bool MIRParserImpl::initializeFrameInfo(
const Function &F, MachineFrameInfo &MFI,
const yaml::MachineFunction &YamlMF,
DenseMap<unsigned, int> &StackObjectSlots,
DenseMap<unsigned, int> &FixedStackObjectSlots) {
const yaml::MachineFrameInfo &YamlMFI = YamlMF.FrameInfo;
MFI.setFrameAddressIsTaken(YamlMFI.IsFrameAddressTaken);
MFI.setReturnAddressIsTaken(YamlMFI.IsReturnAddressTaken);
MFI.setHasStackMap(YamlMFI.HasStackMap);
MFI.setHasPatchPoint(YamlMFI.HasPatchPoint);
MFI.setStackSize(YamlMFI.StackSize);
MFI.setOffsetAdjustment(YamlMFI.OffsetAdjustment);
if (YamlMFI.MaxAlignment)
MFI.ensureMaxAlignment(YamlMFI.MaxAlignment);
MFI.setAdjustsStack(YamlMFI.AdjustsStack);
MFI.setHasCalls(YamlMFI.HasCalls);
MFI.setMaxCallFrameSize(YamlMFI.MaxCallFrameSize);
MFI.setHasOpaqueSPAdjustment(YamlMFI.HasOpaqueSPAdjustment);
MFI.setHasVAStart(YamlMFI.HasVAStart);
MFI.setHasMustTailInVarArgFunc(YamlMFI.HasMustTailInVarArgFunc);
// Initialize the fixed frame objects.
for (const auto &Object : YamlMF.FixedStackObjects) {
int ObjectIdx;
if (Object.Type != yaml::FixedMachineStackObject::SpillSlot)
ObjectIdx = MFI.CreateFixedObject(Object.Size, Object.Offset,
Object.IsImmutable, Object.IsAliased);
else
ObjectIdx = MFI.CreateFixedSpillStackObject(Object.Size, Object.Offset);
MFI.setObjectAlignment(ObjectIdx, Object.Alignment);
// TODO: Report an error when objects are redefined.
FixedStackObjectSlots.insert(std::make_pair(Object.ID, ObjectIdx));
}
// Initialize the ordinary frame objects.
for (const auto &Object : YamlMF.StackObjects) {
int ObjectIdx;
const AllocaInst *Alloca = nullptr;
const yaml::StringValue &Name = Object.Name;
if (!Name.Value.empty()) {
Alloca = dyn_cast_or_null<AllocaInst>(
F.getValueSymbolTable().lookup(Name.Value));
if (!Alloca)
return error(Name.SourceRange.Start,
"alloca instruction named '" + Name.Value +
"' isn't defined in the function '" + F.getName() +
"'");
}
if (Object.Type == yaml::MachineStackObject::VariableSized)
ObjectIdx = MFI.CreateVariableSizedObject(Object.Alignment, Alloca);
else
ObjectIdx = MFI.CreateStackObject(
Object.Size, Object.Alignment,
Object.Type == yaml::MachineStackObject::SpillSlot, Alloca);
MFI.setObjectOffset(ObjectIdx, Object.Offset);
// TODO: Report an error when objects are redefined.
StackObjectSlots.insert(std::make_pair(Object.ID, ObjectIdx));
}
return false;
}
void AArch64FrameLowering::emitPrologue(MachineFunction &MF) const {
AArch64MachineFunctionInfo *FuncInfo =
MF.getInfo<AArch64MachineFunctionInfo>();
MachineBasicBlock &MBB = MF.front();
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
DebugLoc DL = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
MachineModuleInfo &MMI = MF.getMMI();
const MCRegisterInfo *MRI = MMI.getContext().getRegisterInfo();
bool NeedsFrameMoves = MMI.hasDebugInfo()
|| MF.getFunction()->needsUnwindTableEntry();
uint64_t NumInitialBytes, NumResidualBytes;
// Currently we expect the stack to be laid out by
// sub sp, sp, #initial
// stp x29, x30, [sp, #offset]
// ...
// str xxx, [sp, #offset]
// sub sp, sp, #rest (possibly via extra instructions).
if (MFI->getCalleeSavedInfo().size()) {
// If there are callee-saved registers, we want to store them efficiently as
// a block, and virtual base assignment happens too early to do it for us so
// we adjust the stack in two phases: first just for callee-saved fiddling,
// then to allocate the rest of the frame.
splitSPAdjustments(MFI->getStackSize(), NumInitialBytes, NumResidualBytes);
} else {
// If there aren't any callee-saved registers, two-phase adjustment is
// inefficient. It's more efficient to adjust with NumInitialBytes too
// because when we're in a "callee pops argument space" situation, that pop
// must be tacked onto Initial for correctness.
NumInitialBytes = MFI->getStackSize();
NumResidualBytes = 0;
}
// Tell everyone else how much adjustment we're expecting them to use. In
// particular if an adjustment is required for a tail call the epilogue could
// have a different view of things.
FuncInfo->setInitialStackAdjust(NumInitialBytes);
emitSPUpdate(MBB, MBBI, DL, TII, AArch64::X16, -NumInitialBytes,
MachineInstr::FrameSetup);
if (NeedsFrameMoves && NumInitialBytes) {
// We emit this update even if the CFA is set from a frame pointer later so
// that the CFA is valid in the interim.
MachineLocation Dst(MachineLocation::VirtualFP);
unsigned Reg = MRI->getDwarfRegNum(AArch64::XSP, true);
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfa(nullptr, Reg, -NumInitialBytes));
BuildMI(MBB, MBBI, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex);
}
// Otherwise we need to set the frame pointer and/or add a second stack
// adjustment.
bool FPNeedsSetting = hasFP(MF);
for (; MBBI != MBB.end(); ++MBBI) {
// Note that this search makes strong assumptions about the operation used
// to store the frame-pointer: it must be "STP x29, x30, ...". This could
// change in future, but until then there's no point in implementing
// untestable more generic cases.
if (FPNeedsSetting && MBBI->getOpcode() == AArch64::LSPair64_STR
&& MBBI->getOperand(0).getReg() == AArch64::X29) {
int64_t X29FrameIdx = MBBI->getOperand(2).getIndex();
FuncInfo->setFramePointerOffset(MFI->getObjectOffset(X29FrameIdx));
++MBBI;
emitRegUpdate(MBB, MBBI, DL, TII, AArch64::X29, AArch64::XSP,
AArch64::X29,
NumInitialBytes + MFI->getObjectOffset(X29FrameIdx),
MachineInstr::FrameSetup);
// The offset adjustment used when emitting debugging locations relative
// to whatever frame base is set. AArch64 uses the default frame base (FP
// or SP) and this adjusts the calculations to be correct.
