void AlphaRegisterInfo::
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I) const {
const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
if (TFI->hasFP(MF)) {
// If we have a frame pointer, turn the adjcallstackup instruction into a
// 'sub ESP, <amt>' and the adjcallstackdown instruction into 'add ESP,
// <amt>'
MachineInstr *Old = I;
uint64_t Amount = Old->getOperand(0).getImm();
if (Amount != 0) {
// We need to keep the stack aligned properly. To do this, we round the
// amount of space needed for the outgoing arguments up to the next
// alignment boundary.
unsigned Align = TFI->getStackAlignment();
Amount = (Amount+Align-1)/Align*Align;
MachineInstr *New;
if (Old->getOpcode() == Alpha::ADJUSTSTACKDOWN) {
New=BuildMI(MF, Old->getDebugLoc(), TII.get(Alpha::LDA), Alpha::R30)
.addImm(-Amount).addReg(Alpha::R30);
} else {
assert(Old->getOpcode() == Alpha::ADJUSTSTACKUP);
New=BuildMI(MF, Old->getDebugLoc(), TII.get(Alpha::LDA), Alpha::R30)
.addImm(Amount).addReg(Alpha::R30);
}
// Replace the pseudo instruction with a new instruction...
MBB.insert(I, New);
}
}
MBB.erase(I);
}
/// ProcessSDDbgValues - Process SDDbgValues associated with this node.
static void ProcessSDDbgValues(SDNode *N, SelectionDAG *DAG,
InstrEmitter &Emitter,
SmallVector<std::pair<unsigned, MachineInstr*>, 32> &Orders,
DenseMap<SDValue, unsigned> &VRBaseMap,
unsigned Order) {
if (!N->getHasDebugValue())
return;
// Opportunistically insert immediate dbg_value uses, i.e. those with source
// order number right after the N.
MachineBasicBlock *BB = Emitter.getBlock();
MachineBasicBlock::iterator InsertPos = Emitter.getInsertPos();
ArrayRef<SDDbgValue*> DVs = DAG->GetDbgValues(N);
for (unsigned i = 0, e = DVs.size(); i != e; ++i) {
if (DVs[i]->isInvalidated())
continue;
unsigned DVOrder = DVs[i]->getOrder();
if (!Order || DVOrder == ++Order) {
MachineInstr *DbgMI = Emitter.EmitDbgValue(DVs[i], VRBaseMap);
if (DbgMI) {
Orders.push_back(std::make_pair(DVOrder, DbgMI));
BB->insert(InsertPos, DbgMI);
}
DVs[i]->setIsInvalidated();
}
}
}
bool PatmosInstrInfo::moveTo(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &Target,
MachineBasicBlock::iterator &Source,
SmallVectorImpl<MachineOperand> *Pred,
bool Negate) const
{
if (Target->isBundle()) return false;
if (Target->getOpcode() == Patmos::NOP) {
// replace the NOP with the source instruction
Source = MBB.insert(Target, MBB.remove(Source));
MBB.erase(Target);
if (Pred) {
PredicateInstruction(&*Source, *Pred);
}
if (Negate) {
NegatePredicate(&*Source);
}
return true;
}
// TODO check if we can bundle the target and the source instruction, do so
return false;
}
bool MIRParserImpl::initializeMachineBasicBlock(
MachineFunction &MF, MachineBasicBlock &MBB,
const yaml::MachineBasicBlock &YamlMBB,
const PerFunctionMIParsingState &PFS) {
MBB.setAlignment(YamlMBB.Alignment);
if (YamlMBB.AddressTaken)
MBB.setHasAddressTaken();
MBB.setIsLandingPad(YamlMBB.IsLandingPad);
SMDiagnostic Error;
// Parse the successors.
for (const auto &MBBSource : YamlMBB.Successors) {
MachineBasicBlock *SuccMBB = nullptr;
if (parseMBBReference(SuccMBB, MBBSource, MF, PFS))
return true;
// TODO: Report an error when adding the same successor more than once.
MBB.addSuccessor(SuccMBB);
}
// Parse the liveins.
for (const auto &LiveInSource : YamlMBB.LiveIns) {
unsigned Reg = 0;
if (parseNamedRegisterReference(Reg, SM, MF, LiveInSource.Value, PFS,
IRSlots, Error))
return error(Error, LiveInSource.SourceRange);
MBB.addLiveIn(Reg);
}
// Parse the instructions.
for (const auto &MISource : YamlMBB.Instructions) {
MachineInstr *MI = nullptr;
if (parseMachineInstr(MI, SM, MF, MISource.Value, PFS, IRSlots, Error))
return error(Error, MISource.SourceRange);
MBB.insert(MBB.end(), MI);
}
return false;
}
// MI is a 128-bit load or store. Split it into two 64-bit loads or stores,
// each having the opcode given by NewOpcode.
void SystemZInstrInfo::splitMove(MachineBasicBlock::iterator MI,
unsigned NewOpcode) const {
MachineBasicBlock *MBB = MI->getParent();
MachineFunction &MF = *MBB->getParent();
// Get two load or store instructions. Use the original instruction for one
// of them (arbitarily the second here) and create a clone for the other.
MachineInstr *EarlierMI = MF.CloneMachineInstr(MI);
MBB->insert(MI, EarlierMI);
// Set up the two 64-bit registers.
MachineOperand &HighRegOp = EarlierMI->getOperand(0);
MachineOperand &LowRegOp = MI->getOperand(0);
HighRegOp.setReg(RI.getSubReg(HighRegOp.getReg(), SystemZ::subreg_high));
LowRegOp.setReg(RI.getSubReg(LowRegOp.getReg(), SystemZ::subreg_low));
// The address in the first (high) instruction is already correct.
// Adjust the offset in the second (low) instruction.
MachineOperand &HighOffsetOp = EarlierMI->getOperand(2);
MachineOperand &LowOffsetOp = MI->getOperand(2);
LowOffsetOp.setImm(LowOffsetOp.getImm() + 8);
// Set the opcodes.
unsigned HighOpcode = getOpcodeForOffset(NewOpcode, HighOffsetOp.getImm());
unsigned LowOpcode = getOpcodeForOffset(NewOpcode, LowOffsetOp.getImm());
assert(HighOpcode && LowOpcode && "Both offsets should be in range");
EarlierMI->setDesc(get(HighOpcode));
MI->setDesc(get(LowOpcode));
}
void
PPCInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
unsigned DestReg, int FrameIdx,
const TargetRegisterClass *RC,
const TargetRegisterInfo *TRI) const {
MachineFunction &MF = *MBB.getParent();
SmallVector<MachineInstr*, 4> NewMIs;
DebugLoc DL;
if (MI != MBB.end()) DL = MI->getDebugLoc();
if (LoadRegFromStackSlot(MF, DL, DestReg, FrameIdx, RC, NewMIs)) {
PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
FuncInfo->setSpillsCR();
}
for (unsigned i = 0, e = NewMIs.size(); i != e; ++i)
MBB.insert(MI, NewMIs[i]);
const MachineFrameInfo &MFI = *MF.getFrameInfo();
MachineMemOperand *MMO =
MF.getMachineMemOperand(MachinePointerInfo::getFixedStack(FrameIdx),
MachineMemOperand::MOLoad,
MFI.getObjectSize(FrameIdx),
MFI.getObjectAlignment(FrameIdx));
NewMIs.back()->addMemOperand(MF, MMO);
}
void
PPCInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
unsigned SrcReg, bool isKill, int FrameIdx,
const TargetRegisterClass *RC,
const TargetRegisterInfo *TRI) const {
MachineFunction &MF = *MBB.getParent();
SmallVector<MachineInstr*, 4> NewMIs;
if (StoreRegToStackSlot(MF, SrcReg, isKill, FrameIdx, RC, NewMIs)) {
PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
FuncInfo->setSpillsCR();
}
for (unsigned i = 0, e = NewMIs.size(); i != e; ++i)
MBB.insert(MI, NewMIs[i]);
const MachineFrameInfo &MFI = *MF.getFrameInfo();
MachineMemOperand *MMO =
MF.getMachineMemOperand(MachinePointerInfo::getFixedStack(FrameIdx),
MachineMemOperand::MOStore,
MFI.getObjectSize(FrameIdx),
MFI.getObjectAlignment(FrameIdx));
NewMIs.back()->addMemOperand(MF, MMO);
}
// Eliminate ADJCALLSTACKDOWN/ADJCALLSTACKUP pseudo instructions
void XTCFrameLowering::
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I) const {
const XTCInstrInfo &TII =
*static_cast<const XTCInstrInfo*>(MF.getTarget().getInstrInfo());
if (!hasReservedCallFrame(MF)) {
MachineInstr *Old = I;
int Amount = Old->getOperand(0).getImm() + 4;
if (Amount != 0) {
// We need to keep the stack aligned properly. To do this, we round the
// amount of space needed for the outgoing arguments up to the next
// alignment boundary.
