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/APFloat.h"
15 #include "llvm/ADT/SmallString.h"
16 #include "llvm/ADT/StringMap.h"
17 #include "llvm/ADT/StringSwitch.h"
18 #include "llvm/ADT/Twine.h"
19 #include "llvm/MC/MCAsmInfo.h"
20 #include "llvm/MC/MCContext.h"
21 #include "llvm/MC/MCExpr.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/MemoryBuffer.h"
31 #include "llvm/Support/SourceMgr.h"
32 #include "llvm/Support/raw_ostream.h"
33 #include "llvm/Target/TargetAsmParser.h"
39 /// \brief Helper class for tracking macro definitions.
45 Macro(StringRef N, StringRef B) : Name(N), Body(B) {}
48 /// \brief Helper class for storing information about an active macro
50 struct MacroInstantiation {
51 /// The macro being instantiated.
52 const Macro *TheMacro;
54 /// The macro instantiation with substitutions.
55 MemoryBuffer *Instantiation;
57 /// The location of the instantiation.
58 SMLoc InstantiationLoc;
60 /// The location where parsing should resume upon instantiation completion.
64 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
65 const std::vector<std::vector<AsmToken> > &A);
68 /// \brief The concrete assembly parser instance.
69 class AsmParser : public MCAsmParser {
70 friend class GenericAsmParser;
72 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
73 void operator=(const AsmParser &); // DO NOT IMPLEMENT
79 MCAsmParserExtension *GenericParser;
80 MCAsmParserExtension *PlatformParser;
82 /// This is the current buffer index we're lexing from as managed by the
87 std::vector<AsmCond> TheCondStack;
89 /// DirectiveMap - This is a table handlers for directives. Each handler is
90 /// invoked after the directive identifier is read and is responsible for
91 /// parsing and validating the rest of the directive. The handler is passed
92 /// in the directive name and the location of the directive keyword.
93 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
95 /// MacroMap - Map of currently defined macros.
96 StringMap<Macro*> MacroMap;
98 /// ActiveMacros - Stack of active macro instantiations.
99 std::vector<MacroInstantiation*> ActiveMacros;
101 /// Boolean tracking whether macro substitution is enabled.
102 unsigned MacrosEnabled : 1;
104 /// Flag tracking whether any errors have been encountered.
105 unsigned HadError : 1;
108 AsmParser(const Target &T, SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
109 const MCAsmInfo &MAI);
112 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
114 void AddDirectiveHandler(MCAsmParserExtension *Object,
116 DirectiveHandler Handler) {
117 DirectiveMap[Directive] = std::make_pair(Object, Handler);
121 /// @name MCAsmParser Interface
124 virtual SourceMgr &getSourceManager() { return SrcMgr; }
125 virtual MCAsmLexer &getLexer() { return Lexer; }
126 virtual MCContext &getContext() { return Ctx; }
127 virtual MCStreamer &getStreamer() { return Out; }
129 virtual void Warning(SMLoc L, const Twine &Meg);
130 virtual bool Error(SMLoc L, const Twine &Msg);
132 const AsmToken &Lex();
134 bool ParseExpression(const MCExpr *&Res);
135 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
136 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
137 virtual bool ParseAbsoluteExpression(int64_t &Res);
142 void CheckForValidSection();
144 bool ParseStatement();
146 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
147 void HandleMacroExit();
149 void PrintMacroInstantiations();
150 void PrintMessage(SMLoc Loc, const Twine &Msg, const char *Type) const {
151 SrcMgr.PrintMessage(Loc, Msg, Type);
154 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
155 bool EnterIncludeFile(const std::string &Filename);
157 /// \brief Reset the current lexer position to that given by \arg Loc. The
158 /// current token is not set; clients should ensure Lex() is called
160 void JumpToLoc(SMLoc Loc);
162 void EatToEndOfStatement();
164 /// \brief Parse up to the end of statement and a return the contents from the
165 /// current token until the end of the statement; the current token on exit
166 /// will be either the EndOfStatement or EOF.
167 StringRef ParseStringToEndOfStatement();
169 bool ParseAssignment(StringRef Name);
171 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
172 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
173 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
175 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
176 /// and set \arg Res to the identifier contents.
177 bool ParseIdentifier(StringRef &Res);
179 // Directive Parsing.
180 bool ParseDirectiveAscii(bool ZeroTerminated); // ".ascii", ".asciiz"
181 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
182 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
183 bool ParseDirectiveFill(); // ".fill"
184 bool ParseDirectiveSpace(); // ".space"
185 bool ParseDirectiveZero(); // ".zero"
186 bool ParseDirectiveSet(); // ".set"
187 bool ParseDirectiveOrg(); // ".org"
188 // ".align{,32}", ".p2align{,w,l}"
189 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
191 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
192 /// accepts a single symbol (which should be a label or an external).
193 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
194 bool ParseDirectiveELFType(); // ELF specific ".type"
196 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
198 bool ParseDirectiveAbort(); // ".abort"
199 bool ParseDirectiveInclude(); // ".include"
201 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
202 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
203 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
204 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
206 /// ParseEscapedString - Parse the current token as a string which may include
207 /// escaped characters and return the string contents.
208 bool ParseEscapedString(std::string &Data);
210 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
211 MCSymbolRefExpr::VariantKind Variant);
214 /// \brief Generic implementations of directive handling, etc. which is shared
215 /// (or the default, at least) for all assembler parser.
216 class GenericAsmParser : public MCAsmParserExtension {
217 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
218 void AddDirectiveHandler(StringRef Directive) {
219 getParser().AddDirectiveHandler(this, Directive,
220 HandleDirective<GenericAsmParser, Handler>);
224 GenericAsmParser() {}
226 AsmParser &getParser() {
227 return (AsmParser&) this->MCAsmParserExtension::getParser();
230 virtual void Initialize(MCAsmParser &Parser) {
231 // Call the base implementation.
232 this->MCAsmParserExtension::Initialize(Parser);
234 // Debugging directives.
235 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
236 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
237 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
238 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
241 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
243 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
245 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
246 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
247 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
249 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
250 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
253 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
254 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
255 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
256 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc);
258 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
259 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
260 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
262 void ParseUleb128(uint64_t Value);
263 void ParseSleb128(int64_t Value);
264 bool ParseDirectiveLEB128(StringRef, SMLoc);
271 extern MCAsmParserExtension *createDarwinAsmParser();
272 extern MCAsmParserExtension *createELFAsmParser();
273 extern MCAsmParserExtension *createCOFFAsmParser();
277 enum { DEFAULT_ADDRSPACE = 0 };
279 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
280 MCStreamer &_Out, const MCAsmInfo &_MAI)
281 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
282 GenericParser(new GenericAsmParser), PlatformParser(0),
283 CurBuffer(0), MacrosEnabled(true) {
284 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
286 // Initialize the generic parser.
