diff options
Diffstat (limited to 'gnu/llvm/lib/Target/X86/X86FrameLowering.cpp')
| -rw-r--r-- | gnu/llvm/lib/Target/X86/X86FrameLowering.cpp | 1123 |
1 files changed, 392 insertions, 731 deletions
diff --git a/gnu/llvm/lib/Target/X86/X86FrameLowering.cpp b/gnu/llvm/lib/Target/X86/X86FrameLowering.cpp index 11808f8995f..3348e56bc77 100644 --- a/gnu/llvm/lib/Target/X86/X86FrameLowering.cpp +++ b/gnu/llvm/lib/Target/X86/X86FrameLowering.cpp @@ -29,8 +29,8 @@ #include "llvm/IR/Function.h" #include "llvm/MC/MCAsmInfo.h" #include "llvm/MC/MCSymbol.h" -#include "llvm/Support/Debug.h" #include "llvm/Target/TargetOptions.h" +#include "llvm/Support/Debug.h" #include <cstdlib> using namespace llvm; @@ -50,7 +50,7 @@ X86FrameLowering::X86FrameLowering(const X86Subtarget &STI, } bool X86FrameLowering::hasReservedCallFrame(const MachineFunction &MF) const { - return !MF.getFrameInfo().hasVarSizedObjects() && + return !MF.getFrameInfo()->hasVarSizedObjects() && !MF.getInfo<X86MachineFunctionInfo>()->getHasPushSequences(); } @@ -74,7 +74,7 @@ X86FrameLowering::canSimplifyCallFramePseudos(const MachineFunction &MF) const { // when there are no stack objects. bool X86FrameLowering::needsFrameIndexResolution(const MachineFunction &MF) const { - return MF.getFrameInfo().hasStackObjects() || + return MF.getFrameInfo()->hasStackObjects() || MF.getInfo<X86MachineFunctionInfo>()->getHasPushSequences(); } @@ -82,15 +82,17 @@ X86FrameLowering::needsFrameIndexResolution(const MachineFunction &MF) const { /// pointer register. This is true if the function has variable sized allocas /// or if frame pointer elimination is disabled. bool X86FrameLowering::hasFP(const MachineFunction &MF) const { - const MachineFrameInfo &MFI = MF.getFrameInfo(); + const MachineFrameInfo *MFI = MF.getFrameInfo(); + const MachineModuleInfo &MMI = MF.getMMI(); + return (MF.getTarget().Options.DisableFramePointerElim(MF) || TRI->needsStackRealignment(MF) || - MFI.hasVarSizedObjects() || - MFI.isFrameAddressTaken() || MFI.hasOpaqueSPAdjustment() || + MFI->hasVarSizedObjects() || + MFI->isFrameAddressTaken() || MFI->hasOpaqueSPAdjustment() || MF.getInfo<X86MachineFunctionInfo>()->getForceFramePointer() || - MF.callsUnwindInit() || MF.hasEHFunclets() || MF.callsEHReturn() || - MFI.hasStackMap() || MFI.hasPatchPoint() || - MFI.hasCopyImplyingStackAdjustment()); + MMI.callsUnwindInit() || MMI.hasEHFunclets() || MMI.callsEHReturn() || + MFI->hasStackMap() || MFI->hasPatchPoint() || + MFI->hasCopyImplyingStackAdjustment()); } static unsigned getSUBriOpcode(unsigned IsLP64, int64_t Imm) { @@ -148,18 +150,15 @@ static unsigned findDeadCallerSavedReg(MachineBasicBlock &MBB, const X86RegisterInfo *TRI, bool Is64Bit) { const MachineFunction *MF = MBB.getParent(); - if (MF->callsEHReturn()) + const Function *F = MF->getFunction(); + if (!F || MF->getMMI().callsEHReturn()) return 0; const TargetRegisterClass &AvailableRegs = *TRI->getGPRsForTailCall(*MF); - if (MBBI == MBB.end()) - return 0; - - switch (MBBI->getOpcode()) { + unsigned Opc = MBBI->getOpcode(); + switch (Opc) { default: return 0; - case TargetOpcode::PATCHABLE_RET: - case X86::RET: case X86::RETL: case X86::RETQ: case X86::RETIL: @@ -251,76 +250,40 @@ void X86FrameLowering::emitSPUpdate(MachineBasicBlock &MBB, int64_t NumBytes, bool InEpilogue) const { bool isSub = NumBytes < 0; uint64_t Offset = isSub ? -NumBytes : NumBytes; - MachineInstr::MIFlag Flag = - isSub ? MachineInstr::FrameSetup : MachineInstr::FrameDestroy; uint64_t Chunk = (1LL << 31) - 1; DebugLoc DL = MBB.findDebugLoc(MBBI); - if (Offset > Chunk) { - // Rather than emit a long series of instructions for large offsets, - // load the offset into a register and do one sub/add - unsigned Reg = 0; - unsigned Rax = (unsigned)(Is64Bit ? X86::RAX : X86::EAX); + while (Offset) { + if (Offset > Chunk) { + // Rather than emit a long series of instructions for large offsets, + // load the offset into a register and do one sub/add + unsigned Reg = 0; - if (isSub && !isEAXLiveIn(MBB)) - Reg = Rax; - else - Reg = findDeadCallerSavedReg(MBB, MBBI, TRI, Is64Bit); - - unsigned MovRIOpc = Is64Bit ? X86::MOV64ri : X86::MOV32ri; - unsigned AddSubRROpc = - isSub ? getSUBrrOpcode(Is64Bit) : getADDrrOpcode(Is64Bit); - if (Reg) { - BuildMI(MBB, MBBI, DL, TII.get(MovRIOpc), Reg) - .addImm(Offset) - .setMIFlag(Flag); - MachineInstr *MI = BuildMI(MBB, MBBI, DL, TII.get(AddSubRROpc), StackPtr) - .addReg(StackPtr) - .addReg(Reg); - MI->getOperand(3).setIsDead(); // The EFLAGS implicit def is dead. - return; - } else if (Offset > 8 * Chunk) { - // If we would need more than 8 add or sub instructions (a >16GB stack - // frame), it's worth spilling RAX to materialize this immediate. - // pushq %rax - // movabsq +-$Offset+-SlotSize, %rax - // addq %rsp, %rax - // xchg %rax, (%rsp) - // movq (%rsp), %rsp - assert(Is64Bit && "can't have 32-bit 16GB stack frame"); - BuildMI(MBB, MBBI, DL, TII.get(X86::PUSH64r)) - .addReg(Rax, RegState::Kill) - .setMIFlag(Flag); - // Subtract is not commutative, so negate the offset and always use add. - // Subtract 8 less and add 8 more to account for the PUSH we just did. - if (isSub) - Offset = -(Offset - SlotSize); + if (isSub && !isEAXLiveIn(MBB)) + Reg = (unsigned)(Is64Bit ? X86::RAX : X86::EAX); else - Offset = Offset + SlotSize; - BuildMI(MBB, MBBI, DL, TII.get(MovRIOpc), Rax) - .addImm(Offset) - .setMIFlag(Flag); - MachineInstr *MI = BuildMI(MBB, MBBI, DL, TII.get(X86::ADD64rr), Rax) - .addReg(Rax) - .addReg(StackPtr); - MI->getOperand(3).setIsDead(); // The EFLAGS implicit def is dead. - // Exchange the new SP in RAX with the top of the stack. - addRegOffset( - BuildMI(MBB, MBBI, DL, TII.get(X86::XCHG64rm), Rax).addReg(Rax), - StackPtr, false, 0); - // Load new SP from the top of the stack into RSP. - addRegOffset(BuildMI(MBB, MBBI, DL, TII.get(X86::MOV64rm), StackPtr), - StackPtr, false, 0); - return; + Reg = findDeadCallerSavedReg(MBB, MBBI, TRI, Is64Bit); + + if (Reg) { + unsigned Opc = Is64Bit ? X86::MOV64ri : X86::MOV32ri; + BuildMI(MBB, MBBI, DL, TII.get(Opc), Reg) + .addImm(Offset); + Opc = isSub + ? getSUBrrOpcode(Is64Bit) + : getADDrrOpcode(Is64Bit); + MachineInstr *MI = BuildMI(MBB, MBBI, DL, TII.get(Opc), StackPtr) + .addReg(StackPtr) + .addReg(Reg); + MI->getOperand(3).setIsDead(); // The EFLAGS implicit def is dead. + Offset = 0; + continue; + } } - } - while (Offset) { uint64_t ThisVal = std::min(Offset, Chunk); - if (ThisVal == SlotSize) { - // Use push / pop for slot sized adjustments as a size optimization. We - // need to find a dead register when using pop. + if (ThisVal == (Is64Bit ? 8 : 4)) { + // Use push / pop instead. unsigned Reg = isSub ? (unsigned)(Is64Bit ? X86::RAX : X86::EAX) : findDeadCallerSavedReg(MBB, MBBI, TRI, Is64Bit); @@ -328,24 +291,31 @@ void X86FrameLowering::emitSPUpdate(MachineBasicBlock &MBB, unsigned Opc = isSub ? (Is64Bit ? X86::PUSH64r : X86::PUSH32r) : (Is64Bit ? X86::POP64r : X86::POP32r); - BuildMI(MBB, MBBI, DL, TII.get(Opc)) - .addReg(Reg, getDefRegState(!isSub) | getUndefRegState(isSub)) - .setMIFlag(Flag); + MachineInstr *MI = BuildMI(MBB, MBBI, DL, TII.get(Opc)) + .addReg(Reg, getDefRegState(!isSub) | getUndefRegState(isSub)); + if (isSub) + MI->setFlag(MachineInstr::FrameSetup); + else + MI->setFlag(MachineInstr::FrameDestroy); Offset -= ThisVal; continue; } } - BuildStackAdjustment(MBB, MBBI, DL, isSub ? -ThisVal : ThisVal, InEpilogue) - .setMIFlag(Flag); + MachineInstrBuilder MI = BuildStackAdjustment( + MBB, MBBI, DL, isSub ? -ThisVal : ThisVal, InEpilogue); + if (isSub) + MI.setMIFlag(MachineInstr::FrameSetup); + else + MI.setMIFlag(MachineInstr::FrameDestroy); Offset -= ThisVal; } } MachineInstrBuilder X86FrameLowering::BuildStackAdjustment( - MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, int64_t Offset, bool InEpilogue) const { + MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI, DebugLoc DL, + int64_t Offset, bool InEpilogue) const { assert(Offset != 0 && "zero offset stack adjustment requested"); // On Atom, using LEA to adjust SP is preferred, but using it in the epilogue @@ -401,40 +371,19 @@ int X86FrameLowering::mergeSPUpdates(MachineBasicBlock &MBB, MachineBasicBlock::iterator PI = doMergeWithPrevious ? std::prev(MBBI) : MBBI; MachineBasicBlock::iterator NI = doMergeWithPrevious ? nullptr : std::next(MBBI); - PI = skipDebugInstructionsBackward(PI, MBB.begin()); - if (NI != nullptr) - NI = skipDebugInstructionsForward(NI, MBB.end()); - unsigned Opc = PI->getOpcode(); int Offset = 0; - if (!doMergeWithPrevious && NI != MBB.end() && - NI->getOpcode() == TargetOpcode::CFI_INSTRUCTION) { - // Don't merge with the next instruction if it has CFI. - return Offset; - } - if ((Opc == X86::ADD64ri32 || Opc == X86::ADD64ri8 || - Opc == X86::ADD32ri || Opc == X86::ADD32ri8) && + Opc == X86::ADD32ri || Opc == X86::ADD32ri8 || + Opc == X86::LEA32r || Opc == X86::LEA64_32r) && PI->getOperand(0).getReg() == StackPtr){ - assert(PI->getOperand(1).getReg() == StackPtr); Offset += PI->getOperand(2).getImm(); MBB.erase(PI); if (!doMergeWithPrevious) MBBI = NI; - } else if ((Opc == X86::LEA32r || Opc == X86::LEA64_32r) && - PI->getOperand(0).getReg() == StackPtr && - PI->getOperand(1).getReg() == StackPtr && - PI->getOperand(2).getImm() == 1 && - PI->getOperand(3).getReg() == X86::NoRegister && - PI->getOperand(5).getReg() == X86::NoRegister) { - // For LEAs we have: def = lea SP, FI, noreg, Offset, noreg. - Offset += PI->getOperand(4).