unsigned Align = getStackAlignment();
Amount = (Amount+Align-1)/Align*Align;
MachineInstr *New;
if (Old->getOpcode() == XTC::ADJCALLSTACKDOWN) {
New = BuildMI(MF,Old->getDebugLoc(), TII.get(XTC::ADDI),XTC::r15)
.addReg(XTC::r15).addImm(-Amount);
} else {
assert(Old->getOpcode() == XTC::ADJCALLSTACKUP);
New = BuildMI(MF,Old->getDebugLoc(), TII.get(XTC::ADDI),XTC::r15)
.addReg(XTC::r15).addImm(Amount);
}
// Replace the pseudo instruction with a new instruction...
MBB.insert(I, New);
}
}
// Simply discard ADJCALLSTACKDOWN, ADJCALLSTACKUP instructions.
MBB.erase(I);
}
bool RAFast::handleGCRegRoot(MachineInstr *MI) {
MachineOperand &MO = MI->getOperand(0);
if (!MO.isReg())
return false;
unsigned Reg = MO.getReg();
if (!TargetRegisterInfo::isVirtualRegister(Reg))
return false;
LiveRegMap::iterator LRI = findLiveVirtReg(Reg);
if (LRI != LiveVirtRegs.end()) {
setPhysReg(MI, 0, LRI->PhysReg);
return true;
}
int FrameIndex = StackSlotForVirtReg[Reg];
unsigned RootIndex = MI->getOperand(1).getImm();
DebugLoc DL = MI->getDebugLoc();
MachineInstr *NewMI = TII->emitFrameIndexGCRegRoot(*MF, FrameIndex,
RootIndex, DL);
MachineBasicBlock *MBB = MI->getParent();
MBB->insert(MBB->erase(MI), NewMI);
GCFunctionInfo &GCFI = MF->getGMI()->getFunctionInfo(*MF->getFunction());
GCFI.spillRegRoot(RootIndex, FrameIndex);
return true;
}
static void moveInstsAfter(MachineBasicBlock::iterator I,
ArrayRef<MachineInstr *> InstsToMove) {
MachineBasicBlock *MBB = I->getParent();
++I;
for (MachineInstr *MI : InstsToMove) {
MI->removeFromParent();
MBB->insert(I, MI);
}
}
void TargetInstrInfo::reMaterialize(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned DestReg, unsigned SubIdx,
const MachineInstr &Orig,
const TargetRegisterInfo &TRI) const {
MachineInstr *MI = MBB.getParent()->CloneMachineInstr(&Orig);
MI->substituteRegister(MI->getOperand(0).getReg(), DestReg, SubIdx, TRI);
MBB.insert(I, MI);
}
/// UnfoldAndRewriteInstruction - Rewrite specified instruction by unfolding
/// folded memory references and replacing those references with register
/// references instead.
void
StackSlotColoring::UnfoldAndRewriteInstruction(MachineInstr *MI, int OldFI,
unsigned Reg,
const TargetRegisterClass *RC,
SmallSet<unsigned, 4> &Defs,
MachineFunction &MF) {
MachineBasicBlock *MBB = MI->getParent();
if (unsigned DstReg = TII->isLoadFromStackSlot(MI, OldFI)) {
if (PropagateForward(MI, MBB, DstReg, Reg)) {
DEBUG(dbgs() << "Eliminated load: ");
DEBUG(MI->dump());
++NumLoadElim;
} else {
BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(TargetOpcode::COPY),
DstReg).addReg(Reg);
++NumRegRepl;
}
if (!Defs.count(Reg)) {
// If this is the first use of Reg in this MBB and it wasn't previously
// defined in MBB, add it to livein.
MBB->addLiveIn(Reg);
Defs.insert(Reg);
}
} else if (unsigned SrcReg = TII->isStoreToStackSlot(MI, OldFI)) {
if (MI->killsRegister(SrcReg) && PropagateBackward(MI, MBB, SrcReg, Reg)) {
DEBUG(dbgs() << "Eliminated store: ");
DEBUG(MI->dump());
++NumStoreElim;
} else {
BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(TargetOpcode::COPY), Reg)
.addReg(SrcReg);
++NumRegRepl;
}
// Remember reg has been defined in MBB.
Defs.insert(Reg);
} else {
SmallVector<MachineInstr*, 4> NewMIs;
bool Success = TII->unfoldMemoryOperand(MF, MI, Reg, false, false, NewMIs);
Success = Success; // Silence compiler warning.
assert(Success && "Failed to unfold!");
MachineInstr *NewMI = NewMIs[0];
MBB->insert(MI, NewMI);
++NumRegRepl;
if (NewMI->readsRegister(Reg)) {
if (!Defs.count(Reg))
// If this is the first use of Reg in this MBB and it wasn't previously
// defined in MBB, add it to livein.
MBB->addLiveIn(Reg);
Defs.insert(Reg);
}
}
MBB->erase(MI);
}
/// Unfold a load from the given machineinstr if the load itself could be
/// hoisted. Return the unfolded and hoistable load, or null if the load
/// couldn't be unfolded or if it wouldn't be hoistable.
MachineInstr *MachineLICM::ExtractHoistableLoad(MachineInstr *MI) {
// Don't unfold simple loads.
if (MI->canFoldAsLoad())
return nullptr;
// If not, we may be able to unfold a load and hoist that.
// First test whether the instruction is loading from an amenable
// memory location.
if (!MI->isInvariantLoad(AA))
return nullptr;
// Next determine the register class for a temporary register.
unsigned LoadRegIndex;
unsigned NewOpc =
TII->getOpcodeAfterMemoryUnfold(MI->getOpcode(),
/*UnfoldLoad=*/true,
/*UnfoldStore=*/false,
&LoadRegIndex);
if (NewOpc == 0) return nullptr;
const MCInstrDesc &MID = TII->get(NewOpc);
if (MID.getNumDefs() != 1) return nullptr;
MachineFunction &MF = *MI->getParent()->getParent();
const TargetRegisterClass *RC = TII->getRegClass(MID, LoadRegIndex, TRI, MF);
// Ok, we're unfolding. Create a temporary register and do the unfold.
unsigned Reg = MRI->createVirtualRegister(RC);
SmallVector<MachineInstr *, 2> NewMIs;
bool Success =
TII->unfoldMemoryOperand(MF, MI, Reg,
/*UnfoldLoad=*/true, /*UnfoldStore=*/false,
NewMIs);
(void)Success;
assert(Success &&
"unfoldMemoryOperand failed when getOpcodeAfterMemoryUnfold "
"succeeded!");
assert(NewMIs.size() == 2 &&
"Unfolded a load into multiple instructions!");
MachineBasicBlock *MBB = MI->getParent();
MachineBasicBlock::iterator Pos = MI;
MBB->insert(Pos, NewMIs[0]);
MBB->insert(Pos, NewMIs[1]);
// If unfolding produced a load that wasn't loop-invariant or profitable to
// hoist, discard the new instructions and bail.
if (!IsLoopInvariantInst(*NewMIs[0]) || !IsProfitableToHoist(*NewMIs[0])) {
NewMIs[0]->eraseFromParent();
NewMIs[1]->eraseFromParent();
return nullptr;
}
// Update register pressure for the unfolded instruction.
