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/STLExtras.h"
16 #include "llvm/ADT/SmallString.h"
17 #include "llvm/ADT/StringMap.h"
18 #include "llvm/ADT/Twine.h"
19 #include "llvm/MC/MCAsmInfo.h"
20 #include "llvm/MC/MCContext.h"
21 #include "llvm/MC/MCDwarf.h"
22 #include "llvm/MC/MCExpr.h"
23 #include "llvm/MC/MCInstPrinter.h"
24 #include "llvm/MC/MCInstrInfo.h"
25 #include "llvm/MC/MCObjectFileInfo.h"
26 #include "llvm/MC/MCParser/AsmCond.h"
27 #include "llvm/MC/MCParser/AsmLexer.h"
28 #include "llvm/MC/MCParser/MCAsmParser.h"
29 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
30 #include "llvm/MC/MCRegisterInfo.h"
31 #include "llvm/MC/MCSectionMachO.h"
32 #include "llvm/MC/MCStreamer.h"
33 #include "llvm/MC/MCSymbol.h"
34 #include "llvm/MC/MCTargetAsmParser.h"
35 #include "llvm/Support/CommandLine.h"
36 #include "llvm/Support/ErrorHandling.h"
37 #include "llvm/Support/MathExtras.h"
38 #include "llvm/Support/MemoryBuffer.h"
39 #include "llvm/Support/SourceMgr.h"
40 #include "llvm/Support/raw_ostream.h"
49 FatalAssemblerWarnings("fatal-assembler-warnings",
50 cl::desc("Consider warnings as error"));
52 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() {}
55 /// \brief Helper types for tracking macro definitions.
56 typedef std::vector<AsmToken> MCAsmMacroArgument;
57 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
59 struct MCAsmMacroParameter {
61 MCAsmMacroArgument Value;
65 MCAsmMacroParameter() : Required(false), Vararg(false) {}
68 typedef std::vector<MCAsmMacroParameter> MCAsmMacroParameters;
73 MCAsmMacroParameters Parameters;
76 MCAsmMacro(StringRef N, StringRef B, ArrayRef<MCAsmMacroParameter> P) :
77 Name(N), Body(B), Parameters(P) {}
80 /// \brief Helper class for storing information about an active macro
82 struct MacroInstantiation {
83 /// The macro being instantiated.
84 const MCAsmMacro *TheMacro;
86 /// The macro instantiation with substitutions.
87 MemoryBuffer *Instantiation;
89 /// The location of the instantiation.
90 SMLoc InstantiationLoc;
92 /// The buffer where parsing should resume upon instantiation completion.
95 /// The location where parsing should resume upon instantiation completion.
99 MacroInstantiation(const MCAsmMacro *M, SMLoc IL, int EB, SMLoc EL,
103 struct ParseStatementInfo {
104 /// \brief The parsed operands from the last parsed statement.
105 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands;
107 /// \brief The opcode from the last parsed instruction.
110 /// \brief Was there an error parsing the inline assembly?
113 SmallVectorImpl<AsmRewrite> *AsmRewrites;
115 ParseStatementInfo() : Opcode(~0U), ParseError(false), AsmRewrites(nullptr) {}
116 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
117 : Opcode(~0), ParseError(false), AsmRewrites(rewrites) {}
120 /// \brief The concrete assembly parser instance.
121 class AsmParser : public MCAsmParser {
122 AsmParser(const AsmParser &) LLVM_DELETED_FUNCTION;
123 void operator=(const AsmParser &) LLVM_DELETED_FUNCTION;
128 const MCAsmInfo &MAI;
130 SourceMgr::DiagHandlerTy SavedDiagHandler;
131 void *SavedDiagContext;
132 MCAsmParserExtension *PlatformParser;
134 /// This is the current buffer index we're lexing from as managed by the
135 /// SourceMgr object.
138 AsmCond TheCondState;
139 std::vector<AsmCond> TheCondStack;
141 /// \brief maps directive names to handler methods in parser
142 /// extensions. Extensions register themselves in this map by calling
143 /// addDirectiveHandler.
144 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
146 /// \brief Map of currently defined macros.
147 StringMap<MCAsmMacro*> MacroMap;
149 /// \brief Stack of active macro instantiations.
150 std::vector<MacroInstantiation*> ActiveMacros;
152 /// \brief List of bodies of anonymous macros.
153 std::deque<MCAsmMacro> MacroLikeBodies;
155 /// Boolean tracking whether macro substitution is enabled.
156 unsigned MacrosEnabledFlag : 1;
158 /// Flag tracking whether any errors have been encountered.
159 unsigned HadError : 1;
161 /// The values from the last parsed cpp hash file line comment if any.
162 StringRef CppHashFilename;
163 int64_t CppHashLineNumber;
166 /// When generating dwarf for assembly source files we need to calculate the
167 /// logical line number based on the last parsed cpp hash file line comment
168 /// and current line. Since this is slow and messes up the SourceMgr's
169 /// cache we save the last info we queried with SrcMgr.FindLineNumber().
170 SMLoc LastQueryIDLoc;
172 unsigned LastQueryLine;
174 /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
175 unsigned AssemblerDialect;
177 /// \brief is Darwin compatibility enabled?
180 /// \brief Are we parsing ms-style inline assembly?
181 bool ParsingInlineAsm;
184 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
185 const MCAsmInfo &MAI);
186 virtual ~AsmParser();
188 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
190 void addDirectiveHandler(StringRef Directive,
191 ExtensionDirectiveHandler Handler) override {
192 ExtensionDirectiveMap[Directive] = Handler;
196 /// @name MCAsmParser Interface
199 SourceMgr &getSourceManager() override { return SrcMgr; }
200 MCAsmLexer &getLexer() override { return Lexer; }
201 MCContext &getContext() override { return Ctx; }
202 MCStreamer &getStreamer() override { return Out; }
203 unsigned getAssemblerDialect() override {
204 if (AssemblerDialect == ~0U)
205 return MAI.getAssemblerDialect();
207 return AssemblerDialect;
209 void setAssemblerDialect(unsigned i) override {
210 AssemblerDialect = i;
213 void Note(SMLoc L, const Twine &Msg,
214 ArrayRef<SMRange> Ranges = None) override;
215 bool Warning(SMLoc L, const Twine &Msg,
216 ArrayRef<SMRange> Ranges = None) override;
217 bool Error(SMLoc L, const Twine &Msg,
218 ArrayRef<SMRange> Ranges = None) override;
220 const AsmToken &Lex() override;
222 void setParsingInlineAsm(bool V) override { ParsingInlineAsm = V; }
223 bool isParsingInlineAsm() override { return ParsingInlineAsm; }
225 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString,
226 unsigned &NumOutputs, unsigned &NumInputs,
227 SmallVectorImpl<std::pair<void *,bool> > &OpDecls,
228 SmallVectorImpl<std::string> &Constraints,
229 SmallVectorImpl<std::string> &Clobbers,
230 const MCInstrInfo *MII, const MCInstPrinter *IP,
231 MCAsmParserSemaCallback &SI) override;
233 bool parseExpression(const MCExpr *&Res);
234 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
235 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override;
236 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
237 bool parseAbsoluteExpression(int64_t &Res) override;
239 /// \brief Parse an identifier or string (as a quoted identifier)
240 /// and set \p Res to the identifier contents.
241 bool parseIdentifier(StringRef &Res) override;
242 void eatToEndOfStatement() override;
244 void checkForValidSection() override;
249 bool parseStatement(ParseStatementInfo &Info);
250 void eatToEndOfLine();
251 bool parseCppHashLineFilenameComment(const SMLoc &L);
253 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body,
254 ArrayRef<MCAsmMacroParameter> Parameters);
255 bool expandMacro(raw_svector_ostream &OS, StringRef Body,
256 ArrayRef<MCAsmMacroParameter> Parameters,
257 ArrayRef<MCAsmMacroArgument> A,
260 /// \brief Are macros enabled in the parser?
261 bool areMacrosEnabled() {return MacrosEnabledFlag;}
263 /// \brief Control a flag in the parser that enables or disables macros.
264 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;}
266 /// \brief Lookup a previously defined macro.
267 /// \param Name Macro name.
268 /// \returns Pointer to macro. NULL if no such macro was defined.
269 const MCAsmMacro* lookupMacro(StringRef Name);
271 /// \brief Define a new macro with the given name and information.
272 void defineMacro(StringRef Name, const MCAsmMacro& Macro);
274 /// \brief Undefine a macro. If no such macro was defined, it's a no-op.
275 void undefineMacro(StringRef Name);
277 /// \brief Are we inside a macro instantiation?
278 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();}
280 /// \brief Handle entry to macro instantiation.
282 /// \param M The macro.
283 /// \param NameLoc Instantiation location.
284 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc);
286 /// \brief Handle exit from macro instantiation.
287 void handleMacroExit();
289 /// \brief Extract AsmTokens for a macro argument.
290 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg);
292 /// \brief Parse all macro arguments for a given macro.
293 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A);
295 void printMacroInstantiations();
296 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
297 ArrayRef<SMRange> Ranges = None) const {
298 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
300 static void DiagHandler(const SMDiagnostic &Diag, void *Context);
302 /// \brief Enter the specified file. This returns true on failure.
303 bool enterIncludeFile(const std::string &Filename);
305 /// \brief Process the specified file for the .incbin directive.
306 /// This returns true on failure.
307 bool processIncbinFile(const std::string &Filename);
309 /// \brief Reset the current lexer position to that given by \p Loc. The
310 /// current token is not set; clients should ensure Lex() is called
313 /// \param InBuffer If not -1, should be the known buffer id that contains the
315 void jumpToLoc(SMLoc Loc, int InBuffer=-1);
317 /// \brief Parse up to the end of statement and a return the contents from the
318 /// current token until the end of the statement; the current token on exit
319 /// will be either the EndOfStatement or EOF.
320 StringRef parseStringToEndOfStatement() override;
322 /// \brief Parse until the end of a statement or a comma is encountered,
323 /// return the contents from the current token up to the end or comma.
324 StringRef parseStringToComma();
326 bool parseAssignment(StringRef Name, bool allow_redef,
327 bool NoDeadStrip = false);
329 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
330 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
331 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
333 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
335 // Generic (target and platform independent) directive parsing.
