[AMDGPU] Enable whole wave register copy

[cdevadas]

Summary

So far, we haven't exposed the allocation of whole-wave
registers to regalloc. We hand-picked them for various
whole wave mode operations. With a future patch, we
want the allocator to efficiently allocate them rather
than using the custom pre-allocation pass.

Any liverange split of virtual registers involved in
whole-wave operations require the resulting COPY
introduced with the split to be performed for all
lanes. It isn't implemented in the compiler yet.

This patch would identify all such copies and
manipulate the exec mask around them to enable all
lanes without affecting the value of exec mask
elsewhere.

[yassingh]

• Rebase
• Incorporated the downstream code

[yassingh]

fix file comment

[yassingh]

Rebase and merge D150390 ([AMDGPU] Introduce and use the new PRED_COPY opcode) into this.

[yassingh]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

1939–1941

I wanted to keep this in SILowerPredicatedCopy pass but it wasn't working there, it needed virtregRewriter to be run before it.

[yassingh]

• Revert back to PRED_COPY to COPY lowering mechanism
• Rebase

[yassingh]

• Review comments (use getWaveMaskRegClass(), getExec() functions)

[yassingh]

In D143762#4438232, @yassingh wrote:

• Review comments (use getWaveMaskRegClass(), getExec() functions)

Accidentally updated this one instead of D143759. Not reverting as the extra diff will go away once it is rebased.

[arsenm]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

8007–8010

Don't need to query TRI, it's already the RI member. You can also just do RI.isSGPRReg.

Also, you're not checking the flags? The point was to check the flags and switch to the WWM copy. You don't care about SGPR or VGPR, you care about WWM VGPR or not

[yassingh]

• Check vreg flag in getLiveRangeSplitOpcode

[arsenm]

llvm/lib/CodeGen/SplitKit.cpp

567 ↗

(On Diff #533516)

You're missing the new opcode here. buildSingleSubRegCopy should gain a new MCInstrDesc & parameter for the opcode to use

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

2419

Regular copy shouldn't reach here?

8007

SrcReg.isVirtual should be implied

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

140 ↗

(On Diff #533516)

Can you early return false if the WWM reg set is empty?

143 ↗

(On Diff #533516)

continue and reduce indentation

144 ↗

(On Diff #533516)

Should add a TODO to try to reduce the saveexec/restore exec pairs for adjacent WWM ops

159 ↗

(On Diff #533516)

Do these need to gain an implicit exec use?

[yassingh]

• review comments

[yassingh]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

2419

Yes it won't, will remove.

8007

Added an assert.

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

140 ↗

(On Diff #533516)

Done, had to add a new helper function.

159 ↗

(On Diff #533516)

Do you mean we should add the implicit exec? In that case, SIFixVGPRCopies will take care?

[arsenm]

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

159 ↗

(On Diff #533516)

This isn't something that can be taken care of later. SIFixVGPRCopies is a horribly broken hack, the less we depend on it the better

159 ↗

(On Diff #533516)

Actually these split copies might have been the original problem which caused it to be added. Maybe we have a way to drop it now?

[arsenm]

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

159 ↗

(On Diff #533516)

That was the reasoning given in D28874 when it was added. How about as a next step we make sure all the VGPR splits end up with exec reads

[yassingh]

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

159 ↗

(On Diff #533516)

Will look into it, don't have sufficient expertise to answer this right now. Adding @AMDGPU.

159 ↗

(On Diff #533516)

I can try and see what comes. I can see 2 ways to go about it, add the exec operand here itself or, in tablegen definition of PRED_COPY. What do you suggest?

[arsenm]

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

159 ↗

(On Diff #533516)

Add to the instruction definition. Though we're still stuck with COPY=PRED_COPY 'sometimes'

[yassingh]

Rebase

[arsenm]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

8009

I guess the name is wrong now. It's technically an unpredicated copy. Rename to WWM_COPY or COPY_WWM?

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

82 ↗

(On Diff #534588)

SIInstrInfo

92 ↗

(On Diff #534588)

Either this check is unnecessary or should have happened earlier to avoid crashing on the checkFlag

[cdevadas]

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

8009

Make sense. I prefer WWM_COPY.

llvm/lib/Target/AMDGPU/SILowerPredicatedCopies.cpp

82 ↗

(On Diff #534588)

This function is no longer needed. Now the PRED_COPY is inserted only for wwm-regs after we fine-tuned AMDGPU's getLiveRangeSplitOpcode implementation.

151 ↗

(On Diff #534588)

I guess this whole check now can be entirely avoided. PRED_COPY is inserted only for wwm-regs. If we see any such copy, insert the EXEC mask manipulation unconditionally.

[yassingh]

• Rename PRED_COPY to WWM_COPY
• Review comments delete redundant isWWMCopy() check

[cdevadas]

Remove the test llvm/test/CodeGen/AMDGPU/skip-subreg-copy-from-iswwmcopy-check.mir. It is no longer needed as you have dropped isWWMCopy function.

llvm/include/llvm/CodeGen/TargetInstrInfo.h

1974 ↗

(On Diff #534849)

Did you rebase this patch after the prototype change in D150388?
This diff shouldn't be here.

llvm/lib/CodeGen/SplitKit.cpp

539 ↗

(On Diff #534849)

Ditto

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

2419

Still see the regular COPY here.

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

94

Remove the VRM check here. We no longer include this pass in the O0 pipeline.

126

I think this assertion is redundant. We are inserting WWM_COPY only during the live range split.

[yassingh]

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

94

This pass is being invoked in O0 too currently(added while exploring an independent lowering mechanism for PRED_COPY). Will remove that too.

[arsenm]

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

130

avoid DebugLoc copy, use const ref

[arsenm]

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

145

In a follow up could release the reserved register

[yassingh]

Review commnets and remove pass from O0.

[yassingh]

Reupload

[yassingh]

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

145

I don't understand, since sgpr allocation is already done do we have to free the reserved sgprs? Also, can you refer me to some other reserved registers to see how do I release it.