UpdateRegPressure(NewMIs[1]);
// Otherwise we successfully unfolded a load that we can hoist.
MI->eraseFromParent();
return NewMIs[0];
}
/// foldMemoryOperand - Attempt to fold a load or store of the specified stack
/// slot into the specified machine instruction for the specified operand(s).
/// If this is possible, a new instruction is returned with the specified
/// operand folded, otherwise NULL is returned. The client is responsible for
/// removing the old instruction and adding the new one in the instruction
/// stream.
MachineInstr*
TargetInstrInfo::foldMemoryOperand(MachineBasicBlock::iterator MI,
const SmallVectorImpl<unsigned> &Ops,
int FI) const {
unsigned Flags = 0;
for (unsigned i = 0, e = Ops.size(); i != e; ++i)
if (MI->getOperand(Ops[i]).isDef())
Flags |= MachineMemOperand::MOStore;
else
Flags |= MachineMemOperand::MOLoad;
MachineBasicBlock *MBB = MI->getParent();
assert(MBB && "foldMemoryOperand needs an inserted instruction");
MachineFunction &MF = *MBB->getParent();
// Ask the target to do the actual folding.
if (MachineInstr *NewMI = foldMemoryOperandImpl(MF, MI, Ops, FI)) {
// Add a memory operand, foldMemoryOperandImpl doesn't do that.
assert((!(Flags & MachineMemOperand::MOStore) ||
NewMI->getDesc().mayStore()) &&
"Folded a def to a non-store!");
assert((!(Flags & MachineMemOperand::MOLoad) ||
NewMI->getDesc().mayLoad()) &&
"Folded a use to a non-load!");
const MachineFrameInfo &MFI = *MF.getFrameInfo();
assert(MFI.getObjectOffset(FI) != -1);
MachineMemOperand *MMO =
MF.getMachineMemOperand(
MachinePointerInfo(PseudoSourceValue::getFixedStack(FI)),
Flags, MFI.getObjectSize(FI),
MFI.getObjectAlignment(FI));
NewMI->addMemOperand(MF, MMO);
// FIXME: change foldMemoryOperandImpl semantics to also insert NewMI.
return MBB->insert(MI, NewMI);
}
// Straight COPY may fold as load/store.
if (!MI->isCopy() || Ops.size() != 1)
return 0;
const TargetRegisterClass *RC = canFoldCopy(MI, Ops[0]);
if (!RC)
return 0;
const MachineOperand &MO = MI->getOperand(1-Ops[0]);
MachineBasicBlock::iterator Pos = MI;
const TargetRegisterInfo *TRI = MF.getTarget().getRegisterInfo();
if (Flags == MachineMemOperand::MOStore)
storeRegToStackSlot(*MBB, Pos, MO.getReg(), MO.isKill(), FI, RC, TRI);
else
loadRegFromStackSlot(*MBB, Pos, MO.getReg(), FI, RC, TRI);
return --Pos;
}
/// spillVirtReg - Do the actual work of spilling.
void RAFast::spillVirtReg(MachineBasicBlock::iterator MI,
LiveRegMap::iterator LRI) {
LiveReg &LR = *LRI;
assert(PhysRegState[LR.PhysReg] == LRI->VirtReg && "Broken RegState mapping");
if (LR.Dirty) {
// If this physreg is used by the instruction, we want to kill it on the
// instruction, not on the spill.
bool SpillKill = LR.LastUse != MI;
LR.Dirty = false;
DEBUG(dbgs() << "Spilling " << PrintReg(LRI->VirtReg, TRI)
<< " in " << PrintReg(LR.PhysReg, TRI));
const TargetRegisterClass *RC = MRI->getRegClass(LRI->VirtReg);
int FI = getStackSpaceFor(LRI->VirtReg, RC);
DEBUG(dbgs() << " to stack slot #" << FI << "\n");
TII->storeRegToStackSlot(*MBB, MI, LR.PhysReg, SpillKill, FI, RC, TRI);
++NumStores; // Update statistics
// If this register is used by DBG_VALUE then insert new DBG_VALUE to
// identify spilled location as the place to find corresponding variable's
// value.
SmallVector<MachineInstr *, 4> &LRIDbgValues =
LiveDbgValueMap[LRI->VirtReg];
for (unsigned li = 0, le = LRIDbgValues.size(); li != le; ++li) {
MachineInstr *DBG = LRIDbgValues[li];
const MDNode *MDPtr =
DBG->getOperand(DBG->getNumOperands()-1).getMetadata();
int64_t Offset = 0;
if (DBG->getOperand(1).isImm())
Offset = DBG->getOperand(1).getImm();
DebugLoc DL;
if (MI == MBB->end()) {
// If MI is at basic block end then use last instruction's location.
MachineBasicBlock::iterator EI = MI;
DL = (--EI)->getDebugLoc();
}
else
DL = MI->getDebugLoc();
if (MachineInstr *NewDV =
TII->emitFrameIndexDebugValue(*MF, FI, Offset, MDPtr, DL)) {
MachineBasicBlock *MBB = DBG->getParent();
MBB->insert(MI, NewDV);
DEBUG(dbgs() << "Inserting debug info due to spill:" << "\n" << *NewDV);
}
}
// Now this register is spilled there is should not be any DBG_VALUE
// pointing to this register because they are all pointing to spilled value
// now.
LRIDbgValues.clear();
if (SpillKill)
LR.LastUse = 0; // Don't kill register again
}
killVirtReg(LRI);
}
bool Thumb1FrameLowering::
restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI,
const TargetRegisterInfo *TRI) const {
if (CSI.empty())
return false;
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const TargetInstrInfo &TII = *STI.getInstrInfo();
bool isVarArg = AFI->getArgRegsSaveSize() > 0;
DebugLoc DL = MI != MBB.end() ? MI->getDebugLoc() : DebugLoc();
MachineInstrBuilder MIB = BuildMI(MF, DL, TII.get(ARM::tPOP));
AddDefaultPred(MIB);
bool NeedsPop = false;
for (unsigned i = CSI.size(); i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
if (Reg == ARM::LR) {
if (MBB.succ_empty()) {
// Special epilogue for vararg functions. See emitEpilogue
if (isVarArg)
continue;
// ARMv4T requires BX, see emitEpilogue
if (!STI.hasV5TOps())
continue;
Reg = ARM::PC;
(*MIB).setDesc(TII.get(ARM::tPOP_RET));
if (MI != MBB.end())
MIB.copyImplicitOps(*MI);
MI = MBB.erase(MI);
} else
// LR may only be popped into PC, as part of return sequence.
// If this isn't the return sequence, we'll need emitPopSpecialFixUp
// to restore LR the hard way.
continue;
}
MIB.addReg(Reg, getDefRegState(true));
NeedsPop = true;
}
// It's illegal to emit pop instruction without operands.
if (NeedsPop)
MBB.insert(MI, &*MIB);
else
MF.DeleteMachineInstr(MIB);
return true;
}
// This function eliminates ADJCALLSTACKDOWN,
// ADJCALLSTACKUP pseudo instructions
void XCoreRegisterInfo::
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I) const {
const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
if (!TFI->hasReservedCallFrame(MF)) {
// Turn the adjcallstackdown instruction into 'extsp <amt>' and the
// adjcallstackup instruction into 'ldaw sp, sp[<amt>]'
MachineInstr *Old = I;
uint64_t Amount = Old->getOperand(0).getImm();
if (Amount != 0) {
// We need to keep the stack aligned properly. To do this, we round the
// amount of space needed for the outgoing arguments up to the next
// alignment boundary.
unsigned Align = TFI->getStackAlignment();
Amount = (Amount+Align-1)/Align*Align;
assert(Amount%4 == 0);
Amount /= 4;
bool isU6 = isImmU6(Amount);
if (!isU6 && !isImmU16(Amount)) {
// FIX could emit multiple instructions in this case.