MFI->setOffsetAdjustment(- MFI->getObjectOffset(X29FrameIdx)
- MFI->getStackSize());
if (NeedsFrameMoves) {
unsigned Reg = MRI->getDwarfRegNum(AArch64::X29, true);
unsigned Offset = MFI->getObjectOffset(X29FrameIdx);
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfa(nullptr, Reg, Offset));
BuildMI(MBB, MBBI, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex);
}
FPNeedsSetting = false;
}
if (!MBBI->getFlag(MachineInstr::FrameSetup))
break;
}
assert(!FPNeedsSetting && "Frame pointer couldn't be set");
emitSPUpdate(MBB, MBBI, DL, TII, AArch64::X16, -NumResidualBytes,
MachineInstr::FrameSetup);
// Now we emit the rest of the frame setup information, if necessary: we've
// already noted the FP and initial SP moves so we're left with the prologue's
// final SP update and callee-saved register locations.
if (!NeedsFrameMoves)
return;
// The rest of the stack adjustment
if (!hasFP(MF) && NumResidualBytes) {
MachineLocation Dst(MachineLocation::VirtualFP);
unsigned Reg = MRI->getDwarfRegNum(AArch64::XSP, true);
unsigned Offset = NumResidualBytes + NumInitialBytes;
unsigned CFIIndex =
MMI.addFrameInst(MCCFIInstruction::createDefCfa(nullptr, Reg, -Offset));
BuildMI(MBB, MBBI, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex);
}
// And any callee-saved registers (it's fine to leave them to the end here,
// because the old values are still valid at this point.
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
if (CSI.size()) {
for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
E = CSI.end(); I != E; ++I) {
unsigned Offset = MFI->getObjectOffset(I->getFrameIdx());
unsigned Reg = MRI->getDwarfRegNum(I->getReg(), true);
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createOffset(nullptr, Reg, Offset));
BuildMI(MBB, MBBI, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex);
}
}
}
void MSP430RegisterInfo::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB
MachineFrameInfo *MFI = MF.getFrameInfo();
MSP430MachineFunctionInfo *MSP430FI = MF.getInfo<MSP430MachineFunctionInfo>();
MachineBasicBlock::iterator MBBI = MBB.begin();
DebugLoc DL = (MBBI != MBB.end() ? MBBI->getDebugLoc() :
DebugLoc::getUnknownLoc());
// Get the number of bytes to allocate from the FrameInfo.
uint64_t StackSize = MFI->getStackSize();
uint64_t NumBytes = 0;
if (hasFP(MF)) {
// Calculate required stack adjustment
uint64_t FrameSize = StackSize - 2;
NumBytes = FrameSize - MSP430FI->getCalleeSavedFrameSize();
// Get the offset of the stack slot for the EBP register... which is
// guaranteed to be the last slot by processFunctionBeforeFrameFinalized.
// Update the frame offset adjustment.
MFI->setOffsetAdjustment(-NumBytes);
// Save FPW into the appropriate stack slot...
BuildMI(MBB, MBBI, DL, TII.get(MSP430::PUSH16r))
.addReg(MSP430::FPW, RegState::Kill);
// Update FPW with the new base value...
BuildMI(MBB, MBBI, DL, TII.get(MSP430::MOV16rr), MSP430::FPW)
.addReg(MSP430::SPW);
// Mark the FramePtr as live-in in every block except the entry.
for (MachineFunction::iterator I = next(MF.begin()), E = MF.end();
I != E; ++I)
I->addLiveIn(MSP430::FPW);
} else
NumBytes = StackSize - MSP430FI->getCalleeSavedFrameSize();
// Skip the callee-saved push instructions.
while (MBBI != MBB.end() && (MBBI->getOpcode() == MSP430::PUSH16r))
++MBBI;
if (MBBI != MBB.end())
DL = MBBI->getDebugLoc();
if (NumBytes) { // adjust stack pointer: SPW -= numbytes
// If there is an SUB16ri of SPW immediately before this instruction, merge
// the two.
//NumBytes -= mergeSPUpdates(MBB, MBBI, true);
// If there is an ADD16ri or SUB16ri of SPW immediately after this
// instruction, merge the two instructions.
// mergeSPUpdatesDown(MBB, MBBI, &NumBytes);
if (NumBytes) {
MachineInstr *MI =
BuildMI(MBB, MBBI, DL, TII.get(MSP430::SUB16ri), MSP430::SPW)
.addReg(MSP430::SPW).addImm(NumBytes);
// The SRW implicit def is dead.
MI->getOperand(3).setIsDead();
}
}
}
void Thumb1FrameLowering::emitPrologue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
MachineModuleInfo &MMI = MF.getMMI();
const MCRegisterInfo *MRI = MMI.getContext().getRegisterInfo();
const ThumbRegisterInfo *RegInfo =
static_cast<const ThumbRegisterInfo *>(STI.getRegisterInfo());
const Thumb1InstrInfo &TII =
*static_cast<const Thumb1InstrInfo *>(STI.getInstrInfo());
unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
unsigned NumBytes = MFI->getStackSize();
assert(NumBytes >= ArgRegsSaveSize &&
"ArgRegsSaveSize is included in NumBytes");
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
// Debug location must be unknown since the first debug location is used
// to determine the end of the prologue.
DebugLoc dl;
unsigned FramePtr = RegInfo->getFrameRegister(MF);
unsigned BasePtr = RegInfo->getBaseRegister();
int CFAOffset = 0;
// Thumb add/sub sp, imm8 instructions implicitly multiply the offset by 4.
