void SPURegisterInfo::emitPrologue(MachineFunction &MF) const
{
MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
MachineModuleInfo *MMI = MFI->getMachineModuleInfo();
DebugLoc dl = (MBBI != MBB.end() ?
MBBI->getDebugLoc() : DebugLoc::getUnknownLoc());
// Prepare for debug frame info.
bool hasDebugInfo = MMI && MMI->hasDebugInfo();
MCSymbol *FrameLabel = 0;
// Move MBBI back to the beginning of the function.
MBBI = MBB.begin();
// Work out frame sizes.
determineFrameLayout(MF);
int FrameSize = MFI->getStackSize();
assert((FrameSize & 0xf) == 0
&& "SPURegisterInfo::emitPrologue: FrameSize not aligned");
if (FrameSize > 0 || MFI->hasCalls()) {
FrameSize = -(FrameSize + SPUFrameInfo::minStackSize());
if (hasDebugInfo) {
// Mark effective beginning of when frame pointer becomes valid.
FrameLabel = MMI->getContext().CreateTempSymbol();
BuildMI(MBB, MBBI, dl, TII.get(SPU::DBG_LABEL)).addSym(FrameLabel);
}
// Adjust stack pointer, spilling $lr -> 16($sp) and $sp -> -FrameSize($sp)
// for the ABI
BuildMI(MBB, MBBI, dl, TII.get(SPU::STQDr32), SPU::R0).addImm(16)
.addReg(SPU::R1);
if (isS10Constant(FrameSize)) {
// Spill $sp to adjusted $sp
BuildMI(MBB, MBBI, dl, TII.get(SPU::STQDr32), SPU::R1).addImm(FrameSize)
.addReg(SPU::R1);
// Adjust $sp by required amout
BuildMI(MBB, MBBI, dl, TII.get(SPU::AIr32), SPU::R1).addReg(SPU::R1)
.addImm(FrameSize);
} else if (FrameSize <= (1 << 16) - 1 && FrameSize >= -(1 << 16)) {
// Frame size can be loaded into ILr32n, so temporarily spill $r2 and use
// $r2 to adjust $sp:
BuildMI(MBB, MBBI, dl, TII.get(SPU::STQDr128), SPU::R2)
.addImm(-16)
.addReg(SPU::R1);
BuildMI(MBB, MBBI, dl, TII.get(SPU::ILr32), SPU::R2)
.addImm(FrameSize);
BuildMI(MBB, MBBI, dl, TII.get(SPU::STQDr32), SPU::R1)
.addReg(SPU::R2)
.addReg(SPU::R1);
BuildMI(MBB, MBBI, dl, TII.get(SPU::Ar32), SPU::R1)
.addReg(SPU::R1)
.addReg(SPU::R2);
BuildMI(MBB, MBBI, dl, TII.get(SPU::SFIr32), SPU::R2)
.addReg(SPU::R2)
.addImm(16);
BuildMI(MBB, MBBI, dl, TII.get(SPU::LQXr128), SPU::R2)
.addReg(SPU::R2)
.addReg(SPU::R1);
} else {
std::string msg;
raw_string_ostream Msg(msg);
Msg << "Unhandled frame size: " << FrameSize;
llvm_report_error(Msg.str());
}
if (hasDebugInfo) {
std::vector<MachineMove> &Moves = MMI->getFrameMoves();
// Show update of SP.
MachineLocation SPDst(MachineLocation::VirtualFP);
MachineLocation SPSrc(MachineLocation::VirtualFP, -FrameSize);
Moves.push_back(MachineMove(FrameLabel, SPDst, SPSrc));
// Add callee saved registers to move list.
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
for (unsigned I = 0, E = CSI.size(); I != E; ++I) {
int Offset = MFI->getObjectOffset(CSI[I].getFrameIdx());
unsigned Reg = CSI[I].getReg();
if (Reg == SPU::R0) continue;
MachineLocation CSDst(MachineLocation::VirtualFP, Offset);
MachineLocation CSSrc(Reg);
Moves.push_back(MachineMove(FrameLabel, CSDst, CSSrc));
}
// Mark effective beginning of when frame pointer is ready.
