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.
181 // ".ascii", ".asciiz", ".string"
182 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
183 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
184 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
185 bool ParseDirectiveFill(); // ".fill"
186 bool ParseDirectiveSpace(); // ".space"
187 bool ParseDirectiveZero(); // ".zero"
188 bool ParseDirectiveSet(StringRef IDVal); // ".set" or ".equ"
189 bool ParseDirectiveOrg(); // ".org"
190 // ".align{,32}", ".p2align{,w,l}"
191 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
193 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
194 /// accepts a single symbol (which should be a label or an external).
195 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
197 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
199 bool ParseDirectiveAbort(); // ".abort"
200 bool ParseDirectiveInclude(); // ".include"
202 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
203 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
204 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
205 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
207 /// ParseEscapedString - Parse the current token as a string which may include
208 /// escaped characters and return the string contents.
209 bool ParseEscapedString(std::string &Data);
211 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
212 MCSymbolRefExpr::VariantKind Variant);
215 /// \brief Generic implementations of directive handling, etc. which is shared
216 /// (or the default, at least) for all assembler parser.
217 class GenericAsmParser : public MCAsmParserExtension {
218 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
219 void AddDirectiveHandler(StringRef Directive) {
220 getParser().AddDirectiveHandler(this, Directive,
221 HandleDirective<GenericAsmParser, Handler>);
225 GenericAsmParser() {}
227 AsmParser &getParser() {
228 return (AsmParser&) this->MCAsmParserExtension::getParser();
231 virtual void Initialize(MCAsmParser &Parser) {
232 // Call the base implementation.
233 this->MCAsmParserExtension::Initialize(Parser);
235 // Debugging directives.
236 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
237 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
238 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
239 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
242 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
244 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
246 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
247 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
248 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
250 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
251 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
254 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
255 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
256 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
257 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc);
259 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
260 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
261 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
263 bool ParseDirectiveLEB128(StringRef, SMLoc);
270 extern MCAsmParserExtension *createDarwinAsmParser();
271 extern MCAsmParserExtension *createELFAsmParser();
272 extern MCAsmParserExtension *createCOFFAsmParser();
276 enum { DEFAULT_ADDRSPACE = 0 };
278 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
279 MCStreamer &_Out, const MCAsmInfo &_MAI)
280 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
281 GenericParser(new GenericAsmParser), PlatformParser(0),
282 CurBuffer(0), MacrosEnabled(true) {
283 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
285 // Initialize the generic parser.
286 GenericParser->Initialize(*this);
288 // Initialize the platform / file format parser.
290 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
292 if (_MAI.hasMicrosoftFastStdCallMangling()) {
293 PlatformParser = createCOFFAsmParser();
294 PlatformParser->Initialize(*this);
295 } else if (_MAI.hasSubsectionsViaSymbols()) {
296 PlatformParser = createDarwinAsmParser();
297 PlatformParser->Initialize(*this);
299 PlatformParser = createELFAsmParser();
300 PlatformParser->Initialize(*this);
304 AsmParser::~AsmParser() {
305 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
307 // Destroy any macros.
308 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
309 ie = MacroMap.end(); it != ie; ++it)
310 delete it->getValue();
312 delete PlatformParser;
313 delete GenericParser;
316 void AsmParser::PrintMacroInstantiations() {
317 // Print the active macro instantiation stack.
318 for (std::vector<MacroInstantiation*>::const_reverse_iterator
319 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
320 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
324 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
325 PrintMessage(L, Msg, "warning");
326 PrintMacroInstantiations();
329 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
331 PrintMessage(L, Msg, "error");
332 PrintMacroInstantiations();
336 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
337 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
343 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
348 void AsmParser::JumpToLoc(SMLoc Loc) {
349 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
350 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
353 const AsmToken &AsmParser::Lex() {
354 const AsmToken *tok = &Lexer.Lex();
356 if (tok->is(AsmToken::Eof)) {
357 // If this is the end of an included file, pop the parent file off the
359 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
360 if (ParentIncludeLoc != SMLoc()) {
361 JumpToLoc(ParentIncludeLoc);
366 if (tok->is(AsmToken::Error))
367 Error(Lexer.getErrLoc(), Lexer.getErr());
372 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
373 // Create the initial section, if requested.
374 if (!NoInitialTextSection)
381 AsmCond StartingCondState = TheCondState;
383 // While we have input, parse each statement.
384 while (Lexer.isNot(AsmToken::Eof)) {
385 if (!ParseStatement()) continue;
387 // We had an error, validate that one was emitted and recover by skipping to
389 assert(HadError && "Parse statement returned an error, but none emitted!");
390 EatToEndOfStatement();
393 if (TheCondState.TheCond != StartingCondState.TheCond ||
394 TheCondState.Ignore != StartingCondState.Ignore)
395 return TokError("unmatched .ifs or .elses");
397 // Check to see there are no empty DwarfFile slots.
398 const std::vector<MCDwarfFile *> &MCDwarfFiles =
399 getContext().getMCDwarfFiles();
400 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
401 if (!MCDwarfFiles[i])
402 TokError("unassigned file number: " + Twine(i) + " for .file directives");
405 // Finalize the output stream if there are no errors and if the client wants
407 if (!HadError && !NoFinalize)
413 void AsmParser::CheckForValidSection() {
414 if (!getStreamer().getCurrentSection()) {
415 TokError("expected section directive before assembly directive");
416 Out.SwitchSection(Ctx.getMachOSection(
418 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
419 0, SectionKind::getText()));
423 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
424 void AsmParser::EatToEndOfStatement() {
425 while (Lexer.isNot(AsmToken::EndOfStatement) &&
426 Lexer.isNot(AsmToken::Eof))
430 if (Lexer.is(AsmToken::EndOfStatement))
434 StringRef AsmParser::ParseStringToEndOfStatement() {
435 const char *Start = getTok().getLoc().getPointer();
437 while (Lexer.isNot(AsmToken::EndOfStatement) &&
438 Lexer.isNot(AsmToken::Eof))
441 const char *End = getTok().getLoc().getPointer();
442 return StringRef(Start, End - Start);
445 /// ParseParenExpr - Parse a paren expression and return it.
446 /// NOTE: This assumes the leading '(' has already been consumed.