[yassingh]

rebase

[yassingh]

re-upload

[arsenm]

LGTM with nit

llvm/lib/Target/AMDGPU/SIInstructions.td

3311

This should not be AMDGPUGenericInstruction, it's not a generic instruction. Should subclass from AMDGPUInst, and move to the other non-generic pseudos. AMDGPUGenericInstruction is for the G_AMDGPU_ globalisel pre-isel pseudos

[arsenm]

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

121

Typos Club adjancent

136

Missing newline

llvm/test/CodeGen/AMDGPU/llc-pipeline.ll

364

I think I lost track of what was going on with -O0, did this lose the -O0 run?

[arsenm]

Also not sure how this has no test changes

[arsenm]

llvm/lib/Target/AMDGPU/SIInstructions.td

3318

Should be marked convergent

[yassingh]

Review comments

[yassingh]

In D143762#4456567, @arsenm wrote:

Also not sure how this has no test changes

This patch was originally submitted with 0 tests. My understating is we won't see any effects of these changes until WWM copies are inserted in code by the next patch(Spill SGPRs to virtual VGPRs).
cc @cdevadas

llvm/test/CodeGen/AMDGPU/llc-pipeline.ll

364

It was introduced when we moved from PRED_COPY "simplification" to proper lowering mechanism. Was removed again after the approach changed back to lowering WWM_COPY to COPY first. I have added the O0 invocation back but it's not doing anything.

[cdevadas]

In D143762#4471029, @yassingh wrote:

In D143762#4456567, @arsenm wrote:

Also not sure how this has no test changes

This patch was originally submitted with 0 tests. My understating is we won't see any effects of these changes until WWM copies are inserted in code by the next patch(Spill SGPRs to virtual VGPRs).
cc @cdevadas

Yes. This is only a pre-patch for D124196 where the actual test was included, llvm/test/CodeGen/AMDGPU/whole-wave-register-copy.ll

[arsenm]

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

133

Back to back debug printing prints same instruction, just remove the second one?

[yassingh]

Removed extra debug statement

[cdevadas]

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

1356

I'm still not convinced why this is needed in the -O0 flow?
By now, the VGPR allocation is done in the -O0 flow, and we no longer have any virtual registers. This pass act on virtual registers to see if wwm copies needed exec manipulation.

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

91

If this pass is enabled for for -O0 as well, put back the !VRM check to ensure an early return from here for -O0 flow.
And may be a comment too?

[yassingh]

Update instruction class of WWM_COPY

[arsenm]

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

1356

It's conceptually needed and it's an implementation detail of current regalloc fast that these aren't introduced. Plus I think in general we should have other WWM copies for general WWM support in the future

[cdevadas]

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

1356

ok

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

112

Also, do an early return if MRI.getNumVirtRegs() is zero. This would avoid iterating the whole function when there are no virtual registers at all (would take care of the -O0 path when physRegs are already assigned).

[yassingh]

Rebase before merge

[yassingh]

Patch relanded with 4d42e8b5d1fa87e49768d100dd1bc53515391e89

llvm/lib/Target/AMDGPU/AMDGPU.h

 FunctionPass *createSILowerI1CopiesPass();
 FunctionPass *createSIShrinkInstructionsPass();
 FunctionPass *createSILoadStoreOptimizerPass();
 FunctionPass *createSIWholeQuadModePass();
 FunctionPass *createSIFixControlFlowLiveIntervalsPass();
 FunctionPass *createSIOptimizeExecMaskingPreRAPass();
 FunctionPass *createSIOptimizeVGPRLiveRangePass();
 FunctionPass *createSIFixSGPRCopiesPass();
+FunctionPass *createLowerWWMCopiesPass();
 FunctionPass *createSIMemoryLegalizerPass();
 FunctionPass *createSIInsertWaitcntsPass();
 FunctionPass *createSIPreAllocateWWMRegsPass();
 FunctionPass *createSIFormMemoryClausesPass();
 
 FunctionPass *createSIPostRABundlerPass();
 FunctionPass *createAMDGPUSimplifyLibCallsPass(const TargetMachine *);
 FunctionPass *createAMDGPUUseNativeCallsPass();

 extern char &SIShrinkInstructionsID;
 
 void initializeSIFixSGPRCopiesPass(PassRegistry &);
 extern char &SIFixSGPRCopiesID;
 
 void initializeSIFixVGPRCopiesPass(PassRegistry &);
 extern char &SIFixVGPRCopiesID;
 
+void initializeSILowerWWMCopiesPass(PassRegistry &);
+extern char &SILowerWWMCopiesID;
+
 void initializeSILowerI1CopiesPass(PassRegistry &);
 extern char &SILowerI1CopiesID;
 
 void initializeSILowerSGPRSpillsPass(PassRegistry &);
 extern char &SILowerSGPRSpillsID;
 
 void initializeSILoadStoreOptimizerPass(PassRegistry &);
 extern char &SILoadStoreOptimizerID;

llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp

extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   initializeR600ControlFlowFinalizerPass(*PR);
   initializeR600PacketizerPass(*PR);
   initializeR600ExpandSpecialInstrsPassPass(*PR);
   initializeR600VectorRegMergerPass(*PR);
   initializeGlobalISel(*PR);
   initializeAMDGPUDAGToDAGISelPass(*PR);
   initializeGCNDPPCombinePass(*PR);
   initializeSILowerI1CopiesPass(*PR);
+  initializeSILowerWWMCopiesPass(*PR);
   initializeSILowerSGPRSpillsPass(*PR);
   initializeSIFixSGPRCopiesPass(*PR);
   initializeSIFixVGPRCopiesPass(*PR);
   initializeSIFoldOperandsPass(*PR);
   initializeSIPeepholeSDWAPass(*PR);
   initializeSIShrinkInstructionsPass(*PR);
   initializeSIOptimizeExecMaskingPreRAPass(*PR);
   initializeSIOptimizeVGPRLiveRangePass(*PR);
void GCNPassConfig::addOptimizedRegAlloc() {
 
   if (EnableDCEInRA)
     insertPass(&DetectDeadLanesID, &DeadMachineInstructionElimID);
 
