MachineInstr *
LanaiInstrInfo::optimizeSelect(MachineInstr &MI,
SmallPtrSetImpl<MachineInstr *> &SeenMIs,
bool PreferFalse) const {
assert(MI.getOpcode() == Lanai::SELECT && "unknown select instruction");
MachineRegisterInfo &MRI = MI.getParent()->getParent()->getRegInfo();
MachineInstr *DefMI = canFoldIntoSelect(MI.getOperand(1).getReg(), MRI, this);
bool Invert = !DefMI;
if (!DefMI)
DefMI = canFoldIntoSelect(MI.getOperand(2).getReg(), MRI, this);
if (!DefMI)
return nullptr;
// Find new register class to use.
MachineOperand FalseReg = MI.getOperand(Invert ? 1 : 2);
unsigned DestReg = MI.getOperand(0).getReg();
const TargetRegisterClass *PreviousClass = MRI.getRegClass(FalseReg.getReg());
if (!MRI.constrainRegClass(DestReg, PreviousClass))
return nullptr;
// Create a new predicated version of DefMI.
MachineInstrBuilder NewMI =
BuildMI(*MI.getParent(), MI, MI.getDebugLoc(), DefMI->getDesc(), DestReg);
// Copy all the DefMI operands, excluding its (null) predicate.
const MCInstrDesc &DefDesc = DefMI->getDesc();
for (unsigned i = 1, e = DefDesc.getNumOperands();
i != e && !DefDesc.OpInfo[i].isPredicate(); ++i)
NewMI.addOperand(DefMI->getOperand(i));
unsigned CondCode = MI.getOperand(3).getImm();
if (Invert)
NewMI.addImm(getOppositeCondition(LPCC::CondCode(CondCode)));
else
NewMI.addImm(CondCode);
NewMI.copyImplicitOps(MI);
// The output register value when the predicate is false is an implicit
// register operand tied to the first def. The tie makes the register
// allocator ensure the FalseReg is allocated the same register as operand 0.
FalseReg.setImplicit();
NewMI.addOperand(FalseReg);
NewMI->tieOperands(0, NewMI->getNumOperands() - 1);
// Update SeenMIs set: register newly created MI and erase removed DefMI.
SeenMIs.insert(NewMI);
SeenMIs.erase(DefMI);
// If MI is inside a loop, and DefMI is outside the loop, then kill flags on
// DefMI would be invalid when transferred inside the loop. Checking for a
// loop is expensive, but at least remove kill flags if they are in different
// BBs.
if (DefMI->getParent() != MI.getParent())
NewMI->clearKillInfo();
// The caller will erase MI, but not DefMI.
DefMI->eraseFromParent();
return NewMI;
}
void HexagonFrameLowering::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = std::prev(MBB.end());
DebugLoc dl = MBBI->getDebugLoc();
//
// Only insert deallocframe if we need to. Also at -O0. See comment
// in emitPrologue above.
//
if (hasFP(MF) || MF.getTarget().getOptLevel() == CodeGenOpt::None) {
MachineBasicBlock::iterator MBBI = std::prev(MBB.end());
MachineBasicBlock::iterator MBBI_end = MBB.end();
const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
// Handle EH_RETURN.
if (MBBI->getOpcode() == Hexagon::EH_RETURN_JMPR) {
assert(MBBI->getOperand(0).isReg() && "Offset should be in register!");
BuildMI(MBB, MBBI, dl, TII.get(Hexagon::L2_deallocframe));
BuildMI(MBB, MBBI, dl, TII.get(Hexagon::A2_add),
Hexagon::R29).addReg(Hexagon::R29).addReg(Hexagon::R28);
return;
}
// Replace 'jumpr r31' instruction with dealloc_return for V4 and higher
// versions.
if (MF.getSubtarget<HexagonSubtarget>().hasV4TOps() &&
MBBI->getOpcode() == Hexagon::JMPret && !DisableDeallocRet) {
// Check for RESTORE_DEALLOC_RET_JMP_V4 call. Don't emit an extra DEALLOC
// instruction if we encounter it.