337 DK_NO_DIRECTIVE, // Placeholder
338 DK_SET, DK_EQU, DK_EQUIV, DK_ASCII, DK_ASCIZ, DK_STRING, DK_BYTE, DK_SHORT,
339 DK_VALUE, DK_2BYTE, DK_LONG, DK_INT, DK_4BYTE, DK_QUAD, DK_8BYTE, DK_OCTA,
340 DK_SINGLE, DK_FLOAT, DK_DOUBLE, DK_ALIGN, DK_ALIGN32, DK_BALIGN, DK_BALIGNW,
341 DK_BALIGNL, DK_P2ALIGN, DK_P2ALIGNW, DK_P2ALIGNL, DK_ORG, DK_FILL, DK_ENDR,
342 DK_BUNDLE_ALIGN_MODE, DK_BUNDLE_LOCK, DK_BUNDLE_UNLOCK,
343 DK_ZERO, DK_EXTERN, DK_GLOBL, DK_GLOBAL,
344 DK_LAZY_REFERENCE, DK_NO_DEAD_STRIP, DK_SYMBOL_RESOLVER, DK_PRIVATE_EXTERN,
345 DK_REFERENCE, DK_WEAK_DEFINITION, DK_WEAK_REFERENCE,
346 DK_WEAK_DEF_CAN_BE_HIDDEN, DK_COMM, DK_COMMON, DK_LCOMM, DK_ABORT,
347 DK_INCLUDE, DK_INCBIN, DK_CODE16, DK_CODE16GCC, DK_REPT, DK_IRP, DK_IRPC,
348 DK_IF, DK_IFEQ, DK_IFGE, DK_IFGT, DK_IFLE, DK_IFLT, DK_IFNE, DK_IFB,
349 DK_IFNB, DK_IFC, DK_IFEQS, DK_IFNC, DK_IFDEF, DK_IFNDEF, DK_IFNOTDEF,
350 DK_ELSEIF, DK_ELSE, DK_ENDIF,
351 DK_SPACE, DK_SKIP, DK_FILE, DK_LINE, DK_LOC, DK_STABS,
352 DK_CFI_SECTIONS, DK_CFI_STARTPROC, DK_CFI_ENDPROC, DK_CFI_DEF_CFA,
353 DK_CFI_DEF_CFA_OFFSET, DK_CFI_ADJUST_CFA_OFFSET, DK_CFI_DEF_CFA_REGISTER,
354 DK_CFI_OFFSET, DK_CFI_REL_OFFSET, DK_CFI_PERSONALITY, DK_CFI_LSDA,
355 DK_CFI_REMEMBER_STATE, DK_CFI_RESTORE_STATE, DK_CFI_SAME_VALUE,
356 DK_CFI_RESTORE, DK_CFI_ESCAPE, DK_CFI_SIGNAL_FRAME, DK_CFI_UNDEFINED,
357 DK_CFI_REGISTER, DK_CFI_WINDOW_SAVE,
358 DK_MACROS_ON, DK_MACROS_OFF, DK_MACRO, DK_ENDM, DK_ENDMACRO, DK_PURGEM,
359 DK_SLEB128, DK_ULEB128,
364 /// \brief Maps directive name --> DirectiveKind enum, for
365 /// directives parsed by this class.
366 StringMap<DirectiveKind> DirectiveKindMap;
368 // ".ascii", ".asciz", ".string"
369 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
370 bool parseDirectiveValue(unsigned Size); // ".byte", ".long", ...
371 bool parseDirectiveOctaValue(); // ".octa"
372 bool parseDirectiveRealValue(const fltSemantics &); // ".single", ...
373 bool parseDirectiveFill(); // ".fill"
374 bool parseDirectiveZero(); // ".zero"
375 // ".set", ".equ", ".equiv"
376 bool parseDirectiveSet(StringRef IDVal, bool allow_redef);
377 bool parseDirectiveOrg(); // ".org"
378 // ".align{,32}", ".p2align{,w,l}"
379 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
381 // ".file", ".line", ".loc", ".stabs"
382 bool parseDirectiveFile(SMLoc DirectiveLoc);
383 bool parseDirectiveLine();
384 bool parseDirectiveLoc();
385 bool parseDirectiveStabs();
388 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
389 bool parseDirectiveCFIWindowSave();
390 bool parseDirectiveCFISections();
391 bool parseDirectiveCFIStartProc();
392 bool parseDirectiveCFIEndProc();
393 bool parseDirectiveCFIDefCfaOffset();
394 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
395 bool parseDirectiveCFIAdjustCfaOffset();
396 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
397 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
398 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
399 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
400 bool parseDirectiveCFIRememberState();
401 bool parseDirectiveCFIRestoreState();
402 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
403 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
404 bool parseDirectiveCFIEscape();
405 bool parseDirectiveCFISignalFrame();
406 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
409 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
410 bool parseDirectiveEndMacro(StringRef Directive);
411 bool parseDirectiveMacro(SMLoc DirectiveLoc);
412 bool parseDirectiveMacrosOnOff(StringRef Directive);
414 // ".bundle_align_mode"
415 bool parseDirectiveBundleAlignMode();
417 bool parseDirectiveBundleLock();
419 bool parseDirectiveBundleUnlock();
422 bool parseDirectiveSpace(StringRef IDVal);
424 // .sleb128 (Signed=true) and .uleb128 (Signed=false)
425 bool parseDirectiveLEB128(bool Signed);
427 /// \brief Parse a directive like ".globl" which
428 /// accepts a single symbol (which should be a label or an external).
429 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
431 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
433 bool parseDirectiveAbort(); // ".abort"
434 bool parseDirectiveInclude(); // ".include"
435 bool parseDirectiveIncbin(); // ".incbin"
437 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne"
438 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
439 // ".ifb" or ".ifnb", depending on ExpectBlank.
440 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
441 // ".ifc" or ".ifnc", depending on ExpectEqual.
442 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
444 bool parseDirectiveIfeqs(SMLoc DirectiveLoc);
445 // ".ifdef" or ".ifndef", depending on expect_defined
446 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
447 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
448 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else"
449 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
450 bool parseEscapedString(std::string &Data) override;
452 const MCExpr *applyModifierToExpr(const MCExpr *E,
453 MCSymbolRefExpr::VariantKind Variant);
455 // Macro-like directives
456 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
457 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
458 raw_svector_ostream &OS);
459 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive);
460 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp"
461 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
462 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
464 // "_emit" or "__emit"
465 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
469 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
472 bool parseDirectiveEnd(SMLoc DirectiveLoc);
474 // ".err" or ".error"
475 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage);
477 void initializeDirectiveKindMap();
483 extern MCAsmParserExtension *createDarwinAsmParser();
484 extern MCAsmParserExtension *createELFAsmParser();
485 extern MCAsmParserExtension *createCOFFAsmParser();
489 enum { DEFAULT_ADDRSPACE = 0 };
491 AsmParser::AsmParser(SourceMgr &_SM, MCContext &_Ctx, MCStreamer &_Out,
492 const MCAsmInfo &_MAI)
493 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM),
494 PlatformParser(nullptr), CurBuffer(0), MacrosEnabledFlag(true),
495 HadError(false), CppHashLineNumber(0), AssemblerDialect(~0U),
496 IsDarwin(false), ParsingInlineAsm(false) {
497 // Save the old handler.
498 SavedDiagHandler = SrcMgr.getDiagHandler();
499 SavedDiagContext = SrcMgr.getDiagContext();
500 // Set our own handler which calls the saved handler.
501 SrcMgr.setDiagHandler(DiagHandler, this);
502 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
504 // Initialize the platform / file format parser.
505 switch (_Ctx.getObjectFileInfo()->getObjectFileType()) {
506 case MCObjectFileInfo::IsCOFF:
507 PlatformParser = createCOFFAsmParser();
508 PlatformParser->Initialize(*this);
510 case MCObjectFileInfo::IsMachO:
511 PlatformParser = createDarwinAsmParser();
512 PlatformParser->Initialize(*this);
515 case MCObjectFileInfo::IsELF:
516 PlatformParser = createELFAsmParser();
517 PlatformParser->Initialize(*this);
521 initializeDirectiveKindMap();
524 AsmParser::~AsmParser() {
525 assert((HadError || ActiveMacros.empty()) &&
526 "Unexpected active macro instantiation!");
528 // Destroy any macros.
529 for (StringMap<MCAsmMacro *>::iterator it = MacroMap.begin(),
532 delete it->getValue();
534 delete PlatformParser;
537 void AsmParser::printMacroInstantiations() {
538 // Print the active macro instantiation stack.
539 for (std::vector<MacroInstantiation *>::const_reverse_iterator
540 it = ActiveMacros.rbegin(),
541 ie = ActiveMacros.rend();
543 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
544 "while in macro instantiation");
547 void AsmParser::Note(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
548 printMessage(L, SourceMgr::DK_Note, Msg, Ranges);
549 printMacroInstantiations();
552 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
553 if (FatalAssemblerWarnings)
554 return Error(L, Msg, Ranges);
555 printMessage(L, SourceMgr::DK_Warning, Msg, Ranges);
556 printMacroInstantiations();
560 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
562 printMessage(L, SourceMgr::DK_Error, Msg, Ranges);
563 printMacroInstantiations();
567 bool AsmParser::enterIncludeFile(const std::string &Filename) {
568 std::string IncludedFile;
569 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
575 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
580 /// Process the specified .incbin file by searching for it in the include paths
581 /// then just emitting the byte contents of the file to the streamer. This
582 /// returns true on failure.
583 bool AsmParser::processIncbinFile(const std::string &Filename) {
584 std::string IncludedFile;
585 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
589 // Pick up the bytes from the file and emit them.
590 getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer());
594 void AsmParser::jumpToLoc(SMLoc Loc, int InBuffer) {
595 if (InBuffer != -1) {
596 CurBuffer = InBuffer;
598 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
600 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
603 const AsmToken &AsmParser::Lex() {
604 const AsmToken *tok = &Lexer.Lex();
606 if (tok->is(AsmToken::Eof)) {
607 // If this is the end of an included file, pop the parent file off the
609 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
610 if (ParentIncludeLoc != SMLoc()) {
611 jumpToLoc(ParentIncludeLoc);
616 if (tok->is(AsmToken::Error))
617 Error(Lexer.getErrLoc(), Lexer.getErr());
622 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
623 // Create the initial section, if requested.
624 if (!NoInitialTextSection)
631 AsmCond StartingCondState = TheCondState;
633 // If we are generating dwarf for assembly source files save the initial text
634 // section and generate a .file directive.
635 if (getContext().getGenDwarfForAssembly()) {
636 MCSymbol *SectionStartSym = getContext().CreateTempSymbol();
637 getStreamer().EmitLabel(SectionStartSym);
638 auto InsertResult = getContext().addGenDwarfSection(
639 getStreamer().getCurrentSection().first);
640 assert(InsertResult.second && ".text section should not have debug info yet");
641 InsertResult.first->second.first = SectionStartSym;
642 getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective(
643 0, StringRef(), getContext().getMainFileName()));
646 // While we have input, parse each statement.