287 GenericParser->Initialize(*this);
289 // Initialize the platform / file format parser.
291 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
293 if (_MAI.hasMicrosoftFastStdCallMangling()) {
294 PlatformParser = createCOFFAsmParser();
295 PlatformParser->Initialize(*this);
296 } else if (_MAI.hasSubsectionsViaSymbols()) {
297 PlatformParser = createDarwinAsmParser();
298 PlatformParser->Initialize(*this);
300 PlatformParser = createELFAsmParser();
301 PlatformParser->Initialize(*this);
305 AsmParser::~AsmParser() {
306 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
308 // Destroy any macros.
309 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
310 ie = MacroMap.end(); it != ie; ++it)
311 delete it->getValue();
313 delete PlatformParser;
314 delete GenericParser;
317 void AsmParser::PrintMacroInstantiations() {
318 // Print the active macro instantiation stack.
319 for (std::vector<MacroInstantiation*>::const_reverse_iterator
320 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
321 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
325 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
326 PrintMessage(L, Msg, "warning");
327 PrintMacroInstantiations();
330 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
332 PrintMessage(L, Msg, "error");
333 PrintMacroInstantiations();
337 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
338 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
344 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
349 void AsmParser::JumpToLoc(SMLoc Loc) {
350 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
351 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
354 const AsmToken &AsmParser::Lex() {
355 const AsmToken *tok = &Lexer.Lex();
357 if (tok->is(AsmToken::Eof)) {
358 // If this is the end of an included file, pop the parent file off the
360 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
361 if (ParentIncludeLoc != SMLoc()) {
362 JumpToLoc(ParentIncludeLoc);
367 if (tok->is(AsmToken::Error))
368 Error(Lexer.getErrLoc(), Lexer.getErr());
373 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
374 // Create the initial section, if requested.
375 if (!NoInitialTextSection)
382 AsmCond StartingCondState = TheCondState;
384 // While we have input, parse each statement.
385 while (Lexer.isNot(AsmToken::Eof)) {
386 if (!ParseStatement()) continue;
388 // We had an error, validate that one was emitted and recover by skipping to
390 assert(HadError && "Parse statement returned an error, but none emitted!");
391 EatToEndOfStatement();
394 if (TheCondState.TheCond != StartingCondState.TheCond ||
395 TheCondState.Ignore != StartingCondState.Ignore)
396 return TokError("unmatched .ifs or .elses");
398 // Check to see there are no empty DwarfFile slots.
399 const std::vector<MCDwarfFile *> &MCDwarfFiles =
400 getContext().getMCDwarfFiles();
401 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
402 if (!MCDwarfFiles[i])
403 TokError("unassigned file number: " + Twine(i) + " for .file directives");
406 // Finalize the output stream if there are no errors and if the client wants
408 if (!HadError && !NoFinalize)
414 void AsmParser::CheckForValidSection() {
415 if (!getStreamer().getCurrentSection()) {
416 TokError("expected section directive before assembly directive");
417 Out.SwitchSection(Ctx.getMachOSection(
419 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
420 0, SectionKind::getText()));
424 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
425 void AsmParser::EatToEndOfStatement() {
426 while (Lexer.isNot(AsmToken::EndOfStatement) &&
427 Lexer.isNot(AsmToken::Eof))
431 if (Lexer.is(AsmToken::EndOfStatement))
435 StringRef AsmParser::ParseStringToEndOfStatement() {
436 const char *Start = getTok().getLoc().getPointer();
438 while (Lexer.isNot(AsmToken::EndOfStatement) &&
439 Lexer.isNot(AsmToken::Eof))
442 const char *End = getTok().getLoc().getPointer();
443 return StringRef(Start, End - Start);
446 /// ParseParenExpr - Parse a paren expression and return it.
447 /// NOTE: This assumes the leading '(' has already been consumed.
449 /// parenexpr ::= expr)
451 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
452 if (ParseExpression(Res)) return true;
453 if (Lexer.isNot(AsmToken::RParen))
454 return TokError("expected ')' in parentheses expression");
455 EndLoc = Lexer.getLoc();
460 /// ParsePrimaryExpr - Parse a primary expression and return it.
461 /// primaryexpr ::= (parenexpr
462 /// primaryexpr ::= symbol
463 /// primaryexpr ::= number
464 /// primaryexpr ::= '.'
465 /// primaryexpr ::= ~,+,- primaryexpr
466 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
467 switch (Lexer.getKind()) {
469 return TokError("unknown token in expression");
470 case AsmToken::Exclaim:
471 Lex(); // Eat the operator.
472 if (ParsePrimaryExpr(Res, EndLoc))
474 Res = MCUnaryExpr::CreateLNot(Res, getContext());
476 case AsmToken::Dollar:
477 case AsmToken::String:
478 case AsmToken::Identifier: {
479 EndLoc = Lexer.getLoc();
481 StringRef Identifier;
482 if (ParseIdentifier(Identifier))
485 // This is a symbol reference.
486 std::pair<StringRef, StringRef> Split = Identifier.split('@');
487 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
489 // Mark the symbol as used in an expression.
490 Sym->setUsedInExpr(true);
492 // Lookup the symbol variant if used.
493 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
494 if (Split.first.size() != Identifier.size()) {
495 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
496 if (Variant == MCSymbolRefExpr::VK_Invalid) {
497 Variant = MCSymbolRefExpr::VK_None;
498 TokError("invalid variant '" + Split.second + "'");
502 // If this is an absolute variable reference, substitute it now to preserve
503 // semantics in the face of reassignment.
504 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
506 return Error(EndLoc, "unexpected modified on variable reference");
508 Res = Sym->getVariableValue();
512 // Otherwise create a symbol ref.
513 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
516 case AsmToken::Integer: {
517 SMLoc Loc = getTok().getLoc();
518 int64_t IntVal = getTok().getIntVal();
519 Res = MCConstantExpr::Create(IntVal, getContext());
520 EndLoc = Lexer.getLoc();
522 // Look for 'b' or 'f' following an Integer as a directional label
523 if (Lexer.getKind() == AsmToken::Identifier) {
524 StringRef IDVal = getTok().getString();
525 if (IDVal == "f" || IDVal == "b"){
526 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
527 IDVal == "f" ? 1 : 0);
528 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
530 if(IDVal == "b" && Sym->isUndefined())
531 return Error(Loc, "invalid reference to undefined symbol");
532 EndLoc = Lexer.getLoc();
533 Lex(); // Eat identifier.
538 case AsmToken::Dot: {
539 // This is a '.' reference, which references the current PC. Emit a
540 // temporary label to the streamer and refer to it.
