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/Twine.h"
18 #include "llvm/MC/MCAsmInfo.h"
19 #include "llvm/MC/MCContext.h"
20 #include "llvm/MC/MCDwarf.h"
21 #include "llvm/MC/MCExpr.h"
22 #include "llvm/MC/MCInstPrinter.h"
23 #include "llvm/MC/MCInstrInfo.h"
24 #include "llvm/MC/MCParser/AsmCond.h"
25 #include "llvm/MC/MCParser/AsmLexer.h"
26 #include "llvm/MC/MCParser/MCAsmParser.h"
27 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
28 #include "llvm/MC/MCRegisterInfo.h"
29 #include "llvm/MC/MCSectionMachO.h"
30 #include "llvm/MC/MCStreamer.h"
31 #include "llvm/MC/MCSymbol.h"
32 #include "llvm/MC/MCTargetAsmParser.h"
33 #include "llvm/Support/CommandLine.h"
34 #include "llvm/Support/ErrorHandling.h"
35 #include "llvm/Support/MathExtras.h"
36 #include "llvm/Support/MemoryBuffer.h"
37 #include "llvm/Support/SourceMgr.h"
38 #include "llvm/Support/raw_ostream.h"
46 FatalAssemblerWarnings("fatal-assembler-warnings",
47 cl::desc("Consider warnings as error"));
49 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() {}
53 /// \brief Helper class for tracking macro definitions.
54 typedef std::vector<AsmToken> MacroArgument;
55 typedef std::vector<MacroArgument> MacroArguments;
56 typedef std::pair<StringRef, MacroArgument> MacroParameter;
57 typedef std::vector<MacroParameter> MacroParameters;
62 MacroParameters Parameters;
65 Macro(StringRef N, StringRef B, const MacroParameters &P) :
66 Name(N), Body(B), Parameters(P) {}
69 /// \brief Helper class for storing information about an active macro
71 struct MacroInstantiation {
72 /// The macro being instantiated.
73 const Macro *TheMacro;
75 /// The macro instantiation with substitutions.
76 MemoryBuffer *Instantiation;
78 /// The location of the instantiation.
79 SMLoc InstantiationLoc;
81 /// The location where parsing should resume upon instantiation completion.
85 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
90 struct ParseStatementInfo {
91 /// ParsedOperands - The parsed operands from the last parsed statement.
92 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
94 /// Opcode - The opcode from the last parsed instruction.
97 SmallVectorImpl<AsmRewrite> *AsmRewrites;
99 ParseStatementInfo() : Opcode(~0U), AsmRewrites(0) {}
100 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
101 : Opcode(~0), AsmRewrites(rewrites) {}
103 ~ParseStatementInfo() {
104 // Free any parsed operands.
105 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
106 delete ParsedOperands[i];
107 ParsedOperands.clear();
111 /// \brief The concrete assembly parser instance.
112 class AsmParser : public MCAsmParser {
113 friend class GenericAsmParser;
115 AsmParser(const AsmParser &) LLVM_DELETED_FUNCTION;
116 void operator=(const AsmParser &) LLVM_DELETED_FUNCTION;
121 const MCAsmInfo &MAI;
123 SourceMgr::DiagHandlerTy SavedDiagHandler;
124 void *SavedDiagContext;
125 MCAsmParserExtension *GenericParser;
126 MCAsmParserExtension *PlatformParser;
128 /// This is the current buffer index we're lexing from as managed by the
129 /// SourceMgr object.
132 AsmCond TheCondState;
133 std::vector<AsmCond> TheCondStack;
135 /// DirectiveMap - This is a table handlers for directives. Each handler is
136 /// invoked after the directive identifier is read and is responsible for
137 /// parsing and validating the rest of the directive. The handler is passed
138 /// in the directive name and the location of the directive keyword.
139 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
141 /// MacroMap - Map of currently defined macros.
142 StringMap<Macro*> MacroMap;
144 /// ActiveMacros - Stack of active macro instantiations.
145 std::vector<MacroInstantiation*> ActiveMacros;
147 /// Boolean tracking whether macro substitution is enabled.
148 unsigned MacrosEnabled : 1;
150 /// Flag tracking whether any errors have been encountered.
151 unsigned HadError : 1;
153 /// The values from the last parsed cpp hash file line comment if any.
154 StringRef CppHashFilename;
155 int64_t CppHashLineNumber;
158 /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
159 unsigned AssemblerDialect;
161 /// IsDarwin - is Darwin compatibility enabled?
164 /// ParsingInlineAsm - Are we parsing ms-style inline assembly?
165 bool ParsingInlineAsm;
168 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
169 const MCAsmInfo &MAI);
170 virtual ~AsmParser();
172 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
174 virtual void AddDirectiveHandler(MCAsmParserExtension *Object,
176 DirectiveHandler Handler) {
177 DirectiveMap[Directive] = std::make_pair(Object, Handler);
181 /// @name MCAsmParser Interface
184 virtual SourceMgr &getSourceManager() { return SrcMgr; }
185 virtual MCAsmLexer &getLexer() { return Lexer; }
186 virtual MCContext &getContext() { return Ctx; }
187 virtual MCStreamer &getStreamer() { return Out; }
188 virtual unsigned getAssemblerDialect() {
189 if (AssemblerDialect == ~0U)
190 return MAI.getAssemblerDialect();
192 return AssemblerDialect;
194 virtual void setAssemblerDialect(unsigned i) {
195 AssemblerDialect = i;
198 virtual bool Warning(SMLoc L, const Twine &Msg,
199 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>());
200 virtual bool Error(SMLoc L, const Twine &Msg,
201 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>());
203 virtual const AsmToken &Lex();
205 void setParsingInlineAsm(bool V) { ParsingInlineAsm = V; }
206 bool isParsingInlineAsm() { return ParsingInlineAsm; }
208 bool ParseMSInlineAsm(void *AsmLoc, std::string &AsmString,
209 unsigned &NumOutputs, unsigned &NumInputs,
210 SmallVectorImpl<std::pair<void *,bool> > &OpDecls,
211 SmallVectorImpl<std::string> &Constraints,
212 SmallVectorImpl<std::string> &Clobbers,
213 const MCInstrInfo *MII,
214 const MCInstPrinter *IP,
215 MCAsmParserSemaCallback &SI);
217 bool ParseExpression(const MCExpr *&Res);
218 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
219 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
220 virtual bool ParseAbsoluteExpression(int64_t &Res);
225 void CheckForValidSection();
227 bool ParseStatement(ParseStatementInfo &Info);
228 void EatToEndOfLine();
229 bool ParseCppHashLineFilenameComment(const SMLoc &L);
231 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
232 bool expandMacro(raw_svector_ostream &OS, StringRef Body,
233 const MacroParameters &Parameters,
234 const MacroArguments &A,
236 void HandleMacroExit();
238 void PrintMacroInstantiations();
239 void PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
240 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()) const {
241 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
243 static void DiagHandler(const SMDiagnostic &Diag, void *Context);
245 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
246 bool EnterIncludeFile(const std::string &Filename);
247 /// ProcessIncbinFile - Process the specified file for the .incbin directive.
248 /// This returns true on failure.
249 bool ProcessIncbinFile(const std::string &Filename);
251 /// \brief Reset the current lexer position to that given by \p Loc. The
252 /// current token is not set; clients should ensure Lex() is called
254 void JumpToLoc(SMLoc Loc);
256 virtual void EatToEndOfStatement();
258 bool ParseMacroArgument(MacroArgument &MA,
259 AsmToken::TokenKind &ArgumentDelimiter);
260 bool ParseMacroArguments(const Macro *M, MacroArguments &A);
262 /// \brief Parse up to the end of statement and a return the contents from the
263 /// current token until the end of the statement; the current token on exit
264 /// will be either the EndOfStatement or EOF.
265 virtual StringRef ParseStringToEndOfStatement();
267 /// \brief Parse until the end of a statement or a comma is encountered,
268 /// return the contents from the current token up to the end or comma.
269 StringRef ParseStringToComma();
271 bool ParseAssignment(StringRef Name, bool allow_redef,
272 bool NoDeadStrip = false);
274 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
275 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
276 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
277 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
279 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
280 /// and set \p Res to the identifier contents.
281 virtual bool ParseIdentifier(StringRef &Res);
283 // Directive Parsing.
285 // ".ascii", ".asciiz", ".string"
286 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
287 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
288 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
289 bool ParseDirectiveFill(); // ".fill"
290 bool ParseDirectiveSpace(); // ".space"
291 bool ParseDirectiveZero(); // ".zero"
292 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv"
293 bool ParseDirectiveOrg(); // ".org"
294 // ".align{,32}", ".p2align{,w,l}"
295 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
297 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
298 /// accepts a single symbol (which should be a label or an external).
299 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
301 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
303 bool ParseDirectiveAbort(); // ".abort"
304 bool ParseDirectiveInclude(); // ".include"
305 bool ParseDirectiveIncbin(); // ".incbin"
307 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
308 // ".ifb" or ".ifnb", depending on ExpectBlank.
309 bool ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
310 // ".ifc" or ".ifnc", depending on ExpectEqual.
311 bool ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
312 // ".ifdef" or ".ifndef", depending on expect_defined
313 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
314 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
315 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
316 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
318 /// ParseEscapedString - Parse the current token as a string which may include
319 /// escaped characters and return the string contents.
320 bool ParseEscapedString(std::string &Data);
322 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
323 MCSymbolRefExpr::VariantKind Variant);
325 // Macro-like directives
326 Macro *ParseMacroLikeBody(SMLoc DirectiveLoc);
327 void InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc,
328 raw_svector_ostream &OS);
329 bool ParseDirectiveRept(SMLoc DirectiveLoc); // ".rept"
330 bool ParseDirectiveIrp(SMLoc DirectiveLoc); // ".irp"
331 bool ParseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
332 bool ParseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
335 bool ParseDirectiveEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info);
338 /// \brief Generic implementations of directive handling, etc. which is shared
339 /// (or the default, at least) for all assembler parser.
340 class GenericAsmParser : public MCAsmParserExtension {
341 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
342 void AddDirectiveHandler(StringRef Directive) {
343 getParser().AddDirectiveHandler(this, Directive,
344 HandleDirective<GenericAsmParser, Handler>);
347 GenericAsmParser() {}
349 AsmParser &getParser() {
350 return (AsmParser&) this->MCAsmParserExtension::getParser();
353 virtual void Initialize(MCAsmParser &Parser) {
354 // Call the base implementation.
355 this->MCAsmParserExtension::Initialize(Parser);
357 // Debugging directives.
358 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
359 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
360 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
361 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
364 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFISections>(
366 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>(
368 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>(
370 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>(
372 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>(
373 ".cfi_def_cfa_offset");
374 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset>(
375 ".cfi_adjust_cfa_offset");
376 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>(
377 ".cfi_def_cfa_register");
378 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>(
380 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>(
383 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality");
385 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda");
387 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state");
389 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state");
391 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value");
393 &GenericAsmParser::ParseDirectiveCFIRestore>(".cfi_restore");
395 &GenericAsmParser::ParseDirectiveCFIEscape>(".cfi_escape");
397 &GenericAsmParser::ParseDirectiveCFISignalFrame>(".cfi_signal_frame");
400 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
402 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
404 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
405 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
406 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
407 AddDirectiveHandler<&GenericAsmParser::ParseDirectivePurgeMacro>(".purgem");
409 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
410 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
413 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
415 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
416 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
417 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
418 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc);
419 bool ParseDirectiveCFISections(StringRef, SMLoc DirectiveLoc);
420 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc);
421 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc);
422 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc);
423 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc);
424 bool ParseDirectiveCFIAdjustCfaOffset(StringRef, SMLoc DirectiveLoc);
425 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc);
426 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc);
427 bool ParseDirectiveCFIRelOffset(StringRef, SMLoc DirectiveLoc);
428 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc);
429 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc);
430 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc);
431 bool ParseDirectiveCFISameValue(StringRef, SMLoc DirectiveLoc);
432 bool ParseDirectiveCFIRestore(StringRef, SMLoc DirectiveLoc);
433 bool ParseDirectiveCFIEscape(StringRef, SMLoc DirectiveLoc);
434 bool ParseDirectiveCFISignalFrame(StringRef, SMLoc DirectiveLoc);
436 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
437 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
438 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
439 bool ParseDirectivePurgeMacro(StringRef, SMLoc DirectiveLoc);
441 bool ParseDirectiveLEB128(StringRef, SMLoc);
448 extern MCAsmParserExtension *createDarwinAsmParser();
449 extern MCAsmParserExtension *createELFAsmParser();
450 extern MCAsmParserExtension *createCOFFAsmParser();
454 enum { DEFAULT_ADDRSPACE = 0 };
456 AsmParser::AsmParser(SourceMgr &_SM, MCContext &_Ctx,
457 MCStreamer &_Out, const MCAsmInfo &_MAI)
458 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM),
459 GenericParser(new GenericAsmParser), PlatformParser(0),
460 CurBuffer(0), MacrosEnabled(true), CppHashLineNumber(0),
461 AssemblerDialect(~0U), IsDarwin(false), ParsingInlineAsm(false) {
462 // Save the old handler.
