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/Compiler.h"
31 #include "llvm/Support/MemoryBuffer.h"
32 #include "llvm/Support/SourceMgr.h"
33 #include "llvm/Support/raw_ostream.h"
34 #include "llvm/Target/TargetAsmParser.h"
40 /// \brief Helper class for tracking macro definitions.
46 Macro(StringRef N, StringRef B) : Name(N), Body(B) {}
49 /// \brief Helper class for storing information about an active macro
51 struct MacroInstantiation {
52 /// The macro being instantiated.
53 const Macro *TheMacro;
55 /// The macro instantiation with substitutions.
56 MemoryBuffer *Instantiation;
58 /// The location of the instantiation.
59 SMLoc InstantiationLoc;
61 /// The location where parsing should resume upon instantiation completion.
65 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
66 const std::vector<std::vector<AsmToken> > &A);
69 /// \brief The concrete assembly parser instance.
70 class AsmParser : public MCAsmParser {
71 friend class GenericAsmParser;
73 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
74 void operator=(const AsmParser &); // DO NOT IMPLEMENT
80 MCAsmParserExtension *GenericParser;
81 MCAsmParserExtension *PlatformParser;
83 /// This is the current buffer index we're lexing from as managed by the
88 std::vector<AsmCond> TheCondStack;
90 /// DirectiveMap - This is a table handlers for directives. Each handler is
91 /// invoked after the directive identifier is read and is responsible for
92 /// parsing and validating the rest of the directive. The handler is passed
93 /// in the directive name and the location of the directive keyword.
94 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
96 /// MacroMap - Map of currently defined macros.
97 StringMap<Macro*> MacroMap;
99 /// ActiveMacros - Stack of active macro instantiations.
100 std::vector<MacroInstantiation*> ActiveMacros;
102 /// Boolean tracking whether macro substitution is enabled.
103 unsigned MacrosEnabled : 1;
105 /// Flag tracking whether any errors have been encountered.
106 unsigned HadError : 1;
109 AsmParser(const Target &T, SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
110 const MCAsmInfo &MAI);
113 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
115 void AddDirectiveHandler(MCAsmParserExtension *Object,
117 DirectiveHandler Handler) {
118 DirectiveMap[Directive] = std::make_pair(Object, Handler);
122 /// @name MCAsmParser Interface
125 virtual SourceMgr &getSourceManager() { return SrcMgr; }
126 virtual MCAsmLexer &getLexer() { return Lexer; }
127 virtual MCContext &getContext() { return Ctx; }
128 virtual MCStreamer &getStreamer() { return Out; }
130 virtual void Warning(SMLoc L, const Twine &Meg);
131 virtual bool Error(SMLoc L, const Twine &Msg);
133 const AsmToken &Lex();
135 bool ParseExpression(const MCExpr *&Res);
136 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
137 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
138 virtual bool ParseAbsoluteExpression(int64_t &Res);
143 void CheckForValidSection();
145 bool ParseStatement();
147 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
148 void HandleMacroExit();
150 void PrintMacroInstantiations();
151 void PrintMessage(SMLoc Loc, const Twine &Msg, const char *Type) const {
152 SrcMgr.PrintMessage(Loc, Msg, Type);
155 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
156 bool EnterIncludeFile(const std::string &Filename);
158 /// \brief Reset the current lexer position to that given by \arg Loc. The
159 /// current token is not set; clients should ensure Lex() is called
161 void JumpToLoc(SMLoc Loc);
163 void EatToEndOfStatement();
165 /// \brief Parse up to the end of statement and a return the contents from the
166 /// current token until the end of the statement; the current token on exit
167 /// will be either the EndOfStatement or EOF.
168 StringRef ParseStringToEndOfStatement();
170 bool ParseAssignment(StringRef Name);
172 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
173 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
174 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
176 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
177 /// and set \arg Res to the identifier contents.
178 bool ParseIdentifier(StringRef &Res);
180 // Directive Parsing.
181 bool ParseDirectiveAscii(bool ZeroTerminated); // ".ascii", ".asciiz"
182 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
183 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
184 bool ParseDirectiveFill(); // ".fill"
185 bool ParseDirectiveSpace(); // ".space"
186 bool ParseDirectiveZero(); // ".zero"
187 bool ParseDirectiveSet(); // ".set"
188 bool ParseDirectiveOrg(); // ".org"
189 // ".align{,32}", ".p2align{,w,l}"
190 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
192 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
193 /// accepts a single symbol (which should be a label or an external).
194 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
195 bool ParseDirectiveELFType(); // ELF specific ".type"
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 void ParseUleb128(uint64_t Value);
264 void ParseSleb128(int64_t Value);
265 bool ParseDirectiveLEB128(StringRef, SMLoc);
272 extern MCAsmParserExtension *createDarwinAsmParser();
273 extern MCAsmParserExtension *createELFAsmParser();
274 extern MCAsmParserExtension *createCOFFAsmParser();
278 enum { DEFAULT_ADDRSPACE = 0 };
280 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
281 MCStreamer &_Out, const MCAsmInfo &_MAI)
282 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
283 GenericParser(new GenericAsmParser), PlatformParser(0),
284 CurBuffer(0), MacrosEnabled(true) {
285 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
287 // Initialize the generic parser.
288 GenericParser->Initialize(*this);
290 // Initialize the platform / file format parser.
292 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
294 if (_MAI.hasMicrosoftFastStdCallMangling()) {
295 PlatformParser = createCOFFAsmParser();
296 PlatformParser->Initialize(*this);
297 } else if (_MAI.hasSubsectionsViaSymbols()) {
298 PlatformParser = createDarwinAsmParser();
299 PlatformParser->Initialize(*this);
301 PlatformParser = createELFAsmParser();
302 PlatformParser->Initialize(*this);
306 AsmParser::~AsmParser() {
307 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
309 // Destroy any macros.
310 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
311 ie = MacroMap.end(); it != ie; ++it)
312 delete it->getValue();
314 delete PlatformParser;
315 delete GenericParser;
318 void AsmParser::PrintMacroInstantiations() {
319 // Print the active macro instantiation stack.