541 MCSymbol *Sym = Ctx.CreateTempSymbol();
543 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
544 EndLoc = Lexer.getLoc();
545 Lex(); // Eat identifier.
549 case AsmToken::LParen:
550 Lex(); // Eat the '('.
551 return ParseParenExpr(Res, EndLoc);
552 case AsmToken::Minus:
553 Lex(); // Eat the operator.
554 if (ParsePrimaryExpr(Res, EndLoc))
556 Res = MCUnaryExpr::CreateMinus(Res, getContext());
559 Lex(); // Eat the operator.
560 if (ParsePrimaryExpr(Res, EndLoc))
562 Res = MCUnaryExpr::CreatePlus(Res, getContext());
564 case AsmToken::Tilde:
565 Lex(); // Eat the operator.
566 if (ParsePrimaryExpr(Res, EndLoc))
568 Res = MCUnaryExpr::CreateNot(Res, getContext());
573 bool AsmParser::ParseExpression(const MCExpr *&Res) {
575 return ParseExpression(Res, EndLoc);
579 AsmParser::ApplyModifierToExpr(const MCExpr *E,
580 MCSymbolRefExpr::VariantKind Variant) {
581 // Recurse over the given expression, rebuilding it to apply the given variant
582 // if there is exactly one symbol.
583 switch (E->getKind()) {
585 case MCExpr::Constant:
588 case MCExpr::SymbolRef: {
589 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
591 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
592 TokError("invalid variant on expression '" +
593 getTok().getIdentifier() + "' (already modified)");
597 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
600 case MCExpr::Unary: {
601 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
602 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
605 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
608 case MCExpr::Binary: {
609 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
610 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
611 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
616 if (!LHS) LHS = BE->getLHS();
617 if (!RHS) RHS = BE->getRHS();
619 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
623 assert(0 && "Invalid expression kind!");
627 /// ParseExpression - Parse an expression and return it.
629 /// expr ::= expr +,- expr -> lowest.
630 /// expr ::= expr |,^,&,! expr -> middle.
631 /// expr ::= expr *,/,%,<<,>> expr -> highest.
632 /// expr ::= primaryexpr
634 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
635 // Parse the expression.
637 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
640 // As a special case, we support 'a op b @ modifier' by rewriting the
641 // expression to include the modifier. This is inefficient, but in general we
642 // expect users to use 'a@modifier op b'.
643 if (Lexer.getKind() == AsmToken::At) {
646 if (Lexer.isNot(AsmToken::Identifier))
647 return TokError("unexpected symbol modifier following '@'");
649 MCSymbolRefExpr::VariantKind Variant =
650 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
651 if (Variant == MCSymbolRefExpr::VK_Invalid)
652 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
654 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
656 return TokError("invalid modifier '" + getTok().getIdentifier() +
657 "' (no symbols present)");
665 // Try to constant fold it up front, if possible.
667 if (Res->EvaluateAsAbsolute(Value))
668 Res = MCConstantExpr::Create(Value, getContext());
673 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
675 return ParseParenExpr(Res, EndLoc) ||
676 ParseBinOpRHS(1, Res, EndLoc);
679 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
682 SMLoc StartLoc = Lexer.getLoc();
683 if (ParseExpression(Expr))
686 if (!Expr->EvaluateAsAbsolute(Res))
687 return Error(StartLoc, "expected absolute expression");
692 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
693 MCBinaryExpr::Opcode &Kind) {
696 return 0; // not a binop.
698 // Lowest Precedence: &&, ||, @
699 case AsmToken::AmpAmp:
700 Kind = MCBinaryExpr::LAnd;
702 case AsmToken::PipePipe:
703 Kind = MCBinaryExpr::LOr;
707 // Low Precedence: |, &, ^
709 // FIXME: gas seems to support '!' as an infix operator?
711 Kind = MCBinaryExpr::Or;
713 case AsmToken::Caret:
714 Kind = MCBinaryExpr::Xor;
717 Kind = MCBinaryExpr::And;
720 // Intermediate Precedence: +, -, ==, !=, <>, <, <=, >, >=
722 Kind = MCBinaryExpr::Add;
724 case AsmToken::Minus:
725 Kind = MCBinaryExpr::Sub;
727 case AsmToken::EqualEqual:
728 Kind = MCBinaryExpr::EQ;
730 case AsmToken::ExclaimEqual:
731 case AsmToken::LessGreater:
732 Kind = MCBinaryExpr::NE;
735 Kind = MCBinaryExpr::LT;
737 case AsmToken::LessEqual:
738 Kind = MCBinaryExpr::LTE;
740 case AsmToken::Greater:
741 Kind = MCBinaryExpr::GT;
743 case AsmToken::GreaterEqual:
744 Kind = MCBinaryExpr::GTE;
747 // Highest Precedence: *, /, %, <<, >>
749 Kind = MCBinaryExpr::Mul;
751 case AsmToken::Slash:
752 Kind = MCBinaryExpr::Div;
754 case AsmToken::Percent:
755 Kind = MCBinaryExpr::Mod;
757 case AsmToken::LessLess:
758 Kind = MCBinaryExpr::Shl;
760 case AsmToken::GreaterGreater:
761 Kind = MCBinaryExpr::Shr;
767 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
768 /// Res contains the LHS of the expression on input.
769 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
772 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
773 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
775 // If the next token is lower precedence than we are allowed to eat, return
776 // successfully with what we ate already.
777 if (TokPrec < Precedence)
782 // Eat the next primary expression.
784 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
786 // If BinOp binds less tightly with RHS than the operator after RHS, let
787 // the pending operator take RHS as its LHS.
788 MCBinaryExpr::Opcode Dummy;
789 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
790 if (TokPrec < NextTokPrec) {
791 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
794 // Merge LHS and RHS according to operator.
795 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
803 /// ::= EndOfStatement
804 /// ::= Label* Directive ...Operands... EndOfStatement
805 /// ::= Label* Identifier OperandList* EndOfStatement
806 bool AsmParser::ParseStatement() {
807 if (Lexer.is(AsmToken::EndOfStatement)) {
813 // Statements always start with an identifier.
814 AsmToken ID = getTok();
815 SMLoc IDLoc = ID.getLoc();
817 int64_t LocalLabelVal = -1;
818 // GUESS allow an integer followed by a ':' as a directional local label
819 if (Lexer.is(AsmToken::Integer)) {
820 LocalLabelVal = getTok().getIntVal();
821 if (LocalLabelVal < 0) {
822 if (!TheCondState.Ignore)
823 return TokError("unexpected token at start of statement");
827 IDVal = getTok().getString();
828 Lex(); // Consume the integer token to be used as an identifier token.
829 if (Lexer.getKind() != AsmToken::Colon) {
830 if (!TheCondState.Ignore)
831 return TokError("unexpected token at start of statement");
835 else if (ParseIdentifier(IDVal)) {
836 if (!TheCondState.Ignore)
837 return TokError("unexpected token at start of statement");
841 // Handle conditional assembly here before checking for skipping. We
842 // have to do this so that .endif isn't skipped in a ".if 0" block for
845 return ParseDirectiveIf(IDLoc);
846 if (IDVal == ".elseif")
847 return ParseDirectiveElseIf(IDLoc);
848 if (IDVal == ".else")
849 return ParseDirectiveElse(IDLoc);
850 if (IDVal == ".endif")
851 return ParseDirectiveEndIf(IDLoc);
853 // If we are in a ".if 0" block, ignore this statement.
