X-Git-Url: http://plrg.eecs.uci.edu/git/?p=oota-llvm.git;a=blobdiff_plain;f=lib%2FTarget%2FX86%2FX86MCInstLower.cpp;h=85cf5144f0cc853e57604171785f34b17a619162;hp=04d1ad57a75ffa2da1fde114a193217e9496b98e;hb=7476f9c5134a7170ca53561d20e9a19adef251f8;hpb=3d74dea4bddc84d1881efc21eb5eefbddbfa9aed diff --git a/lib/Target/X86/X86MCInstLower.cpp b/lib/Target/X86/X86MCInstLower.cpp index 04d1ad57a75..85cf5144f0c 100644 --- a/lib/Target/X86/X86MCInstLower.cpp +++ b/lib/Target/X86/X86MCInstLower.cpp @@ -13,21 +13,28 @@ //===----------------------------------------------------------------------===// #include "X86AsmPrinter.h" +#include "X86RegisterInfo.h" #include "InstPrinter/X86ATTInstPrinter.h" -#include "X86COFFMachineModuleInfo.h" +#include "MCTargetDesc/X86BaseInfo.h" +#include "Utils/X86ShuffleDecode.h" #include "llvm/ADT/SmallString.h" +#include "llvm/CodeGen/MachineFunction.h" +#include "llvm/CodeGen/MachineConstantPool.h" +#include "llvm/CodeGen/MachineOperand.h" #include "llvm/CodeGen/MachineModuleInfoImpls.h" #include "llvm/CodeGen/StackMaps.h" -#include "llvm/IR/Type.h" +#include "llvm/IR/DataLayout.h" +#include "llvm/IR/GlobalValue.h" +#include "llvm/IR/Mangler.h" #include "llvm/MC/MCAsmInfo.h" +#include "llvm/MC/MCCodeEmitter.h" #include "llvm/MC/MCContext.h" #include "llvm/MC/MCExpr.h" #include "llvm/MC/MCInst.h" #include "llvm/MC/MCInstBuilder.h" #include "llvm/MC/MCStreamer.h" #include "llvm/MC/MCSymbol.h" -#include "llvm/Support/FormattedStream.h" -#include "llvm/Target/Mangler.h" +#include "llvm/Support/TargetRegistry.h" using namespace llvm; namespace { @@ -56,6 +63,53 @@ private: } // end anonymous namespace +// Emit a minimal sequence of nops spanning NumBytes bytes. +static void EmitNops(MCStreamer &OS, unsigned NumBytes, bool Is64Bit, + const MCSubtargetInfo &STI); + +namespace llvm { + X86AsmPrinter::StackMapShadowTracker::StackMapShadowTracker(TargetMachine &TM) + : TM(TM), InShadow(false), RequiredShadowSize(0), CurrentShadowSize(0) {} + + X86AsmPrinter::StackMapShadowTracker::~StackMapShadowTracker() {} + + void + X86AsmPrinter::StackMapShadowTracker::startFunction(MachineFunction &MF) { + CodeEmitter.reset(TM.getTarget().createMCCodeEmitter( + *TM.getSubtargetImpl()->getInstrInfo(), + *TM.getSubtargetImpl()->getRegisterInfo(), *TM.getSubtargetImpl(), + MF.getContext())); + } + + void X86AsmPrinter::StackMapShadowTracker::count(MCInst &Inst, + const MCSubtargetInfo &STI) { + if (InShadow) { + SmallString<256> Code; + SmallVector Fixups; + raw_svector_ostream VecOS(Code); + CodeEmitter->EncodeInstruction(Inst, VecOS, Fixups, STI); + VecOS.flush(); + CurrentShadowSize += Code.size(); + if (CurrentShadowSize >= RequiredShadowSize) + InShadow = false; // The shadow is big enough. Stop counting. + } + } + + void X86AsmPrinter::StackMapShadowTracker::emitShadowPadding( + MCStreamer &OutStreamer, const MCSubtargetInfo &STI) { + if (InShadow && CurrentShadowSize < RequiredShadowSize) { + InShadow = false; + EmitNops(OutStreamer, RequiredShadowSize - CurrentShadowSize, + TM.getSubtarget().is64Bit(), STI); + } + } + + void X86AsmPrinter::EmitAndCountInstruction(MCInst &Inst) { + OutStreamer.EmitInstruction(Inst, getSubtargetInfo()); + SMShadowTracker.count(Inst, getSubtargetInfo()); + } +} // end llvm namespace + X86MCInstLower::X86MCInstLower(const MachineFunction &mf, X86AsmPrinter &asmprinter) : Ctx(mf.getContext()), MF(mf), TM(mf.getTarget()), @@ -70,70 +124,77 @@ MachineModuleInfoMachO &X86MCInstLower::getMachOMMI() const { /// operand to an MCSymbol. MCSymbol *X86MCInstLower:: GetSymbolFromOperand(const MachineOperand &MO) const { + const DataLayout *DL = TM.getSubtargetImpl()->getDataLayout(); assert((MO.isGlobal() || MO.isSymbol() || MO.isMBB()) && "Isn't a symbol reference"); SmallString<128> Name; + StringRef Suffix; + + switch (MO.getTargetFlags()) { + case X86II::MO_DLLIMPORT: + // Handle dllimport linkage. + Name += "__imp_"; + break; + case X86II::MO_DARWIN_STUB: + Suffix = "$stub"; + break; + case X86II::MO_DARWIN_NONLAZY: + case X86II::MO_DARWIN_NONLAZY_PIC_BASE: + case X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE: + Suffix = "$non_lazy_ptr"; + break; + } + + if (!Suffix.empty()) + Name += DL->getPrivateGlobalPrefix(); + + unsigned PrefixLen = Name.size(); if (MO.isGlobal()) { const GlobalValue *GV = MO.getGlobal(); - bool isImplicitlyPrivate = false; - if (MO.getTargetFlags() == X86II::MO_DARWIN_STUB || - MO.getTargetFlags() == X86II::MO_DARWIN_NONLAZY || - MO.getTargetFlags() == X86II::MO_DARWIN_NONLAZY_PIC_BASE || - MO.getTargetFlags() == X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE) - isImplicitlyPrivate = true; - - getMang()->getNameWithPrefix(Name, GV, isImplicitlyPrivate); + AsmPrinter.getNameWithPrefix(Name, GV); } else if (MO.isSymbol()) { - Name += MAI.getGlobalPrefix(); - Name += MO.getSymbolName(); + getMang()->getNameWithPrefix(Name, MO.getSymbolName()); } else if (MO.isMBB()) { Name += MO.getMBB()->getSymbol()->getName(); } + unsigned OrigLen = Name.size() - PrefixLen; + + Name += Suffix; + MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name); + + StringRef OrigName = StringRef(Name).substr(PrefixLen, OrigLen); // If the target flags on the operand changes the name of the symbol, do that // before we return the symbol. switch (MO.getTargetFlags()) { default: break; - case X86II::MO_DLLIMPORT: { - // Handle dllimport linkage. - const char *Prefix = "__imp_"; - Name.insert(Name.begin(), Prefix, Prefix+strlen(Prefix)); - break; - } case X86II::MO_DARWIN_NONLAZY: case X86II::MO_DARWIN_NONLAZY_PIC_BASE: { - Name += "$non_lazy_ptr"; - MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str()); - MachineModuleInfoImpl::StubValueTy &StubSym = getMachOMMI().getGVStubEntry(Sym); - if (StubSym.getPointer() == 0) { + if (!StubSym.getPointer()) { assert(MO.isGlobal() && "Extern symbol not handled yet"); StubSym = MachineModuleInfoImpl:: StubValueTy(AsmPrinter.getSymbol(MO.getGlobal()), !MO.getGlobal()->hasInternalLinkage()); } - return Sym; + break; } case X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE: { - Name += "$non_lazy_ptr"; - MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str()); MachineModuleInfoImpl::StubValueTy &StubSym = getMachOMMI().getHiddenGVStubEntry(Sym); - if (StubSym.getPointer() == 0) { + if (!StubSym.getPointer()) { assert(MO.isGlobal() && "Extern symbol not handled yet"); StubSym = MachineModuleInfoImpl:: StubValueTy(AsmPrinter.getSymbol(MO.getGlobal()), !MO.getGlobal()->hasInternalLinkage()); } - return Sym; + break; } case X86II::MO_DARWIN_STUB: { - Name += "$stub"; - MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str()); MachineModuleInfoImpl::StubValueTy &StubSym = getMachOMMI().getFnStubEntry(Sym); if (StubSym.getPointer()) @@ -145,23 +206,22 @@ GetSymbolFromOperand(const MachineOperand &MO) const { StubValueTy(AsmPrinter.getSymbol(MO.getGlobal()), !MO.getGlobal()->hasInternalLinkage()); } else { - Name.erase(Name.end()-5, Name.end()); StubSym = MachineModuleInfoImpl:: - StubValueTy(Ctx.GetOrCreateSymbol(Name.str()), false); + StubValueTy(Ctx.GetOrCreateSymbol(OrigName), false); } - return Sym; + break; } } - return Ctx.GetOrCreateSymbol(Name.str()); + return Sym; } MCOperand X86MCInstLower::LowerSymbolOperand(const MachineOperand &MO, MCSymbol *Sym) const { // FIXME: We would like an efficient form for this, so we don't have to do a // lot of extra uniquing. - const MCExpr *Expr = 0; + const MCExpr *Expr = nullptr; MCSymbolRefExpr::VariantKind RefKind = MCSymbolRefExpr::VK_None; switch (MO.getTargetFlags()) { @@ -204,7 +264,8 @@ MCOperand X86MCInstLower::LowerSymbolOperand(const MachineOperand &MO, Expr = MCBinaryExpr::CreateSub(Expr, MCSymbolRefExpr::Create(MF.getPICBaseSymbol(), Ctx), Ctx); - if (MO.isJTI() && MAI.hasSetDirective()) { + if (MO.isJTI()) { + assert(MAI.doesSetDirectiveSuppressesReloc()); // If .set directive is supported, use it to reduce the number of // relocations the assembler will generate for differences between // local labels. This is only safe when the symbols are in the same @@ -216,7 +277,7 @@ MCOperand X86MCInstLower::LowerSymbolOperand(const MachineOperand &MO, break; } - if (Expr == 0) + if (!Expr) Expr = MCSymbolRefExpr::Create(Sym, RefKind, Ctx); if (!MO.isJTI() && !MO.isMBB() && MO.getOffset()) @@ -227,13 +288,6 @@ MCOperand X86MCInstLower::LowerSymbolOperand(const MachineOperand &MO, } -/// LowerUnaryToTwoAddr - R = setb -> R = sbb R, R -static void LowerUnaryToTwoAddr(MCInst &OutMI, unsigned NewOpc) { - OutMI.setOpcode(NewOpc); - OutMI.addOperand(OutMI.getOperand(0)); - OutMI.addOperand(OutMI.getOperand(0)); -} - /// \brief Simplify FOO $imm, %{al,ax,eax,rax} to FOO $imm, for instruction with /// a short fixed-register form. static void SimplifyShortImmForm(MCInst &Inst, unsigned Opcode) { @@ -297,12 +351,12 @@ static void SimplifyShortMoveForm(X86AsmPrinter &Printer, MCInst &Inst, unsigned RegOp = IsStore ? 