X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FX86%2FAsmParser%2FX86AsmParser.cpp;h=cbddec42cbbd77df9420a1cd0bc1089ba8e85bc3;hb=b6e0946d407acf62a3f55ae22ce578d5af3bc5cc;hp=4418a3c5e221c7197a50e08248e389bba507ab4e;hpb=89ab4e4cd0627be8b779c6a1e7023e9ea58c4b09;p=oota-llvm.git diff --git a/lib/Target/X86/AsmParser/X86AsmParser.cpp b/lib/Target/X86/AsmParser/X86AsmParser.cpp index 4418a3c5e22..cbddec42cbb 100644 --- a/lib/Target/X86/AsmParser/X86AsmParser.cpp +++ b/lib/Target/X86/AsmParser/X86AsmParser.cpp @@ -9,6 +9,7 @@ #include "MCTargetDesc/X86BaseInfo.h" #include "llvm/ADT/APFloat.h" +#include "llvm/ADT/STLExtras.h" #include "llvm/ADT/SmallString.h" #include "llvm/ADT/SmallVector.h" #include "llvm/ADT/StringSwitch.h" @@ -49,6 +50,12 @@ class X86AsmParser : public MCTargetAsmParser { MCAsmParser &Parser; ParseInstructionInfo *InstInfo; private: + SMLoc consumeToken() { + SMLoc Result = Parser.getTok().getLoc(); + Parser.Lex(); + return Result; + } + enum InfixCalculatorTok { IC_PLUS = 0, IC_MINUS, @@ -79,7 +86,7 @@ private: PostfixStack.push_back(std::make_pair(Op, Val)); } - void popOperator() { InfixOperatorStack.pop_back_val(); } + void popOperator() { InfixOperatorStack.pop_back(); } void pushOperator(InfixCalculatorTok Op) { // Push the new operator if the stack is empty. if (InfixOperatorStack.empty()) { @@ -117,12 +124,12 @@ private: if (StackOp == IC_RPAREN) { ++ParenCount; - InfixOperatorStack.pop_back_val(); + InfixOperatorStack.pop_back(); } else if (StackOp == IC_LPAREN) { --ParenCount; - InfixOperatorStack.pop_back_val(); + InfixOperatorStack.pop_back(); } else { - InfixOperatorStack.pop_back_val(); + InfixOperatorStack.pop_back(); PostfixStack.push_back(std::make_pair(StackOp, 0)); } } @@ -206,11 +213,12 @@ private: StringRef SymName; bool StopOnLBrac, AddImmPrefix; InfixCalculator IC; + InlineAsmIdentifierInfo Info; public: IntelExprStateMachine(int64_t imm, bool stoponlbrac, bool addimmprefix) : State(IES_PLUS), PrevState(IES_ERROR), BaseReg(0), IndexReg(0), TmpReg(0), Scale(1), Imm(imm), Sym(0), StopOnLBrac(stoponlbrac), - AddImmPrefix(addimmprefix) {} + AddImmPrefix(addimmprefix) { Info.clear(); } unsigned getBaseReg() { return BaseReg; } unsigned getIndexReg() { return IndexReg; } @@ -218,11 +226,17 @@ private: const MCExpr *getSym() { return Sym; } StringRef getSymName() { return SymName; } int64_t getImm() { return Imm + IC.execute(); } - bool isValidEndState() { return State == IES_RBRAC; } + bool isValidEndState() { + return State == IES_RBRAC || State == IES_INTEGER; + } bool getStopOnLBrac() { return StopOnLBrac; } bool getAddImmPrefix() { return AddImmPrefix; } bool hadError() { return State == IES_ERROR; } + InlineAsmIdentifierInfo &getIdentifierInfo() { + return Info; + } + void onPlus() { IntelExprState CurrState = State; switch (State) { @@ -314,7 +328,7 @@ private: } PrevState = CurrState; } - void onDispExpr(const MCExpr *SymRef, StringRef SymRefName) { + void onIdentifierExpr(const MCExpr *SymRef, StringRef SymRefName) { PrevState = State; switch (State) { default: @@ -472,7 +486,7 @@ private: MCAsmLexer &getLexer() const { return Parser.getLexer(); } bool Error(SMLoc L, const Twine &Msg, - ArrayRef Ranges = ArrayRef(), + ArrayRef Ranges = None, bool MatchingInlineAsm = false) { if (MatchingInlineAsm) return true; return Parser.Error(L, Msg, Ranges); @@ -487,21 +501,25 @@ private: X86Operand *ParseATTOperand(); X86Operand *ParseIntelOperand(); X86Operand *ParseIntelOffsetOfOperator(); - X86Operand *ParseIntelDotOperator(const MCExpr *Disp, const MCExpr *&NewDisp); + bool ParseIntelDotOperator(const MCExpr *Disp, const MCExpr *&NewDisp); X86Operand *ParseIntelOperator(unsigned OpKind); - X86Operand *ParseIntelMemOperand(unsigned SegReg, int64_t ImmDisp, - SMLoc StartLoc); - X86Operand *ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End); + X86Operand *ParseIntelSegmentOverride(unsigned SegReg, SMLoc Start, unsigned Size); + X86Operand *ParseIntelMemOperand(int64_t ImmDisp, SMLoc StartLoc, + unsigned Size); + bool ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End); X86Operand *ParseIntelBracExpression(unsigned SegReg, SMLoc Start, int64_t ImmDisp, unsigned Size); - X86Operand *ParseIntelVarWithQualifier(const MCExpr *&Disp, - StringRef &Identifier); + bool ParseIntelIdentifier(const MCExpr *&Val, StringRef &Identifier, + InlineAsmIdentifierInfo &Info, + bool IsUnevaluatedOperand, SMLoc &End); + X86Operand *ParseMemOperand(unsigned SegReg, SMLoc StartLoc); X86Operand *CreateMemForInlineAsm(unsigned SegReg, const MCExpr *Disp, unsigned BaseReg, unsigned IndexReg, unsigned Scale, SMLoc Start, SMLoc End, - unsigned Size, StringRef SymName); + unsigned Size, StringRef Identifier, + InlineAsmIdentifierInfo &Info); bool ParseDirectiveWord(unsigned Size, SMLoc L); bool ParseDirectiveCode(StringRef IDVal, SMLoc L); @@ -526,9 +544,21 @@ private: // FIXME: Can tablegen auto-generate this? return (STI.getFeatureBits() & X86::Mode64Bit) != 0; } - void SwitchMode() { - unsigned FB = ComputeAvailableFeatures(STI.ToggleFeature(X86::Mode64Bit)); + bool is32BitMode() const { + // FIXME: Can tablegen auto-generate this? + return (STI.getFeatureBits() & X86::Mode32Bit) != 0; + } + bool is16BitMode() const { + // FIXME: Can tablegen auto-generate this? + return (STI.getFeatureBits() & X86::Mode16Bit) != 0; + } + void SwitchMode(uint64_t mode) { + uint64_t oldMode = STI.getFeatureBits() & + (X86::Mode64Bit | X86::Mode32Bit | X86::Mode16Bit); + unsigned FB = ComputeAvailableFeatures(STI.ToggleFeature(oldMode | mode)); setAvailableFeatures(FB); + assert(mode == (STI.getFeatureBits() & + (X86::Mode64Bit | X86::Mode32Bit | X86::Mode16Bit))); } bool isParsingIntelSyntax() { @@ -544,8 +574,9 @@ private: /// } public: - X86AsmParser(MCSubtargetInfo &sti, MCAsmParser &parser) - : MCTargetAsmParser(), STI(sti), Parser(parser), InstInfo(0) { + X86AsmParser(MCSubtargetInfo &sti, MCAsmParser &parser, + const MCInstrInfo &MII) + : MCTargetAsmParser(), STI(sti), Parser(parser), InstInfo(0) { // Initialize the set of available features. setAvailableFeatures(ComputeAvailableFeatures(STI.getFeatureBits())); @@ -607,6 +638,7 @@ struct X86Operand : public MCParsedAsmOperand { SMLoc StartLoc, EndLoc; SMLoc OffsetOfLoc; StringRef SymName; + void *OpDecl; bool AddressOf; struct TokOp { @@ -642,6 +674,7 @@ struct X86Operand : public MCParsedAsmOperand { : Kind(K), StartLoc(Start), EndLoc(End) {} StringRef getSymName() { return SymName; } + void *getOpDecl() { return OpDecl; } /// getStartLoc - Get the location of the first token of this operand. SMLoc getStartLoc() const { return StartLoc; } @@ -801,6 +834,9 @@ struct X86Operand : public MCParsedAsmOperand { bool isMem256() const { return Kind == Memory && (!Mem.Size || Mem.Size == 256); } + bool isMem512() const { + return Kind == Memory && (!Mem.Size || Mem.Size == 512); + } bool isMemVX32() const { return Kind == Memory && (!Mem.Size || Mem.Size == 32) && @@ -818,14 +854,45 @@ struct X86Operand : public MCParsedAsmOperand { return Kind == Memory && (!Mem.Size || Mem.Size == 64) && getMemIndexReg() >= X86::YMM0 && getMemIndexReg() <= X86::YMM15; } + bool isMemVZ32() const { + return Kind == Memory && (!Mem.Size || Mem.Size == 32) && + getMemIndexReg() >= X86::ZMM0 && getMemIndexReg() <= X86::ZMM31; + } + bool isMemVZ64() const { + return Kind == Memory && (!Mem.Size || Mem.Size == 64) && + getMemIndexReg() >= X86::ZMM0 && getMemIndexReg() <= X86::ZMM31; + } bool isAbsMem() const { return Kind == Memory && !getMemSegReg() && !getMemBaseReg() && !getMemIndexReg() && getMemScale() == 1; } + bool isMemOffs8() const { + return Kind == Memory && !getMemSegReg() && !getMemBaseReg() && + !getMemIndexReg() && getMemScale() == 1 && (!Mem.Size || Mem.Size == 8); + } + bool isMemOffs16() const { + return Kind == Memory && !getMemSegReg() && !getMemBaseReg() && + !getMemIndexReg() && getMemScale() == 1 && (!Mem.Size || Mem.Size == 16); + } + bool isMemOffs32() const { + return Kind == Memory && !getMemSegReg() && !getMemBaseReg() && + !getMemIndexReg() && getMemScale() == 1 && (!Mem.Size || Mem.Size == 32); + } + bool isMemOffs64() const { + return Kind == Memory && !getMemSegReg() && !getMemBaseReg() && + !getMemIndexReg() && getMemScale() == 1 && (!Mem.Size || Mem.Size == 64); + } + bool isReg() const { return Kind == Register; } + bool isGR32orGR64() const { + return Kind == Register && + (X86MCRegisterClasses[X86::GR32RegClassID].contains(getReg()) || + X86MCRegisterClasses[X86::GR64RegClassID].contains(getReg())); + } + void addExpr(MCInst &Inst, const MCExpr *Expr) const { // Add as immediates when possible. if (const MCConstantExpr *CE = dyn_cast(Expr)) @@ -839,43 +906,40 @@ struct X86Operand : public MCParsedAsmOperand { Inst.addOperand(MCOperand::CreateReg(getReg())); } - void addImmOperands(MCInst &Inst, unsigned N) const { - assert(N == 1 && "Invalid number of operands!"); - addExpr(Inst, getImm()); + static unsigned getGR32FromGR64(unsigned RegNo) { + switch (RegNo) { + default: llvm_unreachable("Unexpected register"); + case X86::RAX: return X86::EAX; + case X86::RCX: return X86::ECX; + case X86::RDX: return X86::EDX; + case X86::RBX: return X86::EBX; + case X86::RBP: return X86::EBP; + case X86::RSP: return X86::ESP; + case X86::RSI: return X86::ESI; + case X86::RDI: return X86::EDI; + case X86::R8: return X86::R8D; + case X86::R9: return X86::R9D; + case X86::R10: return X86::R10D; + case X86::R11: return X86::R11D; + case X86::R12: return X86::R12D; + case X86::R13: return X86::R13D; + case X86::R14: return X86::R14D; + case X86::R15: return X86::R15D; + case X86::RIP: return X86::EIP; + } } - void addMem8Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMem16Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMem32Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMem64Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMem80Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMem128Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMem256Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMemVX32Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMemVY32Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); - } - void addMemVX64Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); + void addGR32orGR64Operands(MCInst &Inst, unsigned N) const { + assert(N == 1 && "Invalid number of operands!"); + unsigned RegNo = getReg(); + if (X86MCRegisterClasses[X86::GR64RegClassID].contains(RegNo)) + RegNo = getGR32FromGR64(RegNo); + Inst.addOperand(MCOperand::CreateReg(RegNo)); } - void addMemVY64Operands(MCInst &Inst, unsigned N) const { - addMemOperands(Inst, N); + + void addImmOperands(MCInst &Inst, unsigned N) const { + assert(N == 1 && "Invalid number of operands!"); + addExpr(Inst, getImm()); } void addMemOperands(MCInst &Inst, unsigned N) const { @@ -896,6 +960,15 @@ struct X86Operand : public MCParsedAsmOperand { Inst.addOperand(MCOperand::CreateExpr(getMemDisp())); } + void addMemOffsOperands(MCInst &Inst, unsigned N) const { + assert((N == 1) && "Invalid number of operands!"); + // Add as immediates when possible. + if (const MCConstantExpr *CE = dyn_cast(getMemDisp())) + Inst.addOperand(MCOperand::CreateImm(CE->getValue())); + else + Inst.addOperand(MCOperand::CreateExpr(getMemDisp())); + } + static X86Operand *CreateToken(StringRef Str, SMLoc Loc) { SMLoc EndLoc = SMLoc::getFromPointer(Loc.getPointer() + Str.size()); X86Operand *Res = new X86Operand(Token, Loc, EndLoc); @@ -907,12 +980,14 @@ struct X86Operand : public MCParsedAsmOperand { static X86Operand *CreateReg(unsigned RegNo, SMLoc StartLoc, SMLoc EndLoc, bool AddressOf = false, SMLoc OffsetOfLoc = SMLoc(), - StringRef SymName = StringRef()) { + StringRef SymName = StringRef(), + void *OpDecl = 0) { X86Operand *Res = new X86Operand(Register, StartLoc, EndLoc); Res->Reg.RegNo = RegNo; Res->AddressOf = AddressOf; Res->OffsetOfLoc = OffsetOfLoc; Res->SymName = SymName; + Res->OpDecl = OpDecl; return Res; } @@ -924,8 +999,8 @@ struct X86Operand : public MCParsedAsmOperand { /// Create an absolute memory operand. static X86Operand *CreateMem(const MCExpr *Disp, SMLoc StartLoc, SMLoc EndLoc, - unsigned Size = 0, - StringRef SymName = StringRef()) { + unsigned Size = 0, StringRef SymName = StringRef(), + void *OpDecl = 0) { X86Operand *Res = new X86Operand(Memory, StartLoc, EndLoc); Res->Mem.SegReg = 0; Res->Mem.Disp = Disp; @@ -933,8 +1008,9 @@ struct X86Operand : public MCParsedAsmOperand { Res->Mem.IndexReg = 0; Res->Mem.Scale = 1; Res->Mem.Size = Size; - Res->SymName = SymName; - Res->AddressOf = false; + Res->SymName = SymName; + Res->OpDecl = OpDecl; + Res->AddressOf = false; return Res; } @@ -943,7 +1019,8 @@ struct X86Operand : public MCParsedAsmOperand { unsigned BaseReg, unsigned IndexReg, unsigned Scale, SMLoc StartLoc, SMLoc EndLoc, unsigned Size = 0, - StringRef SymName = StringRef()) { + StringRef SymName = StringRef(), + void *OpDecl = 0) { // We should never just have a displacement, that should be parsed as an // absolute memory operand. assert((SegReg || BaseReg || IndexReg) && "Invalid memory operand!"); @@ -958,8 +1035,9 @@ struct X86Operand : public MCParsedAsmOperand { Res->Mem.IndexReg = IndexReg; Res->Mem.Scale = Scale; Res->Mem.Size = Size; - Res->SymName = SymName; - Res->AddressOf = false; + Res->SymName = SymName; + Res->OpDecl = OpDecl; + Res->AddressOf = false; return Res; } }; @@ -967,7 +1045,8 @@ struct X86Operand : public MCParsedAsmOperand { } // end anonymous namespace. bool X86AsmParser::isSrcOp(X86Operand &Op) { - unsigned basereg = is64BitMode() ? X86::RSI : X86::ESI; + unsigned basereg = + is64BitMode() ? X86::RSI : (is32BitMode() ? X86::ESI : X86::SI); return (Op.isMem() && (Op.Mem.SegReg == 0 || Op.Mem.SegReg == X86::DS) && @@ -977,7 +1056,8 @@ bool X86AsmParser::isSrcOp(X86Operand &Op) { } bool X86AsmParser::isDstOp(X86Operand &Op) { - unsigned basereg = is64BitMode() ? X86::RDI : X86::EDI; + unsigned basereg = + is64BitMode() ? X86::RDI : (is32BitMode() ? X86::EDI : X86::DI); return Op.isMem() && (Op.Mem.SegReg == 0 || Op.Mem.SegReg == X86::ES) && @@ -1013,7 +1093,7 @@ bool X86AsmParser::ParseRegister(unsigned &RegNo, RegNo = MatchRegisterName(Tok.getString().lower()); if (!is64BitMode()) { - // FIXME: This should be done using Requires and + // FIXME: This should be done using Requires and // Requires so "eiz" usage in 64-bit instructions can be also // checked. // FIXME: Check AH, CH, DH, BH cannot be used in an instruction requiring a @@ -1119,42 +1199,33 @@ X86Operand * X86AsmParser::CreateMemForInlineAsm(unsigned SegReg, const MCExpr *Disp, unsigned BaseReg, unsigned IndexReg, unsigned Scale, SMLoc Start, SMLoc End, - unsigned Size, StringRef SymName) { - bool NeedSizeDir = false; - if (const MCSymbolRefExpr *SymRef = dyn_cast(Disp)) { - const MCSymbol &Sym = SymRef->getSymbol(); - // FIXME: The SemaLookup will fail if the name is anything other then an - // identifier. - // FIXME: Pass a valid SMLoc. - bool IsVarDecl = false; - unsigned tLength, tSize, tType; - SemaCallback->LookupInlineAsmIdentifier(Sym.getName(), NULL, tLength, tSize, - tType, IsVarDecl); - if (!Size) { - Size = tType * 8; // Size is in terms of bits in this context. - NeedSizeDir = Size > 0; - } + unsigned Size, StringRef Identifier, + InlineAsmIdentifierInfo &Info){ + if (isa(Disp)) { // If this is not a VarDecl then assume it is a FuncDecl or some other label // reference. We need an 'r' constraint here, so we need to create register // operand to ensure proper matching. Just pick a GPR based on the size of // a pointer. - if (!IsVarDecl) { - unsigned RegNo = is64BitMode() ? X86::RBX : X86::EBX; + if (!Info.IsVarDecl) { + unsigned RegNo = + is64BitMode() ? X86::RBX : (is32BitMode() ? X86::EBX : X86::BX); return X86Operand::CreateReg(RegNo, Start, End, /*AddressOf=*/true, - SMLoc(), SymName); + SMLoc(), Identifier, Info.OpDecl); + } + if (!Size) { + Size = Info.Type * 8; // Size is in terms of bits in this context. + if (Size) + InstInfo->AsmRewrites->push_back(AsmRewrite(AOK_SizeDirective, Start, + /*Len=*/0, Size)); } } - if (NeedSizeDir) - InstInfo->AsmRewrites->push_back(AsmRewrite(AOK_SizeDirective, Start, - /*Len=*/0, Size)); - // When parsing inline assembly we set the base register to a non-zero value // if we don't know the actual value at this time. This is necessary to // get the matching correct in some cases. BaseReg = BaseReg ? BaseReg : 1; return X86Operand::CreateMem(SegReg, Disp, BaseReg, IndexReg, Scale, Start, - End, Size, SymName); + End, Size, Identifier, Info.OpDecl); } static void @@ -1189,6 +1260,7 @@ RewriteIntelBracExpression(SmallVectorImpl *AsmRewrites, } } assert (Found && "Unable to rewrite ImmDisp."); + (void)Found; } else { // We have a symbolic and an immediate displacement, but no displacement // before the bracketed expression. Put the immediate displacement @@ -1218,8 +1290,7 @@ RewriteIntelBracExpression(SmallVectorImpl *AsmRewrites, } } -X86Operand * -X86AsmParser::ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End) { +bool X86AsmParser::ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End) { const AsmToken &Tok = Parser.getTok(); bool Done = false; @@ -1241,7 +1312,7 @@ X86AsmParser::ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End) { Done = true; break; } - return ErrorOperand(Tok.getLoc(), "Unexpected token!"); + return Error(Tok.getLoc(), "unknown token in expression"); } case AsmToken::EndOfStatement: { Done = true; @@ -1250,29 +1321,60 @@ X86AsmParser::ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End) { case AsmToken::Identifier: { // This could be a register or a symbolic displacement. unsigned TmpReg; - const MCExpr *Disp = 0; + const MCExpr *Val; SMLoc IdentLoc = Tok.getLoc(); StringRef Identifier = Tok.getString(); if(!ParseRegister(TmpReg, IdentLoc, End)) { SM.onRegister(TmpReg); UpdateLocLex = false; break; - } else if (!getParser().parsePrimaryExpr(Disp, End)) { - if (X86Operand *Err = ParseIntelVarWithQualifier(Disp, Identifier)) - return Err; - - SM.onDispExpr(Disp, Identifier); + } else { + if (!isParsingInlineAsm()) { + if (getParser().parsePrimaryExpr(Val, End)) + return Error(Tok.getLoc(), "Unexpected identifier!"); + } else { + InlineAsmIdentifierInfo &Info = SM.getIdentifierInfo(); + if (ParseIntelIdentifier(Val, Identifier, Info, + /*Unevaluated=*/false, End)) + return true; + } + SM.onIdentifierExpr(Val, Identifier); UpdateLocLex = false; break; } - return ErrorOperand(Tok.getLoc(), "Unexpected identifier!"); + return Error(Tok.getLoc(), "Unexpected identifier!"); } - case AsmToken::Integer: + case AsmToken::Integer: { if (isParsingInlineAsm() && SM.getAddImmPrefix()) InstInfo->AsmRewrites->push_back(AsmRewrite(AOK_ImmPrefix, Tok.getLoc())); - SM.onInteger(Tok.getIntVal()); + // Look for 'b' or 'f' following an Integer as a directional label + SMLoc Loc = getTok().getLoc(); + int64_t IntVal = getTok().getIntVal(); + End = consumeToken(); + UpdateLocLex = false; + if (getLexer().getKind() == AsmToken::Identifier) { + StringRef IDVal = getTok().getString(); + if (IDVal == "f" || IDVal == "b") { + MCSymbol *Sym = + getContext().GetDirectionalLocalSymbol(IntVal, + IDVal == "f" ? 1 : 0); + MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; + const MCExpr *Val = + MCSymbolRefExpr::Create(Sym, Variant, getContext()); + if (IDVal == "b" && Sym->isUndefined()) + return Error(Loc, "invalid reference to undefined symbol"); + StringRef Identifier = Sym->getName(); + SM.onIdentifierExpr(Val, Identifier); + End = consumeToken(); + } else { + SM.onInteger(IntVal); + } + } else { + SM.onInteger(IntVal); + } break; + } case AsmToken::Plus: SM.onPlus(); break; case AsmToken::Minus: SM.onMinus(); break; case AsmToken::Star: SM.onStar(); break; @@ -1283,14 +1385,12 @@ X86AsmParser::ParseIntelExpression(IntelExprStateMachine &SM, SMLoc &End) { case AsmToken::RParen: SM.onRParen(); break; } if (SM.hadError()) - return ErrorOperand(Tok.getLoc(), "Unexpected token!"); + return Error(Tok.getLoc(), "unknown token in expression"); - if (!Done && UpdateLocLex) { - End = Tok.getLoc(); - Parser.Lex(); // Consume the token. - } + if (!Done && UpdateLocLex) + End = consumeToken(); } - return 0; + return false; } X86Operand *X86AsmParser::ParseIntelBracExpression(unsigned SegReg, SMLoc Start, @@ -1307,8 +1407,8 @@ X86Operand *X86AsmParser::ParseIntelBracExpression(unsigned SegReg, SMLoc Start, // may have already parsed an immediate displacement before the bracketed // expression. IntelExprStateMachine SM(ImmDisp, /*StopOnLBrac=*/false, /*AddImmPrefix=*/true); - if (X86Operand *Err = ParseIntelExpression(SM, End)) - return Err; + if (ParseIntelExpression(SM, End)) + return 0; const MCExpr *Disp; if (const MCExpr *Sym = SM.getSym()) { @@ -1326,8 +1426,8 @@ X86Operand *X86AsmParser::ParseIntelBracExpression(unsigned SegReg, SMLoc Start, // Parse the dot operator (e.g., [ebx].foo.bar). if (Tok.getString().startswith(".")) { const MCExpr *NewDisp; - if (X86Operand *Err = ParseIntelDotOperator(Disp, NewDisp)) - return Err; + if (ParseIntelDotOperator(Disp, NewDisp)) + return 0; End = Tok.getEndLoc(); Parser.Lex(); // Eat the field. @@ -1337,121 +1437,135 @@ X86Operand *X86AsmParser::ParseIntelBracExpression(unsigned SegReg, SMLoc Start, int BaseReg = SM.getBaseReg(); int IndexReg = SM.getIndexReg(); int Scale = SM.getScale(); - - if (isParsingInlineAsm()) - return CreateMemForInlineAsm(SegReg, Disp, BaseReg, IndexReg, Scale, Start, - End, Size, SM.getSymName()); - - // handle [-42] - if (!BaseReg && !IndexReg) { - if (!SegReg) - return X86Operand::CreateMem(Disp, Start, End, Size); - else - return X86Operand::CreateMem(SegReg, Disp, 0, 0, 1, Start, End, Size); + if (!isParsingInlineAsm()) { + // handle [-42] + if (!BaseReg && !IndexReg) { + if (!SegReg) + return X86Operand::CreateMem(Disp, Start, End, Size); + else + return X86Operand::CreateMem(SegReg, Disp, 0, 0, 1, Start, End, Size); + } + return X86Operand::CreateMem(SegReg, Disp, BaseReg, IndexReg, Scale, Start, + End, Size); } - return X86Operand::CreateMem(SegReg, Disp, BaseReg, IndexReg, Scale, Start, - End, Size); + + InlineAsmIdentifierInfo &Info = SM.getIdentifierInfo(); + return CreateMemForInlineAsm(SegReg, Disp, BaseReg, IndexReg, Scale, Start, + End, Size, SM.getSymName(), Info); } // Inline assembly may use variable names with namespace alias qualifiers. -X86Operand *X86AsmParser::ParseIntelVarWithQualifier(const MCExpr *&Disp, - StringRef &Identifier) { - // We should only see Foo::Bar if we're parsing inline assembly. - if (!isParsingInlineAsm()) - return 0; +bool X86AsmParser::ParseIntelIdentifier(const MCExpr *&Val, + StringRef &Identifier, + InlineAsmIdentifierInfo &Info, + bool IsUnevaluatedOperand, SMLoc &End) { + assert (isParsingInlineAsm() && "Expected to be parsing inline assembly."); + Val = 0; - // If we don't see a ':' then there can't be a qualifier. - if (getLexer().isNot(AsmToken::Colon)) - return 0; + StringRef LineBuf(Identifier.data()); + SemaCallback->LookupInlineAsmIdentifier(LineBuf, Info, IsUnevaluatedOperand); - bool Done = false; const AsmToken &Tok = Parser.getTok(); - AsmToken IdentEnd = Tok; - while (!Done) { - switch (getLexer().getKind()) { - default: - Done = true; - break; - case AsmToken::Colon: - getLexer().Lex(); // Consume ':'. - if (getLexer().isNot(AsmToken::Colon)) - return ErrorOperand(Tok.getLoc(), "Expected ':' token!"); - getLexer().Lex(); // Consume second ':'. - if (getLexer().isNot(AsmToken::Identifier)) - return ErrorOperand(Tok.getLoc(), "Expected an identifier token!"); - break; - case AsmToken::Identifier: - IdentEnd = Tok; - getLexer().Lex(); // Consume the identifier. - break; - } + + // Advance the token stream until the end of the current token is + // after the end of what the frontend claimed. + const char *EndPtr = Tok.getLoc().getPointer() + LineBuf.size(); + while (true) { + End = Tok.getEndLoc(); + getLexer().Lex(); + + assert(End.getPointer() <= EndPtr && "frontend claimed part of a token?"); + if (End.getPointer() == EndPtr) break; } - unsigned Len = IdentEnd.getLoc().getPointer() - Identifier.data(); - Identifier = StringRef(Identifier.data(), Len + IdentEnd.getString().size()); + // Create the symbol reference. + Identifier = LineBuf; MCSymbol *Sym = getContext().GetOrCreateSymbol(Identifier); MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; - Disp = MCSymbolRefExpr::Create(Sym, Variant, getParser().getContext()); - return 0; + Val = MCSymbolRefExpr::Create(Sym, Variant, getParser().getContext()); + return false; } -/// ParseIntelMemOperand - Parse intel style memory operand. -X86Operand *X86AsmParser::ParseIntelMemOperand(unsigned SegReg, - int64_t ImmDisp, - SMLoc Start) { - const AsmToken &Tok = Parser.getTok(); - SMLoc End; - - unsigned Size = getIntelMemOperandSize(Tok.getString()); - if (Size) { - Parser.Lex(); - assert ((Tok.getString() == "PTR" || Tok.getString() == "ptr") && - "Unexpected token!"); - Parser.Lex(); - } - - // Parse ImmDisp [ BaseReg + Scale*IndexReg + Disp ]. +/// \brief Parse intel style segment override. +X86Operand *X86AsmParser::ParseIntelSegmentOverride(unsigned SegReg, + SMLoc Start, + unsigned Size) { + assert(SegReg != 0 && "Tried to parse a segment override without a segment!"); + const AsmToken &Tok = Parser.getTok(); // Eat colon. + if (Tok.isNot(AsmToken::Colon)) + return ErrorOperand(Tok.getLoc(), "Expected ':' token!"); + Parser.Lex(); // Eat ':' + + int64_t ImmDisp = 0; if (getLexer().is(AsmToken::Integer)) { + ImmDisp = Tok.getIntVal(); + AsmToken ImmDispToken = Parser.Lex(); // Eat the integer. + if (isParsingInlineAsm()) - InstInfo->AsmRewrites->push_back(AsmRewrite(AOK_ImmPrefix, - Tok.getLoc())); - int64_t ImmDisp = Tok.getIntVal(); - Parser.Lex(); // Eat the integer. - if (getLexer().isNot(AsmToken::LBrac)) - return ErrorOperand(Start, "Expected '[' token!"); - return ParseIntelBracExpression(SegReg, Start, ImmDisp, Size); + InstInfo->AsmRewrites->push_back( + AsmRewrite(AOK_ImmPrefix, ImmDispToken.getLoc())); + + if (getLexer().isNot(AsmToken::LBrac)) { + // An immediate following a 'segment register', 'colon' token sequence can + // be followed by a bracketed expression. If it isn't we know we have our + // final segment override. + const MCExpr *Disp = MCConstantExpr::Create(ImmDisp, getContext()); + return X86Operand::CreateMem(SegReg, Disp, /*BaseReg=*/0, /*IndexReg=*/0, + /*Scale=*/1, Start, ImmDispToken.getEndLoc(), + Size); + } } if (getLexer().is(AsmToken::LBrac)) return ParseIntelBracExpression(SegReg, Start, ImmDisp, Size); - if (!ParseRegister(SegReg, Start, End)) { - // Handel SegReg : [ ... ] - if (getLexer().isNot(AsmToken::Colon)) - return ErrorOperand(Start, "Expected ':' token!"); - Parser.Lex(); // Eat : - if (getLexer().isNot(AsmToken::LBrac)) - return ErrorOperand(Start, "Expected '[' token!"); - return ParseIntelBracExpression(SegReg, Start, ImmDisp, Size); + const MCExpr *Val; + SMLoc End; + if (!isParsingInlineAsm()) { + if (getParser().parsePrimaryExpr(Val, End)) + return ErrorOperand(Tok.getLoc(), "unknown token in expression"); + + return X86Operand::CreateMem(Val, Start, End, Size); } - const MCExpr *Disp = 0; + InlineAsmIdentifierInfo Info; StringRef Identifier = Tok.getString(); - if (getParser().parsePrimaryExpr(Disp, End)) + if (ParseIntelIdentifier(Val, Identifier, Info, + /*Unevaluated=*/false, End)) return 0; + return CreateMemForInlineAsm(/*SegReg=*/0, Val, /*BaseReg=*/0,/*IndexReg=*/0, + /*Scale=*/1, Start, End, Size, Identifier, Info); +} + +/// ParseIntelMemOperand - Parse intel style memory operand. +X86Operand *X86AsmParser::ParseIntelMemOperand(int64_t ImmDisp, SMLoc Start, + unsigned Size) { + const AsmToken &Tok = Parser.getTok(); + SMLoc End; + + // Parse ImmDisp [ BaseReg + Scale*IndexReg + Disp ]. + if (getLexer().is(AsmToken::LBrac)) + return ParseIntelBracExpression(/*SegReg=*/0, Start, ImmDisp, Size); - if (!isParsingInlineAsm()) - return X86Operand::CreateMem(Disp, Start, End, Size); + const MCExpr *Val; + if (!isParsingInlineAsm()) { + if (getParser().parsePrimaryExpr(Val, End)) + return ErrorOperand(Tok.getLoc(), "unknown token in expression"); - if (X86Operand *Err = ParseIntelVarWithQualifier(Disp, Identifier)) - return Err; + return X86Operand::CreateMem(Val, Start, End, Size); + } - return CreateMemForInlineAsm(/*SegReg=*/0, Disp, /*BaseReg=*/0,/*IndexReg=*/0, - /*Scale=*/1, Start, End, Size, Identifier); + InlineAsmIdentifierInfo Info; + StringRef Identifier = Tok.getString(); + if (ParseIntelIdentifier(Val, Identifier, Info, + /*Unevaluated=*/false, End)) + return 0; + return CreateMemForInlineAsm(/*SegReg=*/0, Val, /*BaseReg=*/0, /*IndexReg=*/0, + /*Scale=*/1, Start, End, Size, Identifier, Info); } /// Parse the '.' operator. -X86Operand *X86AsmParser::ParseIntelDotOperator(const MCExpr *Disp, +bool X86AsmParser::ParseIntelDotOperator(const MCExpr *Disp, const MCExpr *&NewDisp) { const AsmToken &Tok = Parser.getTok(); int64_t OrigDispVal, DotDispVal; @@ -1460,7 +1574,7 @@ X86Operand *X86AsmParser::ParseIntelDotOperator(const MCExpr *Disp, if (const MCConstantExpr *OrigDisp = dyn_cast(Disp)) OrigDispVal = OrigDisp->getValue(); else - return ErrorOperand(Tok.getLoc(), "Non-constant offsets are not supported!"); + return Error(Tok.getLoc(), "Non-constant offsets are not supported!"); // Drop the '.'