   TargetPassConfig::addOptimizedRegAlloc();
 }
 
 bool GCNPassConfig::addPreRewrite() {
+  addPass(&SILowerWWMCopiesID);
   if (EnableRegReassign)
     addPass(&GCNNSAReassignID);
   return true;
 }
 
 FunctionPass *GCNPassConfig::createSGPRAllocPass(bool Optimized) {
   // Initialize the global default.
   llvm::call_once(InitializeDefaultSGPRRegisterAllocatorFlag,
bool GCNPassConfig::addRegAssignAndRewriteFast() {
   addPass(&GCNPreRALongBranchRegID);
 
   addPass(createSGPRAllocPass(false));
 
   // Equivalent of PEI for SGPRs.
   addPass(&SILowerSGPRSpillsID);
 
   addPass(createVGPRAllocPass(false));
+
+  addPass(&SILowerWWMCopiesID);
   return true;
 }
 
 bool GCNPassConfig::addRegAssignAndRewriteOptimized() {
   if (!usingDefaultRegAlloc())
     report_fatal_error(RegAllocOptNotSupportedMessage);
 
   addPass(&GCNPreRALongBranchRegID);

llvm/lib/Target/AMDGPU/CMakeLists.txt

add_llvm_target(AMDGPUCodeGen
   SIInsertHardClauses.cpp
   SIInsertWaitcnts.cpp
   SIInstrInfo.cpp
   SIISelLowering.cpp
   SILateBranchLowering.cpp
   SILoadStoreOptimizer.cpp
   SILowerControlFlow.cpp
   SILowerI1Copies.cpp
+  SILowerWWMCopies.cpp
   SILowerSGPRSpills.cpp
   SIMachineFunctionInfo.cpp
   SIMachineScheduler.cpp
   SIMemoryLegalizer.cpp
   SIModeRegister.cpp
   SIModeRegisterDefaults.cpp
   SIOptimizeExecMasking.cpp
   SIOptimizeExecMaskingPreRA.cpp

llvm/lib/Target/AMDGPU/SIInstrInfo.h

private:
   getDestEquivalentVGPRClass(const MachineInstr &Inst) const;
 
   bool checkInstOffsetsDoNotOverlap(const MachineInstr &MIa,
                                     const MachineInstr &MIb) const;
 
   Register findUsedSGPR(const MachineInstr &MI, int OpIndices[3]) const;
 
 protected:
+  /// If the specific machine instruction is a instruction that moves/copies
+  /// value from one register to another register return destination and source
+  /// registers as machine operands.
+  std::optional<DestSourcePair>
+  isCopyInstrImpl(const MachineInstr &MI) const override;
+
   bool swapSourceModifiers(MachineInstr &MI,
                            MachineOperand &Src0, unsigned Src0OpName,
                            MachineOperand &Src1, unsigned Src1OpName) const;
 
   MachineInstr *commuteInstructionImpl(MachineInstr &MI, bool NewMI,
                                        unsigned OpIdx0,
                                        unsigned OpIdx1) const override;
 
static bool doesNotReadTiedSource(const MachineInstr &MI) {
     return MI.getDesc().TSFlags & SIInstrFlags::TiedSourceNotRead;
   }
 
   bool doesNotReadTiedSource(uint16_t Opcode) const {
     return get(Opcode).TSFlags & SIInstrFlags::TiedSourceNotRead;
   }
 
   bool isVGPRCopy(const MachineInstr &MI) const {
-    assert(MI.isCopy());
+    assert(isCopyInstr(MI));
     Register Dest = MI.getOperand(0).getReg();
     const MachineFunction &MF = *MI.getParent()->getParent();
     const MachineRegisterInfo &MRI = MF.getRegInfo();
     return !RI.isSGPRReg(MRI, Dest);
   }
 
   bool hasVGPRUses(const MachineInstr &MI) const {
     const MachineFunction &MF = *MI.getParent()->getParent();
bool isInlineConstant(const MachineInstr &MI, unsigned OpIdx) const {
     return isInlineConstant(MO, MI.getDesc().operands()[OpIdx].OperandType);
   }
 
   bool isInlineConstant(const MachineInstr &MI, unsigned OpIdx,
                         const MachineOperand &MO) const {
     if (OpIdx >= MI.getDesc().NumOperands)
       return false;
 
-    if (MI.isCopy()) {
+    if (isCopyInstr(MI)) {
       unsigned Size = getOpSize(MI, OpIdx);
       assert(Size == 8 || Size == 4);
 
       uint8_t OpType = (Size == 8) ?
         AMDGPU::OPERAND_REG_IMM_INT64 : AMDGPU::OPERAND_REG_IMM_INT32;
       return isInlineConstant(MO, OpType);
     }
 
public:
 
   bool verifyInstruction(const MachineInstr &MI,
                          StringRef &ErrInfo) const override;
 
   unsigned getVALUOp(const MachineInstr &MI) const;
 
   void insertScratchExecCopy(MachineFunction &MF, MachineBasicBlock &MBB,
                              MachineBasicBlock::iterator MBBI,
-                             const DebugLoc &DL, Register Reg,
-                             bool IsSCCLive) const;
+                             const DebugLoc &DL, Register Reg, bool IsSCCLive,
+                             SlotIndexes *Indexes = nullptr) const;
 
   void restoreExec(MachineFunction &MF, MachineBasicBlock &MBB,
                    MachineBasicBlock::iterator MBBI, const DebugLoc &DL,
-                   Register Reg) const;
+                   Register Reg, SlotIndexes *Indexes = nullptr) const;
 
   /// Return the correct register class for \p OpNo.  For target-specific
   /// instructions, this will return the register class that has been defined
   /// in tablegen.  For generic instructions, like REG_SEQUENCE it will return
   /// the register class of its machine operand.
   /// to infer the correct register class base on the other operands.
   const TargetRegisterClass *getOpRegClass(const MachineInstr &MI,
                                            unsigned OpNo) const;
public:
 