0 : 5; unsigned AddrOp = AddrBase + 3; assert(Inst.getNumOperands() == 6 && Inst.getOperand(RegOp).isReg() && - Inst.getOperand(AddrBase + 0).isReg() && // base - Inst.getOperand(AddrBase + 1).isImm() && // scale - Inst.getOperand(AddrBase + 2).isReg() && // index register - (Inst.getOperand(AddrOp).isExpr() || // address - Inst.getOperand(AddrOp).isImm())&& - Inst.getOperand(AddrBase + 4).isReg() && // segment + Inst.getOperand(AddrBase + X86::AddrBaseReg).isReg() && + Inst.getOperand(AddrBase + X86::AddrScaleAmt).isImm() && + Inst.getOperand(AddrBase + X86::AddrIndexReg).isReg() && + Inst.getOperand(AddrBase + X86::AddrSegmentReg).isReg() && + (Inst.getOperand(AddrOp).isExpr() || + Inst.getOperand(AddrOp).isImm()) && "Unexpected instruction!"); // Check whether the destination register can be fixed. @@ -322,17 +376,23 @@ static void SimplifyShortMoveForm(X86AsmPrinter &Printer, MCInst &Inst, } if (Absolute && - (Inst.getOperand(AddrBase + 0).getReg() != 0 || - Inst.getOperand(AddrBase + 2).getReg() != 0 || - Inst.getOperand(AddrBase + 4).getReg() != 0 || - Inst.getOperand(AddrBase + 1).getImm() != 1)) + (Inst.getOperand(AddrBase + X86::AddrBaseReg).getReg() != 0 || + Inst.getOperand(AddrBase + X86::AddrScaleAmt).getImm() != 1 || + Inst.getOperand(AddrBase + X86::AddrIndexReg).getReg() != 0)) return; // If so, rewrite the instruction. MCOperand Saved = Inst.getOperand(AddrOp); + MCOperand Seg = Inst.getOperand(AddrBase + X86::AddrSegmentReg); Inst = MCInst(); Inst.setOpcode(Opcode); Inst.addOperand(Saved); + Inst.addOperand(Seg); +} + +static unsigned getRetOpcode(const X86Subtarget &Subtarget) +{ + return Subtarget.is64Bit() ? X86::RETQ : X86::RETL; } void X86MCInstLower::Lower(const MachineInstr *MI, MCInst &OutMI) const { @@ -390,7 +450,6 @@ ReSimplify: assert(OutMI.getOperand(1+X86::AddrSegmentReg).getReg() == 0 && "LEA has segment specified!"); break; - case X86::MOV32r0: LowerUnaryToTwoAddr(OutMI, X86::XOR32rr); break; case X86::MOV32ri64: OutMI.setOpcode(X86::MOV32ri); @@ -464,7 +523,7 @@ ReSimplify: case X86::EH_RETURN: case X86::EH_RETURN64: { OutMI = MCInst(); - OutMI.setOpcode(X86::RET); + OutMI.setOpcode(getRetOpcode(AsmPrinter.getSubtarget())); break; } @@ -525,14 +584,38 @@ ReSimplify: // Atomic load and store require a separate pseudo-inst because Acquire // implies mayStore and Release implies mayLoad; fix these to regular MOV // instructions here - case X86::ACQUIRE_MOV8rm: OutMI.setOpcode(X86::MOV8rm); goto ReSimplify; - case X86::ACQUIRE_MOV16rm: OutMI.setOpcode(X86::MOV16rm); goto ReSimplify; - case X86::ACQUIRE_MOV32rm: OutMI.setOpcode(X86::MOV32rm); goto ReSimplify; - case X86::ACQUIRE_MOV64rm: OutMI.setOpcode(X86::MOV64rm); goto ReSimplify; - case X86::RELEASE_MOV8mr: OutMI.setOpcode(X86::MOV8mr); goto ReSimplify; - case X86::RELEASE_MOV16mr: OutMI.setOpcode(X86::MOV16mr); goto ReSimplify; - case X86::RELEASE_MOV32mr: OutMI.setOpcode(X86::MOV32mr); goto ReSimplify; - case X86::RELEASE_MOV64mr: OutMI.setOpcode(X86::MOV64mr); goto ReSimplify; + case X86::ACQUIRE_MOV8rm: OutMI.setOpcode(X86::MOV8rm); goto ReSimplify; + case X86::ACQUIRE_MOV16rm: OutMI.setOpcode(X86::MOV16rm); goto ReSimplify; + case X86::ACQUIRE_MOV32rm: OutMI.setOpcode(X86::MOV32rm); goto ReSimplify; + case X86::ACQUIRE_MOV64rm: OutMI.setOpcode(X86::MOV64rm); goto ReSimplify; + case X86::RELEASE_MOV8mr: OutMI.setOpcode(X86::MOV8mr); goto ReSimplify; + case X86::RELEASE_MOV16mr: OutMI.setOpcode(X86::MOV16mr); goto ReSimplify; + case X86::RELEASE_MOV32mr: OutMI.setOpcode(X86::MOV32mr); goto ReSimplify; + case X86::RELEASE_MOV64mr: OutMI.setOpcode(X86::MOV64mr); goto ReSimplify; + case X86::RELEASE_MOV8mi: OutMI.setOpcode(X86::MOV8mi); goto ReSimplify; + case X86::RELEASE_MOV16mi: OutMI.setOpcode(X86::MOV16mi); goto ReSimplify; + case X86::RELEASE_MOV32mi: OutMI.setOpcode(X86::MOV32mi); goto ReSimplify; + case