MachineBasicBlock::iterator BeforeJMPR =
MBB.begin() == MBBI ? MBBI : std::prev(MBBI);
if (BeforeJMPR != MBBI &&
BeforeJMPR->getOpcode() == Hexagon::RESTORE_DEALLOC_RET_JMP_V4) {
// Remove the JMPR node.
MBB.erase(MBBI);
return;
}
// Add dealloc_return.
MachineInstrBuilder MIB =
BuildMI(MBB, MBBI_end, dl, TII.get(Hexagon::L4_return));
// Transfer the function live-out registers.
MIB->copyImplicitOps(*MBB.getParent(), &*MBBI);
// Remove the JUMPR node.
MBB.erase(MBBI);
} else { // Add deallocframe for V2 and V3, and V4 tail calls.
// Check for RESTORE_DEALLOC_BEFORE_TAILCALL_V4. We don't need an extra
// DEALLOCFRAME instruction after it.
MachineBasicBlock::iterator Term = MBB.getFirstTerminator();
MachineBasicBlock::iterator I =
Term == MBB.begin() ? MBB.end() : std::prev(Term);
if (I != MBB.end() &&
I->getOpcode() == Hexagon::RESTORE_DEALLOC_BEFORE_TAILCALL_V4)
return;
BuildMI(MBB, MBBI, dl, TII.get(Hexagon::L2_deallocframe));
}
}
}
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;
}
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;
}
MachineInstrBuilder
MipsInstrInfo::genInstrWithNewOpc(unsigned NewOpc,
MachineBasicBlock::iterator I) const {
MachineInstrBuilder MIB;
// Certain branches have two forms: e.g beq $1, $zero, dst vs beqz $1, dest
// Pick the zero form of the branch for readable assembly and for greater
// branch distance in non-microMIPS mode.
// FIXME: Certain atomic sequences on mips64 generate 32bit references to
// Mips::ZERO, which is incorrect. This test should be updated to use
// Subtarget.getABI().GetZeroReg() when those atomic sequences and others
// are fixed.
bool BranchWithZeroOperand =
(I->isBranch() && !I->isPseudo() && I->getOperand(1).isReg() &&
(I->getOperand(1).getReg() == Mips::ZERO ||
I->getOperand(1).getReg() == Mips::ZERO_64));
if (BranchWithZeroOperand) {
switch (NewOpc) {
case Mips::BEQC:
NewOpc = Mips::BEQZC;
break;
case Mips::BNEC:
NewOpc = Mips::BNEZC;
break;
case Mips::BGEC:
NewOpc = Mips::BGEZC;
break;
case Mips::BLTC:
NewOpc = Mips::BLTZC;
break;
}
}
MIB = BuildMI(*I->getParent(), I, I->getDebugLoc(), get(NewOpc));
// For MIPSR6 JI*C requires an immediate 0 as an operand, JIALC(64) an
// immediate 0 as an operand and requires the removal of it's %RA<imp-def>
// implicit operand as copying the implicit operations of the instructio we're
// looking at will give us the correct flags.
if (NewOpc == Mips::JIC || NewOpc == Mips::JIALC || NewOpc == Mips::JIC64 ||
NewOpc == Mips::JIALC64) {
if (NewOpc == Mips::JIALC || NewOpc == Mips::JIALC64)
MIB->RemoveOperand(0);
for (unsigned J = 0, E = I->getDesc().getNumOperands(); J < E; ++J) {
MIB.addOperand(I->getOperand(J));
}
MIB.addImm(0);
} else if (BranchWithZeroOperand) {
// For MIPSR6 and microMIPS branches with an explicit zero operand, copy
// everything after the zero.
