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");
241 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
243 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
245 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
246 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
247 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
249 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
250 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
253 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
254 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
255 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
257 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
258 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
259 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
261 void ParseUleb128(uint64_t Value);
262 void ParseSleb128(int64_t Value);
263 bool ParseDirectiveLEB128(StringRef, SMLoc);
270 extern MCAsmParserExtension *createDarwinAsmParser();
271 extern MCAsmParserExtension *createELFAsmParser();
275 enum { DEFAULT_ADDRSPACE = 0 };
277 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
278 MCStreamer &_Out, const MCAsmInfo &_MAI)
279 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
280 GenericParser(new GenericAsmParser), PlatformParser(0),
281 CurBuffer(0), MacrosEnabled(true) {
282 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
284 // Initialize the generic parser.
285 GenericParser->Initialize(*this);
287 // Initialize the platform / file format parser.
289 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
291 if (_MAI.hasSubsectionsViaSymbols()) {
292 PlatformParser = createDarwinAsmParser();
293 PlatformParser->Initialize(*this);
295 PlatformParser = createELFAsmParser();
296 PlatformParser->Initialize(*this);
300 AsmParser::~AsmParser() {
301 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
303 // Destroy any macros.
304 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
305 ie = MacroMap.end(); it != ie; ++it)
306 delete it->getValue();
308 delete PlatformParser;
309 delete GenericParser;
312 void AsmParser::PrintMacroInstantiations() {
313 // Print the active macro instantiation stack.
314 for (std::vector<MacroInstantiation*>::const_reverse_iterator
315 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
316 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
320 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
321 PrintMessage(L, Msg, "warning");
322 PrintMacroInstantiations();
325 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
327 PrintMessage(L, Msg, "error");
328 PrintMacroInstantiations();
332 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
333 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
339 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
344 void AsmParser::JumpToLoc(SMLoc Loc) {
345 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
346 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
349 const AsmToken &AsmParser::Lex() {
350 const AsmToken *tok = &Lexer.Lex();
352 if (tok->is(AsmToken::Eof)) {
353 // If this is the end of an included file, pop the parent file off the
355 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
356 if (ParentIncludeLoc != SMLoc()) {
357 JumpToLoc(ParentIncludeLoc);
362 if (tok->is(AsmToken::Error))
363 Error(Lexer.getErrLoc(), Lexer.getErr());
368 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
369 // Create the initial section, if requested.
370 if (!NoInitialTextSection)
377 AsmCond StartingCondState = TheCondState;
379 // While we have input, parse each statement.
380 while (Lexer.isNot(AsmToken::Eof)) {
381 if (!ParseStatement()) continue;
383 // We had an error, validate that one was emitted and recover by skipping to
385 assert(HadError && "Parse statement returned an error, but none emitted!");
386 EatToEndOfStatement();
389 if (TheCondState.TheCond != StartingCondState.TheCond ||
390 TheCondState.Ignore != StartingCondState.Ignore)
391 return TokError("unmatched .ifs or .elses");
393 // Check to see there are no empty DwarfFile slots.
394 const std::vector<MCDwarfFile *> &MCDwarfFiles =
395 getContext().getMCDwarfFiles();
396 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
397 if (!MCDwarfFiles[i])
398 TokError("unassigned file number: " + Twine(i) + " for .file directives");
401 // Finalize the output stream if there are no errors and if the client wants
403 if (!HadError && !NoFinalize)
409 void AsmParser::CheckForValidSection() {
410 if (!getStreamer().getCurrentSection()) {
411 TokError("expected section directive before assembly directive");
412 Out.SwitchSection(Ctx.getMachOSection(
414 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
415 0, SectionKind::getText()));
419 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
420 void AsmParser::EatToEndOfStatement() {
421 while (Lexer.isNot(AsmToken::EndOfStatement) &&
422 Lexer.isNot(AsmToken::Eof))
426 if (Lexer.is(AsmToken::EndOfStatement))
430 StringRef AsmParser::ParseStringToEndOfStatement() {
431 const char *Start = getTok().getLoc().getPointer();
433 while (Lexer.isNot(AsmToken::EndOfStatement) &&
434 Lexer.isNot(AsmToken::Eof))
437 const char *End = getTok().getLoc().getPointer();
438 return StringRef(Start, End - Start);
441 /// ParseParenExpr - Parse a paren expression and return it.
442 /// NOTE: This assumes the leading '(' has already been consumed.
444 /// parenexpr ::= expr)
446 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
447 if (ParseExpression(Res)) return true;
448 if (Lexer.isNot(AsmToken::RParen))
449 return TokError("expected ')' in parentheses expression");
450 EndLoc = Lexer.getLoc();
455 /// ParsePrimaryExpr - Parse a primary expression and return it.
456 /// primaryexpr ::= (parenexpr
457 /// primaryexpr ::= symbol
458 /// primaryexpr ::= number
459 /// primaryexpr ::= '.'
460 /// primaryexpr ::= ~,+,- primaryexpr
461 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
462 switch (Lexer.getKind()) {
464 return TokError("unknown token in expression");
465 case AsmToken::Exclaim:
466 Lex(); // Eat the operator.
467 if (ParsePrimaryExpr(Res, EndLoc))
469 Res = MCUnaryExpr::CreateLNot(Res, getContext());
471 case AsmToken::Dollar:
472 case AsmToken::String:
473 case AsmToken::Identifier: {
474 EndLoc = Lexer.getLoc();
476 StringRef Identifier;
477 if (ParseIdentifier(Identifier))
480 // This is a symbol reference.
481 std::pair<StringRef, StringRef> Split = Identifier.split('@');
482 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
484 // Mark the symbol as used in an expression.
485 Sym->setUsedInExpr(true);
487 // Lookup the symbol variant if used.
488 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
489 if (Split.first.size() != Identifier.size()) {
490 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
491 if (Variant == MCSymbolRefExpr::VK_Invalid) {
492 Variant = MCSymbolRefExpr::VK_None;
493 TokError("invalid variant '" + Split.second + "'");
497 // If this is an absolute variable reference, substitute it now to preserve
498 // semantics in the face of reassignment.
499 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
501 return Error(EndLoc, "unexpected modified on variable reference");
503 Res = Sym->getVariableValue();
507 // Otherwise create a symbol ref.