X86::RELEASE_MOV64mi32: OutMI.setOpcode(X86::MOV64mi32); goto ReSimplify; + case X86::RELEASE_ADD8mi: OutMI.setOpcode(X86::ADD8mi); goto ReSimplify; + case X86::RELEASE_ADD32mi: OutMI.setOpcode(X86::ADD32mi); goto ReSimplify; + case X86::RELEASE_ADD64mi32: OutMI.setOpcode(X86::ADD64mi32); goto ReSimplify; + case X86::RELEASE_AND8mi: OutMI.setOpcode(X86::AND8mi); goto ReSimplify; + case X86::RELEASE_AND32mi: OutMI.setOpcode(X86::AND32mi); goto ReSimplify; + case X86::RELEASE_AND64mi32: OutMI.setOpcode(X86::AND64mi32); goto ReSimplify; + case X86::RELEASE_OR8mi: OutMI.setOpcode(X86::OR8mi); goto ReSimplify; + case X86::RELEASE_OR32mi: OutMI.setOpcode(X86::OR32mi); goto ReSimplify; + case X86::RELEASE_OR64mi32: OutMI.setOpcode(X86::OR64mi32); goto ReSimplify; + case X86::RELEASE_XOR8mi: OutMI.setOpcode(X86::XOR8mi); goto ReSimplify; + case X86::RELEASE_XOR32mi: OutMI.setOpcode(X86::XOR32mi); goto ReSimplify; + case X86::RELEASE_XOR64mi32: OutMI.setOpcode(X86::XOR64mi32); goto ReSimplify; + case X86::RELEASE_INC8m: OutMI.setOpcode(X86::INC8m); goto ReSimplify; + case X86::RELEASE_INC16m: OutMI.setOpcode(X86::INC16m); goto ReSimplify; + case X86::RELEASE_INC32m: OutMI.setOpcode(X86::INC32m); goto ReSimplify; + case X86::RELEASE_INC64m: OutMI.setOpcode(X86::INC64m); goto ReSimplify; + case X86::RELEASE_DEC8m: OutMI.setOpcode(X86::DEC8m); goto ReSimplify; + case X86::RELEASE_DEC16m: OutMI.setOpcode(X86::DEC16m); goto ReSimplify; + case X86::RELEASE_DEC32m: OutMI.setOpcode(X86::DEC32m); goto ReSimplify; + case X86::RELEASE_DEC64m: OutMI.setOpcode(X86::DEC64m); goto ReSimplify; // We don't currently select the correct instruction form for instructions // which have a short %eax, etc. form. Handle this by custom lowering, for @@ -593,24 +676,11 @@ ReSimplify: case X86::MOVSX64rr32: SimplifyMOVSX(OutMI); break; - - case X86::MORESTACK_RET: - OutMI.setOpcode(X86::RET); - break; - - case X86::MORESTACK_RET_RESTORE_R10: - OutMI.setOpcode(X86::MOV64rr); - OutMI.addOperand(MCOperand::CreateReg(X86::R10)); - OutMI.addOperand(MCOperand::CreateReg(X86::RAX)); - - AsmPrinter.OutStreamer.EmitInstruction(MCInstBuilder(X86::RET)); - break; } } -static void LowerTlsAddr(MCStreamer &OutStreamer, - X86MCInstLower &MCInstLowering, - const MachineInstr &MI) { +void X86AsmPrinter::LowerTlsAddr(X86MCInstLower &MCInstLowering, + const MachineInstr &MI) { bool is64Bits = MI.getOpcode() == X86::TLS_addr64 || MI.getOpcode() == X86::TLS_base_addr64; @@ -620,7 +690,7 @@ static void LowerTlsAddr(MCStreamer &OutStreamer, MCContext &context = OutStreamer.getContext(); if (needsPadding) - OutStreamer.EmitInstruction(MCInstBuilder(X86::DATA16_PREFIX)); + EmitAndCountInstruction(MCInstBuilder(X86::DATA16_PREFIX)); MCSymbolRefExpr::VariantKind SRVK; switch (MI.getOpcode()) { @@ -667,12 +737,12 @@ static void LowerTlsAddr(MCStreamer &OutStreamer, LEA.addOperand(MCOperand::CreateExpr(symRef)); // disp LEA.addOperand(MCOperand::CreateReg(0)); // seg } - OutStreamer.EmitInstruction(LEA); + EmitAndCountInstruction(LEA); if (needsPadding) { - OutStreamer.EmitInstruction(MCInstBuilder(X86::DATA16_PREFIX)); - OutStreamer.EmitInstruction(MCInstBuilder(X86::DATA16_PREFIX)); - OutStreamer.EmitInstruction(MCInstBuilder(X86::REX64_PREFIX)); + EmitAndCountInstruction(MCInstBuilder(X86::DATA16_PREFIX)); + EmitAndCountInstruction(MCInstBuilder(X86::DATA16_PREFIX)); + EmitAndCountInstruction(MCInstBuilder(X86::REX64_PREFIX)); } StringRef name = is64Bits ? "__tls_get_addr" : "___tls_get_addr"; @@ -682,138 +752,205 @@ static void LowerTlsAddr(MCStreamer &OutStreamer, MCSymbolRefExpr::VK_PLT, context); - OutStreamer.EmitInstruction(MCInstBuilder(is64Bits ? X86::CALL64pcrel32 - : X86::CALLpcrel32) - .addExpr(tlsRef)); + EmitAndCountInstruction(MCInstBuilder(is64Bits ? X86::CALL64pcrel32 + : X86::CALLpcrel32) + .addExpr(tlsRef)); } -static std::pair -parseMemoryOperand(StackMaps::Location::LocationType LocTy, - MachineInstr::const_mop_iterator MOI, - MachineInstr::const_mop_iterator MOE) { - - typedef StackMaps::Location Location; - - assert(std::distance(MOI, MOE) >= 5 && "Too few operands to encode mem op."); +/// \brief Emit the optimal amount of multi-byte nops on X86. +static void EmitNops(MCStreamer &OS, unsigned NumBytes, bool Is64Bit, const