MIB.addOperand(I->getOperand(0));
for (unsigned J = 2, E = I->getDesc().getNumOperands(); J < E; ++J) {
MIB.addOperand(I->getOperand(J));
}
} else {
// All other cases copy all other operands.
for (unsigned J = 0, E = I->getDesc().getNumOperands(); J < E; ++J) {
MIB.addOperand(I->getOperand(J));
}
}
MIB.copyImplicitOps(*I);
MIB.setMemRefs(I->memoperands_begin(), I->memoperands_end());
return MIB;
}
bool Thumb1FrameLowering::emitPopSpecialFixUp(MachineBasicBlock &MBB,
bool DoIt) const {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
const TargetInstrInfo &TII = *STI.getInstrInfo();
const ThumbRegisterInfo *RegInfo =
static_cast<const ThumbRegisterInfo *>(STI.getRegisterInfo());
// If MBBI is a return instruction, we may be able to directly restore
// LR in the PC.
// This is possible if we do not need to emit any SP update.
// Otherwise, we need a temporary register to pop the value
// and copy that value into LR.
auto MBBI = MBB.getFirstTerminator();
if (!ArgRegsSaveSize && MBBI != MBB.end() &&
MBBI->getOpcode() == ARM::tBX_RET) {
if (!DoIt)
return true;
MachineInstrBuilder MIB =
AddDefaultPred(
BuildMI(MBB, MBBI, MBBI->getDebugLoc(), TII.get(ARM::tPOP_RET)))
.addReg(ARM::PC, RegState::Define);
MIB.copyImplicitOps(&*MBBI);
// erase the old tBX_RET instruction
MBB.erase(MBBI);
return true;
}
// Look for a temporary register to use.
// First, compute the liveness information.
LivePhysRegs UsedRegs(STI.getRegisterInfo());
UsedRegs.addLiveOuts(&MBB, /*AddPristines*/ true);
// The semantic of pristines changed recently and now,
// the callee-saved registers that are touched in the function
// are not part of the pristines set anymore.
// Add those callee-saved now.
const TargetRegisterInfo *TRI = STI.getRegisterInfo();
const MCPhysReg *CSRegs = TRI->getCalleeSavedRegs(&MF);
for (unsigned i = 0; CSRegs[i]; ++i)
UsedRegs.addReg(CSRegs[i]);
DebugLoc dl = DebugLoc();
if (MBBI != MBB.end()) {
dl = MBBI->getDebugLoc();
auto InstUpToMBBI = MBB.end();
// The post-decrement is on purpose here.
// We want to have the liveness right before MBBI.
while (InstUpToMBBI-- != MBBI)
UsedRegs.stepBackward(*InstUpToMBBI);
}
// Look for a register that can be directly use in the POP.
unsigned PopReg = 0;
// And some temporary register, just in case.
unsigned TemporaryReg = 0;
BitVector PopFriendly =
TRI->getAllocatableSet(MF, TRI->getRegClass(ARM::tGPRRegClassID));
assert(PopFriendly.any() && "No allocatable pop-friendly register?!");
// Rebuild the GPRs from the high registers because they are removed
// form the GPR reg class for thumb1.
BitVector GPRsNoLRSP =
TRI->getAllocatableSet(MF, TRI->getRegClass(ARM::hGPRRegClassID));
GPRsNoLRSP |= PopFriendly;
GPRsNoLRSP.reset(ARM::LR);
GPRsNoLRSP.reset(ARM::SP);
GPRsNoLRSP.reset(ARM::PC);
for (int Register = GPRsNoLRSP.find_first(); Register != -1;
Register = GPRsNoLRSP.find_next(Register)) {
if (!UsedRegs.contains(Register)) {
// Remember the first pop-friendly register and exit.
if (PopFriendly.test(Register)) {
PopReg = Register;
TemporaryReg = 0;
break;
}
// Otherwise, remember that the register will be available to
// save a pop-friendly register.