508 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
511 case AsmToken::Integer: {
512 SMLoc Loc = getTok().getLoc();
513 int64_t IntVal = getTok().getIntVal();
514 Res = MCConstantExpr::Create(IntVal, getContext());
515 EndLoc = Lexer.getLoc();
517 // Look for 'b' or 'f' following an Integer as a directional label
518 if (Lexer.getKind() == AsmToken::Identifier) {
519 StringRef IDVal = getTok().getString();
520 if (IDVal == "f" || IDVal == "b"){
521 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
522 IDVal == "f" ? 1 : 0);
523 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
525 if(IDVal == "b" && Sym->isUndefined())
526 return Error(Loc, "invalid reference to undefined symbol");
527 EndLoc = Lexer.getLoc();
528 Lex(); // Eat identifier.
533 case AsmToken::Dot: {
534 // This is a '.' reference, which references the current PC. Emit a
535 // temporary label to the streamer and refer to it.
536 MCSymbol *Sym = Ctx.CreateTempSymbol();
538 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
539 EndLoc = Lexer.getLoc();
540 Lex(); // Eat identifier.
544 case AsmToken::LParen:
545 Lex(); // Eat the '('.
546 return ParseParenExpr(Res, EndLoc);
547 case AsmToken::Minus:
548 Lex(); // Eat the operator.
549 if (ParsePrimaryExpr(Res, EndLoc))
551 Res = MCUnaryExpr::CreateMinus(Res, getContext());
554 Lex(); // Eat the operator.
555 if (ParsePrimaryExpr(Res, EndLoc))
557 Res = MCUnaryExpr::CreatePlus(Res, getContext());
559 case AsmToken::Tilde:
560 Lex(); // Eat the operator.
561 if (ParsePrimaryExpr(Res, EndLoc))
563 Res = MCUnaryExpr::CreateNot(Res, getContext());
568 bool AsmParser::ParseExpression(const MCExpr *&Res) {
570 return ParseExpression(Res, EndLoc);
574 AsmParser::ApplyModifierToExpr(const MCExpr *E,
575 MCSymbolRefExpr::VariantKind Variant) {
576 // Recurse over the given expression, rebuilding it to apply the given variant
577 // if there is exactly one symbol.
578 switch (E->getKind()) {
580 case MCExpr::Constant:
583 case MCExpr::SymbolRef: {
584 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
586 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
587 TokError("invalid variant on expression '" +
588 getTok().getIdentifier() + "' (already modified)");
592 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
595 case MCExpr::Unary: {
596 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
597 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
600 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
603 case MCExpr::Binary: {
604 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
605 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
606 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
611 if (!LHS) LHS = BE->getLHS();
612 if (!RHS) RHS = BE->getRHS();
614 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
618 assert(0 && "Invalid expression kind!");
622 /// ParseExpression - Parse an expression and return it.
624 /// expr ::= expr +,- expr -> lowest.
625 /// expr ::= expr |,^,&,! expr -> middle.
626 /// expr ::= expr *,/,%,<<,>> expr -> highest.
627 /// expr ::= primaryexpr
629 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
630 // Parse the expression.
632 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
635 // As a special case, we support 'a op b @ modifier' by rewriting the
636 // expression to include the modifier. This is inefficient, but in general we
637 // expect users to use 'a@modifier op b'.
638 if (Lexer.getKind() == AsmToken::At) {
641 if (Lexer.isNot(AsmToken::Identifier))
642 return TokError("unexpected symbol modifier following '@'");
644 MCSymbolRefExpr::VariantKind Variant =
645 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
646 if (Variant == MCSymbolRefExpr::VK_Invalid)
647 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
649 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
651 return TokError("invalid modifier '" + getTok().getIdentifier() +
652 "' (no symbols present)");
660 // Try to constant fold it up front, if possible.
662 if (Res->EvaluateAsAbsolute(Value))
663 Res = MCConstantExpr::Create(Value, getContext());
668 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
670 return ParseParenExpr(Res, EndLoc) ||
671 ParseBinOpRHS(1, Res, EndLoc);
674 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
677 SMLoc StartLoc = Lexer.getLoc();
678 if (ParseExpression(Expr))
681 if (!Expr->EvaluateAsAbsolute(Res))
682 return Error(StartLoc, "expected absolute expression");
687 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
688 MCBinaryExpr::Opcode &Kind) {
691 return 0; // not a binop.
693 // Lowest Precedence: &&, ||, @
694 case AsmToken::AmpAmp:
695 Kind = MCBinaryExpr::LAnd;
697 case AsmToken::PipePipe:
698 Kind = MCBinaryExpr::LOr;
702 // Low Precedence: |, &, ^
704 // FIXME: gas seems to support '!' as an infix operator?
706 Kind = MCBinaryExpr::Or;
708 case AsmToken::Caret:
709 Kind = MCBinaryExpr::Xor;
712 Kind = MCBinaryExpr::And;
715 // Intermediate Precedence: +, -, ==, !=, <>, <, <=, >, >=
717 Kind = MCBinaryExpr::Add;
719 case AsmToken::Minus:
720 Kind = MCBinaryExpr::Sub;
722 case AsmToken::EqualEqual:
723 Kind = MCBinaryExpr::EQ;
725 case AsmToken::ExclaimEqual:
726 case AsmToken::LessGreater:
727 Kind = MCBinaryExpr::NE;
730 Kind = MCBinaryExpr::LT;
732 case AsmToken::LessEqual:
733 Kind = MCBinaryExpr::LTE;
735 case AsmToken::Greater:
736 Kind = MCBinaryExpr::GT;
738 case AsmToken::GreaterEqual:
739 Kind = MCBinaryExpr::GTE;
742 // Highest Precedence: *, /, %, <<, >>
744 Kind = MCBinaryExpr::Mul;
746 case AsmToken::Slash:
747 Kind = MCBinaryExpr::Div;
749 case AsmToken::Percent:
750 Kind = MCBinaryExpr::Mod;
752 case AsmToken::LessLess:
753 Kind = MCBinaryExpr::Shl;
755 case AsmToken::GreaterGreater:
756 Kind = MCBinaryExpr::Shr;
762 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
763 /// Res contains the LHS of the expression on input.
764 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
767 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
768 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
770 // If the next token is lower precedence than we are allowed to eat, return
771 // successfully with what we ate already.