MCSubtargetInfo &STI) { + // This works only for 64bit. For 32bit we have to do additional checking if + // the CPU supports multi-byte nops. + assert(Is64Bit && "EmitNops only supports X86-64"); + while (NumBytes) { + unsigned Opc, BaseReg, ScaleVal, IndexReg, Displacement, SegmentReg; + Opc = IndexReg = Displacement = SegmentReg = 0; + BaseReg = X86::RAX; ScaleVal = 1; + switch (NumBytes) { + case 0: llvm_unreachable("Zero nops?"); break; + case 1: NumBytes -= 1; Opc = X86::NOOP; break; + case 2: NumBytes -= 2; Opc = X86::XCHG16ar; break; + case 3: NumBytes -= 3; Opc = X86::NOOPL; break; + case 4: NumBytes -= 4; Opc = X86::NOOPL; Displacement = 8; break; + case 5: NumBytes -= 5; Opc = X86::NOOPL; Displacement = 8; + IndexReg = X86::RAX; break; + case 6: NumBytes -= 6; Opc = X86::NOOPW; Displacement = 8; + IndexReg = X86::RAX; break; + case 7: NumBytes -= 7; Opc = X86::NOOPL; Displacement = 512; break; + case 8: NumBytes -= 8; Opc = X86::NOOPL; Displacement = 512; + IndexReg = X86::RAX; break; + case 9: NumBytes -= 9; Opc = X86::NOOPW; Displacement = 512; + IndexReg = X86::RAX; break; + default: NumBytes -= 10; Opc = X86::NOOPW; Displacement = 512; + IndexReg = X86::RAX; SegmentReg = X86::CS; break; + } - const MachineOperand &Base = *MOI; - const MachineOperand &Scale = *(++MOI); - const MachineOperand &Index = *(++MOI); - const MachineOperand &Disp = *(++MOI); - const MachineOperand &ZeroReg = *(++MOI); + unsigned NumPrefixes = std::min(NumBytes, 5U); + NumBytes -= NumPrefixes; + for (unsigned i = 0; i != NumPrefixes; ++i) + OS.EmitBytes("\x66"); - // Sanity check for supported operand format. - assert(Base.isReg() && - Scale.isImm() && Scale.getImm() == 1 && - Index.isReg() && Index.getReg() == 0 && - Disp.isImm() && ZeroReg.isReg() && (ZeroReg.getReg() == 0) && - "Unsupported x86 memory operand sequence."); + switch (Opc) { + default: llvm_unreachable("Unexpected opcode"); break; + case X86::NOOP: + OS.EmitInstruction(MCInstBuilder(Opc), STI); + break; + case X86::XCHG16ar: + OS.EmitInstruction(MCInstBuilder(Opc).addReg(X86::AX), STI); + break; + case X86::NOOPL: + case X86::NOOPW: + OS.EmitInstruction(MCInstBuilder(Opc).addReg(BaseReg) + .addImm(ScaleVal).addReg(IndexReg) + .addImm(Displacement).addReg(SegmentReg), STI); + break; + } + } // while (NumBytes) +} - return std::make_pair( - Location(LocTy, Base.getReg(), Disp.getImm()), ++MOI); +// Lower a stackmap of the form: +// , , ... +void X86AsmPrinter::LowerSTACKMAP(const MachineInstr &MI) { + SMShadowTracker.emitShadowPadding(OutStreamer, getSubtargetInfo()); + SM.recordStackMap(MI); + unsigned NumShadowBytes = MI.getOperand(1).getImm(); + SMShadowTracker.reset(NumShadowBytes); } -std::pair -X86AsmPrinter::stackmapOperandParser(MachineInstr::const_mop_iterator MOI, - MachineInstr::const_mop_iterator MOE) { - - typedef StackMaps::Location Location; - - const MachineOperand &MOP = *MOI; - assert(!MOP.isRegMask() && (!MOP.isReg() || !MOP.isImplicit()) && - "Register mask and implicit operands should not be processed."); - - if (MOP.isImm()) { - switch (MOP.getImm()) { - default: llvm_unreachable("Unrecognized operand type."); - case StackMaps::DirectMemRefOp: - return parseMemoryOperand(StackMaps::Location::Direct, - llvm::next(MOI), MOE); - case StackMaps::IndirectMemRefOp: - return parseMemoryOperand(StackMaps::Location::Indirect, - llvm::next(MOI), MOE); - case StackMaps::ConstantOp: { - ++MOI; - assert(MOI->isImm() && "Expected constant operand."); - int64_t Imm = MOI->getImm(); - return std::make_pair(Location(Location::Constant, 0, Imm), ++MOI); - } - } +// Lower a patchpoint of the form: +// [], , , , , , ... +void X86AsmPrinter::LowerPATCHPOINT(const MachineInstr &MI) { + assert(Subtarget->is64Bit() && "Patchpoint currently only supports X86-64"); + + SMShadowTracker.emitShadowPadding(OutStreamer, getSubtargetInfo()); + + SM.recordPatchPoint(MI); + + PatchPointOpers opers(&MI); + unsigned ScratchIdx = opers.getNextScratchIdx(); + unsigned EncodedBytes = 0; + int64_t CallTarget = opers.getMetaOper(PatchPointOpers::TargetPos).getImm(); + if (CallTarget) { + // Emit MOV to materialize the target address and the CALL to target. + // This is encoded with 12-13 bytes, depending on which register is used. + unsigned ScratchReg = MI.getOperand(ScratchIdx).getReg(); + if (X86II::isX86_64ExtendedReg(ScratchReg)) + EncodedBytes = 13; + else + EncodedBytes = 12; + EmitAndCountInstruction(MCInstBuilder(X86::MOV64ri).addReg(ScratchReg) + .addImm(CallTarget)); + EmitAndCountInstruction(MCInstBuilder(X86::CALL64r).addReg(ScratchReg)); } + // Emit padding. + unsigned NumBytes = opers.getMetaOper(PatchPointOpers::NBytesPos).getImm(); + assert(NumBytes >= EncodedBytes && + "Patchpoint can't request size less than the length of a call."); - // Otherwise this is a reg operand. - assert(MOP.isReg() && "Expected register operand here."); - assert(TargetRegisterInfo::isPhysicalRegister(MOP.getReg()) && - "Virtreg operands should have been rewritten before now."); - return std::make_pair(Location(Location::Register, MOP.getReg(), 0), ++MOI); + EmitNops(OutStreamer, NumBytes - EncodedBytes, Subtarget->is64Bit(), + getSubtargetInfo()); } -static MachineInstr::const_mop_iterator -getStackMapEndMOP(MachineInstr::const_mop_iterator MOI, - MachineInstr::const_mop_iterator MOE) { - for (; MOI != MOE; ++MOI) - if (MOI->isRegMask() || (MOI->isReg() && MOI->isImplicit())) - break; - - return MOI; +// Returns instruction preceding MBBI in MachineFunction. +// If MBBI is the first instruction of the first basic block, returns null. +static MachineBasicBlock::const_iterator +PrevCrossBBInst(MachineBasicBlock::const_iterator MBBI) { + const MachineBasicBlock *MBB = MBBI->getParent(); + while (MBBI == MBB->begin()) { + if (MBB == MBB->getParent()->begin()) + return nullptr; + MBB = MBB->getPrevNode(); + MBBI = MBB->end(); + } + return --MBBI; } -static void LowerSTACKMAP(MCStreamer &OutStreamer, - X86MCInstLower &MCInstLowering, - StackMaps &SM, - const MachineInstr &MI) -{ - int64_t ID = MI.getOperand(0).getImm(); - unsigned NumNOPBytes = MI.getOperand(1).getImm(); - - assert((int32_t)ID == ID && "Stack maps hold 32-bit IDs"); - SM.recordStackMap(MI, ID, llvm::next(MI.operands_begin(), 2), - getStackMapEndMOP(MI.operands_begin(), MI.operands_end())); - // Emit padding. - for (unsigned i = 0; i < NumNOPBytes; ++i) - OutStreamer.EmitInstruction(MCInstBuilder(X86::NOOP)); +static const Constant *getConstantFromPool(const MachineInstr &MI, + const MachineOperand &Op) { + if (!Op.isCPI()) + return nullptr; + + ArrayRef Constants = + MI.getParent()->getParent()->getConstantPool()->getConstants(); + const MachineConstantPoolEntry &ConstantEntry = + Constants[Op.getIndex()]; + + // Bail if this is a machine constant pool entry, we won't be able to dig out + // anything useful. + if (ConstantEntry.isMachineConstantPoolEntry()) + return nullptr; + + auto *C = dyn_cast(ConstantEntry.Val.ConstVal); + assert((!C || ConstantEntry.getType() == C->getType()) && + "Expected a constant of the same type!"); + return C; } -static void LowerPATCHPOINT(MCStreamer &OutStreamer, - X86MCInstLower &MCInstLowering, - StackMaps &SM, - const MachineInstr &MI) -{ - int64_t ID = MI.getOperand(0).getImm(); - assert((int32_t)ID == ID && "Stack maps hold 32-bit IDs"); - - // Get the number of arguments participating in the call. This number was - // adjusted during call lowering by subtracting stack args. - int64_t StackMapIdx = MI.getOperand(3).getImm() + 4; - assert(StackMapIdx <= MI.getNumOperands() && "Patchpoint dropped args."); - - SM.recordStackMap(MI, ID, llvm::next(MI.operands_begin(), StackMapIdx), - getStackMapEndMOP(MI.operands_begin(), MI.operands_end())); - - // Emit call. We need to know how many bytes we encoded here. - unsigned EncodedBytes = 2; - OutStreamer.EmitInstruction(MCInstBuilder(X86::CALL64r) - .addReg(MI.getOperand(2).getReg())); - - // Emit padding. - unsigned NumNOPBytes = MI.getOperand(1).getImm(); - assert(NumNOPBytes >= EncodedBytes && - "Patchpoint can't request size less than the length of a call."); +static std::string getShuffleComment(const MachineOperand &DstOp, + const MachineOperand &SrcOp, + ArrayRef Mask) { + std::string Comment; + + // Compute the name for a register. This is really goofy because we have + // multiple instruction printers that could (in theory) use different + // names. Fortunately most people use the ATT style (outside of Windows) + // and they actually agree on register naming here. Ultimately, this is + // a