320 for (std::vector<MacroInstantiation*>::const_reverse_iterator
321 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
322 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
326 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
327 PrintMessage(L, Msg, "warning");
328 PrintMacroInstantiations();
331 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
333 PrintMessage(L, Msg, "error");
334 PrintMacroInstantiations();
338 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
339 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
345 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
350 void AsmParser::JumpToLoc(SMLoc Loc) {
351 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
352 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
355 const AsmToken &AsmParser::Lex() {
356 const AsmToken *tok = &Lexer.Lex();
358 if (tok->is(AsmToken::Eof)) {
359 // If this is the end of an included file, pop the parent file off the
361 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
362 if (ParentIncludeLoc != SMLoc()) {
363 JumpToLoc(ParentIncludeLoc);
368 if (tok->is(AsmToken::Error))
369 Error(Lexer.getErrLoc(), Lexer.getErr());
374 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
375 // Create the initial section, if requested.
376 if (!NoInitialTextSection)
383 AsmCond StartingCondState = TheCondState;
385 // While we have input, parse each statement.
386 while (Lexer.isNot(AsmToken::Eof)) {
387 if (!ParseStatement()) continue;
389 // We had an error, validate that one was emitted and recover by skipping to
391 assert(HadError && "Parse statement returned an error, but none emitted!");
392 EatToEndOfStatement();
395 if (TheCondState.TheCond != StartingCondState.TheCond ||
396 TheCondState.Ignore != StartingCondState.Ignore)
397 return TokError("unmatched .ifs or .elses");
399 // Check to see there are no empty DwarfFile slots.
400 const std::vector<MCDwarfFile *> &MCDwarfFiles =
401 getContext().getMCDwarfFiles();
402 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
403 if (!MCDwarfFiles[i])
404 TokError("unassigned file number: " + Twine(i) + " for .file directives");
407 // Finalize the output stream if there are no errors and if the client wants
409 if (!HadError && !NoFinalize)
415 void AsmParser::CheckForValidSection() {
416 if (!getStreamer().getCurrentSection()) {
417 TokError("expected section directive before assembly directive");
418 Out.SwitchSection(Ctx.getMachOSection(
420 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
421 0, SectionKind::getText()));
425 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
426 void AsmParser::EatToEndOfStatement() {
427 while (Lexer.isNot(AsmToken::EndOfStatement) &&
428 Lexer.isNot(AsmToken::Eof))
432 if (Lexer.is(AsmToken::EndOfStatement))
436 StringRef AsmParser::ParseStringToEndOfStatement() {
437 const char *Start = getTok().getLoc().getPointer();
439 while (Lexer.isNot(AsmToken::EndOfStatement) &&
440 Lexer.isNot(AsmToken::Eof))
443 const char *End = getTok().getLoc().getPointer();
444 return StringRef(Start, End - Start);
447 /// ParseParenExpr - Parse a paren expression and return it.
448 /// NOTE: This assumes the leading '(' has already been consumed.
450 /// parenexpr ::= expr)
452 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
453 if (ParseExpression(Res)) return true;
454 if (Lexer.isNot(AsmToken::RParen))
455 return TokError("expected ')' in parentheses expression");
456 EndLoc = Lexer.getLoc();
461 /// ParsePrimaryExpr - Parse a primary expression and return it.
462 /// primaryexpr ::= (parenexpr
463 /// primaryexpr ::= symbol
464 /// primaryexpr ::= number
465 /// primaryexpr ::= '.'
466 /// primaryexpr ::= ~,+,- primaryexpr
467 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
468 switch (Lexer.getKind()) {
470 return TokError("unknown token in expression");
471 case AsmToken::Exclaim:
472 Lex(); // Eat the operator.
473 if (ParsePrimaryExpr(Res, EndLoc))
475 Res = MCUnaryExpr::CreateLNot(Res, getContext());
477 case AsmToken::Dollar:
478 case AsmToken::String:
479 case AsmToken::Identifier: {
480 EndLoc = Lexer.getLoc();
482 StringRef Identifier;
483 if (ParseIdentifier(Identifier))
486 // This is a symbol reference.
487 std::pair<StringRef, StringRef> Split = Identifier.split('@');
488 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
490 // Mark the symbol as used in an expression.
491 Sym->setUsedInExpr(true);
493 // Lookup the symbol variant if used.
494 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
495 if (Split.first.size() != Identifier.size()) {
496 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
497 if (Variant == MCSymbolRefExpr::VK_Invalid) {
498 Variant = MCSymbolRefExpr::VK_None;
499 TokError("invalid variant '" + Split.second + "'");
503 // If this is an absolute variable reference, substitute it now to preserve
504 // semantics in the face of reassignment.
505 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
507 return Error(EndLoc, "unexpected modified on variable reference");
509 Res = Sym->getVariableValue();
513 // Otherwise create a symbol ref.
514 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
517 case AsmToken::Integer: {
518 SMLoc Loc = getTok().getLoc();
519 int64_t IntVal = getTok().getIntVal();
520 Res = MCConstantExpr::Create(IntVal, getContext());
521 EndLoc = Lexer.getLoc();
523 // Look for 'b' or 'f' following an Integer as a directional label
524 if (Lexer.getKind() == AsmToken::Identifier) {
525 StringRef IDVal = getTok().getString();
526 if (IDVal == "f" || IDVal == "b"){
527 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
528 IDVal == "f" ? 1 : 0);
529 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
531 if(IDVal == "b" && Sym->isUndefined())
532 return Error(Loc, "invalid reference to undefined symbol");
533 EndLoc = Lexer.getLoc();
534 Lex(); // Eat identifier.
539 case AsmToken::Dot: {
540 // This is a '.' reference, which references the current PC. Emit a
541 // temporary label to the streamer and refer to it.
