Commit 8e9cc63b authored by Marek Olsak's avatar Marek Olsak
Browse files

AMDGPU/SI: Add s_waitcnt at the end of non-void functions 

Summary:
v2: Make ReturnsVoid private, so that I can another 8 lines of code and
    look more productive.

Reviewers: tstellarAMD, arsenm

Subscribers: arsenm

Differential Revision: http://reviews.llvm.org/D16034

llvm-svn: 257622
parent 8a0f335a

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

+2 −0
Original line number Diff line number Diff line
@@ -893,6 +893,8 @@ SDValue SITargetLowering::LowerReturn(SDValue Chain, 
    return AMDGPUTargetLowering::LowerReturn(Chain, CallConv, isVarArg, Outs,

                                             OutVals, DL, DAG);



  Info->setIfReturnsVoid(Outs.size() == 0);



  SmallVector<ISD::OutputArg, 48> Splits;

  SmallVector<SDValue, 48> SplitVals;

llvm/lib/Target/AMDGPU/SIInsertWaits.cpp

+7 −1
Original line number Diff line number Diff line
@@ -84,6 +84,9 @@ private: 


  bool LastInstWritesM0;



  /// \brief Whether the machine function returns void

  bool ReturnsVoid;



  /// \brief Get increment/decrement amount for this instruction.

  Counters getHwCounts(MachineInstr &MI);
@@ -322,7 +325,9 @@ bool SIInsertWaits::insertWait(MachineBasicBlock &MBB, 
                               const Counters &Required) {



  // End of program? No need to wait on anything

  if (I != MBB.end() && I->getOpcode() == AMDGPU::S_ENDPGM)

  // A function not returning void needs to wait, because other bytecode will

  // be appended after it and we don't know what it will be.

  if (I != MBB.end() && I->getOpcode() == AMDGPU::S_ENDPGM && ReturnsVoid)

    return false;



  // Figure out if the async instructions execute in order
@@ -465,6 +470,7 @@ bool SIInsertWaits::runOnMachineFunction(MachineFunction &MF) { 
  LastIssued = ZeroCounts;

  LastOpcodeType = OTHER;

  LastInstWritesM0 = false;

  ReturnsVoid = MF.getInfo<SIMachineFunctionInfo>()->returnsVoid();



  memset(&UsedRegs, 0, sizeof(UsedRegs));

  memset(&DefinedRegs, 0, sizeof(DefinedRegs));

llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp

+1 −0
Original line number Diff line number Diff line
@@ -47,6 +47,7 @@ SIMachineFunctionInfo::SIMachineFunctionInfo(const MachineFunction &MF) 
    WorkGroupInfoSystemSGPR(AMDGPU::NoRegister),

    PrivateSegmentWaveByteOffsetSystemSGPR(AMDGPU::NoRegister),

    PSInputAddr(0),

    ReturnsVoid(true),

    LDSWaveSpillSize(0),

    PSInputEna(0),

    NumUserSGPRs(0),

llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.h

+9 −0
Original line number Diff line number Diff line
@@ -59,6 +59,7 @@ class SIMachineFunctionInfo : public AMDGPUMachineFunction { 


  // Graphics info.

  unsigned PSInputAddr;

  bool ReturnsVoid;



public:

  // FIXME: Make private
@@ -288,6 +289,14 @@ public: 
    PSInputAddr |= 1 << Index;

  }



  bool returnsVoid() const {

    return ReturnsVoid;

  }



  void setIfReturnsVoid(bool Value) {

    ReturnsVoid = Value;

  }



  unsigned getMaximumWorkGroupSize(const MachineFunction &MF) const;

};