implicit-def $vgpr0
   ; GCN:   [[COPY21:%[0-9]+]]:_(s32) = COPY $vgpr0
   ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
   ; GCN:   [[GV1:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @sibling_call_i32_fastcc_i32_i32
   ; GCN:   [[COPY22:%[0-9]+]]:_(p4) = COPY [[COPY7]]
   ; GCN:   [[COPY23:%[0-9]+]]:_(p4) = COPY [[COPY6]]
   ; GCN:   [[COPY24:%[0-9]+]]:_(p4) = COPY [[COPY5]]
   ; GCN:   [[COPY25:%[0-9]+]]:_(s64) = COPY [[COPY4]]
   ; GCN:   [[COPY26:%[0-9]+]]:_(s32) = COPY [[COPY3]]

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll

; GCN-NEXT:    ; return to shader part epilog
   %v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i64.v4f16(i64 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x half> %ray_dir, <4 x half> %ray_inv_dir, <4 x i32> %tdescr)
   %r = bitcast <4 x i32> %v to <4 x float>
   ret <4 x float> %r
 }
 
 define amdgpu_ps <4 x float> @image_bvh_intersect_ray_vgpr_descr(i32 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x float> %ray_dir, <4 x float> %ray_inv_dir, <4 x i32> %tdescr) {
 ; GFX1030-LABEL: image_bvh_intersect_ray_vgpr_descr:
 ; GFX1030:       ; %bb.0:
-; GFX1030-NEXT:    v_mov_b32_e32 v5, v0
-; GFX1030-NEXT:    v_mov_b32_e32 v9, v1
-; GFX1030-NEXT:    v_mov_b32_e32 v13, v2
-; GFX1030-NEXT:    v_mov_b32_e32 v18, v3
+; GFX1030-NEXT:    v_mov_b32_e32 v18, v0
+; GFX1030-NEXT:    v_mov_b32_e32 v19, v1
+; GFX1030-NEXT:    v_mov_b32_e32 v20, v2
+; GFX1030-NEXT:    v_mov_b32_e32 v21, v3
+; GFX1030-NEXT:    v_mov_b32_e32 v22, v4
+; GFX1030-NEXT:    v_mov_b32_e32 v23, v6
+; GFX1030-NEXT:    v_mov_b32_e32 v24, v7
+; GFX1030-NEXT:    v_mov_b32_e32 v25, v8
+; GFX1030-NEXT:    v_mov_b32_e32 v26, v10
+; GFX1030-NEXT:    v_mov_b32_e32 v27, v11
+; GFX1030-NEXT:    v_mov_b32_e32 v28, v12
 ; GFX1030-NEXT:    s_mov_b32 s1, exec_lo
 ; GFX1030-NEXT:  BB6_1: ; =>This Inner Loop Header: Depth=1
 ; GFX1030-NEXT:    v_readfirstlane_b32 s4, v14
 ; GFX1030-NEXT:    v_readfirstlane_b32 s5, v15
 ; GFX1030-NEXT:    v_readfirstlane_b32 s6, v16
 ; GFX1030-NEXT:    v_readfirstlane_b32 s7, v17
 ; GFX1030-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[14:15]
-; GFX1030-NEXT:    image_bvh_intersect_ray v[0:3], [v5, v9, v13, v18, v4, v6, v7, v8, v10, v11, v12], s[4:7]
 ; GFX1030-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[16:17]
 ; GFX1030-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1030-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1030-NEXT:    image_bvh_intersect_ray v[0:3], v[18:33], s[4:7]
+; GFX1030-NEXT:    ; implicit-def: $vgpr14_vgpr15
+; GFX1030-NEXT:    ; implicit-def: $vgpr18
+; GFX1030-NEXT:    ; implicit-def: $vgpr19
+; GFX1030-NEXT:    ; implicit-def: $vgpr20
+; GFX1030-NEXT:    ; implicit-def: $vgpr21
+; GFX1030-NEXT:    ; implicit-def: $vgpr22
+; GFX1030-NEXT:    ; implicit-def: $vgpr23
+; GFX1030-NEXT:    ; implicit-def: $vgpr24
+; GFX1030-NEXT:    ; implicit-def: $vgpr25
+; GFX1030-NEXT:    ; implicit-def: $vgpr26
+; GFX1030-NEXT:    ; implicit-def: $vgpr27
+; GFX1030-NEXT:    ; implicit-def: $vgpr28
+; GFX1030-NEXT:    ; implicit-def: $vgpr14_vgpr15_vgpr16_vgpr17
 ; GFX1030-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1030-NEXT:    s_cbranch_execnz BB6_1
 ; GFX1030-NEXT:  ; %bb.2:
 ; GFX1030-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1030-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1030-NEXT:    ; return to shader part epilog
 ;
 ; GFX1013-LABEL: image_bvh_intersect_ray_vgpr_descr:
 ; GFX1013:       ; %bb.0:
 ; GFX1013-NEXT:    v_mov_b32_e32 v5, v6
 ; GFX1013-NEXT:    v_mov_b32_e32 v6, v7
 ; GFX1013-NEXT:    v_mov_b32_e32 v7, v8
 ; GFX1013-NEXT:    v_mov_b32_e32 v8, v10
 ; GFX1013-NEXT:    v_mov_b32_e32 v9, v11
 ; GFX1013-NEXT:    v_mov_b32_e32 v10, v12
+; GFX1013-NEXT:    v_mov_b32_e32 v18, v14
+; GFX1013-NEXT:    v_mov_b32_e32 v19, v15
 ; GFX1013-NEXT:    s_mov_b32 s1, exec_lo
 ; GFX1013-NEXT:  BB6_1: ; =>This Inner Loop Header: Depth=1
-; GFX1013-NEXT:    v_readfirstlane_b32 s4, v14
-; GFX1013-NEXT:    v_readfirstlane_b32 s5, v15
+; GFX1013-NEXT:    v_readfirstlane_b32 s4, v18
+; GFX1013-NEXT:    v_readfirstlane_b32 s5, v19
 ; GFX1013-NEXT:    v_readfirstlane_b32 s6, v16
 ; GFX1013-NEXT:    v_readfirstlane_b32 s7, v17
-; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[14:15]
-; GFX1013-NEXT:    image_bvh_intersect_ray v[18:21], v[0:15], s[4:7]
+; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[18:19]
 ; GFX1013-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[16:17]
 ; GFX1013-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1013-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1013-NEXT:    image_bvh_intersect_ray v[20:23], v[0:15], s[4:7]
+; GFX1013-NEXT:    ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
+; GFX1013-NEXT:    ; implicit-def: $vgpr18_vgpr19
+; GFX1013-NEXT:    ; implicit-def: $vgpr14_vgpr15_vgpr16_vgpr17
+; GFX1013-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1013-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1013-NEXT:    s_cbranch_execnz BB6_1
 ; GFX1013-NEXT:  ; %bb.2:
 ; GFX1013-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1013-NEXT:    s_waitcnt vmcnt(0)
-; GFX1013-NEXT:    v_mov_b32_e32 v0, v18
-; GFX1013-NEXT:    v_mov_b32_e32 v1, v19
-; GFX1013-NEXT:    v_mov_b32_e32 v2, v20
-; GFX1013-NEXT:    v_mov_b32_e32 v3, v21
+; GFX1013-NEXT:    v_mov_b32_e32 v0, v20
+; GFX1013-NEXT:    v_mov_b32_e32 v1, v21
+; GFX1013-NEXT:    v_mov_b32_e32 v2, v22
+; GFX1013-NEXT:    v_mov_b32_e32 v3, v23
 ; GFX1013-NEXT:    ; return to shader part epilog
   %v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i32.v4f32(i32 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x float> %ray_dir, <4 x float> %ray_inv_dir, <4 x i32> %tdescr)
   %r = bitcast <4 x i32> %v to <4 x float>
   ret <4 x float> %r
 }
 
 define amdgpu_ps <4 x float> @image_bvh_intersect_ray_a16_vgpr_descr(i32 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x half> %ray_dir, <4 x half> %ray_inv_dir, <4 x i32> %tdescr) {
 ; GFX1030-LABEL: image_bvh_intersect_ray_a16_vgpr_descr:
 ; GFX1030:       ; %bb.0:
 ; GFX1030-NEXT:    s_mov_b32 s0, 0xffff
-; GFX1030-NEXT:    v_mov_b32_e32 v5, v0
+; GFX1030-NEXT:    v_mov_b32_e32 v14, v0
 ; GFX1030-NEXT:    v_lshrrev_b32_e32 v0, 16, v6
-; GFX1030-NEXT:    v_mov_b32_e32 v14, v1
+; GFX1030-NEXT:    v_mov_b32_e32 v15, v1
 ; GFX1030-NEXT:    v_and_b32_e32 v1, s0, v8
-; GFX1030-NEXT:    v_mov_b32_e32 v15, v2
-; GFX1030-NEXT:    v_mov_b32_e32 v16, v3
+; GFX1030-NEXT:    v_mov_b32_e32 v16, v2
+; GFX1030-NEXT:    v_mov_b32_e32 v17, v3
 ; GFX1030-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX1030-NEXT:    v_lshrrev_b32_e32 v2, 16, v8
 ; GFX1030-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX1030-NEXT:    v_and_b32_e32 v3, s0, v9
+; GFX1030-NEXT:    v_mov_b32_e32 v18, v4
+; GFX1030-NEXT:    v_and_or_b32 v19, v6, s0, v0
 ; GFX1030-NEXT:    s_mov_b32 s1, exec_lo
-; GFX1030-NEXT:    v_and_or_b32 v6, v6, s0, v0
-; GFX1030-NEXT:    v_and_or_b32 v7, v7, s0, v1
-; GFX1030-NEXT:    v_lshl_or_b32 v8, v3, 16, v2
+; GFX1030-NEXT:    v_and_or_b32 v20, v7, s0, v1
+; GFX1030-NEXT:    v_lshl_or_b32 v21, v3, 16, v2
 ; GFX1030-NEXT:  BB7_1: ; =>This Inner Loop Header: Depth=1
 ; GFX1030-NEXT:    v_readfirstlane_b32 s4, v10
 ; GFX1030-NEXT:    v_readfirstlane_b32 s5, v11
 ; GFX1030-NEXT:    v_readfirstlane_b32 s6, v12
 ; GFX1030-NEXT:    v_readfirstlane_b32 s7, v13
 ; GFX1030-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[10:11]
-; GFX1030-NEXT:    image_bvh_intersect_ray v[0:3], [v5, v14, v15, v16, v4, v6, v7, v8], s[4:7] a16
 ; GFX1030-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[12:13]
 ; GFX1030-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1030-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1030-NEXT:    image_bvh_intersect_ray v[0:3], v[14:21], s[4:7] a16
+; GFX1030-NEXT:    ; implicit-def: $vgpr10_vgpr11
+; GFX1030-NEXT:    ; implicit-def: $vgpr14
+; GFX1030-NEXT:    ; implicit-def: $vgpr15
+; GFX1030-NEXT:    ; implicit-def: $vgpr16
+; GFX1030-NEXT:    ; implicit-def: $vgpr17
+; GFX1030-NEXT:    ; implicit-def: $vgpr18
+; GFX1030-NEXT:    ; implicit-def: $vgpr19
+; GFX1030-NEXT:    ; implicit-def: $vgpr20
+; GFX1030-NEXT:    ; implicit-def: $vgpr21
+; GFX1030-NEXT:    ; implicit-def: $vgpr10_vgpr11_vgpr12_vgpr13
 ; GFX1030-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1030-NEXT:    s_cbranch_execnz BB7_1
 ; GFX1030-NEXT:  ; %bb.2:
 ; GFX1030-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1030-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1030-NEXT:    ; return to shader part epilog
 ;
 ; GFX1013-LABEL: image_bvh_intersect_ray_a16_vgpr_descr:

 ; GFX1013-NEXT:    v_and_or_b32 v6, v7, s0, v14
 ; GFX1013-NEXT:    v_lshl_or_b32 v7, v9, 16, v8
 ; GFX1013-NEXT:  BB7_1: ; =>This Inner Loop Header: Depth=1
 ; GFX1013-NEXT:    v_readfirstlane_b32 s4, v10
 ; GFX1013-NEXT:    v_readfirstlane_b32 s5, v11
 ; GFX1013-NEXT:    v_readfirstlane_b32 s6, v12
 ; GFX1013-NEXT:    v_readfirstlane_b32 s7, v13
 ; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[10:11]
-; GFX1013-NEXT:    image_bvh_intersect_ray v[14:17], v[0:7], s[4:7] a16
 ; GFX1013-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[12:13]
 ; GFX1013-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1013-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1013-NEXT:    image_bvh_intersect_ray v[14:17], v[0:7], s[4:7] a16
+; GFX1013-NEXT:    ; implicit-def: $vgpr10_vgpr11
+; GFX1013-NEXT:    ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7
+; GFX1013-NEXT:    ; implicit-def: $vgpr10_vgpr11_vgpr12_vgpr13
+; GFX1013-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1013-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1013-NEXT:    s_cbranch_execnz BB7_1
 ; GFX1013-NEXT:  ; %bb.2:
 ; GFX1013-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1013-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1013-NEXT:    v_mov_b32_e32 v0, v14
 ; GFX1013-NEXT:    v_mov_b32_e32 v1, v15
 ; GFX1013-NEXT:    v_mov_b32_e32 v2, v16
 ; GFX1013-NEXT:    v_mov_b32_e32 v3, v17
 ; GFX1013-NEXT:    ; return to shader part epilog
   %v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i32.v4f16(i32 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x half> %ray_dir, <4 x half> %ray_inv_dir, <4 x i32> %tdescr)
   %r = bitcast <4 x i32> %v to <4 x float>
   ret <4 x float> %r
 }
 
 define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_vgpr_descr(i64 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x float> %ray_dir, <4 x float> %ray_inv_dir, <4 x i32> %tdescr) {
 ; GFX1030-LABEL: image_bvh64_intersect_ray_vgpr_descr:
 ; GFX1030:       ; %bb.0:
-; GFX1030-NEXT:    v_mov_b32_e32 v6, v0
-; GFX1030-NEXT:    v_mov_b32_e32 v10, v1
-; GFX1030-NEXT:    v_mov_b32_e32 v14, v2
-; GFX1030-NEXT:    v_mov_b32_e32 v19, v3
+; GFX1030-NEXT:    v_mov_b32_e32 v19, v0
+; GFX1030-NEXT:    v_mov_b32_e32 v20, v1
+; GFX1030-NEXT:    v_mov_b32_e32 v21, v2
+; GFX1030-NEXT:    v_mov_b32_e32 v22, v3
+; GFX1030-NEXT:    v_mov_b32_e32 v23, v4
+; GFX1030-NEXT:    v_mov_b32_e32 v24, v5
+; GFX1030-NEXT:    v_mov_b32_e32 v25, v7
+; GFX1030-NEXT:    v_mov_b32_e32 v26, v8
+; GFX1030-NEXT:    v_mov_b32_e32 v27, v9
+; GFX1030-NEXT:    v_mov_b32_e32 v28, v11
+; GFX1030-NEXT:    v_mov_b32_e32 v29, v12
+; GFX1030-NEXT:    v_mov_b32_e32 v30, v13
 ; GFX1030-NEXT:    s_mov_b32 s1, exec_lo
 ; GFX1030-NEXT:  BB8_1: ; =>This Inner Loop Header: Depth=1
 ; GFX1030-NEXT:    v_readfirstlane_b32 s4, v15
 ; GFX1030-NEXT:    v_readfirstlane_b32 s5, v16
 ; GFX1030-NEXT:    v_readfirstlane_b32 s6, v17
 ; GFX1030-NEXT:    v_readfirstlane_b32 s7, v18
 ; GFX1030-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[15:16]
-; GFX1030-NEXT:    image_bvh64_intersect_ray v[0:3], [v6, v10, v14, v19, v4, v5, v7, v8, v9, v11, v12, v13], s[4:7]
 ; GFX1030-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[17:18]
 ; GFX1030-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1030-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1030-NEXT:    image_bvh64_intersect_ray v[0:3], v[19:34], s[4:7]
+; GFX1030-NEXT:    ; implicit-def: $vgpr15_vgpr16
+; GFX1030-NEXT:    ; implicit-def: $vgpr19
+; GFX1030-NEXT:    ; implicit-def: $vgpr20
+; GFX1030-NEXT:    ; implicit-def: $vgpr21
+; GFX1030-NEXT:    ; implicit-def: $vgpr22
+; GFX1030-NEXT:    ; implicit-def: $vgpr23
+; GFX1030-NEXT:    ; implicit-def: $vgpr24
+; GFX1030-NEXT:    ; implicit-def: $vgpr25
+; GFX1030-NEXT:    ; implicit-def: $vgpr26
+; GFX1030-NEXT:    ; implicit-def: $vgpr27
+; GFX1030-NEXT:    ; implicit-def: $vgpr28
+; GFX1030-NEXT:    ; implicit-def: $vgpr29
+; GFX1030-NEXT:    ; implicit-def: $vgpr30
+; GFX1030-NEXT:    ; implicit-def: $vgpr15_vgpr16_vgpr17_vgpr18
 ; GFX1030-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1030-NEXT:    s_cbranch_execnz BB8_1
 ; GFX1030-NEXT:  ; %bb.2:
 ; GFX1030-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1030-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1030-NEXT:    ; return to shader part epilog
 ;
 ; GFX1013-LABEL: image_bvh64_intersect_ray_vgpr_descr:
 ; GFX1013:       ; %bb.0:
 ; GFX1013-NEXT:    v_mov_b32_e32 v6, v7
 ; GFX1013-NEXT:    v_mov_b32_e32 v7, v8
 ; GFX1013-NEXT:    v_mov_b32_e32 v8, v9
 ; GFX1013-NEXT:    v_mov_b32_e32 v9, v11
 ; GFX1013-NEXT:    v_mov_b32_e32 v10, v12
 ; GFX1013-NEXT:    v_mov_b32_e32 v11, v13
+; GFX1013-NEXT:    v_mov_b32_e32 v19, v15
+; GFX1013-NEXT:    v_mov_b32_e32 v20, v16
 ; GFX1013-NEXT:    s_mov_b32 s1, exec_lo
 ; GFX1013-NEXT:  BB8_1: ; =>This Inner Loop Header: Depth=1
-; GFX1013-NEXT:    v_readfirstlane_b32 s4, v15
-; GFX1013-NEXT:    v_readfirstlane_b32 s5, v16
+; GFX1013-NEXT:    v_readfirstlane_b32 s4, v19
+; GFX1013-NEXT:    v_readfirstlane_b32 s5, v20
 ; GFX1013-NEXT:    v_readfirstlane_b32 s6, v17
 ; GFX1013-NEXT:    v_readfirstlane_b32 s7, v18
-; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[15:16]
-; GFX1013-NEXT:    image_bvh64_intersect_ray v[19:22], v[0:15], s[4:7]
+; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[19:20]
 ; GFX1013-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[17:18]
 ; GFX1013-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1013-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1013-NEXT:    image_bvh64_intersect_ray v[21:24], v[0:15], s[4:7]
+; GFX1013-NEXT:    ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
+; GFX1013-NEXT:    ; implicit-def: $vgpr19_vgpr20
+; GFX1013-NEXT:    ; implicit-def: $vgpr15_vgpr16_vgpr17_vgpr18
+; GFX1013-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1013-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1013-NEXT:    s_cbranch_execnz BB8_1
 ; GFX1013-NEXT:  ; %bb.2:
 ; GFX1013-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1013-NEXT:    s_waitcnt vmcnt(0)
-; GFX1013-NEXT:    v_mov_b32_e32 v0, v19
-; GFX1013-NEXT:    v_mov_b32_e32 v1, v20
-; GFX1013-NEXT:    v_mov_b32_e32 v2, v21
-; GFX1013-NEXT:    v_mov_b32_e32 v3, v22
+; GFX1013-NEXT:    v_mov_b32_e32 v0, v21
+; GFX1013-NEXT:    v_mov_b32_e32 v1, v22
+; GFX1013-NEXT:    v_mov_b32_e32 v2, v23
+; GFX1013-NEXT:    v_mov_b32_e32 v3, v24
 ; GFX1013-NEXT:    ; return to shader part epilog
   %v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i64.v4f32(i64 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x float> %ray_dir, <4 x float> %ray_inv_dir, <4 x i32> %tdescr)
   %r = bitcast <4 x i32> %v to <4 x float>
   ret <4 x float> %r
 }
 