542 MCSymbol *Sym = Ctx.CreateTempSymbol();
544 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
545 EndLoc = Lexer.getLoc();
546 Lex(); // Eat identifier.
550 case AsmToken::LParen:
551 Lex(); // Eat the '('.
552 return ParseParenExpr(Res, EndLoc);
553 case AsmToken::Minus:
554 Lex(); // Eat the operator.
555 if (ParsePrimaryExpr(Res, EndLoc))
557 Res = MCUnaryExpr::CreateMinus(Res, getContext());
560 Lex(); // Eat the operator.
561 if (ParsePrimaryExpr(Res, EndLoc))
563 Res = MCUnaryExpr::CreatePlus(Res, getContext());
565 case AsmToken::Tilde:
566 Lex(); // Eat the operator.
567 if (ParsePrimaryExpr(Res, EndLoc))
569 Res = MCUnaryExpr::CreateNot(Res, getContext());
574 bool AsmParser::ParseExpression(const MCExpr *&Res) {
576 return ParseExpression(Res, EndLoc);
580 AsmParser::ApplyModifierToExpr(const MCExpr *E,
581 MCSymbolRefExpr::VariantKind Variant) {
582 // Recurse over the given expression, rebuilding it to apply the given variant
583 // if there is exactly one symbol.
584 switch (E->getKind()) {
586 case MCExpr::Constant:
589 case MCExpr::SymbolRef: {
590 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
592 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
593 TokError("invalid variant on expression '" +
594 getTok().getIdentifier() + "' (already modified)");
598 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
601 case MCExpr::Unary: {
602 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
603 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
606 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
609 case MCExpr::Binary: {
610 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
611 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
612 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
617 if (!LHS) LHS = BE->getLHS();
618 if (!RHS) RHS = BE->getRHS();
620 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
624 assert(0 && "Invalid expression kind!");
628 /// ParseExpression - Parse an expression and return it.
630 /// expr ::= expr +,- expr -> lowest.
631 /// expr ::= expr |,^,&,! expr -> middle.
632 /// expr ::= expr *,/,%,<<,>> expr -> highest.
633 /// expr ::= primaryexpr
635 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
636 // Parse the expression.
638 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
641 // As a special case, we support 'a op b @ modifier' by rewriting the
642 // expression to include the modifier. This is inefficient, but in general we
643 // expect users to use 'a@modifier op b'.
644 if (Lexer.getKind() == AsmToken::At) {
647 if (Lexer.isNot(AsmToken::Identifier))
648 return TokError("unexpected symbol modifier following '@'");
650 MCSymbolRefExpr::VariantKind Variant =
651 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
652 if (Variant == MCSymbolRefExpr::VK_Invalid)
653 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
655 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
657 return TokError("invalid modifier '" + getTok().getIdentifier() +
658 "' (no symbols present)");
666 // Try to constant fold it up front, if possible.
668 if (Res->EvaluateAsAbsolute(Value))
669 Res = MCConstantExpr::Create(Value, getContext());
674 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
676 return ParseParenExpr(Res, EndLoc) ||
677 ParseBinOpRHS(1, Res, EndLoc);
680 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
683 SMLoc StartLoc = Lexer.getLoc();
684 if (ParseExpression(Expr))
687 if (!Expr->EvaluateAsAbsolute(Res))
688 return Error(StartLoc, "expected absolute expression");
693 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
694 MCBinaryExpr::Opcode &Kind) {
697 return 0; // not a binop.
699 // Lowest Precedence: &&, ||, @
700 case AsmToken::AmpAmp:
701 Kind = MCBinaryExpr::LAnd;
703 case AsmToken::PipePipe:
704 Kind = MCBinaryExpr::LOr;
708 // Low Precedence: |, &, ^
710 // FIXME: gas seems to support '!' as an infix operator?
712 Kind = MCBinaryExpr::Or;
714 case AsmToken::Caret:
715 Kind = MCBinaryExpr::Xor;
718 Kind = MCBinaryExpr::And;
721 // Intermediate Precedence: +, -, ==, !=, <>, <, <=, >, >=
723 Kind = MCBinaryExpr::Add;
725 case AsmToken::Minus:
726 Kind = MCBinaryExpr::Sub;
728 case AsmToken::EqualEqual:
729 Kind = MCBinaryExpr::EQ;
731 case AsmToken::ExclaimEqual:
732 case AsmToken::LessGreater:
733 Kind = MCBinaryExpr::NE;
736 Kind = MCBinaryExpr::LT;
738 case AsmToken::LessEqual:
739 Kind = MCBinaryExpr::LTE;
741 case AsmToken::Greater:
742 Kind = MCBinaryExpr::GT;
744 case AsmToken::GreaterEqual:
745 Kind = MCBinaryExpr::GTE;
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
916 return ParseDirectiveSet();
920 if (IDVal == ".ascii")
921 return ParseDirectiveAscii(false);
922 if (IDVal == ".asciz")
923 return ParseDirectiveAscii(true);
925 if (IDVal == ".byte")
926 return ParseDirectiveValue(1);
927 if (IDVal == ".short")
928 return ParseDirectiveValue(2);
929 if (IDVal == ".long")
930 return ParseDirectiveValue(4);
931 if (IDVal == ".quad")
932 return ParseDirectiveValue(8);
933 if (IDVal == ".single")
934 return ParseDirectiveRealValue(APFloat::IEEEsingle);
935 if (IDVal == ".double")
936 return ParseDirectiveRealValue(APFloat::IEEEdouble);
938 if (IDVal == ".align") {
939 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
940 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
942 if (IDVal == ".align32") {
943 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
944 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
946 if (IDVal == ".balign")
947 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
948 if (IDVal == ".balignw")
949 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
950 if (IDVal == ".balignl")
951 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
952 if (IDVal == ".p2align")
953 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
954 if (IDVal == ".p2alignw")
955 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
956 if (IDVal == ".p2alignl")
957 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
960 return ParseDirectiveOrg();
962 if (IDVal == ".fill")
963 return ParseDirectiveFill();
964 if (IDVal == ".space")
965 return ParseDirectiveSpace();
966 if (IDVal == ".zero")
967 return ParseDirectiveZero();
969 // Symbol attribute directives
971 if (IDVal == ".globl" || IDVal == ".global")
972 return ParseDirectiveSymbolAttribute(MCSA_Global);
973 // ELF only? Should it be here?
