1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This class implements the parser for assembly files.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/ADT/SmallString.h"
15 #include "llvm/ADT/StringMap.h"
16 #include "llvm/ADT/StringSwitch.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/MC/MCAsmInfo.h"
19 #include "llvm/MC/MCContext.h"
20 #include "llvm/MC/MCExpr.h"
21 #include "llvm/MC/MCInst.h"
22 #include "llvm/MC/MCParser/AsmCond.h"
23 #include "llvm/MC/MCParser/AsmLexer.h"
24 #include "llvm/MC/MCParser/MCAsmParser.h"
25 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
26 #include "llvm/MC/MCSectionMachO.h"
27 #include "llvm/MC/MCStreamer.h"
28 #include "llvm/MC/MCSymbol.h"
29 #include "llvm/MC/MCDwarf.h"
30 #include "llvm/Support/Compiler.h"
31 #include "llvm/Support/MemoryBuffer.h"
32 #include "llvm/Support/SourceMgr.h"
33 #include "llvm/Support/raw_ostream.h"
34 #include "llvm/Target/TargetAsmParser.h"
40 /// \brief Helper class for tracking macro definitions.
46 Macro(StringRef N, StringRef B) : Name(N), Body(B) {}
49 /// \brief Helper class for storing information about an active macro
51 struct MacroInstantiation {
52 /// The macro being instantiated.
53 const Macro *TheMacro;
55 /// The macro instantiation with substitutions.
56 MemoryBuffer *Instantiation;
58 /// The location of the instantiation.
59 SMLoc InstantiationLoc;
61 /// The location where parsing should resume upon instantiation completion.
65 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
66 const std::vector<std::vector<AsmToken> > &A);
69 /// \brief The concrete assembly parser instance.
70 class AsmParser : public MCAsmParser {
71 friend class GenericAsmParser;
73 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
74 void operator=(const AsmParser &); // DO NOT IMPLEMENT
80 MCAsmParserExtension *GenericParser;
81 MCAsmParserExtension *PlatformParser;
83 /// This is the current buffer index we're lexing from as managed by the
88 std::vector<AsmCond> TheCondStack;
90 /// DirectiveMap - This is a table handlers for directives. Each handler is
91 /// invoked after the directive identifier is read and is responsible for
92 /// parsing and validating the rest of the directive. The handler is passed
93 /// in the directive name and the location of the directive keyword.
94 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
96 /// MacroMap - Map of currently defined macros.
97 StringMap<Macro*> MacroMap;
99 /// ActiveMacros - Stack of active macro instantiations.
100 std::vector<MacroInstantiation*> ActiveMacros;
102 /// Boolean tracking whether macro substitution is enabled.
103 unsigned MacrosEnabled : 1;
105 /// Flag tracking whether any errors have been encountered.
106 unsigned HadError : 1;
109 AsmParser(const Target &T, SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
110 const MCAsmInfo &MAI);
113 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
115 void AddDirectiveHandler(MCAsmParserExtension *Object,
117 DirectiveHandler Handler) {
118 DirectiveMap[Directive] = std::make_pair(Object, Handler);
122 /// @name MCAsmParser Interface
125 virtual SourceMgr &getSourceManager() { return SrcMgr; }
126 virtual MCAsmLexer &getLexer() { return Lexer; }
127 virtual MCContext &getContext() { return Ctx; }
128 virtual MCStreamer &getStreamer() { return Out; }
130 virtual void Warning(SMLoc L, const Twine &Meg);
131 virtual bool Error(SMLoc L, const Twine &Msg);
133 const AsmToken &Lex();
135 bool ParseExpression(const MCExpr *&Res);
136 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
137 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
138 virtual bool ParseAbsoluteExpression(int64_t &Res);
143 void CheckForValidSection();
145 bool ParseStatement();
147 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
148 void HandleMacroExit();
150 void PrintMacroInstantiations();
151 void PrintMessage(SMLoc Loc, const std::string &Msg, const char *Type) const;
153 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
154 bool EnterIncludeFile(const std::string &Filename);
156 /// \brief Reset the current lexer position to that given by \arg Loc. The
157 /// current token is not set; clients should ensure Lex() is called
159 void JumpToLoc(SMLoc Loc);
161 void EatToEndOfStatement();
163 /// \brief Parse up to the end of statement and a return the contents from the
164 /// current token until the end of the statement; the current token on exit
165 /// will be either the EndOfStatement or EOF.
166 StringRef ParseStringToEndOfStatement();
168 bool ParseAssignment(StringRef Name);
170 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
171 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
172 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
174 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
175 /// and set \arg Res to the identifier contents.
176 bool ParseIdentifier(StringRef &Res);
178 // Directive Parsing.
179 bool ParseDirectiveAscii(bool ZeroTerminated); // ".ascii", ".asciiz"
180 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
181 bool ParseDirectiveFill(); // ".fill"
182 bool ParseDirectiveSpace(); // ".space"
183 bool ParseDirectiveZero(); // ".zero"
184 bool ParseDirectiveSet(); // ".set"
185 bool ParseDirectiveOrg(); // ".org"
186 // ".align{,32}", ".p2align{,w,l}"
187 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
189 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
190 /// accepts a single symbol (which should be a label or an external).
191 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
192 bool ParseDirectiveELFType(); // ELF specific ".type"
194 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
196 bool ParseDirectiveAbort(); // ".abort"
197 bool ParseDirectiveInclude(); // ".include"
199 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
200 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
201 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
202 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
204 /// ParseEscapedString - Parse the current token as a string which may include
205 /// escaped characters and return the string contents.
206 bool ParseEscapedString(std::string &Data);
208 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
209 MCSymbolRefExpr::VariantKind Variant);
212 /// \brief Generic implementations of directive handling, etc. which is shared
213 /// (or the default, at least) for all assembler parser.
214 class GenericAsmParser : public MCAsmParserExtension {
215 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
216 void AddDirectiveHandler(StringRef Directive) {
217 getParser().AddDirectiveHandler(this, Directive,
218 HandleDirective<GenericAsmParser, Handler>);
222 GenericAsmParser() {}
224 AsmParser &getParser() {
225 return (AsmParser&) this->MCAsmParserExtension::getParser();
228 virtual void Initialize(MCAsmParser &Parser) {
229 // Call the base implementation.
230 this->MCAsmParserExtension::Initialize(Parser);
232 // Debugging directives.
233 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
234 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
235 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
238 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
240 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
242 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
243 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
244 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
246 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
247 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
250 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
251 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
252 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
254 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
255 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
256 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
258 void ParseUleb128(uint64_t Value);
259 void ParseSleb128(int64_t Value);
260 bool ParseDirectiveLEB128(StringRef, SMLoc);
267 extern MCAsmParserExtension *createDarwinAsmParser();
268 extern MCAsmParserExtension *createELFAsmParser();
272 enum { DEFAULT_ADDRSPACE = 0 };
274 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
275 MCStreamer &_Out, const MCAsmInfo &_MAI)
276 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
277 GenericParser(new GenericAsmParser), PlatformParser(0),
278 CurBuffer(0), MacrosEnabled(true) {
279 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
281 // Initialize the generic parser.