MCSymbol *ReadyLabel = MMI->getContext().CreateTempSymbol();
BuildMI(MBB, MBBI, dl, TII.get(SPU::DBG_LABEL)).addSym(ReadyLabel);
MachineLocation FPDst(SPU::R1);
MachineLocation FPSrc(MachineLocation::VirtualFP);
Moves.push_back(MachineMove(ReadyLabel, FPDst, FPSrc));
}
} else {
// This is a leaf function -- insert a branch hint iff there are
// sufficient number instructions in the basic block. Note that
// this is just a best guess based on the basic block's size.
if (MBB.size() >= (unsigned) SPUFrameInfo::branchHintPenalty()) {
MachineBasicBlock::iterator MBBI = prior(MBB.end());
dl = MBBI->getDebugLoc();
// Insert terminator label
BuildMI(MBB, MBBI, dl, TII.get(SPU::DBG_LABEL))
.addSym(MMI->getContext().CreateTempSymbol());
}
}
}
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);
}
void XCoreRegisterInfo::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
MachineModuleInfo *MMI = MFI->getMachineModuleInfo();
XCoreFunctionInfo *XFI = MF.getInfo<XCoreFunctionInfo>();
bool FP = hasFP(MF);
// Work out frame sizes.
int FrameSize = MFI->getStackSize();
assert(FrameSize%4 == 0 && "Misaligned frame size");
FrameSize/=4;
bool isU6 = isImmU6(FrameSize);
if (!isU6 && !isImmU16(FrameSize)) {
// FIXME could emit multiple instructions.
cerr << "emitPrologue Frame size too big: " << FrameSize << "\n";
abort();
}
bool emitFrameMoves = needsFrameMoves(MF);
// Do we need to allocate space on the stack?
if (FrameSize) {
bool saveLR = XFI->getUsesLR();
bool LRSavedOnEntry = false;
int Opcode;
if (saveLR && (MFI->getObjectOffset(XFI->getLRSpillSlot()) == 0)) {
Opcode = (isU6) ? XCore::ENTSP_u6 : XCore::ENTSP_lu6;
MBB.addLiveIn(XCore::LR);
saveLR = false;
LRSavedOnEntry = true;
} else {
Opcode = (isU6) ? XCore::EXTSP_u6 : XCore::EXTSP_lu6;
}
BuildMI(MBB, MBBI, TII.get(Opcode)).addImm(FrameSize);
if (emitFrameMoves) {
std::vector<MachineMove> &Moves = MMI->getFrameMoves();
// Show update of SP.
unsigned FrameLabelId = MMI->NextLabelID();
BuildMI(MBB, MBBI, TII.get(XCore::DBG_LABEL)).addImm(FrameLabelId);
MachineLocation SPDst(MachineLocation::VirtualFP);
MachineLocation SPSrc(MachineLocation::VirtualFP, -FrameSize * 4);
Moves.push_back(MachineMove(FrameLabelId, SPDst, SPSrc));
if (LRSavedOnEntry) {
MachineLocation CSDst(MachineLocation::VirtualFP, 0);
MachineLocation CSSrc(XCore::LR);
Moves.push_back(MachineMove(FrameLabelId, CSDst, CSSrc));
}
}
if (saveLR) {
int LRSpillOffset = MFI->getObjectOffset(XFI->getLRSpillSlot());
storeToStack(MBB, MBBI, XCore::LR, LRSpillOffset + FrameSize*4);
MBB.addLiveIn(XCore::LR);
if (emitFrameMoves) {
unsigned SaveLRLabelId = MMI->NextLabelID();
BuildMI(MBB, MBBI, TII.get(XCore::DBG_LABEL)).addImm(SaveLRLabelId);
MachineLocation CSDst(MachineLocation::VirtualFP, LRSpillOffset);
MachineLocation CSSrc(XCore::LR);
MMI->getFrameMoves().push_back(MachineMove(SaveLRLabelId,
CSDst, CSSrc));
}
}
}
if (FP) {
// Save R10 to the stack.
int FPSpillOffset = MFI->getObjectOffset(XFI->getFPSpillSlot());
storeToStack(MBB, MBBI, XCore::R10, FPSpillOffset + FrameSize*4);
// R10 is live-in. It is killed at the spill.
MBB.addLiveIn(XCore::R10);
if (emitFrameMoves) {
unsigned SaveR10LabelId = MMI->NextLabelID();
BuildMI(MBB, MBBI, TII.get(XCore::DBG_LABEL)).addImm(SaveR10LabelId);
MachineLocation CSDst(MachineLocation::VirtualFP, FPSpillOffset);
MachineLocation CSSrc(XCore::R10);
MMI->getFrameMoves().push_back(MachineMove(SaveR10LabelId,
CSDst, CSSrc));
}
// Set the FP from the SP.
unsigned FramePtr = XCore::R10;
BuildMI(MBB, MBBI, TII.get(XCore::LDAWSP_ru6), FramePtr)
.addImm(0);
if (emitFrameMoves) {
// Show FP is now valid.