974 if (IDVal == ".local")
975 return ParseDirectiveSymbolAttribute(MCSA_Local);
976 if (IDVal == ".hidden")
977 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
978 if (IDVal == ".indirect_symbol")
979 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
980 if (IDVal == ".internal")
981 return ParseDirectiveSymbolAttribute(MCSA_Internal);
982 if (IDVal == ".lazy_reference")
983 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
984 if (IDVal == ".no_dead_strip")
985 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
986 if (IDVal == ".private_extern")
987 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
988 if (IDVal == ".protected")
989 return ParseDirectiveSymbolAttribute(MCSA_Protected);
990 if (IDVal == ".reference")
991 return ParseDirectiveSymbolAttribute(MCSA_Reference);
992 if (IDVal == ".weak")
993 return ParseDirectiveSymbolAttribute(MCSA_Weak);
994 if (IDVal == ".weak_definition")
995 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
996 if (IDVal == ".weak_reference")
997 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
998 if (IDVal == ".weak_def_can_be_hidden")
999 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1001 if (IDVal == ".comm")
1002 return ParseDirectiveComm(/*IsLocal=*/false);
1003 if (IDVal == ".lcomm")
1004 return ParseDirectiveComm(/*IsLocal=*/true);
1006 if (IDVal == ".abort")
1007 return ParseDirectiveAbort();
1008 if (IDVal == ".include")
1009 return ParseDirectiveInclude();
1011 // Look up the handler in the handler table.
1012 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1013 DirectiveMap.lookup(IDVal);
1015 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1017 // Target hook for parsing target specific directives.
1018 if (!getTargetParser().ParseDirective(ID))
1021 Warning(IDLoc, "ignoring directive for now");
1022 EatToEndOfStatement();
1026 CheckForValidSection();
1028 // Canonicalize the opcode to lower case.
1029 SmallString<128> Opcode;
1030 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1031 Opcode.push_back(tolower(IDVal[i]));
1033 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1034 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1037 // Dump the parsed representation, if requested.
1038 if (getShowParsedOperands()) {
1039 SmallString<256> Str;
1040 raw_svector_ostream OS(Str);
1041 OS << "parsed instruction: [";
1042 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1045 ParsedOperands[i]->dump(OS);
1049 PrintMessage(IDLoc, OS.str(), "note");
1052 // If parsing succeeded, match the instruction.
1054 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1057 // Free any parsed operands.
1058 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1059 delete ParsedOperands[i];
1061 // Don't skip the rest of the line, the instruction parser is responsible for
1066 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1067 const std::vector<std::vector<AsmToken> > &A)
1068 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1070 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1071 // to hold the macro body with substitutions.
1072 SmallString<256> Buf;
1073 raw_svector_ostream OS(Buf);
1075 StringRef Body = M->Body;
1076 while (!Body.empty()) {
1077 // Scan for the next substitution.
1078 std::size_t End = Body.size(), Pos = 0;
1079 for (; Pos != End; ++Pos) {
1080 // Check for a substitution or escape.
1081 if (Body[Pos] != '$' || Pos + 1 == End)
1084 char Next = Body[Pos + 1];
1085 if (Next == '$' || Next == 'n' || isdigit(Next))
1090 OS << Body.slice(0, Pos);
1092 // Check if we reached the end.
1096 switch (Body[Pos+1]) {
1102 // $n => number of arguments
1107 // $[0-9] => argument
1109 // Missing arguments are ignored.
1110 unsigned Index = Body[Pos+1] - '0';
1111 if (Index >= A.size())
1114 // Otherwise substitute with the token values, with spaces eliminated.
1115 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1116 ie = A[Index].end(); it != ie; ++it)
1117 OS << it->getString();
1122 // Update the scan point.
1123 Body = Body.substr(Pos + 2);
1126 // We include the .endmacro in the buffer as our queue to exit the macro
1128 OS << ".endmacro\n";
1130 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1133 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1135 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1136 // this, although we should protect against infinite loops.
1137 if (ActiveMacros.size() == 20)
1138 return TokError("macros cannot be nested more than 20 levels deep");
1140 // Parse the macro instantiation arguments.
1141 std::vector<std::vector<AsmToken> > MacroArguments;
1142 MacroArguments.push_back(std::vector<AsmToken>());
1143 unsigned ParenLevel = 0;
1145 if (Lexer.is(AsmToken::Eof))
1146 return TokError("unexpected token in macro instantiation");
1147 if (Lexer.is(AsmToken::EndOfStatement))
1150 // If we aren't inside parentheses and this is a comma, start a new token
1152 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1153 MacroArguments.push_back(std::vector<AsmToken>());
1155 // Adjust the current parentheses level.
1156 if (Lexer.is(AsmToken::LParen))
1158 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1161 // Append the token to the current argument list.
1162 MacroArguments.back().push_back(getTok());
1167 // Create the macro instantiation object and add to the current macro
1168 // instantiation stack.
1169 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1172 ActiveMacros.push_back(MI);
1174 // Jump to the macro instantiation and prime the lexer.
1175 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1176 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1182 void AsmParser::HandleMacroExit() {
1183 // Jump to the EndOfStatement we should return to, and consume it.
1184 JumpToLoc(ActiveMacros.back()->ExitLoc);
1187 // Pop the instantiation entry.
1188 delete ActiveMacros.back();
1189 ActiveMacros.pop_back();
1192 bool AsmParser::ParseAssignment(StringRef Name) {
1193 // FIXME: Use better location, we should use proper tokens.
1194 SMLoc EqualLoc = Lexer.getLoc();
1196 const MCExpr *Value;
1197 if (ParseExpression(Value))
1200 if (Lexer.isNot(AsmToken::EndOfStatement))
1201 return TokError("unexpected token in assignment");
1203 // Eat the end of statement marker.
1206 // Validate that the LHS is allowed to be a variable (either it has not been
1207 // used as a symbol, or it is an absolute symbol).
1208 MCSymbol *Sym = getContext().LookupSymbol(Name);
1210 // Diagnose assignment to a label.
1212 // FIXME: Diagnostics. Note the location of the definition as a label.
1213 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1214 if (Sym->isUndefined() && !Sym->isUsedInExpr())
1215 ; // Allow redefinitions of undefined symbols only used in directives.