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) { - assert(PI->getOperand(1).getReg() == StackPtr); Offset -= PI->getOperand(2).getImm(); MBB.erase(PI); if (!doMergeWithPrevious) MBBI = NI; @@ -444,31 +393,31 @@ int X86FrameLowering::mergeSPUpdates(MachineBasicBlock &MBB, } void X86FrameLowering::BuildCFI(MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, - const MCCFIInstruction &CFIInst) const { + MachineBasicBlock::iterator MBBI, DebugLoc DL, + MCCFIInstruction CFIInst) const { MachineFunction &MF = *MBB.getParent(); - unsigned CFIIndex = MF.addFrameInst(CFIInst); + unsigned CFIIndex = MF.getMMI().addFrameInst(CFIInst); BuildMI(MBB, MBBI, DL, TII.get(TargetOpcode::CFI_INSTRUCTION)) .addCFIIndex(CFIIndex); } -void X86FrameLowering::emitCalleeSavedFrameMoves( - MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI, - const DebugLoc &DL) const { +void +X86FrameLowering::emitCalleeSavedFrameMoves(MachineBasicBlock &MBB, + MachineBasicBlock::iterator MBBI, + DebugLoc DL) const { MachineFunction &MF = *MBB.getParent(); - MachineFrameInfo &MFI = MF.getFrameInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); MachineModuleInfo &MMI = MF.getMMI(); const MCRegisterInfo *MRI = MMI.getContext().getRegisterInfo(); // Add callee saved registers to move list. - const std::vector<CalleeSavedInfo> &CSI = MFI.getCalleeSavedInfo(); + const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo(); if (CSI.empty()) return; // Calculate offsets. for (std::vector<CalleeSavedInfo>::const_iterator I = CSI.begin(), E = CSI.end(); I != E; ++I) { - int64_t Offset = MFI.getObjectOffset(I->getFrameIdx()); + int64_t Offset = MFI->getObjectOffset(I->getFrameIdx()); unsigned Reg = I->getReg(); unsigned DwarfReg = MRI->getDwarfRegNum(Reg, true); @@ -477,19 +426,20 @@ void X86FrameLowering::emitCalleeSavedFrameMoves( } } -void X86FrameLowering::emitStackProbe(MachineFunction &MF, - MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, bool InProlog) const { +MachineInstr *X86FrameLowering::emitStackProbe(MachineFunction &MF, + MachineBasicBlock &MBB, + MachineBasicBlock::iterator MBBI, + DebugLoc DL, + bool InProlog) const { const X86Subtarget &STI = MF.getSubtarget<X86Subtarget>(); if (STI.isTargetWindowsCoreCLR()) { if (InProlog) { - emitStackProbeInlineStub(MF, MBB, MBBI, DL, true); + return emitStackProbeInlineStub(MF, MBB, MBBI, DL, true); } else { - emitStackProbeInline(MF, MBB, MBBI, DL, false); + return emitStackProbeInline(MF, MBB, MBBI, DL, false); } } else { - emitStackProbeCall(MF, MBB, MBBI, DL, InProlog); + return emitStackProbeCall(MF, MBB, MBBI, DL, InProlog); } } @@ -507,22 +457,18 @@ void X86FrameLowering::inlineStackProbe(MachineFunction &MF, } if (ChkStkStub != nullptr) { - assert(!ChkStkStub->isBundled() && - "Not expecting bundled instructions here"); MachineBasicBlock::iterator MBBI = std::next(ChkStkStub->getIterator()); - assert(std::prev(MBBI) == ChkStkStub && - "MBBI expected after __chkstk_stub."); + assert(std::prev(MBBI).operator==(ChkStkStub) && + "MBBI expected after __chkstk_stub."); DebugLoc DL = PrologMBB.findDebugLoc(MBBI); emitStackProbeInline(MF, PrologMBB, MBBI, DL, true); ChkStkStub->eraseFromParent(); } } -void X86FrameLowering::emitStackProbeInline(MachineFunction &MF, - MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, - bool InProlog) const { +MachineInstr *X86FrameLowering::emitStackProbeInline( + MachineFunction &MF, MachineBasicBlock &MBB, + MachineBasicBlock::iterator MBBI, DebugLoc DL, bool InProlog) const { const X86Subtarget &STI = MF.getSubtarget<X86Subtarget>(); assert(STI.is64Bit() && "different expansion needed for 32 bit"); assert(STI.isTargetWindowsCoreCLR() && "custom expansion expects CoreCLR"); @@ -645,7 +591,7 @@ void X86FrameLowering::emitStackProbeInline(MachineFunction &MF, // lowest touched page on the stack, not the point at which the OS // will cause an overflow exception, so this is just an optimization // to avoid unnecessarily touching pages that are below the current - // SP but already committed to the stack by the OS. + // SP but already commited to the stack by the OS. BuildMI(&MBB, DL, TII.get(X86::MOV64rm), LimitReg) .addReg(0) .addImm(1) @@ -732,27 +678,32 @@ void X86FrameLowering::emitStackProbeInline(MachineFunction &MF, } // Possible TODO: physreg liveness for InProlog case. + + return ContinueMBBI; } -void X86FrameLowering::emitStackProbeCall(MachineFunction &MF, - MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, - bool InProlog) const { +MachineInstr *X86FrameLowering::emitStackProbeCall( + MachineFunction &MF, MachineBasicBlock &MBB, + MachineBasicBlock::iterator MBBI, DebugLoc DL, bool InProlog) const { bool IsLargeCodeModel = MF.getTarget().getCodeModel() == CodeModel::Large; - // FIXME: Add retpoline support and remove this. - if (Is64Bit && IsLargeCodeModel && STI.useRetpoline()) - report_fatal_error("Emitting stack probe calls on 64-bit with the large " - "code model and retpoline not yet implemented."); - unsigned CallOp; if (Is64Bit) CallOp = IsLargeCodeModel ? X86::CALL64r : X86::CALL64pcrel32; else CallOp = X86::CALLpcrel32; - StringRef Symbol = STI.getTargetLowering()->getStackProbeSymbolName(MF); + const char *Symbol; + if (Is64Bit) { + if (STI.isTargetCygMing()) { + Symbol = "___chkstk_ms"; + } else { + Symbol = "__chkstk"; + } + } else if (STI.isTargetCygMing()) + Symbol = "_alloca"; + else + Symbol = "_chkstk"; MachineInstrBuilder CI; MachineBasicBlock::iterator ExpansionMBBI = std::prev(MBBI); @@ -763,11 +714,10 @@ void X86FrameLowering::emitStackProbeCall(MachineFunction &MF, // For the large code model, we have to call through a register. Use R11, // as it is scratch in all supported calling conventions. BuildMI(MBB, MBBI, DL, TII.get(X86::MOV64ri), X86::R11) - .addExternalSymbol(MF.createExternalSymbolName(Symbol)); + .addExternalSymbol(Symbol); CI = BuildMI(MBB, MBBI, DL, TII.get(CallOp)).addReg(X86::R11); } else { - CI = BuildMI(MBB, MBBI, DL, TII.get(CallOp)) - .addExternalSymbol(MF.createExternalSymbolName(Symbol)); + CI = BuildMI(MBB, MBBI, DL, TII.get(CallOp)).addExternalSymbol(Symbol); } unsigned AX = Is64Bit ? X86::RAX : X86::EAX; @@ -778,16 +728,13 @@ void X86FrameLowering::emitStackProbeCall(MachineFunction &MF, .addReg(SP, RegState::Define | RegState::Implicit) .addReg(X86::EFLAGS, RegState::Define | RegState::Implicit); - if (STI.isTargetWin64() || !STI.isOSWindows()) { - // MSVC x32's _chkstk and cygwin/mingw's _alloca adjust %esp themselves. + if (Is64Bit) { // MSVC x64's __chkstk and cygwin/mingw's ___chkstk_ms do not adjust %rsp - // themselves. They also does not clobber %rax so we can reuse it when + // themselves. It also does not clobber %rax so we can reuse it when // adjusting %rsp. - // All other platforms do not specify a particular ABI for the stack probe - // function, so we arbitrarily define it to not adjust %esp/%rsp itself. - BuildMI(MBB, MBBI, DL, TII.get(getSUBrrOpcode(Is64Bit)), SP) - .addReg(SP) - .addReg(AX); + BuildMI(MBB, MBBI, DL, TII.get(X86::SUB64rr), X86::RSP) + .addReg(X86::RSP) + .addReg(X86::RAX); } if (InProlog) { @@ -795,16 +742,20 @@ void X86FrameLowering::emitStackProbeCall(MachineFunction &MF, for (++ExpansionMBBI; ExpansionMBBI != MBBI; ++ExpansionMBBI) ExpansionMBBI->setFlag(MachineInstr::FrameSetup); } + + return MBBI; } -void X86FrameLowering::emitStackProbeInlineStub( +MachineInstr *X86FrameLowering::emitStackProbeInlineStub( MachineFunction &MF, MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, const DebugLoc &DL, bool InProlog) const { + MachineBasicBlock::iterator MBBI, DebugLoc DL, bool InProlog) const { assert(InProlog && "ChkStkStub called outside prolog!"); BuildMI(MBB, MBBI, DL, TII.get(X86::CALLpcrel32)) .addExternalSymbol("__chkstk_stub"); + + return MBBI; } static unsigned calculateSetFPREG(uint64_t SPAdjust) { @@ -821,11 +772,11 @@ static unsigned calculateSetFPREG(uint64_t SPAdjust) { // have a call out. Otherwise just make sure we have some alignment - we'll // go with the minimum SlotSize. uint64_t X86FrameLowering::calculateMaxStackAlign(const MachineFunction &MF) const { - const MachineFrameInfo &MFI = MF.getFrameInfo(); - uint64_t MaxAlign = MFI.getMaxAlignment(); // Desired stack alignment. + const MachineFrameInfo *MFI = MF.getFrameInfo(); + uint64_t MaxAlign = MFI->getMaxAlignment(); // Desired stack alignment. unsigned StackAlign = getStackAlignment(); - if (MF.getFunction().hasFnAttribute("stackrealign")) { - if (MFI.hasCalls()) + if (MF.getFunction()->hasFnAttribute("stackrealign")) { + if (MFI->hasCalls()) MaxAlign = (StackAlign > MaxAlign) ? StackAlign : MaxAlign; else if (MaxAlign < SlotSize) MaxAlign = SlotSize; @@ -835,7 +786,7 @@ uint64_t X86FrameLowering::calculateMaxStackAlign(const MachineFunction &MF) con void X86FrameLowering::BuildStackAlignAND(MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, unsigned Reg, + DebugLoc DL, unsigned Reg, uint64_t MaxAlign) const { uint64_t Val = -MaxAlign; unsigned AndOp = getANDriOpcode(Uses64BitFramePtr, Val); @@ -928,7 +879,6 @@ void X86FrameLowering::BuildStackAlignAND(MachineBasicBlock &MBB, Notes: - .seh directives are emitted only for Windows 64 ABI - - .cv_fpo directives are emitted on win32 when emitting CodeView - .cfi directives are emitted for all other ABIs - for 32-bit code, substitute %e?? registers for %r?? */ @@ -938,36 +888,31 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, assert(&STI == &MF.getSubtarget<X86Subtarget>() && "MF used frame lowering for wrong subtarget"); MachineBasicBlock::iterator MBBI = MBB.begin(); - MachineFrameInfo &MFI = MF.getFrameInfo(); - const Function &Fn = MF.getFunction(); + MachineFrameInfo *MFI = MF.getFrameInfo(); + const Function *Fn = MF.getFunction(); MachineModuleInfo &MMI = MF.getMMI(); X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); uint64_t MaxAlign = calculateMaxStackAlign(MF); // Desired stack alignment. - uint64_t StackSize = MFI.getStackSize(); // Number of bytes to allocate. + uint64_t StackSize = MFI->getStackSize(); // Number of bytes to allocate. bool IsFunclet = MBB.isEHFuncletEntry(); EHPersonality Personality = EHPersonality::Unknown; - if (Fn.hasPersonalityFn()) - Personality = classifyEHPersonality(Fn.getPersonalityFn()); + if (Fn->hasPersonalityFn()) + Personality = classifyEHPersonality(Fn->getPersonalityFn()); bool FnHasClrFunclet = - MF.hasEHFunclets() && Personality == EHPersonality::CoreCLR; + MMI.hasEHFunclets() && Personality == EHPersonality::CoreCLR; bool IsClrFunclet = IsFunclet && FnHasClrFunclet; bool HasFP = hasFP(MF); - bool IsWin64CC = STI.isCallingConvWin64(Fn.getCallingConv()); + bool IsWin64CC = STI.isCallingConvWin64(Fn->getCallingConv()); bool IsWin64Prologue = MF.getTarget().getMCAsmInfo()->usesWindowsCFI(); - bool NeedsWin64CFI = IsWin64Prologue && Fn.needsUnwindTableEntry(); - // FIXME: Emit FPO data for EH funclets. - bool NeedsWinFPO = - !IsFunclet && STI.isTargetWin32() && MMI.getModule()->getCodeViewFlag(); - bool NeedsWinCFI = NeedsWin64CFI || NeedsWinFPO; + bool NeedsWinCFI = IsWin64Prologue && Fn->needsUnwindTableEntry(); bool NeedsDwarfCFI = - !IsWin64Prologue && (MMI.hasDebugInfo() || Fn.needsUnwindTableEntry()); + !IsWin64Prologue && (MMI.hasDebugInfo() || Fn->needsUnwindTableEntry()); unsigned FramePtr = TRI->getFrameRegister(MF); const unsigned MachineFramePtr = STI.isTarget64BitILP32() ? getX86SubSuperRegister(FramePtr, 64) : FramePtr; unsigned BasePtr = TRI->getBaseRegister(); - bool HasWinCFI = false; - + // Debug location must be unknown since the first debug location is used // to determine the end of the prologue. DebugLoc DL; @@ -981,44 +926,32 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, X86FI->setCalleeSavedFrameSize( X86FI->getCalleeSavedFrameSize() - TailCallReturnAddrDelta); - bool UseStackProbe = !STI.getTargetLowering()->getStackProbeSymbolName(MF).empty(); + bool UseStackProbe = (STI.isOSWindows() && !STI.isTargetMachO()); // The default stack probe size is 4096 if the function has no stackprobesize // attribute. unsigned StackProbeSize = 4096; - if (Fn.hasFnAttribute("stack-probe-size")) - Fn.getFnAttribute("stack-probe-size") + if (Fn->hasFnAttribute("stack-probe-size")) + Fn->getFnAttribute("stack-probe-size") .getValueAsString() .getAsInteger(0, StackProbeSize); - // Re-align the stack on 64-bit if the x86-interrupt calling convention is - // used and an error code was pushed, since the x86-64 ABI requires a 16-byte - // stack alignment. - if (Fn.getCallingConv() == CallingConv::X86_INTR && Is64Bit && - Fn.arg_size() == 2) { - StackSize += 8; - MFI.setStackSize(StackSize); - emitSPUpdate(MBB, MBBI, -8, /*InEpilogue=*/false); - } - // 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). We also check that we don't // push and pop from the stack. - if (Is64Bit && !Fn.hasFnAttribute(Attribute::NoRedZone) && + if (Is64Bit && !Fn->hasFnAttribute(Attribute::NoRedZone) && !TRI->needsStackRealignment(MF) && - !MFI.hasVarSizedObjects() && // No dynamic alloca. - !MFI.adjustsStack() && // No calls. - !UseStackProbe && // No stack probes. - !IsWin64CC && // Win64 has no Red Zone - !MFI.hasCopyImplyingStackAdjustment() && // Don't push and pop. - !MF.shouldSplitStack()) { // Regular stack + !MFI->hasVarSizedObjects() && // No dynamic alloca. + !MFI->adjustsStack() && // No calls. + !IsWin64CC && // Win64 has no Red Zone + !MFI->hasCopyImplyingStackAdjustment() && // Don't push and pop. + !MF.shouldSplitStack()) { // Regular stack uint64_t MinSize = X86FI->getCalleeSavedFrameSize(); if (HasFP) MinSize += SlotSize; - X86FI->setUsesRedZone(MinSize > 0 || StackSize > 0); StackSize = std::max(MinSize, StackSize > 128 ? StackSize - 128 : 0); - MFI.setStackSize(StackSize); + MFI->setStackSize(StackSize); } // Insert stack pointer adjustment for later moving of return addr. Only @@ -1066,8 +999,6 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, } if (HasFP) { - assert(MF.getRegInfo().isReserved(MachineFramePtr) && "FP reserved"); - // Calculate required stack adjustment. uint64_t FrameSize = StackSize - SlotSize; // If required, include space for extra hidden slot for stashing base pointer. @@ -1078,15 +1009,15 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, // Callee-saved registers are pushed on stack before the stack is realigned. if (TRI->needsStackRealignment(MF) && !IsWin64Prologue) - NumBytes = alignTo(NumBytes, MaxAlign); + NumBytes = RoundUpToAlignment(NumBytes, MaxAlign); // 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. if (!IsFunclet) - MFI.setOffsetAdjustment(-NumBytes); + MFI->setOffsetAdjustment(-NumBytes); else - assert(MFI.getOffsetAdjustment() == -(int)NumBytes && + assert(MFI->getOffsetAdjustment() == -(int)NumBytes && "should calculate same local variable offset for funclets"); // Save EBP/RBP into the appropriate stack slot. @@ -1108,7 +1039,6 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, } if (NeedsWinCFI) { - HasWinCFI = true; BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_PushReg)) .addImm(FramePtr) .setMIFlag(MachineInstr::FrameSetup); @@ -1129,15 +1059,13 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, BuildCFI(MBB, MBBI, DL, MCCFIInstruction::createDefCfaRegister( nullptr, DwarfFramePtr)); } + } - if (NeedsWinFPO) { - // .cv_fpo_setframe $FramePtr - HasWinCFI = true; - BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_SetFrame)) - .addImm(FramePtr) - .addImm(0) - .setMIFlag(MachineInstr::FrameSetup); - } + // Mark the FramePtr as live-in in every block. Don't do this again for + // funclet prologues. + if (!IsFunclet) { + for (MachineBasicBlock &EveryMBB : MF) + EveryMBB.addLiveIn(MachineFramePtr); } } else { assert(!IsFunclet && "funclets without FPs not yet implemented"); @@ -1172,10 +1100,8 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, } if (NeedsWinCFI) { - HasWinCFI = true; - BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_PushReg)) - .addImm(Reg) - .setMIFlag(MachineInstr::FrameSetup); + BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_PushReg)).addImm(Reg).setMIFlag( + MachineInstr::FrameSetup); } } @@ -1204,11 +1130,8 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, // virtual memory manager are allocated in correct sequence. uint64_t AlignedNumBytes = NumBytes; if (IsWin64Prologue && !IsFunclet && TRI->needsStackRealignment(MF)) - AlignedNumBytes = alignTo(AlignedNumBytes, MaxAlign); + AlignedNumBytes = RoundUpToAlignment(AlignedNumBytes, MaxAlign); if (AlignedNumBytes >= StackProbeSize && UseStackProbe) { - assert(!X86FI->getUsesRedZone() && - "The Red Zone is not accounted for in stack probes"); - // Check whether EAX is livein for this block. bool isEAXAlive = isEAXLiveIn(MBB); @@ -1262,12 +1185,10 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, emitSPUpdate(MBB, MBBI, -(int64_t)NumBytes, /*InEpilogue=*/false); } - if (NeedsWinCFI && NumBytes) { - HasWinCFI = true; + if (NeedsWinCFI && NumBytes) BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_StackAlloc)) .addImm(NumBytes) .setMIFlag(MachineInstr::FrameSetup); - } int SEHFrameOffset = 0; unsigned SPOrEstablisher; @@ -1314,8 +1235,6 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, // If this is not a funclet, emit the CFI describing our frame pointer. if (NeedsWinCFI && !IsFunclet) { - assert(!NeedsWinFPO && "this setframe incompatible with FPO data"); - HasWinCFI = true; BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_SetFrame)) .addImm(FramePtr) .addImm(SEHFrameOffset) @@ -1341,7 +1260,7 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, } while (MBBI != MBB.end() && MBBI->getFlag(MachineInstr::FrameSetup)) { - const MachineInstr &FrameInstr = *MBBI; + const MachineInstr *FrameInstr = &*MBBI; ++MBBI; if (NeedsWinCFI) { @@ -1352,8 +1271,6 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, int Offset = getFrameIndexReference(MF, FI, IgnoredFrameReg); Offset += SEHFrameOffset; - HasWinCFI = true; - assert(!NeedsWinFPO && "SEH_SaveXMM incompatible with FPO data"); BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_SaveXMM)) .addImm(Reg) .addImm(Offset) @@ -1363,7 +1280,7 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, } } - if (NeedsWinCFI && HasWinCFI) + if (NeedsWinCFI) BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_EndPrologue)) .setMIFlag(MachineInstr::FrameSetup); @@ -1440,38 +1357,24 @@ void X86FrameLowering::emitPrologue(MachineFunction &MF, } // Emit DWARF info specifying the offsets of the callee-saved registers. - emitCalleeSavedFrameMoves(MBB, MBBI, DL); + if (PushedRegs) + emitCalleeSavedFrameMoves(MBB, MBBI, DL); } - - // X86 Interrupt handling function cannot assume anything about the direction - // flag (DF in EFLAGS register). Clear this flag by creating "cld" instruction - // in each prologue of interrupt handler function. - // - // FIXME: Create "cld" instruction only in these cases: - // 1. The interrupt handling function uses any of the "rep" instructions. - // 2. Interrupt handling function calls another function. - // - if (Fn.getCallingConv() == CallingConv::X86_INTR) - BuildMI(MBB, MBBI, DL, TII.get(X86::CLD)) - .setMIFlag(MachineInstr::FrameSetup); - - // At this point we know if the function has WinCFI or not. - MF.setHasWinCFI(HasWinCFI); } bool X86FrameLowering::canUseLEAForSPInEpilogue( const MachineFunction &MF) const { - // We can't use LEA instructions for adjusting the stack pointer if we don't - // have a frame pointer in the Win64 ABI. Only ADD instructions may be used - // to deallocate the stack. + // We can't use LEA instructions for adjusting the stack pointer if this is a + // leaf function in the Win64 ABI. Only ADD instructions