 define amdgpu_ps <4 x float> @image_bvh64_intersect_ray_a16_vgpr_descr(i64 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x half> %ray_dir, <4 x half> %ray_inv_dir, <4 x i32> %tdescr) {
 ; GFX1030-LABEL: image_bvh64_intersect_ray_a16_vgpr_descr:
 ; GFX1030:       ; %bb.0:
 ; GFX1030-NEXT:    s_mov_b32 s0, 0xffff
-; GFX1030-NEXT:    v_mov_b32_e32 v6, v0
+; GFX1030-NEXT:    v_mov_b32_e32 v15, v0
 ; GFX1030-NEXT:    v_lshrrev_b32_e32 v0, 16, v7
-; GFX1030-NEXT:    v_mov_b32_e32 v15, v1
+; GFX1030-NEXT:    v_mov_b32_e32 v16, v1
 ; GFX1030-NEXT:    v_and_b32_e32 v1, s0, v9
-; GFX1030-NEXT:    v_mov_b32_e32 v16, v2
-; GFX1030-NEXT:    v_mov_b32_e32 v17, v3
+; GFX1030-NEXT:    v_mov_b32_e32 v17, v2
+; GFX1030-NEXT:    v_mov_b32_e32 v18, v3
 ; GFX1030-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX1030-NEXT:    v_lshrrev_b32_e32 v2, 16, v9
 ; GFX1030-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX1030-NEXT:    v_and_b32_e32 v3, s0, v10
+; GFX1030-NEXT:    v_mov_b32_e32 v19, v4
+; GFX1030-NEXT:    v_mov_b32_e32 v20, v5
+; GFX1030-NEXT:    v_and_or_b32 v21, v7, s0, v0
+; GFX1030-NEXT:    v_and_or_b32 v22, v8, s0, v1
+; GFX1030-NEXT:    v_lshl_or_b32 v23, v3, 16, v2
 ; GFX1030-NEXT:    s_mov_b32 s1, exec_lo
-; GFX1030-NEXT:    v_and_or_b32 v7, v7, s0, v0
-; GFX1030-NEXT:    v_and_or_b32 v8, v8, s0, v1
-; GFX1030-NEXT:    v_lshl_or_b32 v9, v3, 16, v2
 ; GFX1030-NEXT:  BB9_1: ; =>This Inner Loop Header: Depth=1
 ; GFX1030-NEXT:    v_readfirstlane_b32 s4, v11
 ; GFX1030-NEXT:    v_readfirstlane_b32 s5, v12
 ; GFX1030-NEXT:    v_readfirstlane_b32 s6, v13
 ; GFX1030-NEXT:    v_readfirstlane_b32 s7, v14
 ; GFX1030-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[11:12]
-; GFX1030-NEXT:    image_bvh64_intersect_ray v[0:3], [v6, v15, v16, v17, v4, v5, v7, v8, v9], s[4:7] a16
 ; GFX1030-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[13:14]
 ; GFX1030-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1030-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1030-NEXT:    image_bvh64_intersect_ray v[0:3], v[15:30], s[4:7] a16
+; GFX1030-NEXT:    ; implicit-def: $vgpr11_vgpr12
+; GFX1030-NEXT:    ; implicit-def: $vgpr15
+; GFX1030-NEXT:    ; implicit-def: $vgpr16
+; GFX1030-NEXT:    ; implicit-def: $vgpr17
+; GFX1030-NEXT:    ; implicit-def: $vgpr18
+; GFX1030-NEXT:    ; implicit-def: $vgpr19
+; GFX1030-NEXT:    ; implicit-def: $vgpr20
+; GFX1030-NEXT:    ; implicit-def: $vgpr21
+; GFX1030-NEXT:    ; implicit-def: $vgpr22
+; GFX1030-NEXT:    ; implicit-def: $vgpr23
+; GFX1030-NEXT:    ; implicit-def: $vgpr11_vgpr12_vgpr13_vgpr14
 ; GFX1030-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1030-NEXT:    s_cbranch_execnz BB9_1
 ; GFX1030-NEXT:  ; %bb.2:
 ; GFX1030-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1030-NEXT:    s_waitcnt vmcnt(0)
 ; GFX1030-NEXT:    ; return to shader part epilog
 ;
 ; GFX1013-LABEL: image_bvh64_intersect_ray_a16_vgpr_descr:
 ; GFX1013:       ; %bb.0:
 ; GFX1013-NEXT:    s_mov_b32 s0, 0xffff
 ; GFX1013-NEXT:    v_lshrrev_b32_e32 v6, 16, v7
-; GFX1013-NEXT:    v_and_b32_e32 v15, s0, v9
+; GFX1013-NEXT:    v_mov_b32_e32 v16, v11
+; GFX1013-NEXT:    v_and_b32_e32 v11, s0, v9
 ; GFX1013-NEXT:    v_lshrrev_b32_e32 v9, 16, v9
 ; GFX1013-NEXT:    v_and_b32_e32 v10, s0, v10
-; GFX1013-NEXT:    s_mov_b32 s1, exec_lo
 ; GFX1013-NEXT:    v_lshlrev_b32_e32 v6, 16, v6
-; GFX1013-NEXT:    v_lshlrev_b32_e32 v15, 16, v15
+; GFX1013-NEXT:    v_mov_b32_e32 v17, v12
+; GFX1013-NEXT:    v_lshlrev_b32_e32 v11, 16, v11
+; GFX1013-NEXT:    v_mov_b32_e32 v18, v13
+; GFX1013-NEXT:    v_mov_b32_e32 v19, v14
 ; GFX1013-NEXT:    v_and_or_b32 v6, v7, s0, v6
-; GFX1013-NEXT:    v_and_or_b32 v7, v8, s0, v15
+; GFX1013-NEXT:    s_mov_b32 s1, exec_lo
+; GFX1013-NEXT:    v_and_or_b32 v7, v8, s0, v11
 ; GFX1013-NEXT:    v_lshl_or_b32 v8, v10, 16, v9
 ; GFX1013-NEXT:  BB9_1: ; =>This Inner Loop Header: Depth=1
-; GFX1013-NEXT:    v_readfirstlane_b32 s4, v11
-; GFX1013-NEXT:    v_readfirstlane_b32 s5, v12
-; GFX1013-NEXT:    v_readfirstlane_b32 s6, v13
-; GFX1013-NEXT:    v_readfirstlane_b32 s7, v14
-; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[11:12]
-; GFX1013-NEXT:    s_waitcnt vmcnt(0)
-; GFX1013-NEXT:    image_bvh64_intersect_ray v[15:18], v[0:15], s[4:7] a16
-; GFX1013-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[13:14]
+; GFX1013-NEXT:    v_readfirstlane_b32 s4, v16
+; GFX1013-NEXT:    v_readfirstlane_b32 s5, v17
+; GFX1013-NEXT:    v_readfirstlane_b32 s6, v18
+; GFX1013-NEXT:    v_readfirstlane_b32 s7, v19
+; GFX1013-NEXT:    v_cmp_eq_u64_e32 vcc_lo, s[4:5], v[16:17]
+; GFX1013-NEXT:    v_cmp_eq_u64_e64 s0, s[6:7], v[18:19]
 ; GFX1013-NEXT:    s_and_b32 s0, s0, vcc_lo
 ; GFX1013-NEXT:    s_and_saveexec_b32 s0, s0
+; GFX1013-NEXT:    image_bvh64_intersect_ray v[20:23], v[0:15], s[4:7] a16
+; GFX1013-NEXT:    ; implicit-def: $vgpr16_vgpr17
+; GFX1013-NEXT:    ; implicit-def: $vgpr0_vgpr1_vgpr2_vgpr3_vgpr4_vgpr5_vgpr6_vgpr7_vgpr8_vgpr9_vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15
+; GFX1013-NEXT:    ; implicit-def: $vgpr16_vgpr17_vgpr18_vgpr19
+; GFX1013-NEXT:    s_waitcnt_depctr 0xffe3
 ; GFX1013-NEXT:    s_xor_b32 exec_lo, exec_lo, s0
 ; GFX1013-NEXT:    s_cbranch_execnz BB9_1
 ; GFX1013-NEXT:  ; %bb.2:
 ; GFX1013-NEXT:    s_mov_b32 exec_lo, s1
 ; GFX1013-NEXT:    s_waitcnt vmcnt(0)
-; GFX1013-NEXT:    v_mov_b32_e32 v0, v15
-; GFX1013-NEXT:    v_mov_b32_e32 v1, v16
-; GFX1013-NEXT:    v_mov_b32_e32 v2, v17
-; GFX1013-NEXT:    v_mov_b32_e32 v3, v18
+; GFX1013-NEXT:    v_mov_b32_e32 v0, v20
+; GFX1013-NEXT:    v_mov_b32_e32 v1, v21
+; GFX1013-NEXT:    v_mov_b32_e32 v2, v22
+; GFX1013-NEXT:    v_mov_b32_e32 v3, v23
 ; GFX1013-NEXT:    ; return to shader part epilog
   %v = call <4 x i32> @llvm.amdgcn.image.bvh.intersect.ray.i64.v4f16(i64 %node_ptr, float %ray_extent, <4 x float> %ray_origin, <4 x half> %ray_dir, <4 x half> %ray_inv_dir, <4 x i32> %tdescr)
   %r = bitcast <4 x i32> %v to <4 x float>
   ret <4 x float> %r
 }
 
 define amdgpu_kernel void @image_bvh_intersect_ray_nsa_reassign(i32* %p_node_ptr, float* %p_ray, <4 x i32> inreg %tdescr) {
 ; GFX1030-LABEL: image_bvh_intersect_ray_nsa_reassign:

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.atomic.add.ll

define amdgpu_ps float @raw_buffer_atomic_add_i32__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   [[BUFFER_ATOMIC_ADD_OFFEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_OFFEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_ATOMIC_ADD_OFFEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_OFFEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_ATOMIC_ADD_OFFEN_RTN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %ret = call i32 @llvm.amdgcn.raw.buffer.atomic.add.i32(i32 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   %cast = bitcast i32 %ret to float
define amdgpu_ps void @raw_buffer_atomic_add_i32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   [[BUFFER_ATOMIC_ADD_OFFEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_OFFEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_ATOMIC_ADD_OFFEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_OFFEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %ret = call i32 @llvm.amdgcn.raw.buffer.atomic.add.i32(i32 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.atomic.cmpswap.ll

define amdgpu_ps float @raw_buffer_atomic_cmpswap_i32__sgpr_val__sgpr_cmp__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, i32 inreg %cmp, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY9]], %subreg.sub1
   ; CHECK:   [[BUFFER_ATOMIC_CMPSWAP_OFFEN_RTN:%[0-9]+]]:vreg_64 = BUFFER_ATOMIC_CMPSWAP_OFFEN_RTN [[REG_SEQUENCE4]], [[COPY10]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_ATOMIC_CMPSWAP_OFFEN_RTN]].sub0
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[COPY13]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %ret = call i32 @llvm.amdgcn.raw.buffer.atomic.cmpswap.i32(i32 %val, i32 %cmp, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   %cast = bitcast i32 %ret to float
define amdgpu_ps void @raw_buffer_atomic_cmpswap_i32_noret__sgpr_val__sgpr_cmp__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, i32 inreg %cmp, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY8]], %subreg.sub0, [[COPY9]], %subreg.sub1
   ; CHECK:   [[BUFFER_ATOMIC_CMPSWAP_OFFEN_RTN:%[0-9]+]]:vreg_64 = BUFFER_ATOMIC_CMPSWAP_OFFEN_RTN [[REG_SEQUENCE4]], [[COPY10]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_ATOMIC_CMPSWAP_OFFEN_RTN]].sub0
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %ret = call i32 @llvm.amdgcn.raw.buffer.atomic.cmpswap.i32(i32 %val, i32 %cmp, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.atomic.fadd.ll

define amdgpu_ps void @raw_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; GFX908:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; GFX908:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX908:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; GFX908:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX908:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX908:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; GFX908:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; GFX908:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; GFX908:   BUFFER_ATOMIC_ADD_F32_OFFEN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX908:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX908:   BUFFER_ATOMIC_ADD_F32_OFFEN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX908:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX908:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX908:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX908: bb.3:
   ; GFX908:   successors: %bb.4(0x80000000)
   ; GFX908:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX908: bb.4:
   ; GFX908:   S_ENDPGM 0
   ; GFX90A-LABEL: name: raw_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset
   ; GFX90A: bb.1 (%ir-block.0):
   ; GFX90A:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; GFX90A:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; GFX90A:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX90A:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; GFX90A:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX90A:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX90A:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; GFX90A:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; GFX90A:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_OFFEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_OFFEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX90A:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_OFFEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_OFFEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX90A:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX90A:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX90A:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX90A: bb.3:
   ; GFX90A:   successors: %bb.4(0x80000000)
   ; GFX90A:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX90A: bb.4:
   ; GFX90A:   S_ENDPGM 0
   %ret = call float @llvm.amdgcn.raw.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__0_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 %soffset) {
   ; GFX908:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; GFX908:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX908:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; GFX908:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX908:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX908:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; GFX908:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; GFX908:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; GFX908:   BUFFER_ATOMIC_ADD_F32_OFFSET [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX908:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX908:   BUFFER_ATOMIC_ADD_F32_OFFSET [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX908:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX908:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX908:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX908: bb.3:
   ; GFX908:   successors: %bb.4(0x80000000)
   ; GFX908:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX908: bb.4:
   ; GFX908:   S_ENDPGM 0
   ; GFX90A-LABEL: name: raw_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__0_voffset__vgpr_soffset
   ; GFX90A: bb.1 (%ir-block.0):
   ; GFX90A:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__0_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 %soffset) {
   ; GFX90A:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; GFX90A:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX90A:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; GFX90A:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX90A:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX90A:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; GFX90A:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; GFX90A:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_OFFSET_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_OFFSET_RTN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX90A:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_OFFSET_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_OFFSET_RTN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX90A:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX90A:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX90A:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX90A: bb.3:
   ; GFX90A:   successors: %bb.4(0x80000000)
   ; GFX90A:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX90A: bb.4:
   ; GFX90A:   S_ENDPGM 0
   %ret = call float @llvm.amdgcn.raw.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 0, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.load.format.f16.ll

define amdgpu_ps half @raw_buffer_load_format_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; PACKED:   [[BUFFER_LOAD_FORMAT_D16_X_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_D16_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   [[BUFFER_LOAD_FORMAT_D16_X_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_D16_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   $vgpr0 = COPY [[BUFFER_LOAD_FORMAT_D16_X_OFFEN]]
   ; PACKED:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; UNPACKED-LABEL: name: raw_buffer_load_format_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset
   ; UNPACKED: bb.1 (%ir-block.0):
define amdgpu_ps half @raw_buffer_load_format_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; UNPACKED:   [[BUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   [[BUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   $vgpr0 = COPY [[BUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN]]
   ; UNPACKED:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call half @llvm.amdgcn.raw.buffer.load.format.f16(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret half %val

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.load.format.ll

define amdgpu_ps float @raw_buffer_load_format_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   [[BUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_FORMAT_X_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.buffer.load.format.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret float %val