854 if (TheCondState.Ignore) {
855 EatToEndOfStatement();
859 // FIXME: Recurse on local labels?
861 // See what kind of statement we have.
862 switch (Lexer.getKind()) {
863 case AsmToken::Colon: {
864 CheckForValidSection();
866 // identifier ':' -> Label.
869 // Diagnose attempt to use a variable as a label.
871 // FIXME: Diagnostics. Note the location of the definition as a label.
872 // FIXME: This doesn't diagnose assignment to a symbol which has been
873 // implicitly marked as external.
875 if (LocalLabelVal == -1)
876 Sym = getContext().GetOrCreateSymbol(IDVal);
878 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
879 if (!Sym->isUndefined() || Sym->isVariable())
880 return Error(IDLoc, "invalid symbol redefinition");
885 // Consume any end of statement token, if present, to avoid spurious
886 // AddBlankLine calls().
887 if (Lexer.is(AsmToken::EndOfStatement)) {
889 if (Lexer.is(AsmToken::Eof))
893 return ParseStatement();
896 case AsmToken::Equal:
897 // identifier '=' ... -> assignment statement
900 return ParseAssignment(IDVal);
902 default: // Normal instruction or directive.
906 // If macros are enabled, check to see if this is a macro instantiation.
908 if (const Macro *M = MacroMap.lookup(IDVal))
909 return HandleMacroEntry(IDVal, IDLoc, M);
911 // Otherwise, we have a normal instruction or directive.
912 if (IDVal[0] == '.') {
913 // Assembler features
915 return ParseDirectiveSet();
919 if (IDVal == ".ascii")
920 return ParseDirectiveAscii(false);
921 if (IDVal == ".asciz")
922 return ParseDirectiveAscii(true);
924 if (IDVal == ".byte")
925 return ParseDirectiveValue(1);
926 if (IDVal == ".short")
927 return ParseDirectiveValue(2);
928 if (IDVal == ".long")
929 return ParseDirectiveValue(4);
930 if (IDVal == ".quad")
931 return ParseDirectiveValue(8);
932 if (IDVal == ".single")
933 return ParseDirectiveRealValue(APFloat::IEEEsingle);
934 if (IDVal == ".double")
935 return ParseDirectiveRealValue(APFloat::IEEEdouble);
937 if (IDVal == ".align") {
938 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
939 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
941 if (IDVal == ".align32") {
942 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
943 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
945 if (IDVal == ".balign")
946 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
947 if (IDVal == ".balignw")
948 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
949 if (IDVal == ".balignl")
950 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
951 if (IDVal == ".p2align")
952 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
953 if (IDVal == ".p2alignw")
954 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
955 if (IDVal == ".p2alignl")
956 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
959 return ParseDirectiveOrg();
961 if (IDVal == ".fill")
962 return ParseDirectiveFill();
963 if (IDVal == ".space")
964 return ParseDirectiveSpace();
965 if (IDVal == ".zero")
966 return ParseDirectiveZero();
968 // Symbol attribute directives
970 if (IDVal == ".globl" || IDVal == ".global")
971 return ParseDirectiveSymbolAttribute(MCSA_Global);
972 // ELF only? Should it be here?
973 if (IDVal == ".local")
974 return ParseDirectiveSymbolAttribute(MCSA_Local);
975 if (IDVal == ".hidden")
976 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
977 if (IDVal == ".indirect_symbol")
978 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
979 if (IDVal == ".internal")
980 return ParseDirectiveSymbolAttribute(MCSA_Internal);
981 if (IDVal == ".lazy_reference")
982 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
983 if (IDVal == ".no_dead_strip")
984 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
985 if (IDVal == ".private_extern")
986 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
987 if (IDVal == ".protected")
988 return ParseDirectiveSymbolAttribute(MCSA_Protected);
989 if (IDVal == ".reference")
990 return ParseDirectiveSymbolAttribute(MCSA_Reference);
991 if (IDVal == ".weak")
992 return ParseDirectiveSymbolAttribute(MCSA_Weak);
993 if (IDVal == ".weak_definition")
994 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
995 if (IDVal == ".weak_reference")
996 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
997 if (IDVal == ".weak_def_can_be_hidden")
998 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1000 if (IDVal == ".comm")
1001 return ParseDirectiveComm(/*IsLocal=*/false);
1002 if (IDVal == ".lcomm")
1003 return ParseDirectiveComm(/*IsLocal=*/true);
1005 if (IDVal == ".abort")
1006 return ParseDirectiveAbort();
1007 if (IDVal == ".include")
1008 return ParseDirectiveInclude();
1010 // Look up the handler in the handler table.
1011 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1012 DirectiveMap.lookup(IDVal);
1014 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1016 // Target hook for parsing target specific directives.
1017 if (!getTargetParser().ParseDirective(ID))
1020 Warning(IDLoc, "ignoring directive for now");
1021 EatToEndOfStatement();
1025 CheckForValidSection();
1027 // Canonicalize the opcode to lower case.
1028 SmallString<128> Opcode;
1029 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1030 Opcode.push_back(tolower(IDVal[i]));
1032 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1033 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1036 // Dump the parsed representation, if requested.
1037 if (getShowParsedOperands()) {
1038 SmallString<256> Str;
1039 raw_svector_ostream OS(Str);
1040 OS << "parsed instruction: [";
1041 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1044 ParsedOperands[i]->dump(OS);
1048 PrintMessage(IDLoc, OS.str(), "note");
1051 // If parsing succeeded, match the instruction.
1053 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1056 // Free any parsed operands.
1057 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1058 delete ParsedOperands[i];
1060 // Don't skip the rest of the line, the instruction parser is responsible for
1065 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1066 const std::vector<std::vector<AsmToken> > &A)
1067 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1069 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1070 // to hold the macro body with substitutions.
1071 SmallString<256> Buf;
1072 raw_svector_ostream OS(Buf);
1074 StringRef Body = M->Body;
1075 while (!Body.empty()) {
1076 // Scan for the next substitution.
1077 std::size_t End = Body.size(), Pos = 0;
1078 for (; Pos != End; ++Pos) {
1079 // Check for a substitution or escape.
1080 if (Body[Pos] != '$' || Pos + 1 == End)
1083 char Next = Body[Pos + 1];
1084 if (Next == '$' || Next == 'n' || isdigit(Next))
1089 OS << Body.slice(0, Pos);
1091 // Check if we reached the end.