1216 else if (!Sym->isUndefined() && !Sym->isAbsolute())
1217 return Error(EqualLoc, "redefinition of '" + Name + "'");
1218 else if (!Sym->isVariable())
1219 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1220 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1221 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1224 Sym = getContext().GetOrCreateSymbol(Name);
1226 // FIXME: Handle '.'.
1228 Sym->setUsedInExpr(true);
1230 // Do the assignment.
1231 Out.EmitAssignment(Sym, Value);
1236 /// ParseIdentifier:
1239 bool AsmParser::ParseIdentifier(StringRef &Res) {
1240 // The assembler has relaxed rules for accepting identifiers, in particular we
1241 // allow things like '.globl $foo', which would normally be separate
1242 // tokens. At this level, we have already lexed so we cannot (currently)
1243 // handle this as a context dependent token, instead we detect adjacent tokens
1244 // and return the combined identifier.
1245 if (Lexer.is(AsmToken::Dollar)) {
1246 SMLoc DollarLoc = getLexer().getLoc();
1248 // Consume the dollar sign, and check for a following identifier.
1250 if (Lexer.isNot(AsmToken::Identifier))
1253 // We have a '$' followed by an identifier, make sure they are adjacent.
1254 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1257 // Construct the joined identifier and consume the token.
1258 Res = StringRef(DollarLoc.getPointer(),
1259 getTok().getIdentifier().size() + 1);
1264 if (Lexer.isNot(AsmToken::Identifier) &&
1265 Lexer.isNot(AsmToken::String))
1268 Res = getTok().getIdentifier();
1270 Lex(); // Consume the identifier token.
1275 /// ParseDirectiveSet:
1276 /// ::= .set identifier ',' expression
1277 bool AsmParser::ParseDirectiveSet() {
1280 if (ParseIdentifier(Name))
1281 return TokError("expected identifier after '.set' directive");
1283 if (getLexer().isNot(AsmToken::Comma))
1284 return TokError("unexpected token in '.set'");
1287 return ParseAssignment(Name);
1290 bool AsmParser::ParseEscapedString(std::string &Data) {
1291 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1294 StringRef Str = getTok().getStringContents();
1295 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1296 if (Str[i] != '\\') {
1301 // Recognize escaped characters. Note that this escape semantics currently
1302 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1305 return TokError("unexpected backslash at end of string");
1307 // Recognize octal sequences.
1308 if ((unsigned) (Str[i] - '0') <= 7) {
1309 // Consume up to three octal characters.
1310 unsigned Value = Str[i] - '0';
1312 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1314 Value = Value * 8 + (Str[i] - '0');
1316 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1318 Value = Value * 8 + (Str[i] - '0');
1323 return TokError("invalid octal escape sequence (out of range)");
1325 Data += (unsigned char) Value;
1329 // Otherwise recognize individual escapes.
1332 // Just reject invalid escape sequences for now.
1333 return TokError("invalid escape sequence (unrecognized character)");
1335 case 'b': Data += '\b'; break;
1336 case 'f': Data += '\f'; break;
1337 case 'n': Data += '\n'; break;
1338 case 'r': Data += '\r'; break;
1339 case 't': Data += '\t'; break;
1340 case '"': Data += '"'; break;
1341 case '\\': Data += '\\'; break;
1348 /// ParseDirectiveAscii:
1349 /// ::= ( .ascii | .asciz ) [ "string" ( , "string" )* ]
1350 bool AsmParser::ParseDirectiveAscii(bool ZeroTerminated) {
1351 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1352 CheckForValidSection();
1355 if (getLexer().isNot(AsmToken::String))
1356 return TokError("expected string in '.ascii' or '.asciz' directive");
1359 if (ParseEscapedString(Data))
1362 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1364 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1368 if (getLexer().is(AsmToken::EndOfStatement))
1371 if (getLexer().isNot(AsmToken::Comma))
1372 return TokError("unexpected token in '.ascii' or '.asciz' directive");
1381 /// ParseDirectiveValue
1382 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1383 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1384 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1385 CheckForValidSection();
1388 const MCExpr *Value;
1389 SMLoc ATTRIBUTE_UNUSED StartLoc = getLexer().getLoc();
1390 if (ParseExpression(Value))
1393 // Special case constant expressions to match code generator.
1394 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1395 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1397 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1399 if (getLexer().is(AsmToken::EndOfStatement))
1402 // FIXME: Improve diagnostic.
1403 if (getLexer().isNot(AsmToken::Comma))
1404 return TokError("unexpected token in directive");
1413 /// ParseDirectiveRealValue
1414 /// ::= (.single | .double) [ expression (, expression)* ]
1415 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1416 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1417 CheckForValidSection();
1420 // We don't truly support arithmetic on floating point expressions, so we
1421 // have to manually parse unary prefixes.
1423 if (getLexer().is(AsmToken::Minus)) {
1426 } else if (getLexer().is(AsmToken::Plus))
1429 if (getLexer().isNot(AsmToken::Integer) &&
1430 getLexer().isNot(AsmToken::Real))
1431 return TokError("unexpected token in directive");
1433 // Convert to an APFloat.
1434 APFloat Value(Semantics);
1435 if (Value.convertFromString(getTok().getString(),
1436 APFloat::rmNearestTiesToEven) ==
1437 APFloat::opInvalidOp)
1438 return TokError("invalid floating point literal");
1442 // Consume the numeric token.
1445 // Emit the value as an integer.