282 GenericParser->Initialize(*this);
284 // Initialize the platform / file format parser.
286 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
288 if (_MAI.hasSubsectionsViaSymbols()) {
289 PlatformParser = createDarwinAsmParser();
290 PlatformParser->Initialize(*this);
292 PlatformParser = createELFAsmParser();
293 PlatformParser->Initialize(*this);
297 AsmParser::~AsmParser() {
298 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
300 // Destroy any macros.
301 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
302 ie = MacroMap.end(); it != ie; ++it)
303 delete it->getValue();
305 delete PlatformParser;
306 delete GenericParser;
309 void AsmParser::PrintMacroInstantiations() {
310 // Print the active macro instantiation stack.
311 for (std::vector<MacroInstantiation*>::const_reverse_iterator
312 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
313 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
317 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
318 PrintMessage(L, Msg.str(), "warning");
319 PrintMacroInstantiations();
322 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
324 PrintMessage(L, Msg.str(), "error");
325 PrintMacroInstantiations();
329 void AsmParser::PrintMessage(SMLoc Loc, const std::string &Msg,
330 const char *Type) const {
331 SrcMgr.PrintMessage(Loc, Msg, Type);
334 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
335 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
341 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
346 void AsmParser::JumpToLoc(SMLoc Loc) {
347 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
348 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
351 const AsmToken &AsmParser::Lex() {
352 const AsmToken *tok = &Lexer.Lex();
354 if (tok->is(AsmToken::Eof)) {
355 // If this is the end of an included file, pop the parent file off the
357 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
358 if (ParentIncludeLoc != SMLoc()) {
359 JumpToLoc(ParentIncludeLoc);
364 if (tok->is(AsmToken::Error))
365 Error(Lexer.getErrLoc(), Lexer.getErr());
370 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
371 // Create the initial section, if requested.
372 if (!NoInitialTextSection)
379 AsmCond StartingCondState = TheCondState;
381 // While we have input, parse each statement.
382 while (Lexer.isNot(AsmToken::Eof)) {
383 if (!ParseStatement()) continue;
385 // We had an error, validate that one was emitted and recover by skipping to
387 assert(HadError && "Parse statement returned an error, but none emitted!");
388 EatToEndOfStatement();
391 if (TheCondState.TheCond != StartingCondState.TheCond ||
392 TheCondState.Ignore != StartingCondState.Ignore)
393 return TokError("unmatched .ifs or .elses");
395 // Check to see there are no empty DwarfFile slots.
396 const std::vector<MCDwarfFile *> &MCDwarfFiles =
397 getContext().getMCDwarfFiles();
398 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
399 if (!MCDwarfFiles[i])
400 TokError("unassigned file number: " + Twine(i) + " for .file directives");
403 // Finalize the output stream if there are no errors and if the client wants
405 if (!HadError && !NoFinalize)
411 void AsmParser::CheckForValidSection() {
412 if (!getStreamer().getCurrentSection()) {
413 TokError("expected section directive before assembly directive");
414 Out.SwitchSection(Ctx.getMachOSection(
416 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
417 0, SectionKind::getText()));
421 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
422 void AsmParser::EatToEndOfStatement() {
423 while (Lexer.isNot(AsmToken::EndOfStatement) &&
424 Lexer.isNot(AsmToken::Eof))
428 if (Lexer.is(AsmToken::EndOfStatement))
432 StringRef AsmParser::ParseStringToEndOfStatement() {
433 const char *Start = getTok().getLoc().getPointer();
435 while (Lexer.isNot(AsmToken::EndOfStatement) &&
436 Lexer.isNot(AsmToken::Eof))
439 const char *End = getTok().getLoc().getPointer();
440 return StringRef(Start, End - Start);
443 /// ParseParenExpr - Parse a paren expression and return it.
444 /// NOTE: This assumes the leading '(' has already been consumed.
446 /// parenexpr ::= expr)
448 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
449 if (ParseExpression(Res)) return true;
450 if (Lexer.isNot(AsmToken::RParen))
451 return TokError("expected ')' in parentheses expression");
452 EndLoc = Lexer.getLoc();
457 /// ParsePrimaryExpr - Parse a primary expression and return it.
458 /// primaryexpr ::= (parenexpr
459 /// primaryexpr ::= symbol
460 /// primaryexpr ::= number
461 /// primaryexpr ::= '.'
462 /// primaryexpr ::= ~,+,- primaryexpr
463 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
464 switch (Lexer.getKind()) {
466 return TokError("unknown token in expression");
467 case AsmToken::Exclaim:
468 Lex(); // Eat the operator.
469 if (ParsePrimaryExpr(Res, EndLoc))
471 Res = MCUnaryExpr::CreateLNot(Res, getContext());
473 case AsmToken::Dollar:
474 case AsmToken::String:
475 case AsmToken::Identifier: {
476 EndLoc = Lexer.getLoc();
478 StringRef Identifier;
479 if (ParseIdentifier(Identifier))
482 // This is a symbol reference.
483 std::pair<StringRef, StringRef> Split = Identifier.split('@');
484 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
486 // Mark the symbol as used in an expression.
487 Sym->setUsedInExpr(true);
489 // Lookup the symbol variant if used.
490 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
491 if (Split.first.size() != Identifier.size()) {
492 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
493 if (Variant == MCSymbolRefExpr::VK_Invalid) {
494 Variant = MCSymbolRefExpr::VK_None;
495 TokError("invalid variant '" + Split.second + "'");
499 // If this is an absolute variable reference, substitute it now to preserve
500 // semantics in the face of reassignment.
501 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
503 return Error(EndLoc, "unexpected modified on variable reference");
505 Res = Sym->getVariableValue();
509 // Otherwise create a symbol ref.
510 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
513 case AsmToken::Integer: {
514 SMLoc Loc = getTok().getLoc();
515 int64_t IntVal = getTok().getIntVal();
516 Res = MCConstantExpr::Create(IntVal, getContext());
517 EndLoc = Lexer.getLoc();
519 // Look for 'b' or 'f' following an Integer as a directional label
520 if (Lexer.getKind() == AsmToken::Identifier) {
521 StringRef IDVal = getTok().getString();
522 if (IDVal == "f" || IDVal == "b"){
523 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
524 IDVal == "f" ? 1 : 0);
525 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
527 if(IDVal == "b" && Sym->isUndefined())
528 return Error(Loc, "invalid reference to undefined symbol");
529 EndLoc = Lexer.getLoc();
530 Lex(); // Eat identifier.
535 case AsmToken::Dot: {
536 // This is a '.' reference, which references the current PC. Emit a
537 // temporary label to the streamer and refer to it.