may be used to + // deallocate the stack. // This means that we can use LEA for SP in two situations: // 1. We *aren't* using the Win64 ABI which means we are free to use LEA. // 2. We *have* a frame pointer which means we are permitted to use LEA. return !MF.getTarget().getMCAsmInfo()->usesWindowsCFI() || hasFP(MF); } -static bool isFuncletReturnInstr(MachineInstr &MI) { - switch (MI.getOpcode()) { +static bool isFuncletReturnInstr(MachineInstr *MI) { + switch (MI->getOpcode()) { case X86::CATCHRET: case X86::CLEANUPRET: return true; @@ -1497,10 +1400,11 @@ static bool isFuncletReturnInstr(MachineInstr &MI) { unsigned X86FrameLowering::getPSPSlotOffsetFromSP(const MachineFunction &MF) const { const WinEHFuncInfo &Info = *MF.getWinEHFuncInfo(); + // getFrameIndexReferenceFromSP has an out ref parameter for the stack + // pointer register; pass a dummy that we ignore unsigned SPReg; - int Offset = getFrameIndexReferencePreferSP(MF, Info.PSPSymFrameIdx, SPReg, - /*IgnoreSPUpdates*/ true); - assert(Offset >= 0 && SPReg == TRI->getStackRegister()); + int Offset = getFrameIndexReferenceFromSP(MF, Info.PSPSymFrameIdx, SPReg); + assert(Offset >= 0); return static_cast<unsigned>(Offset); } @@ -1512,7 +1416,7 @@ X86FrameLowering::getWinEHFuncletFrameSize(const MachineFunction &MF) const { // This is the amount of stack a funclet needs to allocate. unsigned UsedSize; EHPersonality Personality = - classifyEHPersonality(MF.getFunction().getPersonalityFn()); + classifyEHPersonality(MF.getFunction()->getPersonalityFn()); if (Personality == EHPersonality::CoreCLR) { // CLR funclets need to hold enough space to include the PSPSym, at the // same offset from the stack pointer (immediately after the prolog) as it @@ -1520,30 +1424,23 @@ X86FrameLowering::getWinEHFuncletFrameSize(const MachineFunction &MF) const { UsedSize = getPSPSlotOffsetFromSP(MF) + SlotSize; } else { // Other funclets just need enough stack for outgoing call arguments. - UsedSize = MF.getFrameInfo().getMaxCallFrameSize(); + UsedSize = MF.getFrameInfo()->getMaxCallFrameSize(); } // RBP is not included in the callee saved register block. After pushing RBP, // everything is 16 byte aligned. Everything we allocate before an outgoing // call must also be 16 byte aligned. - unsigned FrameSizeMinusRBP = alignTo(CSSize + UsedSize, getStackAlignment()); + unsigned FrameSizeMinusRBP = + RoundUpToAlignment(CSSize + UsedSize, getStackAlignment()); // Subtract out the size of the callee saved registers. This is how much stack // each funclet will allocate. return FrameSizeMinusRBP - CSSize; } -static bool isTailCallOpcode(unsigned Opc) { - return Opc == X86::TCRETURNri || Opc == X86::TCRETURNdi || - Opc == X86::TCRETURNmi || - Opc == X86::TCRETURNri64 || Opc == X86::TCRETURNdi64 || - Opc == X86::TCRETURNmi64; -} - void X86FrameLowering::emitEpilogue(MachineFunction &MF, MachineBasicBlock &MBB) const { - const MachineFrameInfo &MFI = MF.getFrameInfo(); + const MachineFrameInfo *MFI = MF.getFrameInfo(); X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); - MachineBasicBlock::iterator Terminator = MBB.getFirstTerminator(); - MachineBasicBlock::iterator MBBI = Terminator; + MachineBasicBlock::iterator MBBI = MBB.getFirstTerminator(); DebugLoc DL; if (MBBI != MBB.end()) DL = MBBI->getDebugLoc(); @@ -1554,21 +1451,38 @@ void X86FrameLowering::emitEpilogue(MachineFunction &MF, Is64BitILP32 ? getX86SubSuperRegister(FramePtr, 64) : FramePtr; bool IsWin64Prologue = MF.getTarget().getMCAsmInfo()->usesWindowsCFI(); - bool NeedsWin64CFI = - IsWin64Prologue && MF.getFunction().needsUnwindTableEntry(); - bool IsFunclet = MBBI == MBB.end() ? false : isFuncletReturnInstr(*MBBI); + bool NeedsWinCFI = + IsWin64Prologue && MF.getFunction()->needsUnwindTableEntry(); + bool IsFunclet = isFuncletReturnInstr(MBBI); + MachineBasicBlock *TargetMBB = nullptr; // Get the number of bytes to allocate from the FrameInfo. - uint64_t StackSize = MFI.getStackSize(); + uint64_t StackSize = MFI->getStackSize(); uint64_t MaxAlign = calculateMaxStackAlign(MF); unsigned CSSize = X86FI->getCalleeSavedFrameSize(); - bool HasFP = hasFP(MF); uint64_t NumBytes = 0; - if (IsFunclet) { - assert(HasFP && "EH funclets without FP not yet implemented"); + if (MBBI->getOpcode() == X86::CATCHRET) { + // SEH shouldn't use catchret. + assert(!isAsynchronousEHPersonality( + classifyEHPersonality(MF.getFunction()->getPersonalityFn())) && + "SEH should not use CATCHRET"); + NumBytes = getWinEHFuncletFrameSize(MF); - } else if (HasFP) { + assert(hasFP(MF) && "EH funclets without FP not yet implemented"); + TargetMBB = MBBI->getOperand(0).getMBB(); + + // Pop EBP. + BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::POP64r : X86::POP32r), + MachineFramePtr) + .setMIFlag(MachineInstr::FrameDestroy); + } else if (MBBI->getOpcode() == X86::CLEANUPRET) { + NumBytes = getWinEHFuncletFrameSize(MF); + assert(hasFP(MF) && "EH funclets without FP not yet implemented"); + BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::POP64r : X86::POP32r), + MachineFramePtr) + .setMIFlag(MachineInstr::FrameDestroy); + } else if (hasFP(MF)) { // Calculate required stack adjustment. uint64_t FrameSize = StackSize - SlotSize; NumBytes = FrameSize - CSSize; @@ -1576,52 +1490,65 @@ void X86FrameLowering::emitEpilogue(MachineFunction &MF, // Callee-saved registers were pushed on stack before the stack was // realigned. if (TRI->needsStackRealignment(MF) && !IsWin64Prologue) - NumBytes = alignTo(FrameSize, MaxAlign); - } else { - NumBytes = StackSize - CSSize; - } - uint64_t SEHStackAllocAmt = NumBytes; + NumBytes = RoundUpToAlignment(FrameSize, MaxAlign); - if (HasFP) { // Pop EBP. - BuildMI(MBB, MBBI, DL, TII.get(Is64Bit ? X86::POP64r : X86::POP32r), - MachineFramePtr) + BuildMI(MBB, MBBI, DL, + TII.get(Is64Bit ? X86::POP64r : X86::POP32r), MachineFramePtr) .setMIFlag(MachineInstr::FrameDestroy); + } else { + NumBytes = StackSize - CSSize; } + uint64_t SEHStackAllocAmt = NumBytes; - MachineBasicBlock::iterator FirstCSPop = MBBI; // Skip the callee-saved pop instructions. while (MBBI != MBB.begin()) { MachineBasicBlock::iterator PI = std::prev(MBBI); unsigned Opc = PI->getOpcode(); - if (Opc != X86::DBG_VALUE && !PI->isTerminator()) { - if ((Opc != X86::POP32r || !PI->getFlag(MachineInstr::FrameDestroy)) && - (Opc != X86::POP64r || !PI->getFlag(MachineInstr::FrameDestroy))) - break; - FirstCSPop = PI; - } + if ((Opc != X86::POP32r || !PI->getFlag(MachineInstr::FrameDestroy)) && + (Opc != X86::POP64r || !PI->getFlag(MachineInstr::FrameDestroy)) && + Opc != X86::DBG_VALUE && !PI->isTerminator()) + break; --MBBI; } - MBBI = FirstCSPop; + MachineBasicBlock::iterator FirstCSPop = MBBI; - if (IsFunclet && Terminator->getOpcode() == X86::CATCHRET) - emitCatchRetReturnValue(MBB, FirstCSPop, &*Terminator); + if (TargetMBB) { + // Fill EAX/RAX with the address of the target block. + unsigned ReturnReg = STI.is64Bit() ? X86::RAX : X86::EAX; + if (STI.is64Bit()) { + // LEA64r TargetMBB(%rip), %rax + BuildMI(MBB, FirstCSPop, DL, TII.get(X86::LEA64r), ReturnReg) + .addReg(X86::RIP) + .addImm(0) + .addReg(0) + .addMBB(TargetMBB) + .addReg(0); + } else { + // MOV32ri $TargetMBB, %eax + BuildMI(MBB, FirstCSPop, DL, TII.get(X86::MOV32ri), ReturnReg) + .addMBB(TargetMBB); + } + // Record that we've taken the address of TargetMBB and no longer just + // reference it in a terminator. + TargetMBB->setHasAddressTaken(); + } if (MBBI != MBB.end()) DL = MBBI->getDebugLoc(); // If there is an ADD32ri or SUB32ri of ESP immediately before this // instruction, merge the two instructions. - if (NumBytes || MFI.hasVarSizedObjects()) + if (NumBytes || MFI->hasVarSizedObjects()) NumBytes += mergeSPUpdates(MBB, MBBI, true); // 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. Don't do this if this was a funclet epilogue, since the funclets // will not do realignment or dynamic stack allocation. - if ((TRI->needsStackRealignment(MF) || MFI.hasVarSizedObjects()) && + if ((TRI->needsStackRealignment(MF) || MFI->hasVarSizedObjects()) && !IsFunclet) { if (TRI->needsStackRealignment(MF)) MBBI = FirstCSPop; @@ -1659,33 +1586,36 @@ void X86FrameLowering::emitEpilogue(MachineFunction &MF, // into the epilogue. To cope with that, we insert an epilogue marker here, // then replace it with a 'nop' if it ends up immediately after a CALL in the // final emitted code. - if (NeedsWin64CFI && MF.hasWinCFI()) + if (NeedsWinCFI) BuildMI(MBB, MBBI, DL, TII.get(X86::SEH_Epilogue)); - if (Terminator == MBB.end() || !isTailCallOpcode(Terminator->getOpcode())) { - // Add the return addr area delta back since we are not tail calling. - int Offset = -1 * X86FI->getTCReturnAddrDelta(); - assert(Offset >= 0 && "TCDelta should never be positive"); - if (Offset) { - // Check for possible merge with preceding ADD instruction. - Offset += mergeSPUpdates(MBB, Terminator, true); - emitSPUpdate(MBB, Terminator, Offset, /*InEpilogue=*/true); - } + // Add the return addr area delta back since we are not tail calling. + int Offset = -1 * X86FI->getTCReturnAddrDelta(); + assert(Offset >= 0 && "TCDelta should never be positive"); + if (Offset) { + MBBI = MBB.getFirstTerminator(); + + // Check for possible merge with preceding ADD instruction. + Offset += mergeSPUpdates(MBB, MBBI, true); + emitSPUpdate(MBB, MBBI, Offset, /*InEpilogue=*/true); } } +// NOTE: this only has a subset of the full frame index logic. In +// particular, the FI < 0 and AfterFPPop logic is handled in +// X86RegisterInfo::eliminateFrameIndex, but not here. Possibly +// (probably?) it should be moved into here. int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI, unsigned &FrameReg) const { - const MachineFrameInfo &MFI = MF.getFrameInfo(); + const MachineFrameInfo *MFI = MF.getFrameInfo(); - bool IsFixed = MFI.isFixedObjectIndex(FI); // We can't calculate offset from frame pointer if the stack is realigned, // so enforce usage of stack/base pointer. The base pointer is used when we // have dynamic allocas in addition to dynamic realignment. if (TRI->hasBasePointer(MF)) - FrameReg = IsFixed ? TRI->getFramePtr() : TRI->getBaseRegister(); + FrameReg = TRI->getBaseRegister(); else if (TRI->needsStackRealignment(MF)) - FrameReg = IsFixed ? TRI->getFramePtr() : TRI->getStackRegister(); + FrameReg = TRI->getStackRegister(); else FrameReg = TRI->getFrameRegister(MF); @@ -1693,16 +1623,16 @@ int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI, // object. // We need to factor in additional offsets applied during the prologue to the // frame, base, and stack pointer depending on which is used. - int Offset = MFI.getObjectOffset(FI) - getOffsetOfLocalArea(); + int Offset = MFI->getObjectOffset(FI) - getOffsetOfLocalArea(); const X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); unsigned CSSize = X86FI->getCalleeSavedFrameSize(); - uint64_t StackSize = MFI.getStackSize(); + uint64_t StackSize = MFI->getStackSize(); bool HasFP = hasFP(MF); bool IsWin64Prologue = MF.getTarget().getMCAsmInfo()->usesWindowsCFI(); int64_t FPDelta = 0; if (IsWin64Prologue) { - assert(!MFI.hasCalls() || (StackSize % 16) == 8); + assert(!MFI->hasCalls() || (StackSize % 16) == 8); // Calculate required stack adjustment. uint64_t FrameSize = StackSize - SlotSize; @@ -1720,7 +1650,7 @@ int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI, // restricted Win64 prologue. // Add FPDelta to all offsets below that go through the frame pointer. FPDelta = FrameSize - SEHFrameOffset; - assert((!MFI.hasCalls() || (FPDelta % 16) == 0) && + assert((!MFI->hasCalls() || (FPDelta % 16) == 0) && "FPDelta isn't aligned per the Win64 ABI!"); } @@ -1731,7 +1661,7 @@ int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI, // Skip the saved EBP. return Offset + SlotSize + FPDelta; } else { - assert((-(Offset + StackSize)) % MFI.getObjectAlignment(FI) == 0); + assert((-(Offset + StackSize)) % MFI->getObjectAlignment(FI) == 0); return Offset + StackSize; } } else if (TRI->needsStackRealignment(MF)) { @@ -1739,7 +1669,7 @@ int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI, // Skip the saved EBP. return Offset + SlotSize + FPDelta; } else { - assert((-(Offset + StackSize)) % MFI.getObjectAlignment(FI) == 0); + assert((-(Offset + StackSize)) % MFI->getObjectAlignment(FI) == 0); return Offset + StackSize; } // FIXME: Support tail calls @@ -1759,69 +1689,61 @@ int X86FrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI, return Offset + FPDelta; } -int X86FrameLowering::getFrameIndexReferenceSP(const MachineFunction &MF, - int FI, unsigned &FrameReg, - int Adjustment) const { - const MachineFrameInfo &MFI = MF.getFrameInfo(); - FrameReg = TRI->getStackRegister(); - return MFI.getObjectOffset(FI) - getOffsetOfLocalArea() + Adjustment; -} - -int -X86FrameLowering::getFrameIndexReferencePreferSP(const MachineFunction &MF, - int FI, unsigned &FrameReg, - bool IgnoreSPUpdates) const { - - const MachineFrameInfo &MFI = MF.getFrameInfo(); +// Simplified from getFrameIndexReference keeping only StackPointer cases +int X86FrameLowering::getFrameIndexReferenceFromSP(const MachineFunction &MF, + int FI, + unsigned &FrameReg) const { + const MachineFrameInfo *MFI = MF.getFrameInfo(); // Does not include any dynamic realign. - const uint64_t StackSize = MFI.getStackSize(); - // LLVM arranges the stack as follows: - // ... - // ARG2 - // ARG1 - // RETADDR - // PUSH RBP <-- RBP points here - // PUSH CSRs - // ~~~~~~~ <-- possible stack realignment (non-win64) - // ... - // STACK OBJECTS - // ... <-- RSP after prologue points here - // ~~~~~~~ <-- possible stack realignment (win64) - // - // if (hasVarSizedObjects()): - // ... <-- "base pointer" (ESI/RBX) points here - // DYNAMIC ALLOCAS - // ... <-- RSP points here - // - // Case 1: In the simple case of no stack realignment and no dynamic - // allocas, both "fixed" stack objects (arguments and CSRs) are addressable - // with fixed offsets from RSP. - // - // Case 2: In the case of stack realignment with no dynamic allocas, fixed - // stack objects are addressed with RBP and regular stack objects with RSP. - // - // Case 3: In the case of dynamic allocas and stack realignment, RSP is used - // to address stack arguments for outgoing calls and nothing else. The "base - // pointer" points to local variables, and RBP points to fixed objects. - // - // In cases 2 and 3, we can only answer for non-fixed stack objects, and the - // answer we give is relative to the SP after the prologue, and not the - // SP in the middle of the function. - - if (MFI.isFixedObjectIndex(FI) && TRI->needsStackRealignment(MF) && - !STI.isTargetWin64()) - return getFrameIndexReference(MF, FI, FrameReg); - - // If !hasReservedCallFrame the function might have SP adjustement in the - // body. So, even though the offset is statically known, it depends on where - // we are in the function. - const TargetFrameLowering *TFI = MF.getSubtarget().getFrameLowering(); - if (!IgnoreSPUpdates && !TFI->hasReservedCallFrame(MF)) - return getFrameIndexReference(MF, FI, FrameReg); + const uint64_t StackSize = MFI->getStackSize(); + { +#ifndef NDEBUG + // LLVM arranges the stack as follows: + // ... + // ARG2 + // ARG1 + // RETADDR + // PUSH RBP <-- RBP points here + // PUSH CSRs + // ~~~~~~~ <-- possible stack realignment (non-win64) + // ... + // STACK OBJECTS + // ... <-- RSP after prologue points here + // ~~~~~~~ <-- possible stack realignment (win64) + // + // if (hasVarSizedObjects()): + // ... <-- "base pointer" (ESI/RBX) points here + // DYNAMIC ALLOCAS + // ... <-- RSP points here + // + // Case 1: In the simple case of no stack realignment and no dynamic + // allocas, both "fixed" stack objects (arguments and CSRs) are addressable + // with fixed offsets from RSP. + // + // Case 2: In the case of stack realignment with no dynamic allocas, fixed + // stack objects are addressed with RBP and regular stack objects with RSP. + // + // Case 3: In the case of dynamic allocas and stack realignment, RSP is used + // to address stack arguments for outgoing calls and nothing else. The "base + // pointer" points to local variables, and RBP points to fixed objects. + // + // In cases 2 and 3, we can only answer for non-fixed stack objects, and the + // answer we give is relative to the SP after the prologue, and not the + // SP in the middle of the function. + + assert((!MFI->isFixedObjectIndex(FI) || !TRI->needsStackRealignment(MF) || + STI.isTargetWin64()) && + "offset from fixed object to SP is not static"); + + // We don't handle tail calls, and shouldn't be seeing them either. + int TailCallReturnAddrDelta = + MF.getInfo<X86MachineFunctionInfo>()->getTCReturnAddrDelta(); + assert(!(TailCallReturnAddrDelta < 0) && "we don't handle this case!"); +#endif + } - // We don't handle tail calls, and shouldn't be seeing them either. - assert(MF.getInfo<X86MachineFunctionInfo>()->getTCReturnAddrDelta() >= 0 && - "we don't handle this case!"); + // Fill in FrameReg output argument. + FrameReg = TRI->getStackRegister(); // This is how the math works out: // @@ -1837,7 +1759,7 @@ X86FrameLowering::getFrameIndexReferencePreferSP(const MachineFunction &MF, // // A is the incoming stack pointer. // (B - A) is the local area offset (-8 for x86-64) [1] - // (C - A) is the Offset returned by MFI.getObjectOffset for Obj0 [2] + // (C - A) is the Offset returned by MFI->getObjectOffset for Obj0 [2] // // |(E - B)| is the StackSize (absolute value, positive). For a // stack that grown down, this works out to be (B - E). [3] @@ -1847,14 +1769,18 @@ X86FrameLowering::getFrameIndexReferencePreferSP(const MachineFunction &MF, // (C - E) == (C - A) - (B - A) + (B - E) // { Using [1], [2] and [3] above } // == getObjectOffset - LocalAreaOffset + StackSize + // + + // Get the Offset from the StackPointer + int Offset = MFI->getObjectOffset(FI) - getOffsetOfLocalArea(); - return getFrameIndexReferenceSP(MF, FI, FrameReg, StackSize); + return Offset + StackSize; } bool X86FrameLowering::assignCalleeSavedSpillSlots( MachineFunction &MF, const TargetRegisterInfo *TRI, std::vector<CalleeSavedInfo> &CSI) const { - MachineFrameInfo &MFI = MF.getFrameInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); unsigned CalleeSavedFrameSize = 0; @@ -1863,7 +1789,7 @@ bool X86FrameLowering::assignCalleeSavedSpillSlots( if (hasFP(MF)) { // emitPrologue always spills frame register the first thing. SpillSlotOffset -= SlotSize; - MFI.CreateFixedSpillStackObject(SlotSize, SpillSlotOffset); + MFI->CreateFixedSpillStackObject(SlotSize, SpillSlotOffset); // Since emitPrologue and emitEpilogue will handle spilling and restoring of // the frame register, we can delete it from CSI list and not have to worry @@ -1887,7 +1813,7 @@ bool X86FrameLowering::assignCalleeSavedSpillSlots( SpillSlotOffset -= SlotSize; CalleeSavedFrameSize += SlotSize; - int SlotIndex = MFI.CreateFixedSpillStackObject(SlotSize, SpillSlotOffset); + int SlotIndex = MFI->CreateFixedSpillStackObject(SlotSize, SpillSlotOffset); CSI[i - 1].setFrameIdx(SlotIndex); } @@ -1900,15 +1826,14 @@ bool X86FrameLowering::assignCalleeSavedSpillSlots( continue; const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg); - unsigned Size = TRI->getSpillSize(*RC); - unsigned Align = TRI->getSpillAlignment(*RC); // ensure alignment - SpillSlotOffset -= std::abs(SpillSlotOffset) % Align; + SpillSlotOffset -= std::abs(SpillSlotOffset) % RC->getAlignment(); // spill into