463 SavedDiagHandler = SrcMgr.getDiagHandler();
464 SavedDiagContext = SrcMgr.getDiagContext();
465 // Set our own handler which calls the saved handler.
466 SrcMgr.setDiagHandler(DiagHandler, this);
467 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
469 // Initialize the generic parser.
470 GenericParser->Initialize(*this);
472 // Initialize the platform / file format parser.
474 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
476 if (_MAI.hasMicrosoftFastStdCallMangling()) {
477 PlatformParser = createCOFFAsmParser();
478 PlatformParser->Initialize(*this);
479 } else if (_MAI.hasSubsectionsViaSymbols()) {
480 PlatformParser = createDarwinAsmParser();
481 PlatformParser->Initialize(*this);
484 PlatformParser = createELFAsmParser();
485 PlatformParser->Initialize(*this);
489 AsmParser::~AsmParser() {
490 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
492 // Destroy any macros.
493 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
494 ie = MacroMap.end(); it != ie; ++it)
495 delete it->getValue();
497 delete PlatformParser;
498 delete GenericParser;
501 void AsmParser::PrintMacroInstantiations() {
502 // Print the active macro instantiation stack.
503 for (std::vector<MacroInstantiation*>::const_reverse_iterator
504 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
505 PrintMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
506 "while in macro instantiation");
509 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
510 if (FatalAssemblerWarnings)
511 return Error(L, Msg, Ranges);
512 PrintMessage(L, SourceMgr::DK_Warning, Msg, Ranges);
513 PrintMacroInstantiations();
517 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
519 PrintMessage(L, SourceMgr::DK_Error, Msg, Ranges);
520 PrintMacroInstantiations();
524 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
525 std::string IncludedFile;
526 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
532 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
537 /// Process the specified .incbin file by seaching for it in the include paths
538 /// then just emitting the byte contents of the file to the streamer. This
539 /// returns true on failure.
540 bool AsmParser::ProcessIncbinFile(const std::string &Filename) {
541 std::string IncludedFile;
542 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
546 // Pick up the bytes from the file and emit them.
547 getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(),
552 void AsmParser::JumpToLoc(SMLoc Loc) {
553 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
554 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
557 const AsmToken &AsmParser::Lex() {
558 const AsmToken *tok = &Lexer.Lex();
560 if (tok->is(AsmToken::Eof)) {
561 // If this is the end of an included file, pop the parent file off the
563 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
564 if (ParentIncludeLoc != SMLoc()) {
565 JumpToLoc(ParentIncludeLoc);
570 if (tok->is(AsmToken::Error))
571 Error(Lexer.getErrLoc(), Lexer.getErr());
576 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
577 // Create the initial section, if requested.
578 if (!NoInitialTextSection)
585 AsmCond StartingCondState = TheCondState;
587 // If we are generating dwarf for assembly source files save the initial text
588 // section and generate a .file directive.
589 if (getContext().getGenDwarfForAssembly()) {
590 getContext().setGenDwarfSection(getStreamer().getCurrentSection());
591 MCSymbol *SectionStartSym = getContext().CreateTempSymbol();
592 getStreamer().EmitLabel(SectionStartSym);
593 getContext().setGenDwarfSectionStartSym(SectionStartSym);
594 getStreamer().EmitDwarfFileDirective(getContext().nextGenDwarfFileNumber(),
595 StringRef(), SrcMgr.getMemoryBuffer(CurBuffer)->getBufferIdentifier());
598 // While we have input, parse each statement.
599 while (Lexer.isNot(AsmToken::Eof)) {
600 ParseStatementInfo Info;
601 if (!ParseStatement(Info)) continue;
603 // We had an error, validate that one was emitted and recover by skipping to
605 assert(HadError && "Parse statement returned an error, but none emitted!");
606 EatToEndOfStatement();
609 if (TheCondState.TheCond != StartingCondState.TheCond ||
610 TheCondState.Ignore != StartingCondState.Ignore)
611 return TokError("unmatched .ifs or .elses");
613 // Check to see there are no empty DwarfFile slots.
614 const std::vector<MCDwarfFile *> &MCDwarfFiles =
615 getContext().getMCDwarfFiles();
616 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
617 if (!MCDwarfFiles[i])
618 TokError("unassigned file number: " + Twine(i) + " for .file directives");
621 // Check to see that all assembler local symbols were actually defined.
622 // Targets that don't do subsections via symbols may not want this, though,
623 // so conservatively exclude them. Only do this if we're finalizing, though,
624 // as otherwise we won't necessarilly have seen everything yet.
625 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) {
626 const MCContext::SymbolTable &Symbols = getContext().getSymbols();
627 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(),
630 MCSymbol *Sym = i->getValue();
631 // Variable symbols may not be marked as defined, so check those
632 // explicitly. If we know it's a variable, we have a definition for
633 // the purposes of this check.
634 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
635 // FIXME: We would really like to refer back to where the symbol was
636 // first referenced for a source location. We need to add something
637 // to track that. Currently, we just point to the end of the file.
638 PrintMessage(getLexer().getLoc(), SourceMgr::DK_Error,
639 "assembler local symbol '" + Sym->getName() +
645 // Finalize the output stream if there are no errors and if the client wants
647 if (!HadError && !NoFinalize)
653 void AsmParser::CheckForValidSection() {
654 if (!ParsingInlineAsm && !getStreamer().getCurrentSection()) {
655 TokError("expected section directive before assembly directive");
656 Out.SwitchSection(Ctx.getMachOSection(
658 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
659 0, SectionKind::getText()));
663 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
664 void AsmParser::EatToEndOfStatement() {
665 while (Lexer.isNot(AsmToken::EndOfStatement) &&
666 Lexer.isNot(AsmToken::Eof))
670 if (Lexer.is(AsmToken::EndOfStatement))
674 StringRef AsmParser::ParseStringToEndOfStatement() {
675 const char *Start = getTok().getLoc().getPointer();
677 while (Lexer.isNot(AsmToken::EndOfStatement) &&
678 Lexer.isNot(AsmToken::Eof))
681 const char *End = getTok().getLoc().getPointer();
682 return StringRef(Start, End - Start);
685 StringRef AsmParser::ParseStringToComma() {
686 const char *Start = getTok().getLoc().getPointer();
688 while (Lexer.isNot(AsmToken::EndOfStatement) &&
689 Lexer.isNot(AsmToken::Comma) &&
690 Lexer.isNot(AsmToken::Eof))
693 const char *End = getTok().getLoc().getPointer();
694 return StringRef(Start, End - Start);
697 /// ParseParenExpr - Parse a paren expression and return it.
698 /// NOTE: This assumes the leading '(' has already been consumed.
700 /// parenexpr ::= expr)
702 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
703 if (ParseExpression(Res)) return true;
704 if (Lexer.isNot(AsmToken::RParen))
705 return TokError("expected ')' in parentheses expression");
706 EndLoc = Lexer.getLoc();
711 /// ParseBracketExpr - Parse a bracket expression and return it.
712 /// NOTE: This assumes the leading '[' has already been consumed.
714 /// bracketexpr ::= expr]
716 bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
717 if (ParseExpression(Res)) return true;
718 if (Lexer.isNot(AsmToken::RBrac))
719 return TokError("expected ']' in brackets expression");
720 EndLoc = Lexer.getLoc();
725 /// ParsePrimaryExpr - Parse a primary expression and return it.
726 /// primaryexpr ::= (parenexpr
727 /// primaryexpr ::= symbol
728 /// primaryexpr ::= number
729 /// primaryexpr ::= '.'
730 /// primaryexpr ::= ~,+,- primaryexpr
731 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
732 switch (Lexer.getKind()) {
734 return TokError("unknown token in expression");
735 // If we have an error assume that we've already handled it.
736 case AsmToken::Error:
738 case AsmToken::Exclaim:
739 Lex(); // Eat the operator.
740 if (ParsePrimaryExpr(Res, EndLoc))
742 Res = MCUnaryExpr::CreateLNot(Res, getContext());
744 case AsmToken::Dollar:
745 case AsmToken::String:
746 case AsmToken::Identifier: {
747 EndLoc = Lexer.getLoc();
749 StringRef Identifier;
750 if (ParseIdentifier(Identifier))
753 // This is a symbol reference.
754 std::pair<StringRef, StringRef> Split = Identifier.split('@');
755 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
757 // Lookup the symbol variant if used.
758 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
759 if (Split.first.size() != Identifier.size()) {
760 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
761 if (Variant == MCSymbolRefExpr::VK_Invalid) {
762 Variant = MCSymbolRefExpr::VK_None;
763 return TokError("invalid variant '" + Split.second + "'");
767 // If this is an absolute variable reference, substitute it now to preserve
768 // semantics in the face of reassignment.
769 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
771 return Error(EndLoc, "unexpected modifier on variable reference");
773 Res = Sym->getVariableValue();
777 // Otherwise create a symbol ref.
778 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
781 case AsmToken::Integer: {
782 SMLoc Loc = getTok().getLoc();
783 int64_t IntVal = getTok().getIntVal();
784 Res = MCConstantExpr::Create(IntVal, getContext());
785 EndLoc = Lexer.getLoc();
787 // Look for 'b' or 'f' following an Integer as a directional label
788 if (Lexer.getKind() == AsmToken::Identifier) {
789 StringRef IDVal = getTok().getString();
790 if (IDVal == "f" || IDVal == "b"){
791 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
792 IDVal == "f" ? 1 : 0);
793 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
795 if (IDVal == "b" && Sym->isUndefined())
796 return Error(Loc, "invalid reference to undefined symbol");
797 EndLoc = Lexer.getLoc();
798 Lex(); // Eat identifier.
803 case AsmToken::Real: {
804 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
805 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
806 Res = MCConstantExpr::Create(IntVal, getContext());
810 case AsmToken::Dot: {
811 // This is a '.' reference, which references the current PC. Emit a
812 // temporary label to the streamer and refer to it.
813 MCSymbol *Sym = Ctx.CreateTempSymbol();
815 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
816 EndLoc = Lexer.getLoc();
817 Lex(); // Eat identifier.
820 case AsmToken::LParen:
821 Lex(); // Eat the '('.
822 return ParseParenExpr(Res, EndLoc);
823 case AsmToken::LBrac:
824 if (!PlatformParser->HasBracketExpressions())
825 return TokError("brackets expression not supported on this target");
826 Lex(); // Eat the '['.
827 return ParseBracketExpr(Res, EndLoc);
828 case AsmToken::Minus:
829 Lex(); // Eat the operator.
830 if (ParsePrimaryExpr(Res, EndLoc))
832 Res = MCUnaryExpr::CreateMinus(Res, getContext());
835 Lex(); // Eat the operator.
836 if (ParsePrimaryExpr(Res, EndLoc))
838 Res = MCUnaryExpr::CreatePlus(Res, getContext());
840 case AsmToken::Tilde:
841 Lex(); // Eat the operator.
842 if (ParsePrimaryExpr(Res, EndLoc))
844 Res = MCUnaryExpr::CreateNot(Res, getContext());
849 bool AsmParser::ParseExpression(const MCExpr *&Res) {
851 return ParseExpression(Res, EndLoc);
855 AsmParser::ApplyModifierToExpr(const MCExpr *E,
856 MCSymbolRefExpr::VariantKind Variant) {
857 // Recurse over the given expression, rebuilding it to apply the given variant
858 // if there is exactly one symbol.
859 switch (E->getKind()) {
861 case MCExpr::Constant:
864 case MCExpr::SymbolRef: {
865 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
867 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
868 TokError("invalid variant on expression '" +
869 getTok().getIdentifier() + "' (already modified)");
873 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
876 case MCExpr::Unary: {
877 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
878 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
881 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
884 case MCExpr::Binary: {
885 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
886 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
887 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
892 if (!LHS) LHS = BE->getLHS();
893 if (!RHS) RHS = BE->getRHS();
895 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
899 llvm_unreachable("Invalid expression kind!");
902 /// ParseExpression - Parse an expression and return it.
904 /// expr ::= expr &&,|| expr -> lowest.
905 /// expr ::= expr |,^,&,! expr
906 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr
907 /// expr ::= expr <<,>> expr
908 /// expr ::= expr +,- expr
909 /// expr ::= expr *,/,% expr -> highest.
910 /// expr ::= primaryexpr
912 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
913 // Parse the expression.
915 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
918 // As a special case, we support 'a op b @ modifier' by rewriting the
919 // expression to include the modifier. This is inefficient, but in general we
920 // expect users to use 'a@modifier op b'.