comment, and so its OK if it isn't perfect. + auto GetRegisterName = [](unsigned RegNum) -> StringRef { + return X86ATTInstPrinter::getRegisterName(RegNum); + }; + + StringRef DstName = DstOp.isReg() ? GetRegisterName(DstOp.getReg()) : "mem"; + StringRef SrcName = SrcOp.isReg() ? GetRegisterName(SrcOp.getReg()) : "mem"; + + raw_string_ostream CS(Comment); + CS << DstName << " = "; + bool NeedComma = false; + bool InSrc = false; + for (int M : Mask) { + // Wrap up any prior entry... + if (M == SM_SentinelZero && InSrc) { + InSrc = false; + CS << "]"; + } + if (NeedComma) + CS << ","; + else + NeedComma = true; + + // Print this shuffle... + if (M == SM_SentinelZero) { + CS << "zero"; + } else { + if (!InSrc) { + InSrc = true; + CS << SrcName << "["; + } + if (M == SM_SentinelUndef) + CS << "u"; + else + CS << M; + } + } + if (InSrc) + CS << "]"; + CS.flush(); - for (unsigned i = EncodedBytes; i < NumNOPBytes; ++i) - OutStreamer.EmitInstruction(MCInstBuilder(X86::NOOP)); + return Comment; } void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) { X86MCInstLower MCInstLowering(*MF, *this); + const X86RegisterInfo *RI = static_cast( + TM.getSubtargetImpl()->getRegisterInfo()); + switch (MI->getOpcode()) { case TargetOpcode::DBG_VALUE: llvm_unreachable("Should be handled target independently"); // Emit nothing here but a comment if we can. case X86::Int_MemBarrier: - if (OutStreamer.hasRawTextSupport()) - OutStreamer.EmitRawText(StringRef("\t#MEMBARRIER")); + OutStreamer.emitRawComment("MEMBARRIER"); return; @@ -836,7 +973,7 @@ void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) { case X86::TLS_addr64: case X86::TLS_base_addr32: case X86::TLS_base_addr64: - return LowerTlsAddr(OutStreamer, MCInstLowering, *MI); + return LowerTlsAddr(MCInstLowering, *MI); case X86::MOVPC32r: { // This is a pseudo op for a two instruction sequence with a label, which @@ -849,15 +986,15 @@ void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) { MCSymbol *PICBase = MF->getPICBaseSymbol(); // FIXME: We would like an efficient form for this, so we don't have to do a // lot of extra uniquing. - OutStreamer.EmitInstruction(MCInstBuilder(X86::CALLpcrel32) + EmitAndCountInstruction(MCInstBuilder(X86::CALLpcrel32) .addExpr(MCSymbolRefExpr::Create(PICBase, OutContext))); // Emit the label. OutStreamer.EmitLabel(PICBase); // popl $reg - OutStreamer.EmitInstruction(MCInstBuilder(X86::POP32r) - .addReg(MI->getOperand(0).getReg())); + EmitAndCountInstruction(MCInstBuilder(X86::POP32r) + .addReg(MI->getOperand(0).getReg())); return; } @@ -887,7 +1024,7 @@ void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) { DotExpr = MCBinaryExpr::CreateAdd(MCSymbolRefExpr::Create(OpSym,OutContext), DotExpr, OutContext); - OutStreamer.EmitInstruction(MCInstBuilder(X86::ADD32ri) + EmitAndCountInstruction(MCInstBuilder(X86::ADD32ri) .addReg(MI->getOperand(0).getReg()) .addReg(MI->getOperand(1).getReg()) .addExpr(DotExpr)); @@ -895,13 +1032,188 @@ void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) { } case TargetOpcode::STACKMAP: - return LowerSTACKMAP(OutStreamer, MCInstLowering, SM, *MI); + return LowerSTACKMAP(*MI); case TargetOpcode::PATCHPOINT: - return LowerPATCHPOINT(OutStreamer, MCInstLowering, SM, *MI); + return LowerPATCHPOINT(*MI); + + case X86::MORESTACK_RET: + EmitAndCountInstruction(MCInstBuilder(getRetOpcode(*Subtarget))); + return; + + case X86::MORESTACK_RET_RESTORE_R10: + // Return, then restore R10. + EmitAndCountInstruction(MCInstBuilder(getRetOpcode(*Subtarget))); + EmitAndCountInstruction(MCInstBuilder(X86::MOV64rr) + .addReg(X86::R10) + .addReg(X86::RAX)); + return; + + case X86::SEH_PushReg: + OutStreamer.EmitWinCFIPushReg(RI->getSEHRegNum(MI->getOperand(0).getImm())); + return; + + case X86::SEH_SaveReg: + OutStreamer.EmitWinCFISaveReg(RI->getSEHRegNum(MI->getOperand(0).getImm()), + MI->getOperand(1).getImm()); + return; + + case X86::SEH_SaveXMM: + OutStreamer.EmitWinCFISaveXMM(RI->getSEHRegNum(MI->getOperand(0).getImm()), + MI->getOperand(1).getImm()); + return; + + case X86::SEH_StackAlloc: + OutStreamer.EmitWinCFIAllocStack(MI->getOperand(0).getImm()); + return; + + case X86::SEH_SetFrame: + OutStreamer.EmitWinCFISetFrame(RI->getSEHRegNum(MI->getOperand(0).getImm()), + MI->getOperand(1).getImm()); + return; + + case