TemporaryReg = Register;
}
}
if (!DoIt && !PopReg && !TemporaryReg)
return false;
assert((PopReg || TemporaryReg) && "Cannot get LR");
if (TemporaryReg) {
assert(!PopReg && "Unnecessary MOV is about to be inserted");
PopReg = PopFriendly.find_first();
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr))
.addReg(TemporaryReg, RegState::Define)
.addReg(PopReg, RegState::Kill));
}
assert(PopReg && "Do not know how to get LR");
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tPOP)))
.addReg(PopReg, RegState::Define);
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, ArgRegsSaveSize);
if (!TemporaryReg && MBBI != MBB.end() && MBBI->getOpcode() == ARM::tBX_RET) {
MachineInstrBuilder MIB = BuildMI(MBB, MBBI, dl, TII.get(ARM::tBX))
.addReg(PopReg, RegState::Kill);
AddDefaultPred(MIB);
MIB.copyImplicitOps(&*MBBI);
// erase the old tBX_RET instruction
MBB.erase(MBBI);
return true;
}
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr))
.addReg(ARM::LR, RegState::Define)
.addReg(PopReg, RegState::Kill));
if (TemporaryReg) {
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr))
.addReg(PopReg, RegState::Define)
.addReg(TemporaryReg, RegState::Kill));
}
return true;
}
MachineInstrBuilder
MipsInstrInfo::genInstrWithNewOpc(unsigned NewOpc,
MachineBasicBlock::iterator I) const {
MachineInstrBuilder MIB;
// Certain branches have two forms: e.g beq $1, $zero, dest vs beqz $1, dest
// Pick the zero form of the branch for readable assembly and for greater
// branch distance in non-microMIPS mode.
// Additional MIPSR6 does not permit the use of register $zero for compact
// branches.
// FIXME: Certain atomic sequences on mips64 generate 32bit references to
// Mips::ZERO, which is incorrect. This test should be updated to use
// Subtarget.getABI().GetZeroReg() when those atomic sequences and others
// are fixed.
int ZeroOperandPosition = -1;
bool BranchWithZeroOperand = false;
if (I->isBranch() && !I->isPseudo()) {
auto TRI = I->getParent()->getParent()->getSubtarget().getRegisterInfo();
ZeroOperandPosition = I->findRegisterUseOperandIdx(Mips::ZERO, false, TRI);
BranchWithZeroOperand = ZeroOperandPosition != -1;
}
if (BranchWithZeroOperand) {
switch (NewOpc) {
case Mips::BEQC:
NewOpc = Mips::BEQZC;
break;
case Mips::BNEC:
NewOpc = Mips::BNEZC;
break;
case Mips::BGEC:
NewOpc = Mips::BGEZC;
break;
case Mips::BLTC:
NewOpc = Mips::BLTZC;
break;
case Mips::BEQC64:
NewOpc = Mips::BEQZC64;
break;
case Mips::BNEC64:
NewOpc = Mips::BNEZC64;
break;
}
}
MIB = BuildMI(*I->getParent(), I, I->getDebugLoc(), get(NewOpc));
// For MIPSR6 JI*C requires an immediate 0 as an operand, JIALC(64) an
// immediate 0 as an operand and requires the removal of it's %RA<imp-def>
// implicit operand as copying the implicit operations of the instructio we're
// looking at will give us the correct flags.