921 if (Lexer.getKind() == AsmToken::At) {
924 if (Lexer.isNot(AsmToken::Identifier))
925 return TokError("unexpected symbol modifier following '@'");
927 MCSymbolRefExpr::VariantKind Variant =
928 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
929 if (Variant == MCSymbolRefExpr::VK_Invalid)
930 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
932 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
934 return TokError("invalid modifier '" + getTok().getIdentifier() +
935 "' (no symbols present)");
942 // Try to constant fold it up front, if possible.
944 if (Res->EvaluateAsAbsolute(Value))
945 Res = MCConstantExpr::Create(Value, getContext());
950 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
952 return ParseParenExpr(Res, EndLoc) ||
953 ParseBinOpRHS(1, Res, EndLoc);
956 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
959 SMLoc StartLoc = Lexer.getLoc();
960 if (ParseExpression(Expr))
963 if (!Expr->EvaluateAsAbsolute(Res))
964 return Error(StartLoc, "expected absolute expression");
969 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
970 MCBinaryExpr::Opcode &Kind) {
973 return 0; // not a binop.
975 // Lowest Precedence: &&, ||
976 case AsmToken::AmpAmp:
977 Kind = MCBinaryExpr::LAnd;
979 case AsmToken::PipePipe:
980 Kind = MCBinaryExpr::LOr;
984 // Low Precedence: |, &, ^
986 // FIXME: gas seems to support '!' as an infix operator?
988 Kind = MCBinaryExpr::Or;
990 case AsmToken::Caret:
991 Kind = MCBinaryExpr::Xor;
994 Kind = MCBinaryExpr::And;
997 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
998 case AsmToken::EqualEqual:
999 Kind = MCBinaryExpr::EQ;
1001 case AsmToken::ExclaimEqual:
1002 case AsmToken::LessGreater:
1003 Kind = MCBinaryExpr::NE;
1005 case AsmToken::Less:
1006 Kind = MCBinaryExpr::LT;
1008 case AsmToken::LessEqual:
1009 Kind = MCBinaryExpr::LTE;
1011 case AsmToken::Greater:
1012 Kind = MCBinaryExpr::GT;
1014 case AsmToken::GreaterEqual:
1015 Kind = MCBinaryExpr::GTE;
1018 // Intermediate Precedence: <<, >>
1019 case AsmToken::LessLess:
1020 Kind = MCBinaryExpr::Shl;
1022 case AsmToken::GreaterGreater:
1023 Kind = MCBinaryExpr::Shr;
1026 // High Intermediate Precedence: +, -
1027 case AsmToken::Plus:
1028 Kind = MCBinaryExpr::Add;
1030 case AsmToken::Minus:
1031 Kind = MCBinaryExpr::Sub;
1034 // Highest Precedence: *, /, %
1035 case AsmToken::Star:
1036 Kind = MCBinaryExpr::Mul;
1038 case AsmToken::Slash:
1039 Kind = MCBinaryExpr::Div;
1041 case AsmToken::Percent:
1042 Kind = MCBinaryExpr::Mod;
1048 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
1049 /// Res contains the LHS of the expression on input.
1050 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1053 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1054 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1056 // If the next token is lower precedence than we are allowed to eat, return
1057 // successfully with what we ate already.
1058 if (TokPrec < Precedence)
1063 // Eat the next primary expression.
1065 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
1067 // If BinOp binds less tightly with RHS than the operator after RHS, let
1068 // the pending operator take RHS as its LHS.
1069 MCBinaryExpr::Opcode Dummy;
1070 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1071 if (TokPrec < NextTokPrec) {
1072 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
1075 // Merge LHS and RHS according to operator.
1076 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
1081 /// ::= EndOfStatement
1082 /// ::= Label* Directive ...Operands... EndOfStatement
1083 /// ::= Label* Identifier OperandList* EndOfStatement
1084 bool AsmParser::ParseStatement(ParseStatementInfo &Info) {
1085 if (Lexer.is(AsmToken::EndOfStatement)) {
1091 // Statements always start with an identifier or are a full line comment.
1092 AsmToken ID = getTok();
1093 SMLoc IDLoc = ID.getLoc();
1095 int64_t LocalLabelVal = -1;
1096 // A full line comment is a '#' as the first token.
1097 if (Lexer.is(AsmToken::Hash))
1098 return ParseCppHashLineFilenameComment(IDLoc);
1100 // Allow an integer followed by a ':' as a directional local label.
1101 if (Lexer.is(AsmToken::Integer)) {
1102 LocalLabelVal = getTok().getIntVal();
1103 if (LocalLabelVal < 0) {
1104 if (!TheCondState.Ignore)
1105 return TokError("unexpected token at start of statement");
1109 IDVal = getTok().getString();
1110 Lex(); // Consume the integer token to be used as an identifier token.
1111 if (Lexer.getKind() != AsmToken::Colon) {
1112 if (!TheCondState.Ignore)
1113 return TokError("unexpected token at start of statement");
1117 } else if (Lexer.is(AsmToken::Dot)) {
1118 // Treat '.' as a valid identifier in this context.
1122 } else if (ParseIdentifier(IDVal)) {
1123 if (!TheCondState.Ignore)
1124 return TokError("unexpected token at start of statement");
1129 // Handle conditional assembly here before checking for skipping. We
1130 // have to do this so that .endif isn't skipped in a ".if 0" block for
1133 return ParseDirectiveIf(IDLoc);
1134 if (IDVal == ".ifb")
1135 return ParseDirectiveIfb(IDLoc, true);
1136 if (IDVal == ".ifnb")
1137 return ParseDirectiveIfb(IDLoc, false);
1138 if (IDVal == ".ifc")
1139 return ParseDirectiveIfc(IDLoc, true);
1140 if (IDVal == ".ifnc")
1141 return ParseDirectiveIfc(IDLoc, false);
1142 if (IDVal == ".ifdef")
1143 return ParseDirectiveIfdef(IDLoc, true);
1144 if (IDVal == ".ifndef" || IDVal == ".ifnotdef")
1145 return ParseDirectiveIfdef(IDLoc, false);
1146 if (IDVal == ".elseif")
1147 return ParseDirectiveElseIf(IDLoc);
1148 if (IDVal == ".else")
1149 return ParseDirectiveElse(IDLoc);
1150 if (IDVal == ".endif")
1151 return ParseDirectiveEndIf(IDLoc);
1153 // If we are in a ".if 0" block, ignore this statement.
1154 if (TheCondState.Ignore) {
1155 EatToEndOfStatement();
1159 // FIXME: Recurse on local labels?
1161 // See what kind of statement we have.
1162 switch (Lexer.getKind()) {
1163 case AsmToken::Colon: {
1164 CheckForValidSection();
1166 // identifier ':' -> Label.
1169 // Diagnose attempt to use '.' as a label.
1171 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1173 // Diagnose attempt to use a variable as a label.
1175 // FIXME: Diagnostics. Note the location of the definition as a label.
1176 // FIXME: This doesn't diagnose assignment to a symbol which has been
1177 // implicitly marked as external.
1179 if (LocalLabelVal == -1)
1180 Sym = getContext().GetOrCreateSymbol(IDVal);
1182 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
1183 if (!Sym->isUndefined() || Sym->isVariable())
1184 return Error(IDLoc, "invalid symbol redefinition");
1189 // If we are generating dwarf for assembly source files then gather the
1190 // info to make a dwarf label entry for this label if needed.
1191 if (getContext().getGenDwarfForAssembly())
1192 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1195 // Consume any end of statement token, if present, to avoid spurious
1196 // AddBlankLine calls().
1197 if (Lexer.is(AsmToken::EndOfStatement)) {
1199 if (Lexer.is(AsmToken::Eof))
1206 case AsmToken::Equal:
1207 // identifier '=' ... -> assignment statement
1210 return ParseAssignment(IDVal, true);
1212 default: // Normal instruction or directive.
1216 // If macros are enabled, check to see if this is a macro instantiation.
1218 if (const Macro *M = MacroMap.lookup(IDVal))
1219 return HandleMacroEntry(IDVal, IDLoc, M);
1221 // Otherwise, we have a normal instruction or directive.
1222 if (IDVal[0] == '.' && IDVal != ".") {
1224 // Target hook for parsing target specific directives.
1225 if (!getTargetParser().ParseDirective(ID))
1228 // Assembler features
1229 if (IDVal == ".set" || IDVal == ".equ")
1230 return ParseDirectiveSet(IDVal, true);
1231 if (IDVal == ".equiv")
1232 return ParseDirectiveSet(IDVal, false);
1236 if (IDVal == ".ascii")
1237 return ParseDirectiveAscii(IDVal, false);
1238 if (IDVal == ".asciz" || IDVal == ".string")
1239 return ParseDirectiveAscii(IDVal, true);
1241 if (IDVal == ".byte")
1242 return ParseDirectiveValue(1);
1243 if (IDVal == ".short")
1244 return ParseDirectiveValue(2);
1245 if (IDVal == ".value")
1246 return ParseDirectiveValue(2);
1247 if (IDVal == ".2byte")
1248 return ParseDirectiveValue(2);
1249 if (IDVal == ".long")
1250 return ParseDirectiveValue(4);
1251 if (IDVal == ".int")
1252 return ParseDirectiveValue(4);
1253 if (IDVal == ".4byte")
1254 return ParseDirectiveValue(4);
1255 if (IDVal == ".quad")
1256 return ParseDirectiveValue(8);
1257 if (IDVal == ".8byte")
1258 return ParseDirectiveValue(8);
1259 if (IDVal == ".single" || IDVal == ".float")
1260 return ParseDirectiveRealValue(APFloat::IEEEsingle);
1261 if (IDVal == ".double")
1262 return ParseDirectiveRealValue(APFloat::IEEEdouble);
1264 if (IDVal == ".align") {
1265 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1266 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1268 if (IDVal == ".align32") {
1269 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1270 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1272 if (IDVal == ".balign")
1273 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1274 if (IDVal == ".balignw")
1275 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1276 if (IDVal == ".balignl")
1277 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1278 if (IDVal == ".p2align")
1279 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1280 if (IDVal == ".p2alignw")
1281 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1282 if (IDVal == ".p2alignl")
1283 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1285 if (IDVal == ".org")
1286 return ParseDirectiveOrg();
1288 if (IDVal == ".fill")
1289 return ParseDirectiveFill();
1290 if (IDVal == ".space" || IDVal == ".skip")
1291 return ParseDirectiveSpace();
1292 if (IDVal == ".zero")
1293 return ParseDirectiveZero();
1295 // Symbol attribute directives
1297 if (IDVal == ".extern") {
1298 EatToEndOfStatement(); // .extern is the default, ignore it.
1301 if (IDVal == ".globl" || IDVal == ".global")
1302 return ParseDirectiveSymbolAttribute(MCSA_Global);
1303 if (IDVal == ".indirect_symbol")
1304 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
1305 if (IDVal == ".lazy_reference")
1306 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
1307 if (IDVal == ".no_dead_strip")
1308 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1309 if (IDVal == ".symbol_resolver")
1310 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1311 if (IDVal == ".private_extern")
1312 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1313 if (IDVal == ".reference")
1314 return ParseDirectiveSymbolAttribute(MCSA_Reference);
1315 if (IDVal == ".weak_definition")
1316 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1317 if (IDVal == ".weak_reference")
1318 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
1319 if (IDVal == ".weak_def_can_be_hidden")
1320 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1322 if (IDVal == ".comm" || IDVal == ".common")
1323 return ParseDirectiveComm(/*IsLocal=*/false);
1324 if (IDVal == ".lcomm")
1325 return ParseDirectiveComm(/*IsLocal=*/true);
1327 if (IDVal == ".abort")
1328 return ParseDirectiveAbort();
1329 if (IDVal == ".include")
1330 return ParseDirectiveInclude();
1331 if (IDVal == ".incbin")
1332 return ParseDirectiveIncbin();
1334 if (IDVal == ".code16" || IDVal == ".code16gcc")
1335 return TokError(Twine(IDVal) + " not supported yet");
1337 // Macro-like directives
1338 if (IDVal == ".rept")
1339 return ParseDirectiveRept(IDLoc);
1340 if (IDVal == ".irp")
1341 return ParseDirectiveIrp(IDLoc);
1342 if (IDVal == ".irpc")
1343 return ParseDirectiveIrpc(IDLoc);
1344 if (IDVal == ".endr")
1345 return ParseDirectiveEndr(IDLoc);
1347 // Look up the handler in the handler table.
1348 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1349 DirectiveMap.lookup(IDVal);
1351 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1354 return Error(IDLoc, "unknown directive");
1358 if (ParsingInlineAsm && IDVal == "_emit")
1359 return ParseDirectiveEmit(IDLoc, Info);
1361 CheckForValidSection();
1363 // Canonicalize the opcode to lower case.