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.load.ll

define amdgpu_ps float @raw_buffer_load_f32__vgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[COPY5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[COPY5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps float @raw_buffer_load_f32__vgpr_rsrc__vgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps half @raw_buffer_load_f16__vgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   [[BUFFER_LOAD_USHORT_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_USHORT_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[COPY5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_USHORT_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_USHORT_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[COPY5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_USHORT_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call half @llvm.amdgcn.raw.buffer.load.f16(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret half %val
define amdgpu_ps float @raw_buffer_load_i8__vgpr_rsrc__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   [[BUFFER_LOAD_UBYTE_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_UBYTE_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[COPY5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s8), addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_UBYTE_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_UBYTE_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[COPY5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s8), addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_UBYTE_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call i8 @llvm.amdgcn.raw.buffer.load.i8(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   %zext = zext i8 %val to i32
define amdgpu_ps float @raw_buffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add5000(<4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset.base) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_ADD_I32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_ADD_I32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %soffset = add i32 %soffset.base, 5000
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
define amdgpu_ps float @raw_buffer_load_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add5000(<4 x i32> %rsrc, i32 %voffset.base, i32 inreg %soffset) {
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY $vgpr3
   ; CHECK:   [[REG_SEQUENCE:%[0-9]+]]:vreg_128 = REG_SEQUENCE [[COPY]], %subreg.sub0, [[COPY1]], %subreg.sub1, [[COPY2]], %subreg.sub2, [[COPY3]], %subreg.sub3
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr4
   ; CHECK:   [[COPY5:%[0-9]+]]:sreg_32 = COPY $sgpr2
   ; CHECK:   [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 4096
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[S_MOV_B32_]]
-  ; CHECK:   %14:vgpr_32, dead %35:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY4]], [[COPY6]], 0, implicit $exec
+  ; CHECK:   %14:vgpr_32, dead %33:sreg_64_xexec = V_ADD_CO_U32_e64 [[COPY4]], [[COPY6]], 0, implicit $exec
   ; CHECK:   [[COPY7:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub0_sub1
   ; CHECK:   [[COPY8:%[0-9]+]]:vreg_64 = COPY [[REG_SEQUENCE]].sub2_sub3
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN %14, [[REG_SEQUENCE3]], [[COPY5]], 904, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN %14, [[REG_SEQUENCE3]], [[COPY5]], 904, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %voffset = add i32 %voffset.base, 5000
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.store.format.f16.ll

define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16(<4 x i32> %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub0, implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY11]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; UNPACKED:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16(<4 x i32> %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub0, implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY9]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; PACKED:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.format.v4f16(<4 x half> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16_add_4096(<4 x i32> %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY11]], implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub0, implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY12]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; UNPACKED:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE1]], %13, [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   BUFFER_STORE_FORMAT_D16_XYZW_gfx80_OFFEN_exact [[REG_SEQUENCE1]], %13, [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16_add_4096
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16_add_4096(<4 x i32> %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub0, implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY10]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; PACKED:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE1]], %13, [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   BUFFER_STORE_FORMAT_D16_XYZW_OFFEN_exact [[REG_SEQUENCE1]], %13, [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   %voffset.add = add i32 %voffset, 4096
   call void @llvm.amdgcn.raw.buffer.store.format.v4f16(<4 x half> %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
   ret void
 }
 
 declare void @llvm.amdgcn.raw.buffer.store.format.f16(half, <4 x i32>, i32, i32, i32 immarg)
 declare void @llvm.amdgcn.raw.buffer.store.format.v2f16(<2 x half>, <4 x i32>, i32, i32, i32 immarg)
 declare void @llvm.amdgcn.raw.buffer.store.format.v4f16(<4 x half>, <4 x i32>, i32, i32, i32 immarg)

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.store.format.f32.ll

define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32(<4 x i32> %rsrc, <4 x float> %val, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY11]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY11]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], [[COPY8]], [[REG_SEQUENCE4]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], [[COPY8]], [[REG_SEQUENCE4]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_store_format__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f32_add_4096(<4 x i32> %rsrc, <4 x float> %val, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY11]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY12]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], %15, [[REG_SEQUENCE4]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_FORMAT_XYZW_OFFEN_exact [[REG_SEQUENCE1]], %15, [[REG_SEQUENCE4]], [[COPY9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %voffset.add = add i32 %voffset, 4096
   call void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
   ret void
 }
 
 declare void @llvm.amdgcn.raw.buffer.store.format.f32(float, <4 x i32>, i32, i32, i32 immarg)
 declare void @llvm.amdgcn.raw.buffer.store.format.v2f32(<2 x float>, <4 x i32>, i32, i32, i32 immarg)
 declare void @llvm.amdgcn.raw.buffer.store.format.v3f32(<3 x float>, <4 x i32>, i32, i32, i32 immarg)
 declare void @llvm.amdgcn.raw.buffer.store.format.v4f32(<4 x float>, <4 x i32>, i32, i32, i32 immarg)

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.buffer.store.ll

define amdgpu_ps void @raw_buffer_store__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, float %val, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_store__sgpr_rsrc__vgpr_val__vgpr_voffset__vgpr_soffset(<4 x i32> inreg %rsrc, float %val, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY $vgpr0
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr_32 = COPY $vgpr1
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY $vgpr2
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_]], [[COPY6]], implicit $exec
-  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[V_READFIRSTLANE_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[V_CMP_EQ_U32_e64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE]], [[V_READFIRSTLANE_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_store__vgpr_rsrc__vgpr_val__vgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, float %val, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_store__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_v4f16(<4 x i32> %rsrc, <4 x half> %val, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY9]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   BUFFER_STORE_DWORDX2_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_DWORDX2_OFFEN_exact [[REG_SEQUENCE1]], [[COPY6]], [[REG_SEQUENCE4]], [[COPY7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s16>), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.v4f16(<4 x half> %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @raw_buffer_store__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset_offset_add_5000(<4 x i32> %rsrc, float %val, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY8]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], %14, [[REG_SEQUENCE3]], [[COPY6]], 904, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], %14, [[REG_SEQUENCE3]], [[COPY6]], 904, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %voffset.add = add i32 %voffset, 5000
   call void @llvm.amdgcn.raw.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset.add, i32 %soffset, i32 0)
   ret void
define amdgpu_ps void @raw_buffer_store__vgpr_rsrc__vgpr_val__5000_voffset__sgpr_soffset_offset(<4 x i32> %rsrc, float %val, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY6]], [[REG_SEQUENCE3]], [[COPY5]], 904, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   BUFFER_STORE_DWORD_OFFEN_exact [[COPY4]], [[COPY6]], [[REG_SEQUENCE3]], [[COPY5]], 904, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 5000, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.tbuffer.load.f16.ll

define amdgpu_ps half @raw_tbuffer_load_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; UNPACKED:   [[TBUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   [[TBUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_D16_X_gfx80_OFFEN]]
   ; UNPACKED:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; PACKED-LABEL: name: raw_tbuffer_load_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
define amdgpu_ps half @raw_tbuffer_load_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; PACKED:   [[TBUFFER_LOAD_FORMAT_D16_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_D16_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   [[TBUFFER_LOAD_FORMAT_D16_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_D16_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s16), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_D16_X_OFFEN]]
   ; PACKED:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call half @llvm.amdgcn.raw.tbuffer.load.f16(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
   ret half %val

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.tbuffer.load.ll

define amdgpu_ps float @raw_tbuffer_load_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY5]], implicit $exec
   ; CHECK:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U32_e64_]], [[S_AND_B32_]], implicit-def $scc
-  ; CHECK:   [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[TBUFFER_LOAD_FORMAT_X_OFFEN:%[0-9]+]]:vgpr_32 = TBUFFER_LOAD_FORMAT_X_OFFEN [[COPY6]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[TBUFFER_LOAD_FORMAT_X_OFFEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.tbuffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
   ret float %val

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.tbuffer.store.f16.ll

define amdgpu_ps void @raw_tbuffer_store_f16__vgpr_rsrc__vgpr_voffset__sgpr_soffset(half %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; UNPACKED:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; UNPACKED:   TBUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   TBUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_tbuffer_store_f16__vgpr_rsrc__vgpr_voffset__sgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_tbuffer_store_f16__vgpr_rsrc__vgpr_voffset__sgpr_soffset(half %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; PACKED:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; PACKED:   TBUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   TBUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.f16(half %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_f16__vgpr_rsrc__vgpr_voffset__vgpr_soffset(half %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; UNPACKED:   TBUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   TBUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_tbuffer_store_f16__vgpr_rsrc__vgpr_voffset__vgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_tbuffer_store_f16__vgpr_rsrc__vgpr_voffset__vgpr_soffset(half %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; PACKED:   TBUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   TBUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.f16(half %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset(half %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; UNPACKED:   TBUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   TBUFFER_STORE_FORMAT_D16_X_gfx80_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_tbuffer_store_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_tbuffer_store_f16__vgpr_rsrc__sgpr_voffset__vgpr_soffset(half %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; PACKED:   TBUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   TBUFFER_STORE_FORMAT_D16_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.f16(half %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.tbuffer.store.i8.ll

define amdgpu_ps void @raw_tbuffer_store_i8__vgpr_rsrc__vgpr_voffset__sgpr_soffset(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; UNPACKED:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; UNPACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_tbuffer_store_i8__vgpr_rsrc__vgpr_voffset__sgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_tbuffer_store_i8__vgpr_rsrc__vgpr_voffset__sgpr_soffset(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; PACKED:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; PACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.i8(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_i8__vgpr_rsrc__vgpr_voffset__vgpr_soffset(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; UNPACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_tbuffer_store_i8__vgpr_rsrc__vgpr_voffset__vgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_tbuffer_store_i8__vgpr_rsrc__vgpr_voffset__vgpr_soffset(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; PACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.i8(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_i8__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i8 %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; UNPACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; UNPACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: raw_tbuffer_store_i8__vgpr_rsrc__sgpr_voffset__vgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @raw_tbuffer_store_i8__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i8 %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; PACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; PACKED:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s8), addrspace 4)
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.i8(i8 %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
 }
 
 declare void @llvm.amdgcn.raw.tbuffer.store.i8(i8, <4 x i32>, i32, i32, i32 immarg, i32 immarg)

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.raw.tbuffer.store.ll

define amdgpu_ps void @raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__sgpr_soffset(float %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 1, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[COPY6]], 0, 94, 1, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 1)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_f32__vgpr_rsrc__vgpr_voffset__vgpr_soffset(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U32_e64_]], [[S_AND_B32_]], implicit-def $scc
-  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_f32__vgpr_rsrc__sgpr_voffset__vgpr_soffset(float %val, <4 x i32> %rsrc, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; CHECK:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B32_1:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U32_e64_]], [[S_AND_B32_]], implicit-def $scc
-  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY7]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 78, i32 0)
   ret void
 }
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_soffset_add5000(float %val, <4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset.base) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY7]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[S_ADD_I32_]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[COPY5]], [[REG_SEQUENCE3]], [[S_ADD_I32_]], 0, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %soffset = add i32 %soffset.base, 5000
   call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
define amdgpu_ps void @raw_tbuffer_store_f32__sgpr_rsrc__vgpr_voffset__sgpr_soffset_voffset_add5000(float %val, <4 x i32> %rsrc, i32 %voffset.base, i32 inreg %soffset) {
   ; CHECK:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY8]], implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub0, implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY9]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY9]], implicit $exec
   ; CHECK:   [[S_AND_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_B32 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[V_ADD_U32_e64_]], [[REG_SEQUENCE3]], [[COPY6]], 904, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B32_:%[0-9]+]]:sreg_32_xm0_xexec = S_AND_SAVEEXEC_B32 killed [[S_AND_B32_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   TBUFFER_STORE_FORMAT_X_OFFEN_exact [[COPY]], [[V_ADD_U32_e64_]], [[REG_SEQUENCE3]], [[COPY6]], 904, 94, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec_lo = S_XOR_B32_term $exec_lo, [[S_AND_SAVEEXEC_B32_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec_lo = S_MOV_B32_term [[S_MOV_B32_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %voffset = add i32 %voffset.base, 5000
   call void @llvm.amdgcn.raw.tbuffer.store.f32(float %val, <4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 94, i32 0)
   ret void
 }
 
 declare void @llvm.amdgcn.raw.tbuffer.store.f32(float, <4 x i32>, i32, i32, i32 immarg, i32 immarg)
 declare void @llvm.amdgcn.raw.tbuffer.store.v2f32(<2 x float>, <4 x i32>, i32, i32, i32 immarg, i32 immarg)
 declare void @llvm.amdgcn.raw.tbuffer.store.v3f32(<3 x float>, <4 x i32>, i32, i32, i32 immarg, i32 immarg)
 declare void @llvm.amdgcn.raw.tbuffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i32 immarg, i32 immarg)

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

define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc(<4 x i32> %rsrc, i32 inreg %soffset) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX6:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX7-LABEL: name: s_buffer_load_f32_vgpr_rsrc
   ; GFX7: bb.1 (%ir-block.0):
   ; GFX7:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc(<4 x i32> %rsrc, i32 inreg %soffset) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX7:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX8-LABEL: name: s_buffer_load_f32_vgpr_rsrc
   ; GFX8: bb.1 (%ir-block.0):
   ; GFX8:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc(<4 x i32> %rsrc, i32 inreg %soffset) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX8:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret float %val
 }
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4092(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4092, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4092, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX6:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX7-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4092
   ; GFX7: bb.1 (%ir-block.0):
   ; GFX7:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4092(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4092, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4092, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX7:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX8-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4092
   ; GFX8: bb.1 (%ir-block.0):
   ; GFX8:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4092(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4092, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4092, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX8:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %soffset = add i32 %soffset.base, 4092
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX6:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX7-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4096
   ; GFX7: bb.1 (%ir-block.0):
   ; GFX7:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX7:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX8-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4096
   ; GFX8: bb.1 (%ir-block.0):
   ; GFX8:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX8:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %soffset = add i32 %soffset.base, 4096
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4095(<4 x i32> %rsrc) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX6:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX7-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4095
   ; GFX7: bb.1 (%ir-block.0):
   ; GFX7:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4095(<4 x i32> %rsrc) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX7:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX8-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4095
   ; GFX8: bb.1 (%ir-block.0):
   ; GFX8:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4095(<4 x i32> %rsrc) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX8:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 4095, i32 0)
   ret float %val
 }
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4096(<4 x i32> %rsrc) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX6:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX6:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX7-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4096
   ; GFX7: bb.1 (%ir-block.0):
   ; GFX7:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4096(<4 x i32> %rsrc) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX7:   [[BUFFER_LOAD_DWORD_OFFEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32))
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFEN]]
   ; GFX7:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GFX8-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4096
   ; GFX8: bb.1 (%ir-block.0):
   ; GFX8:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4096(<4 x i32> %rsrc) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
-  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4096)
   ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX8:   [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4095, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s32) from unknown-address + 4096)
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_OFFSET]]
   ; GFX8:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 4096, i32 0)
   ret float %val
 }
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4064(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFSET1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFSET]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFSET1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4064(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFSET1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFSET]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFSET1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4064(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFSET1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[COPY4]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFSET]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFSET1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4068(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4068(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4068(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY6]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY7]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY5]], [[REG_SEQUENCE3]], [[S_MOV_B32_1]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_5000(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_5000(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_5000(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 936, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 952, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4076(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4076(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4076(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4080(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4080(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 16, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4080(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY5]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY6]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFEN1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFEN [[COPY4]], [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128), align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFEN]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFEN1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY10:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_offset_4064(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX6:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX6:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX6:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX6:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX6:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX6:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
   ; GFX6:   [[BUFFER_LOAD_DWORDX4_OFFSET1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
-  ; GFX6:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX6:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX6:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX6:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX6: bb.3:
   ; GFX6:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX6: bb.4:
   ; GFX6:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFSET]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFSET1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX6:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX6:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX6:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX6:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_offset_4064(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX7:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX7:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX7:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX7:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX7:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX7:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
   ; GFX7:   [[BUFFER_LOAD_DWORDX4_OFFSET1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
-  ; GFX7:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX7:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX7:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX7:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX7: bb.3:
   ; GFX7:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX7: bb.4:
   ; GFX7:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFSET]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFSET1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX7:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX7:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX7:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX7:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_offset_4064(<4 x i32> %rsrc, i32 %offset.base) {
   ; GFX8:   [[REG_SEQUENCE1:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE1]], [[COPY4]], implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub0, implicit $exec
   ; GFX8:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY5]].sub1, implicit $exec
   ; GFX8:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX8:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY5]], implicit $exec
   ; GFX8:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX8:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
+  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFSET:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4064, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
   ; GFX8:   [[BUFFER_LOAD_DWORDX4_OFFSET1:%[0-9]+]]:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET [[REG_SEQUENCE3]], [[S_MOV_B32_]], 4080, 0, 0, 0, implicit $exec :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
-  ; GFX8:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX8:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX8:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX8:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX8: bb.3:
   ; GFX8:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX8: bb.4:
   ; GFX8:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_256 = REG_SEQUENCE [[BUFFER_LOAD_DWORDX4_OFFSET]], %subreg.sub0_sub1_sub2_sub3, [[BUFFER_LOAD_DWORDX4_OFFSET1]], %subreg.sub4_sub5_sub6_sub7
   ; GFX8:   [[COPY6:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub0
   ; GFX8:   [[COPY7:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub1
   ; GFX8:   [[COPY8:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub2
   ; GFX8:   [[COPY9:%[0-9]+]]:vgpr_32 = COPY [[REG_SEQUENCE4]].sub3