772 if (TokPrec < Precedence)
777 // Eat the next primary expression.
779 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
781 // If BinOp binds less tightly with RHS than the operator after RHS, let
782 // the pending operator take RHS as its LHS.
783 MCBinaryExpr::Opcode Dummy;
784 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
785 if (TokPrec < NextTokPrec) {
786 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
789 // Merge LHS and RHS according to operator.
790 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
798 /// ::= EndOfStatement
799 /// ::= Label* Directive ...Operands... EndOfStatement
800 /// ::= Label* Identifier OperandList* EndOfStatement
801 bool AsmParser::ParseStatement() {
802 if (Lexer.is(AsmToken::EndOfStatement)) {
808 // Statements always start with an identifier.
809 AsmToken ID = getTok();
810 SMLoc IDLoc = ID.getLoc();
812 int64_t LocalLabelVal = -1;
813 // GUESS allow an integer followed by a ':' as a directional local label
814 if (Lexer.is(AsmToken::Integer)) {
815 LocalLabelVal = getTok().getIntVal();
816 if (LocalLabelVal < 0) {
817 if (!TheCondState.Ignore)
818 return TokError("unexpected token at start of statement");
822 IDVal = getTok().getString();
823 Lex(); // Consume the integer token to be used as an identifier token.
824 if (Lexer.getKind() != AsmToken::Colon) {
825 if (!TheCondState.Ignore)
826 return TokError("unexpected token at start of statement");
830 else if (ParseIdentifier(IDVal)) {
831 if (!TheCondState.Ignore)
832 return TokError("unexpected token at start of statement");
836 // Handle conditional assembly here before checking for skipping. We
837 // have to do this so that .endif isn't skipped in a ".if 0" block for
840 return ParseDirectiveIf(IDLoc);
841 if (IDVal == ".elseif")
842 return ParseDirectiveElseIf(IDLoc);
843 if (IDVal == ".else")
844 return ParseDirectiveElse(IDLoc);
845 if (IDVal == ".endif")
846 return ParseDirectiveEndIf(IDLoc);
848 // If we are in a ".if 0" block, ignore this statement.
849 if (TheCondState.Ignore) {
850 EatToEndOfStatement();
854 // FIXME: Recurse on local labels?
856 // See what kind of statement we have.
857 switch (Lexer.getKind()) {
858 case AsmToken::Colon: {
859 CheckForValidSection();
861 // identifier ':' -> Label.
864 // Diagnose attempt to use a variable as a label.
866 // FIXME: Diagnostics. Note the location of the definition as a label.
867 // FIXME: This doesn't diagnose assignment to a symbol which has been
868 // implicitly marked as external.
870 if (LocalLabelVal == -1)
871 Sym = getContext().GetOrCreateSymbol(IDVal);
873 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
874 if (!Sym->isUndefined() || Sym->isVariable())
875 return Error(IDLoc, "invalid symbol redefinition");
880 // Consume any end of statement token, if present, to avoid spurious
881 // AddBlankLine calls().
882 if (Lexer.is(AsmToken::EndOfStatement)) {
884 if (Lexer.is(AsmToken::Eof))
888 return ParseStatement();
891 case AsmToken::Equal:
892 // identifier '=' ... -> assignment statement
895 return ParseAssignment(IDVal);
897 default: // Normal instruction or directive.
901 // If macros are enabled, check to see if this is a macro instantiation.
903 if (const Macro *M = MacroMap.lookup(IDVal))
904 return HandleMacroEntry(IDVal, IDLoc, M);
906 // Otherwise, we have a normal instruction or directive.
907 if (IDVal[0] == '.') {
908 // Assembler features
910 return ParseDirectiveSet();
914 if (IDVal == ".ascii")
915 return ParseDirectiveAscii(false);
916 if (IDVal == ".asciz")
917 return ParseDirectiveAscii(true);
919 if (IDVal == ".byte")
920 return ParseDirectiveValue(1);
921 if (IDVal == ".short")
922 return ParseDirectiveValue(2);
923 if (IDVal == ".long")
924 return ParseDirectiveValue(4);
925 if (IDVal == ".quad")
926 return ParseDirectiveValue(8);
927 if (IDVal == ".single")
928 return ParseDirectiveRealValue(APFloat::IEEEsingle);
929 if (IDVal == ".double")
930 return ParseDirectiveRealValue(APFloat::IEEEdouble);
932 if (IDVal == ".align") {
933 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
934 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
936 if (IDVal == ".align32") {
937 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
938 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
940 if (IDVal == ".balign")
941 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
942 if (IDVal == ".balignw")
943 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
944 if (IDVal == ".balignl")
945 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
946 if (IDVal == ".p2align")
947 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
948 if (IDVal == ".p2alignw")
949 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
950 if (IDVal == ".p2alignl")
951 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
954 return ParseDirectiveOrg();
956 if (IDVal == ".fill")
957 return ParseDirectiveFill();
958 if (IDVal == ".space")
959 return ParseDirectiveSpace();
960 if (IDVal == ".zero")
961 return ParseDirectiveZero();
963 // Symbol attribute directives
965 if (IDVal == ".globl" || IDVal == ".global")
966 return ParseDirectiveSymbolAttribute(MCSA_Global);
967 // ELF only? Should it be here?
968 if (IDVal == ".local")
969 return ParseDirectiveSymbolAttribute(MCSA_Local);
970 if (IDVal == ".hidden")
971 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
972 if (IDVal == ".indirect_symbol")
973 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
974 if (IDVal == ".internal")
975 return ParseDirectiveSymbolAttribute(MCSA_Internal);
976 if (IDVal == ".lazy_reference")
977 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
978 if (IDVal == ".no_dead_strip")
979 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
980 if (IDVal == ".private_extern")
981 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
982 if (IDVal == ".protected")
983 return ParseDirectiveSymbolAttribute(MCSA_Protected);
984 if (IDVal == ".reference")
985 return ParseDirectiveSymbolAttribute(MCSA_Reference);
986 if (IDVal == ".type")
987 return ParseDirectiveELFType();
988 if (IDVal == ".weak")
989 return ParseDirectiveSymbolAttribute(MCSA_Weak);
990 if (IDVal == ".weak_definition")
991 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
992 if (IDVal == ".weak_reference")
993 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
994 if (IDVal == ".weak_def_can_be_hidden")
995 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
997 if (IDVal == ".comm")
998 return ParseDirectiveComm(/*IsLocal=*/false);
999 if (IDVal == ".lcomm")
1000 return ParseDirectiveComm(/*IsLocal=*/true);
1002 if (IDVal == ".abort")
1003 return ParseDirectiveAbort();
1004 if (IDVal == ".include")
1005 return ParseDirectiveInclude();
1007 // Look up the handler in the handler table.