1095 switch (Body[Pos+1]) {
1101 // $n => number of arguments
1106 // $[0-9] => argument
1108 // Missing arguments are ignored.
1109 unsigned Index = Body[Pos+1] - '0';
1110 if (Index >= A.size())
1113 // Otherwise substitute with the token values, with spaces eliminated.
1114 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1115 ie = A[Index].end(); it != ie; ++it)
1116 OS << it->getString();
1121 // Update the scan point.
1122 Body = Body.substr(Pos + 2);
1125 // We include the .endmacro in the buffer as our queue to exit the macro
1127 OS << ".endmacro\n";
1129 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1132 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1134 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1135 // this, although we should protect against infinite loops.
1136 if (ActiveMacros.size() == 20)
1137 return TokError("macros cannot be nested more than 20 levels deep");
1139 // Parse the macro instantiation arguments.
1140 std::vector<std::vector<AsmToken> > MacroArguments;
1141 MacroArguments.push_back(std::vector<AsmToken>());
1142 unsigned ParenLevel = 0;
1144 if (Lexer.is(AsmToken::Eof))
1145 return TokError("unexpected token in macro instantiation");
1146 if (Lexer.is(AsmToken::EndOfStatement))
1149 // If we aren't inside parentheses and this is a comma, start a new token
1151 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1152 MacroArguments.push_back(std::vector<AsmToken>());
1154 // Adjust the current parentheses level.
1155 if (Lexer.is(AsmToken::LParen))
1157 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1160 // Append the token to the current argument list.
1161 MacroArguments.back().push_back(getTok());
1166 // Create the macro instantiation object and add to the current macro
1167 // instantiation stack.
1168 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1171 ActiveMacros.push_back(MI);
1173 // Jump to the macro instantiation and prime the lexer.
1174 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1175 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1181 void AsmParser::HandleMacroExit() {
1182 // Jump to the EndOfStatement we should return to, and consume it.
1183 JumpToLoc(ActiveMacros.back()->ExitLoc);
1186 // Pop the instantiation entry.
1187 delete ActiveMacros.back();
1188 ActiveMacros.pop_back();
1191 bool AsmParser::ParseAssignment(StringRef Name) {
1192 // FIXME: Use better location, we should use proper tokens.
1193 SMLoc EqualLoc = Lexer.getLoc();
1195 const MCExpr *Value;
1196 if (ParseExpression(Value))
1199 if (Lexer.isNot(AsmToken::EndOfStatement))
1200 return TokError("unexpected token in assignment");
1202 // Eat the end of statement marker.
1205 // Validate that the LHS is allowed to be a variable (either it has not been
1206 // used as a symbol, or it is an absolute symbol).
1207 MCSymbol *Sym = getContext().LookupSymbol(Name);
1209 // Diagnose assignment to a label.
1211 // FIXME: Diagnostics. Note the location of the definition as a label.
1212 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1213 if (Sym->isUndefined() && !Sym->isUsedInExpr())
1214 ; // Allow redefinitions of undefined symbols only used in directives.
1215 else if (!Sym->isUndefined() && !Sym->isAbsolute())
1216 return Error(EqualLoc, "redefinition of '" + Name + "'");
1217 else if (!Sym->isVariable())
1218 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1219 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1220 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1223 Sym = getContext().GetOrCreateSymbol(Name);
1225 // FIXME: Handle '.'.
1227 Sym->setUsedInExpr(true);
1229 // Do the assignment.
1230 Out.EmitAssignment(Sym, Value);
1235 /// ParseIdentifier:
1238 bool AsmParser::ParseIdentifier(StringRef &Res) {
1239 // The assembler has relaxed rules for accepting identifiers, in particular we
1240 // allow things like '.globl $foo', which would normally be separate
1241 // tokens. At this level, we have already lexed so we cannot (currently)
1242 // handle this as a context dependent token, instead we detect adjacent tokens
1243 // and return the combined identifier.
1244 if (Lexer.is(AsmToken::Dollar)) {
1245 SMLoc DollarLoc = getLexer().getLoc();
1247 // Consume the dollar sign, and check for a following identifier.
1249 if (Lexer.isNot(AsmToken::Identifier))
1252 // We have a '$' followed by an identifier, make sure they are adjacent.
1253 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1256 // Construct the joined identifier and consume the token.
1257 Res = StringRef(DollarLoc.getPointer(),
1258 getTok().getIdentifier().size() + 1);
1263 if (Lexer.isNot(AsmToken::Identifier) &&
1264 Lexer.isNot(AsmToken::String))
1267 Res = getTok().getIdentifier();
1269 Lex(); // Consume the identifier token.
1274 /// ParseDirectiveSet:
1275 /// ::= .set identifier ',' expression
1276 bool AsmParser::ParseDirectiveSet() {
1279 if (ParseIdentifier(Name))
1280 return TokError("expected identifier after '.set' directive");
1282 if (getLexer().isNot(AsmToken::Comma))
1283 return TokError("unexpected token in '.set'");
1286 return ParseAssignment(Name);
1289 bool AsmParser::ParseEscapedString(std::string &Data) {
1290 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1293 StringRef Str = getTok().getStringContents();
1294 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1295 if (Str[i] != '\\') {
1300 // Recognize escaped characters. Note that this escape semantics currently
1301 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1304 return TokError("unexpected backslash at end of string");
1306 // Recognize octal sequences.
1307 if ((unsigned) (Str[i] - '0') <= 7) {
1308 // Consume up to three octal characters.
1309 unsigned Value = Str[i] - '0';
1311 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1313 Value = Value * 8 + (Str[i] - '0');
1315 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1317 Value = Value * 8 + (Str[i] - '0');
1322 return TokError("invalid octal escape sequence (out of range)");
1324 Data += (unsigned char) Value;
1328 // Otherwise recognize individual escapes.
1331 // Just reject invalid escape sequences for now.
1332 return TokError("invalid escape sequence (unrecognized character)");
1334 case 'b': Data += '\b'; break;
1335 case 'f': Data += '\f'; break;
1336 case 'n': Data += '\n'; break;
1337 case 'r': Data += '\r'; break;
1338 case 't': Data += '\t'; break;
1339 case '"': Data += '"'; break;
1340 case '\\': Data += '\\'; break;
1347 /// ParseDirectiveAscii:
1348 /// ::= ( .ascii | .asciz ) [ "string" ( , "string" )* ]
1349 bool AsmParser::ParseDirectiveAscii(bool ZeroTerminated) {
1350 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1351 CheckForValidSection();
1354 if (getLexer().isNot(AsmToken::String))
1355 return TokError("expected string in '.ascii' or '.asciz' directive");
1358 if (ParseEscapedString(Data))
1361 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1363 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1367 if (getLexer().is(AsmToken::EndOfStatement))
1370 if (getLexer().isNot(AsmToken::Comma))
1371 return TokError("unexpected token in '.ascii' or '.asciz' directive");
1380 /// ParseDirectiveValue
1381 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1382 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1383 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1384 CheckForValidSection();
1387 const MCExpr *Value;
1388 if (ParseExpression(Value))
1391 // Special case constant expressions to match code generator.