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.atomic.add.ll

define amdgpu_ps float @struct_buffer_atomic_add_i32__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; CHECK:   [[BUFFER_ATOMIC_ADD_BOTHEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_BOTHEN_RTN [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_ATOMIC_ADD_BOTHEN_RTN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %ret = call i32 @llvm.amdgcn.struct.buffer.atomic.add.i32(i32 %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   %cast = bitcast i32 %ret to float
define amdgpu_ps void @struct_buffer_atomic_add_i32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; CHECK:   [[BUFFER_ATOMIC_ADD_BOTHEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_BOTHEN_RTN [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %ret = call i32 @llvm.amdgcn.struct.buffer.atomic.add.i32(i32 %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.atomic.cmpswap.ll

define amdgpu_ps float @struct_buffer_atomic_cmpswap_i32__sgpr_val__sgpr_cmp__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, i32 inreg %cmp, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY14]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY14]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY11]], %subreg.sub0, [[COPY12]], %subreg.sub1
   ; CHECK:   [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; CHECK:   [[BUFFER_ATOMIC_CMPSWAP_BOTHEN_RTN:%[0-9]+]]:vreg_64 = BUFFER_ATOMIC_CMPSWAP_BOTHEN_RTN [[REG_SEQUENCE5]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   [[COPY15:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_ATOMIC_CMPSWAP_BOTHEN_RTN]].sub0
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[COPY15]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %ret = call i32 @llvm.amdgcn.struct.buffer.atomic.cmpswap.i32(i32 %val, i32 %cmp, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   %cast = bitcast i32 %ret to float
define amdgpu_ps void @struct_buffer_atomic_cmpswap_i32_noret__sgpr_val__sgpr_cmp__vgpr_rsrc__sgpr_voffset__vgpr_soffset(i32 inreg %val, i32 inreg %cmp, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY14]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY14]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY8]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY8]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY11]], %subreg.sub0, [[COPY12]], %subreg.sub1
   ; CHECK:   [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; CHECK:   [[BUFFER_ATOMIC_CMPSWAP_BOTHEN_RTN:%[0-9]+]]:vreg_64 = BUFFER_ATOMIC_CMPSWAP_BOTHEN_RTN [[REG_SEQUENCE5]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; CHECK:   [[COPY15:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_ATOMIC_CMPSWAP_BOTHEN_RTN]].sub0
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   %ret = call i32 @llvm.amdgcn.struct.buffer.atomic.cmpswap.i32(i32 %val, i32 %cmp, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.atomic.fadd.ll

define amdgpu_ps void @struct_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; GFX908:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; GFX908:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX908:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; GFX908:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX908:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX908:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; GFX908:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; GFX908:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; GFX908:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX908:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; GFX908:   BUFFER_ATOMIC_ADD_F32_BOTHEN [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
-  ; GFX908:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX908:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX908:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX908:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX908: bb.3:
   ; GFX908:   successors: %bb.4(0x80000000)
   ; GFX908:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX908: bb.4:
   ; GFX908:   S_ENDPGM 0
   ; GFX90A-LABEL: name: struct_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset
   ; GFX90A: bb.1 (%ir-block.0):
   ; GFX90A:   successors: %bb.2(0x80000000)
define amdgpu_ps void @struct_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__sgpr_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; GFX90A:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; GFX90A:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX90A:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; GFX90A:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX90A:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX90A:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; GFX90A:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; GFX90A:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; GFX90A:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX90A:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64_align2 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_BOTHEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_BOTHEN_RTN [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
-  ; GFX90A:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GFX90A:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX90A:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX90A:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX90A: bb.3:
   ; GFX90A:   successors: %bb.4(0x80000000)
   ; GFX90A:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX90A: bb.4:
   ; GFX90A:   S_ENDPGM 0
   %ret = call float @llvm.amdgcn.struct.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @struct_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__0_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 %soffset) {
   ; GFX908:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; GFX908:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX908:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; GFX908:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX908:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX908:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; GFX908:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; GFX908:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; GFX908:   BUFFER_ATOMIC_ADD_F32_IDXEN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX908:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX908:   BUFFER_ATOMIC_ADD_F32_IDXEN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX908:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX908:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX908:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX908: bb.3:
   ; GFX908:   successors: %bb.4(0x80000000)
   ; GFX908:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX908: bb.4:
   ; GFX908:   S_ENDPGM 0
   ; GFX90A-LABEL: name: struct_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__0_voffset__vgpr_soffset
   ; GFX90A: bb.1 (%ir-block.0):
   ; GFX90A:   successors: %bb.2(0x80000000)
define amdgpu_ps void @struct_buffer_atomic_add_f32_noret__sgpr_val__vgpr_rsrc__0_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 %soffset) {
   ; GFX90A:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; GFX90A:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; GFX90A:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; GFX90A:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GFX90A:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; GFX90A:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; GFX90A:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; GFX90A:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_IDXEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_IDXEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX90A:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GFX90A:   [[BUFFER_ATOMIC_ADD_F32_IDXEN_RTN:%[0-9]+]]:vgpr_32 = BUFFER_ATOMIC_ADD_F32_IDXEN_RTN [[COPY7]], [[COPY8]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 1, implicit $exec :: (volatile dereferenceable load store (s32), align 1, addrspace 4)
   ; GFX90A:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GFX90A:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GFX90A:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GFX90A: bb.3:
   ; GFX90A:   successors: %bb.4(0x80000000)
   ; GFX90A:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GFX90A: bb.4:
   ; GFX90A:   S_ENDPGM 0
   %ret = call float @llvm.amdgcn.struct.buffer.atomic.fadd.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 0, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.load.format.f16.ll

define amdgpu_ps <4 x half> @struct_buffer_load_format_v4f16__vpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; UNPACKED:   [[BUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (<4 x s16>), align 1, addrspace 4)
-  ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub0
   ; UNPACKED:   [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub1
   ; UNPACKED:   [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub2
   ; UNPACKED:   [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub3
define amdgpu_ps <4 x half> @struct_buffer_load_format_v4f16__vpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; PACKED:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; PACKED:   [[BUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN:%[0-9]+]]:vreg_64 = BUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (<4 x s16>), align 1, addrspace 4)
-  ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN]].sub0
   ; PACKED:   [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN]].sub1
   ; PACKED:   $vgpr0 = COPY [[COPY11]]
   ; PACKED:   $vgpr1 = COPY [[COPY12]]

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.load.format.ll

define amdgpu_ps <4 x float> @struct_buffer_load_format_v4f32__vpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; CHECK:   [[BUFFER_LOAD_FORMAT_XYZW_BOTHEN:%[0-9]+]]:vreg_128 = BUFFER_LOAD_FORMAT_XYZW_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (<4 x s32>), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub0
   ; CHECK:   [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub1
   ; CHECK:   [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub2
   ; CHECK:   [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[BUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub3

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.load.ll

define amdgpu_ps float @struct_buffer_load_f32__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; CHECK:   [[BUFFER_LOAD_DWORD_BOTHEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[BUFFER_LOAD_DWORD_BOTHEN]]
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret float %val

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.store.format.f16.ll

define amdgpu_ps void @struct_buffer_store_format_f16__sgpr_val__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(half inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; UNPACKED:   BUFFER_STORE_FORMAT_D16_X_gfx80_BOTHEN_exact [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
-  ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   S_ENDPGM 0
   ; PACKED-LABEL: name: struct_buffer_store_format_f16__sgpr_val__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset
   ; PACKED: bb.1 (%ir-block.0):
   ; PACKED:   successors: %bb.2(0x80000000)
define amdgpu_ps void @struct_buffer_store_format_f16__sgpr_val__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(half inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; PACKED:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; PACKED:   BUFFER_STORE_FORMAT_D16_X_BOTHEN_exact [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s16), align 1, addrspace 4)
-  ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   S_ENDPGM 0
   call void @llvm.amdgcn.struct.buffer.store.format.f16(half %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.store.format.f32.ll

define amdgpu_ps void @struct_buffer_store_format_f32__sgpr_val__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(float inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY12]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY12]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY7]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY7]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY9]], %subreg.sub0, [[COPY10]], %subreg.sub1
   ; CHECK:   BUFFER_STORE_FORMAT_X_BOTHEN_exact [[COPY8]], [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s32), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.struct.buffer.store.format.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.buffer.store.ll

define amdgpu_ps void @struct_buffer_store_v4f32_vgpr_rsrc__sgpr_val__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x float> inreg %val, <4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY15]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE3]], [[COPY15]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY10]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE5:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY12]], %subreg.sub0, [[COPY13]], %subreg.sub1
   ; CHECK:   BUFFER_STORE_DWORDX4_BOTHEN_exact [[COPY11]], [[REG_SEQUENCE5]], [[REG_SEQUENCE4]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable store (<4 x s32>), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.struct.buffer.store.v4f32(<4 x float> %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.tbuffer.load.f16.ll

define amdgpu_ps <4 x half> @struct_tbuffer_load_v4f16__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; PACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; PACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; PACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; PACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; PACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; PACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; PACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; PACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; PACKED:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; PACKED:   [[TBUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN:%[0-9]+]]:vreg_64 = TBUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (<4 x s16>), align 1, addrspace 4)
-  ; PACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; PACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; PACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; PACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; PACKED: bb.3:
   ; PACKED:   successors: %bb.4(0x80000000)
   ; PACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; PACKED: bb.4:
   ; PACKED:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN]].sub0
   ; PACKED:   [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_D16_XYZW_BOTHEN]].sub1
   ; PACKED:   $vgpr0 = COPY [[COPY11]]
   ; PACKED:   $vgpr1 = COPY [[COPY12]]
define amdgpu_ps <4 x half> @struct_tbuffer_load_v4f16__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; UNPACKED:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; UNPACKED:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; UNPACKED:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; UNPACKED:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; UNPACKED:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; UNPACKED:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; UNPACKED:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; UNPACKED:   [[TBUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (<4 x s16>), align 1, addrspace 4)
-  ; UNPACKED:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; UNPACKED:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; UNPACKED:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; UNPACKED:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; UNPACKED: bb.3:
   ; UNPACKED:   successors: %bb.4(0x80000000)
   ; UNPACKED:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; UNPACKED: bb.4:
   ; UNPACKED:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub0
   ; UNPACKED:   [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub1
   ; UNPACKED:   [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub2
   ; UNPACKED:   [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_D16_XYZW_gfx80_BOTHEN]].sub3

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.struct.tbuffer.load.ll

define amdgpu_ps <4 x float> @struct_tbuffer_load_v4f32__vgpr_rsrc__sgpr_vindex__sgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 inreg %vindex, i32 inreg %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY10]].sub1, implicit $exec
   ; CHECK:   [[REG_SEQUENCE2:%[0-9]+]]:sreg_64_xexec = REG_SEQUENCE [[V_READFIRSTLANE_B32_2]], %subreg.sub0, [[V_READFIRSTLANE_B32_3]], %subreg.sub1
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[REG_SEQUENCE2]], [[COPY10]], implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[REG_SEQUENCE3:%[0-9]+]]:sgpr_128 = REG_SEQUENCE [[V_READFIRSTLANE_B32_]], %subreg.sub0, [[V_READFIRSTLANE_B32_1]], %subreg.sub1, [[V_READFIRSTLANE_B32_2]], %subreg.sub2, [[V_READFIRSTLANE_B32_3]], %subreg.sub3
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0 = V_READFIRSTLANE_B32 [[COPY6]], implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]], [[COPY6]], implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[REG_SEQUENCE4:%[0-9]+]]:vreg_64 = REG_SEQUENCE [[COPY7]], %subreg.sub0, [[COPY8]], %subreg.sub1
   ; CHECK:   [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN:%[0-9]+]]:vreg_128 = TBUFFER_LOAD_FORMAT_XYZW_BOTHEN [[REG_SEQUENCE4]], [[REG_SEQUENCE3]], [[V_READFIRSTLANE_B32_4]], 0, 78, 0, 0, 0, implicit $exec :: (dereferenceable load (<4 x s32>), align 1, addrspace 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[COPY11:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub0
   ; CHECK:   [[COPY12:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub1
   ; CHECK:   [[COPY13:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub2
   ; CHECK:   [[COPY14:%[0-9]+]]:vgpr_32 = COPY [[TBUFFER_LOAD_FORMAT_XYZW_BOTHEN]].sub3

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

bb.0:
     ; CHECK: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
     ; CHECK: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
     ; CHECK: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
     ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY2]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
     ; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY2]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
     ; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-    ; CHECK: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+    ; CHECK: SI_WATERFALL_LOOP %bb.1, implicit $exec
     ; CHECK: .2:
     ; CHECK: successors: %bb.3(0x80000000)
     ; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
     ; CHECK: .3:
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = COPY $sgpr0
     %2:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD %0, %1, 0
 
bb.0:
     ; CHECK: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
     ; CHECK: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
     ; CHECK: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
     ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY1]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
     ; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+    ; CHECK: [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[C1]](s32), [[COPY1]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
     ; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-    ; CHECK: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+    ; CHECK: SI_WATERFALL_LOOP %bb.1, implicit $exec
     ; CHECK: .2:
     ; CHECK: successors: %bb.3(0x80000000)
     ; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
     ; CHECK: .3:
     %0:_(<4 x s32>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
     %1:_(s32) = COPY $vgpr4
     %2:_(<4 x s32>) = G_AMDGPU_S_BUFFER_LOAD %0, %1, 0
 
 ...

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

define amdgpu_ps void @load_1d_vgpr_vaddr__vgpr_srsrc(<8 x i32> %rsrc, i32 %s) {
   ; FAST:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; FAST:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; FAST:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
   ; FAST:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY $vgpr6
   ; FAST:   [[COPY7:%[0-9]+]]:vgpr(s32) = COPY $vgpr7
   ; FAST:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; FAST:   [[COPY8:%[0-9]+]]:vgpr(s32) = COPY $vgpr8
   ; FAST:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; FAST:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; FAST:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; FAST:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; FAST: bb.2:
   ; FAST:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %19, %bb.2
-  ; FAST:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %12(<4 x s32>), %bb.2
+  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
   ; FAST:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; FAST:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; FAST:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
   ; FAST:   [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
   ; FAST:   [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
   ; FAST:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY8]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY8]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; FAST:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; FAST:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; FAST: bb.3:
   ; FAST:   successors: %bb.4(0x80000000)
   ; FAST:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; FAST: bb.4:
   ; FAST:   [[COPY9:%[0-9]+]]:vgpr(p1) = COPY [[DEF]](p1)
   ; FAST:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[COPY9]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; FAST:   S_ENDPGM 0
   ; GREEDY-LABEL: name: load_1d_vgpr_vaddr__vgpr_srsrc
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   successors: %bb.2(0x80000000)
   ; GREEDY:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; GREEDY:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
   ; GREEDY:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY $vgpr6
   ; GREEDY:   [[COPY7:%[0-9]+]]:vgpr(s32) = COPY $vgpr7
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GREEDY:   [[COPY8:%[0-9]+]]:vgpr(s32) = COPY $vgpr8
   ; GREEDY:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %19, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %12(<4 x s32>), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY8]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY8]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   successors: %bb.4(0x80000000)
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[COPY9:%[0-9]+]]:vgpr(p1) = COPY [[DEF]](p1)
   ; GREEDY:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[COPY9]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; GREEDY:   S_ENDPGM 0
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
define amdgpu_ps void @load_1d_sgpr_vaddr__vgpr_srsrc(<8 x i32> %rsrc, i32 inreg %s) {
   ; FAST:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; FAST:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
   ; FAST:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY $vgpr6
   ; FAST:   [[COPY7:%[0-9]+]]:vgpr(s32) = COPY $vgpr7
   ; FAST:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; FAST:   [[COPY8:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; FAST:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
   ; FAST:   [[COPY9:%[0-9]+]]:vgpr(s32) = COPY [[COPY8]](s32)
-  ; FAST:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; FAST:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; FAST:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; FAST:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; FAST: bb.2:
   ; FAST:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %20, %bb.2
-  ; FAST:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %12(<4 x s32>), %bb.2
+  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %18, %bb.2
   ; FAST:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; FAST:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; FAST:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
   ; FAST:   [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
   ; FAST:   [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
   ; FAST:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY9]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY9]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; FAST:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; FAST:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; FAST: bb.3:
   ; FAST:   successors: %bb.4(0x80000000)
   ; FAST:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; FAST: bb.4:
   ; FAST:   [[COPY10:%[0-9]+]]:vgpr(p1) = COPY [[DEF]](p1)
   ; FAST:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[COPY10]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; FAST:   S_ENDPGM 0
   ; GREEDY-LABEL: name: load_1d_sgpr_vaddr__vgpr_srsrc
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   successors: %bb.2(0x80000000)
   ; GREEDY:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; GREEDY:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
   ; GREEDY:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY $vgpr6
   ; GREEDY:   [[COPY7:%[0-9]+]]:vgpr(s32) = COPY $vgpr7
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GREEDY:   [[COPY8:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
   ; GREEDY:   [[COPY9:%[0-9]+]]:vgpr(s32) = COPY [[COPY8]](s32)
-  ; GREEDY:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %20, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %12(<4 x s32>), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %18, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY9]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY9]](s32), [[BUILD_VECTOR1]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   successors: %bb.4(0x80000000)
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[COPY10:%[0-9]+]]:vgpr(p1) = COPY [[DEF]](p1)
   ; GREEDY:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[COPY10]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; GREEDY:   S_ENDPGM 0
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   store <4 x float> %v, <4 x float> addrspace(1)* undef
   ret void
 }
 
 declare <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 immarg, i32, <8 x i32>, i32 immarg, i32 immarg) #0
 