1008 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1009 DirectiveMap.lookup(IDVal);
1011 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1013 // Target hook for parsing target specific directives.
1014 if (!getTargetParser().ParseDirective(ID))
1017 Warning(IDLoc, "ignoring directive for now");
1018 EatToEndOfStatement();
1022 CheckForValidSection();
1024 // Canonicalize the opcode to lower case.
1025 SmallString<128> Opcode;
1026 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1027 Opcode.push_back(tolower(IDVal[i]));
1029 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1030 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1033 // Dump the parsed representation, if requested.
1034 if (getShowParsedOperands()) {
1035 SmallString<256> Str;
1036 raw_svector_ostream OS(Str);
1037 OS << "parsed instruction: [";
1038 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1041 ParsedOperands[i]->dump(OS);
1045 PrintMessage(IDLoc, OS.str(), "note");
1048 // If parsing succeeded, match the instruction.
1050 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1053 // Free any parsed operands.
1054 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1055 delete ParsedOperands[i];
1057 // Don't skip the rest of the line, the instruction parser is responsible for
1062 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1063 const std::vector<std::vector<AsmToken> > &A)
1064 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1066 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1067 // to hold the macro body with substitutions.
1068 SmallString<256> Buf;
1069 raw_svector_ostream OS(Buf);
1071 StringRef Body = M->Body;
1072 while (!Body.empty()) {
1073 // Scan for the next substitution.
1074 std::size_t End = Body.size(), Pos = 0;
1075 for (; Pos != End; ++Pos) {
1076 // Check for a substitution or escape.
1077 if (Body[Pos] != '$' || Pos + 1 == End)
1080 char Next = Body[Pos + 1];
1081 if (Next == '$' || Next == 'n' || isdigit(Next))
1086 OS << Body.slice(0, Pos);
1088 // Check if we reached the end.
1092 switch (Body[Pos+1]) {
1098 // $n => number of arguments
1103 // $[0-9] => argument
1105 // Missing arguments are ignored.
1106 unsigned Index = Body[Pos+1] - '0';
1107 if (Index >= A.size())
1110 // Otherwise substitute with the token values, with spaces eliminated.
1111 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1112 ie = A[Index].end(); it != ie; ++it)
1113 OS << it->getString();
1118 // Update the scan point.
1119 Body = Body.substr(Pos + 2);
1122 // We include the .endmacro in the buffer as our queue to exit the macro
1124 OS << ".endmacro\n";
1126 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1129 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1131 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1132 // this, although we should protect against infinite loops.
1133 if (ActiveMacros.size() == 20)
1134 return TokError("macros cannot be nested more than 20 levels deep");
1136 // Parse the macro instantiation arguments.
1137 std::vector<std::vector<AsmToken> > MacroArguments;
1138 MacroArguments.push_back(std::vector<AsmToken>());
1139 unsigned ParenLevel = 0;
1141 if (Lexer.is(AsmToken::Eof))
1142 return TokError("unexpected token in macro instantiation");
1143 if (Lexer.is(AsmToken::EndOfStatement))
1146 // If we aren't inside parentheses and this is a comma, start a new token
1148 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1149 MacroArguments.push_back(std::vector<AsmToken>());
1151 // Adjust the current parentheses level.
1152 if (Lexer.is(AsmToken::LParen))
1154 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1157 // Append the token to the current argument list.
1158 MacroArguments.back().push_back(getTok());
1163 // Create the macro instantiation object and add to the current macro
1164 // instantiation stack.
1165 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1168 ActiveMacros.push_back(MI);
1170 // Jump to the macro instantiation and prime the lexer.
1171 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1172 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1178 void AsmParser::HandleMacroExit() {
1179 // Jump to the EndOfStatement we should return to, and consume it.
1180 JumpToLoc(ActiveMacros.back()->ExitLoc);
1183 // Pop the instantiation entry.
1184 delete ActiveMacros.back();
1185 ActiveMacros.pop_back();
1188 bool AsmParser::ParseAssignment(StringRef Name) {
1189 // FIXME: Use better location, we should use proper tokens.
1190 SMLoc EqualLoc = Lexer.getLoc();
1192 const MCExpr *Value;
1193 if (ParseExpression(Value))
1196 if (Lexer.isNot(AsmToken::EndOfStatement))
1197 return TokError("unexpected token in assignment");
1199 // Eat the end of statement marker.
1202 // Validate that the LHS is allowed to be a variable (either it has not been
1203 // used as a symbol, or it is an absolute symbol).
1204 MCSymbol *Sym = getContext().LookupSymbol(Name);
1206 // Diagnose assignment to a label.
1208 // FIXME: Diagnostics. Note the location of the definition as a label.
1209 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1210 if (Sym->isUndefined() && !Sym->isUsedInExpr())
1211 ; // Allow redefinitions of undefined symbols only used in directives.
1212 else if (!Sym->isUndefined() && !Sym->isAbsolute())
1213 return Error(EqualLoc, "redefinition of '" + Name + "'");
1214 else if (!Sym->isVariable())
1215 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1216 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1217 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1220 Sym = getContext().GetOrCreateSymbol(Name);
1222 // FIXME: Handle '.'.
1224 Sym->setUsedInExpr(true);
1226 // Do the assignment.
1227 Out.EmitAssignment(Sym, Value);
1232 /// ParseIdentifier:
1235 bool AsmParser::ParseIdentifier(StringRef &Res) {
1236 // The assembler has relaxed rules for accepting identifiers, in particular we
1237 // allow things like '.globl $foo', which would normally be separate
1238 // tokens. At this level, we have already lexed so we cannot (currently)
1239 // handle this as a context dependent token, instead we detect adjacent tokens
1240 // and return the combined identifier.
1241 if (Lexer.is(AsmToken::Dollar)) {
1242 SMLoc DollarLoc = getLexer().getLoc();
1244 // Consume the dollar sign, and check for a following identifier.