   ScheduleHazardRecognizer *
   CreateTargetPostRAHazardRecognizer(const MachineFunction &MF) const override;
 
   ScheduleHazardRecognizer *
   CreateTargetMIHazardRecognizer(const InstrItineraryData *II,
                                  const ScheduleDAGMI *DAG) const override;
 
+  unsigned getLiveRangeSplitOpcode(Register Reg,
+                                   const MachineFunction &MF) const override;
+
   bool isBasicBlockPrologue(const MachineInstr &MI) const override;
 
   MachineInstr *createPHIDestinationCopy(MachineBasicBlock &MBB,
                                          MachineBasicBlock::iterator InsPt,
                                          const DebugLoc &DL, Register Src,
                                          Register Dst) const override;
 
   MachineInstr *createPHISourceCopy(MachineBasicBlock &MBB,

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

 }
 
 bool SIInstrInfo::expandPostRAPseudo(MachineInstr &MI) const {
   const SIRegisterInfo *TRI = ST.getRegisterInfo();
   MachineBasicBlock &MBB = *MI.getParent();
   DebugLoc DL = MBB.findDebugLoc(MI);
   switch (MI.getOpcode()) {
   default: return TargetInstrInfo::expandPostRAPseudo(MI);
   case AMDGPU::S_MOV_B64_term:
     // This is only a terminator to get the correct spill code placement during
     // register allocation.
     MI.setDesc(get(AMDGPU::S_MOV_B64));
     break;
 
   case AMDGPU::S_MOV_B32_term:
     // This is only a terminator to get the correct spill code placement during
     // register allocation.
     MI.setDesc(get(AMDGPU::S_MOV_B32));
     break;
BuildMI(MBB, MI, DL, get(AMDGPU::REG_SEQUENCE), Dst)
       .addImm(AMDGPU::sub0)
       .addReg(Split[1]->getOperand(0).getReg())
       .addImm(AMDGPU::sub1);
 
   MI.eraseFromParent();
   return std::pair(Split[0], Split[1]);
 }
 
+std::optional<DestSourcePair>
+SIInstrInfo::isCopyInstrImpl(const MachineInstr &MI) const {
+  if (MI.getOpcode() == AMDGPU::WWM_COPY)
+    return DestSourcePair{MI.getOperand(0), MI.getOperand(1)};
+
+  return std::nullopt;
+}
+
 bool SIInstrInfo::swapSourceModifiers(MachineInstr &MI,
                                       MachineOperand &Src0,
                                       unsigned Src0OpName,
                                       MachineOperand &Src1,
                                       unsigned Src1OpName) const {
   MachineOperand *Src0Mods = getNamedOperand(MI, Src0OpName);
   if (!Src0Mods)
     return false;
bool SIInstrInfo::isFoldableCopy(const MachineInstr &MI) {
   case AMDGPU::V_MOV_B32_e32:
   case AMDGPU::V_MOV_B32_e64:
   case AMDGPU::V_MOV_B64_PSEUDO:
   case AMDGPU::V_MOV_B64_e32:
   case AMDGPU::V_MOV_B64_e64:
   case AMDGPU::S_MOV_B32:
   case AMDGPU::S_MOV_B64:
   case AMDGPU::COPY:
+  case AMDGPU::WWM_COPY:
   case AMDGPU::V_ACCVGPR_WRITE_B32_e64:
   case AMDGPU::V_ACCVGPR_READ_B32_e64:
   case AMDGPU::V_ACCVGPR_MOV_B32:
     return true;
   default:
     return false;
   }
 }
unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) const {
   llvm_unreachable(
       "Unexpected scalar opcode without corresponding vector one!");
 }
 
 void SIInstrInfo::insertScratchExecCopy(MachineFunction &MF,
                                         MachineBasicBlock &MBB,
                                         MachineBasicBlock::iterator MBBI,
                                         const DebugLoc &DL, Register Reg,
-                                        bool IsSCCLive) const {
+                                        bool IsSCCLive,
+                                        SlotIndexes *Indexes) const {
   const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
   const SIInstrInfo *TII = ST.getInstrInfo();
   bool IsWave32 = ST.isWave32();
   if (IsSCCLive) {
     // Insert two move instructions, one to save the original value of EXEC and
     // the other to turn on all bits in EXEC. This is required as we can't use
     // the single instruction S_OR_SAVEEXEC that clobbers SCC.
     unsigned MovOpc = IsWave32 ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
     MCRegister Exec = IsWave32 ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
-    BuildMI(MBB, MBBI, DL, TII->get(MovOpc), Reg).addReg(Exec, RegState::Kill);
-    BuildMI(MBB, MBBI, DL, TII->get(MovOpc), Exec).addImm(-1);
+    auto StoreExecMI = BuildMI(MBB, MBBI, DL, TII->get(MovOpc), Reg)
+                           .addReg(Exec, RegState::Kill);
+    auto FlipExecMI = BuildMI(MBB, MBBI, DL, TII->get(MovOpc), Exec).addImm(-1);
+    if (Indexes) {
+      Indexes->insertMachineInstrInMaps(*StoreExecMI);
+      Indexes->insertMachineInstrInMaps(*FlipExecMI);
+    }
   } else {
     const unsigned OrSaveExec =
         IsWave32 ? AMDGPU::S_OR_SAVEEXEC_B32 : AMDGPU::S_OR_SAVEEXEC_B64;
     auto SaveExec =
         BuildMI(MBB, MBBI, DL, TII->get(OrSaveExec), Reg).addImm(-1);
     SaveExec->getOperand(3).setIsDead(); // Mark SCC as dead.
+    if (Indexes)
+      Indexes->insertMachineInstrInMaps(*SaveExec);
   }
 }
 