 attributes #0 = { nounwind readonly }

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.image.sample.1d.ll

define amdgpu_ps void @sample_1d_vgpr_vaddr__vgpr_rsrc__sgpr_samp(<8 x i32> %rsrc, <4 x i32> inreg %samp, float %s) {
   ; FAST:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; FAST:   [[COPY8:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; FAST:   [[COPY9:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; FAST:   [[COPY10:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; FAST:   [[COPY11:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; FAST:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
   ; FAST:   [[COPY12:%[0-9]+]]:vgpr(s32) = COPY $vgpr8
   ; FAST:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; FAST:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; FAST:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; FAST:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; FAST: bb.2:
   ; FAST:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %24, %bb.2
-  ; FAST:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %17(<4 x s32>), %bb.2
+  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %22, %bb.2
   ; FAST:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; FAST:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; FAST:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
   ; FAST:   [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
   ; FAST:   [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
   ; FAST:   [[BUILD_VECTOR2:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; FAST:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; FAST:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; FAST: bb.3:
   ; FAST:   successors: %bb.4(0x80000000)
   ; FAST:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; FAST: bb.4:
   ; FAST:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[DEF]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; FAST:   S_ENDPGM 0
   ; GREEDY-LABEL: name: sample_1d_vgpr_vaddr__vgpr_rsrc__sgpr_samp
   ; GREEDY: bb.1 (%ir-block.0):
define amdgpu_ps void @sample_1d_vgpr_vaddr__vgpr_rsrc__sgpr_samp(<8 x i32> %rsrc, <4 x i32> inreg %samp, float %s) {
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GREEDY:   [[COPY8:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY:   [[COPY9:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; GREEDY:   [[COPY10:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; GREEDY:   [[COPY11:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
   ; GREEDY:   [[COPY12:%[0-9]+]]:vgpr(s32) = COPY $vgpr8
   ; GREEDY:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %24, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %17(<4 x s32>), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %22, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR2:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR1]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   successors: %bb.4(0x80000000)
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[DEF]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; GREEDY:   S_ENDPGM 0
   %v = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   store <4 x float> %v, <4 x float> addrspace(1)* undef
define amdgpu_ps void @sample_1d_vgpr_vaddr__sgpr_rsrc__vgpr_samp(<8 x i32> inreg %rsrc, <4 x i32> %samp, float %s) {
   ; FAST:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; FAST:   [[COPY8:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; FAST:   [[COPY9:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; FAST:   [[COPY10:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; FAST:   [[COPY11:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; FAST:   [[BUILD_VECTOR1:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
   ; FAST:   [[COPY12:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; FAST:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; FAST:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; FAST:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR1]](<4 x s32>)
   ; FAST:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; FAST: bb.2:
   ; FAST:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %24, %bb.2
-  ; FAST:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %17(<4 x s32>), %bb.2
+  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %22, %bb.2
   ; FAST:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; FAST:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; FAST:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; FAST:   [[BUILD_VECTOR2:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR2]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR2]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; FAST:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; FAST:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; FAST: bb.3:
   ; FAST:   successors: %bb.4(0x80000000)
   ; FAST:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; FAST: bb.4:
   ; FAST:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[DEF]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; FAST:   S_ENDPGM 0
   ; GREEDY-LABEL: name: sample_1d_vgpr_vaddr__sgpr_rsrc__vgpr_samp
   ; GREEDY: bb.1 (%ir-block.0):
define amdgpu_ps void @sample_1d_vgpr_vaddr__sgpr_rsrc__vgpr_samp(<8 x i32> inreg %rsrc, <4 x i32> %samp, float %s) {
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GREEDY:   [[COPY8:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY9:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY10:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY11:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
   ; GREEDY:   [[COPY12:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; GREEDY:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR1]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %24, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %17(<4 x s32>), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %22, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR2:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR2]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR]](<8 x s32>), [[BUILD_VECTOR2]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   successors: %bb.4(0x80000000)
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[DEF]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; GREEDY:   S_ENDPGM 0
   %v = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   store <4 x float> %v, <4 x float> addrspace(1)* undef
define amdgpu_ps void @sample_1d_vgpr_vaddr__vgpr_rsrc__vgpr_samp(<8 x i32> %rsrc, <4 x i32> %samp, float %s) {
   ; FAST:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; FAST:   [[COPY8:%[0-9]+]]:vgpr(s32) = COPY $vgpr8
   ; FAST:   [[COPY9:%[0-9]+]]:vgpr(s32) = COPY $vgpr9
   ; FAST:   [[COPY10:%[0-9]+]]:vgpr(s32) = COPY $vgpr10
   ; FAST:   [[COPY11:%[0-9]+]]:vgpr(s32) = COPY $vgpr11
   ; FAST:   [[BUILD_VECTOR1:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
   ; FAST:   [[COPY12:%[0-9]+]]:vgpr(s32) = COPY $vgpr12
   ; FAST:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; FAST:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; FAST:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; FAST:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; FAST:   [[UV4:%[0-9]+]]:vreg_64(s64), [[UV5:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR1]](<4 x s32>)
   ; FAST:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; FAST: bb.2:
   ; FAST:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %24, %bb.2
-  ; FAST:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %17(<4 x s32>), %bb.2
+  ; FAST:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %22, %bb.2
   ; FAST:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; FAST:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; FAST:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
define amdgpu_ps void @sample_1d_vgpr_vaddr__vgpr_rsrc__vgpr_samp(<8 x i32> %rsrc, <4 x i32> %samp, float %s) {
   ; FAST:   [[V_CMP_EQ_U64_e64_4:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV4]](s64), [[UV4]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_3:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_4]], [[S_AND_B64_2]], implicit-def $scc
   ; FAST:   [[V_READFIRSTLANE_B32_10:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV5]].sub0(s64), implicit $exec
   ; FAST:   [[V_READFIRSTLANE_B32_11:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV5]].sub1(s64), implicit $exec
   ; FAST:   [[MV5:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_10]](s32), [[V_READFIRSTLANE_B32_11]](s32)
   ; FAST:   [[V_CMP_EQ_U64_e64_5:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV5]](s64), [[UV5]](s64), implicit $exec
   ; FAST:   [[S_AND_B64_4:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_5]], [[S_AND_B64_3]], implicit-def $scc
   ; FAST:   [[BUILD_VECTOR3:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_8]](s32), [[V_READFIRSTLANE_B32_9]](s32), [[V_READFIRSTLANE_B32_10]](s32), [[V_READFIRSTLANE_B32_11]](s32)
-  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR3]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_4]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; FAST:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR3]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; FAST:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; FAST:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; FAST:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; FAST: bb.3:
   ; FAST:   successors: %bb.4(0x80000000)
   ; FAST:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; FAST: bb.4:
   ; FAST:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[DEF]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; FAST:   S_ENDPGM 0
   ; GREEDY-LABEL: name: sample_1d_vgpr_vaddr__vgpr_rsrc__vgpr_samp
   ; GREEDY: bb.1 (%ir-block.0):
define amdgpu_ps void @sample_1d_vgpr_vaddr__vgpr_rsrc__vgpr_samp(<8 x i32> %rsrc, <4 x i32> %samp, float %s) {
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
   ; GREEDY:   [[COPY8:%[0-9]+]]:vgpr(s32) = COPY $vgpr8
   ; GREEDY:   [[COPY9:%[0-9]+]]:vgpr(s32) = COPY $vgpr9
   ; GREEDY:   [[COPY10:%[0-9]+]]:vgpr(s32) = COPY $vgpr10
   ; GREEDY:   [[COPY11:%[0-9]+]]:vgpr(s32) = COPY $vgpr11
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32), [[COPY10]](s32), [[COPY11]](s32)
   ; GREEDY:   [[COPY12:%[0-9]+]]:vgpr(s32) = COPY $vgpr12
   ; GREEDY:   [[DEF:%[0-9]+]]:sgpr(p1) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF2:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<8 x s32>)
   ; GREEDY:   [[UV4:%[0-9]+]]:vreg_64(s64), [[UV5:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR1]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF2]], %bb.1, %24, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF1]](<4 x s32>), %bb.1, %17(<4 x s32>), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %22, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
define amdgpu_ps void @sample_1d_vgpr_vaddr__vgpr_rsrc__vgpr_samp(<8 x i32> %rsrc, <4 x i32> %samp, float %s) {
   ; GREEDY:   [[V_CMP_EQ_U64_e64_4:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV4]](s64), [[UV4]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_3:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_4]], [[S_AND_B64_2]], implicit-def $scc
   ; GREEDY:   [[V_READFIRSTLANE_B32_10:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV5]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_11:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV5]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV5:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_10]](s32), [[V_READFIRSTLANE_B32_11]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_5:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV5]](s64), [[UV5]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_4:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_5]], [[S_AND_B64_3]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR3:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_8]](s32), [[V_READFIRSTLANE_B32_9]](s32), [[V_READFIRSTLANE_B32_10]](s32), [[V_READFIRSTLANE_B32_11]](s32)
-  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR3]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_4]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.sample.1d), 15, [[COPY12]](s32), [[BUILD_VECTOR2]](<8 x s32>), [[BUILD_VECTOR3]](<4 x s32>), 0, 0, 0, 0 :: (dereferenceable load (<4 x s32>) from custom "ImageResource")
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   successors: %bb.4(0x80000000)
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   G_STORE [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>), [[DEF]](p1) :: (store (<4 x s32>) into `<4 x float> addrspace(1)* undef`, addrspace 1)
   ; GREEDY:   S_ENDPGM 0
   %v = call <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 15, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   store <4 x float> %v, <4 x float> addrspace(1)* undef
   ret void
 }
 
 declare <4 x float> @llvm.amdgcn.image.sample.1d.v4f32.f32(i32 immarg, float, <8 x i32>, <4 x i32>, i1 immarg, i32 immarg, i32 immarg) #0
 
 attributes #0 = { nounwind readonly }

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

define amdgpu_ps float @raw_buffer_load__vgpr_rsrc__vgpr_val__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; CHECK:   [[COPY5:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[C]](s32)
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %9(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY6]](s32), [[COPY4]], [[COPY5]], 0, 0, 0 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY6]](s32), [[COPY4]], [[COPY5]], 0, 0, 0 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps float @raw_buffer_load__sgpr_rsrc__vgpr_val__vgpr_voffset__vgpr_soffset(<4 x i32> inreg %rsrc, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[C]](s32)
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %9(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY5]](s32), implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_]](s32), [[COPY5]](s32), implicit $exec
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY6]](s32), [[COPY4]], [[V_READFIRSTLANE_B32_]], 0, 0, 0 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[V_CMP_EQ_U32_e64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY6]](s32), [[COPY4]], [[V_READFIRSTLANE_B32_]], 0, 0, 0 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps float @raw_buffer_load__vgpr_rsrc__vgpr_val__vgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr5
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr(s32) = COPY [[C]](s32)
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %9(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY5]](s32), implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]](s32), [[COPY5]](s32), implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY6]](s32), [[COPY4]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY6]](s32), [[COPY4]], [[V_READFIRSTLANE_B32_4]], 0, 0, 0 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32> %rsrc, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
 }
 
 declare float @llvm.amdgcn.raw.buffer.load.f32(<4 x i32>, i32, i32, i32 immarg)

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

define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc(<4 x i32> %rsrc, i32 inreg %soffset) {
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[COPY4]](s32)
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3, %bb.2
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %8(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY5]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY5]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GREEDY-LABEL: name: s_buffer_load_f32_vgpr_rsrc
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[COPY4]](s32)
   ; GREEDY:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; GREEDY:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; GREEDY:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3, %bb.2
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %8(s32), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY5]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY5]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; GREEDY:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret float %val
 }
 
 ; Use the offset inside the waterfall loop
 define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4092(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; CHECK-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4092
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4092
   ; CHECK:   [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[COPY4]], [[C]]
   ; CHECK:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[C2:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3, %bb.2
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %18, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %16, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4092, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4092, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GREEDY-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4092
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4092
   ; GREEDY:   [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[COPY4]], [[C]]
   ; GREEDY:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
   ; GREEDY:   [[C2:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; GREEDY:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3, %bb.2
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %18, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %16, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4092, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4092, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; GREEDY:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %soffset = add i32 %soffset.base, 4092
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret float %val
define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_soffset_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4096
   ; CHECK:   [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[COPY4]], [[C]]
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[ADD]](s32)
   ; CHECK:   [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[C2:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3, %bb.2
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %19, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %17, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GREEDY-LABEL: name: s_buffer_load_f32_vgpr_rsrc_soffset_add_4096
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   liveins: $sgpr2, $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY:   [[COPY4:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; GREEDY:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4096
   ; GREEDY:   [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[COPY4]], [[C]]
   ; GREEDY:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY [[ADD]](s32)
   ; GREEDY:   [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; GREEDY:   [[C2:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; GREEDY:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3, %bb.2
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %19, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %17, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; GREEDY:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %soffset = add i32 %soffset.base, 4096
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 %soffset, i32 0)
   ret float %val
 }
 
 ; Waterfall loop, but constant offset
 define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4095(<4 x i32> %rsrc) {
   ; CHECK-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4095
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4095
   ; CHECK:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[C3:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3, %bb.2
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %7(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4095, 0, 0 :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4095, 0, 0 :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GREEDY-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4095
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4095
   ; GREEDY:   [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
   ; GREEDY:   [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; GREEDY:   [[C3:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; GREEDY:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3, %bb.2
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %7(s32), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4095, 0, 0 :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4095, 0, 0 :: (dereferenceable invariant load (s32) from unknown-address + 4095, align 1)
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; GREEDY:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 4095, i32 0)
   ret float %val
 }
 
 ; Waterfall loop, but constant offset
 define amdgpu_ps float @s_buffer_load_f32_vgpr_rsrc_offset_4096(<4 x i32> %rsrc) {
   ; CHECK-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4096
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4096
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY [[C]](s32)
   ; CHECK:   [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; CHECK:   [[C2:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3, %bb.2
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %7(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY4]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY4]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   ; GREEDY-LABEL: name: s_buffer_load_f32_vgpr_rsrc_offset_4096
   ; GREEDY: bb.1 (%ir-block.0):
   ; GREEDY:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3
   ; GREEDY:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; GREEDY:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; GREEDY:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; GREEDY:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; GREEDY:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; GREEDY:   [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 4096
   ; GREEDY:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY [[C]](s32)
   ; GREEDY:   [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0
   ; GREEDY:   [[C2:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0
-  ; GREEDY:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; GREEDY:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; GREEDY:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; GREEDY:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; GREEDY:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; GREEDY: bb.2:
   ; GREEDY:   successors: %bb.3, %bb.2
-  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %17, %bb.2
-  ; GREEDY:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %7(s32), %bb.2
+  ; GREEDY:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %15, %bb.2
   ; GREEDY:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY4]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY4]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s32))
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; GREEDY:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.s.buffer.load.f32(<4 x i32> %rsrc, i32 4096, i32 0)
   ret float %val
 }
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4064(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4064, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4080, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4064(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4064, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[C1]], [[COPY4]], 4080, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4068(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4068(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_rsrc_add_4096(<4 x i32> %rsrc, i32 inreg %soffset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C2]](s32), [[COPY5]], [[C1]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_5000(<4 x i32> %rsrc, i32 %offset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_5000(<4 x i32> %rsrc, i32 %offset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4076(<4 x i32> %rsrc, i32 %offset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4076(<4 x i32> %rsrc, i32 %offset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4080(<4 x i32> %rsrc, i32 %offset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_add_4080(<4 x i32> %rsrc, i32 %offset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 0, 0, 0 :: (dereferenceable invariant load (s128), align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C1]](s32), [[COPY4]], [[C]], 16, 0, 0 :: (dereferenceable invariant load (s128), align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_offset_4064(<4 x i32> %rsrc, i32 %offset.base) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4064, 0, 0 :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
   ; CHECK:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4080, 0, 0 :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
-  ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; CHECK:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; CHECK:   $vgpr0 = COPY [[UV2]](s32)
   ; CHECK:   $vgpr1 = COPY [[UV3]](s32)
   ; CHECK:   $vgpr2 = COPY [[UV4]](s32)
define amdgpu_ps <8 x float> @s_buffer_load_v8f32_vgpr_offset_vgpr_rsrc_offset_4064(<4 x i32> %rsrc, i32 %offset.base) {
   ; GREEDY:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; GREEDY:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; GREEDY:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; GREEDY:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; GREEDY:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; GREEDY:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
+  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4064, 0, 0 :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
   ; GREEDY:   [[AMDGPU_BUFFER_LOAD1:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[C3]](s32), [[C1]], [[C2]], 4080, 0, 0 :: (dereferenceable invariant load (s128) from unknown-address + 4064, align 4)
-  ; GREEDY:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
   ; GREEDY:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; GREEDY:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; GREEDY:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; GREEDY: bb.3:
   ; GREEDY:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; GREEDY: bb.4:
   ; GREEDY:   [[CONCAT_VECTORS:%[0-9]+]]:vgpr(<8 x s32>) = G_CONCAT_VECTORS [[AMDGPU_BUFFER_LOAD]](<4 x s32>), [[AMDGPU_BUFFER_LOAD1]](<4 x s32>)
   ; GREEDY:   [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32), [[UV4:%[0-9]+]]:vgpr(s32), [[UV5:%[0-9]+]]:vgpr(s32), [[UV6:%[0-9]+]]:vgpr(s32), [[UV7:%[0-9]+]]:vgpr(s32), [[UV8:%[0-9]+]]:vgpr(s32), [[UV9:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<8 x s32>)
   ; GREEDY:   $vgpr0 = COPY [[UV2]](s32)
   ; GREEDY:   $vgpr1 = COPY [[UV3]](s32)
   ; GREEDY:   $vgpr2 = COPY [[UV4]](s32)

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

define amdgpu_ps float @struct_buffer_load__vgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
   ; CHECK:   [[COPY6:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %16, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %14, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY4]](s32), [[COPY5]], [[COPY6]], 0, 0, -1 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY4]](s32), [[COPY5]], [[COPY6]], 0, 0, -1 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
 }
 