1246 if (Lexer.isNot(AsmToken::Identifier))
1249 // We have a '$' followed by an identifier, make sure they are adjacent.
1250 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1253 // Construct the joined identifier and consume the token.
1254 Res = StringRef(DollarLoc.getPointer(),
1255 getTok().getIdentifier().size() + 1);
1260 if (Lexer.isNot(AsmToken::Identifier) &&
1261 Lexer.isNot(AsmToken::String))
1264 Res = getTok().getIdentifier();
1266 Lex(); // Consume the identifier token.
1271 /// ParseDirectiveSet:
1272 /// ::= .set identifier ',' expression
1273 bool AsmParser::ParseDirectiveSet() {
1276 if (ParseIdentifier(Name))
1277 return TokError("expected identifier after '.set' directive");
1279 if (getLexer().isNot(AsmToken::Comma))
1280 return TokError("unexpected token in '.set'");
1283 return ParseAssignment(Name);
1286 bool AsmParser::ParseEscapedString(std::string &Data) {
1287 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1290 StringRef Str = getTok().getStringContents();
1291 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1292 if (Str[i] != '\\') {
1297 // Recognize escaped characters. Note that this escape semantics currently
1298 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1301 return TokError("unexpected backslash at end of string");
1303 // Recognize octal sequences.
1304 if ((unsigned) (Str[i] - '0') <= 7) {
1305 // Consume up to three octal characters.
1306 unsigned Value = Str[i] - '0';
1308 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1310 Value = Value * 8 + (Str[i] - '0');
1312 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1314 Value = Value * 8 + (Str[i] - '0');
1319 return TokError("invalid octal escape sequence (out of range)");
1321 Data += (unsigned char) Value;
1325 // Otherwise recognize individual escapes.
1328 // Just reject invalid escape sequences for now.
1329 return TokError("invalid escape sequence (unrecognized character)");
1331 case 'b': Data += '\b'; break;
1332 case 'f': Data += '\f'; break;
1333 case 'n': Data += '\n'; break;
1334 case 'r': Data += '\r'; break;
1335 case 't': Data += '\t'; break;
1336 case '"': Data += '"'; break;
1337 case '\\': Data += '\\'; break;
1344 /// ParseDirectiveAscii:
1345 /// ::= ( .ascii | .asciz ) [ "string" ( , "string" )* ]
1346 bool AsmParser::ParseDirectiveAscii(bool ZeroTerminated) {
1347 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1348 CheckForValidSection();
1351 if (getLexer().isNot(AsmToken::String))
1352 return TokError("expected string in '.ascii' or '.asciz' directive");
1355 if (ParseEscapedString(Data))
1358 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1360 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1364 if (getLexer().is(AsmToken::EndOfStatement))
1367 if (getLexer().isNot(AsmToken::Comma))
1368 return TokError("unexpected token in '.ascii' or '.asciz' directive");
1377 /// ParseDirectiveValue
1378 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1379 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1380 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1381 CheckForValidSection();
1384 const MCExpr *Value;
1385 SMLoc ATTRIBUTE_UNUSED StartLoc = getLexer().getLoc();
1386 if (ParseExpression(Value))
1389 // Special case constant expressions to match code generator.
1390 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1391 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1393 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1395 if (getLexer().is(AsmToken::EndOfStatement))
1398 // FIXME: Improve diagnostic.
1399 if (getLexer().isNot(AsmToken::Comma))
1400 return TokError("unexpected token in directive");
1409 /// ParseDirectiveRealValue
1410 /// ::= (.single | .double) [ expression (, expression)* ]
1411 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1412 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1413 CheckForValidSection();
1416 // We don't truly support arithmetic on floating point expressions, so we
1417 // have to manually parse unary prefixes.
1419 if (getLexer().is(AsmToken::Minus)) {
1422 } else if (getLexer().is(AsmToken::Plus))
1425 if (getLexer().isNot(AsmToken::Integer) &&
1426 getLexer().isNot(AsmToken::Real))
1427 return TokError("unexpected token in directive");
1429 // Convert to an APFloat.
1430 APFloat Value(Semantics);
1431 if (Value.convertFromString(getTok().getString(),
1432 APFloat::rmNearestTiesToEven) ==
1433 APFloat::opInvalidOp)
1434 return TokError("invalid floating point literal");
1438 // Consume the numeric token.
1441 // Emit the value as an integer.
1442 APInt AsInt = Value.bitcastToAPInt();
1443 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1444 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1446 if (getLexer().is(AsmToken::EndOfStatement))
1449 if (getLexer().isNot(AsmToken::Comma))
1450 return TokError("unexpected token in directive");
1459 /// ParseDirectiveSpace
1460 /// ::= .space expression [ , expression ]
1461 bool AsmParser::ParseDirectiveSpace() {
1462 CheckForValidSection();
1465 if (ParseAbsoluteExpression(NumBytes))
1468 int64_t FillExpr = 0;
1469 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1470 if (getLexer().isNot(AsmToken::Comma))
1471 return TokError("unexpected token in '.space' directive");
1474 if (ParseAbsoluteExpression(FillExpr))
1477 if (getLexer().isNot(AsmToken::EndOfStatement))
1478 return TokError("unexpected token in '.space' directive");
1484 return TokError("invalid number of bytes in '.space' directive");
1486 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1487 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1492 /// ParseDirectiveZero
1493 /// ::= .zero expression
1494 bool AsmParser::ParseDirectiveZero() {
1495 CheckForValidSection();
1498 if (ParseAbsoluteExpression(NumBytes))
1501 if (getLexer().isNot(AsmToken::EndOfStatement))
1502 return TokError("unexpected token in '.zero' directive");
1506 getStreamer().EmitFill(NumBytes, 0, DEFAULT_ADDRSPACE);
1511 /// ParseDirectiveFill
1512 /// ::= .fill expression , expression , expression
1513 bool AsmParser::ParseDirectiveFill() {
1514 CheckForValidSection();
1517 if (ParseAbsoluteExpression(NumValues))
1520 if (getLexer().isNot(AsmToken::Comma))
1521 return TokError("unexpected token in '.fill' directive");
1525 if (ParseAbsoluteExpression(FillSize))
1528 if (getLexer().isNot(AsmToken::Comma))
1529 return TokError("unexpected token in '.fill' directive");
1533 if (ParseAbsoluteExpression(FillExpr))
1536 if (getLexer().isNot(AsmToken::EndOfStatement))
1537 return TokError("unexpected token in '.fill' directive");
1541 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1542 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1544 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1545 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1550 /// ParseDirectiveOrg
1551 /// ::= .org expression [ , expression ]
1552 bool AsmParser::ParseDirectiveOrg() {
1553 CheckForValidSection();
1555 const MCExpr *Offset;
1556 if (ParseExpression(Offset))
1559 // Parse optional fill expression.
