diff options
Diffstat (limited to 'contrib/llvm/lib/Target/X86/X86RegisterInfo.cpp')
| -rw-r--r-- | contrib/llvm/lib/Target/X86/X86RegisterInfo.cpp | 955 |
1 files changed, 98 insertions, 857 deletions
diff --git a/contrib/llvm/lib/Target/X86/X86RegisterInfo.cpp b/contrib/llvm/lib/Target/X86/X86RegisterInfo.cpp index fedd49e..2f6bd88 100644 --- a/contrib/llvm/lib/Target/X86/X86RegisterInfo.cpp +++ b/contrib/llvm/lib/Target/X86/X86RegisterInfo.cpp @@ -31,7 +31,7 @@ #include "llvm/CodeGen/MachineModuleInfo.h" #include "llvm/CodeGen/MachineRegisterInfo.h" #include "llvm/MC/MCAsmInfo.h" -#include "llvm/Target/TargetFrameInfo.h" +#include "llvm/Target/TargetFrameLowering.h" #include "llvm/Target/TargetInstrInfo.h" #include "llvm/Target/TargetMachine.h" #include "llvm/Target/TargetOptions.h" @@ -41,7 +41,7 @@ #include "llvm/Support/CommandLine.h" using namespace llvm; -static cl::opt<bool> +cl::opt<bool> ForceStackAlign("force-align-stack", cl::desc("Force align the stack to the minimum alignment" " needed for the function."), @@ -60,7 +60,7 @@ X86RegisterInfo::X86RegisterInfo(X86TargetMachine &tm, const X86Subtarget *Subtarget = &TM.getSubtarget<X86Subtarget>(); Is64Bit = Subtarget->is64Bit(); IsWin64 = Subtarget->isTargetWin64(); - StackAlign = TM.getFrameInfo()->getStackAlignment(); + StackAlign = TM.getFrameLowering()->getStackAlignment(); if (Is64Bit) { SlotSize = 8; @@ -159,46 +159,21 @@ unsigned X86RegisterInfo::getX86RegNum(unsigned RegNo) { case X86::YMM7: case X86::YMM15: case X86::MM7: return 7; - case X86::ES: - return 0; - case X86::CS: - return 1; - case X86::SS: - return 2; - case X86::DS: - return 3; - case X86::FS: - return 4; - case X86::GS: - return 5; - - case X86::CR0: - return 0; - case X86::CR1: - return 1; - case X86::CR2: - return 2; - case X86::CR3: - return 3; - case X86::CR4: - return 4; - - case X86::DR0: - return 0; - case X86::DR1: - return 1; - case X86::DR2: - return 2; - case X86::DR3: - return 3; - case X86::DR4: - return 4; - case X86::DR5: - return 5; - case X86::DR6: - return 6; - case X86::DR7: - return 7; + case X86::ES: return 0; + case X86::CS: return 1; + case X86::SS: return 2; + case X86::DS: return 3; + case X86::FS: return 4; + case X86::GS: return 5; + + case X86::CR0: case X86::CR8 : case X86::DR0: return 0; + case X86::CR1: case X86::CR9 : case X86::DR1: return 1; + case X86::CR2: case X86::CR10: case X86::DR2: return 2; + case X86::CR3: case X86::CR11: case X86::DR3: return 3; + case X86::CR4: case X86::CR12: case X86::DR4: return 4; + case X86::CR5: case X86::CR13: case X86::DR5: return 5; + case X86::CR6: case X86::CR14: case X86::DR6: return 6; + case X86::CR7: case X86::CR15: case X86::DR7: return 7; // Pseudo index registers are equivalent to a "none" // scaled index (See Intel Manual 2A, table 2-3) @@ -295,9 +270,14 @@ X86RegisterInfo::getMatchingSuperRegClass(const TargetRegisterClass *A, } break; case X86::sub_32bit: - if (B == &X86::GR32RegClass || B == &X86::GR32_NOSPRegClass) { + if (B == &X86::GR32RegClass) { if (A->getSize() == 8) return A; + } else if (B == &X86::GR32_NOSPRegClass) { + if (A == &X86::GR64RegClass || A == &X86::GR64_NOSPRegClass) + return &X86::GR64_NOSPRegClass; + if (A->getSize() == 8) + return getCommonSubClass(A, &X86::GR64_NOSPRegClass); } else if (B == &X86::GR32_ABCDRegClass) { if (A == &X86::GR64RegClass || A == &X86::GR64_ABCDRegClass || A == &X86::GR64_NOREXRegClass || @@ -336,10 +316,16 @@ X86RegisterInfo::getPointerRegClass(unsigned Kind) const { if (TM.getSubtarget<X86Subtarget>().is64Bit()) return &X86::GR64RegClass; return &X86::GR32RegClass; - case 1: // Normal GRPs except the stack pointer (for encoding reasons). + case 1: // Normal GPRs except the stack pointer (for encoding reasons). if (TM.getSubtarget<X86Subtarget>().is64Bit()) return &X86::GR64_NOSPRegClass; return &X86::GR32_NOSPRegClass; + case 2: // Available for tailcall (not callee-saved GPRs). + if (TM.getSubtarget<X86Subtarget>().isTargetWin64()) + return &X86::GR64_TCW64RegClass; + if (TM.getSubtarget<X86Subtarget>().is64Bit()) + return &X86::GR64_TCRegClass; + return &X86::GR32_TCRegClass; } } @@ -408,6 +394,8 @@ X86RegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const { BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const { BitVector Reserved(getNumRegs()); + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + // Set the stack-pointer register and its aliases as reserved. Reserved.set(X86::RSP); Reserved.set(X86::ESP); @@ -420,7 +408,7 @@ BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const { Reserved.set(X86::IP); // Set the frame-pointer register and its aliases as reserved if needed. - if (hasFP(MF)) { + if (TFI->hasFP(MF)) { Reserved.set(X86::RBP); Reserved.set(X86::EBP); Reserved.set(X86::BP); @@ -445,21 +433,6 @@ BitVector X86RegisterInfo::getReservedRegs(const MachineFunction &MF) const { // Stack Frame Processing methods //===----------------------------------------------------------------------===// -/// hasFP - Return true if the specified function should have a dedicated frame -/// pointer register. This is true if the function has variable sized allocas -/// or if frame pointer elimination is disabled. -bool X86RegisterInfo::hasFP(const MachineFunction &MF) const { - const MachineFrameInfo *MFI = MF.getFrameInfo(); - const MachineModuleInfo &MMI = MF.getMMI(); - - return (DisableFramePointerElim(MF) || - needsStackRealignment(MF) || - MFI->hasVarSizedObjects() || - MFI->isFrameAddressTaken() || - MF.getInfo<X86MachineFunctionInfo>()->getForceFramePointer() || - MMI.callsUnwindInit()); -} - bool X86RegisterInfo::canRealignStack(const MachineFunction &MF) const { const MachineFrameInfo *MFI = MF.getFrameInfo(); return (RealignStack && @@ -478,62 +451,25 @@ bool X86RegisterInfo::needsStackRealignment(const MachineFunction &MF) const { if (0 && requiresRealignment && MFI->hasVarSizedObjects()) report_fatal_error( "Stack realignment in presense of dynamic allocas is not supported"); - + // If we've requested that we force align the stack do so now. if (ForceStackAlign) return canRealignStack(MF); - - return requiresRealignment && canRealignStack(MF); -} -bool X86RegisterInfo::hasReservedCallFrame(const MachineFunction &MF) const { - return !MF.getFrameInfo()->hasVarSizedObjects(); + return requiresRealignment && canRealignStack(MF); } bool X86RegisterInfo::hasReservedSpillSlot(const MachineFunction &MF, unsigned Reg, int &FrameIdx) const { - if (Reg == FramePtr && hasFP(MF)) { + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + + if (Reg == FramePtr && TFI->hasFP(MF)) { FrameIdx = MF.getFrameInfo()->getObjectIndexBegin(); return true; } return false; } -int -X86RegisterInfo::getFrameIndexOffset(const MachineFunction &MF, int FI) const { - const TargetFrameInfo &TFI = *MF.getTarget().getFrameInfo(); - const MachineFrameInfo *MFI = MF.getFrameInfo(); - int Offset = MFI->getObjectOffset(FI) - TFI.getOffsetOfLocalArea(); - uint64_t StackSize = MFI->getStackSize(); - - if (needsStackRealignment(MF)) { - if (FI < 0) { - // Skip the saved EBP. - Offset += SlotSize; - } else { - unsigned Align = MFI->getObjectAlignment(FI); - assert((-(Offset + StackSize)) % Align == 0); - Align = 0; - return Offset + StackSize; - } - // FIXME: Support tail calls - } else { - if (!hasFP(MF)) - return