1392 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1393 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1395 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1397 if (getLexer().is(AsmToken::EndOfStatement))
1400 // FIXME: Improve diagnostic.
1401 if (getLexer().isNot(AsmToken::Comma))
1402 return TokError("unexpected token in directive");
1411 /// ParseDirectiveRealValue
1412 /// ::= (.single | .double) [ expression (, expression)* ]
1413 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1414 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1415 CheckForValidSection();
1418 // We don't truly support arithmetic on floating point expressions, so we
1419 // have to manually parse unary prefixes.
1421 if (getLexer().is(AsmToken::Minus)) {
1424 } else if (getLexer().is(AsmToken::Plus))
1427 if (getLexer().isNot(AsmToken::Integer) &&
1428 getLexer().isNot(AsmToken::Real))
1429 return TokError("unexpected token in directive");
1431 // Convert to an APFloat.
1432 APFloat Value(Semantics);
1433 if (Value.convertFromString(getTok().getString(),
1434 APFloat::rmNearestTiesToEven) ==
1435 APFloat::opInvalidOp)
1436 return TokError("invalid floating point literal");
1440 // Consume the numeric token.
1443 // Emit the value as an integer.
1444 APInt AsInt = Value.bitcastToAPInt();
1445 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1446 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1448 if (getLexer().is(AsmToken::EndOfStatement))
1451 if (getLexer().isNot(AsmToken::Comma))
1452 return TokError("unexpected token in directive");
1461 /// ParseDirectiveSpace
1462 /// ::= .space expression [ , expression ]
1463 bool AsmParser::ParseDirectiveSpace() {
1464 CheckForValidSection();
1467 if (ParseAbsoluteExpression(NumBytes))
1470 int64_t FillExpr = 0;
1471 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1472 if (getLexer().isNot(AsmToken::Comma))
1473 return TokError("unexpected token in '.space' directive");
1476 if (ParseAbsoluteExpression(FillExpr))
1479 if (getLexer().isNot(AsmToken::EndOfStatement))
1480 return TokError("unexpected token in '.space' directive");
1486 return TokError("invalid number of bytes in '.space' directive");
1488 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1489 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1494 /// ParseDirectiveZero
1495 /// ::= .zero expression
1496 bool AsmParser::ParseDirectiveZero() {
1497 CheckForValidSection();
1500 if (ParseAbsoluteExpression(NumBytes))
1504 if (getLexer().is(AsmToken::Comma)) {
1506 if (ParseAbsoluteExpression(Val))
1510 if (getLexer().isNot(AsmToken::EndOfStatement))
1511 return TokError("unexpected token in '.zero' directive");
1515 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
1520 /// ParseDirectiveFill
1521 /// ::= .fill expression , expression , expression
1522 bool AsmParser::ParseDirectiveFill() {
1523 CheckForValidSection();
1526 if (ParseAbsoluteExpression(NumValues))
1529 if (getLexer().isNot(AsmToken::Comma))
1530 return TokError("unexpected token in '.fill' directive");
1534 if (ParseAbsoluteExpression(FillSize))
1537 if (getLexer().isNot(AsmToken::Comma))
1538 return TokError("unexpected token in '.fill' directive");
1542 if (ParseAbsoluteExpression(FillExpr))
1545 if (getLexer().isNot(AsmToken::EndOfStatement))
1546 return TokError("unexpected token in '.fill' directive");
1550 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1551 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1553 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1554 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1559 /// ParseDirectiveOrg
1560 /// ::= .org expression [ , expression ]
1561 bool AsmParser::ParseDirectiveOrg() {
1562 CheckForValidSection();
1564 const MCExpr *Offset;
1565 if (ParseExpression(Offset))
1568 // Parse optional fill expression.
1569 int64_t FillExpr = 0;
1570 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1571 if (getLexer().isNot(AsmToken::Comma))
1572 return TokError("unexpected token in '.org' directive");
1575 if (ParseAbsoluteExpression(FillExpr))
1578 if (getLexer().isNot(AsmToken::EndOfStatement))
1579 return TokError("unexpected token in '.org' directive");
1584 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1585 // has to be relative to the current section.
1586 getStreamer().EmitValueToOffset(Offset, FillExpr);
1591 /// ParseDirectiveAlign
1592 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1593 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1594 CheckForValidSection();
1596 SMLoc AlignmentLoc = getLexer().getLoc();
1598 if (ParseAbsoluteExpression(Alignment))
1602 bool HasFillExpr = false;
1603 int64_t FillExpr = 0;
1604 int64_t MaxBytesToFill = 0;
1605 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1606 if (getLexer().isNot(AsmToken::Comma))
1607 return TokError("unexpected token in directive");
1610 // The fill expression can be omitted while specifying a maximum number of
1611 // alignment bytes, e.g:
1613 if (getLexer().isNot(AsmToken::Comma)) {
1615 if (ParseAbsoluteExpression(FillExpr))
1619 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1620 if (getLexer().isNot(AsmToken::Comma))
1621 return TokError("unexpected token in directive");
1624 MaxBytesLoc = getLexer().getLoc();
1625 if (ParseAbsoluteExpression(MaxBytesToFill))
1628 if (getLexer().isNot(AsmToken::EndOfStatement))
1629 return TokError("unexpected token in directive");
1638 // Compute alignment in bytes.
1640 // FIXME: Diagnose overflow.
1641 if (Alignment >= 32) {
1642 Error(AlignmentLoc, "invalid alignment value");
1646 Alignment = 1ULL << Alignment;
1649 // Diagnose non-sensical max bytes to align.
1650 if (MaxBytesLoc.isValid()) {
1651 if (MaxBytesToFill < 1) {
1652 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1653 "many bytes, ignoring maximum bytes expression");
1657 if (MaxBytesToFill >= Alignment) {
1658 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1664 // Check whether we should use optimal code alignment for this .align
1666 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
1667 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1668 ValueSize == 1 && UseCodeAlign) {
1669 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1671 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1672 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1679 /// ParseDirectiveSymbolAttribute
1680 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1681 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1682 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1686 if (ParseIdentifier(Name))
1687 return TokError("expected identifier in directive");
1689 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1691 getStreamer().EmitSymbolAttribute(Sym, Attr);
1693 if (getLexer().is(AsmToken::EndOfStatement))
1696 if (getLexer().isNot(AsmToken::Comma))
1697 return TokError("unexpected token in directive");
1706 /// ParseDirectiveComm
1707 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1708 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1709 CheckForValidSection();
1711 SMLoc IDLoc = getLexer().getLoc();
1713 if (ParseIdentifier(Name))
1714 return TokError("expected identifier in directive");
1716 // Handle the identifier as the key symbol.