1364 SmallString<128> OpcodeStr;
1365 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1366 OpcodeStr.push_back(tolower(IDVal[i]));
1368 bool HadError = getTargetParser().ParseInstruction(OpcodeStr.str(), IDLoc,
1369 Info.ParsedOperands);
1371 // Dump the parsed representation, if requested.
1372 if (getShowParsedOperands()) {
1373 SmallString<256> Str;
1374 raw_svector_ostream OS(Str);
1375 OS << "parsed instruction: [";
1376 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
1379 Info.ParsedOperands[i]->print(OS);
1383 PrintMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1386 // If we are generating dwarf for assembly source files and the current
1387 // section is the initial text section then generate a .loc directive for
1389 if (!HadError && getContext().getGenDwarfForAssembly() &&
1390 getContext().getGenDwarfSection() == getStreamer().getCurrentSection() ) {
1391 getStreamer().EmitDwarfLocDirective(getContext().getGenDwarfFileNumber(),
1392 SrcMgr.FindLineNumber(IDLoc, CurBuffer),
1393 0, DWARF2_LINE_DEFAULT_IS_STMT ?
1394 DWARF2_FLAG_IS_STMT : 0, 0, 0,
1398 // If parsing succeeded, match the instruction.
1401 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, Info.Opcode,
1402 Info.ParsedOperands,
1407 // Don't skip the rest of the line, the instruction parser is responsible for
1412 /// EatToEndOfLine uses the Lexer to eat the characters to the end of the line
1413 /// since they may not be able to be tokenized to get to the end of line token.
1414 void AsmParser::EatToEndOfLine() {
1415 if (!Lexer.is(AsmToken::EndOfStatement))
1416 Lexer.LexUntilEndOfLine();
1421 /// ParseCppHashLineFilenameComment as this:
1422 /// ::= # number "filename"
1423 /// or just as a full line comment if it doesn't have a number and a string.
1424 bool AsmParser::ParseCppHashLineFilenameComment(const SMLoc &L) {
1425 Lex(); // Eat the hash token.
1427 if (getLexer().isNot(AsmToken::Integer)) {
1428 // Consume the line since in cases it is not a well-formed line directive,
1429 // as if were simply a full line comment.
1434 int64_t LineNumber = getTok().getIntVal();
1437 if (getLexer().isNot(AsmToken::String)) {
1442 StringRef Filename = getTok().getString();
1443 // Get rid of the enclosing quotes.
1444 Filename = Filename.substr(1, Filename.size()-2);
1446 // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1448 CppHashFilename = Filename;
1449 CppHashLineNumber = LineNumber;
1451 // Ignore any trailing characters, they're just comment.
1456 /// DiagHandler - will use the last parsed cpp hash line filename comment
1457 /// for the Filename and LineNo if any in the diagnostic.
1458 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1459 const AsmParser *Parser = static_cast<const AsmParser*>(Context);
1460 raw_ostream &OS = errs();
1462 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
1463 const SMLoc &DiagLoc = Diag.getLoc();
1464 int DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1465 int CppHashBuf = Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc);
1467 // Like SourceMgr::PrintMessage() we need to print the include stack if any
1468 // before printing the message.
1469 int DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1470 if (!Parser->SavedDiagHandler && DiagCurBuffer > 0) {
1471 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
1472 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
1475 // If we have not parsed a cpp hash line filename comment or the source
1476 // manager changed or buffer changed (like in a nested include) then just
1477 // print the normal diagnostic using its Filename and LineNo.
1478 if (!Parser->CppHashLineNumber ||
1479 &DiagSrcMgr != &Parser->SrcMgr ||
1480 DiagBuf != CppHashBuf) {
1481 if (Parser->SavedDiagHandler)
1482 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
1488 // Use the CppHashFilename and calculate a line number based on the
1489 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for
1491 const std::string Filename = Parser->CppHashFilename;
1493 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
1494 int CppHashLocLineNo =
1495 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf);
1496 int LineNo = Parser->CppHashLineNumber - 1 +
1497 (DiagLocLineNo - CppHashLocLineNo);
1499 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(),
1500 Filename, LineNo, Diag.getColumnNo(),
1501 Diag.getKind(), Diag.getMessage(),
1502 Diag.getLineContents(), Diag.getRanges());
1504 if (Parser->SavedDiagHandler)
1505 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
1507 NewDiag.print(0, OS);
1510 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
1511 // difference being that that function accepts '@' as part of identifiers and
1512 // we can't do that. AsmLexer.cpp should probably be changed to handle
1513 // '@' as a special case when needed.
1514 static bool isIdentifierChar(char c) {
1515 return isalnum(c) || c == '_' || c == '$' || c == '.';
1518 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
1519 const MacroParameters &Parameters,
1520 const MacroArguments &A,
1522 unsigned NParameters = Parameters.size();
1523 if (NParameters != 0 && NParameters != A.size())
1524 return Error(L, "Wrong number of arguments");
1526 // A macro without parameters is handled differently on Darwin:
1527 // gas accepts no arguments and does no substitutions
1528 while (!Body.empty()) {
1529 // Scan for the next substitution.
1530 std::size_t End = Body.size(), Pos = 0;
1531 for (; Pos != End; ++Pos) {
1532 // Check for a substitution or escape.
1534 // This macro has no parameters, look for $0, $1, etc.
1535 if (Body[Pos] != '$' || Pos + 1 == End)
1538 char Next = Body[Pos + 1];
1539 if (Next == '$' || Next == 'n' || isdigit(Next))
1542 // This macro has parameters, look for \foo, \bar, etc.
1543 if (Body[Pos] == '\\' && Pos + 1 != End)
1549 OS << Body.slice(0, Pos);
1551 // Check if we reached the end.
1556 switch (Body[Pos+1]) {
1562 // $n => number of arguments
1567 // $[0-9] => argument
1569 // Missing arguments are ignored.
1570 unsigned Index = Body[Pos+1] - '0';
1571 if (Index >= A.size())
1574 // Otherwise substitute with the token values, with spaces eliminated.
1575 for (MacroArgument::const_iterator it = A[Index].begin(),
1576 ie = A[Index].end(); it != ie; ++it)
1577 OS << it->getString();
1583 unsigned I = Pos + 1;
1584 while (isIdentifierChar(Body[I]) && I + 1 != End)
1587 const char *Begin = Body.data() + Pos +1;
1588 StringRef Argument(Begin, I - (Pos +1));
1590 for (; Index < NParameters; ++Index)
1591 if (Parameters[Index].first == Argument)
1594 if (Index == NParameters) {
1595 if (Body[Pos+1] == '(' && Body[Pos+2] == ')')
1598 OS << '\\' << Argument;
1602 for (MacroArgument::const_iterator it = A[Index].begin(),
1603 ie = A[Index].end(); it != ie; ++it)
1604 if (it->getKind() == AsmToken::String)
1605 OS << it->getStringContents();
1607 OS << it->getString();
1609 Pos += 1 + Argument.size();
1612 // Update the scan point.
1613 Body = Body.substr(Pos);
1619 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1621 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL)
1625 static bool IsOperator(AsmToken::TokenKind kind)
1631 case AsmToken::Plus:
1632 case AsmToken::Minus:
1633 case AsmToken::Tilde:
1634 case AsmToken::Slash:
1635 case AsmToken::Star:
1637 case AsmToken::Equal:
1638 case AsmToken::EqualEqual:
1639 case AsmToken::Pipe:
1640 case AsmToken::PipePipe:
1641 case AsmToken::Caret:
1643 case AsmToken::AmpAmp:
1644 case AsmToken::Exclaim:
1645 case AsmToken::ExclaimEqual:
1646 case AsmToken::Percent:
1647 case AsmToken::Less:
1648 case AsmToken::LessEqual:
1649 case AsmToken::LessLess:
1650 case AsmToken::LessGreater:
1651 case AsmToken::Greater:
1652 case AsmToken::GreaterEqual:
1653 case AsmToken::GreaterGreater:
1658 /// ParseMacroArgument - Extract AsmTokens for a macro argument.
1659 /// This is used for both default macro parameter values and the
1660 /// arguments in macro invocations
1661 bool AsmParser::ParseMacroArgument(MacroArgument &MA,
1662 AsmToken::TokenKind &ArgumentDelimiter) {
1663 unsigned ParenLevel = 0;
1664 unsigned AddTokens = 0;
1666 // gas accepts arguments separated by whitespace, except on Darwin
1668 Lexer.setSkipSpace(false);
1671 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) {
1672 Lexer.setSkipSpace(true);
1673 return TokError("unexpected token in macro instantiation");
1676 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1677 // Spaces and commas cannot be mixed to delimit parameters
1678 if (ArgumentDelimiter == AsmToken::Eof)
1679 ArgumentDelimiter = AsmToken::Comma;
1680 else if (ArgumentDelimiter != AsmToken::Comma) {
1681 Lexer.setSkipSpace(true);
1682 return TokError("expected ' ' for macro argument separator");
1687 if (Lexer.is(AsmToken::Space)) {
1688 Lex(); // Eat spaces
1690 // Spaces can delimit parameters, but could also be part an expression.
1691 // If the token after a space is an operator, add the token and the next
1692 // one into this argument
1693 if (ArgumentDelimiter == AsmToken::Space ||
1694 ArgumentDelimiter == AsmToken::Eof) {
1695 if (IsOperator(Lexer.getKind())) {
1696 // Check to see whether the token is used as an operator,
1697 // or part of an identifier
1698 const char *NextChar = getTok().getEndLoc().getPointer() + 1;
1699 if (*NextChar == ' ')
1703 if (!AddTokens && ParenLevel == 0) {
1704 if (ArgumentDelimiter == AsmToken::Eof &&
1705 !IsOperator(Lexer.getKind()))
1706 ArgumentDelimiter = AsmToken::Space;
1712 // HandleMacroEntry relies on not advancing the lexer here
1713 // to be able to fill in the remaining default parameter values
1714 if (Lexer.is(AsmToken::EndOfStatement))
1717 // Adjust the current parentheses level.
1718 if (Lexer.is(AsmToken::LParen))
1720 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1723 // Append the token to the current argument list.
1724 MA.push_back(getTok());
1730 Lexer.setSkipSpace(true);
1731 if (ParenLevel != 0)
1732 return TokError("unbalanced parentheses in macro argument");
1736 // Parse the macro instantiation arguments.
1737 bool AsmParser::ParseMacroArguments(const Macro *M, MacroArguments &A) {
1738 const unsigned NParameters = M ? M->Parameters.size() : 0;
1739 // Argument delimiter is initially unknown. It will be set by
1740 // ParseMacroArgument()
1741 AsmToken::TokenKind ArgumentDelimiter = AsmToken::Eof;
1743 // Parse two kinds of macro invocations:
1744 // - macros defined without any parameters accept an arbitrary number of them
1745 // - macros defined with parameters accept at most that many of them
1746 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
1750 if (ParseMacroArgument(MA, ArgumentDelimiter))
1753 if (!MA.empty() || !NParameters)
1755 else if (NParameters) {
1756 if (!M->Parameters[Parameter].second.empty())
1757 A.push_back(M->Parameters[Parameter].second);
1760 // At the end of the statement, fill in remaining arguments that have
1761 // default values. If there aren't any, then the next argument is
1762 // required but missing
1763 if (Lexer.is(AsmToken::EndOfStatement)) {
1764 if (NParameters && Parameter < NParameters - 1) {
1765 if (M->Parameters[Parameter + 1].second.empty())
1766 return TokError("macro argument '" +
1767 Twine(M->Parameters[Parameter + 1].first) +
1775 if (Lexer.is(AsmToken::Comma))
1778 return TokError("Too many arguments");
1781 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1783 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1784 // this, although we should protect against infinite loops.
1785 if (ActiveMacros.size() == 20)
1786 return TokError("macros cannot be nested more than 20 levels deep");
1789 if (ParseMacroArguments(M, A))
1792 // Remove any trailing empty arguments. Do this after-the-fact as we have
1793 // to keep empty arguments in the middle of the list or positionality
1794 // gets off. e.g., "foo 1, , 2" vs. "foo 1, 2,"
1795 while (!A.empty() && A.back().empty())
1798 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1799 // to hold the macro body with substitutions.
1800 SmallString<256> Buf;
1801 StringRef Body = M->Body;
1802 raw_svector_ostream OS(Buf);
1804 if (expandMacro(OS, Body, M->Parameters, A, getTok().getLoc()))
1807 // We include the .endmacro in the buffer as our queue to exit the macro
1809 OS << ".endmacro\n";
1811 MemoryBuffer *Instantiation =
1812 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1814 // Create the macro instantiation object and add to the current macro
1815 // instantiation stack.
1816 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1819 ActiveMacros.push_back(MI);
1821 // Jump to the macro instantiation and prime the lexer.
1822 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1823 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1829 void AsmParser::HandleMacroExit() {
1830 // Jump to the EndOfStatement we should return to, and consume it.