Offset + StackSize; - - // Skip the saved EBP. - Offset += SlotSize; - - // Skip the RETADDR move area - const X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); - int TailCallReturnAddrDelta = X86FI->getTCReturnAddrDelta(); - if (TailCallReturnAddrDelta < 0) - Offset -= TailCallReturnAddrDelta; - } - - return Offset; -} - static unsigned getSUBriOpcode(unsigned is64Bit, int64_t Imm) { if (is64Bit) { if (isInt<8>(Imm)) @@ -561,69 +497,70 @@ static unsigned getADDriOpcode(unsigned is64Bit, int64_t Imm) { void X86RegisterInfo:: eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, MachineBasicBlock::iterator I) const { - if (!hasReservedCallFrame(MF)) { + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + bool reseveCallFrame = TFI->hasReservedCallFrame(MF); + int Opcode = I->getOpcode(); + bool isDestroy = Opcode == getCallFrameDestroyOpcode(); + DebugLoc DL = I->getDebugLoc(); + uint64_t Amount = !reseveCallFrame ? I->getOperand(0).getImm() : 0; + uint64_t CalleeAmt = isDestroy ? I->getOperand(1).getImm() : 0; + I = MBB.erase(I); + + if (!reseveCallFrame) { // If the stack pointer can be changed after prologue, turn the // adjcallstackup instruction into a 'sub ESP, <amt>' and the // adjcallstackdown instruction into 'add ESP, <amt>' // TODO: consider using push / pop instead of sub + store / add - 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. - Amount = (Amount + StackAlign - 1) / StackAlign * StackAlign; - - MachineInstr *New = 0; - if (Old->getOpcode() == getCallFrameSetupOpcode()) { - New = BuildMI(MF, Old->getDebugLoc(), - TII.get(getSUBriOpcode(Is64Bit, Amount)), - StackPtr) - .addReg(StackPtr) - .addImm(Amount); - } else { - assert(Old->getOpcode() == getCallFrameDestroyOpcode()); - - // Factor out the amount the callee already popped. - uint64_t CalleeAmt = Old->getOperand(1).getImm(); - Amount -= CalleeAmt; - - if (Amount) { - unsigned Opc = getADDriOpcode(Is64Bit, Amount); - New = BuildMI(MF, Old->getDebugLoc(), TII.get(Opc), StackPtr) - .addReg(StackPtr) - .addImm(Amount); - } - } + if (Amount == 0) + return; + + // 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. + Amount = (Amount + StackAlign - 1) / StackAlign * StackAlign; + + MachineInstr *New = 0; + if (Opcode == getCallFrameSetupOpcode()) { + New = BuildMI(MF, DL, TII.get(getSUBriOpcode(Is64Bit, Amount)), + StackPtr) + .addReg(StackPtr) + .addImm(Amount); + } else { + assert(Opcode == getCallFrameDestroyOpcode()); - if (New) { - // The EFLAGS implicit def is dead. - New->getOperand(3).setIsDead(); + // Factor out the amount the callee already popped. + Amount -= CalleeAmt; - // Replace the pseudo instruction with a new instruction. - MBB.insert(I, New); + if (Amount) { + unsigned Opc = getADDriOpcode(Is64Bit, Amount); + New = BuildMI(MF, DL, TII.get(Opc), StackPtr) + .addReg(StackPtr).addImm(Amount); } } - } else if (I->getOpcode() == getCallFrameDestroyOpcode()) { - // If we are performing frame pointer elimination and if the callee pops - // something off the stack pointer, add it back. We do this until we have - // more advanced stack pointer tracking ability. - if (uint64_t CalleeAmt = I->getOperand(1).getImm()) { - unsigned Opc = getSUBriOpcode(Is64Bit, CalleeAmt); - MachineInstr *Old = I; - MachineInstr *New = - BuildMI(MF, Old->getDebugLoc(), TII.get(Opc), - StackPtr) - .addReg(StackPtr) - .addImm(CalleeAmt); + if (New) { // The EFLAGS implicit def is dead. New->getOperand(3).setIsDead(); + + // Replace the pseudo instruction with a new instruction. MBB.insert(I, New); } + + return; } - MBB.erase(I); + if (Opcode == getCallFrameDestroyOpcode() && CalleeAmt) { + // If we