slot - SpillSlotOffset -= Size; - int SlotIndex = MFI.CreateFixedSpillStackObject(Size, SpillSlotOffset); + SpillSlotOffset -= RC->getSize(); + int SlotIndex = + MFI->CreateFixedSpillStackObject(RC->getSize(), SpillSlotOffset); CSI[i - 1].setFrameIdx(SlotIndex); - MFI.ensureMaxAlignment(Align); + MFI->ensureMaxAlignment(RC->getAlignment()); } return true; @@ -1926,37 +1851,16 @@ bool X86FrameLowering::spillCalleeSavedRegisters( return true; // Push GPRs. It increases frame size. - const MachineFunction &MF = *MBB.getParent(); unsigned Opc = STI.is64Bit() ? X86::PUSH64r : X86::PUSH32r; for (unsigned i = CSI.size(); i != 0; --i) { unsigned Reg = CSI[i - 1].getReg(); if (!X86::GR64RegClass.contains(Reg) && !X86::GR32RegClass.contains(Reg)) continue; + // Add the callee-saved register as live-in. It's killed at the spill. + MBB.addLiveIn(Reg); - const MachineRegisterInfo &MRI = MF.getRegInfo(); - bool isLiveIn = MRI.isLiveIn(Reg); - if (!isLiveIn) - MBB.addLiveIn(Reg); - - // Decide whether we can add a kill flag to the use. - bool CanKill = !isLiveIn; - // Check if any subregister is live-in - if (CanKill) { - for (MCRegAliasIterator AReg(Reg, TRI, false); AReg.isValid(); ++AReg) { - if (MRI.isLiveIn(*AReg)) { - CanKill = false; - break; - } - } - } - - // Do not set a kill flag on values that are also marked as live-in. This - // happens with the @llvm-returnaddress intrinsic and with arguments - // passed in callee saved registers. - // Omitting the kill flags is conservatively correct even if the live-in - // is not used after all. - BuildMI(MBB, MI, DL, TII.get(Opc)).addReg(Reg, getKillRegState(CanKill)) + BuildMI(MBB, MI, DL, TII.get(Opc)).addReg(Reg, RegState::Kill) .setMIFlag(MachineInstr::FrameSetup); } @@ -1980,44 +1884,14 @@ bool X86FrameLowering::spillCalleeSavedRegisters( return true; } -void X86FrameLowering::emitCatchRetReturnValue(MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, - MachineInstr *CatchRet) const { - // SEH shouldn't use catchret. - assert(!isAsynchronousEHPersonality(classifyEHPersonality( - MBB.getParent()->getFunction().getPersonalityFn())) && - "SEH should not use CATCHRET"); - DebugLoc DL = CatchRet->getDebugLoc(); - MachineBasicBlock *CatchRetTarget = CatchRet->getOperand(0).getMBB(); - - // Fill EAX/RAX with the address of the target block. - if (STI.is64Bit()) { - // LEA64r CatchRetTarget(%rip), %rax - BuildMI(MBB, MBBI, DL, TII.get(X86::LEA64r), X86::RAX) - .addReg(X86::RIP) - .addImm(0) - .addReg(0) - .addMBB(CatchRetTarget) - .addReg(0); - } else { - // MOV32ri $CatchRetTarget, %eax - BuildMI(MBB, MBBI, DL, TII.get(X86::MOV32ri), X86::EAX) - .addMBB(CatchRetTarget); - } - - // Record that we've taken the address of CatchRetTarget and no longer just - // reference it in a terminator. - CatchRetTarget->setHasAddressTaken(); -} - bool X86FrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, MachineBasicBlock::iterator MI, - std::vector<CalleeSavedInfo> &CSI, + const std::vector<CalleeSavedInfo> &CSI, const TargetRegisterInfo *TRI) const { if (CSI.empty()) return false; - if (MI != MBB.end() && isFuncletReturnInstr(*MI) && STI.isOSWindows()) { + if (isFuncletReturnInstr(MI) && STI.isOSWindows()) { // Don't restore CSRs in 32-bit EH funclets. Matches // spillCalleeSavedRegisters. if (STI.is32Bit()) @@ -2025,9 +1899,9 @@ bool X86FrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB, // Don't restore CSRs before an SEH catchret. SEH except blocks do not form // funclets. emitEpilogue transforms these to normal jumps. if (MI->getOpcode() == X86::CATCHRET) { - const Function &F = MBB.getParent()->getFunction(); + const Function *Func = MBB.getParent()->getFunction(); bool IsSEH = isAsynchronousEHPersonality( - classifyEHPersonality(F.getPersonalityFn())); + classifyEHPersonality(Func->getPersonalityFn())); if (IsSEH) return true; } @@ -2065,7 +1939,7 @@ void X86FrameLowering::determineCalleeSaves(MachineFunction &MF, RegScavenger *RS) const { TargetFrameLowering::determineCalleeSaves(MF, SavedRegs, RS); - MachineFrameInfo &MFI = MF.getFrameInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); int64_t TailCallReturnAddrDelta = X86FI->getTCReturnAddrDelta(); @@ -2080,7 +1954,7 @@ void X86FrameLowering::determineCalleeSaves(MachineFunction &MF, // ... // } // [EBP] - MFI.CreateFixedObject(-TailCallReturnAddrDelta, + MFI->CreateFixedObject(-TailCallReturnAddrDelta, TailCallReturnAddrDelta - SlotSize, true); } @@ -2089,8 +1963,8 @@ void X86FrameLowering::determineCalleeSaves(MachineFunction &MF, SavedRegs.set(TRI->getBaseRegister()); // Allocate a spill slot for EBP if we have a base pointer and EH funclets. - if (MF.hasEHFunclets()) { - int FI = MFI.CreateSpillStackObject(SlotSize, SlotSize); + if (MF.getMMI().hasEHFunclets()) { + int FI = MFI->CreateSpillStackObject(SlotSize, SlotSize); X86FI->setHasSEHFramePtrSave(true); X86FI->setSEHFramePtrSaveIndex(FI); } @@ -2099,8 +1973,8 @@ void X86FrameLowering::determineCalleeSaves(MachineFunction &MF, static bool HasNestArgument(const MachineFunction *MF) { - const Function &F = MF->getFunction(); - for (Function::const_arg_iterator I = F.arg_begin(), E = F.arg_end(); + const Function *F = MF->getFunction(); + for (Function::const_arg_iterator I = F->arg_begin(), E = F->arg_end(); I != E; I++) { if (I->hasNestAttr()) return true; @@ -2114,7 +1988,7 @@ HasNestArgument(const MachineFunction *MF) { /// needed. Set primary to true for the first register, false for the second. static unsigned GetScratchRegister(bool Is64Bit, bool IsLP64, const MachineFunction &MF, bool Primary) { - CallingConv::ID CallingConvention = MF.getFunction().getCallingConv(); + CallingConv::ID CallingConvention = MF.getFunction()->getCallingConv(); // Erlang stuff. if (CallingConvention == CallingConv::HiPE) { @@ -2151,7 +2025,7 @@ static const uint64_t kSplitStackAvailable = 256; void X86FrameLowering::adjustForSegmentedStacks( MachineFunction &MF, MachineBasicBlock &PrologueMBB) const { - MachineFrameInfo &MFI = MF.getFrameInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); uint64_t StackSize; unsigned TlsReg, TlsOffset; DebugLoc DL; @@ -2164,7 +2038,7 @@ void X86FrameLowering::adjustForSegmentedStacks( assert(!MF.getRegInfo().isLiveIn(ScratchReg) && "Scratch register is live-in"); - if (MF.getFunction().isVarArg()) + if (MF.getFunction()->isVarArg()) report_fatal_error("Segmented stacks do not support vararg functions."); if (!STI.isTargetLinux() && !STI.isTargetDarwin() && !STI.isTargetWin32() && !STI.isTargetWin64() && !STI.isTargetFreeBSD() && @@ -2174,7 +2048,7 @@ void X86FrameLowering::adjustForSegmentedStacks( // Eventually StackSize will be calculated by a link-time pass; which will // also decide whether checking code needs to be injected into this particular // prologue. - StackSize = MFI.getStackSize(); + StackSize = MFI->getStackSize(); // Do not generate a prologue for functions with a stack of size zero if (StackSize == 0) @@ -2350,10 +2224,6 @@ void X86FrameLowering::adjustForSegmentedStacks( // This solution is not perfect, as it assumes that the .rodata section // is laid out within 2^31 bytes of each function body, but this seems // to be sufficient for JIT. - // FIXME: Add retpoline support and remove the error here.. - if (STI.useRetpoline()) - report_fatal_error("Emitting morestack calls on 64-bit with the large " - "code model and retpoline not yet implemented."); BuildMI(allocMBB, DL, TII.get(X86::CALL64m)) .addReg(X86::RIP) .addImm(0) @@ -2380,33 +2250,11 @@ void X86FrameLowering::adjustForSegmentedStacks( checkMBB->addSuccessor(allocMBB); checkMBB->addSuccessor(&PrologueMBB); -#ifdef EXPENSIVE_CHECKS +#ifdef XDEBUG MF.verify(); #endif } -/// Lookup an ERTS parameter in the !hipe.literals named metadata node. -/// HiPE provides Erlang Runtime System-internal parameters, such as PCB offsets -/// to fields it needs, through a named metadata node "hipe.literals" containing -/// name-value pairs. -static unsigned getHiPELiteral( - NamedMDNode *HiPELiteralsMD, const StringRef LiteralName) { - for (int i = 0, e = HiPELiteralsMD->getNumOperands(); i != e; ++i) { - MDNode *Node = HiPELiteralsMD->getOperand(i); - if (Node->getNumOperands() != 2) continue; - MDString *NodeName = dyn_cast<MDString>(Node->getOperand(0)); - ValueAsMetadata *NodeVal = dyn_cast<ValueAsMetadata>(Node->getOperand(1)); - if (!NodeName || !NodeVal) continue; - ConstantInt *ValConst = dyn_cast_or_null<ConstantInt>(NodeVal->getValue()); - if (ValConst && NodeName->getString() == LiteralName) { - return ValConst->getZExtValue(); - } - } - - report_fatal_error("HiPE literal " + LiteralName - + " required but not provided"); -} - /// Erlang programs may need a special prologue to handle the stack size they /// might need at runtime. That is because Erlang/OTP does not implement a C /// stack but uses a custom implementation of hybrid stack/heap architecture. @@ -2424,7 +2272,7 @@ static unsigned getHiPELiteral( /// if( temp0 < SP_LIMIT(P) ) goto IncStack else goto OldStart void X86FrameLowering::adjustForHiPEPrologue( MachineFunction &MF, MachineBasicBlock &PrologueMBB) const { - MachineFrameInfo &MFI = MF.getFrameInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); DebugLoc DL; // To support shrink-wrapping we would need to insert the new blocks @@ -2432,19 +2280,12 @@ void X86FrameLowering::adjustForHiPEPrologue( assert(&(*MF.begin()) == &PrologueMBB && "Shrink-wrapping not supported yet"); // HiPE-specific values - NamedMDNode *HiPELiteralsMD = MF.getMMI().getModule() - ->getNamedMetadata("hipe.literals"); - if (!HiPELiteralsMD) - report_fatal_error( - "Can't generate HiPE prologue without runtime parameters"); - const unsigned HipeLeafWords - = getHiPELiteral(HiPELiteralsMD, - Is64Bit ? "AMD64_LEAF_WORDS" : "X86_LEAF_WORDS"); + const unsigned HipeLeafWords = 24; const unsigned CCRegisteredArgs = Is64Bit ? 