1446 APInt AsInt = Value.bitcastToAPInt();
1447 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1448 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1450 if (getLexer().is(AsmToken::EndOfStatement))
1453 if (getLexer().isNot(AsmToken::Comma))
1454 return TokError("unexpected token in directive");
1463 /// ParseDirectiveSpace
1464 /// ::= .space expression [ , expression ]
1465 bool AsmParser::ParseDirectiveSpace() {
1466 CheckForValidSection();
1469 if (ParseAbsoluteExpression(NumBytes))
1472 int64_t FillExpr = 0;
1473 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1474 if (getLexer().isNot(AsmToken::Comma))
1475 return TokError("unexpected token in '.space' directive");
1478 if (ParseAbsoluteExpression(FillExpr))
1481 if (getLexer().isNot(AsmToken::EndOfStatement))
1482 return TokError("unexpected token in '.space' directive");
1488 return TokError("invalid number of bytes in '.space' directive");
1490 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1491 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1496 /// ParseDirectiveZero
1497 /// ::= .zero expression
1498 bool AsmParser::ParseDirectiveZero() {
1499 CheckForValidSection();
1502 if (ParseAbsoluteExpression(NumBytes))
1506 if (getLexer().is(AsmToken::Comma)) {
1508 if (ParseAbsoluteExpression(Val))
1512 if (getLexer().isNot(AsmToken::EndOfStatement))
1513 return TokError("unexpected token in '.zero' directive");
1517 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
1522 /// ParseDirectiveFill
1523 /// ::= .fill expression , expression , expression
1524 bool AsmParser::ParseDirectiveFill() {
1525 CheckForValidSection();
1528 if (ParseAbsoluteExpression(NumValues))
1531 if (getLexer().isNot(AsmToken::Comma))
1532 return TokError("unexpected token in '.fill' directive");
1536 if (ParseAbsoluteExpression(FillSize))
1539 if (getLexer().isNot(AsmToken::Comma))
1540 return TokError("unexpected token in '.fill' directive");
1544 if (ParseAbsoluteExpression(FillExpr))
1547 if (getLexer().isNot(AsmToken::EndOfStatement))
1548 return TokError("unexpected token in '.fill' directive");
1552 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1553 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1555 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1556 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1561 /// ParseDirectiveOrg
1562 /// ::= .org expression [ , expression ]
1563 bool AsmParser::ParseDirectiveOrg() {
1564 CheckForValidSection();
1566 const MCExpr *Offset;
1567 if (ParseExpression(Offset))
1570 // Parse optional fill expression.
1571 int64_t FillExpr = 0;
1572 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1573 if (getLexer().isNot(AsmToken::Comma))
1574 return TokError("unexpected token in '.org' directive");
1577 if (ParseAbsoluteExpression(FillExpr))
1580 if (getLexer().isNot(AsmToken::EndOfStatement))
1581 return TokError("unexpected token in '.org' directive");
1586 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1587 // has to be relative to the current section.
1588 getStreamer().EmitValueToOffset(Offset, FillExpr);
1593 /// ParseDirectiveAlign
1594 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1595 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1596 CheckForValidSection();
1598 SMLoc AlignmentLoc = getLexer().getLoc();
1600 if (ParseAbsoluteExpression(Alignment))
1604 bool HasFillExpr = false;
1605 int64_t FillExpr = 0;
1606 int64_t MaxBytesToFill = 0;
1607 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1608 if (getLexer().isNot(AsmToken::Comma))
1609 return TokError("unexpected token in directive");
1612 // The fill expression can be omitted while specifying a maximum number of
1613 // alignment bytes, e.g:
1615 if (getLexer().isNot(AsmToken::Comma)) {
1617 if (ParseAbsoluteExpression(FillExpr))
1621 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1622 if (getLexer().isNot(AsmToken::Comma))
1623 return TokError("unexpected token in directive");
1626 MaxBytesLoc = getLexer().getLoc();
1627 if (ParseAbsoluteExpression(MaxBytesToFill))
1630 if (getLexer().isNot(AsmToken::EndOfStatement))
1631 return TokError("unexpected token in directive");
1640 // Compute alignment in bytes.
1642 // FIXME: Diagnose overflow.
1643 if (Alignment >= 32) {
1644 Error(AlignmentLoc, "invalid alignment value");
1648 Alignment = 1ULL << Alignment;
1651 // Diagnose non-sensical max bytes to align.
1652 if (MaxBytesLoc.isValid()) {
1653 if (MaxBytesToFill < 1) {
1654 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1655 "many bytes, ignoring maximum bytes expression");
1659 if (MaxBytesToFill >= Alignment) {
1660 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1666 // Check whether we should use optimal code alignment for this .align
1668 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
1669 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1670 ValueSize == 1 && UseCodeAlign) {
1671 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1673 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1674 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1681 /// ParseDirectiveSymbolAttribute
1682 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1683 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1684 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1688 if (ParseIdentifier(Name))
1689 return TokError("expected identifier in directive");
1691 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1693 getStreamer().EmitSymbolAttribute(Sym, Attr);
1695 if (getLexer().is(AsmToken::EndOfStatement))
1698 if (getLexer().isNot(AsmToken::Comma))
1699 return TokError("unexpected token in directive");
1708 /// ParseDirectiveComm
1709 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1710 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1711 CheckForValidSection();
1713 SMLoc IDLoc = getLexer().getLoc();
1715 if (ParseIdentifier(Name))
1716 return TokError("expected identifier in directive");
1718 // Handle the identifier as the key symbol.
1719 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1721 if (getLexer().isNot(AsmToken::Comma))
1722 return TokError("unexpected token in directive");
1726 SMLoc SizeLoc = getLexer().getLoc();
1727 if (ParseAbsoluteExpression(Size))
1730 int64_t Pow2Alignment = 0;
1731 SMLoc Pow2AlignmentLoc;
1732 if (getLexer().is(AsmToken::Comma)) {
1734 Pow2AlignmentLoc = getLexer().getLoc();
1735 if (ParseAbsoluteExpression(Pow2Alignment))
1738 // If this target takes alignments in bytes (not log) validate and convert.
1739 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1740 if (!isPowerOf2_64(Pow2Alignment))
1741 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1742 Pow2Alignment = Log2_64(Pow2Alignment);
1746 if (getLexer().isNot(AsmToken::EndOfStatement))
1747 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1751 // NOTE: a size of zero for a .comm should create a undefined symbol
1752 // but a size of .lcomm creates a bss symbol of size zero.
1754 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1755 "be less than zero");
1757 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1758 // may internally end up wanting an alignment in bytes.
1759 // FIXME: Diagnose overflow.
1760 if (Pow2Alignment < 0)
1761 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1762 "alignment, can't be less than zero");
1764 if (!Sym->isUndefined())
1765 return Error(IDLoc, "invalid symbol redefinition");
1767 // '.lcomm' is equivalent to '.zerofill'.