 ; Waterfall for soffset
 define amdgpu_ps float @struct_buffer_load__sgpr_rsrc__vgpr_val__vgpr_vindex_vgpr_voffset__vgpr_soffset(<4 x i32> inreg %rsrc, i32 %vindex, i32 %voffset, i32 %soffset) {
   ; CHECK-LABEL: name: struct_buffer_load__sgpr_rsrc__vgpr_val__vgpr_vindex_vgpr_voffset__vgpr_soffset
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   successors: %bb.2(0x80000000)
   ; CHECK:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $vgpr0, $vgpr1, $vgpr2
   ; CHECK:   [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr2
   ; CHECK:   [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr3
   ; CHECK:   [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr4
   ; CHECK:   [[COPY3:%[0-9]+]]:sgpr(s32) = COPY $sgpr5
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %16, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %14, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY6]](s32), implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_]](s32), [[COPY6]](s32), implicit $exec
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY4]](s32), [[COPY5]], [[V_READFIRSTLANE_B32_]], 0, 0, -1 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[V_CMP_EQ_U32_e64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR]](<4 x s32>), [[COPY4]](s32), [[COPY5]], [[V_READFIRSTLANE_B32_]], 0, 0, -1 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
 }
 
 ; Waterfall for rsrc and soffset
 define amdgpu_ps float @struct_buffer_load__vgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset) {
   ; CHECK-LABEL: name: struct_buffer_load__vgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__vgpr_soffset
   ; CHECK: bb.1 (%ir-block.0):
   ; CHECK:   successors: %bb.2(0x80000000)
   ; CHECK:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6
   ; CHECK:   [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
   ; CHECK:   [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1
   ; CHECK:   [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2
   ; CHECK:   [[COPY3:%[0-9]+]]:vgpr(s32) = COPY $vgpr3
   ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:vgpr(<4 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32)
   ; CHECK:   [[COPY4:%[0-9]+]]:vgpr(s32) = COPY $vgpr4
   ; CHECK:   [[COPY5:%[0-9]+]]:vgpr(s32) = COPY $vgpr5
   ; CHECK:   [[COPY6:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr6
-  ; CHECK:   [[DEF:%[0-9]+]]:vgpr(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+  ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[BUILD_VECTOR]](<4 x s32>)
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
-  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.1, %16, %bb.2
-  ; CHECK:   [[PHI1:%[0-9]+]]:vgpr(s32) = G_PHI [[DEF]](s32), %bb.1, %10(s32), %bb.2
+  ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %14, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY6]](s32), implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]](s32), [[COPY6]](s32), implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY4]](s32), [[COPY5]], [[V_READFIRSTLANE_B32_4]], 0, 0, -1 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   [[AMDGPU_BUFFER_LOAD:%[0-9]+]]:vgpr(s32) = G_AMDGPU_BUFFER_LOAD [[BUILD_VECTOR1]](<4 x s32>), [[COPY4]](s32), [[COPY5]], [[V_READFIRSTLANE_B32_4]], 0, 0, -1 :: (dereferenceable load (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   $vgpr0 = COPY [[AMDGPU_BUFFER_LOAD]](s32)
   ; CHECK:   SI_RETURN_TO_EPILOG implicit $vgpr0
   %val = call float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret float %val
 }
 
 declare float @llvm.amdgcn.struct.buffer.load.f32(<4 x i32>, i32, i32, i32, i32 immarg)

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-amdgcn.struct.buffer.store.ll

define amdgpu_ps void @struct_buffer_store__vgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__sgpr_soffset(<4 x i32> %rsrc, float %val, i32 %vindex, i32 %voffset, i32 inreg %soffset) {
   ; CHECK:   [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
-  ; CHECK:   G_AMDGPU_BUFFER_STORE [[COPY4]](s32), [[BUILD_VECTOR1]](<4 x s32>), [[COPY5]](s32), [[COPY6]], [[COPY7]], 0, 0, -1 :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   G_AMDGPU_BUFFER_STORE [[COPY4]](s32), [[BUILD_VECTOR1]](<4 x s32>), [[COPY5]](s32), [[COPY6]], [[COPY7]], 0, 0, -1 :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.struct.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @struct_buffer_store__sgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__vgpr_soffset(<4 x i32> inreg %rsrc, float %val, i32 %vindex, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[COPY7:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr3
   ; CHECK:   [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
   ; CHECK:   [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
   ; CHECK: bb.2:
   ; CHECK:   successors: %bb.3(0x40000000), %bb.2(0x40000000)
   ; CHECK:   [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.1, %14, %bb.2
   ; CHECK:   [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY7]](s32), implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_]](s32), [[COPY7]](s32), implicit $exec
-  ; CHECK:   G_AMDGPU_BUFFER_STORE [[COPY4]](s32), [[BUILD_VECTOR]](<4 x s32>), [[COPY5]](s32), [[COPY6]], [[V_READFIRSTLANE_B32_]], 0, 0, -1 :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[V_CMP_EQ_U32_e64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   G_AMDGPU_BUFFER_STORE [[COPY4]](s32), [[BUILD_VECTOR]](<4 x s32>), [[COPY5]](s32), [[COPY6]], [[V_READFIRSTLANE_B32_]], 0, 0, -1 :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.struct.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
define amdgpu_ps void @struct_buffer_store__vgpr_rsrc__vgpr_val__vgpr_vindex__vgpr_voffset__vgpr_soffset(<4 x i32> %rsrc, float %val, i32 %vindex, i32 %voffset, i32 %soffset) {
   ; CHECK:   [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
   ; CHECK:   [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
   ; CHECK:   [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
   ; CHECK:   [[BUILD_VECTOR1:%[0-9]+]]:sgpr(<4 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
   ; CHECK:   [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY7]](s32), implicit $exec
   ; CHECK:   [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_4]](s32), [[COPY7]](s32), implicit $exec
   ; CHECK:   [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U32_e64_]], [[S_AND_B64_]], implicit-def $scc
-  ; CHECK:   G_AMDGPU_BUFFER_STORE [[COPY4]](s32), [[BUILD_VECTOR1]](<4 x s32>), [[COPY5]](s32), [[COPY6]], [[V_READFIRSTLANE_B32_4]], 0, 0, -1 :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_1]], implicit-def $exec, implicit-def $scc, implicit $exec
+  ; CHECK:   G_AMDGPU_BUFFER_STORE [[COPY4]](s32), [[BUILD_VECTOR1]](<4 x s32>), [[COPY5]](s32), [[COPY6]], [[V_READFIRSTLANE_B32_4]], 0, 0, -1 :: (dereferenceable store (s32), align 1, addrspace 4)
   ; CHECK:   $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-  ; CHECK:   S_CBRANCH_EXECNZ %bb.2, implicit $exec
+  ; CHECK:   SI_WATERFALL_LOOP %bb.2, implicit $exec
   ; CHECK: bb.3:
   ; CHECK:   successors: %bb.4(0x80000000)
   ; CHECK:   $exec = S_MOV_B64_term [[S_MOV_B64_term]]
   ; CHECK: bb.4:
   ; CHECK:   S_ENDPGM 0
   call void @llvm.amdgcn.struct.buffer.store.f32(float %val, <4 x i32> %rsrc, i32 %vindex, i32 %voffset, i32 %soffset, i32 0)
   ret void
 }
 
 declare void @llvm.amdgcn.struct.buffer.store.f32(float, <4 x i32>, i32, i32, i32, i32 immarg)

llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-waterfall-agpr.mir

bb.0:
     ; CHECK: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
     ; CHECK: [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY %agpr(s32)
     ; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
     ; CHECK: .1:
     ; CHECK: successors: %bb.2(0x40000000), %bb.1(0x40000000)
     ; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.0, %9, %bb.1
     ; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[COPY1]](s32), implicit $exec
     ; CHECK: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[V_READFIRSTLANE_B32_]](s32), [[COPY1]](s32), implicit $exec
-    ; CHECK: G_AMDGPU_BUFFER_STORE %val(s32), %rsrc(<4 x s32>), [[COPY]](s32), %voffset, [[V_READFIRSTLANE_B32_]], 0, 0, 0 :: (dereferenceable store (s32), addrspace 4)
     ; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[V_CMP_EQ_U32_e64_]], implicit-def $exec, implicit-def $scc, implicit $exec
+    ; CHECK: G_AMDGPU_BUFFER_STORE %val(s32), %rsrc(<4 x s32>), [[COPY]](s32), %voffset, [[V_READFIRSTLANE_B32_]], 0, 0, 0 :: (dereferenceable store (s32), addrspace 4)
     ; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-    ; CHECK: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+    ; CHECK: SI_WATERFALL_LOOP %bb.1, implicit $exec
     ; CHECK: .2:
     ; CHECK: successors: %bb.3(0x80000000)
     ; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
     ; CHECK: .3:
     ; CHECK: S_ENDPGM 0
     %val:_(s32) = COPY $vgpr0
     %rsrc:_(<4 x s32>) = COPY $sgpr0_sgpr1_sgpr2_sgpr3
     %agpr:_(s32) = COPY $agpr0

 body:             |
   bb.0:
     liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7, $vgpr0
     ; CHECK-LABEL: name: load_1d_vgpr_vaddr__agpr_srsrc
     ; CHECK: successors: %bb.1(0x80000000)
     ; CHECK: liveins: $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7, $vgpr0
     ; CHECK: [[COPY:%[0-9]+]]:agpr(<8 x s32>) = COPY $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7
     ; CHECK: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
-    ; CHECK: [[DEF:%[0-9]+]]:vgpr(<4 x s32>) = G_IMPLICIT_DEF
-    ; CHECK: [[DEF1:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
+    ; CHECK: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF
     ; CHECK: [[COPY2:%[0-9]+]]:vgpr(<8 x s32>) = COPY [[COPY]](<8 x s32>)
     ; CHECK: [[UV:%[0-9]+]]:vreg_64(s64), [[UV1:%[0-9]+]]:vreg_64(s64), [[UV2:%[0-9]+]]:vreg_64(s64), [[UV3:%[0-9]+]]:vreg_64(s64) = G_UNMERGE_VALUES [[COPY2]](<8 x s32>)
     ; CHECK: [[S_MOV_B64_term:%[0-9]+]]:sreg_64_xexec = S_MOV_B64_term $exec
     ; CHECK: .1:
     ; CHECK: successors: %bb.2(0x40000000), %bb.1(0x40000000)
-    ; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF1]], %bb.0, %8, %bb.1
-    ; CHECK: [[PHI1:%[0-9]+]]:vgpr(<4 x s32>) = G_PHI [[DEF]](<4 x s32>), %bb.0, %2(<4 x s32>), %bb.1
+    ; CHECK: [[PHI:%[0-9]+]]:sreg_64_xexec = PHI [[DEF]], %bb.0, %6, %bb.1
     ; CHECK: [[V_READFIRSTLANE_B32_:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub0(s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_1:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV]].sub1(s64), implicit $exec
     ; CHECK: [[MV:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV]](s64), [[UV]](s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_2:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub0(s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_3:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV1]].sub1(s64), implicit $exec
     ; CHECK: [[MV1:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_1:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV1]](s64), [[UV1]](s64), implicit $exec
     ; CHECK: [[S_AND_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_1]], [[V_CMP_EQ_U64_e64_]], implicit-def $scc
     ; CHECK: [[V_READFIRSTLANE_B32_4:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub0(s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_5:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV2]].sub1(s64), implicit $exec
     ; CHECK: [[MV2:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_2:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV2]](s64), [[UV2]](s64), implicit $exec
     ; CHECK: [[S_AND_B64_1:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_2]], [[S_AND_B64_]], implicit-def $scc
     ; CHECK: [[V_READFIRSTLANE_B32_6:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub0(s64), implicit $exec
     ; CHECK: [[V_READFIRSTLANE_B32_7:%[0-9]+]]:sreg_32_xm0(s32) = V_READFIRSTLANE_B32 [[UV3]].sub1(s64), implicit $exec
     ; CHECK: [[MV3:%[0-9]+]]:sreg_64_xexec(s64) = G_MERGE_VALUES [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
     ; CHECK: [[V_CMP_EQ_U64_e64_3:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U64_e64 [[MV3]](s64), [[UV3]](s64), implicit $exec
     ; CHECK: [[S_AND_B64_2:%[0-9]+]]:sreg_64_xexec = S_AND_B64 [[V_CMP_EQ_U64_e64_3]], [[S_AND_B64_1]], implicit-def $scc
     ; CHECK: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<8 x s32>) = G_BUILD_VECTOR [[V_READFIRSTLANE_B32_]](s32), [[V_READFIRSTLANE_B32_1]](s32), [[V_READFIRSTLANE_B32_2]](s32), [[V_READFIRSTLANE_B32_3]](s32), [[V_READFIRSTLANE_B32_4]](s32), [[V_READFIRSTLANE_B32_5]](s32), [[V_READFIRSTLANE_B32_6]](s32), [[V_READFIRSTLANE_B32_7]](s32)
-    ; CHECK: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY1]](s32), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>))
     ; CHECK: [[S_AND_SAVEEXEC_B64_:%[0-9]+]]:sreg_64_xexec = S_AND_SAVEEXEC_B64 killed [[S_AND_B64_2]], implicit-def $exec, implicit-def $scc, implicit $exec
+    ; CHECK: [[AMDGPU_INTRIN_IMAGE_LOAD:%[0-9]+]]:vgpr(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, [[COPY1]](s32), [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>))
     ; CHECK: $exec = S_XOR_B64_term $exec, [[S_AND_SAVEEXEC_B64_]], implicit-def $scc
-    ; CHECK: S_CBRANCH_EXECNZ %bb.1, implicit $exec
+    ; CHECK: SI_WATERFALL_LOOP %bb.1, implicit $exec
     ; CHECK: .2:
     ; CHECK: successors: %bb.3(0x80000000)
     ; CHECK: $exec = S_MOV_B64_term [[S_MOV_B64_term]]
     ; CHECK: .3:
     ; CHECK: S_ENDPGM 0, implicit [[AMDGPU_INTRIN_IMAGE_LOAD]](<4 x s32>)
     %0:_(<8 x s32>) = COPY $agpr0_agpr1_agpr2_agpr3_agpr4_agpr5_agpr6_agpr7
     %1:_(s32) = COPY $vgpr0
     %2:_(<4 x s32>) = G_AMDGPU_INTRIN_IMAGE_LOAD intrinsic(@llvm.amdgcn.image.load.1d), 15, %1(s32), %0(<8 x s32>), 0, 0, 0 :: (dereferenceable load (<4 x s32>))
     S_ENDPGM 0, implicit %2
 ...

llvm/test/CodeGen/AMDGPU/indirect-call.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -amdgpu-fixed-function-abi -mtriple=amdgcn-amd-amdhsa --amdhsa-code-object-version=2 -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
+; RUN: llc -amdgpu-fixed-function-abi -mtriple=amdgcn-amd-amdhsa --amdhsa-code-object-version=2 -verify-machineinstrs -global-isel < %s | FileCheck -check-prefix=GISEL %s
 
 @gv.fptr0 = external hidden unnamed_addr addrspace(4) constant void()*, align 4
 @gv.fptr1 = external hidden unnamed_addr addrspace(4) constant void(i32)*, align 4
 
 define amdgpu_kernel void @test_indirect_call_sgpr_ptr() {
 ; GCN-LABEL: test_indirect_call_sgpr_ptr:
 ; GCN:         .amd_kernel_code_t
 ; GCN-NEXT:     amd_code_version_major = 1