1560 int64_t FillExpr = 0;
1561 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1562 if (getLexer().isNot(AsmToken::Comma))
1563 return TokError("unexpected token in '.org' directive");
1566 if (ParseAbsoluteExpression(FillExpr))
1569 if (getLexer().isNot(AsmToken::EndOfStatement))
1570 return TokError("unexpected token in '.org' directive");
1575 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1576 // has to be relative to the current section.
1577 getStreamer().EmitValueToOffset(Offset, FillExpr);
1582 /// ParseDirectiveAlign
1583 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1584 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1585 CheckForValidSection();
1587 SMLoc AlignmentLoc = getLexer().getLoc();
1589 if (ParseAbsoluteExpression(Alignment))
1593 bool HasFillExpr = false;
1594 int64_t FillExpr = 0;
1595 int64_t MaxBytesToFill = 0;
1596 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1597 if (getLexer().isNot(AsmToken::Comma))
1598 return TokError("unexpected token in directive");
1601 // The fill expression can be omitted while specifying a maximum number of
1602 // alignment bytes, e.g:
1604 if (getLexer().isNot(AsmToken::Comma)) {
1606 if (ParseAbsoluteExpression(FillExpr))
1610 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1611 if (getLexer().isNot(AsmToken::Comma))
1612 return TokError("unexpected token in directive");
1615 MaxBytesLoc = getLexer().getLoc();
1616 if (ParseAbsoluteExpression(MaxBytesToFill))
1619 if (getLexer().isNot(AsmToken::EndOfStatement))
1620 return TokError("unexpected token in directive");
1629 // Compute alignment in bytes.
1631 // FIXME: Diagnose overflow.
1632 if (Alignment >= 32) {
1633 Error(AlignmentLoc, "invalid alignment value");
1637 Alignment = 1ULL << Alignment;
1640 // Diagnose non-sensical max bytes to align.
1641 if (MaxBytesLoc.isValid()) {
1642 if (MaxBytesToFill < 1) {
1643 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1644 "many bytes, ignoring maximum bytes expression");
1648 if (MaxBytesToFill >= Alignment) {
1649 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1655 // Check whether we should use optimal code alignment for this .align
1658 // FIXME: This should be using a target hook.
1659 bool UseCodeAlign = false;
1660 if (const MCSectionMachO *S = dyn_cast<MCSectionMachO>(
1661 getStreamer().getCurrentSection()))
1662 UseCodeAlign = S->hasAttribute(MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS);
1663 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1664 ValueSize == 1 && UseCodeAlign) {
1665 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1667 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1668 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1675 /// ParseDirectiveSymbolAttribute
1676 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1677 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1678 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1682 if (ParseIdentifier(Name))
1683 return TokError("expected identifier in directive");
1685 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1687 getStreamer().EmitSymbolAttribute(Sym, Attr);
1689 if (getLexer().is(AsmToken::EndOfStatement))
1692 if (getLexer().isNot(AsmToken::Comma))
1693 return TokError("unexpected token in directive");
1702 /// ParseDirectiveELFType
1703 /// ::= .type identifier , @attribute
1704 bool AsmParser::ParseDirectiveELFType() {
1706 if (ParseIdentifier(Name))
1707 return TokError("expected identifier in directive");
1709 // Handle the identifier as the key symbol.
1710 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1712 if (getLexer().isNot(AsmToken::Comma))
1713 return TokError("unexpected token in '.type' directive");
1716 if (getLexer().isNot(AsmToken::At))
1717 return TokError("expected '@' before type");
1723 TypeLoc = getLexer().getLoc();
1724 if (ParseIdentifier(Type))
1725 return TokError("expected symbol type in directive");
1727 MCSymbolAttr Attr = StringSwitch<MCSymbolAttr>(Type)
1728 .Case("function", MCSA_ELF_TypeFunction)
1729 .Case("object", MCSA_ELF_TypeObject)
1730 .Case("tls_object", MCSA_ELF_TypeTLS)
1731 .Case("common", MCSA_ELF_TypeCommon)
1732 .Case("notype", MCSA_ELF_TypeNoType)
1733 .Default(MCSA_Invalid);
1735 if (Attr == MCSA_Invalid)
1736 return Error(TypeLoc, "unsupported attribute in '.type' directive");
1738 if (getLexer().isNot(AsmToken::EndOfStatement))
1739 return TokError("unexpected token in '.type' directive");
1743 getStreamer().EmitSymbolAttribute(Sym, Attr);
1748 /// ParseDirectiveComm
1749 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1750 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1751 CheckForValidSection();
1753 SMLoc IDLoc = getLexer().getLoc();
1755 if (ParseIdentifier(Name))
1756 return TokError("expected identifier in directive");
1758 // Handle the identifier as the key symbol.
1759 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1761 if (getLexer().isNot(AsmToken::Comma))
1762 return TokError("unexpected token in directive");
1766 SMLoc SizeLoc = getLexer().getLoc();
1767 if (ParseAbsoluteExpression(Size))
1770 int64_t Pow2Alignment = 0;
1771 SMLoc Pow2AlignmentLoc;
1772 if (getLexer().is(AsmToken::Comma)) {
1774 Pow2AlignmentLoc = getLexer().getLoc();
1775 if (ParseAbsoluteExpression(Pow2Alignment))
1778 // If this target takes alignments in bytes (not log) validate and convert.
1779 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1780 if (!isPowerOf2_64(Pow2Alignment))
1781 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1782 Pow2Alignment = Log2_64(Pow2Alignment);
1786 if (getLexer().isNot(AsmToken::EndOfStatement))
1787 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1791 // NOTE: a size of zero for a .comm should create a undefined symbol
1792 // but a size of .lcomm creates a bss symbol of size zero.