. StringRef DotDispStr = Tok.getString().drop_front(1); @@ -1475,10 +1589,10 @@ X86Operand *X86AsmParser::ParseIntelDotOperator(const MCExpr *Disp, std::pair BaseMember = DotDispStr.split('.'); if (SemaCallback->LookupInlineAsmField(BaseMember.first, BaseMember.second, DotDisp)) - return ErrorOperand(Tok.getLoc(), "Unable to lookup field reference!"); + return Error(Tok.getLoc(), "Unable to lookup field reference!"); DotDispVal = DotDisp; } else - return ErrorOperand(Tok.getLoc(), "Unexpected token type!"); + return Error(Tok.getLoc(), "Unexpected token type!"); if (isParsingInlineAsm() && Tok.is(AsmToken::Identifier)) { SMLoc Loc = SMLoc::getFromPointer(DotDispStr.data()); @@ -1489,7 +1603,7 @@ X86Operand *X86AsmParser::ParseIntelDotOperator(const MCExpr *Disp, } NewDisp = MCConstantExpr::Create(OrigDispVal + DotDispVal, getContext()); - return 0; + return false; } /// Parse the 'offset' operator. This operator is used to specify the @@ -1498,17 +1612,14 @@ X86Operand *X86AsmParser::ParseIntelOffsetOfOperator() { const AsmToken &Tok = Parser.getTok(); SMLoc OffsetOfLoc = Tok.getLoc(); Parser.Lex(); // Eat offset. - assert (Tok.is(AsmToken::Identifier) && "Expected an identifier"); const MCExpr *Val; + InlineAsmIdentifierInfo Info; SMLoc Start = Tok.getLoc(), End; StringRef Identifier = Tok.getString(); - if (getParser().parsePrimaryExpr(Val, End)) - return ErrorOperand(Start, "Unable to parse expression!"); - - const MCExpr *Disp = 0; - if (X86Operand *Err = ParseIntelVarWithQualifier(Disp, Identifier)) - return Err; + if (ParseIntelIdentifier(Val, Identifier, Info, + /*Unevaluated=*/false, End)) + return 0; // Don't emit the offset operator. InstInfo->AsmRewrites->push_back(AsmRewrite(AOK_Skip, OffsetOfLoc, 7)); @@ -1516,9 +1627,10 @@ X86Operand *X86AsmParser::ParseIntelOffsetOfOperator() { // The offset operator will have an 'r' constraint, thus we need to create // register operand to ensure proper matching. Just pick a GPR based on // the size of a pointer. - unsigned RegNo = is64BitMode() ? X86::RBX : X86::EBX; + unsigned RegNo = + is64BitMode() ? X86::RBX : (is32BitMode() ? X86::EBX : X86::BX); return X86Operand::CreateReg(RegNo, Start, End, /*GetAddress=*/true, - OffsetOfLoc, Identifier); + OffsetOfLoc, Identifier, Info.OpDecl); } enum IntelOperatorKind { @@ -1537,39 +1649,24 @@ X86Operand *X86AsmParser::ParseIntelOperator(unsigned OpKind) { const AsmToken &Tok = Parser.getTok(); SMLoc TypeLoc = Tok.getLoc(); Parser.Lex(); // Eat operator. - assert (Tok.is(AsmToken::Identifier) && "Expected an identifier"); - const MCExpr *Val; - AsmToken StartTok = Tok; + const MCExpr *Val = 0; + InlineAsmIdentifierInfo Info; SMLoc Start = Tok.getLoc(), End; StringRef Identifier = Tok.getString(); - if (getParser().parsePrimaryExpr(Val, End)) - return ErrorOperand(Start, "Unable to parse expression!"); - - const MCExpr *Disp = 0; - if (X86Operand *Err = ParseIntelVarWithQualifier(Disp, Identifier)) - return Err; - - unsigned Length = 0, Size = 0, Type = 0; - if (const MCSymbolRefExpr *SymRef = dyn_cast(Val)) { - const MCSymbol &Sym = SymRef->getSymbol(); - // FIXME: The SemaLookup will fail if the name is anything other then an - // identifier. - // FIXME: Pass a valid SMLoc. - bool IsVarDecl; - if (!SemaCallback->LookupInlineAsmIdentifier(Sym.getName(), NULL, Length, - Size, Type, IsVarDecl)) - // FIXME: We don't warn on variables with namespace alias qualifiers - // because support still needs to be added in the frontend. - if (Identifier.equals(StartTok.getString())) - return ErrorOperand(Start, "Unable to lookup expr!"); - } - unsigned CVal; + if (ParseIntelIdentifier(Val, Identifier, Info, + /*Unevaluated=*/true, End)) + return 0; + + if (!Info.OpDecl) + return ErrorOperand(Start, "unable to lookup expression"); + + unsigned CVal = 0; switch(OpKind) { default: llvm_unreachable("Unexpected operand kind!"); - case IOK_LENGTH: CVal = Length; break; - case IOK_SIZE: CVal = Size; break; - case IOK_TYPE: CVal = Type; break; + case IOK_LENGTH: CVal = Info.Length; break; + case IOK_SIZE: CVal = Info.Size; break; + case IOK_TYPE: CVal = Info.Type; break; } // Rewrite the type operator and the C or C++ type or variable in terms of an @@ -1583,11 +1680,11 @@ X86Operand *X86AsmParser::ParseIntelOperator(unsigned OpKind) { X86Operand *X86AsmParser::ParseIntelOperand() { const AsmToken &Tok = Parser.getTok(); - SMLoc Start = Tok.getLoc(), End; - StringRef AsmTokStr = Tok.getString(); + SMLoc Start, End; // Offset, length, type and size operators. if (isParsingInlineAsm()) { + StringRef AsmTokStr = Tok.getString(); if (AsmTokStr == "offset" || AsmTokStr == "OFFSET") return ParseIntelOffsetOfOperator(); if (AsmTokStr == "length" || AsmTokStr == "LENGTH") @@ -1598,14 +1695,23 @@ X86Operand *X86AsmParser::ParseIntelOperand() { return ParseIntelOperator(IOK_TYPE); } + unsigned Size = getIntelMemOperandSize(Tok.getString()); + if (Size) { + Parser.Lex(); // Eat operand size (e.g., byte, word). + if (Tok.getString() != "PTR" && Tok.getString() != "ptr") + return ErrorOperand(Start, "Expected 'PTR' or 'ptr' token!"); + Parser.Lex(); // Eat ptr. + } + Start = Tok.getLoc(); + // Immediate. if (getLexer().is(AsmToken::Integer) || getLexer().is(AsmToken::Minus) || getLexer().is(AsmToken::LParen)) { AsmToken StartTok = Tok; IntelExprStateMachine SM(/*Imm=*/0, /*StopOnLBrac=*/true, /*AddImmPrefix=*/false); - if (X86Operand *Err = ParseIntelExpression(SM, End)) - return Err; + if (ParseIntelExpression(SM, End)) + return 0; int64_t Imm = SM.getImm(); if (isParsingInlineAsm()) { @@ -1619,6 +1725,13 @@ X86Operand *X86AsmParser::ParseIntelOperand() { } if (getLexer().isNot(AsmToken::LBrac)) { + // If a directional label (ie. 1f or 2b) was parsed above from + // ParseIntelExpression() then SM.getSym() was set to a pointer to + // to the MCExpr with the directional local symbol and this is a + // memory operand not an immediate operand. + if (SM.getSym()) + return X86Operand::CreateMem(SM.getSym(), Start, End, Size); + const MCExpr *ImmExpr = MCConstantExpr::Create(Imm, getContext()); return X86Operand::CreateImm(ImmExpr, Start, End); } @@ -1629,23 +1742,22 @@ X86Operand *X86AsmParser::ParseIntelOperand() { "before bracketed expr."); // Parse ImmDisp [ BaseReg + Scale*IndexReg + Disp ]. - return ParseIntelMemOperand(/*SegReg=*/0, Imm, Start); + return ParseIntelMemOperand(Imm, Start, Size); } // Register. unsigned RegNo = 0; if (!ParseRegister(RegNo, Start, End)) { // If this is a segment register followed by a ':', then this is the start - // of a memory reference, otherwise this is a normal register reference. + // of a segment override, otherwise this is a normal register reference. if (getLexer().isNot(AsmToken::Colon)) return X86Operand::CreateReg(RegNo, Start, End); - getParser().Lex(); // Eat the colon. - return ParseIntelMemOperand(/*SegReg=*/RegNo, /*Disp=*/0, Start); + return ParseIntelSegmentOverride(/*SegReg=*/RegNo, Start, Size); } // Memory operand. - return ParseIntelMemOperand(/*SegReg=*/0, /*Disp=*/0, Start); + return ParseIntelMemOperand(/*Disp=*/0, Start, Size); } X86Operand *X86AsmParser::ParseATTOperand() { @@ -1741,10 +1853,11 @@ X86Operand *X86AsmParser::ParseMemOperand(unsigned SegReg, SMLoc MemStart) { // If we reached here, then we just ate the ( of the memory operand. Process // the rest of the memory operand. unsigned BaseReg = 0, IndexReg = 0, Scale = 1; - SMLoc IndexLoc; + SMLoc IndexLoc, BaseLoc; if (getLexer().is(AsmToken::Percent)) { SMLoc StartLoc, EndLoc; + BaseLoc = Parser.getTok().getLoc(); if (ParseRegister(BaseReg, StartLoc, EndLoc)) return 0; if (BaseReg == X86::EIZ || BaseReg == X86::RIZ) { Error(StartLoc, "eiz and riz can only be used as index registers", @@ -1787,6 +1900,11 @@ X86Operand *X86AsmParser::ParseMemOperand(unsigned SegReg, SMLoc MemStart) { } // Validate the scale amount. + if (X86MCRegisterClasses[X86::GR16RegClassID].contains(BaseReg) && + ScaleVal != 1) { + Error(Loc, "scale factor in 16-bit address must be 1"); + return 0; + } if (ScaleVal != 1 && ScaleVal != 2 && ScaleVal != 4 && ScaleVal != 8){ Error(Loc, "scale factor in address must be 1, 2, 4 or 8"); return 0; @@ -1817,6 +1935,21 @@ X86Operand *X86AsmParser::ParseMemOperand(unsigned SegReg, SMLoc MemStart) { SMLoc MemEnd = Parser.getTok().getEndLoc(); Parser.Lex(); // Eat the ')'. + // Check for use of invalid 16-bit registers. Only BX/BP/SI/DI are allowed, + // and then only in non-64-bit modes. Except for DX, which is a special case + // because an unofficial form of in/out instructions uses it. + if (X86MCRegisterClasses[X86::GR16RegClassID].contains(BaseReg) && + (is64BitMode() || (BaseReg != X86::BX && BaseReg != X86::BP && + BaseReg != X86::SI && BaseReg != X86::DI)) && + BaseReg != X86::DX) { + Error(BaseLoc, "invalid 16-bit base register"); + return 0; + } + if (BaseReg == 0 && + X86MCRegisterClasses[X86::GR16RegClassID].contains(IndexReg)) { + Error(IndexLoc, "16-bit memory operand may not include only index register"); + return 0; + } // If we have both a base register and an index register make sure they are // both 64-bit or 32-bit registers. // To support VSIB, IndexReg can be 128-bit or 256-bit registers. @@ -1825,16 +1958,30 @@ X86Operand *X86AsmParser::ParseMemOperand(unsigned SegReg, SMLoc MemStart) { (X86MCRegisterClasses[X86::GR16RegClassID].contains(IndexReg) || X86MCRegisterClasses[X86::GR32RegClassID].contains(IndexReg)) && IndexReg != X86::RIZ) { - Error(IndexLoc, "index register is 32-bit, but base register is 64-bit"); + Error(BaseLoc, "base register is 64-bit, but index register is not"); return 0; } if (X86MCRegisterClasses[X86::GR32RegClassID].contains(BaseReg) && (X86MCRegisterClasses[X86::GR16RegClassID].contains(IndexReg) || X86MCRegisterClasses[X86::GR64RegClassID].contains(IndexReg)) && IndexReg != X86::EIZ){ - Error(IndexLoc, "index register is 64-bit, but base register is 32-bit"); + Error(BaseLoc, "base register is 32-bit, but index register is not"); return 0; } + if (X86MCRegisterClasses[X86::GR16RegClassID].contains(BaseReg)) { + if (X86MCRegisterClasses[X86::GR32RegClassID].contains(IndexReg) || + X86MCRegisterClasses[X86::GR64RegClassID].contains(IndexReg)) { + Error(BaseLoc, "base register is 16-bit, but index register is not"); + return 0; + } + if (((BaseReg == X86::BX || BaseReg == X86::BP) && + IndexReg != X86::SI && IndexReg != X86::DI) || + ((BaseReg == X86::SI || BaseReg == X86::DI) && + IndexReg != X86::BX && IndexReg != X86::BP)) { + Error(BaseLoc, "invalid 16-bit base/index register combination"); + return 0; + } + } } return X86Operand::CreateMem(SegReg, Disp, BaseReg, IndexReg, Scale, @@ -1931,11 +2078,8 @@ ParseInstruction(ParseInstructionInfo &Info, StringRef Name, SMLoc NameLoc, if (getLexer().isNot(AsmToken::EndOfStatement) && !isPrefix) { // Parse '*' modifier. - if (getLexer().is(AsmToken::Star)) { - SMLoc Loc = Parser.getTok().getLoc(); - Operands.push_back(X86Operand::CreateToken("*", Loc)); - Parser.Lex(); // Eat the star. - } + if (getLexer().is(AsmToken::Star)) + Operands.push_back(X86Operand::CreateToken("*", consumeToken())); // Read the first operand. if (X86Operand *Op = ParseOperand()) @@ -1957,6 +2101,42 @@ ParseInstruction(ParseInstructionInfo &Info, StringRef Name, SMLoc NameLoc, } } + if (STI.getFeatureBits() & X86::FeatureAVX512) { + // Parse mask register {%k1} + if (getLexer().is(AsmToken::LCurly)) { + Operands.push_back(X86Operand::CreateToken("{", consumeToken())); + if (X86Operand *Op = ParseOperand()) { + Operands.push_back(Op); + if (!getLexer().is(AsmToken::RCurly)) { + SMLoc Loc = getLexer().getLoc(); + Parser.eatToEndOfStatement(); + return Error(Loc, "Expected } at this point"); + } + Operands.push_back(X86Operand::CreateToken("}", consumeToken())); + } else { + Parser.eatToEndOfStatement(); + return true; + } + } + // TODO: add parsing of broadcasts {1to8}, {1to16} + // Parse "zeroing non-masked" semantic {z} + if (getLexer().is(AsmToken::LCurly)) { + Operands.push_back(X86Operand::CreateToken("{z}", consumeToken())); + if (!getLexer().is(AsmToken::Identifier) || getLexer().getTok().getIdentifier() != "z") { + SMLoc Loc = getLexer().getLoc(); + Parser.eatToEndOfStatement(); + return Error(Loc, "Expected z at this point"); + } + Parser.Lex(); // Eat the z + if (!getLexer().is(AsmToken::RCurly)) { + SMLoc Loc = getLexer().getLoc(); + Parser.eatToEndOfStatement(); + return