NumBytes = (NumBytes + 3) & ~3;
MFI->setStackSize(NumBytes);
// Determine the sizes of each callee-save spill areas and record which frame
// belongs to which callee-save spill areas.
unsigned GPRCS1Size = 0, GPRCS2Size = 0, DPRCSSize = 0;
int FramePtrSpillFI = 0;
if (ArgRegsSaveSize) {
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, -ArgRegsSaveSize,
MachineInstr::FrameSetup);
CFAOffset -= ArgRegsSaveSize;
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
}
if (!AFI->hasStackFrame()) {
if (NumBytes - ArgRegsSaveSize != 0) {
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, -(NumBytes - ArgRegsSaveSize),
MachineInstr::FrameSetup);
CFAOffset -= NumBytes - ArgRegsSaveSize;
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
}
return;
}
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
int FI = CSI[i].getFrameIdx();
switch (Reg) {
case ARM::R8:
case ARM::R9:
case ARM::R10:
case ARM::R11:
if (STI.isTargetMachO()) {
GPRCS2Size += 4;
break;
}
// fallthrough
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
GPRCS1Size += 4;
break;
default:
DPRCSSize += 8;
}
}
if (MBBI != MBB.end() && MBBI->getOpcode() == ARM::tPUSH) {
++MBBI;
}
// Determine starting offsets of spill areas.
unsigned DPRCSOffset = NumBytes - ArgRegsSaveSize - (GPRCS1Size + GPRCS2Size + DPRCSSize);
unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
bool HasFP = hasFP(MF);
if (HasFP)
AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) +
NumBytes);
AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
NumBytes = DPRCSOffset;
int FramePtrOffsetInBlock = 0;
unsigned adjustedGPRCS1Size = GPRCS1Size;
if (tryFoldSPUpdateIntoPushPop(STI, MF, std::prev(MBBI), NumBytes)) {
FramePtrOffsetInBlock = NumBytes;
adjustedGPRCS1Size += NumBytes;
NumBytes = 0;
}
if (adjustedGPRCS1Size) {
CFAOffset -= adjustedGPRCS1Size;
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
}
for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(),
E = CSI.end(); I != E; ++I) {
unsigned Reg = I->getReg();
int FI = I->getFrameIdx();
switch (Reg) {
case ARM::R8:
case ARM::R9:
case ARM::R10:
case ARM::R11:
case ARM::R12:
if (STI.isTargetMachO())
break;
// fallthough
case ARM::R0:
case ARM::R1:
case ARM::R2:
case ARM::R3:
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
nullptr, MRI->getDwarfRegNum(Reg, true), MFI->getObjectOffset(FI)));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
break;
}
}
// Adjust FP so it point to the stack slot that contains the previous FP.
if (HasFP) {
FramePtrOffsetInBlock +=
MFI->getObjectOffset(FramePtrSpillFI) + GPRCS1Size + ArgRegsSaveSize;
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tADDrSPi), FramePtr)
.addReg(ARM::SP).addImm(FramePtrOffsetInBlock / 4)
.setMIFlags(MachineInstr::FrameSetup));
if(FramePtrOffsetInBlock) {
CFAOffset += FramePtrOffsetInBlock;
unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createDefCfa(
nullptr, MRI->getDwarfRegNum(FramePtr, true), CFAOffset));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
} else {
unsigned CFIIndex =
MMI.addFrameInst(MCCFIInstruction::createDefCfaRegister(
nullptr, MRI->getDwarfRegNum(FramePtr, true)));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
}
if (NumBytes > 508)
// If offset is > 508 then sp cannot be adjusted in a single instruction,
// try restoring from fp instead.
AFI->setShouldRestoreSPFromFP(true);
}
if (NumBytes) {
// Insert it after all the callee-save spills.
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, -NumBytes,
MachineInstr::FrameSetup);
if (!HasFP) {
CFAOffset -= NumBytes;
unsigned CFIIndex = MMI.addFrameInst(
MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
BuildMI(MBB, MBBI, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
.addCFIIndex(CFIIndex)
.setMIFlags(MachineInstr::FrameSetup);
}
}
if (STI.isTargetELF() && HasFP)
MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
AFI->getFramePtrSpillOffset());
AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
// Thumb1 does not currently support dynamic stack realignment. Report a
// fatal error rather then silently generate bad code.
if (RegInfo->needsStackRealignment(MF))
report_fatal_error("Dynamic stack realignment not supported for thumb1.");
// If we need a base pointer, set it up here. It's whatever the value
// of the stack pointer is at this point. Any variable size objects
// will be allocated after this, so we can still use the base pointer
// to reference locals.
if (RegInfo->hasBasePointer(MF))
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), BasePtr)
.addReg(ARM::SP));
// If the frame has variable sized objects then the epilogue must restore
// the sp from fp. We can assume there's an FP here since hasFP already
// checks for hasVarSizedObjects.
if (MFI->hasVarSizedObjects())
AFI->setShouldRestoreSPFromFP(true);
}
void Thumb1FrameLowering::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front();
MachineBasicBlock::iterator MBBI = MBB.begin();
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 VARegSaveSize = AFI->getVarArgsRegSaveSize();
unsigned NumBytes = MFI->getStackSize();
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
DebugLoc dl = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
unsigned BasePtr = RegInfo->getBaseRegister();
// Thumb add/sub sp, imm8 instructions implicitly multiply the offset by 4.