1794 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1795 "be less than zero");
1797 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1798 // may internally end up wanting an alignment in bytes.
1799 // FIXME: Diagnose overflow.
1800 if (Pow2Alignment < 0)
1801 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1802 "alignment, can't be less than zero");
1804 if (!Sym->isUndefined())
1805 return Error(IDLoc, "invalid symbol redefinition");
1807 // '.lcomm' is equivalent to '.zerofill'.
1808 // Create the Symbol as a common or local common with Size and Pow2Alignment
1810 getStreamer().EmitZerofill(Ctx.getMachOSection(
1811 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1812 0, SectionKind::getBSS()),
1813 Sym, Size, 1 << Pow2Alignment);
1817 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1821 /// ParseDirectiveAbort
1822 /// ::= .abort [... message ...]
1823 bool AsmParser::ParseDirectiveAbort() {
1824 // FIXME: Use loc from directive.
1825 SMLoc Loc = getLexer().getLoc();
1827 StringRef Str = ParseStringToEndOfStatement();
1828 if (getLexer().isNot(AsmToken::EndOfStatement))
1829 return TokError("unexpected token in '.abort' directive");
1834 Error(Loc, ".abort detected. Assembly stopping.");
1836 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1837 // FIXME: Actually abort assembly here.
1842 /// ParseDirectiveInclude
1843 /// ::= .include "filename"
1844 bool AsmParser::ParseDirectiveInclude() {
1845 if (getLexer().isNot(AsmToken::String))
1846 return TokError("expected string in '.include' directive");
1848 std::string Filename = getTok().getString();
1849 SMLoc IncludeLoc = getLexer().getLoc();
1852 if (getLexer().isNot(AsmToken::EndOfStatement))
1853 return TokError("unexpected token in '.include' directive");
1855 // Strip the quotes.
1856 Filename = Filename.substr(1, Filename.size()-2);
1858 // Attempt to switch the lexer to the included file before consuming the end
1859 // of statement to avoid losing it when we switch.
1860 if (EnterIncludeFile(Filename)) {
1861 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1868 /// ParseDirectiveIf
1869 /// ::= .if expression
1870 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1871 TheCondStack.push_back(TheCondState);
1872 TheCondState.TheCond = AsmCond::IfCond;
1873 if(TheCondState.Ignore) {
1874 EatToEndOfStatement();
1878 if (ParseAbsoluteExpression(ExprValue))
1881 if (getLexer().isNot(AsmToken::EndOfStatement))
1882 return TokError("unexpected token in '.if' directive");
1886 TheCondState.CondMet = ExprValue;
1887 TheCondState.Ignore = !TheCondState.CondMet;
1893 /// ParseDirectiveElseIf
1894 /// ::= .elseif expression
1895 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1896 if (TheCondState.TheCond != AsmCond::IfCond &&
1897 TheCondState.TheCond != AsmCond::ElseIfCond)
1898 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1900 TheCondState.TheCond = AsmCond::ElseIfCond;
1902 bool LastIgnoreState = false;
1903 if (!TheCondStack.empty())
1904 LastIgnoreState = TheCondStack.back().Ignore;
1905 if (LastIgnoreState || TheCondState.CondMet) {
1906 TheCondState.Ignore = true;
1907 EatToEndOfStatement();
1911 if (ParseAbsoluteExpression(ExprValue))
1914 if (getLexer().isNot(AsmToken::EndOfStatement))
1915 return TokError("unexpected token in '.elseif' directive");
1918 TheCondState.CondMet = ExprValue;
1919 TheCondState.Ignore = !TheCondState.CondMet;
1925 /// ParseDirectiveElse
1927 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
1928 if (getLexer().isNot(AsmToken::EndOfStatement))
1929 return TokError("unexpected token in '.else' directive");
1933 if (TheCondState.TheCond != AsmCond::IfCond &&
1934 TheCondState.TheCond != AsmCond::ElseIfCond)
1935 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
1937 TheCondState.TheCond = AsmCond::ElseCond;
1938 bool LastIgnoreState = false;
1939 if (!TheCondStack.empty())
1940 LastIgnoreState = TheCondStack.back().Ignore;
1941 if (LastIgnoreState || TheCondState.CondMet)
1942 TheCondState.Ignore = true;
1944 TheCondState.Ignore = false;
1949 /// ParseDirectiveEndIf
1951 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
1952 if (getLexer().isNot(AsmToken::EndOfStatement))
1953 return TokError("unexpected token in '.endif' directive");
1957 if ((TheCondState.TheCond == AsmCond::NoCond) ||
1958 TheCondStack.empty())
1959 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
1961 if (!TheCondStack.empty()) {
1962 TheCondState = TheCondStack.back();
1963 TheCondStack.pop_back();
1969 /// ParseDirectiveFile
1970 /// ::= .file [number] string
1971 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
1972 // FIXME: I'm not sure what this is.
1973 int64_t FileNumber = -1;
1974 SMLoc FileNumberLoc = getLexer().getLoc();
1975 if (getLexer().is(AsmToken::Integer)) {
1976 FileNumber = getTok().getIntVal();
1980 return TokError("file number less than one");
1983 if (getLexer().isNot(AsmToken::String))
1984 return TokError("unexpected token in '.file' directive");
1986 StringRef Filename = getTok().getString();
1987 Filename = Filename.substr(1, Filename.size()-2);
1990 if (getLexer().isNot(AsmToken::EndOfStatement))
1991 return TokError("unexpected token in '.file' directive");
1993 if (FileNumber == -1)
1994 getStreamer().EmitFileDirective(Filename);
1996 if (getContext().GetDwarfFile(Filename, FileNumber) == 0)
1997 Error(FileNumberLoc, "file number already allocated");
1998 getStreamer().EmitDwarfFileDirective(FileNumber, Filename);
2004 /// ParseDirectiveLine
2005 /// ::= .line [number]
2006 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
2007 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2008 if (getLexer().isNot(AsmToken::Integer))
2009 return TokError("unexpected token in '.line' directive");
2011 int64_t LineNumber = getTok().getIntVal();
2015 // FIXME: Do something with the .line.