 void SIInstrInfo::restoreExec(MachineFunction &MF, MachineBasicBlock &MBB,
                               MachineBasicBlock::iterator MBBI,
-                              const DebugLoc &DL, Register Reg) const {
+                              const DebugLoc &DL, Register Reg,
+                              SlotIndexes *Indexes) const {
   unsigned ExecMov = isWave32() ? AMDGPU::S_MOV_B32 : AMDGPU::S_MOV_B64;
   MCRegister Exec = isWave32() ? AMDGPU::EXEC_LO : AMDGPU::EXEC;
+  auto ExecRestoreMI =
       BuildMI(MBB, MBBI, DL, get(ExecMov), Exec).addReg(Reg, RegState::Kill);
+  if (Indexes)
+    Indexes->insertMachineInstrInMaps(*ExecRestoreMI);
 }
 
 static const TargetRegisterClass *
 adjustAllocatableRegClass(const GCNSubtarget &ST, const SIRegisterInfo &RI,
                           const MachineRegisterInfo &MRI,
                           const MCInstrDesc &TID, unsigned RCID,
                           bool IsAllocatable) {
   if ((IsAllocatable || !ST.hasGFX90AInsts() || !MRI.reservedRegsFrozen()) &&
SIInstrInfo::getSerializableMachineMemOperandTargetFlags() const {
   static const std::pair<MachineMemOperand::Flags, const char *> TargetFlags[] =
       {
           {MONoClobber, "amdgpu-noclobber"},
       };
 
   return ArrayRef(TargetFlags);
 }
 
+unsigned SIInstrInfo::getLiveRangeSplitOpcode(Register SrcReg,
+                                              const MachineFunction &MF) const {
+  const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
+  assert(SrcReg.isVirtual());
+  if (MFI->checkFlag(SrcReg, AMDGPU::VirtRegFlag::WWM_REG))
+    return AMDGPU::WWM_COPY;
+
+  return AMDGPU::COPY;
+}
+
 bool SIInstrInfo::isBasicBlockPrologue(const MachineInstr &MI) const {
   return !MI.isTerminator() && MI.getOpcode() != AMDGPU::COPY &&
          MI.modifiesRegister(AMDGPU::EXEC, &RI);
 }
 
 MachineInstrBuilder
 SIInstrInfo::getAddNoCarry(MachineBasicBlock &MBB,
                            MachineBasicBlock::iterator I,
MachineInstr *SIInstrInfo::foldMemoryOperandImpl(
   // We explicitly chose SReg_32 for the virtual register so such a copy might
   // be eliminated by RegisterCoalescer. However, that may not be possible, and
   // %0 may even spill. We can't spill $m0 normally (it would require copying to
   // a numbered SGPR anyway), and since it is in the SReg_32 register class,
   // TargetInstrInfo::foldMemoryOperand() is going to try.
   // A similar issue also exists with spilling and reloading $exec registers.
   //
   // To prevent that, constrain the %0 register class here.
-  if (MI.isFullCopy()) {
+  if (isFullCopyInstr(MI)) {
     Register DstReg = MI.getOperand(0).getReg();
     Register SrcReg = MI.getOperand(1).getReg();
     if ((DstReg.isVirtual() || SrcReg.isVirtual()) &&
         (DstReg.isVirtual() != SrcReg.isVirtual())) {
       MachineRegisterInfo &MRI = MF.getRegInfo();
       Register VirtReg = DstReg.isVirtual() ? DstReg : SrcReg;
       const TargetRegisterClass *RC = MRI.getRegClass(VirtReg);
       if (RC->hasSuperClassEq(&AMDGPU::SReg_32RegClass)) {
SIInstrInfo::getInstructionUniformity(const MachineInstr &MI) const {
 
   if (isNeverUniform(MI))
     return InstructionUniformity::NeverUniform;
 
   unsigned opcode = MI.getOpcode();
   if (opcode == AMDGPU::V_READLANE_B32 || opcode == AMDGPU::V_READFIRSTLANE_B32)
     return InstructionUniformity::AlwaysUniform;
 
-  if (MI.isCopy()) {
+  if (isCopyInstr(MI)) {
     const MachineOperand &srcOp = MI.getOperand(1);
     if (srcOp.isReg() && srcOp.getReg().isPhysical()) {
       const TargetRegisterClass *regClass =
           RI.getPhysRegBaseClass(srcOp.getReg());
       return RI.isSGPRClass(regClass) ? InstructionUniformity::AlwaysUniform
                                       : InstructionUniformity::NeverUniform;
     }
     return InstructionUniformity::Default;

llvm/lib/Target/AMDGPU/SIInstructions.td

 // accidentally clobber inactive channels of $vdst.
 let Constraints = "@earlyclobber $vdst" in {
 def STRICT_WWM : PseudoInstSI <(outs unknown:$vdst), (ins unknown:$src0)>;
 def STRICT_WQM : PseudoInstSI <(outs unknown:$vdst), (ins unknown:$src0)>;
 }
 
 } // End let hasSideEffects = 0, mayLoad = 0, mayStore = 0, Uses = [EXEC]
 
+def WWM_COPY : SPseudoInstSI <
+  (outs unknown:$dst), (ins unknown:$src)> {
+  let hasSideEffects = 0;
+  let isAsCheapAsAMove = 1;
+  let isConvergent = 1;
+}
+
 def ENTER_STRICT_WWM : SPseudoInstSI <(outs SReg_1:$sdst), (ins i64imm:$src0)> {
   let Uses = [EXEC];
   let Defs = [EXEC, SCC];
   let hasSideEffects = 0;
   let mayLoad = 0;
   let mayStore = 0;
 }
 

 defm : Int16Med3Pat<V_MED3_I16_e64, smin, smax>;
 defm : Int16Med3Pat<V_MED3_U16_e64, umin, umax>;
 } // End Predicates = [isGFX9Plus]
 
 class AMDGPUGenericInstruction : GenericInstruction {
   let Namespace = "AMDGPU";
 }
 