#ifndef NDEBUG
errs() << "eliminateCallFramePseudoInstr size too big: "
<< Amount << "\n";
#endif
llvm_unreachable(0);
}
MachineInstr *New;
if (Old->getOpcode() == XCore::ADJCALLSTACKDOWN) {
int Opcode = isU6 ? XCore::EXTSP_u6 : XCore::EXTSP_lu6;
New=BuildMI(MF, Old->getDebugLoc(), TII.get(Opcode))
.addImm(Amount);
} else {
assert(Old->getOpcode() == XCore::ADJCALLSTACKUP);
int Opcode = isU6 ? XCore::LDAWSP_ru6_RRegs : XCore::LDAWSP_lru6_RRegs;
New=BuildMI(MF, Old->getDebugLoc(), TII.get(Opcode), XCore::SP)
.addImm(Amount);
}
// Replace the pseudo instruction with a new instruction...
MBB.insert(I, New);
}
}
MBB.erase(I);
}
bool Thumb1FrameLowering::
restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI,
const TargetRegisterInfo *TRI) const {
if (CSI.empty())
return false;
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const TargetInstrInfo &TII = *STI.getInstrInfo();
bool isVarArg = AFI->getArgRegsSaveSize() > 0;
DebugLoc DL = MI->getDebugLoc();
MachineInstrBuilder MIB = BuildMI(MF, DL, TII.get(ARM::tPOP));
AddDefaultPred(MIB);
bool NumRegs = false;
for (unsigned i = CSI.size(); i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
if (Reg == ARM::LR) {
// Special epilogue for vararg functions. See emitEpilogue
if (isVarArg)
continue;
// ARMv4T requires BX, see emitEpilogue
if (STI.hasV4TOps() && !STI.hasV5TOps())
continue;
Reg = ARM::PC;
(*MIB).setDesc(TII.get(ARM::tPOP_RET));
MIB.copyImplicitOps(&*MI);
MI = MBB.erase(MI);
}
MIB.addReg(Reg, getDefRegState(true));
NumRegs = true;
}
// It's illegal to emit pop instruction without operands.
if (NumRegs)
MBB.insert(MI, &*MIB);
else
MF.DeleteMachineInstr(MIB);
return true;
}
void SILowerControlFlow::emitLoadM0FromVGPRLoop(MachineBasicBlock &LoopBB,
DebugLoc DL,
MachineInstr *MovRel,
const MachineOperand &IdxReg,
int Offset) {
MachineBasicBlock::iterator I = LoopBB.begin();
// Read the next variant into VCC (lower 32 bits) <- also loop target
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::V_READFIRSTLANE_B32), AMDGPU::VCC_LO)
.addReg(IdxReg.getReg(), getUndefRegState(IdxReg.isUndef()));
// Move index from VCC into M0
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_MOV_B32), AMDGPU::M0)
.addReg(AMDGPU::VCC_LO);
// Compare the just read M0 value to all possible Idx values
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::V_CMP_EQ_U32_e32))
.addReg(AMDGPU::M0)
.addReg(IdxReg.getReg(), getUndefRegState(IdxReg.isUndef()));
// Update EXEC, save the original EXEC value to VCC
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_AND_SAVEEXEC_B64), AMDGPU::VCC)
.addReg(AMDGPU::VCC);
if (Offset) {
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_ADD_I32), AMDGPU::M0)
.addReg(AMDGPU::M0)
.addImm(Offset);
}
// Do the actual move
LoopBB.insert(I, MovRel);
// Update EXEC, switch all done bits to 0 and all todo bits to 1
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_XOR_B64), AMDGPU::EXEC)
.addReg(AMDGPU::EXEC)
.addReg(AMDGPU::VCC);
// Loop back to V_READFIRSTLANE_B32 if there are still variants to cover
BuildMI(LoopBB, I, DL, TII->get(AMDGPU::S_CBRANCH_EXECNZ))
.addMBB(&LoopBB);
}
bool MIRParserImpl::initializeMachineBasicBlock(
MachineFunction &MF, MachineBasicBlock &MBB,
const yaml::MachineBasicBlock &YamlMBB,
const DenseMap<unsigned, MachineBasicBlock *> &MBBSlots) {
MBB.setAlignment(YamlMBB.Alignment);
if (YamlMBB.AddressTaken)
MBB.setHasAddressTaken();
MBB.setIsLandingPad(YamlMBB.IsLandingPad);
// Parse the instructions.
for (const auto &MISource : YamlMBB.Instructions) {
SMDiagnostic Error;
if (auto *MI = parseMachineInstr(SM, MF, MISource.Value, MBBSlots, IRSlots,
Error)) {
MBB.insert(MBB.end(), MI);
continue;
}
reportDiagnostic(diagFromMIStringDiag(Error, MISource.SourceRange));
return true;
}
return false;
}
// ProcessSourceNode - Process nodes with source order numbers. These are added
// to a vector which EmitSchedule uses to determine how to insert dbg_value
// instructions in the right order.
static void ProcessSourceNode(SDNode *N, SelectionDAG *DAG,
InstrEmitter &Emitter,
DenseMap<SDValue, unsigned> &VRBaseMap,
SmallVector<std::pair<unsigned, MachineInstr*>, 32> &Orders,
SmallSet<unsigned, 8> &Seen) {
unsigned Order = DAG->GetOrdering(N);
if (!Order || !Seen.insert(Order))
return;
MachineBasicBlock *BB = Emitter.getBlock();
if (Emitter.getInsertPos() == BB->begin() || BB->back().isPHI()) {
// Did not insert any instruction.
Orders.push_back(std::make_pair(Order, (MachineInstr*)0));
return;
}
Orders.push_back(std::make_pair(Order, prior(Emitter.getInsertPos())));
if (!N->getHasDebugValue())
return;
// Opportunistically insert immediate dbg_value uses, i.e. those with source
// order number right after the N.
MachineBasicBlock::iterator InsertPos = Emitter.getInsertPos();
SmallVector<SDDbgValue*,2> &DVs = DAG->GetDbgValues(N);
for (unsigned i = 0, e = DVs.size(); i != e; ++i) {
if (DVs[i]->isInvalidated())
continue;
unsigned DVOrder = DVs[i]->getOrder();
if (DVOrder == ++Order) {
MachineInstr *DbgMI = Emitter.EmitDbgValue(DVs[i], VRBaseMap);
if (DbgMI) {
Orders.push_back(std::make_pair(DVOrder, DbgMI));
BB->insert(InsertPos, DbgMI);
}
DVs[i]->setIsInvalidated();
}
}
}
/// ProcessSDDbgValues - Process SDDbgValues associated with this node.
static void
ProcessSDDbgValues(SDNode *N, SelectionDAG *DAG, InstrEmitter &Emitter,
SmallVectorImpl<std::pair<unsigned, MachineInstr*> > &Orders,
DenseMap<SDValue, unsigned> &VRBaseMap, unsigned Order) {
if (!N->getHasDebugValue())
return;
// Opportunistically insert immediate dbg_value uses, i.e. those with the same
// source order number as N.