448 /// parenexpr ::= expr)
450 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
451 if (ParseExpression(Res)) return true;
452 if (Lexer.isNot(AsmToken::RParen))
453 return TokError("expected ')' in parentheses expression");
454 EndLoc = Lexer.getLoc();
459 /// ParsePrimaryExpr - Parse a primary expression and return it.
460 /// primaryexpr ::= (parenexpr
461 /// primaryexpr ::= symbol
462 /// primaryexpr ::= number
463 /// primaryexpr ::= '.'
464 /// primaryexpr ::= ~,+,- primaryexpr
465 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
466 switch (Lexer.getKind()) {
468 return TokError("unknown token in expression");
469 case AsmToken::Exclaim:
470 Lex(); // Eat the operator.
471 if (ParsePrimaryExpr(Res, EndLoc))
473 Res = MCUnaryExpr::CreateLNot(Res, getContext());
475 case AsmToken::Dollar:
476 case AsmToken::String:
477 case AsmToken::Identifier: {
478 EndLoc = Lexer.getLoc();
480 StringRef Identifier;
481 if (ParseIdentifier(Identifier))
484 // This is a symbol reference.
485 std::pair<StringRef, StringRef> Split = Identifier.split('@');
486 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
488 // Mark the symbol as used in an expression.
489 Sym->setUsedInExpr(true);
491 // Lookup the symbol variant if used.
492 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
493 if (Split.first.size() != Identifier.size()) {
494 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
495 if (Variant == MCSymbolRefExpr::VK_Invalid) {
496 Variant = MCSymbolRefExpr::VK_None;
497 TokError("invalid variant '" + Split.second + "'");
501 // If this is an absolute variable reference, substitute it now to preserve
502 // semantics in the face of reassignment.
503 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
505 return Error(EndLoc, "unexpected modifier on variable reference");
507 Res = Sym->getVariableValue();
511 // Otherwise create a symbol ref.
512 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
515 case AsmToken::Integer: {
516 SMLoc Loc = getTok().getLoc();
517 int64_t IntVal = getTok().getIntVal();
518 Res = MCConstantExpr::Create(IntVal, getContext());
519 EndLoc = Lexer.getLoc();
521 // Look for 'b' or 'f' following an Integer as a directional label
522 if (Lexer.getKind() == AsmToken::Identifier) {
523 StringRef IDVal = getTok().getString();
524 if (IDVal == "f" || IDVal == "b"){
525 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
526 IDVal == "f" ? 1 : 0);
527 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
529 if(IDVal == "b" && Sym->isUndefined())
530 return Error(Loc, "invalid reference to undefined symbol");
531 EndLoc = Lexer.getLoc();
532 Lex(); // Eat identifier.
537 case AsmToken::Dot: {
538 // This is a '.' reference, which references the current PC. Emit a
539 // temporary label to the streamer and refer to it.
540 MCSymbol *Sym = Ctx.CreateTempSymbol();
542 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
543 EndLoc = Lexer.getLoc();
544 Lex(); // Eat identifier.
548 case AsmToken::LParen:
549 Lex(); // Eat the '('.
550 return ParseParenExpr(Res, EndLoc);
551 case AsmToken::Minus:
552 Lex(); // Eat the operator.
553 if (ParsePrimaryExpr(Res, EndLoc))
555 Res = MCUnaryExpr::CreateMinus(Res, getContext());
558 Lex(); // Eat the operator.
559 if (ParsePrimaryExpr(Res, EndLoc))
561 Res = MCUnaryExpr::CreatePlus(Res, getContext());
563 case AsmToken::Tilde:
564 Lex(); // Eat the operator.
565 if (ParsePrimaryExpr(Res, EndLoc))
567 Res = MCUnaryExpr::CreateNot(Res, getContext());
572 bool AsmParser::ParseExpression(const MCExpr *&Res) {
574 return ParseExpression(Res, EndLoc);
578 AsmParser::ApplyModifierToExpr(const MCExpr *E,
579 MCSymbolRefExpr::VariantKind Variant) {
580 // Recurse over the given expression, rebuilding it to apply the given variant
581 // if there is exactly one symbol.
582 switch (E->getKind()) {
584 case MCExpr::Constant:
587 case MCExpr::SymbolRef: {
588 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
590 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
591 TokError("invalid variant on expression '" +
592 getTok().getIdentifier() + "' (already modified)");
596 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
599 case MCExpr::Unary: {
600 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
601 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
604 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
607 case MCExpr::Binary: {
608 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
609 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
610 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
615 if (!LHS) LHS = BE->getLHS();
616 if (!RHS) RHS = BE->getRHS();
618 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
622 assert(0 && "Invalid expression kind!");
626 /// ParseExpression - Parse an expression and return it.
628 /// expr ::= expr +,- expr -> lowest.
629 /// expr ::= expr |,^,&,! expr -> middle.
630 /// expr ::= expr *,/,%,<<,>> expr -> highest.
631 /// expr ::= primaryexpr
633 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
634 // Parse the expression.
636 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
639 // As a special case, we support 'a op b @ modifier' by rewriting the
640 // expression to include the modifier. This is inefficient, but in general we
641 // expect users to use 'a@modifier op b'.
642 if (Lexer.getKind() == AsmToken::At) {
645 if (Lexer.isNot(AsmToken::Identifier))
646 return TokError("unexpected symbol modifier following '@'");
648 MCSymbolRefExpr::VariantKind Variant =
649 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
650 if (Variant == MCSymbolRefExpr::VK_Invalid)
651 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
653 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
655 return TokError("invalid modifier '" + getTok().getIdentifier() +
656 "' (no symbols present)");
664 // Try to constant fold it up front, if possible.
666 if (Res->EvaluateAsAbsolute(Value))
667 Res = MCConstantExpr::Create(Value, getContext());
672 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
674 return ParseParenExpr(Res, EndLoc) ||
675 ParseBinOpRHS(1, Res, EndLoc);
678 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
681 SMLoc StartLoc = Lexer.getLoc();
682 if (ParseExpression(Expr))
685 if (!Expr->EvaluateAsAbsolute(Res))
686 return Error(StartLoc, "expected absolute expression");
691 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
692 MCBinaryExpr::Opcode &Kind) {
695 return 0; // not a binop.