647 while (Lexer.isNot(AsmToken::Eof)) {
648 ParseStatementInfo Info;
649 if (!parseStatement(Info))
652 // We had an error, validate that one was emitted and recover by skipping to
654 assert(HadError && "Parse statement returned an error, but none emitted!");
655 eatToEndOfStatement();
658 if (TheCondState.TheCond != StartingCondState.TheCond ||
659 TheCondState.Ignore != StartingCondState.Ignore)
660 return TokError("unmatched .ifs or .elses");
662 // Check to see there are no empty DwarfFile slots.
663 const auto &LineTables = getContext().getMCDwarfLineTables();
664 if (!LineTables.empty()) {
666 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
667 if (File.Name.empty() && Index != 0)
668 TokError("unassigned file number: " + Twine(Index) +
669 " for .file directives");
674 // Check to see that all assembler local symbols were actually defined.
675 // Targets that don't do subsections via symbols may not want this, though,
676 // so conservatively exclude them. Only do this if we're finalizing, though,
677 // as otherwise we won't necessarilly have seen everything yet.
678 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) {
679 const MCContext::SymbolTable &Symbols = getContext().getSymbols();
680 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(),
683 MCSymbol *Sym = i->getValue();
684 // Variable symbols may not be marked as defined, so check those
685 // explicitly. If we know it's a variable, we have a definition for
686 // the purposes of this check.
687 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
688 // FIXME: We would really like to refer back to where the symbol was
689 // first referenced for a source location. We need to add something
690 // to track that. Currently, we just point to the end of the file.
692 getLexer().getLoc(), SourceMgr::DK_Error,
693 "assembler local symbol '" + Sym->getName() + "' not defined");
697 // Finalize the output stream if there are no errors and if the client wants
699 if (!HadError && !NoFinalize)
705 void AsmParser::checkForValidSection() {
706 if (!ParsingInlineAsm && !getStreamer().getCurrentSection().first) {
707 TokError("expected section directive before assembly directive");
712 /// \brief Throw away the rest of the line for testing purposes.
713 void AsmParser::eatToEndOfStatement() {
714 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
718 if (Lexer.is(AsmToken::EndOfStatement))
722 StringRef AsmParser::parseStringToEndOfStatement() {
723 const char *Start = getTok().getLoc().getPointer();
725 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
728 const char *End = getTok().getLoc().getPointer();
729 return StringRef(Start, End - Start);
732 StringRef AsmParser::parseStringToComma() {
733 const char *Start = getTok().getLoc().getPointer();
735 while (Lexer.isNot(AsmToken::EndOfStatement) &&
736 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof))
739 const char *End = getTok().getLoc().getPointer();
740 return StringRef(Start, End - Start);
743 /// \brief Parse a paren expression and return it.
744 /// NOTE: This assumes the leading '(' has already been consumed.
746 /// parenexpr ::= expr)
748 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
749 if (parseExpression(Res))
751 if (Lexer.isNot(AsmToken::RParen))
752 return TokError("expected ')' in parentheses expression");
753 EndLoc = Lexer.getTok().getEndLoc();
758 /// \brief Parse a bracket expression and return it.
759 /// NOTE: This assumes the leading '[' has already been consumed.
761 /// bracketexpr ::= expr]
763 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
764 if (parseExpression(Res))
766 if (Lexer.isNot(AsmToken::RBrac))
767 return TokError("expected ']' in brackets expression");
768 EndLoc = Lexer.getTok().getEndLoc();
773 /// \brief Parse a primary expression and return it.
774 /// primaryexpr ::= (parenexpr
775 /// primaryexpr ::= symbol
776 /// primaryexpr ::= number
777 /// primaryexpr ::= '.'
778 /// primaryexpr ::= ~,+,- primaryexpr
779 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
780 SMLoc FirstTokenLoc = getLexer().getLoc();
781 AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
782 switch (FirstTokenKind) {
784 return TokError("unknown token in expression");
785 // If we have an error assume that we've already handled it.
786 case AsmToken::Error:
788 case AsmToken::Exclaim:
789 Lex(); // Eat the operator.
790 if (parsePrimaryExpr(Res, EndLoc))
792 Res = MCUnaryExpr::CreateLNot(Res, getContext());
794 case AsmToken::Dollar:
796 case AsmToken::String:
797 case AsmToken::Identifier: {
798 StringRef Identifier;
799 if (parseIdentifier(Identifier)) {
800 if (FirstTokenKind == AsmToken::Dollar) {
801 if (Lexer.getMAI().getDollarIsPC()) {
802 // This is a '$' reference, which references the current PC. Emit a
803 // temporary label to the streamer and refer to it.
804 MCSymbol *Sym = Ctx.CreateTempSymbol();
806 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
808 EndLoc = FirstTokenLoc;
811 return Error(FirstTokenLoc, "invalid token in expression");
814 // Parse symbol variant
815 std::pair<StringRef, StringRef> Split;
816 if (!MAI.useParensForSymbolVariant()) {
817 if (FirstTokenKind == AsmToken::String) {
818 if (Lexer.is(AsmToken::At)) {
819 Lexer.Lex(); // eat @
820 SMLoc AtLoc = getLexer().getLoc();
822 if (parseIdentifier(VName))
823 return Error(AtLoc, "expected symbol variant after '@'");
825 Split = std::make_pair(Identifier, VName);
828 Split = Identifier.split('@');
830 } else if (Lexer.is(AsmToken::LParen)) {
831 Lexer.Lex(); // eat (
833 parseIdentifier(VName);
834 if (Lexer.isNot(AsmToken::RParen)) {
835 return Error(Lexer.getTok().getLoc(),
836 "unexpected token in variant, expected ')'");
838 Lexer.Lex(); // eat )
839 Split = std::make_pair(Identifier, VName);
842 EndLoc = SMLoc::getFromPointer(Identifier.end());
844 // This is a symbol reference.
845 StringRef SymbolName = Identifier;
846 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
848 // Lookup the symbol variant if used.
849 if (Split.second.size()) {
850 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
851 if (Variant != MCSymbolRefExpr::VK_Invalid) {
852 SymbolName = Split.first;
853 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
854 Variant = MCSymbolRefExpr::VK_None;
856 return Error(SMLoc::getFromPointer(Split.second.begin()),
857 "invalid variant '" + Split.second + "'");
861 MCSymbol *Sym = getContext().GetOrCreateSymbol(SymbolName);
863 // If this is an absolute variable reference, substitute it now to preserve
864 // semantics in the face of reassignment.
865 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
867 return Error(EndLoc, "unexpected modifier on variable reference");
869 Res = Sym->getVariableValue();
873 // Otherwise create a symbol ref.
874 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
877 case AsmToken::BigNum:
878 return TokError("literal value out of range for directive");
879 case AsmToken::Integer: {
880 SMLoc Loc = getTok().getLoc();
881 int64_t IntVal = getTok().getIntVal();
882 Res = MCConstantExpr::Create(IntVal, getContext());
883 EndLoc = Lexer.getTok().getEndLoc();
885 // Look for 'b' or 'f' following an Integer as a directional label
886 if (Lexer.getKind() == AsmToken::Identifier) {
887 StringRef IDVal = getTok().getString();
888 // Lookup the symbol variant if used.
889 std::pair<StringRef, StringRef> Split = IDVal.split('@');
890 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
891 if (Split.first.size() != IDVal.size()) {
892 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
893 if (Variant == MCSymbolRefExpr::VK_Invalid)
894 return TokError("invalid variant '" + Split.second + "'");
897 if (IDVal == "f" || IDVal == "b") {
899 Ctx.GetDirectionalLocalSymbol(IntVal, IDVal == "b");
900 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
901 if (IDVal == "b" && Sym->isUndefined())
902 return Error(Loc, "invalid reference to undefined symbol");
903 EndLoc = Lexer.getTok().getEndLoc();
904 Lex(); // Eat identifier.
909 case AsmToken::Real: {
910 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
911 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
912 Res = MCConstantExpr::Create(IntVal, getContext());
913 EndLoc = Lexer.getTok().getEndLoc();
917 case AsmToken::Dot: {
918 // This is a '.' reference, which references the current PC. Emit a
919 // temporary label to the streamer and refer to it.
920 MCSymbol *Sym = Ctx.CreateTempSymbol();
922 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
923 EndLoc = Lexer.getTok().getEndLoc();
924 Lex(); // Eat identifier.
927 case AsmToken::LParen:
928 Lex(); // Eat the '('.
929 return parseParenExpr(Res, EndLoc);
930 case AsmToken::LBrac:
931 if (!PlatformParser->HasBracketExpressions())
932 return TokError("brackets expression not supported on this target");
933 Lex(); // Eat the '['.
934 return parseBracketExpr(Res, EndLoc);
935 case AsmToken::Minus:
936 Lex(); // Eat the operator.
937 if (parsePrimaryExpr(Res, EndLoc))
939 Res = MCUnaryExpr::CreateMinus(Res, getContext());
942 Lex(); // Eat the operator.
943 if (parsePrimaryExpr(Res, EndLoc))
945 Res = MCUnaryExpr::CreatePlus(Res, getContext());
947 case AsmToken::Tilde:
948 Lex(); // Eat the operator.
949 if (parsePrimaryExpr(Res, EndLoc))
951 Res = MCUnaryExpr::CreateNot(Res, getContext());
956 bool AsmParser::parseExpression(const MCExpr *&Res) {
958 return parseExpression(Res, EndLoc);
962 AsmParser::applyModifierToExpr(const MCExpr *E,
963 MCSymbolRefExpr::VariantKind Variant) {
964 // Ask the target implementation about this expression first.
965 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx);
968 // Recurse over the given expression, rebuilding it to apply the given variant
969 // if there is exactly one symbol.
970 switch (E->getKind()) {
972 case MCExpr::Constant:
975 case MCExpr::SymbolRef: {
976 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
978 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
979 TokError("invalid variant on expression '" + getTok().getIdentifier() +
980 "' (already modified)");
984 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
987 case MCExpr::Unary: {
988 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
989 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant);
992 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
995 case MCExpr::Binary: {
996 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
997 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant);
998 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant);
1008 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
1012 llvm_unreachable("Invalid expression kind!");
1015 /// \brief Parse an expression and return it.
1017 /// expr ::= expr &&,|| expr -> lowest.
1018 /// expr ::= expr |,^,&,! expr
1019 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr
1020 /// expr ::= expr <<,>> expr
1021 /// expr ::= expr +,- expr
1022 /// expr ::= expr *,/,% expr -> highest.
1023 /// expr ::= primaryexpr
1025 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1026 // Parse the expression.
1028 if (parsePrimaryExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc))
1031 // As a special case, we support 'a op b @ modifier' by rewriting the
1032 // expression to include the modifier. This is inefficient, but in general we
1033 // expect users to use 'a@modifier op b'.