538 MCSymbol *Sym = Ctx.CreateTempSymbol();
540 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
541 EndLoc = Lexer.getLoc();
542 Lex(); // Eat identifier.
546 case AsmToken::LParen:
547 Lex(); // Eat the '('.
548 return ParseParenExpr(Res, EndLoc);
549 case AsmToken::Minus:
550 Lex(); // Eat the operator.
551 if (ParsePrimaryExpr(Res, EndLoc))
553 Res = MCUnaryExpr::CreateMinus(Res, getContext());
556 Lex(); // Eat the operator.
557 if (ParsePrimaryExpr(Res, EndLoc))
559 Res = MCUnaryExpr::CreatePlus(Res, getContext());
561 case AsmToken::Tilde:
562 Lex(); // Eat the operator.
563 if (ParsePrimaryExpr(Res, EndLoc))
565 Res = MCUnaryExpr::CreateNot(Res, getContext());
570 bool AsmParser::ParseExpression(const MCExpr *&Res) {
572 return ParseExpression(Res, EndLoc);
576 AsmParser::ApplyModifierToExpr(const MCExpr *E,
577 MCSymbolRefExpr::VariantKind Variant) {
578 // Recurse over the given expression, rebuilding it to apply the given variant
579 // if there is exactly one symbol.
580 switch (E->getKind()) {
582 case MCExpr::Constant:
585 case MCExpr::SymbolRef: {
586 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
588 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
589 TokError("invalid variant on expression '" +
590 getTok().getIdentifier() + "' (already modified)");
594 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
597 case MCExpr::Unary: {
598 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
599 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
602 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
605 case MCExpr::Binary: {
606 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
607 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
608 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
613 if (!LHS) LHS = BE->getLHS();
614 if (!RHS) RHS = BE->getRHS();
616 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
621 /// ParseExpression - Parse an expression and return it.
623 /// expr ::= expr +,- expr -> lowest.
624 /// expr ::= expr |,^,&,! expr -> middle.
625 /// expr ::= expr *,/,%,<<,>> expr -> highest.
626 /// expr ::= primaryexpr
628 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
629 // Parse the expression.
631 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
634 // As a special case, we support 'a op b @ modifier' by rewriting the
635 // expression to include the modifier. This is inefficient, but in general we
636 // expect users to use 'a@modifier op b'.
637 if (Lexer.getKind() == AsmToken::At) {
640 if (Lexer.isNot(AsmToken::Identifier))
641 return TokError("unexpected symbol modifier following '@'");
643 MCSymbolRefExpr::VariantKind Variant =
644 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
645 if (Variant == MCSymbolRefExpr::VK_Invalid)
646 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
648 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
650 return TokError("invalid modifier '" + getTok().getIdentifier() +
651 "' (no symbols present)");
659 // Try to constant fold it up front, if possible.
661 if (Res->EvaluateAsAbsolute(Value))
662 Res = MCConstantExpr::Create(Value, getContext());
667 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
669 return ParseParenExpr(Res, EndLoc) ||
670 ParseBinOpRHS(1, Res, EndLoc);
673 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
676 SMLoc StartLoc = Lexer.getLoc();
677 if (ParseExpression(Expr))
680 if (!Expr->EvaluateAsAbsolute(Res))
681 return Error(StartLoc, "expected absolute expression");
686 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
687 MCBinaryExpr::Opcode &Kind) {
690 return 0; // not a binop.
692 // Lowest Precedence: &&, ||, @
693 case AsmToken::AmpAmp:
694 Kind = MCBinaryExpr::LAnd;
696 case AsmToken::PipePipe:
697 Kind = MCBinaryExpr::LOr;
700 // Low Precedence: +, -, ==, !=, <>, <, <=, >, >=
702 Kind = MCBinaryExpr::Add;
704 case AsmToken::Minus:
705 Kind = MCBinaryExpr::Sub;
707 case AsmToken::EqualEqual:
708 Kind = MCBinaryExpr::EQ;
710 case AsmToken::ExclaimEqual:
711 case AsmToken::LessGreater:
712 Kind = MCBinaryExpr::NE;
715 Kind = MCBinaryExpr::LT;
717 case AsmToken::LessEqual:
718 Kind = MCBinaryExpr::LTE;
720 case AsmToken::Greater:
721 Kind = MCBinaryExpr::GT;
723 case AsmToken::GreaterEqual:
724 Kind = MCBinaryExpr::GTE;
727 // Intermediate Precedence: |, &, ^
729 // FIXME: gas seems to support '!' as an infix operator?
731 Kind = MCBinaryExpr::Or;
733 case AsmToken::Caret:
734 Kind = MCBinaryExpr::Xor;
737 Kind = MCBinaryExpr::And;
740 // Highest Precedence: *, /, %, <<, >>
742 Kind = MCBinaryExpr::Mul;
744 case AsmToken::Slash:
745 Kind = MCBinaryExpr::Div;
747 case AsmToken::Percent:
748 Kind = MCBinaryExpr::Mod;
750 case AsmToken::LessLess:
751 Kind = MCBinaryExpr::Shl;
753 case AsmToken::GreaterGreater:
754 Kind = MCBinaryExpr::Shr;
760 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
761 /// Res contains the LHS of the expression on input.
762 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
765 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
766 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
768 // If the next token is lower precedence than we are allowed to eat, return
769 // successfully with what we ate already.
770 if (TokPrec < Precedence)
775 // Eat the next primary expression.
777 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
779 // If BinOp binds less tightly with RHS than the operator after RHS, let
780 // the pending operator take RHS as its LHS.
781 MCBinaryExpr::Opcode Dummy;
782 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
783 if (TokPrec < NextTokPrec) {
784 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
787 // Merge LHS and RHS according to operator.
788 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
796 /// ::= EndOfStatement
797 /// ::= Label* Directive ...Operands... EndOfStatement
798 /// ::= Label* Identifier OperandList* EndOfStatement
799 bool AsmParser::ParseStatement() {
800 if (Lexer.is(AsmToken::EndOfStatement)) {
806 // Statements always start with an identifier.
807 AsmToken ID = getTok();
808 SMLoc IDLoc = ID.getLoc();
810 int64_t LocalLabelVal = -1;
811 // GUESS allow an integer followed by a ':' as a directional local label
812 if (Lexer.is(AsmToken::Integer)) {
813 LocalLabelVal = getTok().getIntVal();
814 if (LocalLabelVal < 0) {
815 if (!TheCondState.Ignore)
816 return TokError("unexpected token at start of statement");
820 IDVal = getTok().getString();
821 Lex(); // Consume the integer token to be used as an identifier token.
822 if (Lexer.getKind() != AsmToken::Colon) {
823 if (!TheCondState.Ignore)
824 return TokError("unexpected token at start of statement");
828 else if (ParseIdentifier(IDVal)) {
829 if (!TheCondState.Ignore)
830 return TokError("unexpected token at start of statement");
834 // Handle conditional assembly here before checking for skipping. We
835 // have to do this so that .endif isn't skipped in a ".if 0" block for
838 return ParseDirectiveIf(IDLoc);
839 if (IDVal == ".elseif")
840 return ParseDirectiveElseIf(IDLoc);
841 if (IDVal == ".else")
842 return ParseDirectiveElse(IDLoc);
843 if (IDVal == ".endif")
844 return ParseDirectiveEndIf(IDLoc);
846 // If we are in a ".if 0" block, ignore this statement.