1768 // Create the Symbol as a common or local common with Size and Pow2Alignment
1770 getStreamer().EmitZerofill(Ctx.getMachOSection(
1771 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1772 0, SectionKind::getBSS()),
1773 Sym, Size, 1 << Pow2Alignment);
1777 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1781 /// ParseDirectiveAbort
1782 /// ::= .abort [... message ...]
1783 bool AsmParser::ParseDirectiveAbort() {
1784 // FIXME: Use loc from directive.
1785 SMLoc Loc = getLexer().getLoc();
1787 StringRef Str = ParseStringToEndOfStatement();
1788 if (getLexer().isNot(AsmToken::EndOfStatement))
1789 return TokError("unexpected token in '.abort' directive");
1794 Error(Loc, ".abort detected. Assembly stopping.");
1796 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1797 // FIXME: Actually abort assembly here.
1802 /// ParseDirectiveInclude
1803 /// ::= .include "filename"
1804 bool AsmParser::ParseDirectiveInclude() {
1805 if (getLexer().isNot(AsmToken::String))
1806 return TokError("expected string in '.include' directive");
1808 std::string Filename = getTok().getString();
1809 SMLoc IncludeLoc = getLexer().getLoc();
1812 if (getLexer().isNot(AsmToken::EndOfStatement))
1813 return TokError("unexpected token in '.include' directive");
1815 // Strip the quotes.
1816 Filename = Filename.substr(1, Filename.size()-2);
1818 // Attempt to switch the lexer to the included file before consuming the end
1819 // of statement to avoid losing it when we switch.
1820 if (EnterIncludeFile(Filename)) {
1821 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1828 /// ParseDirectiveIf
1829 /// ::= .if expression
1830 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1831 TheCondStack.push_back(TheCondState);
1832 TheCondState.TheCond = AsmCond::IfCond;
1833 if(TheCondState.Ignore) {
1834 EatToEndOfStatement();
1838 if (ParseAbsoluteExpression(ExprValue))
1841 if (getLexer().isNot(AsmToken::EndOfStatement))
1842 return TokError("unexpected token in '.if' directive");
1846 TheCondState.CondMet = ExprValue;
1847 TheCondState.Ignore = !TheCondState.CondMet;
1853 /// ParseDirectiveElseIf
1854 /// ::= .elseif expression
1855 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1856 if (TheCondState.TheCond != AsmCond::IfCond &&
1857 TheCondState.TheCond != AsmCond::ElseIfCond)
1858 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1860 TheCondState.TheCond = AsmCond::ElseIfCond;
1862 bool LastIgnoreState = false;
1863 if (!TheCondStack.empty())
1864 LastIgnoreState = TheCondStack.back().Ignore;
1865 if (LastIgnoreState || TheCondState.CondMet) {
1866 TheCondState.Ignore = true;
1867 EatToEndOfStatement();
1871 if (ParseAbsoluteExpression(ExprValue))
1874 if (getLexer().isNot(AsmToken::EndOfStatement))
1875 return TokError("unexpected token in '.elseif' directive");
1878 TheCondState.CondMet = ExprValue;
1879 TheCondState.Ignore = !TheCondState.CondMet;
1885 /// ParseDirectiveElse
1887 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
1888 if (getLexer().isNot(AsmToken::EndOfStatement))
1889 return TokError("unexpected token in '.else' directive");
1893 if (TheCondState.TheCond != AsmCond::IfCond &&
1894 TheCondState.TheCond != AsmCond::ElseIfCond)
1895 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
1897 TheCondState.TheCond = AsmCond::ElseCond;
1898 bool LastIgnoreState = false;
1899 if (!TheCondStack.empty())
1900 LastIgnoreState = TheCondStack.back().Ignore;
1901 if (LastIgnoreState || TheCondState.CondMet)
1902 TheCondState.Ignore = true;
1904 TheCondState.Ignore = false;
1909 /// ParseDirectiveEndIf
1911 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
1912 if (getLexer().isNot(AsmToken::EndOfStatement))
1913 return TokError("unexpected token in '.endif' directive");
1917 if ((TheCondState.TheCond == AsmCond::NoCond) ||
1918 TheCondStack.empty())
1919 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
1921 if (!TheCondStack.empty()) {
1922 TheCondState = TheCondStack.back();
1923 TheCondStack.pop_back();
1929 /// ParseDirectiveFile
1930 /// ::= .file [number] string
1931 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
1932 // FIXME: I'm not sure what this is.
1933 int64_t FileNumber = -1;
1934 SMLoc FileNumberLoc = getLexer().getLoc();
1935 if (getLexer().is(AsmToken::Integer)) {
1936 FileNumber = getTok().getIntVal();
1940 return TokError("file number less than one");
1943 if (getLexer().isNot(AsmToken::String))
1944 return TokError("unexpected token in '.file' directive");
1946 StringRef Filename = getTok().getString();
1947 Filename = Filename.substr(1, Filename.size()-2);
1950 if (getLexer().isNot(AsmToken::EndOfStatement))
1951 return TokError("unexpected token in '.file' directive");
1953 if (FileNumber == -1)
1954 getStreamer().EmitFileDirective(Filename);
1956 if (getContext().GetDwarfFile(Filename, FileNumber) == 0)
1957 Error(FileNumberLoc, "file number already allocated");
1958 getStreamer().EmitDwarfFileDirective(FileNumber, Filename);
1964 /// ParseDirectiveLine
1965 /// ::= .line [number]
1966 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
1967 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1968 if (getLexer().isNot(AsmToken::Integer))
1969 return TokError("unexpected token in '.line' directive");
1971 int64_t LineNumber = getTok().getIntVal();
1975 // FIXME: Do something with the .line.
1978 if (getLexer().isNot(AsmToken::EndOfStatement))
1979 return TokError("unexpected token in '.line' directive");
1985 /// ParseDirectiveLoc
1986 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
1987 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
1988 /// The first number is a file number, must have been previously assigned with
1989 /// a .file directive, the second number is the line number and optionally the
1990 /// third number is a column position (zero if not specified). The remaining
1991 /// optional items are .loc sub-directives.