NumBytes = (NumBytes + 3) & ~3;
MFI->setStackSize(NumBytes);
// Determine the sizes of each callee-save spill areas and record which frame
// belongs to which callee-save spill areas.
unsigned GPRCS1Size = 0, GPRCS2Size = 0, DPRCSSize = 0;
int FramePtrSpillFI = 0;
if (VARegSaveSize)
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, -VARegSaveSize,
MachineInstr::FrameSetup);
if (!AFI->hasStackFrame()) {
if (NumBytes != 0)
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, -NumBytes,
MachineInstr::FrameSetup);
return;
}
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
int FI = CSI[i].getFrameIdx();
switch (Reg) {
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
break;
case ARM::R8:
case ARM::R9:
case ARM::R10:
case ARM::R11:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
if (STI.isTargetDarwin()) {
AFI->addGPRCalleeSavedArea2Frame(FI);
GPRCS2Size += 4;
} else {
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
}
break;
default:
AFI->addDPRCalleeSavedAreaFrame(FI);
DPRCSSize += 8;
}
}
if (MBBI != MBB.end() && MBBI->getOpcode() == ARM::tPUSH) {
++MBBI;
if (MBBI != MBB.end())
dl = MBBI->getDebugLoc();
}
// Determine starting offsets of spill areas.
unsigned DPRCSOffset = NumBytes - (GPRCS1Size + GPRCS2Size + DPRCSSize);
unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) + NumBytes);
AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
NumBytes = DPRCSOffset;
// Adjust FP so it point to the stack slot that contains the previous FP.
if (hasFP(MF)) {
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tADDrSPi), FramePtr)
.addFrameIndex(FramePtrSpillFI).addImm(0)
.setMIFlags(MachineInstr::FrameSetup));
if (NumBytes > 508)
// If offset is > 508 then sp cannot be adjusted in a single instruction,
// try restoring from fp instead.
AFI->setShouldRestoreSPFromFP(true);
}
if (NumBytes)
// Insert it after all the callee-save spills.
emitSPUpdate(MBB, MBBI, TII, dl, *RegInfo, -NumBytes,
MachineInstr::FrameSetup);
if (STI.isTargetELF() && hasFP(MF))
MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
AFI->getFramePtrSpillOffset());
AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
// If we need a base pointer, set it up here. It's whatever the value
// of the stack pointer is at this point. Any variable size objects
// will be allocated after this, so we can still use the base pointer
// to reference locals.
if (RegInfo->hasBasePointer(MF))
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), BasePtr)
.addReg(ARM::SP));
// If the frame has variable sized objects then the epilogue must restore
// the sp from fp. We can assume there's an FP here since hasFP already
// checks for hasVarSizedObjects.
if (MFI->hasVarSizedObjects())
AFI->setShouldRestoreSPFromFP(true);
}
void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front();
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
const ARMBaseRegisterInfo *RegInfo =
static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
const ARMBaseInstrInfo &TII =
*static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
assert(!AFI->isThumb1OnlyFunction() &&
"This emitPrologue does not support Thumb1!");
bool isARM = !AFI->isThumbFunction();
unsigned VARegSaveSize = AFI->getVarArgsRegSaveSize();
unsigned NumBytes = MFI->getStackSize();
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
DebugLoc dl = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
// Determine the sizes of each callee-save spill areas and record which frame
// belongs to which callee-save spill areas.
unsigned GPRCS1Size = 0, GPRCS2Size = 0, DPRCSSize = 0;
int FramePtrSpillFI = 0;
// Allocate the vararg register save area. This is not counted in NumBytes.
if (VARegSaveSize)
emitSPUpdate(isARM, MBB, MBBI, dl, TII, -VARegSaveSize,
MachineInstr::FrameSetup);
if (!AFI->hasStackFrame()) {
if (NumBytes != 0)
emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
MachineInstr::FrameSetup);
return;
}
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
int FI = CSI[i].getFrameIdx();
switch (Reg) {
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
break;
case ARM::R8:
case ARM::R9:
case ARM::R10:
case ARM::R11:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
if (STI.isTargetDarwin()) {
AFI->addGPRCalleeSavedArea2Frame(FI);
GPRCS2Size += 4;
} else {
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
}
break;
default:
AFI->addDPRCalleeSavedAreaFrame(FI);
DPRCSSize += 8;
}
}
// Move past area 1.
if (GPRCS1Size > 0) MBBI++;
// Set FP to point to the stack slot that contains the previous FP.
// For Darwin, FP is R7, which has now been stored in spill area 1.
// Otherwise, if this is not Darwin, all the callee-saved registers go
// into spill area 1, including the FP in R11. In either case, it is
// now safe to emit this assignment.
bool HasFP = hasFP(MF);
if (HasFP) {
unsigned ADDriOpc = !AFI->isThumbFunction() ? ARM::ADDri : ARM::t2ADDri;
MachineInstrBuilder MIB =
BuildMI(MBB, MBBI, dl, TII.get(ADDriOpc), FramePtr)
.addFrameIndex(FramePtrSpillFI).addImm(0)
.setMIFlag(MachineInstr::FrameSetup);
AddDefaultCC(AddDefaultPred(MIB));
}
// Move past area 2.
if (GPRCS2Size > 0) MBBI++;
// Determine starting offsets of spill areas.
unsigned DPRCSOffset = NumBytes - (GPRCS1Size + GPRCS2Size + DPRCSSize);
unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
if (HasFP)
AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) +
NumBytes);
AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
// Move past area 3.
if (DPRCSSize > 0) {
MBBI++;
// Since vpush register list cannot have gaps, there may be multiple vpush
// instructions in the prologue.
while (MBBI->getOpcode() == ARM::VSTMDDB_UPD)
MBBI++;
}
NumBytes = DPRCSOffset;
if (NumBytes) {
// Adjust SP after all the callee-save spills.
emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
MachineInstr::FrameSetup);
if (HasFP && isARM)
// Restore from fp only in ARM mode: e.g. sub sp, r7, #24
// Note it's not safe to do this in Thumb2 mode because it would have
// taken two instructions:
// mov sp, r7
// sub sp, #24
// If an interrupt is taken between the two instructions, then sp is in
// an inconsistent state (pointing to the middle of callee-saved area).
// The interrupt handler can end up clobbering the registers.
AFI->setShouldRestoreSPFromFP(true);
}
if (STI.isTargetELF() && hasFP(MF))
MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
AFI->getFramePtrSpillOffset());
AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
// If we need dynamic stack realignment, do it here. Be paranoid and make
// sure if we also have VLAs, we have a base pointer for frame access.
if (RegInfo->needsStackRealignment(MF)) {
unsigned MaxAlign = MFI->getMaxAlignment();
assert (!AFI->isThumb1OnlyFunction());
if (!AFI->isThumbFunction()) {
// Emit bic sp, sp, MaxAlign
AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
TII.get(ARM::BICri), ARM::SP)
.addReg(ARM::SP, RegState::Kill)
.addImm(MaxAlign-1)));
} else {
// We cannot use sp as source/dest register here, thus we're emitting the
// following sequence:
// mov r4, sp
// bic r4, r4, MaxAlign
// mov sp, r4
// FIXME: It will be better just to find spare register here.