if (NewOpc == Mips::JIC || NewOpc == Mips::JIALC || NewOpc == Mips::JIC64 ||
NewOpc == Mips::JIALC64) {
if (NewOpc == Mips::JIALC || NewOpc == Mips::JIALC64)
MIB->RemoveOperand(0);
for (unsigned J = 0, E = I->getDesc().getNumOperands(); J < E; ++J) {
MIB.add(I->getOperand(J));
}
MIB.addImm(0);
} else {
for (unsigned J = 0, E = I->getDesc().getNumOperands(); J < E; ++J) {
if (BranchWithZeroOperand && (unsigned)ZeroOperandPosition == J)
continue;
MIB.add(I->getOperand(J));
}
}
MIB.copyImplicitOps(*I);
MIB.setMemRefs(I->memoperands_begin(), I->memoperands_end());
return MIB;
}
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;
}
void Thumb1FrameLowering::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
assert((MBBI->getOpcode() == ARM::tBX_RET ||
MBBI->getOpcode() == ARM::tPOP_RET) &&
"Can only insert epilog into returning blocks");
DebugLoc dl = MBBI->getDebugLoc();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const ThumbRegisterInfo *RegInfo =
static_cast<const ThumbRegisterInfo *>(STI.getRegisterInfo());
const Thumb1InstrInfo &TII =
*static_cast<const Thumb1InstrInfo *>(STI.getInstrInfo());
unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
int NumBytes = (int)MFI->getStackSize();
assert((unsigned)NumBytes >= ArgRegsSaveSize &&
"ArgRegsSaveSize is included in NumBytes");
const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
unsigned FramePtr = RegInfo->getFrameRegister(MF);
if (!AFI->hasStackFrame()) {
if (NumBytes - ArgRegsSaveSize != 0)
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, NumBytes - ArgRegsSaveSize);
} else {
// Unwind MBBI to point to first LDR / VLDRD.
if (MBBI != MBB.begin()) {
do
--MBBI;
while (MBBI != MBB.begin() && isCSRestore(MBBI, CSRegs));
if (!isCSRestore(MBBI, CSRegs))
++MBBI;
}
// Move SP to start of FP callee save spill area.
NumBytes -= (AFI->getGPRCalleeSavedArea1Size() +
AFI->getGPRCalleeSavedArea2Size() +
AFI->getDPRCalleeSavedAreaSize() +
ArgRegsSaveSize);
if (AFI->shouldRestoreSPFromFP()) {
NumBytes = AFI->getFramePtrSpillOffset() - NumBytes;
// Reset SP based on frame pointer only if the stack frame extends beyond
// frame pointer stack slot, the target is ELF and the function has FP, or
// the target uses var sized objects.
if (NumBytes) {
assert(MF.getRegInfo().isPhysRegUsed(ARM::R4) &&
"No scratch register to restore SP from FP!");
emitThumbRegPlusImmediate(MBB, MBBI, dl, ARM::R4, FramePtr, -NumBytes,
TII, *RegInfo);
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
ARM::SP)
.addReg(ARM::R4));
} else
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
ARM::SP)
.addReg(FramePtr));
} else {
if (MBBI->getOpcode() == ARM::tBX_RET &&
&MBB.front() != MBBI &&
std::prev(MBBI)->getOpcode() == ARM::tPOP) {
MachineBasicBlock::iterator PMBBI = std::prev(MBBI);
if (!tryFoldSPUpdateIntoPushPop(STI, MF, PMBBI, NumBytes))
emitSPUpdate(MBB, PMBBI, TII, dl, *RegInfo, NumBytes);
} else if (!tryFoldSPUpdateIntoPushPop(STI, MF, MBBI, NumBytes))
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, NumBytes);
}
}
bool IsV4PopReturn = false;
for (const CalleeSavedInfo &CSI : MFI->getCalleeSavedInfo())
if (CSI.getReg() == ARM::LR)
IsV4PopReturn = true;
IsV4PopReturn &= STI.hasV4TOps() && !STI.hasV5TOps();
// Unlike T2 and ARM mode, the T1 pop instruction cannot restore
// to LR, and we can't pop the value directly to the PC since
// we need to update the SP after popping the value. So instead
// we have to emit:
// POP {r3}
// ADD sp, #offset
// BX r3
// If this would clobber a return value, then generate this sequence instead:
// MOV ip, r3
// POP {r3}
// ADD sp, #offset
// MOV lr, r3
// MOV r3, ip
// BX lr
if (ArgRegsSaveSize || IsV4PopReturn) {
// Get the last instruction, tBX_RET
MBBI = MBB.getLastNonDebugInstr();
assert (MBBI->getOpcode() == ARM::tBX_RET);
DebugLoc dl = MBBI->getDebugLoc();
if (AFI->getReturnRegsCount() <= 3) {
// Epilogue: pop saved LR to R3 and branch off it.