 ; GCN-NEXT:    v_lshlrev_b32_e32 v2, 20, v2
 ; GCN-NEXT:    v_lshlrev_b32_e32 v1, 10, v1
 ; GCN-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GCN-NEXT:    v_or_b32_e32 v31, v0, v2
 ; GCN-NEXT:    s_mov_b32 s14, s16
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_swappc_b64 s[30:31], s[18:19]
 ; GCN-NEXT:    s_endpgm
+;
+; GISEL-LABEL: test_indirect_call_sgpr_ptr:
+; GISEL:         .amd_kernel_code_t
+; GISEL-NEXT:     amd_code_version_major = 1
+; GISEL-NEXT:     amd_code_version_minor = 2
+; GISEL-NEXT:     amd_machine_kind = 1
+; GISEL-NEXT:     amd_machine_version_major = 7
+; GISEL-NEXT:     amd_machine_version_minor = 0
+; GISEL-NEXT:     amd_machine_version_stepping = 0
+; GISEL-NEXT:     kernel_code_entry_byte_offset = 256
+; GISEL-NEXT:     kernel_code_prefetch_byte_size = 0
+; GISEL-NEXT:     granulated_workitem_vgpr_count = 7
+; GISEL-NEXT:     granulated_wavefront_sgpr_count = 4
+; GISEL-NEXT:     priority = 0
+; GISEL-NEXT:     float_mode = 240
+; GISEL-NEXT:     priv = 0
+; GISEL-NEXT:     enable_dx10_clamp = 1
+; GISEL-NEXT:     debug_mode = 0
+; GISEL-NEXT:     enable_ieee_mode = 1
+; GISEL-NEXT:     enable_wgp_mode = 0
+; GISEL-NEXT:     enable_mem_ordered = 0
+; GISEL-NEXT:     enable_fwd_progress = 0
+; GISEL-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; GISEL-NEXT:     user_sgpr_count = 14
+; GISEL-NEXT:     enable_trap_handler = 0
+; GISEL-NEXT:     enable_sgpr_workgroup_id_x = 1
+; GISEL-NEXT:     enable_sgpr_workgroup_id_y = 1
+; GISEL-NEXT:     enable_sgpr_workgroup_id_z = 1
+; GISEL-NEXT:     enable_sgpr_workgroup_info = 0
+; GISEL-NEXT:     enable_vgpr_workitem_id = 2
+; GISEL-NEXT:     enable_exception_msb = 0
+; GISEL-NEXT:     granulated_lds_size = 0
+; GISEL-NEXT:     enable_exception = 0
+; GISEL-NEXT:     enable_sgpr_private_segment_buffer = 1
+; GISEL-NEXT:     enable_sgpr_dispatch_ptr = 1
+; GISEL-NEXT:     enable_sgpr_queue_ptr = 1
+; GISEL-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; GISEL-NEXT:     enable_sgpr_dispatch_id = 1
+; GISEL-NEXT:     enable_sgpr_flat_scratch_init = 1
+; GISEL-NEXT:     enable_sgpr_private_segment_size = 0
+; GISEL-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; GISEL-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; GISEL-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; GISEL-NEXT:     enable_wavefront_size32 = 0
+; GISEL-NEXT:     enable_ordered_append_gds = 0
+; GISEL-NEXT:     private_element_size = 1
+; GISEL-NEXT:     is_ptr64 = 1
+; GISEL-NEXT:     is_dynamic_callstack = 1
+; GISEL-NEXT:     is_debug_enabled = 0
+; GISEL-NEXT:     is_xnack_enabled = 0
+; GISEL-NEXT:     workitem_private_segment_byte_size = 16384
+; GISEL-NEXT:     workgroup_group_segment_byte_size = 0
+; GISEL-NEXT:     gds_segment_byte_size = 0
+; GISEL-NEXT:     kernarg_segment_byte_size = 0
+; GISEL-NEXT:     workgroup_fbarrier_count = 0
+; GISEL-NEXT:     wavefront_sgpr_count = 37
+; GISEL-NEXT:     workitem_vgpr_count = 32
+; GISEL-NEXT:     reserved_vgpr_first = 0
+; GISEL-NEXT:     reserved_vgpr_count = 0
+; GISEL-NEXT:     reserved_sgpr_first = 0
+; GISEL-NEXT:     reserved_sgpr_count = 0
+; GISEL-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; GISEL-NEXT:     debug_private_segment_buffer_sgpr = 0
+; GISEL-NEXT:     kernarg_segment_alignment = 4
+; GISEL-NEXT:     group_segment_alignment = 4
+; GISEL-NEXT:     private_segment_alignment = 4
+; GISEL-NEXT:     wavefront_size = 6
+; GISEL-NEXT:     call_convention = -1
+; GISEL-NEXT:     runtime_loader_kernel_symbol = 0
+; GISEL-NEXT:    .end_amd_kernel_code_t
+; GISEL-NEXT:  ; %bb.0:
+; GISEL-NEXT:    s_mov_b32 s32, 0
+; GISEL-NEXT:    s_mov_b32 flat_scratch_lo, s13
+; GISEL-NEXT:    s_add_i32 s12, s12, s17
+; GISEL-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
+; GISEL-NEXT:    s_add_u32 s0, s0, s17
+; GISEL-NEXT:    s_addc_u32 s1, s1, 0
+; GISEL-NEXT:    s_mov_b32 s13, s15
+; GISEL-NEXT:    s_mov_b32 s12, s14
+; GISEL-NEXT:    s_getpc_b64 s[14:15]
+; GISEL-NEXT:    s_add_u32 s14, s14, gv.fptr0@rel32@lo+4
+; GISEL-NEXT:    s_addc_u32 s15, s15, gv.fptr0@rel32@hi+12
+; GISEL-NEXT:    s_load_dwordx2 s[18:19], s[14:15], 0x0
+; GISEL-NEXT:    v_lshlrev_b32_e32 v1, 10, v1
+; GISEL-NEXT:    v_or_b32_e32 v0, v0, v1
+; GISEL-NEXT:    v_lshlrev_b32_e32 v1, 20, v2
+; GISEL-NEXT:    v_or_b32_e32 v31, v0, v1
+; GISEL-NEXT:    s_mov_b32 s14, s16
+; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[18:19]
+; GISEL-NEXT:    s_endpgm
   %fptr = load void()*, void()* addrspace(4)* @gv.fptr0
   call void %fptr()
   ret void
 }
 
 define amdgpu_kernel void @test_indirect_call_sgpr_ptr_arg() {
 ; GCN-LABEL: test_indirect_call_sgpr_ptr_arg:
 ; GCN:         .amd_kernel_code_t

 ; GCN-NEXT:    v_lshlrev_b32_e32 v1, 10, v1
 ; GCN-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GCN-NEXT:    v_or_b32_e32 v31, v0, v2
 ; GCN-NEXT:    v_mov_b32_e32 v0, 0x7b
 ; GCN-NEXT:    s_mov_b32 s14, s16
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    s_swappc_b64 s[30:31], s[18:19]
 ; GCN-NEXT:    s_endpgm
+;
+; GISEL-LABEL: test_indirect_call_sgpr_ptr_arg:
+; GISEL:         .amd_kernel_code_t
+; GISEL-NEXT:     amd_code_version_major = 1
+; GISEL-NEXT:     amd_code_version_minor = 2
+; GISEL-NEXT:     amd_machine_kind = 1
+; GISEL-NEXT:     amd_machine_version_major = 7
+; GISEL-NEXT:     amd_machine_version_minor = 0
+; GISEL-NEXT:     amd_machine_version_stepping = 0
+; GISEL-NEXT:     kernel_code_entry_byte_offset = 256
+; GISEL-NEXT:     kernel_code_prefetch_byte_size = 0
+; GISEL-NEXT:     granulated_workitem_vgpr_count = 7
+; GISEL-NEXT:     granulated_wavefront_sgpr_count = 4
+; GISEL-NEXT:     priority = 0
+; GISEL-NEXT:     float_mode = 240
+; GISEL-NEXT:     priv = 0
+; GISEL-NEXT:     enable_dx10_clamp = 1
+; GISEL-NEXT:     debug_mode = 0
+; GISEL-NEXT:     enable_ieee_mode = 1
+; GISEL-NEXT:     enable_wgp_mode = 0
+; GISEL-NEXT:     enable_mem_ordered = 0
+; GISEL-NEXT:     enable_fwd_progress = 0
+; GISEL-NEXT:     enable_sgpr_private_segment_wave_byte_offset = 1
+; GISEL-NEXT:     user_sgpr_count = 14
+; GISEL-NEXT:     enable_trap_handler = 0
+; GISEL-NEXT:     enable_sgpr_workgroup_id_x = 1
+; GISEL-NEXT:     enable_sgpr_workgroup_id_y = 1
+; GISEL-NEXT:     enable_sgpr_workgroup_id_z = 1
+; GISEL-NEXT:     enable_sgpr_workgroup_info = 0
+; GISEL-NEXT:     enable_vgpr_workitem_id = 2
+; GISEL-NEXT:     enable_exception_msb = 0
+; GISEL-NEXT:     granulated_lds_size = 0
+; GISEL-NEXT:     enable_exception = 0
+; GISEL-NEXT:     enable_sgpr_private_segment_buffer = 1
+; GISEL-NEXT:     enable_sgpr_dispatch_ptr = 1
+; GISEL-NEXT:     enable_sgpr_queue_ptr = 1
+; GISEL-NEXT:     enable_sgpr_kernarg_segment_ptr = 1
+; GISEL-NEXT:     enable_sgpr_dispatch_id = 1
+; GISEL-NEXT:     enable_sgpr_flat_scratch_init = 1
+; GISEL-NEXT:     enable_sgpr_private_segment_size = 0
+; GISEL-NEXT:     enable_sgpr_grid_workgroup_count_x = 0
+; GISEL-NEXT:     enable_sgpr_grid_workgroup_count_y = 0
+; GISEL-NEXT:     enable_sgpr_grid_workgroup_count_z = 0
+; GISEL-NEXT:     enable_wavefront_size32 = 0
+; GISEL-NEXT:     enable_ordered_append_gds = 0
+; GISEL-NEXT:     private_element_size = 1
+; GISEL-NEXT:     is_ptr64 = 1
+; GISEL-NEXT:     is_dynamic_callstack = 1
+; GISEL-NEXT:     is_debug_enabled = 0
+; GISEL-NEXT:     is_xnack_enabled = 0
+; GISEL-NEXT:     workitem_private_segment_byte_size = 16384
+; GISEL-NEXT:     workgroup_group_segment_byte_size = 0
+; GISEL-NEXT:     gds_segment_byte_size = 0
+; GISEL-NEXT:     kernarg_segment_byte_size = 0
+; GISEL-NEXT:     workgroup_fbarrier_count = 0
+; GISEL-NEXT:     wavefront_sgpr_count = 37
+; GISEL-NEXT:     workitem_vgpr_count = 32
+; GISEL-NEXT:     reserved_vgpr_first = 0
+; GISEL-NEXT:     reserved_vgpr_count = 0
+; GISEL-NEXT:     reserved_sgpr_first = 0
+; GISEL-NEXT:     reserved_sgpr_count = 0
+; GISEL-NEXT:     debug_wavefront_private_segment_offset_sgpr = 0
+; GISEL-NEXT:     debug_private_segment_buffer_sgpr = 0
+; GISEL-NEXT:     kernarg_segment_alignment = 4
+; GISEL-NEXT:     group_segment_alignment = 4
+; GISEL-NEXT:     private_segment_alignment = 4
+; GISEL-NEXT:     wavefront_size = 6
+; GISEL-NEXT:     call_convention = -1
+; GISEL-NEXT:     runtime_loader_kernel_symbol = 0
+; GISEL-NEXT:    .end_amd_kernel_code_t
+; GISEL-NEXT:  ; %bb.0:
+; GISEL-NEXT:    s_mov_b32 s32, 0
+; GISEL-NEXT:    s_mov_b32 flat_scratch_lo, s13
+; GISEL-NEXT:    s_add_i32 s12, s12, s17
+; GISEL-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
+; GISEL-NEXT:    s_add_u32 s0, s0, s17
+; GISEL-NEXT:    s_addc_u32 s1, s1, 0
+; GISEL-NEXT:    s_mov_b32 s13, s15
+; GISEL-NEXT:    s_mov_b32 s12, s14
+; GISEL-NEXT:    s_getpc_b64 s[14:15]
+; GISEL-NEXT:    s_add_u32 s14, s14, gv.fptr1@rel32@lo+4
+; GISEL-NEXT:    s_addc_u32 s15, s15, gv.fptr1@rel32@hi+12
+; GISEL-NEXT:    v_lshlrev_b32_e32 v1, 10, v1
+; GISEL-NEXT:    s_load_dwordx2 s[18:19], s[14:15], 0x0
+; GISEL-NEXT:    v_or_b32_e32 v0, v0, v1
+; GISEL-NEXT:    v_lshlrev_b32_e32 v1, 20, v2
+; GISEL-NEXT:    v_or_b32_e32 v31, v0, v1
+; GISEL-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GISEL-NEXT:    s_mov_b32 s14, s16
+; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[18:19]
+; GISEL-NEXT:    s_endpgm
   %fptr = load void(i32)*, void(i32)* addrspace(4)* @gv.fptr1
   call void %fptr(i32 123)
   ret void
 }
 
 define void @test_indirect_call_vgpr_ptr(void()* %fptr) {
 ; GCN-LABEL: test_indirect_call_vgpr_ptr:
 ; GCN:       ; %bb.0:

 ; GCN-NEXT:    v_readlane_b32 s34, v40, 0
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 17
 ; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[6:7]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[16:17], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[16:17]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 17
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s38, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s39, 5
+; GISEL-NEXT:    v_writelane_b32 v40, s40, 6
+; GISEL-NEXT:    v_writelane_b32 v40, s41, 7
+; GISEL-NEXT:    v_writelane_b32 v40, s42, 8
+; GISEL-NEXT:    v_writelane_b32 v40, s43, 9
+; GISEL-NEXT:    v_writelane_b32 v40, s44, 10
+; GISEL-NEXT:    v_writelane_b32 v40, s46, 11
+; GISEL-NEXT:    v_writelane_b32 v40, s47, 12
+; GISEL-NEXT:    v_writelane_b32 v40, s48, 13
+; GISEL-NEXT:    v_writelane_b32 v40, s49, 14
+; GISEL-NEXT:    s_mov_b32 s42, s14
+; GISEL-NEXT:    s_mov_b32 s43, s13
+; GISEL-NEXT:    s_mov_b32 s44, s12
+; GISEL-NEXT:    s_mov_b64 s[34:35], s[10:11]
+; GISEL-NEXT:    s_mov_b64 s[36:37], s[8:9]
+; GISEL-NEXT:    s_mov_b64 s[38:39], s[6:7]
+; GISEL-NEXT:    s_mov_b64 s[40:41], s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 15
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 16
+; GISEL-NEXT:    s_mov_b64 s[46:47], exec
+; GISEL-NEXT:  BB2_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s16, v0
+; GISEL-NEXT:    v_readfirstlane_b32 s17, v1
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
+; GISEL-NEXT:    s_and_saveexec_b64 s[48:49], vcc
+; GISEL-NEXT:    s_mov_b64 s[4:5], s[40:41]
+; GISEL-NEXT:    s_mov_b64 s[6:7], s[38:39]
+; GISEL-NEXT:    s_mov_b64 s[8:9], s[36:37]
+; GISEL-NEXT:    s_mov_b64 s[10:11], s[34:35]
+; GISEL-NEXT:    s_mov_b32 s12, s44
+; GISEL-NEXT:    s_mov_b32 s13, s43
+; GISEL-NEXT:    s_mov_b32 s14, s42
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[16:17]
+; GISEL-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GISEL-NEXT:    ; implicit-def: $vgpr31
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[48:49]
+; GISEL-NEXT:    s_cbranch_execnz BB2_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[46:47]
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 15
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 16
+; GISEL-NEXT:    v_readlane_b32 s49, v40, 14
+; GISEL-NEXT:    v_readlane_b32 s48, v40, 13
+; GISEL-NEXT:    v_readlane_b32 s47, v40, 12
+; GISEL-NEXT:    v_readlane_b32 s46, v40, 11
+; GISEL-NEXT:    v_readlane_b32 s44, v40, 10
+; GISEL-NEXT:    v_readlane_b32 s43, v40, 9
+; GISEL-NEXT:    v_readlane_b32 s42, v40, 8
+; GISEL-NEXT:    v_readlane_b32 s41, v40, 7
+; GISEL-NEXT:    v_readlane_b32 s40, v40, 6
+; GISEL-NEXT:    v_readlane_b32 s39, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s38, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 17
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
   call void %fptr()
   ret void
 }
 
 define void @test_indirect_call_vgpr_ptr_arg(void(i32)* %fptr) {
 ; GCN-LABEL: test_indirect_call_vgpr_ptr_arg:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)

 ; GCN-NEXT:    v_readlane_b32 s34, v40, 0
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 17
 ; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[6:7]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr_arg:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[16:17], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[16:17]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 17
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s38, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s39, 5
+; GISEL-NEXT:    v_writelane_b32 v40, s40, 6
+; GISEL-NEXT:    v_writelane_b32 v40, s41, 7
+; GISEL-NEXT:    v_writelane_b32 v40, s42, 8
+; GISEL-NEXT:    v_writelane_b32 v40, s43, 9
+; GISEL-NEXT:    v_writelane_b32 v40, s44, 10
+; GISEL-NEXT:    v_writelane_b32 v40, s46, 11
+; GISEL-NEXT:    v_writelane_b32 v40, s47, 12
+; GISEL-NEXT:    v_writelane_b32 v40, s48, 13
+; GISEL-NEXT:    v_writelane_b32 v40, s49, 14
+; GISEL-NEXT:    s_mov_b32 s42, s14
+; GISEL-NEXT:    s_mov_b32 s43, s13
+; GISEL-NEXT:    s_mov_b32 s44, s12
+; GISEL-NEXT:    s_mov_b64 s[34:35], s[10:11]
+; GISEL-NEXT:    s_mov_b64 s[36:37], s[8:9]
+; GISEL-NEXT:    s_mov_b64 s[38:39], s[6:7]
+; GISEL-NEXT:    s_mov_b64 s[40:41], s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 15
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 16
+; GISEL-NEXT:    s_mov_b64 s[46:47], exec
+; GISEL-NEXT:  BB3_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s16, v0
+; GISEL-NEXT:    v_readfirstlane_b32 s17, v1
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
+; GISEL-NEXT:    s_and_saveexec_b64 s[48:49], vcc
+; GISEL-NEXT:    v_mov_b32_e32 v0, 0x7b
+; GISEL-NEXT:    s_mov_b64 s[4:5], s[40:41]
+; GISEL-NEXT:    s_mov_b64 s[6:7], s[38:39]
+; GISEL-NEXT:    s_mov_b64 s[8:9], s[36:37]
+; GISEL-NEXT:    s_mov_b64 s[10:11], s[34:35]
+; GISEL-NEXT:    s_mov_b32 s12, s44
+; GISEL-NEXT:    s_mov_b32 s13, s43
+; GISEL-NEXT:    s_mov_b32 s14, s42
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[16:17]
+; GISEL-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GISEL-NEXT:    ; implicit-def: $vgpr31
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[48:49]
+; GISEL-NEXT:    s_cbranch_execnz BB3_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[46:47]
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 15
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 16
+; GISEL-NEXT:    v_readlane_b32 s49, v40, 14
+; GISEL-NEXT:    v_readlane_b32 s48, v40, 13
+; GISEL-NEXT:    v_readlane_b32 s47, v40, 12
+; GISEL-NEXT:    v_readlane_b32 s46, v40, 11
+; GISEL-NEXT:    v_readlane_b32 s44, v40, 10
+; GISEL-NEXT:    v_readlane_b32 s43, v40, 9
+; GISEL-NEXT:    v_readlane_b32 s42, v40, 8
+; GISEL-NEXT:    v_readlane_b32 s41, v40, 7
+; GISEL-NEXT:    v_readlane_b32 s40, v40, 6
+; GISEL-NEXT:    v_readlane_b32 s39, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s38, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 17
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
   call void %fptr(i32 123)
   ret void
 }
 
 define i32 @test_indirect_call_vgpr_ptr_ret(i32()* %fptr) {
 ; GCN-LABEL: test_indirect_call_vgpr_ptr_ret:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)