697 // Lowest Precedence: &&, ||, @
698 case AsmToken::AmpAmp:
699 Kind = MCBinaryExpr::LAnd;
701 case AsmToken::PipePipe:
702 Kind = MCBinaryExpr::LOr;
706 // Low Precedence: |, &, ^
708 // FIXME: gas seems to support '!' as an infix operator?
710 Kind = MCBinaryExpr::Or;
712 case AsmToken::Caret:
713 Kind = MCBinaryExpr::Xor;
716 Kind = MCBinaryExpr::And;
719 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
720 case AsmToken::EqualEqual:
721 Kind = MCBinaryExpr::EQ;
723 case AsmToken::ExclaimEqual:
724 case AsmToken::LessGreater:
725 Kind = MCBinaryExpr::NE;
728 Kind = MCBinaryExpr::LT;
730 case AsmToken::LessEqual:
731 Kind = MCBinaryExpr::LTE;
733 case AsmToken::Greater:
734 Kind = MCBinaryExpr::GT;
736 case AsmToken::GreaterEqual:
737 Kind = MCBinaryExpr::GTE;
740 // High Intermediate Precedence: +, -
742 Kind = MCBinaryExpr::Add;
744 case AsmToken::Minus:
745 Kind = MCBinaryExpr::Sub;
748 // Highest Precedence: *, /, %, <<, >>
750 Kind = MCBinaryExpr::Mul;
752 case AsmToken::Slash:
753 Kind = MCBinaryExpr::Div;
755 case AsmToken::Percent:
756 Kind = MCBinaryExpr::Mod;
758 case AsmToken::LessLess:
759 Kind = MCBinaryExpr::Shl;
761 case AsmToken::GreaterGreater:
762 Kind = MCBinaryExpr::Shr;
768 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
769 /// Res contains the LHS of the expression on input.
770 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
773 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
774 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
776 // If the next token is lower precedence than we are allowed to eat, return
777 // successfully with what we ate already.
778 if (TokPrec < Precedence)
783 // Eat the next primary expression.
785 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
787 // If BinOp binds less tightly with RHS than the operator after RHS, let
788 // the pending operator take RHS as its LHS.
789 MCBinaryExpr::Opcode Dummy;
790 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
791 if (TokPrec < NextTokPrec) {
792 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
795 // Merge LHS and RHS according to operator.
796 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
804 /// ::= EndOfStatement
805 /// ::= Label* Directive ...Operands... EndOfStatement
806 /// ::= Label* Identifier OperandList* EndOfStatement
807 bool AsmParser::ParseStatement() {
808 if (Lexer.is(AsmToken::EndOfStatement)) {
814 // Statements always start with an identifier.
815 AsmToken ID = getTok();
816 SMLoc IDLoc = ID.getLoc();
818 int64_t LocalLabelVal = -1;
819 // GUESS allow an integer followed by a ':' as a directional local label
820 if (Lexer.is(AsmToken::Integer)) {
821 LocalLabelVal = getTok().getIntVal();
822 if (LocalLabelVal < 0) {
823 if (!TheCondState.Ignore)
824 return TokError("unexpected token at start of statement");
828 IDVal = getTok().getString();
829 Lex(); // Consume the integer token to be used as an identifier token.
830 if (Lexer.getKind() != AsmToken::Colon) {
831 if (!TheCondState.Ignore)
832 return TokError("unexpected token at start of statement");
836 else if (ParseIdentifier(IDVal)) {
837 if (!TheCondState.Ignore)
838 return TokError("unexpected token at start of statement");
842 // Handle conditional assembly here before checking for skipping. We
843 // have to do this so that .endif isn't skipped in a ".if 0" block for
846 return ParseDirectiveIf(IDLoc);
847 if (IDVal == ".elseif")
848 return ParseDirectiveElseIf(IDLoc);
849 if (IDVal == ".else")
850 return ParseDirectiveElse(IDLoc);
851 if (IDVal == ".endif")
852 return ParseDirectiveEndIf(IDLoc);
854 // If we are in a ".if 0" block, ignore this statement.
855 if (TheCondState.Ignore) {
856 EatToEndOfStatement();
860 // FIXME: Recurse on local labels?
862 // See what kind of statement we have.
863 switch (Lexer.getKind()) {
864 case AsmToken::Colon: {
865 CheckForValidSection();
867 // identifier ':' -> Label.
870 // Diagnose attempt to use a variable as a label.
872 // FIXME: Diagnostics. Note the location of the definition as a label.
873 // FIXME: This doesn't diagnose assignment to a symbol which has been
874 // implicitly marked as external.
876 if (LocalLabelVal == -1)
877 Sym = getContext().GetOrCreateSymbol(IDVal);
879 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
880 if (!Sym->isUndefined() || Sym->isVariable())
881 return Error(IDLoc, "invalid symbol redefinition");
886 // Consume any end of statement token, if present, to avoid spurious
887 // AddBlankLine calls().
888 if (Lexer.is(AsmToken::EndOfStatement)) {
890 if (Lexer.is(AsmToken::Eof))
894 return ParseStatement();
897 case AsmToken::Equal:
898 // identifier '=' ... -> assignment statement
901 return ParseAssignment(IDVal);
903 default: // Normal instruction or directive.
907 // If macros are enabled, check to see if this is a macro instantiation.
909 if (const Macro *M = MacroMap.lookup(IDVal))
910 return HandleMacroEntry(IDVal, IDLoc, M);
912 // Otherwise, we have a normal instruction or directive.