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tPOP)))
.addReg(ARM::R3, RegState::Define);
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, ArgRegsSaveSize);
MachineInstrBuilder MIB =
BuildMI(MBB, MBBI, dl, TII.get(ARM::tBX))
.addReg(ARM::R3, RegState::Kill);
AddDefaultPred(MIB);
MIB.copyImplicitOps(&*MBBI);
// erase the old tBX_RET instruction
MBB.erase(MBBI);
} else {
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr))
.addReg(ARM::R12, RegState::Define)
.addReg(ARM::R3, RegState::Kill));
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tPOP)))
.addReg(ARM::R3, RegState::Define);
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, ArgRegsSaveSize);
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr))
.addReg(ARM::LR, RegState::Define)
.addReg(ARM::R3, RegState::Kill));
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr))
.addReg(ARM::R3, RegState::Define)
.addReg(ARM::R12, RegState::Kill));
// Keep the tBX_RET instruction
}
}
}
void Thumb1FrameLowering::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
assert((MBBI->getOpcode() == ARM::tBX_RET ||
MBBI->getOpcode() == ARM::tPOP_RET) &&
"Can only insert epilog into returning blocks");
DebugLoc dl = MBBI->getDebugLoc();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const Thumb1RegisterInfo *RegInfo =
static_cast<const Thumb1RegisterInfo*>(MF.getTarget().getRegisterInfo());
const Thumb1InstrInfo &TII =
*static_cast<const Thumb1InstrInfo*>(MF.getTarget().getInstrInfo());
unsigned Align = MF.getTarget().getFrameLowering()->getStackAlignment();
unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize(Align);
int NumBytes = (int)MFI->getStackSize();
const uint16_t *CSRegs = RegInfo->getCalleeSavedRegs();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
if (!AFI->hasStackFrame()) {
if (NumBytes != 0)
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, NumBytes);
} else {
// Unwind MBBI to point to first LDR / VLDRD.
if (MBBI != MBB.begin()) {
do
--MBBI;
while (MBBI != MBB.begin() && isCSRestore(MBBI, CSRegs));
if (!isCSRestore(MBBI, CSRegs))
++MBBI;
}
// Move SP to start of FP callee save spill area.
NumBytes -= (AFI->getGPRCalleeSavedArea1Size() +
AFI->getGPRCalleeSavedArea2Size() +
AFI->getDPRCalleeSavedAreaSize());
if (AFI->shouldRestoreSPFromFP()) {
NumBytes = AFI->getFramePtrSpillOffset() - NumBytes;
// Reset SP based on frame pointer only if the stack frame extends beyond
// frame pointer stack slot, the target is ELF and the function has FP, or
// the target uses var sized objects.
if (NumBytes) {
assert(MF.getRegInfo().isPhysRegUsed(ARM::R4) &&
"No scratch register to restore SP from FP!");
emitThumbRegPlusImmediate(MBB, MBBI, dl, ARM::R4, FramePtr, -NumBytes,
TII, *RegInfo);
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
ARM::SP)
.addReg(ARM::R4));
} else
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
ARM::SP)
.addReg(FramePtr));
} else {
if (MBBI->getOpcode() == ARM::tBX_RET &&
&MBB.front() != MBBI &&
std::prev(MBBI)->getOpcode() == ARM::tPOP) {
MachineBasicBlock::iterator PMBBI = std::prev(MBBI);
if (!tryFoldSPUpdateIntoPushPop(STI, MF, PMBBI, NumBytes))
emitSPUpdate(MBB, PMBBI, TII, dl, *RegInfo, NumBytes);
} else if (!tryFoldSPUpdateIntoPushPop(STI, MF, MBBI, NumBytes))
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, NumBytes);
}
}
if (ArgRegsSaveSize) {
// Unlike T2 and ARM mode, the T1 pop instruction cannot restore
// to LR, and we can't pop the value directly to the PC since
// we need to update the SP after popping the value. Therefore, we
// pop the old LR into R3 as a temporary.