unsigned FrameLabelId = MMI->NextLabelID();
BuildMI(MBB, MBBI, TII.get(XCore::DBG_LABEL)).addImm(FrameLabelId);
MachineLocation SPDst(FramePtr);
MachineLocation SPSrc(MachineLocation::VirtualFP);
MMI->getFrameMoves().push_back(MachineMove(FrameLabelId, SPDst, SPSrc));
}
}
if (emitFrameMoves) {
// Frame moves for callee saved.
std::vector<MachineMove> &Moves = MMI->getFrameMoves();
std::vector<std::pair<unsigned, CalleeSavedInfo> >&SpillLabels =
XFI->getSpillLabels();
for (unsigned I = 0, E = SpillLabels.size(); I != E; ++I) {
unsigned SpillLabel = SpillLabels[I].first;
CalleeSavedInfo &CSI = SpillLabels[I].second;
int Offset = MFI->getObjectOffset(CSI.getFrameIdx());
unsigned Reg = CSI.getReg();
MachineLocation CSDst(MachineLocation::VirtualFP, Offset);
MachineLocation CSSrc(Reg);
Moves.push_back(MachineMove(SpillLabel, CSDst, CSSrc));
}
}
}
void X86RegisterInfo::emitFrameMoves(MachineFunction &MF,
unsigned FrameLabelId,
unsigned ReadyLabelId) const {
MachineFrameInfo *MFI = MF.getFrameInfo();
MachineModuleInfo *MMI = MFI->getMachineModuleInfo();
if (!MMI)
return;
uint64_t StackSize = MFI->getStackSize();
std::vector<MachineMove> &Moves = MMI->getFrameMoves();
const TargetData *TD = MF.getTarget().getTargetData();
// Calculate amount of bytes used for return address storing
int stackGrowth =
(MF.getTarget().getFrameInfo()->getStackGrowthDirection() ==
TargetFrameInfo::StackGrowsUp ?
TD->getPointerSize() : -TD->getPointerSize());
if (StackSize) {
// Show update of SP.
if (hasFP(MF)) {
// Adjust SP
MachineLocation SPDst(MachineLocation::VirtualFP);
MachineLocation SPSrc(MachineLocation::VirtualFP, 2*stackGrowth);
Moves.push_back(MachineMove(FrameLabelId, SPDst, SPSrc));
} else {
MachineLocation SPDst(MachineLocation::VirtualFP);
MachineLocation SPSrc(MachineLocation::VirtualFP,
-StackSize+stackGrowth);
Moves.push_back(MachineMove(FrameLabelId, SPDst, SPSrc));
}
} else {
//FIXME: Verify & implement for FP
MachineLocation SPDst(StackPtr);
MachineLocation SPSrc(StackPtr, stackGrowth);
Moves.push_back(MachineMove(FrameLabelId, SPDst, SPSrc));
}
// Add callee saved registers to move list.
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
// FIXME: This is dirty hack. The code itself is pretty mess right now.
// It should be rewritten from scratch and generalized sometimes.
// Determine maximum offset (minumum due to stack growth)
int64_t MaxOffset = 0;
for (unsigned I = 0, E = CSI.size(); I!=E; ++I)
MaxOffset = std::min(MaxOffset,
MFI->getObjectOffset(CSI[I].getFrameIdx()));
// Calculate offsets
int64_t saveAreaOffset = (hasFP(MF) ? 3 : 2)*stackGrowth;
for (unsigned I = 0, E = CSI.size(); I!=E; ++I) {
int64_t Offset = MFI->getObjectOffset(CSI[I].getFrameIdx());
unsigned Reg = CSI[I].getReg();
Offset = (MaxOffset-Offset+saveAreaOffset);
MachineLocation CSDst(MachineLocation::VirtualFP, Offset);
MachineLocation CSSrc(Reg);
Moves.push_back(MachineMove(FrameLabelId, CSDst, CSSrc));
}
if (hasFP(MF)) {
// Save FP
MachineLocation FPDst(MachineLocation::VirtualFP, 2*stackGrowth);
MachineLocation FPSrc(FramePtr);
Moves.push_back(MachineMove(ReadyLabelId, FPDst, FPSrc));
}
MachineLocation FPDst(hasFP(MF) ? FramePtr : StackPtr);
MachineLocation FPSrc(MachineLocation::VirtualFP);
Moves.push_back(MachineMove(ReadyLabelId, FPDst, FPSrc));
}