913 if (IDVal[0] == '.') {
914 // Assembler features
915 if (IDVal == ".set" || IDVal == ".equ")
916 return ParseDirectiveSet(IDVal);
920 if (IDVal == ".ascii")
921 return ParseDirectiveAscii(IDVal, false);
922 if (IDVal == ".asciz" || IDVal == ".string")
923 return ParseDirectiveAscii(IDVal, true);
925 if (IDVal == ".byte")
926 return ParseDirectiveValue(1);
927 if (IDVal == ".short")
928 return ParseDirectiveValue(2);
929 if (IDVal == ".value")
930 return ParseDirectiveValue(2);
931 if (IDVal == ".long")
932 return ParseDirectiveValue(4);
933 if (IDVal == ".quad")
934 return ParseDirectiveValue(8);
935 if (IDVal == ".single")
936 return ParseDirectiveRealValue(APFloat::IEEEsingle);
937 if (IDVal == ".double")
938 return ParseDirectiveRealValue(APFloat::IEEEdouble);
940 if (IDVal == ".align") {
941 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
942 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
944 if (IDVal == ".align32") {
945 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
946 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
948 if (IDVal == ".balign")
949 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
950 if (IDVal == ".balignw")
951 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
952 if (IDVal == ".balignl")
953 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
954 if (IDVal == ".p2align")
955 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
956 if (IDVal == ".p2alignw")
957 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
958 if (IDVal == ".p2alignl")
959 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
962 return ParseDirectiveOrg();
964 if (IDVal == ".fill")
965 return ParseDirectiveFill();
966 if (IDVal == ".space")
967 return ParseDirectiveSpace();
968 if (IDVal == ".zero")
969 return ParseDirectiveZero();
971 // Symbol attribute directives
973 if (IDVal == ".globl" || IDVal == ".global")
974 return ParseDirectiveSymbolAttribute(MCSA_Global);
975 // ELF only? Should it be here?
976 if (IDVal == ".local")
977 return ParseDirectiveSymbolAttribute(MCSA_Local);
978 if (IDVal == ".hidden")
979 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
980 if (IDVal == ".indirect_symbol")
981 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
982 if (IDVal == ".internal")
983 return ParseDirectiveSymbolAttribute(MCSA_Internal);
984 if (IDVal == ".lazy_reference")
985 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
986 if (IDVal == ".no_dead_strip")
987 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
988 if (IDVal == ".private_extern")
989 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
990 if (IDVal == ".protected")
991 return ParseDirectiveSymbolAttribute(MCSA_Protected);
992 if (IDVal == ".reference")
993 return ParseDirectiveSymbolAttribute(MCSA_Reference);
994 if (IDVal == ".weak")
995 return ParseDirectiveSymbolAttribute(MCSA_Weak);
996 if (IDVal == ".weak_definition")
997 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
998 if (IDVal == ".weak_reference")
999 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
1000 if (IDVal == ".weak_def_can_be_hidden")
1001 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1003 if (IDVal == ".comm")
1004 return ParseDirectiveComm(/*IsLocal=*/false);
1005 if (IDVal == ".lcomm")
1006 return ParseDirectiveComm(/*IsLocal=*/true);
1008 if (IDVal == ".abort")
1009 return ParseDirectiveAbort();
1010 if (IDVal == ".include")
1011 return ParseDirectiveInclude();
1013 // Look up the handler in the handler table.
1014 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1015 DirectiveMap.lookup(IDVal);
1017 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1019 // Target hook for parsing target specific directives.
1020 if (!getTargetParser().ParseDirective(ID))
1023 Warning(IDLoc, "ignoring directive for now");
1024 EatToEndOfStatement();
1028 CheckForValidSection();
1030 // Canonicalize the opcode to lower case.
1031 SmallString<128> Opcode;
1032 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1033 Opcode.push_back(tolower(IDVal[i]));
1035 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1036 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1039 // Dump the parsed representation, if requested.
1040 if (getShowParsedOperands()) {
1041 SmallString<256> Str;
1042 raw_svector_ostream OS(Str);
1043 OS << "parsed instruction: [";
1044 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1047 ParsedOperands[i]->dump(OS);
1051 PrintMessage(IDLoc, OS.str(), "note");
1054 // If parsing succeeded, match the instruction.
1056 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1059 // Free any parsed operands.
1060 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1061 delete ParsedOperands[i];
1063 // Don't skip the rest of the line, the instruction parser is responsible for
1068 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1069 const std::vector<std::vector<AsmToken> > &A)
1070 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1072 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1073 // to hold the macro body with substitutions.
1074 SmallString<256> Buf;
1075 raw_svector_ostream OS(Buf);
1077 StringRef Body = M->Body;
1078 while (!Body.empty()) {
1079 // Scan for the next substitution.
1080 std::size_t End = Body.size(), Pos = 0;
1081 for (; Pos != End; ++Pos) {
1082 // Check for a substitution or escape.
1083 if (Body[Pos] != '$' || Pos + 1 == End)
1086 char Next = Body[Pos + 1];
1087 if (Next == '$' || Next == 'n' || isdigit(Next))
1092 OS << Body.slice(0, Pos);
1094 // Check if we reached the end.
1098 switch (Body[Pos+1]) {
1104 // $n => number of arguments
1109 // $[0-9] => argument
1111 // Missing arguments are ignored.
1112 unsigned Index = Body[Pos+1] - '0';
1113 if (Index >= A.size())
1116 // Otherwise substitute with the token values, with spaces eliminated.
1117 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1118 ie = A[Index].end(); it != ie; ++it)
1119 OS << it->getString();
1124 // Update the scan point.
1125 Body = Body.substr(Pos + 2);
1128 // We include the .endmacro in the buffer as our queue to exit the macro
1130 OS << ".endmacro\n";
1132 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1135 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1137 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1138 // this, although we should protect against infinite loops.
1139 if (ActiveMacros.size() == 20)
1140 return TokError("macros cannot be nested more than 20 levels deep");
1142 // Parse the macro instantiation arguments.
1143 std::vector<std::vector<AsmToken> > MacroArguments;
1144 MacroArguments.push_back(std::vector<AsmToken>());
1145 unsigned ParenLevel = 0;
1147 if (Lexer.is(AsmToken::Eof))
1148 return TokError("unexpected token in macro instantiation");
1149 if (Lexer.is(AsmToken::EndOfStatement))
1152 // If we aren't inside parentheses and this is a comma, start a new token
1154 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1155 MacroArguments.push_back(std::vector<AsmToken>());
1157 // Adjust the current parentheses level.
1158 if (Lexer.is(AsmToken::LParen))
1160 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1163 // Append the token to the current argument list.
1164 MacroArguments.back().push_back(getTok());
1169 // Create the macro instantiation object and add to the current macro
1170 // instantiation stack.
1171 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1174 ActiveMacros.push_back(MI);
1176 // Jump to the macro instantiation and prime the lexer.
1177 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1178 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1184 void AsmParser::HandleMacroExit() {
1185 // Jump to the EndOfStatement we should return to, and consume it.