1034 if (Lexer.getKind() == AsmToken::At) {
1037 if (Lexer.isNot(AsmToken::Identifier))
1038 return TokError("unexpected symbol modifier following '@'");
1040 MCSymbolRefExpr::VariantKind Variant =
1041 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
1042 if (Variant == MCSymbolRefExpr::VK_Invalid)
1043 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
1045 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant);
1047 return TokError("invalid modifier '" + getTok().getIdentifier() +
1048 "' (no symbols present)");
1055 // Try to constant fold it up front, if possible.
1057 if (Res->EvaluateAsAbsolute(Value))
1058 Res = MCConstantExpr::Create(Value, getContext());
1063 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1065 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1068 bool AsmParser::parseAbsoluteExpression(int64_t &Res) {
1071 SMLoc StartLoc = Lexer.getLoc();
1072 if (parseExpression(Expr))
1075 if (!Expr->EvaluateAsAbsolute(Res))
1076 return Error(StartLoc, "expected absolute expression");
1081 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
1082 MCBinaryExpr::Opcode &Kind) {
1085 return 0; // not a binop.
1087 // Lowest Precedence: &&, ||
1088 case AsmToken::AmpAmp:
1089 Kind = MCBinaryExpr::LAnd;
1091 case AsmToken::PipePipe:
1092 Kind = MCBinaryExpr::LOr;
1095 // Low Precedence: |, &, ^
1097 // FIXME: gas seems to support '!' as an infix operator?
1098 case AsmToken::Pipe:
1099 Kind = MCBinaryExpr::Or;
1101 case AsmToken::Caret:
1102 Kind = MCBinaryExpr::Xor;
1105 Kind = MCBinaryExpr::And;
1108 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
1109 case AsmToken::EqualEqual:
1110 Kind = MCBinaryExpr::EQ;
1112 case AsmToken::ExclaimEqual:
1113 case AsmToken::LessGreater:
1114 Kind = MCBinaryExpr::NE;
1116 case AsmToken::Less:
1117 Kind = MCBinaryExpr::LT;
1119 case AsmToken::LessEqual:
1120 Kind = MCBinaryExpr::LTE;
1122 case AsmToken::Greater:
1123 Kind = MCBinaryExpr::GT;
1125 case AsmToken::GreaterEqual:
1126 Kind = MCBinaryExpr::GTE;
1129 // Intermediate Precedence: <<, >>
1130 case AsmToken::LessLess:
1131 Kind = MCBinaryExpr::Shl;
1133 case AsmToken::GreaterGreater:
1134 Kind = MCBinaryExpr::Shr;
1137 // High Intermediate Precedence: +, -
1138 case AsmToken::Plus:
1139 Kind = MCBinaryExpr::Add;
1141 case AsmToken::Minus:
1142 Kind = MCBinaryExpr::Sub;
1145 // Highest Precedence: *, /, %
1146 case AsmToken::Star:
1147 Kind = MCBinaryExpr::Mul;
1149 case AsmToken::Slash:
1150 Kind = MCBinaryExpr::Div;
1152 case AsmToken::Percent:
1153 Kind = MCBinaryExpr::Mod;
1158 /// \brief Parse all binary operators with precedence >= 'Precedence'.
1159 /// Res contains the LHS of the expression on input.
1160 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1163 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1164 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1166 // If the next token is lower precedence than we are allowed to eat, return
1167 // successfully with what we ate already.
1168 if (TokPrec < Precedence)
1173 // Eat the next primary expression.
1175 if (parsePrimaryExpr(RHS, EndLoc))
1178 // If BinOp binds less tightly with RHS than the operator after RHS, let
1179 // the pending operator take RHS as its LHS.
1180 MCBinaryExpr::Opcode Dummy;
1181 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1182 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
1185 // Merge LHS and RHS according to operator.
1186 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
1191 /// ::= EndOfStatement
1192 /// ::= Label* Directive ...Operands... EndOfStatement
1193 /// ::= Label* Identifier OperandList* EndOfStatement
1194 bool AsmParser::parseStatement(ParseStatementInfo &Info) {
1195 if (Lexer.is(AsmToken::EndOfStatement)) {
1201 // Statements always start with an identifier or are a full line comment.
1202 AsmToken ID = getTok();
1203 SMLoc IDLoc = ID.getLoc();
1205 int64_t LocalLabelVal = -1;
1206 // A full line comment is a '#' as the first token.
1207 if (Lexer.is(AsmToken::Hash))
1208 return parseCppHashLineFilenameComment(IDLoc);
1210 // Allow an integer followed by a ':' as a directional local label.
1211 if (Lexer.is(AsmToken::Integer)) {
1212 LocalLabelVal = getTok().getIntVal();
1213 if (LocalLabelVal < 0) {
1214 if (!TheCondState.Ignore)
1215 return TokError("unexpected token at start of statement");
1218 IDVal = getTok().getString();
1219 Lex(); // Consume the integer token to be used as an identifier token.
1220 if (Lexer.getKind() != AsmToken::Colon) {
1221 if (!TheCondState.Ignore)
1222 return TokError("unexpected token at start of statement");
1225 } else if (Lexer.is(AsmToken::Dot)) {
1226 // Treat '.' as a valid identifier in this context.
1229 } else if (parseIdentifier(IDVal)) {
1230 if (!TheCondState.Ignore)
1231 return TokError("unexpected token at start of statement");
1235 // Handle conditional assembly here before checking for skipping. We
1236 // have to do this so that .endif isn't skipped in a ".if 0" block for
1238 StringMap<DirectiveKind>::const_iterator DirKindIt =
1239 DirectiveKindMap.find(IDVal);
1240 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
1242 : DirKindIt->getValue();
1253 return parseDirectiveIf(IDLoc, DirKind);
1255 return parseDirectiveIfb(IDLoc, true);
1257 return parseDirectiveIfb(IDLoc, false);
1259 return parseDirectiveIfc(IDLoc, true);
1261 return parseDirectiveIfeqs(IDLoc);
1263 return parseDirectiveIfc(IDLoc, false);
1265 return parseDirectiveIfdef(IDLoc, true);
1268 return parseDirectiveIfdef(IDLoc, false);
1270 return parseDirectiveElseIf(IDLoc);
1272 return parseDirectiveElse(IDLoc);
1274 return parseDirectiveEndIf(IDLoc);
1277 // Ignore the statement if in the middle of inactive conditional
1279 if (TheCondState.Ignore) {
1280 eatToEndOfStatement();
1284 // FIXME: Recurse on local labels?
1286 // See what kind of statement we have.
1287 switch (Lexer.getKind()) {
1288 case AsmToken::Colon: {
1289 checkForValidSection();
1291 // identifier ':' -> Label.
1294 // Diagnose attempt to use '.' as a label.
1296 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1298 // Diagnose attempt to use a variable as a label.
1300 // FIXME: Diagnostics. Note the location of the definition as a label.
1301 // FIXME: This doesn't diagnose assignment to a symbol which has been
1302 // implicitly marked as external.
1304 if (LocalLabelVal == -1)
1305 Sym = getContext().GetOrCreateSymbol(IDVal);
1307 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
1308 if (!Sym->isUndefined() || Sym->isVariable())
1309 return Error(IDLoc, "invalid symbol redefinition");
1312 if (!ParsingInlineAsm)
1315 // If we are generating dwarf for assembly source files then gather the
1316 // info to make a dwarf label entry for this label if needed.
1317 if (getContext().getGenDwarfForAssembly())
1318 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1321 getTargetParser().onLabelParsed(Sym);
1323 // Consume any end of statement token, if present, to avoid spurious
1324 // AddBlankLine calls().
1325 if (Lexer.is(AsmToken::EndOfStatement)) {
1327 if (Lexer.is(AsmToken::Eof))
1334 case AsmToken::Equal:
1335 // identifier '=' ... -> assignment statement
1338 return parseAssignment(IDVal, true);
1340 default: // Normal instruction or directive.
1344 // If macros are enabled, check to see if this is a macro instantiation.
1345 if (areMacrosEnabled())
1346 if (const MCAsmMacro *M = lookupMacro(IDVal)) {
1347 return handleMacroEntry(M, IDLoc);
1350 // Otherwise, we have a normal instruction or directive.
1352 // Directives start with "."
1353 if (IDVal[0] == '.' && IDVal != ".") {
1354 // There are several entities interested in parsing directives:
1356 // 1. The target-specific assembly parser. Some directives are target
1357 // specific or may potentially behave differently on certain targets.
1358 // 2. Asm parser extensions. For example, platform-specific parsers
1359 // (like the ELF parser) register themselves as extensions.
1360 // 3. The generic directive parser implemented by this class. These are
1361 // all the directives that behave in a target and platform independent
1362 // manner, or at least have a default behavior that's shared between
1363 // all targets and platforms.
1365 // First query the target-specific parser. It will return 'true' if it
1366 // isn't interested in this directive.
1367 if (!getTargetParser().ParseDirective(ID))
1370 // Next, check the extension directive map to see if any extension has
1371 // registered itself to parse this directive.
1372 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
1373 ExtensionDirectiveMap.lookup(IDVal);
1375 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1377 // Finally, if no one else is interested in this directive, it must be
1378 // generic and familiar to this class.
1384 return parseDirectiveSet(IDVal, true);
1386 return parseDirectiveSet(IDVal, false);
1388 return parseDirectiveAscii(IDVal, false);
1391 return parseDirectiveAscii(IDVal, true);
1393 return parseDirectiveValue(1);
1397 return parseDirectiveValue(2);
1401 return parseDirectiveValue(4);
1404 return parseDirectiveValue(8);
1406 return parseDirectiveOctaValue();
1409 return parseDirectiveRealValue(APFloat::IEEEsingle);
1411 return parseDirectiveRealValue(APFloat::IEEEdouble);
1413 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1414 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1417 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1418 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1421 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1423 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1425 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1427 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1429 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1431 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1433 return parseDirectiveOrg();
1435 return parseDirectiveFill();
1437 return parseDirectiveZero();
1439 eatToEndOfStatement(); // .extern is the default, ignore it.