847 if (TheCondState.Ignore) {
848 EatToEndOfStatement();
852 // FIXME: Recurse on local labels?
854 // See what kind of statement we have.
855 switch (Lexer.getKind()) {
856 case AsmToken::Colon: {
857 CheckForValidSection();
859 // identifier ':' -> Label.
862 // Diagnose attempt to use a variable as a label.
864 // FIXME: Diagnostics. Note the location of the definition as a label.
865 // FIXME: This doesn't diagnose assignment to a symbol which has been
866 // implicitly marked as external.
868 if (LocalLabelVal == -1)
869 Sym = getContext().GetOrCreateSymbol(IDVal);
871 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
872 if (!Sym->isUndefined() || Sym->isVariable())
873 return Error(IDLoc, "invalid symbol redefinition");
878 // Consume any end of statement token, if present, to avoid spurious
879 // AddBlankLine calls().
880 if (Lexer.is(AsmToken::EndOfStatement)) {
882 if (Lexer.is(AsmToken::Eof))
886 return ParseStatement();
889 case AsmToken::Equal:
890 // identifier '=' ... -> assignment statement
893 return ParseAssignment(IDVal);
895 default: // Normal instruction or directive.
899 // If macros are enabled, check to see if this is a macro instantiation.
901 if (const Macro *M = MacroMap.lookup(IDVal))
902 return HandleMacroEntry(IDVal, IDLoc, M);
904 // Otherwise, we have a normal instruction or directive.
905 if (IDVal[0] == '.') {
906 // Assembler features
908 return ParseDirectiveSet();
912 if (IDVal == ".ascii")
913 return ParseDirectiveAscii(false);
914 if (IDVal == ".asciz")
915 return ParseDirectiveAscii(true);
917 if (IDVal == ".byte")
918 return ParseDirectiveValue(1);
919 if (IDVal == ".short")
920 return ParseDirectiveValue(2);
921 if (IDVal == ".long")
922 return ParseDirectiveValue(4);
923 if (IDVal == ".quad")
924 return ParseDirectiveValue(8);
926 if (IDVal == ".align") {
927 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
928 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
930 if (IDVal == ".align32") {
931 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
932 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
934 if (IDVal == ".balign")
935 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
936 if (IDVal == ".balignw")
937 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
938 if (IDVal == ".balignl")
939 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
940 if (IDVal == ".p2align")
941 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
942 if (IDVal == ".p2alignw")
943 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
944 if (IDVal == ".p2alignl")
945 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
948 return ParseDirectiveOrg();
950 if (IDVal == ".fill")
951 return ParseDirectiveFill();
952 if (IDVal == ".space")
953 return ParseDirectiveSpace();
954 if (IDVal == ".zero")
955 return ParseDirectiveZero();
957 // Symbol attribute directives
959 if (IDVal == ".globl" || IDVal == ".global")
960 return ParseDirectiveSymbolAttribute(MCSA_Global);
961 if (IDVal == ".hidden")
962 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
963 if (IDVal == ".indirect_symbol")
964 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
965 if (IDVal == ".internal")
966 return ParseDirectiveSymbolAttribute(MCSA_Internal);
967 if (IDVal == ".lazy_reference")
968 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
969 if (IDVal == ".no_dead_strip")
970 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
971 if (IDVal == ".private_extern")
972 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
973 if (IDVal == ".protected")
974 return ParseDirectiveSymbolAttribute(MCSA_Protected);
975 if (IDVal == ".reference")
976 return ParseDirectiveSymbolAttribute(MCSA_Reference);
977 if (IDVal == ".type")
978 return ParseDirectiveELFType();
979 if (IDVal == ".weak")
980 return ParseDirectiveSymbolAttribute(MCSA_Weak);
981 if (IDVal == ".weak_definition")
982 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
983 if (IDVal == ".weak_reference")
984 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
985 if (IDVal == ".weak_def_can_be_hidden")
986 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
988 if (IDVal == ".comm")
989 return ParseDirectiveComm(/*IsLocal=*/false);
990 if (IDVal == ".lcomm")
991 return ParseDirectiveComm(/*IsLocal=*/true);
993 if (IDVal == ".abort")
994 return ParseDirectiveAbort();
995 if (IDVal == ".include")
996 return ParseDirectiveInclude();
998 // Look up the handler in the handler table.
999 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1000 DirectiveMap.lookup(IDVal);
1002 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1004 // Target hook for parsing target specific directives.
1005 if (!getTargetParser().ParseDirective(ID))
1008 Warning(IDLoc, "ignoring directive for now");
1009 EatToEndOfStatement();
1013 CheckForValidSection();
1015 // Canonicalize the opcode to lower case.
1016 SmallString<128> Opcode;
1017 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1018 Opcode.push_back(tolower(IDVal[i]));
1020 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1021 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1024 // Dump the parsed representation, if requested.
1025 if (getShowParsedOperands()) {
1026 SmallString<256> Str;
1027 raw_svector_ostream OS(Str);
1028 OS << "parsed instruction: [";
1029 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1032 ParsedOperands[i]->dump(OS);
1036 PrintMessage(IDLoc, OS.str(), "note");
1039 // If parsing succeeded, match the instruction.
1042 if (!getTargetParser().MatchInstruction(IDLoc, ParsedOperands, Inst)) {
1043 // Emit the instruction on success.
1044 Out.EmitInstruction(Inst);
1049 // Free any parsed operands.
1050 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1051 delete ParsedOperands[i];
1053 // Don't skip the rest of the line, the instruction parser is responsible for
1058 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1059 const std::vector<std::vector<AsmToken> > &A)
1060 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1062 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1063 // to hold the macro body with substitutions.
1064 SmallString<256> Buf;
1065 raw_svector_ostream OS(Buf);
1067 StringRef Body = M->Body;
1068 while (!Body.empty()) {
1069 // Scan for the next substitution.
1070 std::size_t End = Body.size(), Pos = 0;
1071 for (; Pos != End; ++Pos) {
1072 // Check for a substitution or escape.
1073 if (Body[Pos] != '$' || Pos + 1 == End)
1076 char Next = Body[Pos + 1];
1077 if (Next == '$' || Next == 'n' || isdigit(Next))
1082 OS << Body.slice(0, Pos);
1084 // Check if we reached the end.
1088 switch (Body[Pos+1]) {
1094 // $n => number of arguments
1099 // $[0-9] => argument
1101 // Missing arguments are ignored.