MachineBasicBlock *BB = Emitter.getBlock();
MachineBasicBlock::iterator InsertPos = Emitter.getInsertPos();
for (auto DV : DAG->GetDbgValues(N)) {
if (DV->isEmitted())
continue;
unsigned DVOrder = DV->getOrder();
if (!Order || DVOrder == Order) {
MachineInstr *DbgMI = Emitter.EmitDbgValue(DV, VRBaseMap);
if (DbgMI) {
Orders.push_back({DVOrder, DbgMI});
BB->insert(InsertPos, DbgMI);
}
}
}
}
// This function eliminate ADJCALLSTACKDOWN/ADJCALLSTACKUP pseudo instructions
void MBlazeRegisterInfo::
eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
MachineBasicBlock::iterator I) const {
const TargetFrameInfo *TFI = MF.getTarget().getFrameInfo();
if (!TFI->hasReservedCallFrame(MF)) {
// If we have a frame pointer, turn the adjcallstackup instruction into a
// 'addi r1, r1, -<amt>' and the adjcallstackdown instruction into
// 'addi r1, r1, <amt>'
MachineInstr *Old = I;
int Amount = Old->getOperand(0).getImm() + 4;
if (Amount != 0) {
// We need to keep the stack aligned properly. To do this, we round the
// amount of space needed for the outgoing arguments up to the next
// alignment boundary.
unsigned Align = MF.getTarget().getFrameInfo()->getStackAlignment();
Amount = (Amount+Align-1)/Align*Align;
MachineInstr *New;
if (Old->getOpcode() == MBlaze::ADJCALLSTACKDOWN) {
New = BuildMI(MF, Old->getDebugLoc(), TII.get(MBlaze::ADDI), MBlaze::R1)
.addReg(MBlaze::R1).addImm(-Amount);
} else {
assert(Old->getOpcode() == MBlaze::ADJCALLSTACKUP);
New = BuildMI(MF, Old->getDebugLoc(), TII.get(MBlaze::ADDI), MBlaze::R1)
.addReg(MBlaze::R1).addImm(Amount);
}
// Replace the pseudo instruction with a new instruction...
MBB.insert(I, New);
}
}
// Simply discard ADJCALLSTACKDOWN, ADJCALLSTACKUP instructions.
MBB.erase(I);
}
bool Thumb1InstrInfo::
restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI) const {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
if (CSI.empty())
return false;
bool isVarArg = AFI->getVarArgsRegSaveSize() > 0;
DebugLoc DL = MI->getDebugLoc();
MachineInstrBuilder MIB = BuildMI(MF, DL, get(ARM::tPOP));
AddDefaultPred(MIB);
MIB.addReg(0); // No write back.
bool NumRegs = 0;
for (unsigned i = CSI.size(); i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
if (Reg == ARM::LR) {
// Special epilogue for vararg functions. See emitEpilogue
if (isVarArg)
continue;
Reg = ARM::PC;
(*MIB).setDesc(get(ARM::tPOP_RET));
MI = MBB.erase(MI);
}
MIB.addReg(Reg, getDefRegState(true));
++NumRegs;
}
// It's illegal to emit pop instruction without operands.
if (NumRegs)
MBB.insert(MI, &*MIB);
return true;
}
void LembergInstrInfo::reMaterialize(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I,
unsigned DestReg,
unsigned SubIdx,
const MachineInstr *Orig,
const TargetRegisterInfo &TRI) const {
MachineInstr *MI = MBB.getParent()->CloneMachineInstr(Orig);
MI->substituteRegister(MI->getOperand(0).getReg(), DestReg, SubIdx, TRI);
// Make sure registers end up in the right register class
if (MI->getOpcode() == Lemberg::LOADga
|| MI->getOpcode() == Lemberg::LOADga_off
|| MI->getOpcode() == Lemberg::LOADuimm19s2) {
assert((TRI.isVirtualRegister(DestReg) || Lemberg::AImmRegClass.contains(DestReg))
&& "Cannot rematerialize this instruction to arbitrary physical register");
if (TRI.isVirtualRegister(DestReg)) {
MachineRegisterInfo &MRI = MBB.getParent()->getRegInfo();
MRI.constrainRegClass(DestReg, Lemberg::AImmRegisterClass);
}
}
MBB.insert(I, MI);
}
void RAFast::AllocateBasicBlock() {
DEBUG(dbgs() << "\nAllocating " << *MBB);
PhysRegState.assign(TRI->getNumRegs(), regDisabled);
assert(LiveVirtRegs.empty() && "Mapping not cleared from last block?");
MachineBasicBlock::iterator MII = MBB->begin();
// Add live-in registers as live.
for (MachineBasicBlock::livein_iterator I = MBB->livein_begin(),
E = MBB->livein_end(); I != E; ++I)
if (RegClassInfo.isAllocatable(*I))
definePhysReg(MII, *I, regReserved);
SmallVector<unsigned, 8> VirtDead;
SmallVector<MachineInstr*, 32> Coalesced;
// Otherwise, sequentially allocate each instruction in the MBB.
while (MII != MBB->end()) {
MachineInstr *MI = MII++;
const MCInstrDesc &MCID = MI->getDesc();
DEBUG({
dbgs() << "\n>> " << *MI << "Regs:";
for (unsigned Reg = 1, E = TRI->getNumRegs(); Reg != E; ++Reg) {
if (PhysRegState[Reg] == regDisabled) continue;
dbgs() << " " << TRI->getName(Reg);
switch(PhysRegState[Reg]) {
case regFree:
break;
case regReserved:
dbgs() << "*";
break;
default: {
dbgs() << '=' << PrintReg(PhysRegState[Reg]);
LiveRegMap::iterator I = findLiveVirtReg(PhysRegState[Reg]);
assert(I != LiveVirtRegs.end() && "Missing VirtReg entry");
if (I->Dirty)
dbgs() << "*";
assert(I->PhysReg == Reg && "Bad inverse map");
break;
}
}
}
dbgs() << '\n';
// Check that LiveVirtRegs is the inverse.
for (LiveRegMap::iterator i = LiveVirtRegs.begin(),
e = LiveVirtRegs.end(); i != e; ++i) {
assert(TargetRegisterInfo::isVirtualRegister(i->VirtReg) &&
"Bad map key");
assert(TargetRegisterInfo::isPhysicalRegister(i->PhysReg) &&
"Bad map value");
assert(PhysRegState[i->PhysReg] == i->VirtReg && "Bad inverse map");
}
});
// Debug values are not allowed to change codegen in any way.
if (MI->isDebugValue()) {
bool ScanDbgValue = true;
while (ScanDbgValue) {
ScanDbgValue = false;
for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg()) continue;
unsigned Reg = MO.getReg();
if (!TargetRegisterInfo::isVirtualRegister(Reg)) continue;
LiveRegMap::iterator LRI = findLiveVirtReg(Reg);
if (LRI != LiveVirtRegs.end())
setPhysReg(MI, i, LRI->PhysReg);
else {
int SS = StackSlotForVirtReg[Reg];
if (SS == -1) {
// We can't allocate a physreg for a DebugValue, sorry!
DEBUG(dbgs() << "Unable to allocate vreg used by DBG_VALUE");
MO.setReg(0);
}
else {
// Modify DBG_VALUE now that the value is in a spill slot.
int64_t Offset = MI->getOperand(1).getImm();
const MDNode *MDPtr =
MI->getOperand(MI->getNumOperands()-1).getMetadata();
DebugLoc DL = MI->getDebugLoc();
if (MachineInstr *NewDV =
TII->emitFrameIndexDebugValue(*MF, SS, Offset, MDPtr, DL)) {
DEBUG(dbgs() << "Modifying debug info due to spill:" <<
"\t" << *MI);
MachineBasicBlock *MBB = MI->getParent();
MBB->insert(MBB->erase(MI), NewDV);
// Scan NewDV operands from the beginning.
MI = NewDV;
ScanDbgValue = true;
break;
} else {
// We can't allocate a physreg for a DebugValue; sorry!
DEBUG(dbgs() << "Unable to allocate vreg used by DBG_VALUE");
MO.setReg(0);
}
}
}
LiveDbgValueMap[Reg].push_back(MI);
}
}
// Next instruction.
continue;
}
// GC register roots need special care here, since inserting reloads can
// break the call sequence.
if (MI->isGCRegRoot()) {
handleGCRegRoot(MI);
continue;
}
// If this is a copy, we may be able to coalesce.
unsigned CopySrc = 0, CopyDst = 0, CopySrcSub = 0, CopyDstSub = 0;
if (MI->isCopy()) {
CopyDst = MI->getOperand(0).getReg();
CopySrc = MI->getOperand(1).getReg();
CopyDstSub = MI->getOperand(0).getSubReg();
CopySrcSub = MI->getOperand(1).getSubReg();
}
// Track registers used by instruction.