1443 return parseDirectiveSymbolAttribute(MCSA_Global);
1444 case DK_LAZY_REFERENCE:
1445 return parseDirectiveSymbolAttribute(MCSA_LazyReference);
1446 case DK_NO_DEAD_STRIP:
1447 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1448 case DK_SYMBOL_RESOLVER:
1449 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1450 case DK_PRIVATE_EXTERN:
1451 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1453 return parseDirectiveSymbolAttribute(MCSA_Reference);
1454 case DK_WEAK_DEFINITION:
1455 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1456 case DK_WEAK_REFERENCE:
1457 return parseDirectiveSymbolAttribute(MCSA_WeakReference);
1458 case DK_WEAK_DEF_CAN_BE_HIDDEN:
1459 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1462 return parseDirectiveComm(/*IsLocal=*/false);
1464 return parseDirectiveComm(/*IsLocal=*/true);
1466 return parseDirectiveAbort();
1468 return parseDirectiveInclude();
1470 return parseDirectiveIncbin();
1473 return TokError(Twine(IDVal) + " not supported yet");
1475 return parseDirectiveRept(IDLoc, IDVal);
1477 return parseDirectiveIrp(IDLoc);
1479 return parseDirectiveIrpc(IDLoc);
1481 return parseDirectiveEndr(IDLoc);
1482 case DK_BUNDLE_ALIGN_MODE:
1483 return parseDirectiveBundleAlignMode();
1484 case DK_BUNDLE_LOCK:
1485 return parseDirectiveBundleLock();
1486 case DK_BUNDLE_UNLOCK:
1487 return parseDirectiveBundleUnlock();
1489 return parseDirectiveLEB128(true);
1491 return parseDirectiveLEB128(false);
1494 return parseDirectiveSpace(IDVal);
1496 return parseDirectiveFile(IDLoc);
1498 return parseDirectiveLine();
1500 return parseDirectiveLoc();
1502 return parseDirectiveStabs();
1503 case DK_CFI_SECTIONS:
1504 return parseDirectiveCFISections();
1505 case DK_CFI_STARTPROC:
1506 return parseDirectiveCFIStartProc();
1507 case DK_CFI_ENDPROC:
1508 return parseDirectiveCFIEndProc();
1509 case DK_CFI_DEF_CFA:
1510 return parseDirectiveCFIDefCfa(IDLoc);
1511 case DK_CFI_DEF_CFA_OFFSET:
1512 return parseDirectiveCFIDefCfaOffset();
1513 case DK_CFI_ADJUST_CFA_OFFSET:
1514 return parseDirectiveCFIAdjustCfaOffset();
1515 case DK_CFI_DEF_CFA_REGISTER:
1516 return parseDirectiveCFIDefCfaRegister(IDLoc);
1518 return parseDirectiveCFIOffset(IDLoc);
1519 case DK_CFI_REL_OFFSET:
1520 return parseDirectiveCFIRelOffset(IDLoc);
1521 case DK_CFI_PERSONALITY:
1522 return parseDirectiveCFIPersonalityOrLsda(true);
1524 return parseDirectiveCFIPersonalityOrLsda(false);
1525 case DK_CFI_REMEMBER_STATE:
1526 return parseDirectiveCFIRememberState();
1527 case DK_CFI_RESTORE_STATE:
1528 return parseDirectiveCFIRestoreState();
1529 case DK_CFI_SAME_VALUE:
1530 return parseDirectiveCFISameValue(IDLoc);
1531 case DK_CFI_RESTORE:
1532 return parseDirectiveCFIRestore(IDLoc);
1534 return parseDirectiveCFIEscape();
1535 case DK_CFI_SIGNAL_FRAME:
1536 return parseDirectiveCFISignalFrame();
1537 case DK_CFI_UNDEFINED:
1538 return parseDirectiveCFIUndefined(IDLoc);
1539 case DK_CFI_REGISTER:
1540 return parseDirectiveCFIRegister(IDLoc);
1541 case DK_CFI_WINDOW_SAVE:
1542 return parseDirectiveCFIWindowSave();
1545 return parseDirectiveMacrosOnOff(IDVal);
1547 return parseDirectiveMacro(IDLoc);
1550 return parseDirectiveEndMacro(IDVal);
1552 return parseDirectivePurgeMacro(IDLoc);
1554 return parseDirectiveEnd(IDLoc);
1556 return parseDirectiveError(IDLoc, false);
1558 return parseDirectiveError(IDLoc, true);
1561 return Error(IDLoc, "unknown directive");
1564 // __asm _emit or __asm __emit
1565 if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
1566 IDVal == "_EMIT" || IDVal == "__EMIT"))
1567 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
1570 if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
1571 return parseDirectiveMSAlign(IDLoc, Info);
1573 checkForValidSection();
1575 // Canonicalize the opcode to lower case.
1576 std::string OpcodeStr = IDVal.lower();
1577 ParseInstructionInfo IInfo(Info.AsmRewrites);
1578 bool HadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, IDLoc,
1579 Info.ParsedOperands);
1580 Info.ParseError = HadError;
1582 // Dump the parsed representation, if requested.
1583 if (getShowParsedOperands()) {
1584 SmallString<256> Str;
1585 raw_svector_ostream OS(Str);
1586 OS << "parsed instruction: [";
1587 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
1590 Info.ParsedOperands[i]->print(OS);
1594 printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1597 // If we are generating dwarf for the current section then generate a .loc
1598 // directive for the instruction.
1599 if (!HadError && getContext().getGenDwarfForAssembly() &&
1600 getContext().getGenDwarfSectionSyms().count(
1601 getStreamer().getCurrentSection().first)) {
1603 unsigned Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
1605 // If we previously parsed a cpp hash file line comment then make sure the
1606 // current Dwarf File is for the CppHashFilename if not then emit the
1607 // Dwarf File table for it and adjust the line number for the .loc.
1608 if (CppHashFilename.size() != 0) {
1609 unsigned FileNumber = getStreamer().EmitDwarfFileDirective(
1610 0, StringRef(), CppHashFilename);
1611 getContext().setGenDwarfFileNumber(FileNumber);
1613 // Since SrcMgr.FindLineNumber() is slow and messes up the SourceMgr's
1614 // cache with the different Loc from the call above we save the last
1615 // info we queried here with SrcMgr.FindLineNumber().
1616 unsigned CppHashLocLineNo;
1617 if (LastQueryIDLoc == CppHashLoc && LastQueryBuffer == CppHashBuf)
1618 CppHashLocLineNo = LastQueryLine;
1620 CppHashLocLineNo = SrcMgr.FindLineNumber(CppHashLoc, CppHashBuf);
1621 LastQueryLine = CppHashLocLineNo;
1622 LastQueryIDLoc = CppHashLoc;
1623 LastQueryBuffer = CppHashBuf;
1625 Line = CppHashLineNumber - 1 + (Line - CppHashLocLineNo);
1628 getStreamer().EmitDwarfLocDirective(
1629 getContext().getGenDwarfFileNumber(), Line, 0,
1630 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0,
1634 // If parsing succeeded, match the instruction.
1637 getTargetParser().MatchAndEmitInstruction(IDLoc, Info.Opcode,
1638 Info.ParsedOperands, Out,
1639 ErrorInfo, ParsingInlineAsm);
1642 // Don't skip the rest of the line, the instruction parser is responsible for
1647 /// eatToEndOfLine uses the Lexer to eat the characters to the end of the line
1648 /// since they may not be able to be tokenized to get to the end of line token.
1649 void AsmParser::eatToEndOfLine() {
1650 if (!Lexer.is(AsmToken::EndOfStatement))
1651 Lexer.LexUntilEndOfLine();
1656 /// parseCppHashLineFilenameComment as this:
1657 /// ::= # number "filename"
1658 /// or just as a full line comment if it doesn't have a number and a string.
1659 bool AsmParser::parseCppHashLineFilenameComment(const SMLoc &L) {
1660 Lex(); // Eat the hash token.
1662 if (getLexer().isNot(AsmToken::Integer)) {
1663 // Consume the line since in cases it is not a well-formed line directive,
1664 // as if were simply a full line comment.
1669 int64_t LineNumber = getTok().getIntVal();
1672 if (getLexer().isNot(AsmToken::String)) {
1677 StringRef Filename = getTok().getString();
1678 // Get rid of the enclosing quotes.
1679 Filename = Filename.substr(1, Filename.size() - 2);
1681 // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1683 CppHashFilename = Filename;
1684 CppHashLineNumber = LineNumber;
1685 CppHashBuf = CurBuffer;
1687 // Ignore any trailing characters, they're just comment.
1692 /// \brief will use the last parsed cpp hash line filename comment
1693 /// for the Filename and LineNo if any in the diagnostic.
1694 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1695 const AsmParser *Parser = static_cast<const AsmParser *>(Context);
1696 raw_ostream &OS = errs();
1698 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
1699 const SMLoc &DiagLoc = Diag.getLoc();
1700 int DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1701 int CppHashBuf = Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc);
1703 // Like SourceMgr::printMessage() we need to print the include stack if any
1704 // before printing the message.
1705 int DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1706 if (!Parser->SavedDiagHandler && DiagCurBuffer > 0) {
1707 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
1708 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
1711 // If we have not parsed a cpp hash line filename comment or the source
1712 // manager changed or buffer changed (like in a nested include) then just
1713 // print the normal diagnostic using its Filename and LineNo.
1714 if (!Parser->CppHashLineNumber || &DiagSrcMgr != &Parser->SrcMgr ||
1715 DiagBuf != CppHashBuf) {
1716 if (Parser->SavedDiagHandler)
1717 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
1719 Diag.print(nullptr, OS);
1723 // Use the CppHashFilename and calculate a line number based on the
1724 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for
1726 const std::string &Filename = Parser->CppHashFilename;
1728 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
1729 int CppHashLocLineNo =
1730 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf);
1732 Parser->CppHashLineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
1734 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
1735 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
1736 Diag.getLineContents(), Diag.getRanges());
1738 if (Parser->SavedDiagHandler)
1739 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
1741 NewDiag.print(nullptr, OS);
1744 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
1745 // difference being that that function accepts '@' as part of identifiers and
1746 // we can't do that. AsmLexer.cpp should probably be changed to handle
1747 // '@' as a special case when needed.
1748 static bool isIdentifierChar(char c) {
1749 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' ||
1753 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
1754 ArrayRef<MCAsmMacroParameter> Parameters,
1755 ArrayRef<MCAsmMacroArgument> A, const SMLoc &L) {
1756 unsigned NParameters = Parameters.size();
1757 bool HasVararg = NParameters ? Parameters.back().Vararg : false;
1758 if ((!IsDarwin || NParameters != 0) && NParameters != A.size())
1759 return Error(L, "Wrong number of arguments");
1761 // A macro without parameters is handled differently on Darwin:
1762 // gas accepts no arguments and does no substitutions
1763 while (!Body.empty()) {
1764 // Scan for the next substitution.
1765 std::size_t End = Body.size(), Pos = 0;
1766 for (; Pos != End; ++Pos) {
1767 // Check for a substitution or escape.
1768 if (IsDarwin && !NParameters) {
1769 // This macro has no parameters, look for $0, $1, etc.
1770 if (Body[Pos] != '$' || Pos + 1 == End)
1773 char Next = Body[Pos + 1];
1774 if (Next == '$' || Next == 'n' ||
1775 isdigit(static_cast<unsigned char>(Next)))
1778 // This macro has parameters, look for \foo, \bar, etc.