are performing frame pointer elimination and if the callee pops + // something off the stack pointer, add it back. We do this until we have + // more advanced stack pointer tracking ability. + unsigned Opc = getSUBriOpcode(Is64Bit, CalleeAmt); + MachineInstr *New = BuildMI(MF, DL, TII.get(Opc), StackPtr) + .addReg(StackPtr).addImm(CalleeAmt); + + // The EFLAGS implicit def is dead. + New->getOperand(3).setIsDead(); + MBB.insert(I, New); + } } void @@ -634,6 +571,7 @@ X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, unsigned i = 0; MachineInstr &MI = *II; MachineFunction &MF = *MI.getParent()->getParent(); + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); while (!MI.getOperand(i).isFI()) { ++i; @@ -650,7 +588,7 @@ X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, else if (AfterFPPop) BasePtr = StackPtr; else - BasePtr = (hasFP(MF) ? FramePtr : StackPtr); + BasePtr = (TFI->hasFP(MF) ? FramePtr : StackPtr); // This must be part of a four operand memory reference. Replace the // FrameIndex with base register with EBP. Add an offset to the offset. @@ -660,11 +598,10 @@ X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, int FIOffset; if (AfterFPPop) { // Tail call jmp happens after FP is popped. - const TargetFrameInfo &TFI = *MF.getTarget().getFrameInfo(); const MachineFrameInfo *MFI = MF.getFrameInfo(); - FIOffset = MFI->getObjectOffset(FrameIndex) - TFI.getOffsetOfLocalArea(); + FIOffset = MFI->getObjectOffset(FrameIndex) - TFI->getOffsetOfLocalArea(); } else - FIOffset = getFrameIndexOffset(MF, FrameIndex); + FIOffset = TFI->getFrameIndexOffset(MF, FrameIndex); if (MI.getOperand(i+3).isImm()) { // Offset is a 32-bit integer. @@ -677,710 +614,14 @@ X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, } } -void -X86RegisterInfo::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, - RegScavenger *RS) const { - MachineFrameInfo *MFI = MF.getFrameInfo(); - - X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); - int32_t TailCallReturnAddrDelta = X86FI->getTCReturnAddrDelta(); - - if (TailCallReturnAddrDelta < 0) { - // create RETURNADDR area - // arg - // arg - // RETADDR - // { ... - // RETADDR area - // ... - // } - // [EBP] - MFI->CreateFixedObject(-TailCallReturnAddrDelta, - (-1U*SlotSize)+TailCallReturnAddrDelta, true); - } - - if (hasFP(MF)) { - assert((TailCallReturnAddrDelta <= 0) && - "The Delta should always be zero or negative"); - const TargetFrameInfo &TFI = *MF.getTarget().getFrameInfo(); - - // Create a frame entry for the EBP register that must be saved. - int FrameIdx = MFI->CreateFixedObject(SlotSize, - -(int)SlotSize + - TFI.getOffsetOfLocalArea() + - TailCallReturnAddrDelta, - true); - assert(FrameIdx == MFI->getObjectIndexBegin() && - "Slot for EBP register must be last in order to be found!"); - FrameIdx = 0; - } -} - -/// emitSPUpdate - Emit a series of instructions to increment / decrement the -/// stack pointer by a constant value. -static -void emitSPUpdate(MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI, - unsigned StackPtr, int64_t NumBytes, bool Is64Bit, - const TargetInstrInfo &TII) { - bool isSub = NumBytes < 0; - uint64_t Offset = isSub ? -NumBytes : NumBytes; - unsigned Opc = isSub ? - getSUBriOpcode(Is64Bit, Offset) : - getADDriOpcode(Is64Bit, Offset); - uint64_t Chunk = (1LL << 31) - 1; - DebugLoc DL = MBB.findDebugLoc(MBBI); - - while (Offset) { - uint64_t ThisVal = (Offset > Chunk) ? Chunk : Offset; - MachineInstr *MI = - BuildMI(MBB, MBBI, DL, TII.get(Opc), StackPtr) - .addReg(StackPtr) - .addImm(ThisVal); - MI->getOperand(3).setIsDead(); // The EFLAGS implicit def is dead. - Offset -= ThisVal; - } -} - -/// mergeSPUpdatesUp - Merge two stack-manipulating instructions upper iterator. -static -void mergeSPUpdatesUp(MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI, - unsigned StackPtr, uint64_t *NumBytes = NULL) { - if (MBBI == MBB.begin()) return; - - MachineBasicBlock::iterator PI = prior(MBBI); - unsigned Opc = PI->getOpcode(); - if ((Opc == X86::ADD64ri32 || Opc == X86::ADD64ri8 || - Opc == X86::ADD32ri || Opc == X86::ADD32ri8) && - PI->getOperand(0).getReg() == StackPtr) { - if (NumBytes) - *NumBytes += PI->getOperand(2).getImm(); - MBB.erase(PI); - } else if ((Opc == X86::SUB64ri32 || Opc == X86::SUB64ri8 || - Opc == X86::SUB32ri || Opc == X86::SUB32ri8) && - PI->getOperand(0).getReg() == StackPtr) { - if (NumBytes) - *NumBytes -= PI->getOperand(2).getImm(); - MBB.erase(PI); - } -} - -/// mergeSPUpdatesDown - Merge two stack-manipulating instructions lower iterator. -static -void mergeSPUpdatesDown(MachineBasicBlock &MBB, - MachineBasicBlock::iterator &MBBI, - unsigned StackPtr, uint64_t *NumBytes = NULL) { - // FIXME: THIS ISN'T RUN!!! - return; - - if (MBBI == MBB.end()) return; - - MachineBasicBlock::iterator NI = llvm::next(MBBI); - if (NI == MBB.end()) return; - - unsigned Opc = NI->getOpcode(); - if ((Opc == X86::ADD64ri32 || Opc == X86::ADD64ri8 || - Opc == X86::ADD32ri || Opc == X86::ADD32ri8) && - NI->getOperand(0).getReg() == StackPtr) { - if (NumBytes) - *NumBytes -= NI->getOperand(2).getImm(); - MBB.erase(NI); - MBBI = NI; - } else if ((Opc == X86::SUB64ri32 || Opc == X86::SUB64ri8 || - Opc == X86::SUB32ri || Opc == X86::SUB32ri8) && - NI->getOperand(0).getReg() == StackPtr) { - if (NumBytes) - *NumBytes += NI->getOperand(2).getImm(); - MBB.erase(NI); - MBBI = NI; - } -} - -/// mergeSPUpdates - Checks the instruction before/after the passed -/// instruction. If it is an ADD/SUB instruction it is deleted argument and the -/// stack adjustment is returned as a positive value for ADD and a negative for -/// SUB. -static int mergeSPUpdates(MachineBasicBlock &MBB, - MachineBasicBlock::iterator &MBBI, - unsigned StackPtr, - bool doMergeWithPrevious) { - if ((doMergeWithPrevious && MBBI == MBB.begin()) || - (!doMergeWithPrevious && MBBI == MBB.end())) - return 0; - - MachineBasicBlock::iterator PI = doMergeWithPrevious ? prior(MBBI) : MBBI; - MachineBasicBlock::iterator NI = doMergeWithPrevious ? 0 : llvm::next(MBBI); - unsigned Opc = PI->getOpcode(); - int Offset = 0; - - if ((Opc == X86::ADD64ri32 || Opc == X86::ADD64ri8 || - Opc == X86::ADD32ri || Opc == X86::ADD32ri8) && - PI->getOperand(0).getReg() == StackPtr){ - Offset += PI->getOperand(2).getImm(); - MBB.erase(PI); - if (!doMergeWithPrevious) MBBI = NI; - } else if ((Opc == X86::SUB64ri32 || Opc == X86::SUB64ri8 || - Opc == X86::SUB32ri || Opc == X86::SUB32ri8) && - PI->getOperand(0).getReg() == StackPtr) { - Offset -= PI->getOperand(2).getImm(); - MBB.erase(PI); - if (!doMergeWithPrevious) MBBI = NI; - } - - return Offset; -} - -void X86RegisterInfo::emitCalleeSavedFrameMoves(MachineFunction &MF, - MCSymbol *Label, - unsigned FramePtr) const { - MachineFrameInfo *MFI = MF.getFrameInfo(); - MachineModuleInfo &MMI = MF.getMMI(); - - // Add callee saved registers to move list. - const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo(); - if (CSI.empty()) return; - - std::vector<MachineMove> &Moves = MMI.getFrameMoves(); - const TargetData *TD = MF.getTarget().getTargetData(); - bool HasFP = hasFP(MF); - - // Calculate amount of bytes used for return address storing. - int stackGrowth = - (MF.getTarget().getFrameInfo()->getStackGrowthDirection() == - TargetFrameInfo::StackGrowsUp ? - TD->getPointerSize() : -TD->getPointerSize()); - - // 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 (std::vector<CalleeSavedInfo>::const_iterator - I = CSI.begin(), E = CSI.end(); I != E; ++I) - MaxOffset = std::min(MaxOffset, - MFI->getObjectOffset(I->getFrameIdx())); - - // Calculate offsets. - int64_t saveAreaOffset = (HasFP ? 