1831 JumpToLoc(ActiveMacros.back()->ExitLoc);
1834 // Pop the instantiation entry.
1835 delete ActiveMacros.back();
1836 ActiveMacros.pop_back();
1839 static bool IsUsedIn(const MCSymbol *Sym, const MCExpr *Value) {
1840 switch (Value->getKind()) {
1841 case MCExpr::Binary: {
1842 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr*>(Value);
1843 return IsUsedIn(Sym, BE->getLHS()) || IsUsedIn(Sym, BE->getRHS());
1846 case MCExpr::Target:
1847 case MCExpr::Constant:
1849 case MCExpr::SymbolRef: {
1850 const MCSymbol &S = static_cast<const MCSymbolRefExpr*>(Value)->getSymbol();
1852 return IsUsedIn(Sym, S.getVariableValue());
1856 return IsUsedIn(Sym, static_cast<const MCUnaryExpr*>(Value)->getSubExpr());
1859 llvm_unreachable("Unknown expr kind!");
1862 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef,
1864 // FIXME: Use better location, we should use proper tokens.
1865 SMLoc EqualLoc = Lexer.getLoc();
1867 const MCExpr *Value;
1868 if (ParseExpression(Value))
1871 // Note: we don't count b as used in "a = b". This is to allow
1875 if (Lexer.isNot(AsmToken::EndOfStatement))
1876 return TokError("unexpected token in assignment");
1878 // Error on assignment to '.'.
1880 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported "
1881 "(use '.space' or '.org').)"));
1884 // Eat the end of statement marker.
1887 // Validate that the LHS is allowed to be a variable (either it has not been
1888 // used as a symbol, or it is an absolute symbol).
1889 MCSymbol *Sym = getContext().LookupSymbol(Name);
1891 // Diagnose assignment to a label.
1893 // FIXME: Diagnostics. Note the location of the definition as a label.
1894 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1895 if (IsUsedIn(Sym, Value))
1896 return Error(EqualLoc, "Recursive use of '" + Name + "'");
1897 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
1898 ; // Allow redefinitions of undefined symbols only used in directives.
1899 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
1900 ; // Allow redefinitions of variables that haven't yet been used.
1901 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
1902 return Error(EqualLoc, "redefinition of '" + Name + "'");
1903 else if (!Sym->isVariable())
1904 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1905 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1906 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1909 // Don't count these checks as uses.
1910 Sym->setUsed(false);
1912 Sym = getContext().GetOrCreateSymbol(Name);
1914 // FIXME: Handle '.'.
1916 // Do the assignment.
1917 Out.EmitAssignment(Sym, Value);
1919 Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip);
1925 /// ParseIdentifier:
1928 bool AsmParser::ParseIdentifier(StringRef &Res) {
1929 // The assembler has relaxed rules for accepting identifiers, in particular we
1930 // allow things like '.globl $foo', which would normally be separate
1931 // tokens. At this level, we have already lexed so we cannot (currently)
1932 // handle this as a context dependent token, instead we detect adjacent tokens
1933 // and return the combined identifier.
1934 if (Lexer.is(AsmToken::Dollar)) {
1935 SMLoc DollarLoc = getLexer().getLoc();
1937 // Consume the dollar sign, and check for a following identifier.
1939 if (Lexer.isNot(AsmToken::Identifier))
1942 // We have a '$' followed by an identifier, make sure they are adjacent.
1943 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1946 // Construct the joined identifier and consume the token.
1947 Res = StringRef(DollarLoc.getPointer(),
1948 getTok().getIdentifier().size() + 1);
1953 if (Lexer.isNot(AsmToken::Identifier) &&
1954 Lexer.isNot(AsmToken::String))
1957 Res = getTok().getIdentifier();
1959 Lex(); // Consume the identifier token.
1964 /// ParseDirectiveSet:
1965 /// ::= .equ identifier ',' expression
1966 /// ::= .equiv identifier ',' expression
1967 /// ::= .set identifier ',' expression
1968 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) {
1971 if (ParseIdentifier(Name))
1972 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1974 if (getLexer().isNot(AsmToken::Comma))
1975 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1978 return ParseAssignment(Name, allow_redef, true);
1981 bool AsmParser::ParseEscapedString(std::string &Data) {
1982 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1985 StringRef Str = getTok().getStringContents();
1986 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1987 if (Str[i] != '\\') {
1992 // Recognize escaped characters. Note that this escape semantics currently
1993 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1996 return TokError("unexpected backslash at end of string");
1998 // Recognize octal sequences.
1999 if ((unsigned) (Str[i] - '0') <= 7) {
2000 // Consume up to three octal characters.
2001 unsigned Value = Str[i] - '0';
2003 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
2005 Value = Value * 8 + (Str[i] - '0');
2007 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
2009 Value = Value * 8 + (Str[i] - '0');
2014 return TokError("invalid octal escape sequence (out of range)");
2016 Data += (unsigned char) Value;
2020 // Otherwise recognize individual escapes.
2023 // Just reject invalid escape sequences for now.
2024 return TokError("invalid escape sequence (unrecognized character)");
2026 case 'b': Data += '\b'; break;
2027 case 'f': Data += '\f'; break;
2028 case 'n': Data += '\n'; break;
2029 case 'r': Data += '\r'; break;
2030 case 't': Data += '\t'; break;
2031 case '"': Data += '"'; break;
2032 case '\\': Data += '\\'; break;
2039 /// ParseDirectiveAscii:
2040 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
2041 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
2042 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2043 CheckForValidSection();
2046 if (getLexer().isNot(AsmToken::String))
2047 return TokError("expected string in '" + Twine(IDVal) + "' directive");
2050 if (ParseEscapedString(Data))
2053 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
2055 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
2059 if (getLexer().is(AsmToken::EndOfStatement))
2062 if (getLexer().isNot(AsmToken::Comma))
2063 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
2072 /// ParseDirectiveValue
2073 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
2074 bool AsmParser::ParseDirectiveValue(unsigned Size) {
2075 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2076 CheckForValidSection();
2079 const MCExpr *Value;
2080 SMLoc ExprLoc = getLexer().getLoc();
2081 if (ParseExpression(Value))
2084 // Special case constant expressions to match code generator.
2085 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2086 assert(Size <= 8 && "Invalid size");
2087 uint64_t IntValue = MCE->getValue();
2088 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
2089 return Error(ExprLoc, "literal value out of range for directive");
2090 getStreamer().EmitIntValue(IntValue, Size, DEFAULT_ADDRSPACE);
2092 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
2094 if (getLexer().is(AsmToken::EndOfStatement))
2097 // FIXME: Improve diagnostic.
2098 if (getLexer().isNot(AsmToken::Comma))
2099 return TokError("unexpected token in directive");
2108 /// ParseDirectiveRealValue
2109 /// ::= (.single | .double) [ expression (, expression)* ]
2110 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
2111 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2112 CheckForValidSection();
2115 // We don't truly support arithmetic on floating point expressions, so we
2116 // have to manually parse unary prefixes.
2118 if (getLexer().is(AsmToken::Minus)) {
2121 } else if (getLexer().is(AsmToken::Plus))
2124 if (getLexer().isNot(AsmToken::Integer) &&
2125 getLexer().isNot(AsmToken::Real) &&
2126 getLexer().isNot(AsmToken::Identifier))
2127 return TokError("unexpected token in directive");
2129 // Convert to an APFloat.
2130 APFloat Value(Semantics);
2131 StringRef IDVal = getTok().getString();
2132 if (getLexer().is(AsmToken::Identifier)) {
2133 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
2134 Value = APFloat::getInf(Semantics);
2135 else if (!IDVal.compare_lower("nan"))
2136 Value = APFloat::getNaN(Semantics, false, ~0);
2138 return TokError("invalid floating point literal");
2139 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
2140 APFloat::opInvalidOp)
2141 return TokError("invalid floating point literal");
2145 // Consume the numeric token.
2148 // Emit the value as an integer.
2149 APInt AsInt = Value.bitcastToAPInt();
2150 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
2151 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
2153 if (getLexer().is(AsmToken::EndOfStatement))
2156 if (getLexer().isNot(AsmToken::Comma))
2157 return TokError("unexpected token in directive");
2166 /// ParseDirectiveSpace
2167 /// ::= .space expression [ , expression ]
2168 bool AsmParser::ParseDirectiveSpace() {
2169 CheckForValidSection();
2172 if (ParseAbsoluteExpression(NumBytes))
2175 int64_t FillExpr = 0;
2176 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2177 if (getLexer().isNot(AsmToken::Comma))
2178 return TokError("unexpected token in '.space' directive");
2181 if (ParseAbsoluteExpression(FillExpr))
2184 if (getLexer().isNot(AsmToken::EndOfStatement))
2185 return TokError("unexpected token in '.space' directive");
2191 return TokError("invalid number of bytes in '.space' directive");
2193 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
2194 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
2199 /// ParseDirectiveZero
2200 /// ::= .zero expression
2201 bool AsmParser::ParseDirectiveZero() {
2202 CheckForValidSection();
2205 if (ParseAbsoluteExpression(NumBytes))
2209 if (getLexer().is(AsmToken::Comma)) {
2211 if (ParseAbsoluteExpression(Val))
2215 if (getLexer().isNot(AsmToken::EndOfStatement))
2216 return TokError("unexpected token in '.zero' directive");
2220 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
2225 /// ParseDirectiveFill
2226 /// ::= .fill expression , expression , expression
2227 bool AsmParser::ParseDirectiveFill() {
2228 CheckForValidSection();
2231 if (ParseAbsoluteExpression(NumValues))
2234 if (getLexer().isNot(AsmToken::Comma))
2235 return TokError("unexpected token in '.fill' directive");
2239 if (ParseAbsoluteExpression(FillSize))
2242 if (getLexer().isNot(AsmToken::Comma))
2243 return TokError("unexpected token in '.fill' directive");
2247 if (ParseAbsoluteExpression(FillExpr))
2250 if (getLexer().isNot(AsmToken::EndOfStatement))
2251 return TokError("unexpected token in '.fill' directive");
2255 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
2256 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
2258 for (uint64_t i = 0, e = NumValues; i != e; ++i)
2259 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
2264 /// ParseDirectiveOrg
2265 /// ::= .org expression [ , expression ]
2266 bool AsmParser::ParseDirectiveOrg() {
2267 CheckForValidSection();
2269 const MCExpr *Offset;
2270 SMLoc Loc = getTok().getLoc();
2271 if (ParseExpression(Offset))
2274 // Parse optional fill expression.
2275 int64_t FillExpr = 0;
2276 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2277 if (getLexer().isNot(AsmToken::Comma))
2278 return TokError("unexpected token in '.org' directive");
2281 if (ParseAbsoluteExpression(FillExpr))
2284 if (getLexer().isNot(AsmToken::EndOfStatement))
2285 return TokError("unexpected token in '.org' directive");
2290 // Only limited forms of relocatable expressions are accepted here, it
2291 // has to be relative to the current section. The streamer will return
2292 // 'true' if the expression wasn't evaluatable.
2293 if (getStreamer().EmitValueToOffset(Offset, FillExpr))
2294 return Error(Loc, "expected assembly-time absolute expression");
2299 /// ParseDirectiveAlign
2300 /// ::= {.align, ...} expression [ , expression [ , expression ]]
2301 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2302 CheckForValidSection();
2304 SMLoc AlignmentLoc = getLexer().getLoc();
2306 if (ParseAbsoluteExpression(Alignment))
2310 bool HasFillExpr = false;
2311 int64_t FillExpr = 0;
2312 int64_t MaxBytesToFill = 0;
2313 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2314 if (getLexer().isNot(AsmToken::Comma))
2315 return TokError("unexpected token in directive");
2318 // The fill expression can be omitted while specifying a maximum number of
2319 // alignment bytes, e.g:
2321 if (getLexer().isNot(AsmToken::Comma)) {
2323 if (ParseAbsoluteExpression(FillExpr))
2327 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2328 if (getLexer().isNot(AsmToken::Comma))
2329 return TokError("unexpected token in directive");
2332 MaxBytesLoc = getLexer().getLoc();
2333 if (ParseAbsoluteExpression(MaxBytesToFill))
2336 if (getLexer().isNot(AsmToken::EndOfStatement))
2337 return TokError("unexpected token in directive");
2346 // Compute alignment in bytes.
2348 // FIXME: Diagnose overflow.
2349 if (Alignment >= 32) {
2350 Error(AlignmentLoc, "invalid alignment value");
2354 Alignment = 1ULL << Alignment;
2357 // Diagnose non-sensical max bytes to align.
2358 if (MaxBytesLoc.isValid()) {
2359 if (MaxBytesToFill < 1) {
2360 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2361 "many bytes, ignoring maximum bytes expression");
2365 if (MaxBytesToFill >= Alignment) {
2366 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2372 // Check whether we should use optimal code alignment for this .align
2374 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
2375 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2376 ValueSize == 1 && UseCodeAlign) {
2377 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2379 // FIXME: Target specific behavior about how the "extra" bytes are filled.