1186 JumpToLoc(ActiveMacros.back()->ExitLoc);
1189 // Pop the instantiation entry.
1190 delete ActiveMacros.back();
1191 ActiveMacros.pop_back();
1194 bool AsmParser::ParseAssignment(StringRef Name) {
1195 // FIXME: Use better location, we should use proper tokens.
1196 SMLoc EqualLoc = Lexer.getLoc();
1198 const MCExpr *Value;
1199 if (ParseExpression(Value))
1202 if (Lexer.isNot(AsmToken::EndOfStatement))
1203 return TokError("unexpected token in assignment");
1205 // Eat the end of statement marker.
1208 // Validate that the LHS is allowed to be a variable (either it has not been
1209 // used as a symbol, or it is an absolute symbol).
1210 MCSymbol *Sym = getContext().LookupSymbol(Name);
1212 // Diagnose assignment to a label.
1214 // FIXME: Diagnostics. Note the location of the definition as a label.
1215 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1216 if (Sym->isUndefined() && !Sym->isUsedInExpr())
1217 ; // Allow redefinitions of undefined symbols only used in directives.
1218 else if (!Sym->isUndefined() && !Sym->isAbsolute())
1219 return Error(EqualLoc, "redefinition of '" + Name + "'");
1220 else if (!Sym->isVariable())
1221 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1222 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1223 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1226 Sym = getContext().GetOrCreateSymbol(Name);
1228 // FIXME: Handle '.'.
1230 Sym->setUsedInExpr(true);
1232 // Do the assignment.
1233 Out.EmitAssignment(Sym, Value);
1238 /// ParseIdentifier:
1241 bool AsmParser::ParseIdentifier(StringRef &Res) {
1242 // The assembler has relaxed rules for accepting identifiers, in particular we
1243 // allow things like '.globl $foo', which would normally be separate
1244 // tokens. At this level, we have already lexed so we cannot (currently)
1245 // handle this as a context dependent token, instead we detect adjacent tokens
1246 // and return the combined identifier.
1247 if (Lexer.is(AsmToken::Dollar)) {
1248 SMLoc DollarLoc = getLexer().getLoc();
1250 // Consume the dollar sign, and check for a following identifier.
1252 if (Lexer.isNot(AsmToken::Identifier))
1255 // We have a '$' followed by an identifier, make sure they are adjacent.
1256 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1259 // Construct the joined identifier and consume the token.
1260 Res = StringRef(DollarLoc.getPointer(),
1261 getTok().getIdentifier().size() + 1);
1266 if (Lexer.isNot(AsmToken::Identifier) &&
1267 Lexer.isNot(AsmToken::String))
1270 Res = getTok().getIdentifier();
1272 Lex(); // Consume the identifier token.
1277 /// ParseDirectiveSet:
1278 /// ::= .set identifier ',' expression
1279 bool AsmParser::ParseDirectiveSet(StringRef IDVal) {
1282 if (ParseIdentifier(Name))
1283 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1285 if (getLexer().isNot(AsmToken::Comma))
1286 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1289 return ParseAssignment(Name);
1292 bool AsmParser::ParseEscapedString(std::string &Data) {
1293 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1296 StringRef Str = getTok().getStringContents();
1297 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1298 if (Str[i] != '\\') {
1303 // Recognize escaped characters. Note that this escape semantics currently
1304 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1307 return TokError("unexpected backslash at end of string");
1309 // Recognize octal sequences.
1310 if ((unsigned) (Str[i] - '0') <= 7) {
1311 // Consume up to three octal characters.
1312 unsigned Value = Str[i] - '0';
1314 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1316 Value = Value * 8 + (Str[i] - '0');
1318 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1320 Value = Value * 8 + (Str[i] - '0');
1325 return TokError("invalid octal escape sequence (out of range)");
1327 Data += (unsigned char) Value;
1331 // Otherwise recognize individual escapes.
1334 // Just reject invalid escape sequences for now.
1335 return TokError("invalid escape sequence (unrecognized character)");
1337 case 'b': Data += '\b'; break;
1338 case 'f': Data += '\f'; break;
1339 case 'n': Data += '\n'; break;
1340 case 'r': Data += '\r'; break;
1341 case 't': Data += '\t'; break;
1342 case '"': Data += '"'; break;
1343 case '\\': Data += '\\'; break;
1350 /// ParseDirectiveAscii:
1351 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
1352 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
1353 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1354 CheckForValidSection();
1357 if (getLexer().isNot(AsmToken::String))
1358 return TokError("expected string in '" + Twine(IDVal) + "' directive");
1361 if (ParseEscapedString(Data))
1364 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1366 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1370 if (getLexer().is(AsmToken::EndOfStatement))
1373 if (getLexer().isNot(AsmToken::Comma))
1374 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
1383 /// ParseDirectiveValue
1384 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1385 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1386 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1387 CheckForValidSection();
1390 const MCExpr *Value;
1391 if (ParseExpression(Value))
1394 // Special case constant expressions to match code generator.
1395 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1396 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1398 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1400 if (getLexer().is(AsmToken::EndOfStatement))
1403 // FIXME: Improve diagnostic.
1404 if (getLexer().isNot(AsmToken::Comma))
1405 return TokError("unexpected token in directive");
1414 /// ParseDirectiveRealValue
1415 /// ::= (.single | .double) [ expression (, expression)* ]
1416 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1417 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1418 CheckForValidSection();
1421 // We don't truly support arithmetic on floating point expressions, so we
1422 // have to manually parse unary prefixes.
1424 if (getLexer().is(AsmToken::Minus)) {
1427 } else if (getLexer().is(AsmToken::Plus))
1430 if (getLexer().isNot(AsmToken::Integer) &&
1431 getLexer().isNot(AsmToken::Real))
1432 return TokError("unexpected token in directive");
1434 // Convert to an APFloat.
1435 APFloat Value(Semantics);
1436 if (Value.convertFromString(getTok().getString(),
1437 APFloat::rmNearestTiesToEven) ==
1438 APFloat::opInvalidOp)
1439 return TokError("invalid floating point literal");
1443 // Consume the numeric token.
1446 // Emit the value as an integer.