1717 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1719 if (getLexer().isNot(AsmToken::Comma))
1720 return TokError("unexpected token in directive");
1724 SMLoc SizeLoc = getLexer().getLoc();
1725 if (ParseAbsoluteExpression(Size))
1728 int64_t Pow2Alignment = 0;
1729 SMLoc Pow2AlignmentLoc;
1730 if (getLexer().is(AsmToken::Comma)) {
1732 Pow2AlignmentLoc = getLexer().getLoc();
1733 if (ParseAbsoluteExpression(Pow2Alignment))
1736 // If this target takes alignments in bytes (not log) validate and convert.
1737 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1738 if (!isPowerOf2_64(Pow2Alignment))
1739 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1740 Pow2Alignment = Log2_64(Pow2Alignment);
1744 if (getLexer().isNot(AsmToken::EndOfStatement))
1745 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1749 // NOTE: a size of zero for a .comm should create a undefined symbol
1750 // but a size of .lcomm creates a bss symbol of size zero.
1752 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1753 "be less than zero");
1755 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1756 // may internally end up wanting an alignment in bytes.
1757 // FIXME: Diagnose overflow.
1758 if (Pow2Alignment < 0)
1759 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1760 "alignment, can't be less than zero");
1762 if (!Sym->isUndefined())
1763 return Error(IDLoc, "invalid symbol redefinition");
1765 // '.lcomm' is equivalent to '.zerofill'.
1766 // Create the Symbol as a common or local common with Size and Pow2Alignment
1768 getStreamer().EmitZerofill(Ctx.getMachOSection(
1769 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1770 0, SectionKind::getBSS()),
1771 Sym, Size, 1 << Pow2Alignment);
1775 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1779 /// ParseDirectiveAbort
1780 /// ::= .abort [... message ...]
1781 bool AsmParser::ParseDirectiveAbort() {
1782 // FIXME: Use loc from directive.
1783 SMLoc Loc = getLexer().getLoc();
1785 StringRef Str = ParseStringToEndOfStatement();
1786 if (getLexer().isNot(AsmToken::EndOfStatement))
1787 return TokError("unexpected token in '.abort' directive");
1792 Error(Loc, ".abort detected. Assembly stopping.");
1794 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1795 // FIXME: Actually abort assembly here.
1800 /// ParseDirectiveInclude
1801 /// ::= .include "filename"
1802 bool AsmParser::ParseDirectiveInclude() {
1803 if (getLexer().isNot(AsmToken::String))
1804 return TokError("expected string in '.include' directive");
1806 std::string Filename = getTok().getString();
1807 SMLoc IncludeLoc = getLexer().getLoc();
1810 if (getLexer().isNot(AsmToken::EndOfStatement))
1811 return TokError("unexpected token in '.include' directive");
1813 // Strip the quotes.
1814 Filename = Filename.substr(1, Filename.size()-2);
1816 // Attempt to switch the lexer to the included file before consuming the end
1817 // of statement to avoid losing it when we switch.
1818 if (EnterIncludeFile(Filename)) {
1819 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1826 /// ParseDirectiveIf
1827 /// ::= .if expression
1828 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1829 TheCondStack.push_back(TheCondState);
1830 TheCondState.TheCond = AsmCond::IfCond;
1831 if(TheCondState.Ignore) {
1832 EatToEndOfStatement();
1836 if (ParseAbsoluteExpression(ExprValue))
1839 if (getLexer().isNot(AsmToken::EndOfStatement))
1840 return TokError("unexpected token in '.if' directive");
1844 TheCondState.CondMet = ExprValue;
1845 TheCondState.Ignore = !TheCondState.CondMet;
1851 /// ParseDirectiveElseIf
1852 /// ::= .elseif expression
1853 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1854 if (TheCondState.TheCond != AsmCond::IfCond &&
1855 TheCondState.TheCond != AsmCond::ElseIfCond)
1856 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1858 TheCondState.TheCond = AsmCond::ElseIfCond;
1860 bool LastIgnoreState = false;
1861 if (!TheCondStack.empty())
1862 LastIgnoreState = TheCondStack.back().Ignore;
1863 if (LastIgnoreState || TheCondState.CondMet) {
1864 TheCondState.Ignore = true;
1865 EatToEndOfStatement();
1869 if (ParseAbsoluteExpression(ExprValue))
1872 if (getLexer().isNot(AsmToken::EndOfStatement))
1873 return TokError("unexpected token in '.elseif' directive");
1876 TheCondState.CondMet = ExprValue;
1877 TheCondState.Ignore = !TheCondState.CondMet;
1883 /// ParseDirectiveElse
1885 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
1886 if (getLexer().isNot(AsmToken::EndOfStatement))
1887 return TokError("unexpected token in '.else' directive");
1891 if (TheCondState.TheCond != AsmCond::IfCond &&
1892 TheCondState.TheCond != AsmCond::ElseIfCond)
1893 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
1895 TheCondState.TheCond = AsmCond::ElseCond;
1896 bool LastIgnoreState = false;
1897 if (!TheCondStack.empty())
1898 LastIgnoreState = TheCondStack.back().Ignore;
1899 if (LastIgnoreState || TheCondState.CondMet)
1900 TheCondState.Ignore = true;
1902 TheCondState.Ignore = false;
1907 /// ParseDirectiveEndIf
1909 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
1910 if (getLexer().isNot(AsmToken::EndOfStatement))
1911 return TokError("unexpected token in '.endif' directive");
1915 if ((TheCondState.TheCond == AsmCond::NoCond) ||
1916 TheCondStack.empty())
1917 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
1919 if (!TheCondStack.empty()) {
1920 TheCondState = TheCondStack.back();
1921 TheCondStack.pop_back();
1927 /// ParseDirectiveFile
1928 /// ::= .file [number] string
1929 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
1930 // FIXME: I'm not sure what this is.
1931 int64_t FileNumber = -1;
1932 SMLoc FileNumberLoc = getLexer().getLoc();
1933 if (getLexer().is(AsmToken::Integer)) {
1934 FileNumber = getTok().getIntVal();
1938 return TokError("file number less than one");
1941 if (getLexer().isNot(AsmToken::String))
1942 return TokError("unexpected token in '.file' directive");
1944 StringRef Filename = getTok().getString();
1945 Filename = Filename.substr(1, Filename.size()-2);
1948 if (getLexer().isNot(AsmToken::EndOfStatement))
1949 return TokError("unexpected token in '.file' directive");
1951 if (FileNumber == -1)
1952 getStreamer().EmitFileDirective(Filename);
1954 if (getContext().GetDwarfFile(Filename, FileNumber) == 0)
1955 Error(FileNumberLoc, "file number already allocated");
1956 getStreamer().EmitDwarfFileDirective(FileNumber, Filename);
1962 /// ParseDirectiveLine
1963 /// ::= .line [number]
1964 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
1965 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1966 if (getLexer().isNot(AsmToken::Integer))
1967 return TokError("unexpected token in '.line' directive");
1969 int64_t LineNumber = getTok().getIntVal();
1973 // FIXME: Do something with the .line.