6 : 5; const unsigned Guaranteed = HipeLeafWords * SlotSize; - unsigned CallerStkArity = MF.getFunction().arg_size() > CCRegisteredArgs ? - MF.getFunction().arg_size() - CCRegisteredArgs : 0; - unsigned MaxStack = MFI.getStackSize() + CallerStkArity*SlotSize + SlotSize; + unsigned CallerStkArity = MF.getFunction()->arg_size() > CCRegisteredArgs ? + MF.getFunction()->arg_size() - CCRegisteredArgs : 0; + unsigned MaxStack = MFI->getStackSize() + CallerStkArity*SlotSize + SlotSize; assert(STI.isTargetLinux() && "HiPE prologue is only supported on Linux operating systems."); @@ -2456,16 +2297,18 @@ void X86FrameLowering::adjustForHiPEPrologue( // b) outgoing on-stack parameter areas, and // c) the minimum stack space this function needs to make available for the // functions it calls (a tunable ABI property). - if (MFI.hasCalls()) { + if (MFI->hasCalls()) { unsigned MoreStackForCalls = 0; - for (auto &MBB : MF) { - for (auto &MI : MBB) { - if (!MI.isCall()) + for (MachineFunction::iterator MBBI = MF.begin(), MBBE = MF.end(); + MBBI != MBBE; ++MBBI) + for (MachineBasicBlock::iterator MI = MBBI->begin(), ME = MBBI->end(); + MI != ME; ++MI) { + if (!MI->isCall()) continue; // Get callee operand. - const MachineOperand &MO = MI.getOperand(0); + const MachineOperand &MO = MI->getOperand(0); // Only take account of global function calls (no closures etc.). if (!MO.isGlobal()) @@ -2491,7 +2334,6 @@ void X86FrameLowering::adjustForHiPEPrologue( MoreStackForCalls = std::max(MoreStackForCalls, (HipeLeafWords - 1 - CalleeStkArity) * SlotSize); } - } MaxStack += MoreStackForCalls; } @@ -2511,19 +2353,20 @@ void X86FrameLowering::adjustForHiPEPrologue( unsigned ScratchReg, SPReg, PReg, SPLimitOffset; unsigned LEAop, CMPop, CALLop; - SPLimitOffset = getHiPELiteral(HiPELiteralsMD, "P_NSP_LIMIT"); if (Is64Bit) { SPReg = X86::RSP; PReg = X86::RBP; LEAop = X86::LEA64r; CMPop = X86::CMP64rm; CALLop = X86::CALL64pcrel32; + SPLimitOffset = 0x90; } else { SPReg = X86::ESP; PReg = X86::EBP; LEAop = X86::LEA32r; CMPop = X86::CMP32rm; CALLop = X86::CALLpcrel32; + SPLimitOffset = 0x4c; } ScratchReg = GetScratchRegister(Is64Bit, IsLP64, MF, true); @@ -2552,15 +2395,13 @@ void X86FrameLowering::adjustForHiPEPrologue( incStackMBB->addSuccessor(&PrologueMBB, {99, 100}); incStackMBB->addSuccessor(incStackMBB, {1, 100}); } -#ifdef EXPENSIVE_CHECKS +#ifdef XDEBUG MF.verify(); #endif } bool X86FrameLowering::adjustStackWithPops(MachineBasicBlock &MBB, - MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, - int Offset) const { + MachineBasicBlock::iterator MBBI, DebugLoc DL, int Offset) const { if (Offset <= 0) return false; @@ -2584,7 +2425,6 @@ bool X86FrameLowering::adjustStackWithPops(MachineBasicBlock &MBB, unsigned Regs[2]; unsigned FoundRegs = 0; - auto &MRI = MBB.getParent()->getRegInfo(); auto RegMask = Prev->getOperand(1); auto &RegClass = @@ -2598,14 +2438,11 @@ bool X86FrameLowering::adjustStackWithPops(MachineBasicBlock &MBB, if (!RegMask.clobbersPhysReg(Candidate)) continue; - // Don't clobber reserved registers - if (MRI.isReserved(Candidate)) - continue; - bool IsDef = false; for (const MachineOperand &MO : Prev->implicit_operands()) { if (MO.isReg() && MO.isDef() && - TRI->isSuperOrSubRegisterEq(MO.getReg(), Candidate)) { + (TRI->isSubRegisterEq(MO.getReg(), Candidate) || + TRI->isSuperRegister(MO.getReg(), Candidate))) { IsDef = true; break; } @@ -2633,17 +2470,16 @@ bool X86FrameLowering::adjustStackWithPops(MachineBasicBlock &MBB, return true; } -MachineBasicBlock::iterator X86FrameLowering:: +void X86FrameLowering:: eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, MachineBasicBlock::iterator I) const { bool reserveCallFrame = hasReservedCallFrame(MF); unsigned Opcode = I->getOpcode(); bool isDestroy = Opcode == TII.getCallFrameDestroyOpcode(); DebugLoc DL = I->getDebugLoc(); - uint64_t Amount = !reserveCallFrame ? TII.getFrameSize(*I) : 0; - uint64_t InternalAmt = (isDestroy || Amount) ? TII.getFrameAdjustment(*I) : 0; + uint64_t Amount = !reserveCallFrame ? I->getOperand(0).getImm() : 0; + uint64_t InternalAmt = (isDestroy || Amount) ? I->getOperand(1).getImm() : 0; I = MBB.erase(I); - auto InsertPos = skipDebugInstructionsForward(I, MBB.end()); if (!reserveCallFrame) { // If the stack pointer can be changed after prologue, turn the @@ -2654,13 +2490,13 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, // amount of space needed for the outgoing arguments up to the next // alignment boundary. unsigned StackAlign = getStackAlignment(); - Amount = alignTo(Amount, StackAlign); + Amount = RoundUpToAlignment(Amount, StackAlign); MachineModuleInfo &MMI = MF.getMMI(); - const Function &F = MF.getFunction(); + const Function *Fn = MF.getFunction(); bool WindowsCFI = MF.getTarget().getMCAsmInfo()->usesWindowsCFI(); - bool DwarfCFI = !WindowsCFI && - (MMI.hasDebugInfo() || F.needsUnwindTableEntry()); + bool DwarfCFI = !WindowsCFI && + (MMI.hasDebugInfo() || Fn->needsUnwindTableEntry()); // If we have any exception handlers in this function, and we adjust // the SP before calls, we may need to indicate this to the unwinder @@ -2669,15 +2505,16 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, // GNU_ARGS_SIZE. // TODO: We don't need to reset this between subsequent functions, // if it didn't change. - bool HasDwarfEHHandlers = !WindowsCFI && !MF.getLandingPads().empty(); + bool HasDwarfEHHandlers = !WindowsCFI && + !MF.getMMI().getLandingPads().empty(); if (HasDwarfEHHandlers && !isDestroy && MF.getInfo<X86MachineFunctionInfo>()->getHasPushSequences()) - BuildCFI(MBB, InsertPos, DL, + BuildCFI(MBB, I, DL, MCCFIInstruction::createGnuArgsSize(nullptr, Amount)); if (Amount == 0) - return I; + return; // Factor out the amount that gets handled inside the sequence // (Pushes of argument for frame setup, callee pops for frame destroy) @@ -2687,26 +2524,16 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, // If this is a callee-pop calling convention, emit a CFA adjust for // the amount the callee popped. if (isDestroy && InternalAmt && DwarfCFI && !hasFP(MF)) - BuildCFI(MBB, InsertPos, DL, + BuildCFI(MBB, I, DL, MCCFIInstruction::createAdjustCfaOffset(nullptr, -InternalAmt)); - // Add Amount to SP to destroy a frame, or subtract to setup. - int64_t StackAdjustment = isDestroy ? Amount : -Amount; - int64_t CfaAdjustment = -StackAdjustment; - - if (StackAdjustment) { - // Merge with any previous or following adjustment instruction. Note: the - // instructions merged with here do not have CFI, so their stack - // adjustments do not feed into CfaAdjustment. - StackAdjustment += mergeSPUpdates(MBB, InsertPos, true); - StackAdjustment += mergeSPUpdates(MBB, InsertPos, false); - - if (StackAdjustment) { - if (!(F.optForMinSize() && - adjustStackWithPops(MBB, InsertPos, DL, StackAdjustment))) - BuildStackAdjustment(MBB, InsertPos, DL, StackAdjustment, - /*InEpilogue=*/false); - } + if (Amount) { + // Add Amount to SP to destroy a frame, and subtract to setup. + int Offset = isDestroy ? Amount : -Amount; + + if (!(Fn->optForMinSize() && + adjustStackWithPops(MBB, I, DL, Offset))) + BuildStackAdjustment(MBB, I, DL, Offset, /*InEpilogue=*/false); } if (DwarfCFI && !hasFP(MF)) { @@ -2716,17 +2543,18 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, // CFI only for EH purposes or for debugging. EH only requires the CFA // offset to be correct at each call site, while for debugging we want // it to be more precise. - + int CFAOffset = Amount; // TODO: When not using precise CFA, we also need to adjust for the // InternalAmt here. - if (CfaAdjustment) { - BuildCFI(MBB, InsertPos, DL, - MCCFIInstruction::createAdjustCfaOffset(nullptr, - CfaAdjustment)); + + if (CFAOffset) { + CFAOffset = isDestroy ? -CFAOffset : CFAOffset; + BuildCFI(MBB, I, DL, + MCCFIInstruction::createAdjustCfaOffset(nullptr, CFAOffset)); } } - return I; + return; } if (isDestroy && InternalAmt) { @@ -2736,14 +2564,11 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, // We are not tracking the stack pointer adjustment by the callee, so make // sure we restore the stack pointer immediately after the call, there may // be spill code inserted between the CALL and ADJCALLSTACKUP instructions. - MachineBasicBlock::iterator CI = I; MachineBasicBlock::iterator B = MBB.begin(); - while (CI != B && !std::prev(CI)->isCall()) - --CI; - BuildStackAdjustment(MBB, CI, DL, -InternalAmt, /*InEpilogue=*/false); + while (I != B && !std::prev(I)->isCall()) + --I; + BuildStackAdjustment(MBB, I, DL, -InternalAmt, /*InEpilogue=*/false); } - - return I; } bool X86FrameLowering::canUseAsPrologue(const MachineBasicBlock &MBB) const { @@ -2775,19 +2600,19 @@ bool X86FrameLowering::canUseAsEpilogue(const MachineBasicBlock &MBB) const { bool X86FrameLowering::enableShrinkWrapping(const MachineFunction &MF) const { // If we may need to emit frameless compact unwind information, give // up as this is currently broken: PR25614. - return (MF.getFunction().hasFnAttribute(Attribute::NoUnwind) || hasFP(MF)) && + return (MF.getFunction()->hasFnAttribute(Attribute::NoUnwind) || hasFP(MF)) && // The lowering of segmented stack and HiPE only support entry blocks // as prologue blocks: PR26107. // This limitation may be lifted if we fix: // - adjustForSegmentedStacks // - adjustForHiPEPrologue - MF.getFunction().getCallingConv() != CallingConv::HiPE && + MF.getFunction()->getCallingConv() != CallingConv::HiPE && !MF.shouldSplitStack(); } MachineBasicBlock::iterator X86FrameLowering::restoreWin32EHStackPointers( MachineBasicBlock &MBB, MachineBasicBlock::iterator MBBI, - const DebugLoc &DL, bool RestoreSP) const { + DebugLoc DL, bool RestoreSP) const { assert(STI.isTargetWindowsMSVC() && "funclets only supported in MSVC env"); assert(STI.isTargetWin32() && "EBP/ESI restoration only required on win32"); assert(STI.is32Bit() && !Uses64BitFramePtr && @@ -2798,12 +2623,12 @@ MachineBasicBlock::iterator X86FrameLowering::restoreWin32EHStackPointers( unsigned BasePtr = TRI->getBaseRegister(); WinEHFuncInfo &FuncInfo = *MF.getWinEHFuncInfo(); X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); - MachineFrameInfo &MFI = MF.getFrameInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); // FIXME: Don't set FrameSetup flag in catchret case. int FI = FuncInfo.EHRegNodeFrameIndex; - int EHRegSize = MFI.getObjectSize(FI); + int EHRegSize = MFI->getObjectSize(FI); if (RestoreSP) { // MOV32rm -EHRegSize(%ebp), %esp @@ -2847,150 +2672,6 @@ MachineBasicBlock::iterator X86FrameLowering::restoreWin32EHStackPointers( return MBBI; } -namespace { -// Struct used by orderFrameObjects to help sort the stack objects. -struct X86FrameSortingObject { - bool IsValid = false; // true if we care about this Object. - unsigned ObjectIndex = 0; // Index of Object into MFI list. - unsigned ObjectSize = 0; // Size of Object in bytes. - unsigned ObjectAlignment = 1; // Alignment of Object in bytes. - unsigned ObjectNumUses = 0; // Object static number of uses. -}; - -// The comparison function we use for std::sort to order our local -// stack symbols. The current algorithm is to use an estimated -// "density". This takes into consideration the size and number of -// uses each object has in order to roughly minimize code size. -// So, for example, an object of size 16B that is referenced 5 times -// will get higher priority than 4 4B objects referenced 1 time each. -// It's not perfect and we may be able to squeeze a few more bytes out of -// it (for example : 0(esp) requires fewer bytes, symbols allocated at the -// fringe end can have special consideration, given their size is less -// important, etc.), but the algorithmic complexity grows too much to be -// worth the extra gains we get. This gets us pretty close. -// The final order leaves us with objects with highest priority going -// at the end of our list. -struct X86FrameSortingComparator { - inline bool operator()(const X86FrameSortingObject &A, - const X86FrameSortingObject &B) { - uint64_t DensityAScaled, DensityBScaled; - - // For consistency in our comparison, all invalid objects are placed - // at the end. This also allows us to stop walking when we hit the - // first invalid item after it's all sorted. - if (!A.IsValid) - return false; - if (!B.IsValid) - return true; - - // The density is calculated by doing : - // (double)DensityA = A.ObjectNumUses / A.ObjectSize - // (double)DensityB = B.ObjectNumUses / B.ObjectSize - // Since this approach may cause inconsistencies in - // the floating point <, >, == comparisons, depending on the floating - // point model with which the compiler was built, we're going - // to scale both sides by multiplying with - // A.ObjectSize * B.ObjectSize. This ends up factoring away - // the division and, with it, the need for any floating point - // arithmetic. - DensityAScaled = static_cast<uint64_t>(A.ObjectNumUses) * - static_cast<uint64_t>(B.ObjectSize); - DensityBScaled = static_cast<uint64_t>(B.ObjectNumUses) * - static_cast<uint64_t>(A.ObjectSize); - - // If the two densities are equal, prioritize highest alignment - // objects. This allows for similar alignment objects - // to be packed together (given the same density). - // There's room for improvement here, also, since we can pack - // similar alignment (different density) objects next to each - // other to save padding. This will also require further - // complexity/iterations, and the overall gain isn't worth it, - // in general. Something to keep in mind, though. - if (DensityAScaled == DensityBScaled) - return A.ObjectAlignment < B.ObjectAlignment; - - return DensityAScaled < DensityBScaled; - } -}; -} // namespace - -// Order the symbols in the local stack. -// We want to place the local stack objects in some sort of sensible order. -// The heuristic we use is to try and pack them according to static number -// of uses and size of object in order to minimize code size. -void X86FrameLowering::orderFrameObjects( - const MachineFunction &MF, SmallVectorImpl<int> &ObjectsToAllocate) const { - const MachineFrameInfo &MFI = MF.getFrameInfo(); - - // Don't waste time if there's nothing to do. - if (ObjectsToAllocate.empty()) - return; - - // Create an array of all MFI objects. We won't need all of these - // objects, but we're going to create a full array of them to make - // it easier to index into when we're counting "uses" down below. - // We want to be able to easily/cheaply access an object by simply - // indexing into it, instead of having to search for it every time. - std::vector<X86FrameSortingObject> SortingObjects(MFI.getObjectIndexEnd()); - - // Walk the objects we care about and mark them as such in our working - // struct. - for (auto &Obj : ObjectsToAllocate) { - SortingObjects[Obj].IsValid = true; - SortingObjects[Obj].ObjectIndex = Obj; - SortingObjects[Obj].ObjectAlignment = MFI.getObjectAlignment(Obj); - // Set the size. - int ObjectSize = MFI.getObjectSize(Obj); - if (ObjectSize == 0) - // Variable size. Just use 4. - SortingObjects[Obj].ObjectSize = 4; - else - SortingObjects[Obj].ObjectSize = ObjectSize; - } - - // Count the number of uses for each object. - for (auto &MBB : MF) { - for (auto &MI : MBB) { - if (MI.isDebugValue()) - continue; - for (const MachineOperand &MO : MI.operands()) { - // Check to see if it's a local stack symbol. - if (!MO.isFI()) - continue; - int Index = MO.getIndex(); - // Check to see if it falls within our range, and is tagged - // to require ordering. - if (Index >= 0 && Index < MFI.getObjectIndexEnd() && - SortingObjects[Index].IsValid) - SortingObjects[Index].ObjectNumUses++; - } - } - } - - // Sort the objects using X86FrameSortingAlgorithm (see its comment for - // info). - std::stable_sort(SortingObjects.begin(), SortingObjects.end(), - X86FrameSortingComparator()); - - // Now modify the original list to represent the final order that - // we want. The order will depend on whether we're going to access them - // from the stack pointer or the frame pointer. For SP, the list should - // end up with the END containing objects that we want with smaller offsets. - // For FP, it should be flipped. - int i = 0; - for (auto &Obj : SortingObjects) { - // All invalid items are sorted at the end, so it's safe to stop. - if (!Obj.IsValid) - break; - ObjectsToAllocate[i++] = Obj.ObjectIndex; - } - - // Flip it if we're accessing off of the FP. - if (!TRI->needsStackRealignment(MF) && hasFP(MF)) - std::reverse(ObjectsToAllocate.begin(), ObjectsToAllocate.end()); -} - - unsigned X86FrameLowering::getWinEHParentFrameOffset(const MachineFunction &MF) const { // RDX, the parent frame pointer, is homed into 16(%rsp) in the prologue. unsigned Offset = 16; @@ -3005,15 +2686,11 @@ unsigned X86FrameLowering::getWinEHParentFrameOffset(const MachineFunction &MF) void X86FrameLowering::processFunctionBeforeFrameFinalized( MachineFunction &MF, RegScavenger *RS) const { - // Mark the function as not having WinCFI. We will set it back to true in - // emitPrologue if it gets called and emits CFI. - MF.setHasWinCFI(false); - // If this function isn't doing Win64-style C++ EH, we don't need to do // anything. - const Function &F = MF.getFunction(); - if (!STI.is64Bit() || !MF.hasEHFunclets() || - classifyEHPersonality(F.getPersonalityFn()) != EHPersonality::MSVC_CXX) + const Function *Fn = MF.getFunction(); + if (!STI.is64Bit() || !MF.getMMI().hasEHFunclets() || + classifyEHPersonality(Fn->getPersonalityFn()) != EHPersonality::MSVC_CXX) return; // Win64 C++ EH needs to allocate the UnwindHelp object at some fixed offset @@ -3021,31 +2698,15 @@ void X86FrameLowering::processFunctionBeforeFrameFinalized( // object, so that we can allocate a slot immediately following it. If there // were no fixed objects, use offset -SlotSize, which is immediately after the // return address. Fixed objects have negative frame indices. - MachineFrameInfo &MFI = MF.getFrameInfo(); - WinEHFuncInfo &EHInfo = *MF.getWinEHFuncInfo(); + MachineFrameInfo *MFI = MF.getFrameInfo(); int64_t MinFixedObjOffset = -SlotSize; - for (int I = MFI.getObjectIndexBegin(); I < 0; ++I) - MinFixedObjOffset = std::min(MinFixedObjOffset, MFI.getObjectOffset(I)); - - for (WinEHTryBlockMapEntry &TBME : EHInfo.TryBlockMap) { - for (WinEHHandlerType &H : TBME.HandlerArray) { - int FrameIndex = H.CatchObj.FrameIndex; - if (FrameIndex != INT_MAX) { - // Ensure alignment. - unsigned Align = MFI.getObjectAlignment(FrameIndex); - MinFixedObjOffset -= std::abs(MinFixedObjOffset) % Align; - MinFixedObjOffset -= MFI.getObjectSize(FrameIndex); - MFI.setObjectOffset(FrameIndex, MinFixedObjOffset); - } - } - } + for (int I = MFI->getObjectIndexBegin(); I < 0; ++I) + MinFixedObjOffset = std::min(MinFixedObjOffset, MFI->getObjectOffset(I)); - // Ensure alignment. - MinFixedObjOffset -= std::abs(MinFixedObjOffset) % 8; int64_t UnwindHelpOffset = MinFixedObjOffset - SlotSize; int UnwindHelpFI = - MFI.CreateFixedObject(SlotSize, UnwindHelpOffset, /*Immutable=*/false); - EHInfo.UnwindHelpFrameIdx = UnwindHelpFI; + MFI->CreateFixedObject(SlotSize, UnwindHelpOffset, /*Immutable=*/false); + MF.getWinEHFuncInfo()->UnwindHelpFrameIdx = UnwindHelpFI; // Store -2 into UnwindHelp on function entry. We have to scan forwards past // other frame setup instructions. |