 ; GCN-NEXT:    v_readlane_b32 s34, v40, 0
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 17
 ; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[6:7]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr_ret:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[16:17], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[16:17]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 17
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s38, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s39, 5
+; GISEL-NEXT:    v_writelane_b32 v40, s40, 6
+; GISEL-NEXT:    v_writelane_b32 v40, s41, 7
+; GISEL-NEXT:    v_writelane_b32 v40, s42, 8
+; GISEL-NEXT:    v_writelane_b32 v40, s43, 9
+; GISEL-NEXT:    v_writelane_b32 v40, s44, 10
+; GISEL-NEXT:    v_writelane_b32 v40, s46, 11
+; GISEL-NEXT:    v_writelane_b32 v40, s47, 12
+; GISEL-NEXT:    v_writelane_b32 v40, s48, 13
+; GISEL-NEXT:    v_writelane_b32 v40, s49, 14
+; GISEL-NEXT:    s_mov_b32 s42, s14
+; GISEL-NEXT:    s_mov_b32 s43, s13
+; GISEL-NEXT:    s_mov_b32 s44, s12
+; GISEL-NEXT:    s_mov_b64 s[34:35], s[10:11]
+; GISEL-NEXT:    s_mov_b64 s[36:37], s[8:9]
+; GISEL-NEXT:    s_mov_b64 s[38:39], s[6:7]
+; GISEL-NEXT:    s_mov_b64 s[40:41], s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 15
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 16
+; GISEL-NEXT:    s_mov_b64 s[46:47], exec
+; GISEL-NEXT:  BB4_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s16, v0
+; GISEL-NEXT:    v_readfirstlane_b32 s17, v1
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
+; GISEL-NEXT:    s_and_saveexec_b64 s[48:49], vcc
+; GISEL-NEXT:    s_mov_b64 s[4:5], s[40:41]
+; GISEL-NEXT:    s_mov_b64 s[6:7], s[38:39]
+; GISEL-NEXT:    s_mov_b64 s[8:9], s[36:37]
+; GISEL-NEXT:    s_mov_b64 s[10:11], s[34:35]
+; GISEL-NEXT:    s_mov_b32 s12, s44
+; GISEL-NEXT:    s_mov_b32 s13, s43
+; GISEL-NEXT:    s_mov_b32 s14, s42
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[16:17]
+; GISEL-NEXT:    v_mov_b32_e32 v2, v0
+; GISEL-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GISEL-NEXT:    ; implicit-def: $vgpr31
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[48:49]
+; GISEL-NEXT:    s_cbranch_execnz BB4_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[46:47]
+; GISEL-NEXT:    v_add_i32_e32 v0, vcc, 1, v2
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 15
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 16
+; GISEL-NEXT:    v_readlane_b32 s49, v40, 14
+; GISEL-NEXT:    v_readlane_b32 s48, v40, 13
+; GISEL-NEXT:    v_readlane_b32 s47, v40, 12
+; GISEL-NEXT:    v_readlane_b32 s46, v40, 11
+; GISEL-NEXT:    v_readlane_b32 s44, v40, 10
+; GISEL-NEXT:    v_readlane_b32 s43, v40, 9
+; GISEL-NEXT:    v_readlane_b32 s42, v40, 8
+; GISEL-NEXT:    v_readlane_b32 s41, v40, 7
+; GISEL-NEXT:    v_readlane_b32 s40, v40, 6
+; GISEL-NEXT:    v_readlane_b32 s39, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s38, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 17
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
   %a = call i32 %fptr()
   %b = add i32 %a, 1
   ret i32 %b
 }
 
 define void @test_indirect_call_vgpr_ptr_in_branch(void()* %fptr, i1 %cond) {
 ; GCN-LABEL: test_indirect_call_vgpr_ptr_in_branch:
 ; GCN:       ; %bb.0: ; %bb0

 ; GCN-NEXT:    v_readlane_b32 s34, v40, 0
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 19
 ; GCN-NEXT:    s_or_saveexec_b64 s[4:5], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[4:5]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[30:31]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr_in_branch:
+; GISEL:       ; %bb.0: ; %bb0
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[16:17], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[16:17]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 19
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s38, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s39, 5
+; GISEL-NEXT:    v_writelane_b32 v40, s40, 6
+; GISEL-NEXT:    v_writelane_b32 v40, s41, 7
+; GISEL-NEXT:    v_writelane_b32 v40, s42, 8
+; GISEL-NEXT:    v_writelane_b32 v40, s43, 9
+; GISEL-NEXT:    v_writelane_b32 v40, s44, 10
+; GISEL-NEXT:    v_writelane_b32 v40, s46, 11
+; GISEL-NEXT:    v_writelane_b32 v40, s47, 12
+; GISEL-NEXT:    v_writelane_b32 v40, s48, 13
+; GISEL-NEXT:    v_writelane_b32 v40, s49, 14
+; GISEL-NEXT:    v_writelane_b32 v40, s50, 15
+; GISEL-NEXT:    v_writelane_b32 v40, s51, 16
+; GISEL-NEXT:    s_mov_b32 s42, s14
+; GISEL-NEXT:    s_mov_b32 s43, s13
+; GISEL-NEXT:    s_mov_b32 s44, s12
+; GISEL-NEXT:    s_mov_b64 s[34:35], s[10:11]
+; GISEL-NEXT:    s_mov_b64 s[36:37], s[8:9]
+; GISEL-NEXT:    s_mov_b64 s[38:39], s[6:7]
+; GISEL-NEXT:    s_mov_b64 s[40:41], s[4:5]
+; GISEL-NEXT:    v_and_b32_e32 v2, 1, v2
+; GISEL-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v2
+; GISEL-NEXT:    s_and_saveexec_b64 s[46:47], vcc
+; GISEL-NEXT:    s_cbranch_execz BB5_4
+; GISEL-NEXT:  ; %bb.1: ; %bb1
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 17
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 18
+; GISEL-NEXT:    s_mov_b64 s[48:49], exec
+; GISEL-NEXT:  BB5_2: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s16, v0
+; GISEL-NEXT:    v_readfirstlane_b32 s17, v1
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[16:17], v[0:1]
+; GISEL-NEXT:    s_and_saveexec_b64 s[50:51], vcc
+; GISEL-NEXT:    s_mov_b64 s[4:5], s[40:41]
+; GISEL-NEXT:    s_mov_b64 s[6:7], s[38:39]
+; GISEL-NEXT:    s_mov_b64 s[8:9], s[36:37]
+; GISEL-NEXT:    s_mov_b64 s[10:11], s[34:35]
+; GISEL-NEXT:    s_mov_b32 s12, s44
+; GISEL-NEXT:    s_mov_b32 s13, s43
+; GISEL-NEXT:    s_mov_b32 s14, s42
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[16:17]
+; GISEL-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GISEL-NEXT:    ; implicit-def: $vgpr31
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[50:51]
+; GISEL-NEXT:    s_cbranch_execnz BB5_2
+; GISEL-NEXT:  ; %bb.3:
+; GISEL-NEXT:    s_mov_b64 exec, s[48:49]
+; GISEL-NEXT:    v_readlane_b32 s30, v40, 17
+; GISEL-NEXT:    v_readlane_b32 s31, v40, 18
+; GISEL-NEXT:  BB5_4: ; %bb2
+; GISEL-NEXT:    s_or_b64 exec, exec, s[46:47]
+; GISEL-NEXT:    v_readlane_b32 s51, v40, 16
+; GISEL-NEXT:    v_readlane_b32 s50, v40, 15
+; GISEL-NEXT:    v_readlane_b32 s49, v40, 14
+; GISEL-NEXT:    v_readlane_b32 s48, v40, 13
+; GISEL-NEXT:    v_readlane_b32 s47, v40, 12
+; GISEL-NEXT:    v_readlane_b32 s46, v40, 11
+; GISEL-NEXT:    v_readlane_b32 s44, v40, 10
+; GISEL-NEXT:    v_readlane_b32 s43, v40, 9
+; GISEL-NEXT:    v_readlane_b32 s42, v40, 8
+; GISEL-NEXT:    v_readlane_b32 s41, v40, 7
+; GISEL-NEXT:    v_readlane_b32 s40, v40, 6
+; GISEL-NEXT:    v_readlane_b32 s39, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s38, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 19
+; GISEL-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[4:5]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[30:31]
 bb0:
   br i1 %cond, label %bb1, label %bb2
 
 bb1:
   call void %fptr()
   br label %bb2
 
 bb2:

 ; GCN-NEXT:    v_readlane_b32 s34, v40, 0
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 6
 ; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[6:7]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr_inreg_arg:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 6
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 5
+; GISEL-NEXT:    s_mov_b64 s[34:35], exec
+; GISEL-NEXT:  BB6_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s6, v0
+; GISEL-NEXT:    v_readfirstlane_b32 s7, v1
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[6:7], v[0:1]
+; GISEL-NEXT:    s_and_saveexec_b64 s[36:37], vcc
+; GISEL-NEXT:    s_movk_i32 s4, 0x7b
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[6:7]
+; GISEL-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[36:37]
+; GISEL-NEXT:    s_cbranch_execnz BB6_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[34:35]
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 6
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
   call amdgpu_gfx void %fptr(i32 inreg 123)
   ret void
 }
 
 define i32 @test_indirect_call_vgpr_ptr_arg_and_reuse(i32 %i, void(i32)* %fptr) {
 ; GCN-LABEL: test_indirect_call_vgpr_ptr_arg_and_reuse:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)

 ; GCN-NEXT:    buffer_load_dword v41, off, s[0:3], s33 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 6
 ; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[6:7]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr_arg_and_reuse:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 offset:4 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 6
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    buffer_store_dword v41, off, s[0:3], s33 ; 4-byte Folded Spill
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_mov_b32_e32 v41, v0
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 5
+; GISEL-NEXT:    s_mov_b64 s[34:35], exec
+; GISEL-NEXT:  BB7_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s4, v1
+; GISEL-NEXT:    v_readfirstlane_b32 s5, v2
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[4:5], v[1:2]
+; GISEL-NEXT:    s_and_saveexec_b64 s[36:37], vcc
+; GISEL-NEXT:    v_mov_b32_e32 v0, v41
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GISEL-NEXT:    ; implicit-def: $vgpr1_vgpr2
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[36:37]
+; GISEL-NEXT:    s_cbranch_execnz BB7_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[34:35]
+; GISEL-NEXT:    v_mov_b32_e32 v0, v41
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    buffer_load_dword v41, off, s[0:3], s33 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 6
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 offset:4 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
   call amdgpu_gfx void %fptr(i32 %i)
   ret i32 %i
 }
 
 ; Use a variable inside a waterfall loop and use the return variable after the loop.
 ; TODO The argument and return variable could be in the same physical register, but the register
 ; allocator is not able to do that because the return value clashes with the liverange of an
 ; IMPLICIT_DEF of the argument.

 ; GCN-NEXT:    v_readlane_b32 s34, v40, 0
 ; GCN-NEXT:    s_addk_i32 s32, 0xfc00
 ; GCN-NEXT:    v_readlane_b32 s33, v40, 6
 ; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
 ; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GCN-NEXT:    s_mov_b64 exec, s[6:7]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_call_vgpr_ptr_arg_and_return:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 6
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 5
+; GISEL-NEXT:    s_mov_b64 s[34:35], exec
+; GISEL-NEXT:  BB8_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s4, v1
+; GISEL-NEXT:    v_readfirstlane_b32 s5, v2
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[4:5], v[1:2]
+; GISEL-NEXT:    s_and_saveexec_b64 s[36:37], vcc
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GISEL-NEXT:    v_mov_b32_e32 v3, v0
+; GISEL-NEXT:    ; implicit-def: $vgpr1_vgpr2
+; GISEL-NEXT:    ; implicit-def: $vgpr0
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[36:37]
+; GISEL-NEXT:    s_cbranch_execnz BB8_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[34:35]
+; GISEL-NEXT:    v_mov_b32_e32 v0, v3
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 6
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
   %ret = call amdgpu_gfx i32 %fptr(i32 %i)
   ret i32 %ret
 }
+
+; Calling a vgpr can never be a tail call.
+define void @test_indirect_tail_call_vgpr_ptr(void()* %fptr) {
+; GCN-LABEL: test_indirect_tail_call_vgpr_ptr:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GCN-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GCN-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GCN-NEXT:    s_mov_b64 exec, s[4:5]
+; GCN-NEXT:    v_writelane_b32 v40, s33, 6
+; GCN-NEXT:    s_mov_b32 s33, s32
+; GCN-NEXT:    s_addk_i32 s32, 0x400
+; GCN-NEXT:    v_writelane_b32 v40, s34, 0
+; GCN-NEXT:    v_writelane_b32 v40, s35, 1
+; GCN-NEXT:    v_writelane_b32 v40, s36, 2
+; GCN-NEXT:    v_writelane_b32 v40, s37, 3
+; GCN-NEXT:    v_writelane_b32 v40, s30, 4
+; GCN-NEXT:    v_writelane_b32 v40, s31, 5
+; GCN-NEXT:    s_mov_b64 s[34:35], exec
+; GCN-NEXT:  BB9_1: ; =>This Inner Loop Header: Depth=1
+; GCN-NEXT:    v_readfirstlane_b32 s4, v0
+; GCN-NEXT:    v_readfirstlane_b32 s5, v1
+; GCN-NEXT:    v_cmp_eq_u64_e32 vcc, s[4:5], v[0:1]
+; GCN-NEXT:    s_and_saveexec_b64 s[36:37], vcc
+; GCN-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GCN-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GCN-NEXT:    s_xor_b64 exec, exec, s[36:37]
+; GCN-NEXT:    s_cbranch_execnz BB9_1
+; GCN-NEXT:  ; %bb.2:
+; GCN-NEXT:    s_mov_b64 exec, s[34:35]
+; GCN-NEXT:    v_readlane_b32 s4, v40, 4
+; GCN-NEXT:    v_readlane_b32 s5, v40, 5
+; GCN-NEXT:    v_readlane_b32 s37, v40, 3
+; GCN-NEXT:    v_readlane_b32 s36, v40, 2
+; GCN-NEXT:    v_readlane_b32 s35, v40, 1
+; GCN-NEXT:    v_readlane_b32 s34, v40, 0
+; GCN-NEXT:    s_addk_i32 s32, 0xfc00
+; GCN-NEXT:    v_readlane_b32 s33, v40, 6
+; GCN-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GCN-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GCN-NEXT:    s_mov_b64 exec, s[6:7]
+; GCN-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NEXT:    s_setpc_b64 s[4:5]
+;
+; GISEL-LABEL: test_indirect_tail_call_vgpr_ptr:
+; GISEL:       ; %bb.0:
+; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GISEL-NEXT:    s_or_saveexec_b64 s[4:5], -1
+; GISEL-NEXT:    buffer_store_dword v40, off, s[0:3], s32 ; 4-byte Folded Spill
+; GISEL-NEXT:    s_mov_b64 exec, s[4:5]
+; GISEL-NEXT:    v_writelane_b32 v40, s33, 6
+; GISEL-NEXT:    s_mov_b32 s33, s32
+; GISEL-NEXT:    s_addk_i32 s32, 0x400
+; GISEL-NEXT:    v_writelane_b32 v40, s34, 0
+; GISEL-NEXT:    v_writelane_b32 v40, s35, 1
+; GISEL-NEXT:    v_writelane_b32 v40, s36, 2
+; GISEL-NEXT:    v_writelane_b32 v40, s37, 3
+; GISEL-NEXT:    v_writelane_b32 v40, s30, 4
+; GISEL-NEXT:    v_writelane_b32 v40, s31, 5
+; GISEL-NEXT:    s_mov_b64 s[34:35], exec
+; GISEL-NEXT:  BB9_1: ; =>This Inner Loop Header: Depth=1
+; GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GISEL-NEXT:    v_readfirstlane_b32 s5, v1
+; GISEL-NEXT:    v_cmp_eq_u64_e32 vcc, s[4:5], v[0:1]
+; GISEL-NEXT:    s_and_saveexec_b64 s[36:37], vcc
+; GISEL-NEXT:    s_swappc_b64 s[30:31], s[4:5]
+; GISEL-NEXT:    ; implicit-def: $vgpr0_vgpr1
+; GISEL-NEXT:    s_xor_b64 exec, exec, s[36:37]
+; GISEL-NEXT:    s_cbranch_execnz BB9_1
+; GISEL-NEXT:  ; %bb.2:
+; GISEL-NEXT:    s_mov_b64 exec, s[34:35]
+; GISEL-NEXT:    v_readlane_b32 s4, v40, 4
+; GISEL-NEXT:    v_readlane_b32 s5, v40, 5
+; GISEL-NEXT:    v_readlane_b32 s37, v40, 3
+; GISEL-NEXT:    v_readlane_b32 s36, v40, 2
+; GISEL-NEXT:    v_readlane_b32 s35, v40, 1
+; GISEL-NEXT:    v_readlane_b32 s34, v40, 0
+; GISEL-NEXT:    s_addk_i32 s32, 0xfc00
+; GISEL-NEXT:    v_readlane_b32 s33, v40, 6
+; GISEL-NEXT:    s_or_saveexec_b64 s[6:7], -1
+; GISEL-NEXT:    buffer_load_dword v40, off, s[0:3], s32 ; 4-byte Folded Reload
+; GISEL-NEXT:    s_mov_b64 exec, s[6:7]
+; GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GISEL-NEXT:    s_setpc_b64 s[4:5]
+  tail call amdgpu_gfx void %fptr()
+  ret void
+}