2380 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2387 /// ParseDirectiveSymbolAttribute
2388 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
2389 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
2390 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2393 SMLoc Loc = getTok().getLoc();
2395 if (ParseIdentifier(Name))
2396 return Error(Loc, "expected identifier in directive");
2398 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2400 // Assembler local symbols don't make any sense here. Complain loudly.
2401 if (Sym->isTemporary())
2402 return Error(Loc, "non-local symbol required in directive");
2404 getStreamer().EmitSymbolAttribute(Sym, Attr);
2406 if (getLexer().is(AsmToken::EndOfStatement))
2409 if (getLexer().isNot(AsmToken::Comma))
2410 return TokError("unexpected token in directive");
2419 /// ParseDirectiveComm
2420 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
2421 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
2422 CheckForValidSection();
2424 SMLoc IDLoc = getLexer().getLoc();
2426 if (ParseIdentifier(Name))
2427 return TokError("expected identifier in directive");
2429 // Handle the identifier as the key symbol.
2430 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2432 if (getLexer().isNot(AsmToken::Comma))
2433 return TokError("unexpected token in directive");
2437 SMLoc SizeLoc = getLexer().getLoc();
2438 if (ParseAbsoluteExpression(Size))
2441 int64_t Pow2Alignment = 0;
2442 SMLoc Pow2AlignmentLoc;
2443 if (getLexer().is(AsmToken::Comma)) {
2445 Pow2AlignmentLoc = getLexer().getLoc();
2446 if (ParseAbsoluteExpression(Pow2Alignment))
2449 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
2450 if (IsLocal && LCOMM == LCOMM::NoAlignment)
2451 return Error(Pow2AlignmentLoc, "alignment not supported on this target");
2453 // If this target takes alignments in bytes (not log) validate and convert.
2454 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
2455 (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
2456 if (!isPowerOf2_64(Pow2Alignment))
2457 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
2458 Pow2Alignment = Log2_64(Pow2Alignment);
2462 if (getLexer().isNot(AsmToken::EndOfStatement))
2463 return TokError("unexpected token in '.comm' or '.lcomm' directive");
2467 // NOTE: a size of zero for a .comm should create a undefined symbol
2468 // but a size of .lcomm creates a bss symbol of size zero.
2470 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
2471 "be less than zero");
2473 // NOTE: The alignment in the directive is a power of 2 value, the assembler
2474 // may internally end up wanting an alignment in bytes.
2475 // FIXME: Diagnose overflow.
2476 if (Pow2Alignment < 0)
2477 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
2478 "alignment, can't be less than zero");
2480 if (!Sym->isUndefined())
2481 return Error(IDLoc, "invalid symbol redefinition");
2483 // Create the Symbol as a common or local common with Size and Pow2Alignment
2485 getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment);
2489 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
2493 /// ParseDirectiveAbort
2494 /// ::= .abort [... message ...]
2495 bool AsmParser::ParseDirectiveAbort() {
2496 // FIXME: Use loc from directive.
2497 SMLoc Loc = getLexer().getLoc();
2499 StringRef Str = ParseStringToEndOfStatement();
2500 if (getLexer().isNot(AsmToken::EndOfStatement))
2501 return TokError("unexpected token in '.abort' directive");
2506 Error(Loc, ".abort detected. Assembly stopping.");
2508 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
2509 // FIXME: Actually abort assembly here.
2514 /// ParseDirectiveInclude
2515 /// ::= .include "filename"
2516 bool AsmParser::ParseDirectiveInclude() {
2517 if (getLexer().isNot(AsmToken::String))
2518 return TokError("expected string in '.include' directive");
2520 std::string Filename = getTok().getString();
2521 SMLoc IncludeLoc = getLexer().getLoc();
2524 if (getLexer().isNot(AsmToken::EndOfStatement))
2525 return TokError("unexpected token in '.include' directive");
2527 // Strip the quotes.
2528 Filename = Filename.substr(1, Filename.size()-2);
2530 // Attempt to switch the lexer to the included file before consuming the end
2531 // of statement to avoid losing it when we switch.
2532 if (EnterIncludeFile(Filename)) {
2533 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
2540 /// ParseDirectiveIncbin
2541 /// ::= .incbin "filename"
2542 bool AsmParser::ParseDirectiveIncbin() {
2543 if (getLexer().isNot(AsmToken::String))
2544 return TokError("expected string in '.incbin' directive");
2546 std::string Filename = getTok().getString();
2547 SMLoc IncbinLoc = getLexer().getLoc();
2550 if (getLexer().isNot(AsmToken::EndOfStatement))
2551 return TokError("unexpected token in '.incbin' directive");
2553 // Strip the quotes.
2554 Filename = Filename.substr(1, Filename.size()-2);
2556 // Attempt to process the included file.
2557 if (ProcessIncbinFile(Filename)) {
2558 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
2565 /// ParseDirectiveIf
2566 /// ::= .if expression
2567 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
2568 TheCondStack.push_back(TheCondState);
2569 TheCondState.TheCond = AsmCond::IfCond;
2570 if (TheCondState.Ignore) {
2571 EatToEndOfStatement();
2574 if (ParseAbsoluteExpression(ExprValue))
2577 if (getLexer().isNot(AsmToken::EndOfStatement))
2578 return TokError("unexpected token in '.if' directive");
2582 TheCondState.CondMet = ExprValue;
2583 TheCondState.Ignore = !TheCondState.CondMet;
2589 /// ParseDirectiveIfb
2591 bool AsmParser::ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
2592 TheCondStack.push_back(TheCondState);
2593 TheCondState.TheCond = AsmCond::IfCond;
2595 if (TheCondState.Ignore) {
2596 EatToEndOfStatement();
2598 StringRef Str = ParseStringToEndOfStatement();
2600 if (getLexer().isNot(AsmToken::EndOfStatement))
2601 return TokError("unexpected token in '.ifb' directive");
2605 TheCondState.CondMet = ExpectBlank == Str.empty();
2606 TheCondState.Ignore = !TheCondState.CondMet;
2612 /// ParseDirectiveIfc
2613 /// ::= .ifc string1, string2
2614 bool AsmParser::ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
2615 TheCondStack.push_back(TheCondState);
2616 TheCondState.TheCond = AsmCond::IfCond;
2618 if (TheCondState.Ignore) {
2619 EatToEndOfStatement();
2621 StringRef Str1 = ParseStringToComma();
2623 if (getLexer().isNot(AsmToken::Comma))
2624 return TokError("unexpected token in '.ifc' directive");
2628 StringRef Str2 = ParseStringToEndOfStatement();
2630 if (getLexer().isNot(AsmToken::EndOfStatement))
2631 return TokError("unexpected token in '.ifc' directive");
2635 TheCondState.CondMet = ExpectEqual == (Str1 == Str2);
2636 TheCondState.Ignore = !TheCondState.CondMet;
2642 /// ParseDirectiveIfdef
2643 /// ::= .ifdef symbol
2644 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
2646 TheCondStack.push_back(TheCondState);
2647 TheCondState.TheCond = AsmCond::IfCond;
2649 if (TheCondState.Ignore) {
2650 EatToEndOfStatement();
2652 if (ParseIdentifier(Name))
2653 return TokError("expected identifier after '.ifdef'");
2657 MCSymbol *Sym = getContext().LookupSymbol(Name);
2660 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined());
2662 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined());
2663 TheCondState.Ignore = !TheCondState.CondMet;
2669 /// ParseDirectiveElseIf
2670 /// ::= .elseif expression
2671 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
2672 if (TheCondState.TheCond != AsmCond::IfCond &&
2673 TheCondState.TheCond != AsmCond::ElseIfCond)
2674 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
2676 TheCondState.TheCond = AsmCond::ElseIfCond;
2678 bool LastIgnoreState = false;
2679 if (!TheCondStack.empty())
2680 LastIgnoreState = TheCondStack.back().Ignore;
2681 if (LastIgnoreState || TheCondState.CondMet) {
2682 TheCondState.Ignore = true;
2683 EatToEndOfStatement();
2687 if (ParseAbsoluteExpression(ExprValue))
2690 if (getLexer().isNot(AsmToken::EndOfStatement))
2691 return TokError("unexpected token in '.elseif' directive");
2694 TheCondState.CondMet = ExprValue;
2695 TheCondState.Ignore = !TheCondState.CondMet;
2701 /// ParseDirectiveElse
2703 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
2704 if (getLexer().isNot(AsmToken::EndOfStatement))
2705 return TokError("unexpected token in '.else' directive");
2709 if (TheCondState.TheCond != AsmCond::IfCond &&
2710 TheCondState.TheCond != AsmCond::ElseIfCond)
2711 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
2713 TheCondState.TheCond = AsmCond::ElseCond;
2714 bool LastIgnoreState = false;
2715 if (!TheCondStack.empty())
2716 LastIgnoreState = TheCondStack.back().Ignore;
2717 if (LastIgnoreState || TheCondState.CondMet)
2718 TheCondState.Ignore = true;
2720 TheCondState.Ignore = false;
2725 /// ParseDirectiveEndIf
2727 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
2728 if (getLexer().isNot(AsmToken::EndOfStatement))
2729 return TokError("unexpected token in '.endif' directive");
2733 if ((TheCondState.TheCond == AsmCond::NoCond) ||
2734 TheCondStack.empty())
2735 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
2737 if (!TheCondStack.empty()) {
2738 TheCondState = TheCondStack.back();
2739 TheCondStack.pop_back();
2745 /// ParseDirectiveFile
2746 /// ::= .file [number] filename
2747 /// ::= .file number directory filename
2748 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
2749 // FIXME: I'm not sure what this is.
2750 int64_t FileNumber = -1;
2751 SMLoc FileNumberLoc = getLexer().getLoc();
2752 if (getLexer().is(AsmToken::Integer)) {
2753 FileNumber = getTok().getIntVal();
2757 return TokError("file number less than one");
2760 if (getLexer().isNot(AsmToken::String))
2761 return TokError("unexpected token in '.file' directive");
2763 // Usually the directory and filename together, otherwise just the directory.
2764 StringRef Path = getTok().getString();
2765 Path = Path.substr(1, Path.size()-2);
2768 StringRef Directory;
2770 if (getLexer().is(AsmToken::String)) {
2771 if (FileNumber == -1)
2772 return TokError("explicit path specified, but no file number");
2773 Filename = getTok().getString();
2774 Filename = Filename.substr(1, Filename.size()-2);
2781 if (getLexer().isNot(AsmToken::EndOfStatement))
2782 return TokError("unexpected token in '.file' directive");
2784 if (FileNumber == -1)
2785 getStreamer().EmitFileDirective(Filename);
2787 if (getContext().getGenDwarfForAssembly() == true)
2788 Error(DirectiveLoc, "input can't have .file dwarf directives when -g is "
2789 "used to generate dwarf debug info for assembly code");
2791 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename))
2792 Error(FileNumberLoc, "file number already allocated");
2798 /// ParseDirectiveLine
2799 /// ::= .line [number]
2800 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
2801 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2802 if (getLexer().isNot(AsmToken::Integer))
2803 return TokError("unexpected token in '.line' directive");
2805 int64_t LineNumber = getTok().getIntVal();
2809 // FIXME: Do something with the .line.
2812 if (getLexer().isNot(AsmToken::EndOfStatement))
2813 return TokError("unexpected token in '.line' directive");
2819 /// ParseDirectiveLoc
2820 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2821 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2822 /// The first number is a file number, must have been previously assigned with
2823 /// a .file directive, the second number is the line number and optionally the
2824 /// third number is a column position (zero if not specified). The remaining
2825 /// optional items are .loc sub-directives.
2826 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2828 if (getLexer().isNot(AsmToken::Integer))
2829 return TokError("unexpected token in '.loc' directive");
2830 int64_t FileNumber = getTok().getIntVal();
2832 return TokError("file number less than one in '.loc' directive");
2833 if (!getContext().isValidDwarfFileNumber(FileNumber))
2834 return TokError("unassigned file number in '.loc' directive");
2837 int64_t LineNumber = 0;
2838 if (getLexer().is(AsmToken::Integer)) {
2839 LineNumber = getTok().getIntVal();
2841 return TokError("line number less than one in '.loc' directive");
2845 int64_t ColumnPos = 0;
2846 if (getLexer().is(AsmToken::Integer)) {
2847 ColumnPos = getTok().getIntVal();
2849 return TokError("column position less than zero in '.loc' directive");
2853 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2855 int64_t Discriminator = 0;
2856 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2858 if (getLexer().is(AsmToken::EndOfStatement))
2862 SMLoc Loc = getTok().getLoc();
2863 if (getParser().ParseIdentifier(Name))
2864 return TokError("unexpected token in '.loc' directive");
2866 if (Name == "basic_block")
2867 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2868 else if (Name == "prologue_end")
2869 Flags |= DWARF2_FLAG_PROLOGUE_END;
2870 else if (Name == "epilogue_begin")
2871 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2872 else if (Name == "is_stmt") {
2873 SMLoc Loc = getTok().getLoc();
2874 const MCExpr *Value;
2875 if (getParser().ParseExpression(Value))
2877 // The expression must be the constant 0 or 1.