3 : 2) * stackGrowth; - for (std::vector<CalleeSavedInfo>::const_iterator - I = CSI.begin(), E = CSI.end(); I != E; ++I) { - int64_t Offset = MFI->getObjectOffset(I->getFrameIdx()); - unsigned Reg = I->getReg(); - Offset = MaxOffset - Offset + saveAreaOffset; - - // Don't output a new machine move if we're re-saving the frame - // pointer. This happens when the PrologEpilogInserter has inserted an extra - // "PUSH" of the frame pointer -- the "emitPrologue" method automatically - // generates one when frame pointers are used. If we generate a "machine - // move" for this extra "PUSH", the linker will lose track of the fact that - // the frame pointer should have the value of the first "PUSH" when it's - // trying to unwind. - // - // FIXME: This looks inelegant. It's possibly correct, but it's covering up - // another bug. I.e., one where we generate a prolog like this: - // - // pushl %ebp - // movl %esp, %ebp - // pushl %ebp - // pushl %esi - // ... - // - // The immediate re-push of EBP is unnecessary. At the least, it's an - // optimization bug. EBP can be used as a scratch register in certain - // cases, but probably not when we have a frame pointer. - if (HasFP && FramePtr == Reg) - continue; - - MachineLocation CSDst(MachineLocation::VirtualFP, Offset); - MachineLocation CSSrc(Reg); - Moves.push_back(MachineMove(Label, CSDst, CSSrc)); - } -} - -/// 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 X86RegisterInfo::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 X86Subtarget *Subtarget = &MF.getTarget().getSubtarget<X86Subtarget>(); - 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); - 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) && - !needsStackRealignment(MF) && - !MFI->hasVarSizedObjects() && // No dynamic alloca. - !MFI->adjustsStack() && // No calls. - !Subtarget->isTargetWin64()) { // 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 (Subtarget->isTargetWin64()) { - // 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 (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 (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); - - // 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 && - (Subtarget->isTargetCygMing() || Subtarget->isTargetWin32())) { - // 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); - } - - - const char *StackProbeSymbol = - Subtarget->isTargetWindows() ? "_chkstk" : "_alloca"; - if (!isEAXAlive) { - BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32ri), X86::EAX) - .addImm(NumBytes); - BuildMI(MBB, MBBI, DL, TII.get(X86::CALLpcrel32)) - .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(X86::CALLpcrel32)) - .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) { - // 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 ((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 X86RegisterInfo::emitEpilogue(MachineFunction &MF, - MachineBasicBlock &MBB) const { - const MachineFrameInfo *MFI = MF.getFrameInfo(); - X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); - MachineBasicBlock::iterator MBBI = prior(MBB.end()); - unsigned RetOpcode = MBBI->getOpcode(); - DebugLoc DL = MBBI->getDebugLoc(); - - switch (RetOpcode) { - default: - llvm_unreachable("Can only insert epilog into returning blocks"); - case X86::RET: - case X86::RETI: - case X86::TCRETURNdi: - case X86::TCRETURNri: - case X86::TCRETURNmi: - case X86::TCRETURNdi64: - case X86::TCRETURNri64: - case X86::TCRETURNmi64: - case X86::EH_RETURN: - case X86::EH_RETURN64: - break; // These are ok - } - - // Get the number of bytes to allocate from the FrameInfo. - uint64_t StackSize = MFI->getStackSize(); - uint64_t MaxAlign = MFI->getMaxAlignment(); - unsigned CSSize = X86FI->getCalleeSavedFrameSize(); - uint64_t NumBytes = 0; - - // 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. - if (ForceStackAlign) { - if (MFI->hasCalls()) - MaxAlign = (StackAlign > MaxAlign) ? StackAlign : MaxAlign; - else - MaxAlign = MaxAlign ? MaxAlign : 4; - } - - if (hasFP(MF)) { - // Calculate required stack adjustment. - uint64_t FrameSize = StackSize - SlotSize; - if (needsStackRealignment(MF)) - FrameSize = (FrameSize + MaxAlign - 1)/MaxAlign*MaxAlign; - - NumBytes = FrameSize - CSSize; - - // Pop EBP. - BuildMI(MBB, MBBI, DL, - TII.get(Is64Bit ? X86::POP64r : X86::POP32r), FramePtr); - } else { - NumBytes = StackSize - CSSize; - } - - // Skip the callee-saved pop instructions. - MachineBasicBlock::iterator LastCSPop = MBBI; - while (MBBI != MBB.begin()) { - MachineBasicBlock::iterator PI = prior(MBBI); - unsigned Opc = PI->getOpcode(); - - if (Opc != X86::POP32r && Opc != X86::POP64r && - !PI->getDesc().isTerminator()) - break; - - --MBBI; - } - - DL = MBBI->getDebugLoc(); - - // If there is an ADD32ri or SUB32ri of ESP immediately before this - // instruction, merge the two instructions. - if (NumBytes || MFI->hasVarSizedObjects()) - mergeSPUpdatesUp(MBB, MBBI, StackPtr, &NumBytes); - - // If dynamic alloca is used, then reset esp to point to the last callee-saved - // slot before popping them off! Same applies for the case, when stack was - // realigned. - if (needsStackRealignment(MF)) { - // We cannot use LEA here, because stack pointer was realigned. We need to - // deallocate local frame back. - if (CSSize) { - emitSPUpdate(MBB, MBBI, StackPtr, NumBytes, Is64Bit, TII); - MBBI = prior(LastCSPop); - } - - BuildMI(MBB, MBBI, DL, - TII.get(Is64Bit ? X86::MOV64rr : X86::MOV32rr), - StackPtr).addReg(FramePtr); - } else if (MFI->hasVarSizedObjects()) { - if (CSSize) { - unsigned Opc = Is64Bit ? X86::LEA64r : X86::LEA32r; - MachineInstr *MI = - addRegOffset(BuildMI(MF, DL, TII.get(Opc), StackPtr), - FramePtr, false, -CSSize); - MBB.insert(MBBI, MI); - } else { - BuildMI(MBB, MBBI, DL, - TII.get(Is64Bit ? X86::MOV64rr : X86::MOV32rr), StackPtr) - .addReg(FramePtr); - } - } else if (NumBytes) { - // Adjust stack pointer back: ESP += numbytes. - emitSPUpdate(MBB, MBBI, StackPtr, NumBytes, Is64Bit, TII); - } - - // We're returning from function via eh_return. - if (RetOpcode == X86::EH_RETURN || RetOpcode == X86::EH_RETURN64) { - MBBI = prior(MBB.end()); - MachineOperand &DestAddr = MBBI->getOperand(0); - assert(DestAddr.isReg() && "Offset should be in register!"); - BuildMI(MBB, MBBI, DL, - TII.get(Is64Bit ? X86::MOV64rr : X86::MOV32rr), - StackPtr).addReg(DestAddr.getReg()); - } else if (RetOpcode == X86::TCRETURNri || RetOpcode == X86::TCRETURNdi || - RetOpcode == X86::TCRETURNmi || - RetOpcode == X86::TCRETURNri64 || RetOpcode == X86::TCRETURNdi64 || - RetOpcode == X86::TCRETURNmi64) { - bool isMem = RetOpcode == X86::TCRETURNmi || RetOpcode == X86::TCRETURNmi64; - // Tail call return: adjust the stack pointer and jump to callee. - MBBI = prior(MBB.end()); - MachineOperand &JumpTarget = MBBI->getOperand(0); - MachineOperand &StackAdjust = MBBI->getOperand(isMem ? 