1779 if (Body[Pos] == '\\' && Pos + 1 != End)
1785 OS << Body.slice(0, Pos);
1787 // Check if we reached the end.
1791 if (IsDarwin && !NParameters) {
1792 switch (Body[Pos + 1]) {
1798 // $n => number of arguments
1803 // $[0-9] => argument
1805 // Missing arguments are ignored.
1806 unsigned Index = Body[Pos + 1] - '0';
1807 if (Index >= A.size())
1810 // Otherwise substitute with the token values, with spaces eliminated.
1811 for (MCAsmMacroArgument::const_iterator it = A[Index].begin(),
1812 ie = A[Index].end();
1814 OS << it->getString();
1820 unsigned I = Pos + 1;
1821 while (isIdentifierChar(Body[I]) && I + 1 != End)
1824 const char *Begin = Body.data() + Pos + 1;
1825 StringRef Argument(Begin, I - (Pos + 1));
1827 for (; Index < NParameters; ++Index)
1828 if (Parameters[Index].Name == Argument)
1831 if (Index == NParameters) {
1832 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
1835 OS << '\\' << Argument;
1839 bool VarargParameter = HasVararg && Index == (NParameters - 1);
1840 for (MCAsmMacroArgument::const_iterator it = A[Index].begin(),
1841 ie = A[Index].end();
1843 // We expect no quotes around the string's contents when
1844 // parsing for varargs.
1845 if (it->getKind() != AsmToken::String || VarargParameter)
1846 OS << it->getString();
1848 OS << it->getStringContents();
1850 Pos += 1 + Argument.size();
1853 // Update the scan point.
1854 Body = Body.substr(Pos);
1860 MacroInstantiation::MacroInstantiation(const MCAsmMacro *M, SMLoc IL, int EB,
1861 SMLoc EL, MemoryBuffer *I)
1862 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitBuffer(EB),
1865 static bool isOperator(AsmToken::TokenKind kind) {
1869 case AsmToken::Plus:
1870 case AsmToken::Minus:
1871 case AsmToken::Tilde:
1872 case AsmToken::Slash:
1873 case AsmToken::Star:
1875 case AsmToken::Equal:
1876 case AsmToken::EqualEqual:
1877 case AsmToken::Pipe:
1878 case AsmToken::PipePipe:
1879 case AsmToken::Caret:
1881 case AsmToken::AmpAmp:
1882 case AsmToken::Exclaim:
1883 case AsmToken::ExclaimEqual:
1884 case AsmToken::Percent:
1885 case AsmToken::Less:
1886 case AsmToken::LessEqual:
1887 case AsmToken::LessLess:
1888 case AsmToken::LessGreater:
1889 case AsmToken::Greater:
1890 case AsmToken::GreaterEqual:
1891 case AsmToken::GreaterGreater:
1897 class AsmLexerSkipSpaceRAII {
1899 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
1900 Lexer.setSkipSpace(SkipSpace);
1903 ~AsmLexerSkipSpaceRAII() {
1904 Lexer.setSkipSpace(true);
1912 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) {
1915 if (Lexer.isNot(AsmToken::EndOfStatement)) {
1916 StringRef Str = parseStringToEndOfStatement();
1917 MA.push_back(AsmToken(AsmToken::String, Str));
1922 unsigned ParenLevel = 0;
1923 unsigned AddTokens = 0;
1925 // Darwin doesn't use spaces to delmit arguments.
1926 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
1929 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
1930 return TokError("unexpected token in macro instantiation");
1932 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma))
1935 if (Lexer.is(AsmToken::Space)) {
1936 Lex(); // Eat spaces
1938 // Spaces can delimit parameters, but could also be part an expression.
1939 // If the token after a space is an operator, add the token and the next
1940 // one into this argument
1942 if (isOperator(Lexer.getKind())) {
1943 // Check to see whether the token is used as an operator,
1944 // or part of an identifier
1945 const char *NextChar = getTok().getEndLoc().getPointer();
1946 if (*NextChar == ' ')
1950 if (!AddTokens && ParenLevel == 0) {
1956 // handleMacroEntry relies on not advancing the lexer here
1957 // to be able to fill in the remaining default parameter values
1958 if (Lexer.is(AsmToken::EndOfStatement))
1961 // Adjust the current parentheses level.
1962 if (Lexer.is(AsmToken::LParen))
1964 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1967 // Append the token to the current argument list.
1968 MA.push_back(getTok());
1974 if (ParenLevel != 0)
1975 return TokError("unbalanced parentheses in macro argument");
1979 // Parse the macro instantiation arguments.
1980 bool AsmParser::parseMacroArguments(const MCAsmMacro *M,
1981 MCAsmMacroArguments &A) {
1982 const unsigned NParameters = M ? M->Parameters.size() : 0;
1983 bool NamedParametersFound = false;
1984 SmallVector<SMLoc, 4> FALocs;
1986 A.resize(NParameters);
1987 FALocs.resize(NParameters);
1989 // Parse two kinds of macro invocations:
1990 // - macros defined without any parameters accept an arbitrary number of them
1991 // - macros defined with parameters accept at most that many of them
1992 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false;
1993 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
1995 SMLoc IDLoc = Lexer.getLoc();
1996 MCAsmMacroParameter FA;
1998 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) {
1999 if (parseIdentifier(FA.Name)) {
2000 Error(IDLoc, "invalid argument identifier for formal argument");
2001 eatToEndOfStatement();
2005 if (!Lexer.is(AsmToken::Equal)) {
2006 TokError("expected '=' after formal parameter identifier");
2007 eatToEndOfStatement();
2012 NamedParametersFound = true;
2015 if (NamedParametersFound && FA.Name.empty()) {
2016 Error(IDLoc, "cannot mix positional and keyword arguments");
2017 eatToEndOfStatement();
2021 bool Vararg = HasVararg && Parameter == (NParameters - 1);
2022 if (parseMacroArgument(FA.Value, Vararg))
2025 unsigned PI = Parameter;
2026 if (!FA.Name.empty()) {
2028 for (FAI = 0; FAI < NParameters; ++FAI)
2029 if (M->Parameters[FAI].Name == FA.Name)
2032 if (FAI >= NParameters) {
2033 assert(M && "expected macro to be defined");
2035 "parameter named '" + FA.Name + "' does not exist for macro '" +
2042 if (!FA.Value.empty()) {
2047 if (FALocs.size() <= PI)
2048 FALocs.resize(PI + 1);
2050 FALocs[PI] = Lexer.getLoc();
2053 // At the end of the statement, fill in remaining arguments that have
2054 // default values. If there aren't any, then the next argument is
2055 // required but missing
2056 if (Lexer.is(AsmToken::EndOfStatement)) {
2057 bool Failure = false;
2058 for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
2059 if (A[FAI].empty()) {
2060 if (M->Parameters[FAI].Required) {
2061 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
2062 "missing value for required parameter "
2063 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
2067 if (!M->Parameters[FAI].Value.empty())
2068 A[FAI] = M->Parameters[FAI].Value;
2074 if (Lexer.is(AsmToken::Comma))
2078 return TokError("too many positional arguments");
2081 const MCAsmMacro *AsmParser::lookupMacro(StringRef Name) {
2082 StringMap<MCAsmMacro *>::iterator I = MacroMap.find(Name);
2083 return (I == MacroMap.end()) ? nullptr : I->getValue();
2086 void AsmParser::defineMacro(StringRef Name, const MCAsmMacro &Macro) {
2087 MacroMap[Name] = new MCAsmMacro(Macro);
2090 void AsmParser::undefineMacro(StringRef Name) {
2091 StringMap<MCAsmMacro *>::iterator I = MacroMap.find(Name);
2092 if (I != MacroMap.end()) {
2093 delete I->getValue();
2098 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) {
2099 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
2100 // this, although we should protect against infinite loops.
2101 if (ActiveMacros.size() == 20)
2102 return TokError("macros cannot be nested more than 20 levels deep");
2104 MCAsmMacroArguments A;
2105 if (parseMacroArguments(M, A))
2108 // Macro instantiation is lexical, unfortunately. We construct a new buffer
2109 // to hold the macro body with substitutions.
2110 SmallString<256> Buf;
2111 StringRef Body = M->Body;
2112 raw_svector_ostream OS(Buf);
2114 if (expandMacro(OS, Body, M->Parameters, A, getTok().getLoc()))
2117 // We include the .endmacro in the buffer as our cue to exit the macro
2119 OS << ".endmacro\n";
2121 MemoryBuffer *Instantiation =
2122 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
2124 // Create the macro instantiation object and add to the current macro
2125 // instantiation stack.
2126 MacroInstantiation *MI = new MacroInstantiation(
2127 M, NameLoc, CurBuffer, getTok().getLoc(), Instantiation);
2128 ActiveMacros.push_back(MI);
2130 // Jump to the macro instantiation and prime the lexer.
2131 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
2132 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
2138 void AsmParser::handleMacroExit() {
2139 // Jump to the EndOfStatement we should return to, and consume it.
2140 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer);
2143 // Pop the instantiation entry.
2144 delete ActiveMacros.back();
2145 ActiveMacros.pop_back();
2148 static bool isUsedIn(const MCSymbol *Sym, const MCExpr *Value) {
2149 switch (Value->getKind()) {
2150 case MCExpr::Binary: {
2151 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value);
2152 return isUsedIn(Sym, BE->getLHS()) || isUsedIn(Sym, BE->getRHS());
2154 case MCExpr::Target:
2155 case MCExpr::Constant:
2157 case MCExpr::SymbolRef: {
2159 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol();
2161 return isUsedIn(Sym, S.getVariableValue());
2165 return isUsedIn(Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr());
2168 llvm_unreachable("Unknown expr kind!");
2171 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef,
2173 // FIXME: Use better location, we should use proper tokens.
2174 SMLoc EqualLoc = Lexer.getLoc();
2176 const MCExpr *Value;
2177 if (parseExpression(Value))
2180 // Note: we don't count b as used in "a = b". This is to allow
2184 if (Lexer.isNot(AsmToken::EndOfStatement))
2185 return TokError("unexpected token in assignment");
2187 // Eat the end of statement marker.
2190 // Validate that the LHS is allowed to be a variable (either it has not been
2191 // used as a symbol, or it is an absolute symbol).
2192 MCSymbol *Sym = getContext().LookupSymbol(Name);
2194 // Diagnose assignment to a label.
2196 // FIXME: Diagnostics. Note the location of the definition as a label.
2197 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
2198 if (isUsedIn(Sym, Value))
2199 return Error(EqualLoc, "Recursive use of '" + Name + "'");
2200 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
2201 ; // Allow redefinitions of undefined symbols only used in directives.
