AMDGPU: Add operand target flags serialization

[arsenm]

r306995

lib/Target/AMDGPU/SIInstrInfo.h

protected:
 
   MachineInstr *commuteInstructionImpl(MachineInstr &MI, bool NewMI,
                                        unsigned OpIdx0,
                                        unsigned OpIdx1) const override;
 
 public:
 
   enum TargetOperandFlags {
+    MO_MASK = 0x7,
+
     MO_NONE = 0,
     // MO_GOTPCREL -> symbol@GOTPCREL -> R_AMDGPU_GOTPCREL.
     MO_GOTPCREL = 1,
     // MO_GOTPCREL32_LO -> symbol@gotpcrel32@lo -> R_AMDGPU_GOTPCREL32_LO.
     MO_GOTPCREL32 = 2,
     MO_GOTPCREL32_LO = 2,
     // MO_GOTPCREL32_HI -> symbol@gotpcrel32@hi -> R_AMDGPU_GOTPCREL32_HI.
     MO_GOTPCREL32_HI = 3,
public:
   bool isNonUniformBranchInstr(MachineInstr &Instr) const;
 
   void convertNonUniformIfRegion(MachineBasicBlock *IfEntry,
                                  MachineBasicBlock *IfEnd) const;
 
   void convertNonUniformLoopRegion(MachineBasicBlock *LoopEntry,
                                    MachineBasicBlock *LoopEnd) const;
 
+  std::pair<unsigned, unsigned>
+  decomposeMachineOperandsTargetFlags(unsigned TF) const override;
+
   ArrayRef<std::pair<int, const char *>>
   getSerializableTargetIndices() const override;
 
+  ArrayRef<std::pair<unsigned, const char *>>
+  getSerializableDirectMachineOperandTargetFlags() const override;
+
   ScheduleHazardRecognizer *
   CreateTargetPostRAHazardRecognizer(const InstrItineraryData *II,
                                  const ScheduleDAG *DAG) const override;
 
   ScheduleHazardRecognizer *
   CreateTargetPostRAHazardRecognizer(const MachineFunction &MF) const override;
 
   bool isBasicBlockPrologue(const MachineInstr &MI) const override;

lib/Target/AMDGPU/SIInstrInfo.cpp

 
 /// This is the hazard recognizer used at -O0 by the PostRAHazardRecognizer
 /// pass.
 ScheduleHazardRecognizer *
 SIInstrInfo::CreateTargetPostRAHazardRecognizer(const MachineFunction &MF) const {
   return new GCNHazardRecognizer(MF);
 }
 
+std::pair<unsigned, unsigned>
+SIInstrInfo::decomposeMachineOperandsTargetFlags(unsigned TF) const {
+  return std::make_pair(TF & MO_MASK, TF & ~MO_MASK);
+}
+
+ArrayRef<std::pair<unsigned, const char *>>
+SIInstrInfo::getSerializableDirectMachineOperandTargetFlags() const {
+  static const std::pair<unsigned, const char *> TargetFlags[] = {
+    { MO_GOTPCREL, "amdgpu-gotprel" },
+    { MO_GOTPCREL32_LO, "amdgpu-gotprel32-lo" },
+    { MO_GOTPCREL32_HI, "amdgpu-gotprel32-hi" },
+    { MO_REL32_LO, "amdgpu-rel32-lo" },
+    { MO_REL32_HI, "amdgpu-rel32-hi" }
+  };
+
+  return makeArrayRef(TargetFlags);
+}
+
 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,

test/CodeGen/MIR/AMDGPU/target-flags.mir

+# RUN: llc -march=amdgcn -run-pass none -o - %s | FileCheck %s
+--- |
+  define amdgpu_kernel void @flags() {
+    ret void
+  }
+
+  declare void @foo()
+...
+---
+
+# CHECK: SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @foo + 4, target-flags(amdgpu-rel32-hi) @foo + 4, implicit-def dead %scc
+# CHECK: %1 = S_MOV_B64 target-flags(amdgpu-gotprel) @foo
+
+name: flags
+liveins:
+  - { reg: '%sgpr0_sgpr1' }
+frameInfo:
+  maxAlignment:  8
+registers:
+  - { id: 0, class: sreg_64, preferred-register: '' }
+  - { id: 1, class: sreg_64, preferred-register: '' }
+body: |
+  bb.0:
+    liveins: %sgpr0_sgpr1
+    %0 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @foo + 4, target-flags(amdgpu-rel32-hi) @foo + 4, implicit-def dead %scc
+  %1 = S_MOV_B64 target-flags(amdgpu-gotprel) @foo
+
+    S_ENDPGM
+...