 // Convert a wave address to a swizzled vector address (i.e. this is
 // for copying the stack pointer to a vector address appropriate to
 // use in the offset field of mubuf instructions).
 def G_AMDGPU_WAVE_ADDRESS : AMDGPUGenericInstruction {
   let OutOperandList = (outs type0:$dst);
   let InOperandList = (ins type0:$src);
   let hasSideEffects = 0;
 }
 
 // Returns -1 if the input is zero.
 def G_AMDGPU_FFBH_U32 : AMDGPUGenericInstruction {
   let OutOperandList = (outs type0:$dst);
   let InOperandList = (ins type1:$src);
   let hasSideEffects = 0;
 }
 

llvm/lib/Target/AMDGPU/SILowerWWMCopies.cpp

+//===-- SILowerWWMCopies.cpp - Lower Copies after regalloc ---===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+/// \file
+/// Lowering the WWM_COPY instructions for various register classes.
+/// AMDGPU target generates WWM_COPY instruction to differentiate WWM
+/// copy from COPY. This pass generates the necessary exec mask manipulation
+/// instructions to replicate 'Whole Wave Mode' and lowers WWM_COPY back to
+/// COPY.
+//
+//===----------------------------------------------------------------------===//
+
+#include "AMDGPU.h"
+#include "GCNSubtarget.h"
+#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
+#include "SIMachineFunctionInfo.h"
+#include "llvm/CodeGen/LiveIntervals.h"
+#include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/VirtRegMap.h"
+#include "llvm/InitializePasses.h"
+
+using namespace llvm;
+
+#define DEBUG_TYPE "si-lower-wwm-copies"
+
+namespace {
+
+class SILowerWWMCopies : public MachineFunctionPass {
+public:
+  static char ID;
+
+  SILowerWWMCopies() : MachineFunctionPass(ID) {
+    initializeSILowerWWMCopiesPass(*PassRegistry::getPassRegistry());
+  }
+
+  bool runOnMachineFunction(MachineFunction &MF) override;
+
+  StringRef getPassName() const override { return "SI Lower WWM Copies"; }
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.setPreservesAll();
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+
+private:
+  bool isSCCLiveAtMI(const MachineInstr &MI);
+  void addToWWMSpills(MachineFunction &MF, Register Reg);
+
+  LiveIntervals *LIS;
+  SlotIndexes *Indexes;
+  VirtRegMap *VRM;
+  const SIRegisterInfo *TRI;
+  const MachineRegisterInfo *MRI;
+  SIMachineFunctionInfo *MFI;
+};
+
+} // End anonymous namespace.
+
+INITIALIZE_PASS_BEGIN(SILowerWWMCopies, DEBUG_TYPE, "SI Lower WWM Copies",
+                      false, false)
+INITIALIZE_PASS_DEPENDENCY(LiveIntervals)
+INITIALIZE_PASS_DEPENDENCY(VirtRegMap)
+INITIALIZE_PASS_END(SILowerWWMCopies, DEBUG_TYPE, "SI Lower WWM Copies", false,
+                    false)
+
+char SILowerWWMCopies::ID = 0;
+
+char &llvm::SILowerWWMCopiesID = SILowerWWMCopies::ID;
+
+bool SILowerWWMCopies::isSCCLiveAtMI(const MachineInstr &MI) {
+  // We can't determine the liveness info if LIS isn't available. Early return
+  // in that case and always assume SCC is live.
+  if (!LIS)
+    return true;
+
+  LiveRange &LR =
+      LIS->getRegUnit(*MCRegUnitIterator(MCRegister::from(AMDGPU::SCC), TRI));
+  SlotIndex Idx = LIS->getInstructionIndex(MI);
+  return LR.liveAt(Idx);
+}
+
+// If \p Reg is assigned with a physical VGPR, add the latter into wwm-spills
+// for preserving its entire lanes at function prolog/epilog.
+void SILowerWWMCopies::addToWWMSpills(MachineFunction &MF, Register Reg) {
+  if (Reg.isPhysical())
+    return;
+
+  Register PhysReg = VRM->getPhys(Reg);
+  assert(PhysReg != VirtRegMap::NO_PHYS_REG &&
+         "should have allocated a physical register");
+
+  MFI->allocateWWMSpill(MF, PhysReg);
+}
+
+bool SILowerWWMCopies::runOnMachineFunction(MachineFunction &MF) {
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+  const SIInstrInfo *TII = ST.getInstrInfo();
+
+  MFI = MF.getInfo<SIMachineFunctionInfo>();
+  LIS = getAnalysisIfAvailable<LiveIntervals>();
+  Indexes = getAnalysisIfAvailable<SlotIndexes>();
+  VRM = getAnalysisIfAvailable<VirtRegMap>();
+  TRI = ST.getRegisterInfo();
+  MRI = &MF.getRegInfo();
+
+  if (!MFI->hasVRegFlags())
+    return false;
+
+  bool Changed = false;
+  for (MachineBasicBlock &MBB : MF) {
+    for (MachineInstr &MI : MBB) {
+      if (MI.getOpcode() != AMDGPU::WWM_COPY)
+        continue;
+
+      // TODO: Club adjacent WWM ops between same exec save/restore
+      assert(TII->isVGPRCopy(MI));
+
+      // For WWM vector copies, manipulate the exec mask around the copy
+      // instruction.
+      const DebugLoc &DL = MI.getDebugLoc();
+      MachineBasicBlock::iterator InsertPt = MI.getIterator();
+      Register RegForExecCopy = MFI->getSGPRForEXECCopy();
+      TII->insertScratchExecCopy(MF, MBB, InsertPt, DL, RegForExecCopy,
+                                 isSCCLiveAtMI(MI), Indexes);
+      TII->restoreExec(MF, MBB, ++InsertPt, DL, RegForExecCopy, Indexes);
+      addToWWMSpills(MF, MI.getOperand(0).getReg());
+      LLVM_DEBUG(dbgs() << "WWM copy manipulation for " << MI);
+
+      // Lower WWM_COPY back to COPY
+      MI.setDesc(TII->get(AMDGPU::COPY));
+      Changed |= true;
+    }
+  }
+
+  return Changed;
+}

llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.h

public:
 
   bool checkFlag(Register Reg, uint8_t Flag) const {
     if (Reg.isPhysical())
       return false;
 
     return VRegFlags.inBounds(Reg) && VRegFlags[Reg] & Flag;
   }
 
+  bool hasVRegFlags() { return VRegFlags.size(); }
+
   void allocateWWMSpill(MachineFunction &MF, Register VGPR, uint64_t Size = 4,
                         Align Alignment = Align(4));
 
   void splitWWMSpillRegisters(
       MachineFunction &MF,
       SmallVectorImpl<std::pair<Register, int>> &CalleeSavedRegs,
       SmallVectorImpl<std::pair<Register, int>> &ScratchRegs) const;
 

llvm/test/CodeGen/AMDGPU/llc-pipeline.ll

 ; GCN-O0-NEXT:        SI Whole Quad Mode
 ; GCN-O0-NEXT:        Virtual Register Map
 ; GCN-O0-NEXT:        Live Register Matrix
 ; GCN-O0-NEXT:        SI Pre-allocate WWM Registers
 ; GCN-O0-NEXT:        AMDGPU Pre-RA Long Branch Reg
 ; GCN-O0-NEXT:        Fast Register Allocator
 ; GCN-O0-NEXT:        SI lower SGPR spill instructions
 ; GCN-O0-NEXT:        Fast Register Allocator
+; GCN-O0-NEXT:        SI Lower WWM Copies
 ; GCN-O0-NEXT:        SI Fix VGPR copies
 ; GCN-O0-NEXT:        Remove Redundant DEBUG_VALUE analysis
 ; GCN-O0-NEXT:        Fixup Statepoint Caller Saved
 ; GCN-O0-NEXT:        Lazy Machine Block Frequency Analysis
 ; GCN-O0-NEXT:        Machine Optimization Remark Emitter
 ; GCN-O0-NEXT:        Prologue/Epilogue Insertion & Frame Finalization
 ; GCN-O0-NEXT:        Post-RA pseudo instruction expansion pass
 ; GCN-O0-NEXT:        SI post-RA bundler

 ; GCN-O1-NEXT:        Lazy Machine Block Frequency Analysis
 ; GCN-O1-NEXT:        Machine Optimization Remark Emitter
 ; GCN-O1-NEXT:        Greedy Register Allocator
 ; GCN-O1-NEXT:        Virtual Register Rewriter
 ; GCN-O1-NEXT:        SI lower SGPR spill instructions
 ; GCN-O1-NEXT:        Virtual Register Map
 ; GCN-O1-NEXT:        Live Register Matrix
 ; GCN-O1-NEXT:        Greedy Register Allocator
+; GCN-O1-NEXT:        SI Lower WWM Copies
 ; GCN-O1-NEXT:        GCN NSA Reassign
 ; GCN-O1-NEXT:        Virtual Register Rewriter
 ; GCN-O1-NEXT:        Stack Slot Coloring
 ; GCN-O1-NEXT:        Machine Copy Propagation Pass
 ; GCN-O1-NEXT:        Machine Loop Invariant Code Motion
 ; GCN-O1-NEXT:        SI Fix VGPR copies
 ; GCN-O1-NEXT:        SI optimize exec mask operations
 ; GCN-O1-NEXT:        Remove Redundant DEBUG_VALUE analysis

 ; GCN-O1-OPTS-NEXT:        Lazy Machine Block Frequency Analysis
 ; GCN-O1-OPTS-NEXT:        Machine Optimization Remark Emitter
 ; GCN-O1-OPTS-NEXT:        Greedy Register Allocator
 ; GCN-O1-OPTS-NEXT:        Virtual Register Rewriter
 ; GCN-O1-OPTS-NEXT:        SI lower SGPR spill instructions
 ; GCN-O1-OPTS-NEXT:        Virtual Register Map
 ; GCN-O1-OPTS-NEXT:        Live Register Matrix
 ; GCN-O1-OPTS-NEXT:        Greedy Register Allocator
+; GCN-O1-OPTS-NEXT:        SI Lower WWM Copies
 ; GCN-O1-OPTS-NEXT:        GCN NSA Reassign
 ; GCN-O1-OPTS-NEXT:        Virtual Register Rewriter
 ; GCN-O1-OPTS-NEXT:        Stack Slot Coloring
 ; GCN-O1-OPTS-NEXT:        Machine Copy Propagation Pass
 ; GCN-O1-OPTS-NEXT:        Machine Loop Invariant Code Motion
 ; GCN-O1-OPTS-NEXT:        SI Fix VGPR copies
 ; GCN-O1-OPTS-NEXT:        SI optimize exec mask operations
 ; GCN-O1-OPTS-NEXT:        Remove Redundant DEBUG_VALUE analysis

 ; GCN-O2-NEXT:        Lazy Machine Block Frequency Analysis
 ; GCN-O2-NEXT:        Machine Optimization Remark Emitter
 ; GCN-O2-NEXT:        Greedy Register Allocator
 ; GCN-O2-NEXT:        Virtual Register Rewriter
 ; GCN-O2-NEXT:        SI lower SGPR spill instructions
 ; GCN-O2-NEXT:        Virtual Register Map
 ; GCN-O2-NEXT:        Live Register Matrix
 ; GCN-O2-NEXT:        Greedy Register Allocator
+; GCN-O2-NEXT:        SI Lower WWM Copies
 ; GCN-O2-NEXT:        GCN NSA Reassign
 ; GCN-O2-NEXT:        Virtual Register Rewriter
 ; GCN-O2-NEXT:        Stack Slot Coloring
 ; GCN-O2-NEXT:        Machine Copy Propagation Pass
 ; GCN-O2-NEXT:        Machine Loop Invariant Code Motion
 ; GCN-O2-NEXT:        SI Fix VGPR copies
 ; GCN-O2-NEXT:        SI optimize exec mask operations
 ; GCN-O2-NEXT:        Remove Redundant DEBUG_VALUE analysis