2202 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
2203 ; // Allow redefinitions of variables that haven't yet been used.
2204 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
2205 return Error(EqualLoc, "redefinition of '" + Name + "'");
2206 else if (!Sym->isVariable())
2207 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
2208 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
2209 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
2212 // Don't count these checks as uses.
2213 Sym->setUsed(false);
2214 } else if (Name == ".") {
2215 if (Out.EmitValueToOffset(Value, 0)) {
2216 Error(EqualLoc, "expected absolute expression");
2217 eatToEndOfStatement();
2221 Sym = getContext().GetOrCreateSymbol(Name);
2223 // Do the assignment.
2224 Out.EmitAssignment(Sym, Value);
2226 Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip);
2231 /// parseIdentifier:
2234 bool AsmParser::parseIdentifier(StringRef &Res) {
2235 // The assembler has relaxed rules for accepting identifiers, in particular we
2236 // allow things like '.globl $foo' and '.def @feat.00', which would normally be
2237 // separate tokens. At this level, we have already lexed so we cannot (currently)
2238 // handle this as a context dependent token, instead we detect adjacent tokens
2239 // and return the combined identifier.
2240 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
2241 SMLoc PrefixLoc = getLexer().getLoc();
2243 // Consume the prefix character, and check for a following identifier.
2245 if (Lexer.isNot(AsmToken::Identifier))
2248 // We have a '$' or '@' followed by an identifier, make sure they are adjacent.
2249 if (PrefixLoc.getPointer() + 1 != getTok().getLoc().getPointer())
2252 // Construct the joined identifier and consume the token.
2254 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1);
2259 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
2262 Res = getTok().getIdentifier();
2264 Lex(); // Consume the identifier token.
2269 /// parseDirectiveSet:
2270 /// ::= .equ identifier ',' expression
2271 /// ::= .equiv identifier ',' expression
2272 /// ::= .set identifier ',' expression
2273 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) {
2276 if (parseIdentifier(Name))
2277 return TokError("expected identifier after '" + Twine(IDVal) + "'");
2279 if (getLexer().isNot(AsmToken::Comma))
2280 return TokError("unexpected token in '" + Twine(IDVal) + "'");
2283 return parseAssignment(Name, allow_redef, true);
2286 bool AsmParser::parseEscapedString(std::string &Data) {
2287 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
2290 StringRef Str = getTok().getStringContents();
2291 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
2292 if (Str[i] != '\\') {
2297 // Recognize escaped characters. Note that this escape semantics currently
2298 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
2301 return TokError("unexpected backslash at end of string");
2303 // Recognize octal sequences.
2304 if ((unsigned)(Str[i] - '0') <= 7) {
2305 // Consume up to three octal characters.
2306 unsigned Value = Str[i] - '0';
2308 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2310 Value = Value * 8 + (Str[i] - '0');
2312 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2314 Value = Value * 8 + (Str[i] - '0');
2319 return TokError("invalid octal escape sequence (out of range)");
2321 Data += (unsigned char)Value;
2325 // Otherwise recognize individual escapes.
2328 // Just reject invalid escape sequences for now.
2329 return TokError("invalid escape sequence (unrecognized character)");
2331 case 'b': Data += '\b'; break;
2332 case 'f': Data += '\f'; break;
2333 case 'n': Data += '\n'; break;
2334 case 'r': Data += '\r'; break;
2335 case 't': Data += '\t'; break;
2336 case '"': Data += '"'; break;
2337 case '\\': Data += '\\'; break;
2344 /// parseDirectiveAscii:
2345 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
2346 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
2347 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2348 checkForValidSection();
2351 if (getLexer().isNot(AsmToken::String))
2352 return TokError("expected string in '" + Twine(IDVal) + "' directive");
2355 if (parseEscapedString(Data))
2358 getStreamer().EmitBytes(Data);
2360 getStreamer().EmitBytes(StringRef("\0", 1));
2364 if (getLexer().is(AsmToken::EndOfStatement))
2367 if (getLexer().isNot(AsmToken::Comma))
2368 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
2377 /// parseDirectiveValue
2378 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
2379 bool AsmParser::parseDirectiveValue(unsigned Size) {
2380 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2381 checkForValidSection();
2384 const MCExpr *Value;
2385 SMLoc ExprLoc = getLexer().getLoc();
2386 if (parseExpression(Value))
2389 // Special case constant expressions to match code generator.
2390 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2391 assert(Size <= 8 && "Invalid size");
2392 uint64_t IntValue = MCE->getValue();
2393 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
2394 return Error(ExprLoc, "literal value out of range for directive");
2395 getStreamer().EmitIntValue(IntValue, Size);
2397 getStreamer().EmitValue(Value, Size, ExprLoc);
2399 if (getLexer().is(AsmToken::EndOfStatement))
2402 // FIXME: Improve diagnostic.
2403 if (getLexer().isNot(AsmToken::Comma))
2404 return TokError("unexpected token in directive");
2413 /// ParseDirectiveOctaValue
2414 /// ::= .octa [ hexconstant (, hexconstant)* ]
2415 bool AsmParser::parseDirectiveOctaValue() {
2416 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2417 checkForValidSection();
2420 if (Lexer.getKind() == AsmToken::Error)
2422 if (Lexer.getKind() != AsmToken::Integer &&
2423 Lexer.getKind() != AsmToken::BigNum)
2424 return TokError("unknown token in expression");
2426 SMLoc ExprLoc = getLexer().getLoc();
2427 APInt IntValue = getTok().getAPIntVal();
2431 if (IntValue.isIntN(64)) {
2433 lo = IntValue.getZExtValue();
2434 } else if (IntValue.isIntN(128)) {
2435 // It might actually have more than 128 bits, but the top ones are zero.
2436 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
2437 lo = IntValue.getLoBits(64).getZExtValue();
2439 return Error(ExprLoc, "literal value out of range for directive");
2441 if (MAI.isLittleEndian()) {
2442 getStreamer().EmitIntValue(lo, 8);
2443 getStreamer().EmitIntValue(hi, 8);
2445 getStreamer().EmitIntValue(hi, 8);
2446 getStreamer().EmitIntValue(lo, 8);
2449 if (getLexer().is(AsmToken::EndOfStatement))
2452 // FIXME: Improve diagnostic.
2453 if (getLexer().isNot(AsmToken::Comma))
2454 return TokError("unexpected token in directive");
2463 /// parseDirectiveRealValue
2464 /// ::= (.single | .double) [ expression (, expression)* ]
2465 bool AsmParser::parseDirectiveRealValue(const fltSemantics &Semantics) {
2466 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2467 checkForValidSection();
2470 // We don't truly support arithmetic on floating point expressions, so we
2471 // have to manually parse unary prefixes.
2473 if (getLexer().is(AsmToken::Minus)) {
2476 } else if (getLexer().is(AsmToken::Plus))
2479 if (getLexer().isNot(AsmToken::Integer) &&
2480 getLexer().isNot(AsmToken::Real) &&
2481 getLexer().isNot(AsmToken::Identifier))
2482 return TokError("unexpected token in directive");
2484 // Convert to an APFloat.
2485 APFloat Value(Semantics);
2486 StringRef IDVal = getTok().getString();
2487 if (getLexer().is(AsmToken::Identifier)) {
2488 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
2489 Value = APFloat::getInf(Semantics);
2490 else if (!IDVal.compare_lower("nan"))
2491 Value = APFloat::getNaN(Semantics, false, ~0);
2493 return TokError("invalid floating point literal");
2494 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
2495 APFloat::opInvalidOp)
2496 return TokError("invalid floating point literal");
2500 // Consume the numeric token.
2503 // Emit the value as an integer.
2504 APInt AsInt = Value.bitcastToAPInt();
2505 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
2506 AsInt.getBitWidth() / 8);
2508 if (getLexer().is(AsmToken::EndOfStatement))
2511 if (getLexer().isNot(AsmToken::Comma))
2512 return TokError("unexpected token in directive");
2521 /// parseDirectiveZero
2522 /// ::= .zero expression
2523 bool AsmParser::parseDirectiveZero() {
2524 checkForValidSection();
2527 if (parseAbsoluteExpression(NumBytes))
2531 if (getLexer().is(AsmToken::Comma)) {
2533 if (parseAbsoluteExpression(Val))
2537 if (getLexer().isNot(AsmToken::EndOfStatement))
2538 return TokError("unexpected token in '.zero' directive");
2542 getStreamer().EmitFill(NumBytes, Val);
2547 /// parseDirectiveFill
2548 /// ::= .fill expression [ , expression [ , expression ] ]
2549 bool AsmParser::parseDirectiveFill() {
2550 checkForValidSection();
2552 SMLoc RepeatLoc = getLexer().getLoc();
2554 if (parseAbsoluteExpression(NumValues))
2557 if (NumValues < 0) {
2559 "'.fill' directive with negative repeat count has no effect");
2563 int64_t FillSize = 1;
2564 int64_t FillExpr = 0;
2566 SMLoc SizeLoc, ExprLoc;
2567 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2568 if (getLexer().isNot(AsmToken::Comma))
2569 return TokError("unexpected token in '.fill' directive");
2572 SizeLoc = getLexer().getLoc();
2573 if (parseAbsoluteExpression(FillSize))
2576 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2577 if (getLexer().isNot(AsmToken::Comma))
2578 return TokError("unexpected token in '.fill' directive");
2581 ExprLoc = getLexer().getLoc();
2582 if (parseAbsoluteExpression(FillExpr))
2585 if (getLexer().isNot(AsmToken::EndOfStatement))
2586 return TokError("unexpected token in '.fill' directive");
2593 Warning(SizeLoc, "'.fill' directive with negative size has no effect");
2597 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8");
2601 if (!isUInt<32>(FillExpr) && FillSize > 4)
2602 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits");
2604 int64_t NonZeroFillSize = FillSize > 4 ? 4 : FillSize;
2605 FillExpr &= ~0ULL >> (64 - NonZeroFillSize * 8);
2607 for (uint64_t i = 0, e = NumValues; i != e; ++i) {
2608 getStreamer().EmitIntValue(FillExpr, NonZeroFillSize);
2609 getStreamer().EmitIntValue(0, FillSize - NonZeroFillSize);
2615 /// parseDirectiveOrg
2616 /// ::= .org expression [ , expression ]
2617 bool AsmParser::parseDirectiveOrg() {
2618 checkForValidSection();
2620 const MCExpr *Offset;
2621 SMLoc Loc = getTok().getLoc();
2622 if (parseExpression(Offset))
2625 // Parse optional fill expression.