// Get the last instruction, tBX_RET
MBBI = MBB.getLastNonDebugInstr();
assert (MBBI->getOpcode() == ARM::tBX_RET);
// Epilogue for vararg functions: pop LR to R3 and branch off it.
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tPOP)))
.addReg(ARM::R3, RegState::Define);
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, ArgRegsSaveSize);
MachineInstrBuilder MIB =
BuildMI(MBB, MBBI, dl, TII.get(ARM::tBX_RET_vararg))
.addReg(ARM::R3, RegState::Kill);
AddDefaultPred(MIB);
MIB.copyImplicitOps(&*MBBI);
// erase the old tBX_RET instruction
MBB.erase(MBBI);
}
}
void ARMFrameLowering::emitPopInst(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI,
unsigned LdmOpc, unsigned LdrOpc,
bool isVarArg, bool NoGap,
bool(*Func)(unsigned, bool)) const {
MachineFunction &MF = *MBB.getParent();
const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
DebugLoc DL = MI->getDebugLoc();
unsigned RetOpcode = MI->getOpcode();
bool isTailCall = (RetOpcode == ARM::TCRETURNdi ||
RetOpcode == ARM::TCRETURNdiND ||
RetOpcode == ARM::TCRETURNri ||
RetOpcode == ARM::TCRETURNriND);
SmallVector<unsigned, 4> Regs;
unsigned i = CSI.size();
while (i != 0) {
unsigned LastReg = 0;
bool DeleteRet = false;
for (; i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
if (!(Func)(Reg, STI.isTargetDarwin())) continue;
if (Reg == ARM::LR && !isTailCall && !isVarArg && STI.hasV5TOps()) {
Reg = ARM::PC;
LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
// Fold the return instruction into the LDM.
DeleteRet = true;
}
// If NoGap is true, pop consecutive registers and then leave the rest
// for other instructions. e.g.
// vpop {d8, d10, d11} -> vpop {d8}, vpop {d10, d11}
if (NoGap && LastReg && LastReg != Reg-1)
break;
LastReg = Reg;
Regs.push_back(Reg);
}
if (Regs.empty())
continue;
if (Regs.size() > 1 || LdrOpc == 0) {
MachineInstrBuilder MIB =
AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(LdmOpc), ARM::SP)
.addReg(ARM::SP));
for (unsigned i = 0, e = Regs.size(); i < e; ++i)
MIB.addReg(Regs[i], getDefRegState(true));
if (DeleteRet) {
MIB->copyImplicitOps(&*MI);
MI->eraseFromParent();
}
MI = MIB;
} else if (Regs.size() == 1) {
// If we adjusted the reg to PC from LR above, switch it back here. We
// only do that for LDM.
if (Regs[0] == ARM::PC)
Regs[0] = ARM::LR;
MachineInstrBuilder MIB =
BuildMI(MBB, MI, DL, TII.get(LdrOpc), Regs[0])
.addReg(ARM::SP, RegState::Define)
.addReg(ARM::SP);
// ARM mode needs an extra reg0 here due to addrmode2. Will go away once
// that refactoring is complete (eventually).
if (LdrOpc == ARM::LDR_POST_REG || LdrOpc == ARM::LDR_POST_IMM) {
MIB.addReg(0);
MIB.addImm(ARM_AM::getAM2Opc(ARM_AM::add, 4, ARM_AM::no_shift));
} else
MIB.addImm(4);
AddDefaultPred(MIB);
}
Regs.clear();
}
}