1102 unsigned Index = Body[Pos+1] - '0';
1103 if (Index >= A.size())
1106 // Otherwise substitute with the token values, with spaces eliminated.
1107 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1108 ie = A[Index].end(); it != ie; ++it)
1109 OS << it->getString();
1114 // Update the scan point.
1115 Body = Body.substr(Pos + 2);
1118 // We include the .endmacro in the buffer as our queue to exit the macro
1120 OS << ".endmacro\n";
1122 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1125 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1127 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1128 // this, although we should protect against infinite loops.
1129 if (ActiveMacros.size() == 20)
1130 return TokError("macros cannot be nested more than 20 levels deep");
1132 // Parse the macro instantiation arguments.
1133 std::vector<std::vector<AsmToken> > MacroArguments;
1134 MacroArguments.push_back(std::vector<AsmToken>());
1135 unsigned ParenLevel = 0;
1137 if (Lexer.is(AsmToken::Eof))
1138 return TokError("unexpected token in macro instantiation");
1139 if (Lexer.is(AsmToken::EndOfStatement))
1142 // If we aren't inside parentheses and this is a comma, start a new token
1144 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1145 MacroArguments.push_back(std::vector<AsmToken>());
1147 // Adjust the current parentheses level.
1148 if (Lexer.is(AsmToken::LParen))
1150 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1153 // Append the token to the current argument list.
1154 MacroArguments.back().push_back(getTok());
1159 // Create the macro instantiation object and add to the current macro
1160 // instantiation stack.
1161 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1164 ActiveMacros.push_back(MI);
1166 // Jump to the macro instantiation and prime the lexer.
1167 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1168 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1174 void AsmParser::HandleMacroExit() {
1175 // Jump to the EndOfStatement we should return to, and consume it.
1176 JumpToLoc(ActiveMacros.back()->ExitLoc);
1179 // Pop the instantiation entry.
1180 delete ActiveMacros.back();
1181 ActiveMacros.pop_back();
1184 bool AsmParser::ParseAssignment(StringRef Name) {
1185 // FIXME: Use better location, we should use proper tokens.
1186 SMLoc EqualLoc = Lexer.getLoc();
1188 const MCExpr *Value;
1189 if (ParseExpression(Value))
1192 if (Lexer.isNot(AsmToken::EndOfStatement))
1193 return TokError("unexpected token in assignment");
1195 // Eat the end of statement marker.
1198 // Validate that the LHS is allowed to be a variable (either it has not been
1199 // used as a symbol, or it is an absolute symbol).
1200 MCSymbol *Sym = getContext().LookupSymbol(Name);
1202 // Diagnose assignment to a label.
1204 // FIXME: Diagnostics. Note the location of the definition as a label.
1205 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1206 if (Sym->isUndefined() && !Sym->isUsedInExpr())
1207 ; // Allow redefinitions of undefined symbols only used in directives.
1208 else if (!Sym->isUndefined() && !Sym->isAbsolute())
1209 return Error(EqualLoc, "redefinition of '" + Name + "'");
1210 else if (!Sym->isVariable())
1211 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1212 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1213 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1216 Sym = getContext().GetOrCreateSymbol(Name);
1218 // FIXME: Handle '.'.
1220 Sym->setUsedInExpr(true);
1222 // Do the assignment.
1223 Out.EmitAssignment(Sym, Value);
1228 /// ParseIdentifier:
1231 bool AsmParser::ParseIdentifier(StringRef &Res) {
1232 // The assembler has relaxed rules for accepting identifiers, in particular we
1233 // allow things like '.globl $foo', which would normally be separate
1234 // tokens. At this level, we have already lexed so we cannot (currently)
1235 // handle this as a context dependent token, instead we detect adjacent tokens
1236 // and return the combined identifier.
1237 if (Lexer.is(AsmToken::Dollar)) {
1238 SMLoc DollarLoc = getLexer().getLoc();
1240 // Consume the dollar sign, and check for a following identifier.
1242 if (Lexer.isNot(AsmToken::Identifier))
1245 // We have a '$' followed by an identifier, make sure they are adjacent.
1246 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1249 // Construct the joined identifier and consume the token.
1250 Res = StringRef(DollarLoc.getPointer(),
1251 getTok().getIdentifier().size() + 1);
1256 if (Lexer.isNot(AsmToken::Identifier) &&
1257 Lexer.isNot(AsmToken::String))
1260 Res = getTok().getIdentifier();
1262 Lex(); // Consume the identifier token.
1267 /// ParseDirectiveSet:
1268 /// ::= .set identifier ',' expression
1269 bool AsmParser::ParseDirectiveSet() {
1272 if (ParseIdentifier(Name))
1273 return TokError("expected identifier after '.set' directive");
1275 if (getLexer().isNot(AsmToken::Comma))
1276 return TokError("unexpected token in '.set'");
1279 return ParseAssignment(Name);
1282 bool AsmParser::ParseEscapedString(std::string &Data) {
1283 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1286 StringRef Str = getTok().getStringContents();
1287 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1288 if (Str[i] != '\\') {
1293 // Recognize escaped characters. Note that this escape semantics currently
1294 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1297 return TokError("unexpected backslash at end of string");
1299 // Recognize octal sequences.
1300 if ((unsigned) (Str[i] - '0') <= 7) {
1301 // Consume up to three octal characters.
1302 unsigned Value = Str[i] - '0';
1304 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1306 Value = Value * 8 + (Str[i] - '0');
1308 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1310 Value = Value * 8 + (Str[i] - '0');
1315 return TokError("invalid octal escape sequence (out of range)");
1317 Data += (unsigned char) Value;
1321 // Otherwise recognize individual escapes.
1324 // Just reject invalid escape sequences for now.
1325 return TokError("invalid escape sequence (unrecognized character)");
1327 case 'b': Data += '\b'; break;
1328 case 'f': Data += '\f'; break;
1329 case 'n': Data += '\n'; break;
1330 case 'r': Data += '\r'; break;
1331 case 't': Data += '\t'; break;
1332 case '"': Data += '"'; break;
1333 case '\\': Data += '\\'; break;
1340 /// ParseDirectiveAscii:
1341 /// ::= ( .ascii | .asciz ) [ "string" ( , "string" )* ]
1342 bool AsmParser::ParseDirectiveAscii(bool ZeroTerminated) {
1343 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1344 CheckForValidSection();
1347 if (getLexer().isNot(AsmToken::String))
1348 return TokError("expected string in '.ascii' or '.asciz' directive");
1351 if (ParseEscapedString(Data))
1354 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1356 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1360 if (getLexer().is(AsmToken::EndOfStatement))
1363 if (getLexer().isNot(AsmToken::Comma))
1364 return TokError("unexpected token in '.ascii' or '.asciz' directive");
1373 /// ParseDirectiveValue
1374 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1375 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1376 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1377 CheckForValidSection();
1380 const MCExpr *Value;
1381 SMLoc ATTRIBUTE_UNUSED StartLoc = getLexer().getLoc();
1382 if (ParseExpression(Value))
1385 // Special case constant expressions to match code generator.