2878 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2879 int Value = MCE->getValue();
2881 Flags &= ~DWARF2_FLAG_IS_STMT;
2882 else if (Value == 1)
2883 Flags |= DWARF2_FLAG_IS_STMT;
2885 return Error(Loc, "is_stmt value not 0 or 1");
2888 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2891 else if (Name == "isa") {
2892 SMLoc Loc = getTok().getLoc();
2893 const MCExpr *Value;
2894 if (getParser().ParseExpression(Value))
2896 // The expression must be a constant greater or equal to 0.
2897 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2898 int Value = MCE->getValue();
2900 return Error(Loc, "isa number less than zero");
2904 return Error(Loc, "isa number not a constant value");
2907 else if (Name == "discriminator") {
2908 if (getParser().ParseAbsoluteExpression(Discriminator))
2912 return Error(Loc, "unknown sub-directive in '.loc' directive");
2915 if (getLexer().is(AsmToken::EndOfStatement))
2920 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2921 Isa, Discriminator, StringRef());
2926 /// ParseDirectiveStabs
2927 /// ::= .stabs string, number, number, number
2928 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2929 SMLoc DirectiveLoc) {
2930 return TokError("unsupported directive '" + Directive + "'");
2933 /// ParseDirectiveCFISections
2934 /// ::= .cfi_sections section [, section]
2935 bool GenericAsmParser::ParseDirectiveCFISections(StringRef,
2936 SMLoc DirectiveLoc) {
2941 if (getParser().ParseIdentifier(Name))
2942 return TokError("Expected an identifier");
2944 if (Name == ".eh_frame")
2946 else if (Name == ".debug_frame")
2949 if (getLexer().is(AsmToken::Comma)) {
2952 if (getParser().ParseIdentifier(Name))
2953 return TokError("Expected an identifier");
2955 if (Name == ".eh_frame")
2957 else if (Name == ".debug_frame")
2961 getStreamer().EmitCFISections(EH, Debug);
2966 /// ParseDirectiveCFIStartProc
2967 /// ::= .cfi_startproc
2968 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef,
2969 SMLoc DirectiveLoc) {
2970 getStreamer().EmitCFIStartProc();
2974 /// ParseDirectiveCFIEndProc
2975 /// ::= .cfi_endproc
2976 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) {
2977 getStreamer().EmitCFIEndProc();
2981 /// ParseRegisterOrRegisterNumber - parse register name or number.
2982 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register,
2983 SMLoc DirectiveLoc) {
2986 if (getLexer().isNot(AsmToken::Integer)) {
2987 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc,
2990 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true);
2992 return getParser().ParseAbsoluteExpression(Register);
2997 /// ParseDirectiveCFIDefCfa
2998 /// ::= .cfi_def_cfa register, offset
2999 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef,
3000 SMLoc DirectiveLoc) {
3001 int64_t Register = 0;
3002 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3005 if (getLexer().isNot(AsmToken::Comma))
3006 return TokError("unexpected token in directive");
3010 if (getParser().ParseAbsoluteExpression(Offset))
3013 getStreamer().EmitCFIDefCfa(Register, Offset);
3017 /// ParseDirectiveCFIDefCfaOffset
3018 /// ::= .cfi_def_cfa_offset offset
3019 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef,
3020 SMLoc DirectiveLoc) {
3022 if (getParser().ParseAbsoluteExpression(Offset))
3025 getStreamer().EmitCFIDefCfaOffset(Offset);
3029 /// ParseDirectiveCFIAdjustCfaOffset
3030 /// ::= .cfi_adjust_cfa_offset adjustment
3031 bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef,
3032 SMLoc DirectiveLoc) {
3033 int64_t Adjustment = 0;
3034 if (getParser().ParseAbsoluteExpression(Adjustment))
3037 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
3041 /// ParseDirectiveCFIDefCfaRegister
3042 /// ::= .cfi_def_cfa_register register
3043 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef,
3044 SMLoc DirectiveLoc) {
3045 int64_t Register = 0;
3046 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3049 getStreamer().EmitCFIDefCfaRegister(Register);
3053 /// ParseDirectiveCFIOffset
3054 /// ::= .cfi_offset register, offset
3055 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) {
3056 int64_t Register = 0;
3059 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3062 if (getLexer().isNot(AsmToken::Comma))
3063 return TokError("unexpected token in directive");
3066 if (getParser().ParseAbsoluteExpression(Offset))
3069 getStreamer().EmitCFIOffset(Register, Offset);
3073 /// ParseDirectiveCFIRelOffset
3074 /// ::= .cfi_rel_offset register, offset
3075 bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef,
3076 SMLoc DirectiveLoc) {
3077 int64_t Register = 0;
3079 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3082 if (getLexer().isNot(AsmToken::Comma))
3083 return TokError("unexpected token in directive");
3087 if (getParser().ParseAbsoluteExpression(Offset))
3090 getStreamer().EmitCFIRelOffset(Register, Offset);
3094 static bool isValidEncoding(int64_t Encoding) {
3095 if (Encoding & ~0xff)
3098 if (Encoding == dwarf::DW_EH_PE_omit)
3101 const unsigned Format = Encoding & 0xf;
3102 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
3103 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
3104 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
3105 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
3108 const unsigned Application = Encoding & 0x70;
3109 if (Application != dwarf::DW_EH_PE_absptr &&
3110 Application != dwarf::DW_EH_PE_pcrel)
3116 /// ParseDirectiveCFIPersonalityOrLsda
3117 /// ::= .cfi_personality encoding, [symbol_name]
3118 /// ::= .cfi_lsda encoding, [symbol_name]
3119 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal,
3120 SMLoc DirectiveLoc) {
3121 int64_t Encoding = 0;
3122 if (getParser().ParseAbsoluteExpression(Encoding))
3124 if (Encoding == dwarf::DW_EH_PE_omit)
3127 if (!isValidEncoding(Encoding))
3128 return TokError("unsupported encoding.");
3130 if (getLexer().isNot(AsmToken::Comma))
3131 return TokError("unexpected token in directive");
3135 if (getParser().ParseIdentifier(Name))
3136 return TokError("expected identifier in directive");
3138 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3140 if (IDVal == ".cfi_personality")
3141 getStreamer().EmitCFIPersonality(Sym, Encoding);
3143 assert(IDVal == ".cfi_lsda");
3144 getStreamer().EmitCFILsda(Sym, Encoding);
3149 /// ParseDirectiveCFIRememberState
3150 /// ::= .cfi_remember_state
3151 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal,
3152 SMLoc DirectiveLoc) {
3153 getStreamer().EmitCFIRememberState();
3157 /// ParseDirectiveCFIRestoreState
3158 /// ::= .cfi_remember_state
3159 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal,
3160 SMLoc DirectiveLoc) {
3161 getStreamer().EmitCFIRestoreState();
3165 /// ParseDirectiveCFISameValue
3166 /// ::= .cfi_same_value register
3167 bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal,
3168 SMLoc DirectiveLoc) {
3169 int64_t Register = 0;
3171 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3174 getStreamer().EmitCFISameValue(Register);
3179 /// ParseDirectiveCFIRestore
3180 /// ::= .cfi_restore register
3181 bool GenericAsmParser::ParseDirectiveCFIRestore(StringRef IDVal,
3182 SMLoc DirectiveLoc) {
3183 int64_t Register = 0;
3184 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3187 getStreamer().EmitCFIRestore(Register);
3192 /// ParseDirectiveCFIEscape
3193 /// ::= .cfi_escape expression[,...]
3194 bool GenericAsmParser::ParseDirectiveCFIEscape(StringRef IDVal,
3195 SMLoc DirectiveLoc) {
3198 if (getParser().ParseAbsoluteExpression(CurrValue))
3201 Values.push_back((uint8_t)CurrValue);
3203 while (getLexer().is(AsmToken::Comma)) {
3206 if (getParser().ParseAbsoluteExpression(CurrValue))
3209 Values.push_back((uint8_t)CurrValue);
3212 getStreamer().EmitCFIEscape(Values);
3216 /// ParseDirectiveCFISignalFrame
3217 /// ::= .cfi_signal_frame
3218 bool GenericAsmParser::ParseDirectiveCFISignalFrame(StringRef Directive,
3219 SMLoc DirectiveLoc) {
3220 if (getLexer().isNot(AsmToken::EndOfStatement))
3221 return Error(getLexer().getLoc(),
3222 "unexpected token in '" + Directive + "' directive");
3224 getStreamer().EmitCFISignalFrame();
3229 /// ParseDirectiveMacrosOnOff
3232 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
3233 SMLoc DirectiveLoc) {
3234 if (getLexer().isNot(AsmToken::EndOfStatement))
3235 return Error(getLexer().getLoc(),
3236 "unexpected token in '" + Directive + "' directive");
3238 getParser().MacrosEnabled = Directive == ".macros_on";
3243 /// ParseDirectiveMacro
3244 /// ::= .macro name [parameters]
3245 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
3246 SMLoc DirectiveLoc) {
3248 if (getParser().ParseIdentifier(Name))
3249 return TokError("expected identifier in '.macro' directive");
3251 MacroParameters Parameters;
3252 // Argument delimiter is initially unknown. It will be set by
3253 // ParseMacroArgument()
3254 AsmToken::TokenKind ArgumentDelimiter = AsmToken::Eof;
3255 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3257 MacroParameter Parameter;
3258 if (getParser().ParseIdentifier(Parameter.first))
3259 return TokError("expected identifier in '.macro' directive");
3261 if (getLexer().is(AsmToken::Equal)) {
3263 if (getParser().ParseMacroArgument(Parameter.second, ArgumentDelimiter))
3267 Parameters.push_back(Parameter);
3269 if (getLexer().is(AsmToken::Comma))
3271 else if (getLexer().is(AsmToken::EndOfStatement))
3276 // Eat the end of statement.
3279 AsmToken EndToken, StartToken = getTok();
3281 // Lex the macro definition.
3283 // Check whether we have reached the end of the file.
3284 if (getLexer().is(AsmToken::Eof))
3285 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3287 // Otherwise, check whether we have reach the .endmacro.
3288 if (getLexer().is(AsmToken::Identifier) &&
3289 (getTok().getIdentifier() == ".endm" ||
3290 getTok().getIdentifier() == ".endmacro")) {
3291 EndToken = getTok();
3293 if (getLexer().isNot(AsmToken::EndOfStatement))
3294 return TokError("unexpected token in '" + EndToken.getIdentifier() +
3299 // Otherwise, scan til the end of the statement.
3300 getParser().EatToEndOfStatement();
3303 if (getParser().MacroMap.lookup(Name)) {
3304 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3307 const char *BodyStart = StartToken.getLoc().getPointer();
3308 const char *BodyEnd = EndToken.getLoc().getPointer();
3309 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3310 getParser().MacroMap[Name] = new Macro(Name, Body, Parameters);
3314 /// ParseDirectiveEndMacro
3317 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
3318 SMLoc DirectiveLoc) {
3319 if (getLexer().isNot(AsmToken::EndOfStatement))
3320 return TokError("unexpected token in '" + Directive + "' directive");
3322 // If we are inside a macro instantiation, terminate the current
3324 if (!getParser().ActiveMacros.empty()) {
3325 getParser().HandleMacroExit();
3329 // Otherwise, this .endmacro is a stray entry in the file; well formed
3330 // .endmacro directives are handled during the macro definition parsing.
3331 return TokError("unexpected '" + Directive + "' in file, "
3332 "no current macro definition");
3335 /// ParseDirectivePurgeMacro
3337 bool GenericAsmParser::ParseDirectivePurgeMacro(StringRef Directive,
3338 SMLoc DirectiveLoc) {
3340 if (getParser().ParseIdentifier(Name))
3341 return TokError("expected identifier in '.purgem' directive");
3343 if (getLexer().isNot(AsmToken::EndOfStatement))
3344 return TokError("unexpected token in '.purgem' directive");
3346 StringMap<Macro*>::iterator I = getParser().MacroMap.find(Name);
3347 if (I == getParser().MacroMap.end())
3348 return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
3350 // Undefine the macro.