UsedInInstr.reset();
// First scan.
// Mark physreg uses and early clobbers as used.
// Find the end of the virtreg operands
unsigned VirtOpEnd = 0;
bool hasTiedOps = false;
bool hasEarlyClobbers = false;
bool hasPartialRedefs = false;
bool hasPhysDefs = false;
for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg()) continue;
unsigned Reg = MO.getReg();
if (!Reg) continue;
if (TargetRegisterInfo::isVirtualRegister(Reg)) {
VirtOpEnd = i+1;
if (MO.isUse()) {
hasTiedOps = hasTiedOps ||
MCID.getOperandConstraint(i, MCOI::TIED_TO) != -1;
} else {
if (MO.isEarlyClobber())
hasEarlyClobbers = true;
if (MO.getSubReg() && MI->readsVirtualRegister(Reg))
hasPartialRedefs = true;
}
continue;
}
if (!RegClassInfo.isAllocatable(Reg)) continue;
if (MO.isUse()) {
usePhysReg(MO);
} else if (MO.isEarlyClobber()) {
definePhysReg(MI, Reg, (MO.isImplicit() || MO.isDead()) ?
regFree : regReserved);
hasEarlyClobbers = true;
} else
hasPhysDefs = true;
}
// The instruction may have virtual register operands that must be allocated
// the same register at use-time and def-time: early clobbers and tied
// operands. If there are also physical defs, these registers must avoid
// both physical defs and uses, making them more constrained than normal
// operands.
// Similarly, if there are multiple defs and tied operands, we must make
// sure the same register is allocated to uses and defs.
// We didn't detect inline asm tied operands above, so just make this extra
// pass for all inline asm.
if (MI->isInlineAsm() || hasEarlyClobbers || hasPartialRedefs ||
(hasTiedOps && (hasPhysDefs || MCID.getNumDefs() > 1))) {
handleThroughOperands(MI, VirtDead);
// Don't attempt coalescing when we have funny stuff going on.
CopyDst = 0;
// Pretend we have early clobbers so the use operands get marked below.
// This is not necessary for the common case of a single tied use.
hasEarlyClobbers = true;
}
// Second scan.
// Allocate virtreg uses.
for (unsigned i = 0; i != VirtOpEnd; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg()) continue;
unsigned Reg = MO.getReg();
if (!TargetRegisterInfo::isVirtualRegister(Reg)) continue;
if (MO.isUse()) {
LiveRegMap::iterator LRI = reloadVirtReg(MI, i, Reg, CopyDst);
unsigned PhysReg = LRI->PhysReg;
CopySrc = (CopySrc == Reg || CopySrc == PhysReg) ? PhysReg : 0;
if (setPhysReg(MI, i, PhysReg))
killVirtReg(LRI);
}
}
MRI->addPhysRegsUsed(UsedInInstr);
// Track registers defined by instruction - early clobbers and tied uses at
// this point.
UsedInInstr.reset();
if (hasEarlyClobbers) {
for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg()) continue;
unsigned Reg = MO.getReg();
if (!Reg || !TargetRegisterInfo::isPhysicalRegister(Reg)) continue;
// Look for physreg defs and tied uses.
if (!MO.isDef() && !MI->isRegTiedToDefOperand(i)) continue;
for (MCRegAliasIterator AI(Reg, TRI, true); AI.isValid(); ++AI)
UsedInInstr.set(*AI);
}
}
unsigned DefOpEnd = MI->getNumOperands();
if (MI->isCall()) {
// Spill all virtregs before a call. This serves two purposes: 1. If an
// exception is thrown, the landing pad is going to expect to find
// registers in their spill slots, and 2. we don't have to wade through
// all the <imp-def> operands on the call instruction.
DefOpEnd = VirtOpEnd;
DEBUG(dbgs() << " Spilling remaining registers before call.\n");
spillAll(MI);
// The imp-defs are skipped below, but we still need to mark those
// registers as used by the function.
SkippedInstrs.insert(&MCID);
}
// Third scan.
// Allocate defs and collect dead defs.
for (unsigned i = 0; i != DefOpEnd; ++i) {
MachineOperand &MO = MI->getOperand(i);
if (!MO.isReg() || !MO.isDef() || !MO.getReg() || MO.isEarlyClobber())
continue;
unsigned Reg = MO.getReg();
if (TargetRegisterInfo::isPhysicalRegister(Reg)) {
if (!RegClassInfo.isAllocatable(Reg)) continue;
definePhysReg(MI, Reg, (MO.isImplicit() || MO.isDead()) ?
regFree : regReserved);
continue;
}
LiveRegMap::iterator LRI = defineVirtReg(MI, i, Reg, CopySrc);
unsigned PhysReg = LRI->PhysReg;
if (setPhysReg(MI, i, PhysReg)) {
VirtDead.push_back(Reg);
CopyDst = 0; // cancel coalescing;
} else
CopyDst = (CopyDst == Reg || CopyDst == PhysReg) ? PhysReg : 0;
}
// Kill dead defs after the scan to ensure that multiple defs of the same
// register are allocated identically. We didn't need to do this for uses
// because we are crerating our own kill flags, and they are always at the
// last use.
for (unsigned i = 0, e = VirtDead.size(); i != e; ++i)
killVirtReg(VirtDead[i]);
VirtDead.clear();
MRI->addPhysRegsUsed(UsedInInstr);
if (CopyDst && CopyDst == CopySrc && CopyDstSub == CopySrcSub) {
DEBUG(dbgs() << "-- coalescing: " << *MI);
Coalesced.push_back(MI);
} else {
DEBUG(dbgs() << "<< " << *MI);
}
}
bool Thumb1FrameLowering::
restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
std::vector<CalleeSavedInfo> &CSI,
const TargetRegisterInfo *TRI) const {
if (CSI.empty())
return false;
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const TargetInstrInfo &TII = *STI.getInstrInfo();
const ARMBaseRegisterInfo *RegInfo = static_cast<const ARMBaseRegisterInfo *>(
MF.getSubtarget().getRegisterInfo());
bool isVarArg = AFI->getArgRegsSaveSize() > 0;
DebugLoc DL = MI != MBB.end() ? MI->getDebugLoc() : DebugLoc();
ARMRegSet LoRegsToRestore;
ARMRegSet HiRegsToRestore;
// Low registers (r0-r7) which can be used to restore the high registers.
ARMRegSet CopyRegs;
for (CalleeSavedInfo I : CSI) {
unsigned Reg = I.getReg();
if (ARM::tGPRRegClass.contains(Reg) || Reg == ARM::LR) {
LoRegsToRestore[Reg] = true;
} else if (ARM::hGPRRegClass.contains(Reg) && Reg != ARM::LR) {
HiRegsToRestore[Reg] = true;
} else {
llvm_unreachable("callee-saved register of unexpected class");
}
// If this is a low register not used as the frame pointer, we may want to
// use it for restoring the high registers.
if ((ARM::tGPRRegClass.contains(Reg)) &&
!(hasFP(MF) && Reg == RegInfo->getFrameRegister(MF)))
CopyRegs[Reg] = true;
}
// If this is a return block, we may be able to use some unused return value
// registers for restoring the high regs.
auto Terminator = MBB.getFirstTerminator();
if (Terminator != MBB.end() && Terminator->getOpcode() == ARM::tBX_RET) {
CopyRegs[ARM::R0] = true;
CopyRegs[ARM::R1] = true;
CopyRegs[ARM::R2] = true;
CopyRegs[ARM::R3] = true;
for (auto Op : Terminator->implicit_operands()) {
if (Op.isReg())
CopyRegs[Op.getReg()] = false;
}
}
static const unsigned AllCopyRegs[] = {ARM::R0, ARM::R1, ARM::R2, ARM::R3,
ARM::R4, ARM::R5, ARM::R6, ARM::R7};
static const unsigned AllHighRegs[] = {ARM::R8, ARM::R9, ARM::R10, ARM::R11};
const unsigned *AllCopyRegsEnd = std::end(AllCopyRegs);
const unsigned *AllHighRegsEnd = std::end(AllHighRegs);
// Find the first register to restore.
auto HiRegToRestore = findNextOrderedReg(std::begin(AllHighRegs),
HiRegsToRestore, AllHighRegsEnd);
while (HiRegToRestore != AllHighRegsEnd) {
assert(!CopyRegs.none());
// Find the first low register to use.
auto CopyReg =
findNextOrderedReg(std::begin(AllCopyRegs), CopyRegs, AllCopyRegsEnd);
// Create the POP instruction.