1386 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1387 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1389 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1391 if (getLexer().is(AsmToken::EndOfStatement))
1394 // FIXME: Improve diagnostic.
1395 if (getLexer().isNot(AsmToken::Comma))
1396 return TokError("unexpected token in directive");
1405 /// ParseDirectiveSpace
1406 /// ::= .space expression [ , expression ]
1407 bool AsmParser::ParseDirectiveSpace() {
1408 CheckForValidSection();
1411 if (ParseAbsoluteExpression(NumBytes))
1414 int64_t FillExpr = 0;
1415 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1416 if (getLexer().isNot(AsmToken::Comma))
1417 return TokError("unexpected token in '.space' directive");
1420 if (ParseAbsoluteExpression(FillExpr))
1423 if (getLexer().isNot(AsmToken::EndOfStatement))
1424 return TokError("unexpected token in '.space' directive");
1430 return TokError("invalid number of bytes in '.space' directive");
1432 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1433 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1438 /// ParseDirectiveZero
1439 /// ::= .zero expression
1440 bool AsmParser::ParseDirectiveZero() {
1441 CheckForValidSection();
1444 if (ParseAbsoluteExpression(NumBytes))
1447 if (getLexer().isNot(AsmToken::EndOfStatement))
1448 return TokError("unexpected token in '.zero' directive");
1452 getStreamer().EmitFill(NumBytes, 0, DEFAULT_ADDRSPACE);
1457 /// ParseDirectiveFill
1458 /// ::= .fill expression , expression , expression
1459 bool AsmParser::ParseDirectiveFill() {
1460 CheckForValidSection();
1463 if (ParseAbsoluteExpression(NumValues))
1466 if (getLexer().isNot(AsmToken::Comma))
1467 return TokError("unexpected token in '.fill' directive");
1471 if (ParseAbsoluteExpression(FillSize))
1474 if (getLexer().isNot(AsmToken::Comma))
1475 return TokError("unexpected token in '.fill' directive");
1479 if (ParseAbsoluteExpression(FillExpr))
1482 if (getLexer().isNot(AsmToken::EndOfStatement))
1483 return TokError("unexpected token in '.fill' directive");
1487 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1488 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1490 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1491 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1496 /// ParseDirectiveOrg
1497 /// ::= .org expression [ , expression ]
1498 bool AsmParser::ParseDirectiveOrg() {
1499 CheckForValidSection();
1501 const MCExpr *Offset;
1502 if (ParseExpression(Offset))
1505 // Parse optional fill expression.
1506 int64_t FillExpr = 0;
1507 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1508 if (getLexer().isNot(AsmToken::Comma))
1509 return TokError("unexpected token in '.org' directive");
1512 if (ParseAbsoluteExpression(FillExpr))
1515 if (getLexer().isNot(AsmToken::EndOfStatement))
1516 return TokError("unexpected token in '.org' directive");
1521 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1522 // has to be relative to the current section.
1523 getStreamer().EmitValueToOffset(Offset, FillExpr);
1528 /// ParseDirectiveAlign
1529 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1530 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1531 CheckForValidSection();
1533 SMLoc AlignmentLoc = getLexer().getLoc();
1535 if (ParseAbsoluteExpression(Alignment))
1539 bool HasFillExpr = false;
1540 int64_t FillExpr = 0;
1541 int64_t MaxBytesToFill = 0;
1542 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1543 if (getLexer().isNot(AsmToken::Comma))
1544 return TokError("unexpected token in directive");
1547 // The fill expression can be omitted while specifying a maximum number of
1548 // alignment bytes, e.g:
1550 if (getLexer().isNot(AsmToken::Comma)) {
1552 if (ParseAbsoluteExpression(FillExpr))
1556 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1557 if (getLexer().isNot(AsmToken::Comma))
1558 return TokError("unexpected token in directive");
1561 MaxBytesLoc = getLexer().getLoc();
1562 if (ParseAbsoluteExpression(MaxBytesToFill))
1565 if (getLexer().isNot(AsmToken::EndOfStatement))
1566 return TokError("unexpected token in directive");
1575 // Compute alignment in bytes.
1577 // FIXME: Diagnose overflow.
1578 if (Alignment >= 32) {
1579 Error(AlignmentLoc, "invalid alignment value");
1583 Alignment = 1ULL << Alignment;
1586 // Diagnose non-sensical max bytes to align.
1587 if (MaxBytesLoc.isValid()) {
1588 if (MaxBytesToFill < 1) {
1589 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1590 "many bytes, ignoring maximum bytes expression");
1594 if (MaxBytesToFill >= Alignment) {
1595 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1601 // Check whether we should use optimal code alignment for this .align
1604 // FIXME: This should be using a target hook.
1605 bool UseCodeAlign = false;
1606 if (const MCSectionMachO *S = dyn_cast<MCSectionMachO>(
1607 getStreamer().getCurrentSection()))
1608 UseCodeAlign = S->hasAttribute(MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS);
1609 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1610 ValueSize == 1 && UseCodeAlign) {
1611 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1613 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1614 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1621 /// ParseDirectiveSymbolAttribute
1622 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1623 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1624 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1628 if (ParseIdentifier(Name))
1629 return TokError("expected identifier in directive");
1631 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1633 getStreamer().EmitSymbolAttribute(Sym, Attr);
1635 if (getLexer().is(AsmToken::EndOfStatement))
1638 if (getLexer().isNot(AsmToken::Comma))
1639 return TokError("unexpected token in directive");
1648 /// ParseDirectiveELFType
1649 /// ::= .type identifier , @attribute
1650 bool AsmParser::ParseDirectiveELFType() {
1652 if (ParseIdentifier(Name))
1653 return TokError("expected identifier in directive");
1655 // Handle the identifier as the key symbol.
1656 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1658 if (getLexer().isNot(AsmToken::Comma))
1659 return TokError("unexpected token in '.type' directive");
1662 if (getLexer().isNot(AsmToken::At))
1663 return TokError("expected '@' before type");
1669 TypeLoc = getLexer().getLoc();
1670 if (ParseIdentifier(Type))
1671 return TokError("expected symbol type in directive");
1673 MCSymbolAttr Attr = StringSwitch<MCSymbolAttr>(Type)
1674 .Case("function", MCSA_ELF_TypeFunction)
1675 .Case("object", MCSA_ELF_TypeObject)
1676 .Case("tls_object", MCSA_ELF_TypeTLS)
1677 .Case("common", MCSA_ELF_TypeCommon)
1678 .Case("notype", MCSA_ELF_TypeNoType)
1679 .Default(MCSA_Invalid);
1681 if (Attr == MCSA_Invalid)
1682 return Error(TypeLoc, "unsupported attribute in '.type' directive");
1684 if (getLexer().isNot(AsmToken::EndOfStatement))
1685 return TokError("unexpected token in '.type' directive");
1689 getStreamer().EmitSymbolAttribute(Sym, Attr);
1694 /// ParseDirectiveComm
1695 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1696 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1697 CheckForValidSection();
1699 SMLoc IDLoc = getLexer().getLoc();
1701 if (ParseIdentifier(Name))
1702 return TokError("expected identifier in directive");
1704 // Handle the identifier as the key symbol.