3351 delete I->getValue();
3352 getParser().MacroMap.erase(I);
3356 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
3357 getParser().CheckForValidSection();
3359 const MCExpr *Value;
3361 if (getParser().ParseExpression(Value))
3364 if (getLexer().isNot(AsmToken::EndOfStatement))
3365 return TokError("unexpected token in directive");
3367 if (DirName[1] == 's')
3368 getStreamer().EmitSLEB128Value(Value);
3370 getStreamer().EmitULEB128Value(Value);
3375 Macro *AsmParser::ParseMacroLikeBody(SMLoc DirectiveLoc) {
3376 AsmToken EndToken, StartToken = getTok();
3378 unsigned NestLevel = 0;
3380 // Check whether we have reached the end of the file.
3381 if (getLexer().is(AsmToken::Eof)) {
3382 Error(DirectiveLoc, "no matching '.endr' in definition");
3386 if (Lexer.is(AsmToken::Identifier) &&
3387 (getTok().getIdentifier() == ".rept")) {
3391 // Otherwise, check whether we have reached the .endr.
3392 if (Lexer.is(AsmToken::Identifier) &&
3393 getTok().getIdentifier() == ".endr") {
3394 if (NestLevel == 0) {
3395 EndToken = getTok();
3397 if (Lexer.isNot(AsmToken::EndOfStatement)) {
3398 TokError("unexpected token in '.endr' directive");
3406 // Otherwise, scan till the end of the statement.
3407 EatToEndOfStatement();
3410 const char *BodyStart = StartToken.getLoc().getPointer();
3411 const char *BodyEnd = EndToken.getLoc().getPointer();
3412 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3414 // We Are Anonymous.
3416 MacroParameters Parameters;
3417 return new Macro(Name, Body, Parameters);
3420 void AsmParser::InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc,
3421 raw_svector_ostream &OS) {
3424 MemoryBuffer *Instantiation =
3425 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
3427 // Create the macro instantiation object and add to the current macro
3428 // instantiation stack.
3429 MacroInstantiation *MI = new MacroInstantiation(M, DirectiveLoc,
3432 ActiveMacros.push_back(MI);
3434 // Jump to the macro instantiation and prime the lexer.
3435 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
3436 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
3440 bool AsmParser::ParseDirectiveRept(SMLoc DirectiveLoc) {
3442 if (ParseAbsoluteExpression(Count))
3443 return TokError("unexpected token in '.rept' directive");
3446 return TokError("Count is negative");
3448 if (Lexer.isNot(AsmToken::EndOfStatement))
3449 return TokError("unexpected token in '.rept' directive");
3451 // Eat the end of statement.
3454 // Lex the rept definition.
3455 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3459 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3460 // to hold the macro body with substitutions.
3461 SmallString<256> Buf;
3462 MacroParameters Parameters;
3464 raw_svector_ostream OS(Buf);
3466 if (expandMacro(OS, M->Body, Parameters, A, getTok().getLoc()))
3469 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3474 /// ParseDirectiveIrp
3475 /// ::= .irp symbol,values
3476 bool AsmParser::ParseDirectiveIrp(SMLoc DirectiveLoc) {
3477 MacroParameters Parameters;
3478 MacroParameter Parameter;
3480 if (ParseIdentifier(Parameter.first))
3481 return TokError("expected identifier in '.irp' directive");
3483 Parameters.push_back(Parameter);
3485 if (Lexer.isNot(AsmToken::Comma))
3486 return TokError("expected comma in '.irp' directive");
3491 if (ParseMacroArguments(0, A))
3494 // Eat the end of statement.
3497 // Lex the irp definition.
3498 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3502 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3503 // to hold the macro body with substitutions.
3504 SmallString<256> Buf;
3505 raw_svector_ostream OS(Buf);
3507 for (MacroArguments::iterator i = A.begin(), e = A.end(); i != e; ++i) {
3508 MacroArguments Args;
3511 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc()))
3515 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3520 /// ParseDirectiveIrpc
3521 /// ::= .irpc symbol,values
3522 bool AsmParser::ParseDirectiveIrpc(SMLoc DirectiveLoc) {
3523 MacroParameters Parameters;
3524 MacroParameter Parameter;
3526 if (ParseIdentifier(Parameter.first))
3527 return TokError("expected identifier in '.irpc' directive");
3529 Parameters.push_back(Parameter);
3531 if (Lexer.isNot(AsmToken::Comma))
3532 return TokError("expected comma in '.irpc' directive");
3537 if (ParseMacroArguments(0, A))
3540 if (A.size() != 1 || A.front().size() != 1)
3541 return TokError("unexpected token in '.irpc' directive");
3543 // Eat the end of statement.
3546 // Lex the irpc definition.
3547 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3551 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3552 // to hold the macro body with substitutions.
3553 SmallString<256> Buf;
3554 raw_svector_ostream OS(Buf);
3556 StringRef Values = A.front().front().getString();
3557 std::size_t I, End = Values.size();
3558 for (I = 0; I < End; ++I) {
3560 Arg.push_back(AsmToken(AsmToken::Identifier, Values.slice(I, I+1)));
3562 MacroArguments Args;
3563 Args.push_back(Arg);
3565 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc()))
3569 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3574 bool AsmParser::ParseDirectiveEndr(SMLoc DirectiveLoc) {
3575 if (ActiveMacros.empty())
3576 return TokError("unmatched '.endr' directive");
3578 // The only .repl that should get here are the ones created by
3579 // InstantiateMacroLikeBody.
3580 assert(getLexer().is(AsmToken::EndOfStatement));
3587 enum AsmRewriteKind {
3597 AsmRewriteKind Kind;
3602 AsmRewrite(AsmRewriteKind kind, SMLoc loc, unsigned len, unsigned size = 0)
3603 : Kind(kind), Loc(loc), Len(len), Size(size) { }
3607 bool AsmParser::ParseDirectiveEmit(SMLoc IDLoc, ParseStatementInfo &Info) {
3608 const MCExpr *Value;
3609 SMLoc ExprLoc = getLexer().getLoc();
3610 if (ParseExpression(Value))
3612 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
3614 return Error(ExprLoc, "unexpected expression in _emit");
3615 uint64_t IntValue = MCE->getValue();
3616 if (!isUIntN(8, IntValue) && !isIntN(8, IntValue))
3617 return Error(ExprLoc, "literal value out of range for directive");
3619 Info.AsmRewrites->push_back(AsmRewrite(AOK_Emit, IDLoc, 5));
3623 bool AsmParser::ParseMSInlineAsm(void *AsmLoc, std::string &AsmString,
3624 unsigned &NumOutputs, unsigned &NumInputs,
3625 SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
3626 SmallVectorImpl<std::string> &Constraints,
3627 SmallVectorImpl<std::string> &Clobbers,
3628 const MCInstrInfo *MII,
3629 const MCInstPrinter *IP,
3630 MCAsmParserSemaCallback &SI) {
3631 SmallVector<void *, 4> InputDecls;
3632 SmallVector<void *, 4> OutputDecls;
3633 SmallVector<bool, 4> InputDeclsOffsetOf;
3634 SmallVector<bool, 4> OutputDeclsOffsetOf;
3635 SmallVector<std::string, 4> InputConstraints;
3636 SmallVector<std::string, 4> OutputConstraints;
3637 std::set<std::string> ClobberRegs;
3639 SmallVector<struct AsmRewrite, 4> AsmStrRewrites;
3644 // While we have input, parse each statement.
3645 unsigned InputIdx = 0;
3646 unsigned OutputIdx = 0;
3647 while (getLexer().isNot(AsmToken::Eof)) {
3648 ParseStatementInfo Info(&AsmStrRewrites);
3649 if (ParseStatement(Info))
3652 if (Info.Opcode != ~0U) {
3653 const MCInstrDesc &Desc = MII->get(Info.Opcode);
3655 // Build the list of clobbers, outputs and inputs.
3656 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
3657 MCParsedAsmOperand *Operand = Info.ParsedOperands[i];
3660 if (Operand->isImm()) {
3661 AsmStrRewrites.push_back(AsmRewrite(AOK_Imm,
3662 Operand->getStartLoc(),
3663 Operand->getNameLen()));
3667 // Register operand.
3668 if (Operand->isReg()) {
3669 unsigned NumDefs = Desc.getNumDefs();
3671 if (NumDefs && Operand->getMCOperandNum() < NumDefs) {
3673 raw_string_ostream OS(Reg);
3674 IP->printRegName(OS, Operand->getReg());
3675 ClobberRegs.insert(StringRef(OS.str()));
3680 // Expr/Input or Output.
3682 void *OpDecl = SI.LookupInlineAsmIdentifier(Operand->getName(), AsmLoc,
3685 bool isOutput = (i == 1) && Desc.mayStore();
3686 if (Operand->needSizeDirective() && !Operand->isOffsetOf())
3687 AsmStrRewrites.push_back(AsmRewrite(AOK_SizeDirective,
3688 Operand->getStartLoc(), 0,
3689 Operand->getMemSize()));
3691 // Don't emit the offset directive.
3692 if (Operand->isOffsetOf())
3693 AsmStrRewrites.push_back(AsmRewrite(AOK_Skip,
3694 Operand->getOffsetOfLoc(), 7));
3697 std::string Constraint = "=";
3699 OutputDecls.push_back(OpDecl);
3700 OutputDeclsOffsetOf.push_back(Operand->isOffsetOf());
3701 Constraint += Operand->getConstraint().str();
3702 OutputConstraints.push_back(Constraint);
3703 AsmStrRewrites.push_back(AsmRewrite(AOK_Output,
3704 Operand->getStartLoc(),
3705 Operand->getNameLen()));
3707 InputDecls.push_back(OpDecl);
3708 InputDeclsOffsetOf.push_back(Operand->isOffsetOf());
3709 InputConstraints.push_back(Operand->getConstraint().str());
3710 AsmStrRewrites.push_back(AsmRewrite(AOK_Input,
3711 Operand->getStartLoc(),
3712 Operand->getNameLen()));
3719 // Set the number of Outputs and Inputs.
3720 NumOutputs = OutputDecls.size();
3721 NumInputs = InputDecls.size();
3723 // Set the unique clobbers.
3724 for (std::set<std::string>::iterator I = ClobberRegs.begin(),
3725 E = ClobberRegs.end(); I != E; ++I)
3726 Clobbers.push_back(*I);
3728 // Merge the various outputs and inputs. Output are expected first.
3729 if (NumOutputs || NumInputs) {
3730 unsigned NumExprs = NumOutputs + NumInputs;
3731 OpDecls.resize(NumExprs);
3732 Constraints.resize(NumExprs);
3733 // FIXME: Constraints are hard coded to 'm', but we need an 'r'
3734 // constraint for offsetof. This needs to be cleaned up!
3735 for (unsigned i = 0; i < NumOutputs; ++i) {
3736 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsOffsetOf[i]);
3737 Constraints[i] = OutputDeclsOffsetOf[i] ? "=r" : OutputConstraints[i];
3739 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
3740 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsOffsetOf[i]);
3741 Constraints[j] = InputDeclsOffsetOf[i] ? "r" : InputConstraints[i];
3745 // Build the IR assembly string.
3746 std::string AsmStringIR;
3747 AsmRewriteKind PrevKind = AOK_Imm;
3748 raw_string_ostream OS(AsmStringIR);
3749 const char *Start = SrcMgr.getMemoryBuffer(0)->getBufferStart();
3750 for (SmallVectorImpl<struct AsmRewrite>::iterator
3751 I = AsmStrRewrites.begin(), E = AsmStrRewrites.end(); I != E; ++I) {
3752 const char *Loc = (*I).Loc.getPointer();
3754 AsmRewriteKind Kind = (*I).Kind;
3756 // Emit everything up to the immediate/expression. If the previous rewrite
3757 // was a size directive, then this has already been done.
3758 if (PrevKind != AOK_SizeDirective)
3759 OS << StringRef(Start, Loc - Start);
3762 // Skip the original expression.
3763 if (Kind == AOK_Skip) {
3764 Start = Loc + (*I).Len;
3768 // Rewrite expressions in $N notation.
3772 OS << Twine("$$") + StringRef(Loc, (*I).Len);
3782 case AOK_SizeDirective:
3785 case 8: OS << "byte ptr "; break;
3786 case 16: OS << "word ptr "; break;
3787 case 32: OS << "dword ptr "; break;
3788 case 64: OS << "qword ptr "; break;
3789 case 80: OS << "xword ptr "; break;
3790 case 128: OS << "xmmword ptr "; break;
3791 case 256: OS << "ymmword ptr "; break;
3799 // Skip the original expression.
3800 if (Kind != AOK_SizeDirective)
3801 Start = Loc + (*I).Len;
3804 // Emit the remainder of the asm string.
3805 const char *AsmEnd = SrcMgr.getMemoryBuffer(0)->getBufferEnd();
3806 if (Start != AsmEnd)
3807 OS << StringRef(Start, AsmEnd - Start);
3809 AsmString = OS.str();
3813 /// \brief Create an MCAsmParser instance.
3814 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM,
3815 MCContext &C, MCStreamer &Out,
3816 const MCAsmInfo &MAI) {
3817 return new AsmParser(SM, C, Out, MAI);