MachineInstrBuilder PopMIB =
BuildMI(MBB, MI, DL, TII.get(ARM::tPOP)).add(predOps(ARMCC::AL));
while (HiRegToRestore != AllHighRegsEnd && CopyReg != AllCopyRegsEnd) {
// Add the low register to the POP.
PopMIB.addReg(*CopyReg, RegState::Define);
// Create the MOV from low to high register.
BuildMI(MBB, MI, DL, TII.get(ARM::tMOVr))
.addReg(*HiRegToRestore, RegState::Define)
.addReg(*CopyReg, RegState::Kill)
.add(predOps(ARMCC::AL));
CopyReg = findNextOrderedReg(++CopyReg, CopyRegs, AllCopyRegsEnd);
HiRegToRestore =
findNextOrderedReg(++HiRegToRestore, HiRegsToRestore, AllHighRegsEnd);
}
}
MachineInstrBuilder MIB =
BuildMI(MF, DL, TII.get(ARM::tPOP)).add(predOps(ARMCC::AL));
bool NeedsPop = false;
for (unsigned i = CSI.size(); i != 0; --i) {
CalleeSavedInfo &Info = CSI[i-1];
unsigned Reg = Info.getReg();
// High registers (excluding lr) have already been dealt with
if (!(ARM::tGPRRegClass.contains(Reg) || Reg == ARM::LR))
continue;
if (Reg == ARM::LR) {
Info.setRestored(false);
if (!MBB.succ_empty() ||
MI->getOpcode() == ARM::TCRETURNdi ||
MI->getOpcode() == ARM::TCRETURNri)
// LR may only be popped into PC, as part of return sequence.
// If this isn't the return sequence, we'll need emitPopSpecialFixUp
// to restore LR the hard way.
// FIXME: if we don't pass any stack arguments it would be actually
// advantageous *and* correct to do the conversion to an ordinary call
// instruction here.
continue;
// Special epilogue for vararg functions. See emitEpilogue
if (isVarArg)
continue;
// ARMv4T requires BX, see emitEpilogue
if (!STI.hasV5TOps())
continue;
// Pop LR into PC.
Reg = ARM::PC;
(*MIB).setDesc(TII.get(ARM::tPOP_RET));
if (MI != MBB.end())
MIB.copyImplicitOps(*MI);
MI = MBB.erase(MI);
}
MIB.addReg(Reg, getDefRegState(true));
NeedsPop = true;
}
// It's illegal to emit pop instruction without operands.
if (NeedsPop)
MBB.insert(MI, &*MIB);
else
MF.DeleteMachineInstr(MIB);
return true;
}
/// spillAroundUses - insert spill code around each use of Reg.
void InlineSpiller::spillAroundUses(unsigned Reg) {
DEBUG(dbgs() << "spillAroundUses " << PrintReg(Reg) << '\n');
LiveInterval &OldLI = LIS.getInterval(Reg);
// Iterate over instructions using Reg.
for (MachineRegisterInfo::reg_iterator RegI = MRI.reg_begin(Reg);
MachineInstr *MI = RegI.skipBundle();) {
// Debug values are not allowed to affect codegen.
if (MI->isDebugValue()) {
// Modify DBG_VALUE now that the value is in a spill slot.
uint64_t Offset = MI->getOperand(1).getImm();
const MDNode *MDPtr = MI->getOperand(2).getMetadata();
DebugLoc DL = MI->getDebugLoc();
if (MachineInstr *NewDV = TII.emitFrameIndexDebugValue(MF, StackSlot,
Offset, MDPtr, DL)) {
DEBUG(dbgs() << "Modifying debug info due to spill:" << "\t" << *MI);
MachineBasicBlock *MBB = MI->getParent();
MBB->insert(MBB->erase(MI), NewDV);
} else {
DEBUG(dbgs() << "Removing debug info due to spill:" << "\t" << *MI);
MI->eraseFromParent();
}
continue;
}
// Ignore copies to/from snippets. We'll delete them.
if (SnippetCopies.count(MI))
continue;
// Stack slot accesses may coalesce away.
if (coalesceStackAccess(MI, Reg))
continue;
// Analyze instruction.
SmallVector<std::pair<MachineInstr*, unsigned>, 8> Ops;
MIBundleOperands::VirtRegInfo RI =
MIBundleOperands(MI).analyzeVirtReg(Reg, &Ops);
// Find the slot index where this instruction reads and writes OldLI.
// This is usually the def slot, except for tied early clobbers.
SlotIndex Idx = LIS.getInstructionIndex(MI).getRegSlot();
if (VNInfo *VNI = OldLI.getVNInfoAt(Idx.getRegSlot(true)))
if (SlotIndex::isSameInstr(Idx, VNI->def))
Idx = VNI->def;
// Check for a sibling copy.
unsigned SibReg = isFullCopyOf(MI, Reg);
if (SibReg && isSibling(SibReg)) {
// This may actually be a copy between snippets.
if (isRegToSpill(SibReg)) {
DEBUG(dbgs() << "Found new snippet copy: " << *MI);
SnippetCopies.insert(MI);
continue;
}
if (RI.Writes) {
// Hoist the spill of a sib-reg copy.
if (hoistSpill(OldLI, MI)) {
// This COPY is now dead, the value is already in the stack slot.
MI->getOperand(0).setIsDead();
DeadDefs.push_back(MI);
continue;
}
} else {
// This is a reload for a sib-reg copy. Drop spills downstream.
LiveInterval &SibLI = LIS.getInterval(SibReg);
eliminateRedundantSpills(SibLI, SibLI.getVNInfoAt(Idx));
// The COPY will fold to a reload below.
}
}
// Attempt to fold memory ops.
if (foldMemoryOperand(Ops))
continue;
// Allocate interval around instruction.
// FIXME: Infer regclass from instruction alone.
LiveInterval &NewLI = Edit->createFrom(Reg);
NewLI.markNotSpillable();
if (RI.Reads)
insertReload(NewLI, Idx, MI);
// Rewrite instruction operands.
bool hasLiveDef = false;
for (unsigned i = 0, e = Ops.size(); i != e; ++i) {
MachineOperand &MO = Ops[i].first->getOperand(Ops[i].second);
MO.setReg(NewLI.reg);
if (MO.isUse()) {
if (!Ops[i].first->isRegTiedToDefOperand(Ops[i].second))
MO.setIsKill();
} else {
if (!MO.isDead())
hasLiveDef = true;
}
}
DEBUG(dbgs() << "\trewrite: " << Idx << '\t' << *MI);
// FIXME: Use a second vreg if instruction has no tied ops.
if (RI.Writes) {
if (hasLiveDef)
insertSpill(NewLI, OldLI, Idx, MI);
else {
// This instruction defines a dead value. We don't need to spill it,
// but do create a live range for the dead value.