1705 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1707 if (getLexer().isNot(AsmToken::Comma))
1708 return TokError("unexpected token in directive");
1712 SMLoc SizeLoc = getLexer().getLoc();
1713 if (ParseAbsoluteExpression(Size))
1716 int64_t Pow2Alignment = 0;
1717 SMLoc Pow2AlignmentLoc;
1718 if (getLexer().is(AsmToken::Comma)) {
1720 Pow2AlignmentLoc = getLexer().getLoc();
1721 if (ParseAbsoluteExpression(Pow2Alignment))
1724 // If this target takes alignments in bytes (not log) validate and convert.
1725 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1726 if (!isPowerOf2_64(Pow2Alignment))
1727 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1728 Pow2Alignment = Log2_64(Pow2Alignment);
1732 if (getLexer().isNot(AsmToken::EndOfStatement))
1733 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1737 // NOTE: a size of zero for a .comm should create a undefined symbol
1738 // but a size of .lcomm creates a bss symbol of size zero.
1740 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1741 "be less than zero");
1743 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1744 // may internally end up wanting an alignment in bytes.
1745 // FIXME: Diagnose overflow.
1746 if (Pow2Alignment < 0)
1747 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1748 "alignment, can't be less than zero");
1750 if (!Sym->isUndefined())
1751 return Error(IDLoc, "invalid symbol redefinition");
1753 // '.lcomm' is equivalent to '.zerofill'.
1754 // Create the Symbol as a common or local common with Size and Pow2Alignment
1756 getStreamer().EmitZerofill(Ctx.getMachOSection(
1757 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1758 0, SectionKind::getBSS()),
1759 Sym, Size, 1 << Pow2Alignment);
1763 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1767 /// ParseDirectiveAbort
1768 /// ::= .abort [... message ...]
1769 bool AsmParser::ParseDirectiveAbort() {
1770 // FIXME: Use loc from directive.
1771 SMLoc Loc = getLexer().getLoc();
1773 StringRef Str = ParseStringToEndOfStatement();
1774 if (getLexer().isNot(AsmToken::EndOfStatement))
1775 return TokError("unexpected token in '.abort' directive");
1780 Error(Loc, ".abort detected. Assembly stopping.");
1782 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1783 // FIXME: Actually abort assembly here.
1788 /// ParseDirectiveInclude
1789 /// ::= .include "filename"
1790 bool AsmParser::ParseDirectiveInclude() {
1791 if (getLexer().isNot(AsmToken::String))
1792 return TokError("expected string in '.include' directive");
1794 std::string Filename = getTok().getString();
1795 SMLoc IncludeLoc = getLexer().getLoc();
1798 if (getLexer().isNot(AsmToken::EndOfStatement))
1799 return TokError("unexpected token in '.include' directive");
1801 // Strip the quotes.
1802 Filename = Filename.substr(1, Filename.size()-2);
1804 // Attempt to switch the lexer to the included file before consuming the end
1805 // of statement to avoid losing it when we switch.
1806 if (EnterIncludeFile(Filename)) {
1807 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1814 /// ParseDirectiveIf
1815 /// ::= .if expression
1816 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1817 TheCondStack.push_back(TheCondState);
1818 TheCondState.TheCond = AsmCond::IfCond;
1819 if(TheCondState.Ignore) {
1820 EatToEndOfStatement();
1824 if (ParseAbsoluteExpression(ExprValue))
1827 if (getLexer().isNot(AsmToken::EndOfStatement))
1828 return TokError("unexpected token in '.if' directive");
1832 TheCondState.CondMet = ExprValue;
1833 TheCondState.Ignore = !TheCondState.CondMet;
1839 /// ParseDirectiveElseIf
1840 /// ::= .elseif expression
1841 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1842 if (TheCondState.TheCond != AsmCond::IfCond &&
1843 TheCondState.TheCond != AsmCond::ElseIfCond)
1844 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1846 TheCondState.TheCond = AsmCond::ElseIfCond;
1848 bool LastIgnoreState = false;
1849 if (!TheCondStack.empty())
1850 LastIgnoreState = TheCondStack.back().Ignore;
1851 if (LastIgnoreState || TheCondState.CondMet) {
1852 TheCondState.Ignore = true;
1853 EatToEndOfStatement();
1857 if (ParseAbsoluteExpression(ExprValue))
1860 if (getLexer().isNot(AsmToken::EndOfStatement))
1861 return TokError("unexpected token in '.elseif' directive");
1864 TheCondState.CondMet = ExprValue;
1865 TheCondState.Ignore = !TheCondState.CondMet;
1871 /// ParseDirectiveElse
1873 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
1874 if (getLexer().isNot(AsmToken::EndOfStatement))
1875 return TokError("unexpected token in '.else' directive");
1879 if (TheCondState.TheCond != AsmCond::IfCond &&
1880 TheCondState.TheCond != AsmCond::ElseIfCond)
1881 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
1883 TheCondState.TheCond = AsmCond::ElseCond;
1884 bool LastIgnoreState = false;
1885 if (!TheCondStack.empty())
1886 LastIgnoreState = TheCondStack.back().Ignore;
1887 if (LastIgnoreState || TheCondState.CondMet)
1888 TheCondState.Ignore = true;
1890 TheCondState.Ignore = false;
1895 /// ParseDirectiveEndIf
1897 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
1898 if (getLexer().isNot(AsmToken::EndOfStatement))
1899 return TokError("unexpected token in '.endif' directive");
1903 if ((TheCondState.TheCond == AsmCond::NoCond) ||
1904 TheCondStack.empty())
1905 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
1907 if (!TheCondStack.empty()) {
1908 TheCondState = TheCondStack.back();
1909 TheCondStack.pop_back();
1915 /// ParseDirectiveFile
1916 /// ::= .file [number] string
1917 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
1918 // FIXME: I'm not sure what this is.
1919 int64_t FileNumber = -1;
1920 SMLoc FileNumberLoc = getLexer().getLoc();
1921 if (getLexer().is(AsmToken::Integer)) {
1922 FileNumber = getTok().getIntVal();
1926 return TokError("file number less than one");
1929 if (getLexer().isNot(AsmToken::String))
1930 return TokError("unexpected token in '.file' directive");
1932 StringRef Filename = getTok().getString();
1933 Filename = Filename.substr(1, Filename.size()-2);
1936 if (getLexer().isNot(AsmToken::EndOfStatement))
1937 return TokError("unexpected token in '.file' directive");
1939 if (FileNumber == -1)
1940 getStreamer().EmitFileDirective(Filename);
1942 if (getContext().GetDwarfFile(Filename, FileNumber) == 0)
1943 Error(FileNumberLoc, "file number already allocated");
1944 getStreamer().EmitDwarfFileDirective(FileNumber, Filename);
1950 /// ParseDirectiveLine
1951 /// ::= .line [number]
1952 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
1953 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1954 if (getLexer().isNot(AsmToken::Integer))
1955 return TokError("unexpected token in '.line' directive");
1957 int64_t LineNumber = getTok().getIntVal();
1961 // FIXME: Do something with the .line.