5 : 1); - assert(StackAdjust.isImm() && "Expecting immediate value."); - - // Adjust stack pointer. - int StackAdj = StackAdjust.getImm(); - int MaxTCDelta = X86FI->getTCReturnAddrDelta(); - int Offset = 0; - assert(MaxTCDelta <= 0 && "MaxTCDelta should never be positive"); - - // Incoporate the retaddr area. - Offset = StackAdj-MaxTCDelta; - assert(Offset >= 0 && "Offset should never be negative"); - - if (Offset) { - // Check for possible merge with preceeding ADD instruction. - Offset += mergeSPUpdates(MBB, MBBI, StackPtr, true); - emitSPUpdate(MBB, MBBI, StackPtr, Offset, Is64Bit, TII); - } - - // Jump to label or value in register. - if (RetOpcode == X86::TCRETURNdi || RetOpcode == X86::TCRETURNdi64) { - BuildMI(MBB, MBBI, DL, TII.get((RetOpcode == X86::TCRETURNdi) - ? X86::TAILJMPd : X86::TAILJMPd64)). - addGlobalAddress(JumpTarget.getGlobal(), JumpTarget.getOffset(), - JumpTarget.getTargetFlags()); - } else if (RetOpcode == X86::TCRETURNmi || RetOpcode == X86::TCRETURNmi64) { - MachineInstrBuilder MIB = - BuildMI(MBB, MBBI, DL, TII.get((RetOpcode == X86::TCRETURNmi) - ? X86::TAILJMPm : X86::TAILJMPm64)); - for (unsigned i = 0; i != 5; ++i) - MIB.addOperand(MBBI->getOperand(i)); - } else if (RetOpcode == X86::TCRETURNri64) { - BuildMI(MBB, MBBI, DL, TII.get(X86::TAILJMPr64)). - addReg(JumpTarget.getReg(), RegState::Kill); - } else { - BuildMI(MBB, MBBI, DL, TII.get(X86::TAILJMPr)). - addReg(JumpTarget.getReg(), RegState::Kill); - } - - MachineInstr *NewMI = prior(MBBI); - for (unsigned i = 2, e = MBBI->getNumOperands(); i != e; ++i) - NewMI->addOperand(MBBI->getOperand(i)); - - // Delete the pseudo instruction TCRETURN. - MBB.erase(MBBI); - } else if ((RetOpcode == X86::RET || RetOpcode == X86::RETI) && - (X86FI->getTCReturnAddrDelta() < 0)) { - // Add the return addr area delta back since we are not tail calling. - int delta = -1*X86FI->getTCReturnAddrDelta(); - MBBI = prior(MBB.end()); - - // Check for possible merge with preceeding ADD instruction. - delta += mergeSPUpdates(MBB, MBBI, StackPtr, true); - emitSPUpdate(MBB, MBBI, StackPtr, delta, Is64Bit, TII); - } -} - unsigned X86RegisterInfo::getRARegister() const { return Is64Bit ? X86::RIP // Should have dwarf #16. : X86::EIP; // Should have dwarf #8. } unsigned X86RegisterInfo::getFrameRegister(const MachineFunction &MF) const { - return hasFP(MF) ? FramePtr : StackPtr; -} - -void -X86RegisterInfo::getInitialFrameState(std::vector<MachineMove> &Moves) const { - // Calculate amount of bytes used for return address storing - int stackGrowth = (Is64Bit ? -8 : -4); - - // Initial state of the frame pointer is esp+stackGrowth. - MachineLocation Dst(MachineLocation::VirtualFP); - MachineLocation Src(StackPtr, stackGrowth); - Moves.push_back(MachineMove(0, Dst, Src)); - - // Add return address to move list - MachineLocation CSDst(StackPtr, stackGrowth); - MachineLocation CSSrc(getRARegister()); - Moves.push_back(MachineMove(0, CSDst, CSSrc)); + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + return TFI->hasFP(MF) ? FramePtr : StackPtr; } unsigned X86RegisterInfo::getEHExceptionRegister() const { @@ -1579,13 +820,13 @@ namespace { // Be over-conservative: scan over all vreg defs and find whether vector // registers are used. If yes, there is a possibility that vector register // will be spilled and thus require dynamic stack realignment. - for (unsigned RegNum = TargetRegisterInfo::FirstVirtualRegister; - RegNum < RI.getLastVirtReg(); ++RegNum) - if (RI.getRegClass(RegNum)->getAlignment() > StackAlignment) { + for (unsigned i = 0, e = RI.getNumVirtRegs(); i != e; ++i) { + unsigned Reg = TargetRegisterInfo::index2VirtReg(i); + if (RI.getRegClass(Reg)->getAlignment() > StackAlignment) { FuncInfo->setReserveFP(true); return true; } - + } // Nothing to do return false; } |
