bool SILowerControlFlow::runOnMachineFunction(MachineFunction &MF) {
const SISubtarget &ST = MF.getSubtarget<SISubtarget>();
TII = ST.getInstrInfo();
TRI = &TII->getRegisterInfo();
SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
bool HaveKill = false;
bool NeedFlat = false;
unsigned Depth = 0;
MachineFunction::iterator NextBB;
for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
BI != BE; BI = NextBB) {
NextBB = std::next(BI);
MachineBasicBlock &MBB = *BI;
MachineBasicBlock *EmptyMBBAtEnd = nullptr;
MachineBasicBlock::iterator I, Next;
bool ExecModified = false;
for (I = MBB.begin(); I != MBB.end(); I = Next) {
Next = std::next(I);
MachineInstr &MI = *I;
// Flat uses m0 in case it needs to access LDS.
if (TII->isFLAT(MI))
NeedFlat = true;
for (const auto &Def : I->defs()) {
if (Def.isReg() && Def.isDef() && Def.getReg() == AMDGPU::EXEC) {
ExecModified = true;
break;
}
}
switch (MI.getOpcode()) {
default: break;
case AMDGPU::SI_IF:
++Depth;
If(MI);
break;
case AMDGPU::SI_ELSE:
Else(MI, ExecModified);
break;
case AMDGPU::SI_BREAK:
Break(MI);
break;
case AMDGPU::SI_IF_BREAK:
IfBreak(MI);
break;
case AMDGPU::SI_ELSE_BREAK:
ElseBreak(MI);
break;
case AMDGPU::SI_LOOP:
++Depth;
Loop(MI);
break;
case AMDGPU::SI_END_CF:
if (--Depth == 0 && HaveKill) {
SkipIfDead(MI);
HaveKill = false;
}
EndCf(MI);
break;
case AMDGPU::SI_KILL:
if (Depth == 0)
SkipIfDead(MI);
else
HaveKill = true;
Kill(MI);
break;
case AMDGPU::S_BRANCH:
Branch(MI);
break;
case AMDGPU::SI_INDIRECT_SRC_V1:
case AMDGPU::SI_INDIRECT_SRC_V2:
case AMDGPU::SI_INDIRECT_SRC_V4:
case AMDGPU::SI_INDIRECT_SRC_V8:
case AMDGPU::SI_INDIRECT_SRC_V16:
if (indirectSrc(MI)) {
// The block was split at this point. We can safely skip the middle
// inserted block to the following which contains the rest of this
// block's instructions.
NextBB = std::next(BI);
BE = MF.end();
Next = MBB.end();
}
break;
case AMDGPU::SI_INDIRECT_DST_V1:
case AMDGPU::SI_INDIRECT_DST_V2:
case AMDGPU::SI_INDIRECT_DST_V4:
case AMDGPU::SI_INDIRECT_DST_V8:
case AMDGPU::SI_INDIRECT_DST_V16:
if (indirectDst(MI)) {
// The block was split at this point. We can safely skip the middle
// inserted block to the following which contains the rest of this
// block's instructions.
NextBB = std::next(BI);
BE = MF.end();
Next = MBB.end();
}
break;
case AMDGPU::S_ENDPGM: {
if (MF.getInfo<SIMachineFunctionInfo>()->returnsVoid())
break;
// Graphics shaders returning non-void shouldn't contain S_ENDPGM,
// because external bytecode will be appended at the end.
if (BI != --MF.end() || I != MBB.getFirstTerminator()) {
// S_ENDPGM is not the last instruction. Add an empty block at
// the end and jump there.
if (!EmptyMBBAtEnd) {
EmptyMBBAtEnd = MF.CreateMachineBasicBlock();
MF.insert(MF.end(), EmptyMBBAtEnd);
}
MBB.addSuccessor(EmptyMBBAtEnd);
BuildMI(*BI, I, MI.getDebugLoc(), TII->get(AMDGPU::S_BRANCH))
.addMBB(EmptyMBBAtEnd);
}
I->eraseFromParent();
break;
}
}
}
}
if (NeedFlat && MFI->IsKernel) {
// TODO: What to use with function calls?
// We will need to Initialize the flat scratch register pair.
if (NeedFlat)
MFI->setHasFlatInstructions(true);
}
return true;
}
bool SILowerControlFlow::runOnMachineFunction(MachineFunction &MF) {
TII = static_cast<const SIInstrInfo *>(MF.getSubtarget().getInstrInfo());
TRI =
static_cast<const SIRegisterInfo *>(MF.getSubtarget().getRegisterInfo());
SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
bool HaveKill = false;
bool NeedWQM = false;
bool NeedFlat = false;
unsigned Depth = 0;
for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
BI != BE; ++BI) {
MachineBasicBlock &MBB = *BI;
MachineBasicBlock::iterator I, Next;
for (I = MBB.begin(); I != MBB.end(); I = Next) {
Next = std::next(I);
MachineInstr &MI = *I;
if (TII->isWQM(MI) || TII->isDS(MI))
NeedWQM = true;
// Flat uses m0 in case it needs to access LDS.
if (TII->isFLAT(MI))
NeedFlat = true;
switch (MI.getOpcode()) {
default: break;
case AMDGPU::SI_IF:
++Depth;
If(MI);
break;
case AMDGPU::SI_ELSE:
Else(MI);
break;
case AMDGPU::SI_BREAK:
Break(MI);
break;
case AMDGPU::SI_IF_BREAK:
IfBreak(MI);
break;
case AMDGPU::SI_ELSE_BREAK:
ElseBreak(MI);
break;
case AMDGPU::SI_LOOP:
++Depth;
Loop(MI);
break;
case AMDGPU::SI_END_CF:
if (--Depth == 0 && HaveKill) {
SkipIfDead(MI);
HaveKill = false;
}
EndCf(MI);
break;
case AMDGPU::SI_KILL:
if (Depth == 0)
SkipIfDead(MI);
else
HaveKill = true;
Kill(MI);
break;
case AMDGPU::S_BRANCH:
Branch(MI);
break;
case AMDGPU::SI_INDIRECT_SRC_V1:
case AMDGPU::SI_INDIRECT_SRC_V2:
case AMDGPU::SI_INDIRECT_SRC_V4:
case AMDGPU::SI_INDIRECT_SRC_V8:
case AMDGPU::SI_INDIRECT_SRC_V16:
IndirectSrc(MI);
break;
case AMDGPU::SI_INDIRECT_DST_V1:
case AMDGPU::SI_INDIRECT_DST_V2:
case AMDGPU::SI_INDIRECT_DST_V4:
case AMDGPU::SI_INDIRECT_DST_V8:
case AMDGPU::SI_INDIRECT_DST_V16:
IndirectDst(MI);
break;
}
}
}
if (NeedWQM && MFI->getShaderType() == ShaderType::PIXEL) {
MachineBasicBlock &MBB = MF.front();
BuildMI(MBB, MBB.getFirstNonPHI(), DebugLoc(), TII->get(AMDGPU::S_WQM_B64),
AMDGPU::EXEC).addReg(AMDGPU::EXEC);
}
if (NeedFlat && MFI->IsKernel) {
// TODO: What to use with function calls?
// We will need to Initialize the flat scratch register pair.
if (NeedFlat)
MFI->setHasFlatInstructions(true);
}
return true;
}