1447 APInt AsInt = Value.bitcastToAPInt();
1448 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1449 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1451 if (getLexer().is(AsmToken::EndOfStatement))
1454 if (getLexer().isNot(AsmToken::Comma))
1455 return TokError("unexpected token in directive");
1464 /// ParseDirectiveSpace
1465 /// ::= .space expression [ , expression ]
1466 bool AsmParser::ParseDirectiveSpace() {
1467 CheckForValidSection();
1470 if (ParseAbsoluteExpression(NumBytes))
1473 int64_t FillExpr = 0;
1474 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1475 if (getLexer().isNot(AsmToken::Comma))
1476 return TokError("unexpected token in '.space' directive");
1479 if (ParseAbsoluteExpression(FillExpr))
1482 if (getLexer().isNot(AsmToken::EndOfStatement))
1483 return TokError("unexpected token in '.space' directive");
1489 return TokError("invalid number of bytes in '.space' directive");
1491 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1492 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1497 /// ParseDirectiveZero
1498 /// ::= .zero expression
1499 bool AsmParser::ParseDirectiveZero() {
1500 CheckForValidSection();
1503 if (ParseAbsoluteExpression(NumBytes))
1507 if (getLexer().is(AsmToken::Comma)) {
1509 if (ParseAbsoluteExpression(Val))
1513 if (getLexer().isNot(AsmToken::EndOfStatement))
1514 return TokError("unexpected token in '.zero' directive");
1518 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
1523 /// ParseDirectiveFill
1524 /// ::= .fill expression , expression , expression
1525 bool AsmParser::ParseDirectiveFill() {
1526 CheckForValidSection();
1529 if (ParseAbsoluteExpression(NumValues))
1532 if (getLexer().isNot(AsmToken::Comma))
1533 return TokError("unexpected token in '.fill' directive");
1537 if (ParseAbsoluteExpression(FillSize))
1540 if (getLexer().isNot(AsmToken::Comma))
1541 return TokError("unexpected token in '.fill' directive");
1545 if (ParseAbsoluteExpression(FillExpr))
1548 if (getLexer().isNot(AsmToken::EndOfStatement))
1549 return TokError("unexpected token in '.fill' directive");
1553 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1554 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1556 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1557 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1562 /// ParseDirectiveOrg
1563 /// ::= .org expression [ , expression ]
1564 bool AsmParser::ParseDirectiveOrg() {
1565 CheckForValidSection();
1567 const MCExpr *Offset;
1568 if (ParseExpression(Offset))
1571 // Parse optional fill expression.
1572 int64_t FillExpr = 0;
1573 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1574 if (getLexer().isNot(AsmToken::Comma))
1575 return TokError("unexpected token in '.org' directive");
1578 if (ParseAbsoluteExpression(FillExpr))
1581 if (getLexer().isNot(AsmToken::EndOfStatement))
1582 return TokError("unexpected token in '.org' directive");
1587 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1588 // has to be relative to the current section.
1589 getStreamer().EmitValueToOffset(Offset, FillExpr);
1594 /// ParseDirectiveAlign
1595 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1596 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1597 CheckForValidSection();
1599 SMLoc AlignmentLoc = getLexer().getLoc();
1601 if (ParseAbsoluteExpression(Alignment))
1605 bool HasFillExpr = false;
1606 int64_t FillExpr = 0;
1607 int64_t MaxBytesToFill = 0;
1608 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1609 if (getLexer().isNot(AsmToken::Comma))
1610 return TokError("unexpected token in directive");
1613 // The fill expression can be omitted while specifying a maximum number of
1614 // alignment bytes, e.g:
1616 if (getLexer().isNot(AsmToken::Comma)) {
1618 if (ParseAbsoluteExpression(FillExpr))
1622 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1623 if (getLexer().isNot(AsmToken::Comma))
1624 return TokError("unexpected token in directive");
1627 MaxBytesLoc = getLexer().getLoc();
1628 if (ParseAbsoluteExpression(MaxBytesToFill))
1631 if (getLexer().isNot(AsmToken::EndOfStatement))
1632 return TokError("unexpected token in directive");
1641 // Compute alignment in bytes.
1643 // FIXME: Diagnose overflow.
1644 if (Alignment >= 32) {
1645 Error(AlignmentLoc, "invalid alignment value");
1649 Alignment = 1ULL << Alignment;
1652 // Diagnose non-sensical max bytes to align.
1653 if (MaxBytesLoc.isValid()) {
1654 if (MaxBytesToFill < 1) {
1655 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1656 "many bytes, ignoring maximum bytes expression");
1660 if (MaxBytesToFill >= Alignment) {
1661 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1667 // Check whether we should use optimal code alignment for this .align
1669 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
1670 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1671 ValueSize == 1 && UseCodeAlign) {
1672 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1674 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1675 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1682 /// ParseDirectiveSymbolAttribute
1683 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1684 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1685 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1689 if (ParseIdentifier(Name))
1690 return TokError("expected identifier in directive");
1692 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1694 getStreamer().EmitSymbolAttribute(Sym, Attr);
1696 if (getLexer().is(AsmToken::EndOfStatement))
1699 if (getLexer().isNot(AsmToken::Comma))
1700 return TokError("unexpected token in directive");
1709 /// ParseDirectiveComm
1710 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1711 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1712 CheckForValidSection();
1714 SMLoc IDLoc = getLexer().getLoc();
1716 if (ParseIdentifier(Name))
1717 return TokError("expected identifier in directive");
1719 // Handle the identifier as the key symbol.
1720 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1722 if (getLexer().isNot(AsmToken::Comma))
1723 return TokError("unexpected token in directive");
1727 SMLoc SizeLoc = getLexer().getLoc();
1728 if (ParseAbsoluteExpression(Size))
1731 int64_t Pow2Alignment = 0;
1732 SMLoc Pow2AlignmentLoc;
1733 if (getLexer().is(AsmToken::Comma)) {
1735 Pow2AlignmentLoc = getLexer().getLoc();
1736 if (ParseAbsoluteExpression(Pow2Alignment))
1739 // If this target takes alignments in bytes (not log) validate and convert.
1740 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1741 if (!isPowerOf2_64(Pow2Alignment))
1742 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1743 Pow2Alignment = Log2_64(Pow2Alignment);
1747 if (getLexer().isNot(AsmToken::EndOfStatement))
1748 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1752 // NOTE: a size of zero for a .comm should create a undefined symbol
1753 // but a size of .lcomm creates a bss symbol of size zero.
