diff options
Diffstat (limited to 'lib/Target/X86/X86FrameInfo.cpp')
-rw-r--r-- | lib/Target/X86/X86FrameInfo.cpp | 736 |
1 files changed, 736 insertions, 0 deletions
diff --git a/lib/Target/X86/X86FrameInfo.cpp b/lib/Target/X86/X86FrameInfo.cpp new file mode 100644 index 0000000..98c579b --- /dev/null +++ b/lib/Target/X86/X86FrameInfo.cpp @@ -0,0 +1,736 @@ +//=======- X86FrameInfo.cpp - X86 Frame Information ------------*- C++ -*-====// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file contains the X86 implementation of TargetFrameInfo class. +// +//===----------------------------------------------------------------------===// + +#include "X86FrameInfo.h" +#include "X86InstrBuilder.h" +#include "X86InstrInfo.h" +#include "X86MachineFunctionInfo.h" +#include "llvm/Function.h" +#include "llvm/CodeGen/MachineFrameInfo.h" +#include "llvm/CodeGen/MachineFunction.h" +#include "llvm/CodeGen/MachineInstrBuilder.h" +#include "llvm/CodeGen/MachineModuleInfo.h" +#include "llvm/CodeGen/MachineRegisterInfo.h" +#include "llvm/Target/TargetData.h" +#include "llvm/Target/TargetOptions.h" +#include "llvm/Support/CommandLine.h" + +using namespace llvm; + +// FIXME: completely move here. +extern cl::opt<bool> ForceStackAlign; + +static unsigned getSUBriOpcode(unsigned is64Bit, int64_t Imm) { + if (is64Bit) { + if (isInt<8>(Imm)) + return X86::SUB64ri8; + return X86::SUB64ri32; + } else { + if (isInt<8>(Imm)) + return X86::SUB32ri8; + return X86::SUB32ri; + } +} + +static unsigned getADDriOpcode(unsigned is64Bit, int64_t Imm) { + if (is64Bit) { + if (isInt<8>(Imm)) + return X86::ADD64ri8; + return X86::ADD64ri32; + } else { + if (isInt<8>(Imm)) + return X86::ADD32ri8; + return X86::ADD32ri; + } +} + +/// 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; +} + +static bool isEAXLiveIn(MachineFunction &MF) { + for (MachineRegisterInfo::livein_iterator II = MF.getRegInfo().livein_begin(), + EE = MF.getRegInfo().livein_end(); II != EE; ++II) { + unsigned Reg = II->first; + + if (Reg == X86::EAX || Reg == X86::AX || + Reg == X86::AH || Reg == X86::AL) + return true; + } + + return false; +} + +void X86FrameInfo::emitCalleeSavedFrameMoves(MachineFunction &MF, + MCSymbol *Label, + unsigned FramePtr) const { + MachineFrameInfo *MFI = MF.getFrameInfo(); + const X86RegisterInfo *RegInfo = + static_cast<const X86RegisterInfo*>(MF.getTarget().getRegisterInfo()); + 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 = RegInfo->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 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 X86Subtarget *Subtarget = &MF.getTarget().getSubtarget<X86Subtarget>(); + const X86RegisterInfo *RegInfo = + static_cast<const X86RegisterInfo*>(MF.getTarget().getRegisterInfo()); + const X86InstrInfo &TII = + *static_cast<const X86InstrInfo*>(MF.getTarget().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 = RegInfo->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 && + (Subtarget->isTargetCygMing() || Subtarget->isTargetWin32())) { + // Check whether EAX is livein for this function. + bool isEAXAlive = isEAXLiveIn(MF); + + const char *StackProbeSymbol = + Subtarget->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 && Subtarget->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 X86FrameInfo::emitEpilogue(MachineFunction &MF, + MachineBasicBlock &MBB) const { + const MachineFrameInfo *MFI = MF.getFrameInfo(); + X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); + const X86RegisterInfo *RegInfo = + static_cast<const X86RegisterInfo*>(MF.getTarget().getRegisterInfo()); + const X86InstrInfo &TII = + *static_cast<const X86InstrInfo*>(MF.getTarget().getInstrInfo()); + MachineBasicBlock::iterator MBBI = prior(MBB.end()); + unsigned RetOpcode = MBBI->getOpcode(); + DebugLoc DL = MBBI->getDebugLoc(); + bool Is64Bit = STI.is64Bit(); + unsigned StackAlign = getStackAlignment(); + unsigned SlotSize = RegInfo->getSlotSize(); + unsigned FramePtr = RegInfo->getFrameRegister(MF); + unsigned StackPtr = RegInfo->getStackRegister(); + + 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 (RegInfo->hasFP(MF)) { + // Calculate required stack adjustment. + uint64_t FrameSize = StackSize - SlotSize; + if (RegInfo->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 (RegInfo->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); + } +} |