VNInfo *VNI = NewLI.getNextValue(Idx, LIS.getVNInfoAllocator());
NewLI.addRange(LiveRange(Idx, Idx.getDeadSlot(), VNI));
}
}
DEBUG(dbgs() << "\tinterval: " << NewLI << '\n');
}
}
/// EmitSchedule - Emit the machine code in scheduled order. Return the new
/// InsertPos and MachineBasicBlock that contains this insertion
/// point. ScheduleDAGSDNodes holds a BB pointer for convenience, but this does
/// not necessarily refer to returned BB. The emitter may split blocks.
MachineBasicBlock *ScheduleDAGSDNodes::
EmitSchedule(MachineBasicBlock::iterator &InsertPos) {
InstrEmitter Emitter(BB, InsertPos);
DenseMap<SDValue, unsigned> VRBaseMap;
DenseMap<SUnit*, unsigned> CopyVRBaseMap;
SmallVector<std::pair<unsigned, MachineInstr*>, 32> Orders;
SmallSet<unsigned, 8> Seen;
bool HasDbg = DAG->hasDebugValues();
// If this is the first BB, emit byval parameter dbg_value's.
if (HasDbg && BB->getParent()->begin() == MachineFunction::iterator(BB)) {
SDDbgInfo::DbgIterator PDI = DAG->ByvalParmDbgBegin();
SDDbgInfo::DbgIterator PDE = DAG->ByvalParmDbgEnd();
for (; PDI != PDE; ++PDI) {
MachineInstr *DbgMI= Emitter.EmitDbgValue(*PDI, VRBaseMap);
if (DbgMI)
BB->insert(InsertPos, DbgMI);
}
}
for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
SUnit *SU = Sequence[i];
if (!SU) {
// Null SUnit* is a noop.
TII->insertNoop(*Emitter.getBlock(), InsertPos);
continue;
}
// For pre-regalloc scheduling, create instructions corresponding to the
// SDNode and any glued SDNodes and append them to the block.
if (!SU->getNode()) {
// Emit a copy.
EmitPhysRegCopy(SU, CopyVRBaseMap, InsertPos);
continue;
}
SmallVector<SDNode *, 4> GluedNodes;
for (SDNode *N = SU->getNode()->getGluedNode(); N;
N = N->getGluedNode())
GluedNodes.push_back(N);
while (!GluedNodes.empty()) {
SDNode *N = GluedNodes.back();
Emitter.EmitNode(GluedNodes.back(), SU->OrigNode != SU, SU->isCloned,
VRBaseMap);
// Remember the source order of the inserted instruction.
if (HasDbg)
ProcessSourceNode(N, DAG, Emitter, VRBaseMap, Orders, Seen);
GluedNodes.pop_back();
}
Emitter.EmitNode(SU->getNode(), SU->OrigNode != SU, SU->isCloned,
VRBaseMap);
// Remember the source order of the inserted instruction.
if (HasDbg)
ProcessSourceNode(SU->getNode(), DAG, Emitter, VRBaseMap, Orders,
Seen);
}
// Insert all the dbg_values which have not already been inserted in source
// order sequence.
if (HasDbg) {
MachineBasicBlock::iterator BBBegin = BB->getFirstNonPHI();
// Sort the source order instructions and use the order to insert debug
// values.
std::sort(Orders.begin(), Orders.end(), OrderSorter());
SDDbgInfo::DbgIterator DI = DAG->DbgBegin();
SDDbgInfo::DbgIterator DE = DAG->DbgEnd();
// Now emit the rest according to source order.
unsigned LastOrder = 0;
for (unsigned i = 0, e = Orders.size(); i != e && DI != DE; ++i) {
unsigned Order = Orders[i].first;
MachineInstr *MI = Orders[i].second;
// Insert all SDDbgValue's whose order(s) are before "Order".
if (!MI)
continue;
for (; DI != DE &&
(*DI)->getOrder() >= LastOrder && (*DI)->getOrder() < Order; ++DI) {
if ((*DI)->isInvalidated())
continue;
MachineInstr *DbgMI = Emitter.EmitDbgValue(*DI, VRBaseMap);
if (DbgMI) {
if (!LastOrder)
// Insert to start of the BB (after PHIs).
BB->insert(BBBegin, DbgMI);
else {
// Insert at the instruction, which may be in a different
// block, if the block was split by a custom inserter.
MachineBasicBlock::iterator Pos = MI;
MI->getParent()->insert(llvm::next(Pos), DbgMI);
}
}
}
LastOrder = Order;
}
// Add trailing DbgValue's before the terminator. FIXME: May want to add
// some of them before one or more conditional branches?
SmallVector<MachineInstr*, 8> DbgMIs;
while (DI != DE) {
if (!(*DI)->isInvalidated())
if (MachineInstr *DbgMI = Emitter.EmitDbgValue(*DI, VRBaseMap))
DbgMIs.push_back(DbgMI);
++DI;
}
MachineBasicBlock *InsertBB = Emitter.getBlock();
MachineBasicBlock::iterator Pos = InsertBB->getFirstTerminator();
InsertBB->insert(Pos, DbgMIs.begin(), DbgMIs.end());
}
InsertPos = Emitter.getInsertPos();
return Emitter.getBlock();
}
bool Thumb2ITBlockPass::InsertITInstructions(MachineBasicBlock &MBB) {
bool Modified = false;
SmallSet<unsigned, 4> Defs;
SmallSet<unsigned, 4> Uses;
MachineBasicBlock::iterator MBBI = MBB.begin(), E = MBB.end();
while (MBBI != E) {
MachineInstr *MI = &*MBBI;
DebugLoc dl = MI->getDebugLoc();
unsigned PredReg = 0;
ARMCC::CondCodes CC = llvm::getITInstrPredicate(MI, PredReg);
if (CC == ARMCC::AL) {
++MBBI;
continue;
}
Defs.clear();
Uses.clear();
TrackDefUses(MI, Defs, Uses, TRI);
// Insert an IT instruction.
MachineInstrBuilder MIB = BuildMI(MBB, MBBI, dl, TII->get(ARM::t2IT))
.addImm(CC);
// Add implicit use of ITSTATE to IT block instructions.
MI->addOperand(MachineOperand::CreateReg(ARM::ITSTATE, false/*ifDef*/,
true/*isImp*/, false/*isKill*/));
MachineInstr *LastITMI = MI;
MachineBasicBlock::iterator InsertPos = MIB;
++MBBI;
// Form IT block.
ARMCC::CondCodes OCC = ARMCC::getOppositeCondition(CC);
unsigned Mask = 0, Pos = 3;
// Branches, including tricky ones like LDM_RET, need to end an IT
// block so check the instruction we just put in the block.
for (; MBBI != E && Pos &&
(!MI->isBranch() && !MI->isReturn()) ; ++MBBI) {
if (MBBI->isDebugValue())
continue;
MachineInstr *NMI = &*MBBI;
MI = NMI;
unsigned NPredReg = 0;
ARMCC::CondCodes NCC = llvm::getITInstrPredicate(NMI, NPredReg);
if (NCC == CC || NCC == OCC) {
Mask |= (NCC & 1) << Pos;
// Add implicit use of ITSTATE.
NMI->addOperand(MachineOperand::CreateReg(ARM::ITSTATE, false/*ifDef*/,
true/*isImp*/, false/*isKill*/));
LastITMI = NMI;
} else {
if (NCC == ARMCC::AL &&
MoveCopyOutOfITBlock(NMI, CC, OCC, Defs, Uses)) {
--MBBI;
MBB.remove(NMI);
MBB.insert(InsertPos, NMI);
++NumMovedInsts;
continue;
}
break;
}
TrackDefUses(NMI, Defs, Uses, TRI);
--Pos;
}
// Finalize IT mask.
Mask |= (1 << Pos);
// Tag along (firstcond[0] << 4) with the mask.
Mask |= (CC & 1) << 4;
MIB.addImm(Mask);
// Last instruction in IT block kills ITSTATE.
LastITMI->findRegisterUseOperand(ARM::ITSTATE)->setIsKill();
Modified = true;
++NumITs;
}
return Modified;
}