Error(Loc, "Expected } at this point"); + } + Parser.Lex(); // Eat the } + } + } + if (getLexer().isNot(AsmToken::EndOfStatement)) { SMLoc Loc = getLexer().getLoc(); Parser.eatToEndOfStatement(); @@ -2208,6 +2388,55 @@ processInstruction(MCInst &Inst, case X86::SBB16i16: return convert16i16to16ri8(Inst, X86::SBB16ri8); case X86::SBB32i32: return convert32i32to32ri8(Inst, X86::SBB32ri8); case X86::SBB64i32: return convert64i32to64ri8(Inst, X86::SBB64ri8); + case X86::VMOVAPDrr: + case X86::VMOVAPDYrr: + case X86::VMOVAPSrr: + case X86::VMOVAPSYrr: + case X86::VMOVDQArr: + case X86::VMOVDQAYrr: + case X86::VMOVDQUrr: + case X86::VMOVDQUYrr: + case X86::VMOVUPDrr: + case X86::VMOVUPDYrr: + case X86::VMOVUPSrr: + case X86::VMOVUPSYrr: { + if (X86II::isX86_64ExtendedReg(Inst.getOperand(0).getReg()) || + !X86II::isX86_64ExtendedReg(Inst.getOperand(1).getReg())) + return false; + + unsigned NewOpc; + switch (Inst.getOpcode()) { + default: llvm_unreachable("Invalid opcode"); + case X86::VMOVAPDrr: NewOpc = X86::VMOVAPDrr_REV; break; + case X86::VMOVAPDYrr: NewOpc = X86::VMOVAPDYrr_REV; break; + case X86::VMOVAPSrr: NewOpc = X86::VMOVAPSrr_REV; break; + case X86::VMOVAPSYrr: NewOpc = X86::VMOVAPSYrr_REV; break; + case X86::VMOVDQArr: NewOpc = X86::VMOVDQArr_REV; break; + case X86::VMOVDQAYrr: NewOpc = X86::VMOVDQAYrr_REV; break; + case X86::VMOVDQUrr: NewOpc = X86::VMOVDQUrr_REV; break; + case X86::VMOVDQUYrr: NewOpc = X86::VMOVDQUYrr_REV; break; + case X86::VMOVUPDrr: NewOpc = X86::VMOVUPDrr_REV; break; + case X86::VMOVUPDYrr: NewOpc = X86::VMOVUPDYrr_REV; break; + case X86::VMOVUPSrr: NewOpc = X86::VMOVUPSrr_REV; break; + case X86::VMOVUPSYrr: NewOpc = X86::VMOVUPSYrr_REV; break; + } + Inst.setOpcode(NewOpc); + return true; + } + case X86::VMOVSDrr: + case X86::VMOVSSrr: { + if (X86II::isX86_64ExtendedReg(Inst.getOperand(0).getReg()) || + !X86II::isX86_64ExtendedReg(Inst.getOperand(2).getReg())) + return false; + unsigned NewOpc; + switch (Inst.getOpcode()) { + default: llvm_unreachable("Invalid opcode"); + case X86::VMOVSDrr: NewOpc = X86::VMOVSDrr_REV; break; + case X86::VMOVSSrr: NewOpc = X86::VMOVSSrr_REV; break; + } + Inst.setOpcode(NewOpc); + return true; + } } } @@ -2220,7 +2449,7 @@ MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode, assert(!Operands.empty() && "Unexpect empty operand list!"); X86Operand *Op = static_cast(Operands[0]); assert(Op->isToken() && "Leading operand should always be a mnemonic!"); - ArrayRef EmptyRanges = ArrayRef(); + ArrayRef EmptyRanges = None; // First, handle aliases that expand to multiple instructions. // FIXME: This should be replaced with a real .td file alias mechanism. @@ -2322,25 +2551,25 @@ MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode, unsigned Match1, Match2, Match3, Match4; Match1 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore, - isParsingIntelSyntax()); + MatchingInlineAsm, isParsingIntelSyntax()); // If this returned as a missing feature failure, remember that. if (Match1 == Match_MissingFeature) ErrorInfoMissingFeature = ErrorInfoIgnore; Tmp[Base.size()] = Suffixes[1]; Match2 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore, - isParsingIntelSyntax()); + MatchingInlineAsm, isParsingIntelSyntax()); // If this returned as a missing feature failure, remember that. if (Match2 == Match_MissingFeature) ErrorInfoMissingFeature = ErrorInfoIgnore; Tmp[Base.size()] = Suffixes[2]; Match3 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore, - isParsingIntelSyntax()); + MatchingInlineAsm, isParsingIntelSyntax()); // If this returned as a missing feature failure, remember that. if (Match3 == Match_MissingFeature) ErrorInfoMissingFeature = ErrorInfoIgnore; Tmp[Base.size()] = Suffixes[3]; Match4 = MatchInstructionImpl(Operands, Inst, ErrorInfoIgnore, - isParsingIntelSyntax()); + MatchingInlineAsm, isParsingIntelSyntax()); // If this returned as a missing feature failure, remember that. if (Match4 == Match_MissingFeature) ErrorInfoMissingFeature = ErrorInfoIgnore; @@ -2464,11 +2693,9 @@ bool X86AsmParser::ParseDirective(AsmToken DirectiveID) { } else if (IDVal.startswith(".intel_syntax")) { getParser().setAssemblerDialect(1); if (getLexer().isNot(AsmToken::EndOfStatement)) { - if(Parser.getTok().getString() == "noprefix") { - // FIXME : Handle noprefix + // FIXME: Handle noprefix + if (Parser.getTok().getString() == "noprefix") Parser.Lex(); - } else - return true; } return false; } @@ -2482,7 +2709,7 @@ bool X86AsmParser::ParseDirectiveWord(unsigned Size, SMLoc L) { for (;;) { const MCExpr *Value; if (getParser().parseExpression(Value)) - return true; + return false; getParser().getStreamer().EmitValue(Value, Size); @@ -2490,8 +2717,10 @@ bool X86AsmParser::ParseDirectiveWord(unsigned Size, SMLoc L) { break; // FIXME: Improve diagnostic. - if (getLexer().isNot(AsmToken::Comma)) - return Error(L, "unexpected token in directive"); + if (getLexer().isNot(AsmToken::Comma)) { + Error(L, "unexpected token in directive"); + return false; + } Parser.Lex(); } } @@ -2501,22 +2730,29 @@ bool X86AsmParser::ParseDirectiveWord(unsigned Size, SMLoc L) { } /// ParseDirectiveCode -/// ::= .code32 | .code64 +/// ::= .code16 | .code32 | .code64 bool X86AsmParser::ParseDirectiveCode(StringRef IDVal, SMLoc L) { - if (IDVal == ".code32") { + if (IDVal == ".code16") { Parser.Lex(); - if (is64BitMode()) { - SwitchMode(); + if (!is16BitMode()) { + SwitchMode(X86::Mode16Bit); + getParser().getStreamer().EmitAssemblerFlag(MCAF_Code16); + } + } else if (IDVal == ".code32") { + Parser.Lex(); + if (!is32BitMode()) { + SwitchMode(X86::Mode32Bit); getParser().getStreamer().EmitAssemblerFlag(MCAF_Code32); } } else if (IDVal == ".code64") { Parser.Lex(); if (!is64BitMode()) { - SwitchMode(); + SwitchMode(X86::Mode64Bit); getParser().getStreamer().EmitAssemblerFlag(MCAF_Code64); } } else { - return Error(L, "unexpected directive " + IDVal); + Error(L, "unknown directive " + IDVal); + return false; } return false;