1964 if (getLexer().isNot(AsmToken::EndOfStatement))
1965 return TokError("unexpected token in '.line' directive");
1971 /// ParseDirectiveLoc
1972 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
1973 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
1974 /// The first number is a file number, must have been previously assigned with
1975 /// a .file directive, the second number is the line number and optionally the
1976 /// third number is a column position (zero if not specified). The remaining
1977 /// optional items are .loc sub-directives.
1978 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
1980 if (getLexer().isNot(AsmToken::Integer))
1981 return TokError("unexpected token in '.loc' directive");
1982 int64_t FileNumber = getTok().getIntVal();
1984 return TokError("file number less than one in '.loc' directive");
1985 if (!getContext().ValidateDwarfFileNumber(FileNumber))
1986 return TokError("unassigned file number in '.loc' directive");
1989 int64_t LineNumber = 0;
1990 if (getLexer().is(AsmToken::Integer)) {
1991 LineNumber = getTok().getIntVal();
1993 return TokError("line number less than one in '.loc' directive");
1997 int64_t ColumnPos = 0;
1998 if (getLexer().is(AsmToken::Integer)) {
1999 ColumnPos = getTok().getIntVal();
2001 return TokError("column position less than zero in '.loc' directive");
2007 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2009 if (getLexer().is(AsmToken::EndOfStatement))
2013 SMLoc Loc = getTok().getLoc();
2014 if (getParser().ParseIdentifier(Name))
2015 return TokError("unexpected token in '.loc' directive");
2017 if (Name == "basic_block")
2018 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2019 else if (Name == "prologue_end")
2020 Flags |= DWARF2_FLAG_PROLOGUE_END;
2021 else if (Name == "epilogue_begin")
2022 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2023 else if (Name == "is_stmt") {
2024 SMLoc Loc = getTok().getLoc();
2025 const MCExpr *Value;
2026 if (getParser().ParseExpression(Value))
2028 // The expression must be the constant 0 or 1.
2029 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2030 int Value = MCE->getValue();
2032 Flags &= ~DWARF2_FLAG_IS_STMT;
2033 else if (Value == 1)
2034 Flags |= DWARF2_FLAG_IS_STMT;
2036 return Error(Loc, "is_stmt value not 0 or 1");
2039 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2042 else if (Name == "isa") {
2043 SMLoc Loc = getTok().getLoc();
2044 const MCExpr *Value;
2045 if (getParser().ParseExpression(Value))
2047 // The expression must be a constant greater or equal to 0.
2048 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2049 int Value = MCE->getValue();
2051 return Error(Loc, "isa number less than zero");
2055 return Error(Loc, "isa number not a constant value");
2059 return Error(Loc, "unknown sub-directive in '.loc' directive");
2062 if (getLexer().is(AsmToken::EndOfStatement))
2067 getContext().setCurrentDwarfLoc(FileNumber, LineNumber, ColumnPos, Flags,Isa);
2072 /// ParseDirectiveMacrosOnOff
2075 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2076 SMLoc DirectiveLoc) {
2077 if (getLexer().isNot(AsmToken::EndOfStatement))
2078 return Error(getLexer().getLoc(),
2079 "unexpected token in '" + Directive + "' directive");
2081 getParser().MacrosEnabled = Directive == ".macros_on";
2086 /// ParseDirectiveMacro
2088 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2089 SMLoc DirectiveLoc) {
2091 if (getParser().ParseIdentifier(Name))
2092 return TokError("expected identifier in directive");
2094 if (getLexer().isNot(AsmToken::EndOfStatement))
2095 return TokError("unexpected token in '.macro' directive");
2097 // Eat the end of statement.
2100 AsmToken EndToken, StartToken = getTok();
2102 // Lex the macro definition.
2104 // Check whether we have reached the end of the file.
2105 if (getLexer().is(AsmToken::Eof))
2106 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2108 // Otherwise, check whether we have reach the .endmacro.
2109 if (getLexer().is(AsmToken::Identifier) &&
2110 (getTok().getIdentifier() == ".endm" ||
2111 getTok().getIdentifier() == ".endmacro")) {
2112 EndToken = getTok();
2114 if (getLexer().isNot(AsmToken::EndOfStatement))
2115 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2120 // Otherwise, scan til the end of the statement.
2121 getParser().EatToEndOfStatement();
2124 if (getParser().MacroMap.lookup(Name)) {
2125 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2128 const char *BodyStart = StartToken.getLoc().getPointer();
2129 const char *BodyEnd = EndToken.getLoc().getPointer();
2130 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2131 getParser().MacroMap[Name] = new Macro(Name, Body);
2135 /// ParseDirectiveEndMacro
2138 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2139 SMLoc DirectiveLoc) {
2140 if (getLexer().isNot(AsmToken::EndOfStatement))
2141 return TokError("unexpected token in '" + Directive + "' directive");
2143 // If we are inside a macro instantiation, terminate the current
2145 if (!getParser().ActiveMacros.empty()) {
2146 getParser().HandleMacroExit();
2150 // Otherwise, this .endmacro is a stray entry in the file; well formed
2151 // .endmacro directives are handled during the macro definition parsing.
2152 return TokError("unexpected '" + Directive + "' in file, "
2153 "no current macro definition");
2156 void GenericAsmParser::ParseUleb128(uint64_t Value) {
2157 const uint64_t Mask = (1 << 7) - 1;
2159 unsigned Byte = Value & Mask;
2161 if (Value) // Not the last one
2163 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2167 void GenericAsmParser::ParseSleb128(int64_t Value) {
2168 const int64_t Mask = (1 << 7) - 1;
2170 unsigned Byte = Value & Mask;
2172 bool Done = ((Value == 0 && (Byte & 0x40) == 0) ||
2173 (Value == -1 && (Byte & 0x40) != 0));
2176 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2182 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2184 if (getParser().ParseAbsoluteExpression(Value))
2187 if (getLexer().isNot(AsmToken::EndOfStatement))
2188 return TokError("unexpected token in directive");
2190 if (DirName[1] == 's')
2191 ParseSleb128(Value);
2193 ParseUleb128(Value);
2198 /// \brief Create an MCAsmParser instance.
2199 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2200 MCContext &C, MCStreamer &Out,
2201 const MCAsmInfo &MAI) {
2202 return new AsmParser(T, SM, C, Out, MAI);