1976 if (getLexer().isNot(AsmToken::EndOfStatement))
1977 return TokError("unexpected token in '.line' directive");
1983 /// ParseDirectiveLoc
1984 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
1985 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
1986 /// The first number is a file number, must have been previously assigned with
1987 /// a .file directive, the second number is the line number and optionally the
1988 /// third number is a column position (zero if not specified). The remaining
1989 /// optional items are .loc sub-directives.
1990 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
1992 if (getLexer().isNot(AsmToken::Integer))
1993 return TokError("unexpected token in '.loc' directive");
1994 int64_t FileNumber = getTok().getIntVal();
1996 return TokError("file number less than one in '.loc' directive");
1997 if (!getContext().isValidDwarfFileNumber(FileNumber))
1998 return TokError("unassigned file number in '.loc' directive");
2001 int64_t LineNumber = 0;
2002 if (getLexer().is(AsmToken::Integer)) {
2003 LineNumber = getTok().getIntVal();
2005 return TokError("line number less than one in '.loc' directive");
2009 int64_t ColumnPos = 0;
2010 if (getLexer().is(AsmToken::Integer)) {
2011 ColumnPos = getTok().getIntVal();
2013 return TokError("column position less than zero in '.loc' directive");
2017 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2019 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2021 if (getLexer().is(AsmToken::EndOfStatement))
2025 SMLoc Loc = getTok().getLoc();
2026 if (getParser().ParseIdentifier(Name))
2027 return TokError("unexpected token in '.loc' directive");
2029 if (Name == "basic_block")
2030 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2031 else if (Name == "prologue_end")
2032 Flags |= DWARF2_FLAG_PROLOGUE_END;
2033 else if (Name == "epilogue_begin")
2034 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2035 else if (Name == "is_stmt") {
2036 SMLoc Loc = getTok().getLoc();
2037 const MCExpr *Value;
2038 if (getParser().ParseExpression(Value))
2040 // The expression must be the constant 0 or 1.
2041 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2042 int Value = MCE->getValue();
2044 Flags &= ~DWARF2_FLAG_IS_STMT;
2045 else if (Value == 1)
2046 Flags |= DWARF2_FLAG_IS_STMT;
2048 return Error(Loc, "is_stmt value not 0 or 1");
2051 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2054 else if (Name == "isa") {
2055 SMLoc Loc = getTok().getLoc();
2056 const MCExpr *Value;
2057 if (getParser().ParseExpression(Value))
2059 // The expression must be a constant greater or equal to 0.
2060 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2061 int Value = MCE->getValue();
2063 return Error(Loc, "isa number less than zero");
2067 return Error(Loc, "isa number not a constant value");
2071 return Error(Loc, "unknown sub-directive in '.loc' directive");
2074 if (getLexer().is(AsmToken::EndOfStatement))
2079 getContext().setCurrentDwarfLoc(FileNumber, LineNumber, ColumnPos, Flags,Isa);
2084 /// ParseDirectiveStabs
2085 /// ::= .stabs string, number, number, number
2086 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2087 SMLoc DirectiveLoc) {
2088 return TokError("unsupported directive '" + Directive + "'");
2091 /// ParseDirectiveMacrosOnOff
2094 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2095 SMLoc DirectiveLoc) {
2096 if (getLexer().isNot(AsmToken::EndOfStatement))
2097 return Error(getLexer().getLoc(),
2098 "unexpected token in '" + Directive + "' directive");
2100 getParser().MacrosEnabled = Directive == ".macros_on";
2105 /// ParseDirectiveMacro
2107 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2108 SMLoc DirectiveLoc) {
2110 if (getParser().ParseIdentifier(Name))
2111 return TokError("expected identifier in directive");
2113 if (getLexer().isNot(AsmToken::EndOfStatement))
2114 return TokError("unexpected token in '.macro' directive");
2116 // Eat the end of statement.
2119 AsmToken EndToken, StartToken = getTok();
2121 // Lex the macro definition.
2123 // Check whether we have reached the end of the file.
2124 if (getLexer().is(AsmToken::Eof))
2125 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2127 // Otherwise, check whether we have reach the .endmacro.
2128 if (getLexer().is(AsmToken::Identifier) &&
2129 (getTok().getIdentifier() == ".endm" ||
2130 getTok().getIdentifier() == ".endmacro")) {
2131 EndToken = getTok();
2133 if (getLexer().isNot(AsmToken::EndOfStatement))
2134 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2139 // Otherwise, scan til the end of the statement.
2140 getParser().EatToEndOfStatement();
2143 if (getParser().MacroMap.lookup(Name)) {
2144 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2147 const char *BodyStart = StartToken.getLoc().getPointer();
2148 const char *BodyEnd = EndToken.getLoc().getPointer();
2149 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2150 getParser().MacroMap[Name] = new Macro(Name, Body);
2154 /// ParseDirectiveEndMacro
2157 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2158 SMLoc DirectiveLoc) {
2159 if (getLexer().isNot(AsmToken::EndOfStatement))
2160 return TokError("unexpected token in '" + Directive + "' directive");
2162 // If we are inside a macro instantiation, terminate the current
2164 if (!getParser().ActiveMacros.empty()) {
2165 getParser().HandleMacroExit();
2169 // Otherwise, this .endmacro is a stray entry in the file; well formed
2170 // .endmacro directives are handled during the macro definition parsing.
2171 return TokError("unexpected '" + Directive + "' in file, "
2172 "no current macro definition");
2175 void GenericAsmParser::ParseUleb128(uint64_t Value) {
2176 const uint64_t Mask = (1 << 7) - 1;
2178 unsigned Byte = Value & Mask;
2180 if (Value) // Not the last one
2182 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2186 void GenericAsmParser::ParseSleb128(int64_t Value) {
2187 const int64_t Mask = (1 << 7) - 1;
2189 unsigned Byte = Value & Mask;
2191 bool Done = ((Value == 0 && (Byte & 0x40) == 0) ||
2192 (Value == -1 && (Byte & 0x40) != 0));
2195 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2201 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2203 if (getParser().ParseAbsoluteExpression(Value))
2206 if (getLexer().isNot(AsmToken::EndOfStatement))
2207 return TokError("unexpected token in directive");
2209 if (DirName[1] == 's')
2210 ParseSleb128(Value);
2212 ParseUleb128(Value);
2217 /// \brief Create an MCAsmParser instance.
2218 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2219 MCContext &C, MCStreamer &Out,
2220 const MCAsmInfo &MAI) {
2221 return new AsmParser(T, SM, C, Out, MAI);