2018 if (getLexer().isNot(AsmToken::EndOfStatement))
2019 return TokError("unexpected token in '.line' directive");
2025 /// ParseDirectiveLoc
2026 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2027 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2028 /// The first number is a file number, must have been previously assigned with
2029 /// a .file directive, the second number is the line number and optionally the
2030 /// third number is a column position (zero if not specified). The remaining
2031 /// optional items are .loc sub-directives.
2032 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2034 if (getLexer().isNot(AsmToken::Integer))
2035 return TokError("unexpected token in '.loc' directive");
2036 int64_t FileNumber = getTok().getIntVal();
2038 return TokError("file number less than one in '.loc' directive");
2039 if (!getContext().ValidateDwarfFileNumber(FileNumber))
2040 return TokError("unassigned file number in '.loc' directive");
2043 int64_t LineNumber = 0;
2044 if (getLexer().is(AsmToken::Integer)) {
2045 LineNumber = getTok().getIntVal();
2047 return TokError("line number less than one in '.loc' directive");
2051 int64_t ColumnPos = 0;
2052 if (getLexer().is(AsmToken::Integer)) {
2053 ColumnPos = getTok().getIntVal();
2055 return TokError("column position less than zero in '.loc' directive");
2059 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2061 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2063 if (getLexer().is(AsmToken::EndOfStatement))
2067 SMLoc Loc = getTok().getLoc();
2068 if (getParser().ParseIdentifier(Name))
2069 return TokError("unexpected token in '.loc' directive");
2071 if (Name == "basic_block")
2072 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2073 else if (Name == "prologue_end")
2074 Flags |= DWARF2_FLAG_PROLOGUE_END;
2075 else if (Name == "epilogue_begin")
2076 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2077 else if (Name == "is_stmt") {
2078 SMLoc Loc = getTok().getLoc();
2079 const MCExpr *Value;
2080 if (getParser().ParseExpression(Value))
2082 // The expression must be the constant 0 or 1.
2083 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2084 int Value = MCE->getValue();
2086 Flags &= ~DWARF2_FLAG_IS_STMT;
2087 else if (Value == 1)
2088 Flags |= DWARF2_FLAG_IS_STMT;
2090 return Error(Loc, "is_stmt value not 0 or 1");
2093 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2096 else if (Name == "isa") {
2097 SMLoc Loc = getTok().getLoc();
2098 const MCExpr *Value;
2099 if (getParser().ParseExpression(Value))
2101 // The expression must be a constant greater or equal to 0.
2102 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2103 int Value = MCE->getValue();
2105 return Error(Loc, "isa number less than zero");
2109 return Error(Loc, "isa number not a constant value");
2113 return Error(Loc, "unknown sub-directive in '.loc' directive");
2116 if (getLexer().is(AsmToken::EndOfStatement))
2121 getContext().setCurrentDwarfLoc(FileNumber, LineNumber, ColumnPos, Flags,Isa);
2126 /// ParseDirectiveMacrosOnOff
2129 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2130 SMLoc DirectiveLoc) {
2131 if (getLexer().isNot(AsmToken::EndOfStatement))
2132 return Error(getLexer().getLoc(),
2133 "unexpected token in '" + Directive + "' directive");
2135 getParser().MacrosEnabled = Directive == ".macros_on";
2140 /// ParseDirectiveMacro
2142 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2143 SMLoc DirectiveLoc) {
2145 if (getParser().ParseIdentifier(Name))
2146 return TokError("expected identifier in directive");
2148 if (getLexer().isNot(AsmToken::EndOfStatement))
2149 return TokError("unexpected token in '.macro' directive");
2151 // Eat the end of statement.
2154 AsmToken EndToken, StartToken = getTok();
2156 // Lex the macro definition.
2158 // Check whether we have reached the end of the file.
2159 if (getLexer().is(AsmToken::Eof))
2160 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2162 // Otherwise, check whether we have reach the .endmacro.
2163 if (getLexer().is(AsmToken::Identifier) &&
2164 (getTok().getIdentifier() == ".endm" ||
2165 getTok().getIdentifier() == ".endmacro")) {
2166 EndToken = getTok();
2168 if (getLexer().isNot(AsmToken::EndOfStatement))
2169 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2174 // Otherwise, scan til the end of the statement.
2175 getParser().EatToEndOfStatement();
2178 if (getParser().MacroMap.lookup(Name)) {
2179 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2182 const char *BodyStart = StartToken.getLoc().getPointer();
2183 const char *BodyEnd = EndToken.getLoc().getPointer();
2184 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2185 getParser().MacroMap[Name] = new Macro(Name, Body);
2189 /// ParseDirectiveEndMacro
2192 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2193 SMLoc DirectiveLoc) {
2194 if (getLexer().isNot(AsmToken::EndOfStatement))
2195 return TokError("unexpected token in '" + Directive + "' directive");
2197 // If we are inside a macro instantiation, terminate the current
2199 if (!getParser().ActiveMacros.empty()) {
2200 getParser().HandleMacroExit();
2204 // Otherwise, this .endmacro is a stray entry in the file; well formed
2205 // .endmacro directives are handled during the macro definition parsing.
2206 return TokError("unexpected '" + Directive + "' in file, "
2207 "no current macro definition");
2210 void GenericAsmParser::ParseUleb128(uint64_t Value) {
2211 const uint64_t Mask = (1 << 7) - 1;
2213 unsigned Byte = Value & Mask;
2215 if (Value) // Not the last one
2217 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2221 void GenericAsmParser::ParseSleb128(int64_t Value) {
2222 const int64_t Mask = (1 << 7) - 1;
2224 unsigned Byte = Value & Mask;
2226 bool Done = ((Value == 0 && (Byte & 0x40) == 0) ||
2227 (Value == -1 && (Byte & 0x40) != 0));
2230 getStreamer().EmitIntValue(Byte, 1, DEFAULT_ADDRSPACE);
2236 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2238 if (getParser().ParseAbsoluteExpression(Value))
2241 if (getLexer().isNot(AsmToken::EndOfStatement))
2242 return TokError("unexpected token in directive");
2244 if (DirName[1] == 's')
2245 ParseSleb128(Value);
2247 ParseUleb128(Value);
2252 /// \brief Create an MCAsmParser instance.
2253 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2254 MCContext &C, MCStreamer &Out,
2255 const MCAsmInfo &MAI) {
2256 return new AsmParser(T, SM, C, Out, MAI);