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::R4)
.addReg(ARM::SP, RegState::Kill));
AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
TII.get(ARM::t2BICri), ARM::R4)
.addReg(ARM::R4, RegState::Kill)
.addImm(MaxAlign-1)));
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::SP)
.addReg(ARM::R4, RegState::Kill));
}
AFI->setShouldRestoreSPFromFP(true);
}
// If we need a base pointer, set it up here. It's whatever the value
// of the stack pointer is at this point. Any variable size objects
// will be allocated after this, so we can still use the base pointer
// to reference locals.
// FIXME: Clarify FrameSetup flags here.
if (RegInfo->hasBasePointer(MF)) {
if (isARM)
BuildMI(MBB, MBBI, dl,
TII.get(ARM::MOVr), RegInfo->getBaseRegister())
.addReg(ARM::SP)
.addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
else
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr),
RegInfo->getBaseRegister())
.addReg(ARM::SP));
}
// If the frame has variable sized objects then the epilogue must restore
// the sp from fp. We can assume there's an FP here since hasFP already
// checks for hasVarSizedObjects.
if (MFI->hasVarSizedObjects())
AFI->setShouldRestoreSPFromFP(true);
}
/// emitPrologue - Push callee-saved registers onto the stack, which
/// automatically adjust the stack pointer. Adjust the stack pointer to allocate
/// space for local variables. Also emit labels used by the exception handler to
/// generate the exception handling frames.
void X86FrameInfo::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front(); // Prologue goes in entry BB.
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
const Function *Fn = MF.getFunction();
const X86RegisterInfo *RegInfo = TM.getRegisterInfo();
const X86InstrInfo &TII = *TM.getInstrInfo();
MachineModuleInfo &MMI = MF.getMMI();
X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>();
bool needsFrameMoves = MMI.hasDebugInfo() ||
!Fn->doesNotThrow() || UnwindTablesMandatory;
uint64_t MaxAlign = MFI->getMaxAlignment(); // Desired stack alignment.
uint64_t StackSize = MFI->getStackSize(); // Number of bytes to allocate.
bool HasFP = hasFP(MF);
bool Is64Bit = STI.is64Bit();
bool IsWin64 = STI.isTargetWin64();
unsigned StackAlign = getStackAlignment();
unsigned SlotSize = RegInfo->getSlotSize();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
unsigned StackPtr = RegInfo->getStackRegister();
DebugLoc DL;
// If we're forcing a stack realignment we can't rely on just the frame
// info, we need to know the ABI stack alignment as well in case we
// have a call out. Otherwise just make sure we have some alignment - we'll
// go with the minimum SlotSize.
if (ForceStackAlign) {
if (MFI->hasCalls())
MaxAlign = (StackAlign > MaxAlign) ? StackAlign : MaxAlign;
else if (MaxAlign < SlotSize)
MaxAlign = SlotSize;
}
// Add RETADDR move area to callee saved frame size.
int TailCallReturnAddrDelta = X86FI->getTCReturnAddrDelta();
if (TailCallReturnAddrDelta < 0)
X86FI->setCalleeSavedFrameSize(
X86FI->getCalleeSavedFrameSize() - TailCallReturnAddrDelta);
// If this is x86-64 and the Red Zone is not disabled, if we are a leaf
// function, and use up to 128 bytes of stack space, don't have a frame
// pointer, calls, or dynamic alloca then we do not need to adjust the
// stack pointer (we fit in the Red Zone).
if (Is64Bit && !Fn->hasFnAttr(Attribute::NoRedZone) &&
!RegInfo->needsStackRealignment(MF) &&
!MFI->hasVarSizedObjects() && // No dynamic alloca.
!MFI->adjustsStack() && // No calls.
!IsWin64) { // Win64 has no Red Zone
uint64_t MinSize = X86FI->getCalleeSavedFrameSize();
if (HasFP) MinSize += SlotSize;
StackSize = std::max(MinSize, StackSize > 128 ? StackSize - 128 : 0);
MFI->setStackSize(StackSize);
} else if (IsWin64) {
// We need to always allocate 32 bytes as register spill area.
// FIXME: We might reuse these 32 bytes for leaf functions.
StackSize += 32;
MFI->setStackSize(StackSize);
}
// Insert stack pointer adjustment for later moving of return addr. Only
// applies to tail call optimized functions where the callee argument stack
// size is bigger than the callers.
if (TailCallReturnAddrDelta < 0) {
MachineInstr *MI =
BuildMI(MBB, MBBI, DL,
TII.get(getSUBriOpcode(Is64Bit, -TailCallReturnAddrDelta)),
StackPtr)
.addReg(StackPtr)
.addImm(-TailCallReturnAddrDelta);
MI->getOperand(3).setIsDead(); // The EFLAGS implicit def is dead.
}
// Mapping for machine moves:
//
// DST: VirtualFP AND
// SRC: VirtualFP => DW_CFA_def_cfa_offset
// ELSE => DW_CFA_def_cfa
//
// SRC: VirtualFP AND
// DST: Register => DW_CFA_def_cfa_register
//
// ELSE
// OFFSET < 0 => DW_CFA_offset_extended_sf
// REG < 64 => DW_CFA_offset + Reg
// ELSE => DW_CFA_offset_extended
std::vector<MachineMove> &Moves = MMI.getFrameMoves();
const TargetData *TD = MF.getTarget().getTargetData();
uint64_t NumBytes = 0;
int stackGrowth = -TD->getPointerSize();
if (HasFP) {
// Calculate required stack adjustment.
uint64_t FrameSize = StackSize - SlotSize;
if (RegInfo->needsStackRealignment(MF))
FrameSize = (FrameSize + MaxAlign - 1) / MaxAlign * MaxAlign;
NumBytes = FrameSize - X86FI->getCalleeSavedFrameSize();
// Get the offset of the stack slot for the EBP register, which is
// guaranteed to be the last slot by processFunctionBeforeFrameFinalized.
// Update the frame offset adjustment.
MFI->setOffsetAdjustment(-NumBytes);
// Save EBP/RBP into the appropriate stack slot.
BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::PUSH64r : X86::PUSH32r))
.addReg(FramePtr, RegState::Kill);
if (needsFrameMoves) {
// Mark the place where EBP/RBP was saved.
MCSymbol *FrameLabel = MMI.getContext().CreateTempSymbol();
BuildMI(MBB, MBBI, DL, TII.get(X86::PROLOG_LABEL)).addSym(FrameLabel);
// Define the current CFA rule to use the provided offset.
if (StackSize) {
MachineLocation SPDst(MachineLocation::VirtualFP);
MachineLocation SPSrc(MachineLocation::VirtualFP, 2 * stackGrowth);
Moves.push_back(MachineMove(FrameLabel, SPDst, SPSrc));
} else {
// FIXME: Verify & implement for FP
MachineLocation SPDst(StackPtr);
MachineLocation SPSrc(StackPtr, stackGrowth);
Moves.push_back(MachineMove(FrameLabel, SPDst, SPSrc));
}
// Change the rule for the FramePtr to be an "offset" rule.
MachineLocation FPDst(MachineLocation::VirtualFP, 2 * stackGrowth);
MachineLocation FPSrc(FramePtr);
Moves.push_back(MachineMove(FrameLabel, FPDst, FPSrc));
}
// Update EBP with the new base value...
BuildMI(MBB, MBBI, DL,
TII.get(Is64Bit ? X86::MOV64rr : X86::MOV32rr), FramePtr)
.addReg(StackPtr);
if (needsFrameMoves) {
// Mark effective beginning of when frame pointer becomes valid.
MCSymbol *FrameLabel = MMI.getContext().CreateTempSymbol();
BuildMI(MBB, MBBI, DL, TII.get(X86::PROLOG_LABEL)).addSym(FrameLabel);
// Define the current CFA to use the EBP/RBP register.
MachineLocation FPDst(FramePtr);
MachineLocation FPSrc(MachineLocation::VirtualFP);
Moves.push_back(MachineMove(FrameLabel, FPDst, FPSrc));
}
// Mark the FramePtr as live-in in every block except the entry.
for (MachineFunction::iterator I = llvm::next(MF.begin()), E = MF.end();
I != E; ++I)
I->addLiveIn(FramePtr);
// Realign stack
if (RegInfo->needsStackRealignment(MF)) {
MachineInstr *MI =
BuildMI(MBB, MBBI, DL,
TII.get(Is64Bit ? X86::AND64ri32 : X86::AND32ri),
StackPtr).addReg(StackPtr).addImm(-MaxAlign);
// The EFLAGS implicit def is dead.
MI->getOperand(3).setIsDead();
}
} else {
NumBytes = StackSize - X86FI->getCalleeSavedFrameSize();
}
// Skip the callee-saved push instructions.
bool PushedRegs = false;
int StackOffset = 2 * stackGrowth;
while (MBBI != MBB.end() &&
(MBBI->getOpcode() == X86::PUSH32r ||
MBBI->getOpcode() == X86::PUSH64r)) {
PushedRegs = true;
++MBBI;
if (!HasFP && needsFrameMoves) {
// Mark callee-saved push instruction.
MCSymbol *Label = MMI.getContext().CreateTempSymbol();
BuildMI(MBB, MBBI, DL, TII.get(X86::PROLOG_LABEL)).addSym(Label);
// Define the current CFA rule to use the provided offset.
unsigned Ptr = StackSize ?
MachineLocation::VirtualFP : StackPtr;
MachineLocation SPDst(Ptr);
MachineLocation SPSrc(Ptr, StackOffset);
Moves.push_back(MachineMove(Label, SPDst, SPSrc));
StackOffset += stackGrowth;
}
}
DL = MBB.findDebugLoc(MBBI);
// If there is an SUB32ri of ESP immediately before this instruction, merge
// the two. This can be the case when tail call elimination is enabled and
// the callee has more arguments then the caller.
NumBytes -= mergeSPUpdates(MBB, MBBI, StackPtr, true);
// If there is an ADD32ri or SUB32ri of ESP immediately after this
// instruction, merge the two instructions.
mergeSPUpdatesDown(MBB, MBBI, StackPtr, &NumBytes);
// Adjust stack pointer: ESP -= numbytes.
// Windows and cygwin/mingw require a prologue helper routine when allocating
// more than 4K bytes on the stack. Windows uses __chkstk and cygwin/mingw
// uses __alloca. __alloca and the 32-bit version of __chkstk will probe the
// stack and adjust the stack pointer in one go. The 64-bit version of
// __chkstk is only responsible for probing the stack. The 64-bit prologue is
// responsible for adjusting the stack pointer. Touching the stack at 4K
// increments is necessary to ensure that the guard pages used by the OS
// virtual memory manager are allocated in correct sequence.
if (NumBytes >= 4096 && (STI.isTargetCygMing() || STI.isTargetWin32())) {
// Check whether EAX is livein for this function.
bool isEAXAlive = isEAXLiveIn(MF);
const char *StackProbeSymbol =
STI.isTargetWindows() ? "_chkstk" : "_alloca";
unsigned CallOp = Is64Bit ? X86::CALL64pcrel32 : X86::CALLpcrel32;
if (!isEAXAlive) {
BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32ri), X86::EAX)
.addImm(NumBytes);
BuildMI(MBB, MBBI, DL, TII.get(CallOp))
.addExternalSymbol(StackProbeSymbol)
.addReg(StackPtr, RegState::Define | RegState::Implicit)
.addReg(X86::EFLAGS, RegState::Define | RegState::Implicit);
} else {
// Save EAX
BuildMI(MBB, MBBI, DL, TII.get(X86::PUSH32r))
.addReg(X86::EAX, RegState::Kill);
// Allocate NumBytes-4 bytes on stack. We'll also use 4 already
// allocated bytes for EAX.
BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32ri), X86::EAX)
.addImm(NumBytes - 4);
BuildMI(MBB, MBBI, DL, TII.get(CallOp))
.addExternalSymbol(StackProbeSymbol)
.addReg(StackPtr, RegState::Define | RegState::Implicit)
.addReg(X86::EFLAGS, RegState::Define | RegState::Implicit);
// Restore EAX
MachineInstr *MI = addRegOffset(BuildMI(MF, DL, TII.get(X86::MOV32rm),
X86::EAX),
StackPtr, false, NumBytes - 4);
MBB.insert(MBBI, MI);
}
} else if (NumBytes >= 4096 && STI.isTargetWin64()) {
// Sanity check that EAX is not livein for this function. It should
// should not be, so throw an assert.
assert(!isEAXLiveIn(MF) && "EAX is livein in the Win64 case!");
// Handle the 64-bit Windows ABI case where we need to call __chkstk.
// Function prologue is responsible for adjusting the stack pointer.
BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32ri), X86::EAX)
.addImm(NumBytes);
BuildMI(MBB, MBBI, DL, TII.get(X86::WINCALL64pcrel32))
.addExternalSymbol("__chkstk")
.addReg(StackPtr, RegState::Define | RegState::Implicit);
emitSPUpdate(MBB, MBBI, StackPtr, -(int64_t)NumBytes, Is64Bit, TII);
} else if (NumBytes)
emitSPUpdate(MBB, MBBI, StackPtr, -(int64_t)NumBytes, Is64Bit, TII);
if ((NumBytes || PushedRegs) && needsFrameMoves) {
// Mark end of stack pointer adjustment.
MCSymbol *Label = MMI.getContext().CreateTempSymbol();
BuildMI(MBB, MBBI, DL, TII.get(X86::PROLOG_LABEL)).addSym(Label);
if (!HasFP && NumBytes) {
// Define the current CFA rule to use the provided offset.
if (StackSize) {
MachineLocation SPDst(MachineLocation::VirtualFP);
MachineLocation SPSrc(MachineLocation::VirtualFP,
-StackSize + stackGrowth);
Moves.push_back(MachineMove(Label, SPDst, SPSrc));
} else {
// FIXME: Verify & implement for FP
MachineLocation SPDst(StackPtr);
MachineLocation SPSrc(StackPtr, stackGrowth);
Moves.push_back(MachineMove(Label, SPDst, SPSrc));
}
}
// Emit DWARF info specifying the offsets of the callee-saved registers.
if (PushedRegs)
emitCalleeSavedFrameMoves(MF, Label, HasFP ? FramePtr : StackPtr);
}
}
void Thumb1RegisterInfo::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front();
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
unsigned VARegSaveSize = AFI->getVarArgsRegSaveSize();
unsigned NumBytes = MFI->getStackSize();
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
DebugLoc dl = (MBBI != MBB.end() ?
MBBI->getDebugLoc() : DebugLoc::getUnknownLoc());
// Thumb add/sub sp, imm8 instructions implicitly multiply the offset by 4.
NumBytes = (NumBytes + 3) & ~3;
MFI->setStackSize(NumBytes);
// Determine the sizes of each callee-save spill areas and record which frame
// belongs to which callee-save spill areas.
unsigned GPRCS1Size = 0, GPRCS2Size = 0, DPRCSSize = 0;
int FramePtrSpillFI = 0;
if (VARegSaveSize)
emitSPUpdate(MBB, MBBI, TII, dl, *this, -VARegSaveSize);
if (!AFI->hasStackFrame()) {
if (NumBytes != 0)
emitSPUpdate(MBB, MBBI, TII, dl, *this, -NumBytes);
return;
}
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
int FI = CSI[i].getFrameIdx();
switch (Reg) {
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::R7:
case ARM::LR:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
break;
case ARM::R8:
case ARM::R9:
case ARM::R10:
case ARM::R11:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
if (STI.isTargetDarwin()) {
AFI->addGPRCalleeSavedArea2Frame(FI);
GPRCS2Size += 4;
} else {
AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
}
break;
default:
AFI->addDPRCalleeSavedAreaFrame(FI);
DPRCSSize += 8;
}
}
if (MBBI != MBB.end() && MBBI->getOpcode() == ARM::tPUSH) {
++MBBI;
if (MBBI != MBB.end())
dl = MBBI->getDebugLoc();
}
// Darwin ABI requires FP to point to the stack slot that contains the
// previous FP.
if (STI.isTargetDarwin() || hasFP(MF)) {
BuildMI(MBB, MBBI, dl, TII.get(ARM::tADDrSPi), FramePtr)
.addFrameIndex(FramePtrSpillFI).addImm(0);
}
// Determine starting offsets of spill areas.
unsigned DPRCSOffset = NumBytes - (GPRCS1Size + GPRCS2Size + DPRCSSize);
unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) + NumBytes);
AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
NumBytes = DPRCSOffset;
if (NumBytes) {
// Insert it after all the callee-save spills.
emitSPUpdate(MBB, MBBI, TII, dl, *this, -NumBytes);
}
if (STI.isTargetELF() && hasFP(MF)) {
MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
AFI->getFramePtrSpillOffset());
}
AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
}
void X86RegisterInfo::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB
MachineFrameInfo *MFI = MF.getFrameInfo();
const Function* Fn = MF.getFunction();
const X86Subtarget* Subtarget = &MF.getTarget().getSubtarget<X86Subtarget>();
MachineModuleInfo *MMI = MFI->getMachineModuleInfo();
X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>();
MachineBasicBlock::iterator MBBI = MBB.begin();
bool needsFrameMoves = (MMI && MMI->hasDebugInfo()) ||
!Fn->doesNotThrow() ||
UnwindTablesMandatory;
// Prepare for frame info.
unsigned FrameLabelId = 0;
// Get the number of bytes to allocate from the FrameInfo.
uint64_t StackSize = MFI->getStackSize();
// Get desired stack alignment
uint64_t MaxAlign = MFI->getMaxAlignment();
// Add RETADDR move area to callee saved frame size.
int TailCallReturnAddrDelta = X86FI->getTCReturnAddrDelta();
if (TailCallReturnAddrDelta < 0)
X86FI->setCalleeSavedFrameSize(
X86FI->getCalleeSavedFrameSize() +(-TailCallReturnAddrDelta));
// If this is x86-64 and the Red Zone is not disabled, if we are a leaf
// function, and use up to 128 bytes of stack space, don't have a frame
// pointer, calls, or dynamic alloca then we do not need to adjust the
// stack pointer (we fit in the Red Zone).
if (Is64Bit && !DisableRedZone &&
!needsStackRealignment(MF) &&
!MFI->hasVarSizedObjects() && // No dynamic alloca.