1755 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1756 "be less than zero");
1758 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1759 // may internally end up wanting an alignment in bytes.
1760 // FIXME: Diagnose overflow.
1761 if (Pow2Alignment < 0)
1762 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1763 "alignment, can't be less than zero");
1765 if (!Sym->isUndefined())
1766 return Error(IDLoc, "invalid symbol redefinition");
1768 // '.lcomm' is equivalent to '.zerofill'.
1769 // Create the Symbol as a common or local common with Size and Pow2Alignment
1771 getStreamer().EmitZerofill(Ctx.getMachOSection(
1772 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1773 0, SectionKind::getBSS()),
1774 Sym, Size, 1 << Pow2Alignment);
1778 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1782 /// ParseDirectiveAbort
1783 /// ::= .abort [... message ...]
1784 bool AsmParser::ParseDirectiveAbort() {
1785 // FIXME: Use loc from directive.
1786 SMLoc Loc = getLexer().getLoc();
1788 StringRef Str = ParseStringToEndOfStatement();
1789 if (getLexer().isNot(AsmToken::EndOfStatement))
1790 return TokError("unexpected token in '.abort' directive");
1795 Error(Loc, ".abort detected. Assembly stopping.");
1797 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1798 // FIXME: Actually abort assembly here.
1803 /// ParseDirectiveInclude
1804 /// ::= .include "filename"
1805 bool AsmParser::ParseDirectiveInclude() {
1806 if (getLexer().isNot(AsmToken::String))
1807 return TokError("expected string in '.include' directive");
1809 std::string Filename = getTok().getString();
1810 SMLoc IncludeLoc = getLexer().getLoc();
1813 if (getLexer().isNot(AsmToken::EndOfStatement))
1814 return TokError("unexpected token in '.include' directive");
1816 // Strip the quotes.
1817 Filename = Filename.substr(1, Filename.size()-2);
1819 // Attempt to switch the lexer to the included file before consuming the end
1820 // of statement to avoid losing it when we switch.
1821 if (EnterIncludeFile(Filename)) {
1822 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1829 /// ParseDirectiveIf
1830 /// ::= .if expression
1831 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1832 TheCondStack.push_back(TheCondState);
1833 TheCondState.TheCond = AsmCond::IfCond;
1834 if(TheCondState.Ignore) {
1835 EatToEndOfStatement();
1839 if (ParseAbsoluteExpression(ExprValue))
1842 if (getLexer().isNot(AsmToken::EndOfStatement))
1843 return TokError("unexpected token in '.if' directive");
1847 TheCondState.CondMet = ExprValue;
1848 TheCondState.Ignore = !TheCondState.CondMet;
1854 /// ParseDirectiveElseIf
1855 /// ::= .elseif expression
1856 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1857 if (TheCondState.TheCond != AsmCond::IfCond &&
1858 TheCondState.TheCond != AsmCond::ElseIfCond)
1859 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1861 TheCondState.TheCond = AsmCond::ElseIfCond;
1863 bool LastIgnoreState = false;
1864 if (!TheCondStack.empty())
1865 LastIgnoreState = TheCondStack.back().Ignore;
1866 if (LastIgnoreState || TheCondState.CondMet) {
1867 TheCondState.Ignore = true;
1868 EatToEndOfStatement();
1872 if (ParseAbsoluteExpression(ExprValue))
1875 if (getLexer().isNot(AsmToken::EndOfStatement))
1876 return TokError("unexpected token in '.elseif' directive");
1879 TheCondState.CondMet = ExprValue;
1880 TheCondState.Ignore = !TheCondState.CondMet;
1886 /// ParseDirectiveElse
1888 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
1889 if (getLexer().isNot(AsmToken::EndOfStatement))
1890 return TokError("unexpected token in '.else' directive");
1894 if (TheCondState.TheCond != AsmCond::IfCond &&
1895 TheCondState.TheCond != AsmCond::ElseIfCond)
1896 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
1898 TheCondState.TheCond = AsmCond::ElseCond;
1899 bool LastIgnoreState = false;
1900 if (!TheCondStack.empty())
1901 LastIgnoreState = TheCondStack.back().Ignore;
1902 if (LastIgnoreState || TheCondState.CondMet)
1903 TheCondState.Ignore = true;
1905 TheCondState.Ignore = false;
1910 /// ParseDirectiveEndIf
1912 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
1913 if (getLexer().isNot(AsmToken::EndOfStatement))
1914 return TokError("unexpected token in '.endif' directive");
1918 if ((TheCondState.TheCond == AsmCond::NoCond) ||
1919 TheCondStack.empty())
1920 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
1922 if (!TheCondStack.empty()) {
1923 TheCondState = TheCondStack.back();
1924 TheCondStack.pop_back();
1930 /// ParseDirectiveFile
1931 /// ::= .file [number] string
1932 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
1933 // FIXME: I'm not sure what this is.
1934 int64_t FileNumber = -1;
1935 SMLoc FileNumberLoc = getLexer().getLoc();
1936 if (getLexer().is(AsmToken::Integer)) {
1937 FileNumber = getTok().getIntVal();
1941 return TokError("file number less than one");
1944 if (getLexer().isNot(AsmToken::String))
1945 return TokError("unexpected token in '.file' directive");
1947 StringRef Filename = getTok().getString();
1948 Filename = Filename.substr(1, Filename.size()-2);
1951 if (getLexer().isNot(AsmToken::EndOfStatement))
1952 return TokError("unexpected token in '.file' directive");
1954 if (FileNumber == -1)
1955 getStreamer().EmitFileDirective(Filename);
1957 if (getContext().GetDwarfFile(Filename, FileNumber) == 0)
1958 Error(FileNumberLoc, "file number already allocated");
1959 getStreamer().EmitDwarfFileDirective(FileNumber, Filename);
1965 /// ParseDirectiveLine
1966 /// ::= .line [number]
1967 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
1968 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1969 if (getLexer().isNot(AsmToken::Integer))
1970 return TokError("unexpected token in '.line' directive");
1972 int64_t LineNumber = getTok().getIntVal();
1976 // FIXME: Do something with the .line.