X86::SEH_PushFrame: + OutStreamer.EmitWinCFIPushFrame(MI->getOperand(0).getImm()); + return; + + case X86::SEH_EndPrologue: + OutStreamer.EmitWinCFIEndProlog(); + return; + + case X86::SEH_Epilogue: { + MachineBasicBlock::const_iterator MBBI(MI); + // Check if preceded by a call and emit nop if so. + for (MBBI = PrevCrossBBInst(MBBI); MBBI; MBBI = PrevCrossBBInst(MBBI)) { + // Conservatively assume that pseudo instructions don't emit code and keep + // looking for a call. We may emit an unnecessary nop in some cases. + if (!MBBI->isPseudo()) { + if (MBBI->isCall()) + EmitAndCountInstruction(MCInstBuilder(X86::NOOP)); + break; + } + } + return; + } + + // Lower PSHUFB and VPERMILP normally but add a comment if we can find + // a constant shuffle mask. We won't be able to do this at the MC layer + // because the mask isn't an immediate. + case X86::PSHUFBrm: + case X86::VPSHUFBrm: + case X86::VPSHUFBYrm: { + if (!OutStreamer.isVerboseAsm()) + break; + assert(MI->getNumOperands() > 5 && + "We should always have at least 5 operands!"); + const MachineOperand &DstOp = MI->getOperand(0); + const MachineOperand &SrcOp = MI->getOperand(1); + const MachineOperand &MaskOp = MI->getOperand(5); + + if (auto *C = getConstantFromPool(*MI, MaskOp)) { + SmallVector Mask; + DecodePSHUFBMask(C, Mask); + if (!Mask.empty()) + OutStreamer.AddComment(getShuffleComment(DstOp, SrcOp, Mask)); + } + break; + } + case X86::VPERMILPSrm: + case X86::VPERMILPDrm: + case X86::VPERMILPSYrm: + case X86::VPERMILPDYrm: { + if (!OutStreamer.isVerboseAsm()) + break; + assert(MI->getNumOperands() > 5 && + "We should always have at least 5 operands!"); + const MachineOperand &DstOp = MI->getOperand(0); + const MachineOperand &SrcOp = MI->getOperand(1); + const MachineOperand &MaskOp = MI->getOperand(5); + + if (auto *C = getConstantFromPool(*MI, MaskOp)) { + SmallVector Mask; + DecodeVPERMILPMask(C, Mask); + if (!Mask.empty()) + OutStreamer.AddComment(getShuffleComment(DstOp, SrcOp, Mask)); + } + break; + } + + // For loads from a constant pool to a vector register, print the constant + // loaded. + case X86::MOVAPDrm: + case X86::VMOVAPDrm: + case X86::VMOVAPDYrm: + case X86::MOVUPDrm: + case X86::VMOVUPDrm: + case X86::VMOVUPDYrm: + case X86::MOVAPSrm: + case X86::VMOVAPSrm: + case X86::VMOVAPSYrm: + case X86::MOVUPSrm: + case X86::VMOVUPSrm: + case X86::VMOVUPSYrm: + case X86::MOVDQArm: + case X86::VMOVDQArm: + case X86::VMOVDQAYrm: + case X86::MOVDQUrm: + case X86::VMOVDQUrm: + case X86::VMOVDQUYrm: + if (!OutStreamer.isVerboseAsm()) + break; + if (MI->getNumOperands() > 4) + if (auto *C = getConstantFromPool(*MI, MI->getOperand(4))) { + std::string Comment; + raw_string_ostream CS(Comment); + const MachineOperand &DstOp = MI->getOperand(0); + CS << X86ATTInstPrinter::getRegisterName(DstOp.getReg()) << " = "; + if (auto *CDS = dyn_cast(C)) { + CS << "["; + for (int i = 0, NumElements = CDS->getNumElements(); i < NumElements; ++i) { + if (i != 0) + CS << ","; + if (CDS->getElementType()->isIntegerTy()) + CS << CDS->getElementAsInteger(i); + else if (CDS->getElementType()->isFloatTy()) + CS << CDS->getElementAsFloat(i); + else if (CDS->getElementType()->isDoubleTy()) + CS << CDS->getElementAsDouble(i); + else + CS << "?"; + } + CS << "]"; + OutStreamer.AddComment(CS.str()); + } else if (auto *CV = dyn_cast(C)) { + CS << "<"; + for (int i = 0, NumOperands = CV->getNumOperands(); i < NumOperands; ++i) { + if (i != 0) + CS << ","; + Constant *COp = CV->getOperand(i); + if (isa(COp)) { + CS << "u"; + } else if (auto *CI = dyn_cast(COp)) { + CS << CI->getZExtValue(); + } else if (auto *CF = dyn_cast(COp)) { + SmallString<32> Str; + CF->getValueAPF().toString(Str); + CS << Str; + } else { + CS << "?"; + } + } + CS << ">"; + OutStreamer.AddComment(CS.str()); + } + } + break; } MCInst TmpInst; MCInstLowering.Lower(MI, TmpInst); - OutStreamer.EmitInstruction(TmpInst); + EmitAndCountInstruction(TmpInst); + + // Stackmap shadows cannot include branch targets, so we can count the bytes + // in a call towards the shadow, but must flush the shadow immediately after + // to account for the return from the call. + if (MI->isCall()) + SMShadowTracker.emitShadowPadding(OutStreamer, getSubtargetInfo()); }