 ; GCN-O3-NEXT:        Lazy Machine Block Frequency Analysis
 ; GCN-O3-NEXT:        Machine Optimization Remark Emitter
 ; GCN-O3-NEXT:        Greedy Register Allocator
 ; GCN-O3-NEXT:        Virtual Register Rewriter
 ; GCN-O3-NEXT:        SI lower SGPR spill instructions
 ; GCN-O3-NEXT:        Virtual Register Map
 ; GCN-O3-NEXT:        Live Register Matrix
 ; GCN-O3-NEXT:        Greedy Register Allocator
+; GCN-O3-NEXT:        SI Lower WWM Copies
 ; GCN-O3-NEXT:        GCN NSA Reassign
 ; GCN-O3-NEXT:        Virtual Register Rewriter
 ; GCN-O3-NEXT:        Stack Slot Coloring
 ; GCN-O3-NEXT:        Machine Copy Propagation Pass
 ; GCN-O3-NEXT:        Machine Loop Invariant Code Motion
 ; GCN-O3-NEXT:        SI Fix VGPR copies
 ; GCN-O3-NEXT:        SI optimize exec mask operations
 ; GCN-O3-NEXT:        Remove Redundant DEBUG_VALUE analysis

llvm/test/CodeGen/AMDGPU/sgpr-regalloc-flags.ll

 ; REGALLOC: -regalloc not supported with amdgcn. Use -sgpr-regalloc and -vgpr-regalloc
 
 ; DEFAULT: Greedy Register Allocator
 ; DEFAULT-NEXT: Virtual Register Rewriter
 ; DEFAULT-NEXT: SI lower SGPR spill instructions
 ; DEFAULT-NEXT: Virtual Register Map
 ; DEFAULT-NEXT: Live Register Matrix
 ; DEFAULT-NEXT: Greedy Register Allocator
+; DEFAULT-NEXT: SI Lower WWM Copies
 ; DEFAULT-NEXT: GCN NSA Reassign
 ; DEFAULT-NEXT: Virtual Register Rewriter
 ; DEFAULT-NEXT: Stack Slot Coloring
 
 ; O0: Fast Register Allocator
 ; O0-NEXT: SI lower SGPR spill instructions
 ; O0-NEXT: Fast Register Allocator
+; O0-NEXT: SI Lower WWM Copies
 ; O0-NEXT: SI Fix VGPR copies
 
 
 
 
 ; BASIC-DEFAULT: Debug Variable Analysis
 ; BASIC-DEFAULT-NEXT: Live Stack Slot Analysis
 ; BASIC-DEFAULT-NEXT: Machine Natural Loop Construction
 ; BASIC-DEFAULT-NEXT: Machine Block Frequency Analysis
 ; BASIC-DEFAULT-NEXT: Virtual Register Map
 ; BASIC-DEFAULT-NEXT: Live Register Matrix
 ; BASIC-DEFAULT-NEXT: Basic Register Allocator
 ; BASIC-DEFAULT-NEXT: Virtual Register Rewriter
 ; BASIC-DEFAULT-NEXT: SI lower SGPR spill instructions
 ; BASIC-DEFAULT-NEXT: Virtual Register Map
 ; BASIC-DEFAULT-NEXT: Live Register Matrix
 ; BASIC-DEFAULT-NEXT: Bundle Machine CFG Edges
 ; BASIC-DEFAULT-NEXT: Spill Code Placement Analysis
 ; BASIC-DEFAULT-NEXT: Lazy Machine Block Frequency Analysis
 ; BASIC-DEFAULT-NEXT: Machine Optimization Remark Emitter
 ; BASIC-DEFAULT-NEXT: Greedy Register Allocator
+; BASIC-DEFAULT-NEXT: SI Lower WWM Copies
 ; BASIC-DEFAULT-NEXT: GCN NSA Reassign
 ; BASIC-DEFAULT-NEXT: Virtual Register Rewriter
 ; BASIC-DEFAULT-NEXT: Stack Slot Coloring
 
 
 
 ; DEFAULT-BASIC: Greedy Register Allocator
 ; DEFAULT-BASIC-NEXT: Virtual Register Rewriter
 ; DEFAULT-BASIC-NEXT: SI lower SGPR spill instructions
 ; DEFAULT-BASIC-NEXT: Virtual Register Map
 ; DEFAULT-BASIC-NEXT: Live Register Matrix
 ; DEFAULT-BASIC-NEXT: Basic Register Allocator
+; DEFAULT-BASIC-NEXT: SI Lower WWM Copies
 ; DEFAULT-BASIC-NEXT: GCN NSA Reassign
 ; DEFAULT-BASIC-NEXT: Virtual Register Rewriter
 ; DEFAULT-BASIC-NEXT: Stack Slot Coloring
 
 
 
 ; BASIC-BASIC: Debug Variable Analysis
 ; BASIC-BASIC-NEXT: Live Stack Slot Analysis
 ; BASIC-BASIC-NEXT: Machine Natural Loop Construction
 ; BASIC-BASIC-NEXT: Machine Block Frequency Analysis
 ; BASIC-BASIC-NEXT: Virtual Register Map
 ; BASIC-BASIC-NEXT: Live Register Matrix
 ; BASIC-BASIC-NEXT: Basic Register Allocator
 ; BASIC-BASIC-NEXT: Virtual Register Rewriter
 ; BASIC-BASIC-NEXT: SI lower SGPR spill instructions
 ; BASIC-BASIC-NEXT: Virtual Register Map
 ; BASIC-BASIC-NEXT: Live Register Matrix
 ; BASIC-BASIC-NEXT: Basic Register Allocator
+; BASIC-BASIC-NEXT: SI Lower WWM Copies
 ; BASIC-BASIC-NEXT: GCN NSA Reassign
 ; BASIC-BASIC-NEXT: Virtual Register Rewriter
 ; BASIC-BASIC-NEXT: Stack Slot Coloring
 
 
 declare void @bar()
 
 ; Something with some CSR SGPR spills