1992 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
1994 if (getLexer().isNot(AsmToken::Integer))
1995 return TokError("unexpected token in '.loc' directive");
1996 int64_t FileNumber = getTok().getIntVal();
1998 return TokError("file number less than one in '.loc' directive");
1999 if (!getContext().isValidDwarfFileNumber(FileNumber))
2000 return TokError("unassigned file number in '.loc' directive");
2003 int64_t LineNumber = 0;
2004 if (getLexer().is(AsmToken::Integer)) {
2005 LineNumber = getTok().getIntVal();
2007 return TokError("line number less than one in '.loc' directive");
2011 int64_t ColumnPos = 0;
2012 if (getLexer().is(AsmToken::Integer)) {
2013 ColumnPos = getTok().getIntVal();
2015 return TokError("column position less than zero in '.loc' directive");
2019 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2021 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2023 if (getLexer().is(AsmToken::EndOfStatement))
2027 SMLoc Loc = getTok().getLoc();
2028 if (getParser().ParseIdentifier(Name))
2029 return TokError("unexpected token in '.loc' directive");
2031 if (Name == "basic_block")
2032 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2033 else if (Name == "prologue_end")
2034 Flags |= DWARF2_FLAG_PROLOGUE_END;
2035 else if (Name == "epilogue_begin")
2036 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2037 else if (Name == "is_stmt") {
2038 SMLoc Loc = getTok().getLoc();
2039 const MCExpr *Value;
2040 if (getParser().ParseExpression(Value))
2042 // The expression must be the constant 0 or 1.
2043 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2044 int Value = MCE->getValue();
2046 Flags &= ~DWARF2_FLAG_IS_STMT;
2047 else if (Value == 1)
2048 Flags |= DWARF2_FLAG_IS_STMT;
2050 return Error(Loc, "is_stmt value not 0 or 1");
2053 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2056 else if (Name == "isa") {
2057 SMLoc Loc = getTok().getLoc();
2058 const MCExpr *Value;
2059 if (getParser().ParseExpression(Value))
2061 // The expression must be a constant greater or equal to 0.
2062 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2063 int Value = MCE->getValue();
2065 return Error(Loc, "isa number less than zero");
2069 return Error(Loc, "isa number not a constant value");
2073 return Error(Loc, "unknown sub-directive in '.loc' directive");
2076 if (getLexer().is(AsmToken::EndOfStatement))
2081 getContext().setCurrentDwarfLoc(FileNumber, LineNumber, ColumnPos, Flags,Isa);
2086 /// ParseDirectiveStabs
2087 /// ::= .stabs string, number, number, number
2088 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2089 SMLoc DirectiveLoc) {
2090 return TokError("unsupported directive '" + Directive + "'");
2093 /// ParseDirectiveMacrosOnOff
2096 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2097 SMLoc DirectiveLoc) {
2098 if (getLexer().isNot(AsmToken::EndOfStatement))
2099 return Error(getLexer().getLoc(),
2100 "unexpected token in '" + Directive + "' directive");
2102 getParser().MacrosEnabled = Directive == ".macros_on";
2107 /// ParseDirectiveMacro
2109 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2110 SMLoc DirectiveLoc) {
2112 if (getParser().ParseIdentifier(Name))
2113 return TokError("expected identifier in directive");
2115 if (getLexer().isNot(AsmToken::EndOfStatement))
2116 return TokError("unexpected token in '.macro' directive");
2118 // Eat the end of statement.
2121 AsmToken EndToken, StartToken = getTok();
2123 // Lex the macro definition.
2125 // Check whether we have reached the end of the file.
2126 if (getLexer().is(AsmToken::Eof))
2127 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2129 // Otherwise, check whether we have reach the .endmacro.
2130 if (getLexer().is(AsmToken::Identifier) &&
2131 (getTok().getIdentifier() == ".endm" ||
2132 getTok().getIdentifier() == ".endmacro")) {
2133 EndToken = getTok();
2135 if (getLexer().isNot(AsmToken::EndOfStatement))
2136 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2141 // Otherwise, scan til the end of the statement.
2142 getParser().EatToEndOfStatement();
2145 if (getParser().MacroMap.lookup(Name)) {
2146 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2149 const char *BodyStart = StartToken.getLoc().getPointer();
2150 const char *BodyEnd = EndToken.getLoc().getPointer();
2151 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2152 getParser().MacroMap[Name] = new Macro(Name, Body);
2156 /// ParseDirectiveEndMacro
2159 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2160 SMLoc DirectiveLoc) {
2161 if (getLexer().isNot(AsmToken::EndOfStatement))
2162 return TokError("unexpected token in '" + Directive + "' directive");
2164 // If we are inside a macro instantiation, terminate the current
2166 if (!getParser().ActiveMacros.empty()) {
2167 getParser().HandleMacroExit();
2171 // Otherwise, this .endmacro is a stray entry in the file; well formed
2172 // .endmacro directives are handled during the macro definition parsing.
2173 return TokError("unexpected '" + Directive + "' in file, "
2174 "no current macro definition");
2177 void GenericAsmParser::ParseUleb128(uint64_t Value) {
2178 const uint64_t Mask = (1 << 7) - 1;
2180 unsigned Byte = Value & Mask;
2182 if (Value) // Not the last one
2184 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2188 void GenericAsmParser::ParseSleb128(int64_t Value) {
2189 const int64_t Mask = (1 << 7) - 1;
2191 unsigned Byte = Value & Mask;
2193 bool Done = ((Value == 0 && (Byte & 0x40) == 0) ||
2194 (Value == -1 && (Byte & 0x40) != 0));
2197 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2203 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2205 if (getParser().ParseAbsoluteExpression(Value))
2208 if (getLexer().isNot(AsmToken::EndOfStatement))
2209 return TokError("unexpected token in directive");
2211 if (DirName[1] == 's')
2212 ParseSleb128(Value);
2214 ParseUleb128(Value);
2219 /// \brief Create an MCAsmParser instance.
2220 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2221 MCContext &C, MCStreamer &Out,
2222 const MCAsmInfo &MAI) {
2223 return new AsmParser(T, SM, C, Out, MAI);