!MFI->hasCalls()) { // No calls.
uint64_t MinSize = X86FI->getCalleeSavedFrameSize();
if (hasFP(MF)) MinSize += SlotSize;
StackSize = std::max(MinSize,
StackSize > 128 ? StackSize - 128 : 0);
MFI->setStackSize(StackSize);
}
// Insert stack pointer adjustment for later moving of return addr. Only
// applies to tail call optimized functions where the callee argument stack
// size is bigger than the callers.
if (TailCallReturnAddrDelta < 0) {
MachineInstr *MI =
BuildMI(MBB, MBBI, TII.get(Is64Bit? X86::SUB64ri32 : X86::SUB32ri),
StackPtr).addReg(StackPtr).addImm(-TailCallReturnAddrDelta);
// The EFLAGS implicit def is dead.
MI->getOperand(3).setIsDead();
}
uint64_t NumBytes = 0;
if (hasFP(MF)) {
// Calculate required stack adjustment
uint64_t FrameSize = StackSize - SlotSize;
if (needsStackRealignment(MF))
FrameSize = (FrameSize + MaxAlign - 1)/MaxAlign*MaxAlign;
NumBytes = FrameSize - X86FI->getCalleeSavedFrameSize();
// Get the offset of the stack slot for the EBP register... which is
// guaranteed to be the last slot by processFunctionBeforeFrameFinalized.
// Update the frame offset adjustment.
MFI->setOffsetAdjustment(-NumBytes);
// Save EBP into the appropriate stack slot...
BuildMI(MBB, MBBI, TII.get(Is64Bit ? X86::PUSH64r : X86::PUSH32r))
.addReg(FramePtr, /*isDef=*/false, /*isImp=*/false, /*isKill=*/true);
if (needsFrameMoves) {
// Mark effective beginning of when frame pointer becomes valid.
FrameLabelId = MMI->NextLabelID();
BuildMI(MBB, MBBI, TII.get(X86::DBG_LABEL)).addImm(FrameLabelId);
}
// Update EBP with the new base value...
BuildMI(MBB, MBBI, TII.get(Is64Bit ? X86::MOV64rr : X86::MOV32rr), FramePtr)
.addReg(StackPtr);
// Mark the FramePtr as live-in in every block except the entry.
for (MachineFunction::iterator I = next(MF.begin()), E = MF.end();
I != E; ++I)
I->addLiveIn(FramePtr);
// Realign stack
if (needsStackRealignment(MF)) {
MachineInstr *MI =
BuildMI(MBB, MBBI,
TII.get(Is64Bit ? X86::AND64ri32 : X86::AND32ri),
StackPtr).addReg(StackPtr).addImm(-MaxAlign);
// The EFLAGS implicit def is dead.
MI->getOperand(3).setIsDead();
}
} else
NumBytes = StackSize - X86FI->getCalleeSavedFrameSize();
unsigned ReadyLabelId = 0;
if (needsFrameMoves) {
// Mark effective beginning of when frame pointer is ready.
ReadyLabelId = MMI->NextLabelID();
BuildMI(MBB, MBBI, TII.get(X86::DBG_LABEL)).addImm(ReadyLabelId);
}
// Skip the callee-saved push instructions.
while (MBBI != MBB.end() &&
(MBBI->getOpcode() == X86::PUSH32r ||
MBBI->getOpcode() == X86::PUSH64r))
++MBBI;
if (NumBytes) { // adjust stack pointer: ESP -= numbytes
if (NumBytes >= 4096 && Subtarget->isTargetCygMing()) {
// Check, whether EAX is livein for this function
bool isEAXAlive = false;
for (MachineRegisterInfo::livein_iterator
II = MF.getRegInfo().livein_begin(),
EE = MF.getRegInfo().livein_end(); (II != EE) && !isEAXAlive; ++II) {
unsigned Reg = II->first;
isEAXAlive = (Reg == X86::EAX || Reg == X86::AX ||
Reg == X86::AH || Reg == X86::AL);
}
// Function prologue calls _alloca to probe the stack when allocating
// more than 4k bytes in one go. Touching the stack at 4K increments is
// necessary to ensure that the guard pages used by the OS virtual memory
// manager are allocated in correct sequence.
if (!isEAXAlive) {
BuildMI(MBB, MBBI, TII.get(X86::MOV32ri), X86::EAX).addImm(NumBytes);
BuildMI(MBB, MBBI, TII.get(X86::CALLpcrel32))
.addExternalSymbol("_alloca");
} else {
// Save EAX
BuildMI(MBB, MBBI, TII.get(X86::PUSH32r))
.addReg(X86::EAX, /*isDef=*/false, /*isImp=*/false, /*isKill=*/true);
// Allocate NumBytes-4 bytes on stack. We'll also use 4 already
// allocated bytes for EAX.
BuildMI(MBB, MBBI, TII.get(X86::MOV32ri), X86::EAX).addImm(NumBytes-4);
BuildMI(MBB, MBBI, TII.get(X86::CALLpcrel32))
.addExternalSymbol("_alloca");
// Restore EAX
MachineInstr *MI = addRegOffset(BuildMI(MF, TII.get(X86::MOV32rm),X86::EAX),
StackPtr, false, NumBytes-4);
MBB.insert(MBBI, MI);
}
} else {
// If there is an SUB32ri of ESP immediately before this instruction,
// merge the two. This can be the case when tail call elimination is
// enabled and the callee has more arguments then the caller.
NumBytes -= mergeSPUpdates(MBB, MBBI, StackPtr, true);
// If there is an ADD32ri or SUB32ri of ESP immediately after this
// instruction, merge the two instructions.
mergeSPUpdatesDown(MBB, MBBI, StackPtr, &NumBytes);
if (NumBytes)
emitSPUpdate(MBB, MBBI, StackPtr, -(int64_t)NumBytes, Is64Bit, TII);
}
}
if (needsFrameMoves)
emitFrameMoves(MF, FrameLabelId, ReadyLabelId);
}