1979 if (getLexer().isNot(AsmToken::EndOfStatement))
1980 return TokError("unexpected token in '.line' directive");
1986 /// ParseDirectiveLoc
1987 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
1988 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
1989 /// The first number is a file number, must have been previously assigned with
1990 /// a .file directive, the second number is the line number and optionally the
1991 /// third number is a column position (zero if not specified). The remaining
1992 /// optional items are .loc sub-directives.
1993 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
1995 if (getLexer().isNot(AsmToken::Integer))
1996 return TokError("unexpected token in '.loc' directive");
1997 int64_t FileNumber = getTok().getIntVal();
1999 return TokError("file number less than one in '.loc' directive");
2000 if (!getContext().isValidDwarfFileNumber(FileNumber))
2001 return TokError("unassigned file number in '.loc' directive");
2004 int64_t LineNumber = 0;
2005 if (getLexer().is(AsmToken::Integer)) {
2006 LineNumber = getTok().getIntVal();
2008 return TokError("line number less than one in '.loc' directive");
2012 int64_t ColumnPos = 0;
2013 if (getLexer().is(AsmToken::Integer)) {
2014 ColumnPos = getTok().getIntVal();
2016 return TokError("column position less than zero in '.loc' directive");
2020 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2022 int64_t Discriminator = 0;
2023 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2025 if (getLexer().is(AsmToken::EndOfStatement))
2029 SMLoc Loc = getTok().getLoc();
2030 if (getParser().ParseIdentifier(Name))
2031 return TokError("unexpected token in '.loc' directive");
2033 if (Name == "basic_block")
2034 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2035 else if (Name == "prologue_end")
2036 Flags |= DWARF2_FLAG_PROLOGUE_END;
2037 else if (Name == "epilogue_begin")
2038 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2039 else if (Name == "is_stmt") {
2040 SMLoc Loc = getTok().getLoc();
2041 const MCExpr *Value;
2042 if (getParser().ParseExpression(Value))
2044 // The expression must be the constant 0 or 1.
2045 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2046 int Value = MCE->getValue();
2048 Flags &= ~DWARF2_FLAG_IS_STMT;
2049 else if (Value == 1)
2050 Flags |= DWARF2_FLAG_IS_STMT;
2052 return Error(Loc, "is_stmt value not 0 or 1");
2055 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2058 else if (Name == "isa") {
2059 SMLoc Loc = getTok().getLoc();
2060 const MCExpr *Value;
2061 if (getParser().ParseExpression(Value))
2063 // The expression must be a constant greater or equal to 0.
2064 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2065 int Value = MCE->getValue();
2067 return Error(Loc, "isa number less than zero");
2071 return Error(Loc, "isa number not a constant value");
2074 else if (Name == "discriminator") {
2075 if (getParser().ParseAbsoluteExpression(Discriminator))
2079 return Error(Loc, "unknown sub-directive in '.loc' directive");
2082 if (getLexer().is(AsmToken::EndOfStatement))
2087 getContext().setCurrentDwarfLoc(FileNumber, LineNumber, ColumnPos, Flags,
2088 Isa, Discriminator);
2093 /// ParseDirectiveStabs
2094 /// ::= .stabs string, number, number, number
2095 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2096 SMLoc DirectiveLoc) {
2097 return TokError("unsupported directive '" + Directive + "'");
2100 /// ParseDirectiveMacrosOnOff
2103 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2104 SMLoc DirectiveLoc) {
2105 if (getLexer().isNot(AsmToken::EndOfStatement))
2106 return Error(getLexer().getLoc(),
2107 "unexpected token in '" + Directive + "' directive");
2109 getParser().MacrosEnabled = Directive == ".macros_on";
2114 /// ParseDirectiveMacro
2116 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2117 SMLoc DirectiveLoc) {
2119 if (getParser().ParseIdentifier(Name))
2120 return TokError("expected identifier in directive");
2122 if (getLexer().isNot(AsmToken::EndOfStatement))
2123 return TokError("unexpected token in '.macro' directive");
2125 // Eat the end of statement.
2128 AsmToken EndToken, StartToken = getTok();
2130 // Lex the macro definition.
2132 // Check whether we have reached the end of the file.
2133 if (getLexer().is(AsmToken::Eof))
2134 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2136 // Otherwise, check whether we have reach the .endmacro.
2137 if (getLexer().is(AsmToken::Identifier) &&
2138 (getTok().getIdentifier() == ".endm" ||
2139 getTok().getIdentifier() == ".endmacro")) {
2140 EndToken = getTok();
2142 if (getLexer().isNot(AsmToken::EndOfStatement))
2143 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2148 // Otherwise, scan til the end of the statement.
2149 getParser().EatToEndOfStatement();
2152 if (getParser().MacroMap.lookup(Name)) {
2153 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2156 const char *BodyStart = StartToken.getLoc().getPointer();
2157 const char *BodyEnd = EndToken.getLoc().getPointer();
2158 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2159 getParser().MacroMap[Name] = new Macro(Name, Body);
2163 /// ParseDirectiveEndMacro
2166 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2167 SMLoc DirectiveLoc) {
2168 if (getLexer().isNot(AsmToken::EndOfStatement))
2169 return TokError("unexpected token in '" + Directive + "' directive");
2171 // If we are inside a macro instantiation, terminate the current
2173 if (!getParser().ActiveMacros.empty()) {
2174 getParser().HandleMacroExit();
2178 // Otherwise, this .endmacro is a stray entry in the file; well formed
2179 // .endmacro directives are handled during the macro definition parsing.
2180 return TokError("unexpected '" + Directive + "' in file, "
2181 "no current macro definition");
2184 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2185 getParser().CheckForValidSection();
2187 const MCExpr *Value;
2189 if (getParser().ParseExpression(Value))
2192 if (getLexer().isNot(AsmToken::EndOfStatement))
2193 return TokError("unexpected token in directive");
2195 if (DirName[1] == 's')
2196 getStreamer().EmitSLEB128Value(Value);
2198 getStreamer().EmitULEB128Value(Value);
2204 /// \brief Create an MCAsmParser instance.
2205 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2206 MCContext &C, MCStreamer &Out,
2207 const MCAsmInfo &MAI) {
2208 return new AsmParser(T, SM, C, Out, MAI);