2626 int64_t FillExpr = 0;
2627 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2628 if (getLexer().isNot(AsmToken::Comma))
2629 return TokError("unexpected token in '.org' directive");
2632 if (parseAbsoluteExpression(FillExpr))
2635 if (getLexer().isNot(AsmToken::EndOfStatement))
2636 return TokError("unexpected token in '.org' directive");
2641 // Only limited forms of relocatable expressions are accepted here, it
2642 // has to be relative to the current section. The streamer will return
2643 // 'true' if the expression wasn't evaluatable.
2644 if (getStreamer().EmitValueToOffset(Offset, FillExpr))
2645 return Error(Loc, "expected assembly-time absolute expression");
2650 /// parseDirectiveAlign
2651 /// ::= {.align, ...} expression [ , expression [ , expression ]]
2652 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2653 checkForValidSection();
2655 SMLoc AlignmentLoc = getLexer().getLoc();
2657 if (parseAbsoluteExpression(Alignment))
2661 bool HasFillExpr = false;
2662 int64_t FillExpr = 0;
2663 int64_t MaxBytesToFill = 0;
2664 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2665 if (getLexer().isNot(AsmToken::Comma))
2666 return TokError("unexpected token in directive");
2669 // The fill expression can be omitted while specifying a maximum number of
2670 // alignment bytes, e.g:
2672 if (getLexer().isNot(AsmToken::Comma)) {
2674 if (parseAbsoluteExpression(FillExpr))
2678 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2679 if (getLexer().isNot(AsmToken::Comma))
2680 return TokError("unexpected token in directive");
2683 MaxBytesLoc = getLexer().getLoc();
2684 if (parseAbsoluteExpression(MaxBytesToFill))
2687 if (getLexer().isNot(AsmToken::EndOfStatement))
2688 return TokError("unexpected token in directive");
2697 // Compute alignment in bytes.
2699 // FIXME: Diagnose overflow.
2700 if (Alignment >= 32) {
2701 Error(AlignmentLoc, "invalid alignment value");
2705 Alignment = 1ULL << Alignment;
2707 // Reject alignments that aren't a power of two, for gas compatibility.
2708 if (!isPowerOf2_64(Alignment))
2709 Error(AlignmentLoc, "alignment must be a power of 2");
2712 // Diagnose non-sensical max bytes to align.
2713 if (MaxBytesLoc.isValid()) {
2714 if (MaxBytesToFill < 1) {
2715 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2716 "many bytes, ignoring maximum bytes expression");
2720 if (MaxBytesToFill >= Alignment) {
2721 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2727 // Check whether we should use optimal code alignment for this .align
2729 const MCSection *Section = getStreamer().getCurrentSection().first;
2730 assert(Section && "must have section to emit alignment");
2731 bool UseCodeAlign = Section->UseCodeAlign();
2732 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2733 ValueSize == 1 && UseCodeAlign) {
2734 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2736 // FIXME: Target specific behavior about how the "extra" bytes are filled.
2737 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2744 /// parseDirectiveFile
2745 /// ::= .file [number] filename
2746 /// ::= .file number directory filename
2747 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
2748 // FIXME: I'm not sure what this is.
2749 int64_t FileNumber = -1;
2750 SMLoc FileNumberLoc = getLexer().getLoc();
2751 if (getLexer().is(AsmToken::Integer)) {
2752 FileNumber = getTok().getIntVal();
2756 return TokError("file number less than one");
2759 if (getLexer().isNot(AsmToken::String))
2760 return TokError("unexpected token in '.file' directive");
2762 // Usually the directory and filename together, otherwise just the directory.
2763 // Allow the strings to have escaped octal character sequence.
2764 std::string Path = getTok().getString();
2765 if (parseEscapedString(Path))
2769 StringRef Directory;
2771 std::string FilenameData;
2772 if (getLexer().is(AsmToken::String)) {
2773 if (FileNumber == -1)
2774 return TokError("explicit path specified, but no file number");
2775 if (parseEscapedString(FilenameData))
2777 Filename = FilenameData;
2784 if (getLexer().isNot(AsmToken::EndOfStatement))
2785 return TokError("unexpected token in '.file' directive");
2787 if (FileNumber == -1)
2788 getStreamer().EmitFileDirective(Filename);
2790 if (getContext().getGenDwarfForAssembly() == true)
2792 "input can't have .file dwarf directives when -g is "
2793 "used to generate dwarf debug info for assembly code");
2795 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename) ==
2797 Error(FileNumberLoc, "file number already allocated");
2803 /// parseDirectiveLine
2804 /// ::= .line [number]
2805 bool AsmParser::parseDirectiveLine() {
2806 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2807 if (getLexer().isNot(AsmToken::Integer))
2808 return TokError("unexpected token in '.line' directive");
2810 int64_t LineNumber = getTok().getIntVal();
2814 // FIXME: Do something with the .line.
2817 if (getLexer().isNot(AsmToken::EndOfStatement))
2818 return TokError("unexpected token in '.line' directive");
2823 /// parseDirectiveLoc
2824 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2825 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2826 /// The first number is a file number, must have been previously assigned with
2827 /// a .file directive, the second number is the line number and optionally the
2828 /// third number is a column position (zero if not specified). The remaining
2829 /// optional items are .loc sub-directives.
2830 bool AsmParser::parseDirectiveLoc() {
2831 if (getLexer().isNot(AsmToken::Integer))
2832 return TokError("unexpected token in '.loc' directive");
2833 int64_t FileNumber = getTok().getIntVal();
2835 return TokError("file number less than one in '.loc' directive");
2836 if (!getContext().isValidDwarfFileNumber(FileNumber))
2837 return TokError("unassigned file number in '.loc' directive");
2840 int64_t LineNumber = 0;
2841 if (getLexer().is(AsmToken::Integer)) {
2842 LineNumber = getTok().getIntVal();
2844 return TokError("line number less than zero in '.loc' directive");
2848 int64_t ColumnPos = 0;
2849 if (getLexer().is(AsmToken::Integer)) {
2850 ColumnPos = getTok().getIntVal();
2852 return TokError("column position less than zero in '.loc' directive");
2856 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2858 int64_t Discriminator = 0;
2859 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2861 if (getLexer().is(AsmToken::EndOfStatement))
2865 SMLoc Loc = getTok().getLoc();
2866 if (parseIdentifier(Name))
2867 return TokError("unexpected token in '.loc' directive");
2869 if (Name == "basic_block")
2870 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2871 else if (Name == "prologue_end")
2872 Flags |= DWARF2_FLAG_PROLOGUE_END;
2873 else if (Name == "epilogue_begin")
2874 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2875 else if (Name == "is_stmt") {
2876 Loc = getTok().getLoc();
2877 const MCExpr *Value;
2878 if (parseExpression(Value))
2880 // The expression must be the constant 0 or 1.
2881 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2882 int Value = MCE->getValue();
2884 Flags &= ~DWARF2_FLAG_IS_STMT;
2885 else if (Value == 1)
2886 Flags |= DWARF2_FLAG_IS_STMT;
2888 return Error(Loc, "is_stmt value not 0 or 1");
2890 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2892 } else if (Name == "isa") {
2893 Loc = getTok().getLoc();
2894 const MCExpr *Value;
2895 if (parseExpression(Value))
2897 // The expression must be a constant greater or equal to 0.
2898 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2899 int Value = MCE->getValue();
2901 return Error(Loc, "isa number less than zero");
2904 return Error(Loc, "isa number not a constant value");
2906 } else if (Name == "discriminator") {
2907 if (parseAbsoluteExpression(Discriminator))
2910 return Error(Loc, "unknown sub-directive in '.loc' directive");
2913 if (getLexer().is(AsmToken::EndOfStatement))
2918 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2919 Isa, Discriminator, StringRef());
2924 /// parseDirectiveStabs
2925 /// ::= .stabs string, number, number, number
2926 bool AsmParser::parseDirectiveStabs() {
2927 return TokError("unsupported directive '.stabs'");
2930 /// parseDirectiveCFISections
2931 /// ::= .cfi_sections section [, section]
2932 bool AsmParser::parseDirectiveCFISections() {
2937 if (parseIdentifier(Name))
2938 return TokError("Expected an identifier");
2940 if (Name == ".eh_frame")
2942 else if (Name == ".debug_frame")
2945 if (getLexer().is(AsmToken::Comma)) {
2948 if (parseIdentifier(Name))
2949 return TokError("Expected an identifier");
2951 if (Name == ".eh_frame")
2953 else if (Name == ".debug_frame")
2957 getStreamer().EmitCFISections(EH, Debug);
2961 /// parseDirectiveCFIStartProc
2962 /// ::= .cfi_startproc [simple]
2963 bool AsmParser::parseDirectiveCFIStartProc() {
2965 if (getLexer().isNot(AsmToken::EndOfStatement))
2966 if (parseIdentifier(Simple) || Simple != "simple")
2967 return TokError("unexpected token in .cfi_startproc directive");
2969 getStreamer().EmitCFIStartProc(!Simple.empty());
2973 /// parseDirectiveCFIEndProc
2974 /// ::= .cfi_endproc
2975 bool AsmParser::parseDirectiveCFIEndProc() {
2976 getStreamer().EmitCFIEndProc();
2980 /// \brief parse register name or number.
2981 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register,
2982 SMLoc DirectiveLoc) {
2985 if (getLexer().isNot(AsmToken::Integer)) {
2986 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc))
2988 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true);
2990 return parseAbsoluteExpression(Register);
2995 /// parseDirectiveCFIDefCfa
2996 /// ::= .cfi_def_cfa register, offset
2997 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) {
2998 int64_t Register = 0;
2999 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3002 if (getLexer().isNot(AsmToken::Comma))
3003 return TokError("unexpected token in directive");
3007 if (parseAbsoluteExpression(Offset))
3010 getStreamer().EmitCFIDefCfa(Register, Offset);
3014 /// parseDirectiveCFIDefCfaOffset
3015 /// ::= .cfi_def_cfa_offset offset
3016 bool AsmParser::parseDirectiveCFIDefCfaOffset() {
3018 if (parseAbsoluteExpression(Offset))
3021 getStreamer().EmitCFIDefCfaOffset(Offset);
3025 /// parseDirectiveCFIRegister
3026 /// ::= .cfi_register register, register
3027 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) {
3028 int64_t Register1 = 0;
3029 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc))
3032 if (getLexer().isNot(AsmToken::Comma))
3033 return TokError("unexpected token in directive");
3036 int64_t Register2 = 0;
3037 if (parseRegisterOrRegisterNumber(Register2, DirectiveLoc))
3040 getStreamer().EmitCFIRegister(Register1, Register2);
3044 /// parseDirectiveCFIWindowSave
3045 /// ::= .cfi_window_save
3046 bool AsmParser::parseDirectiveCFIWindowSave() {
3047 getStreamer().EmitCFIWindowSave();
3051 /// parseDirectiveCFIAdjustCfaOffset
3052 /// ::= .cfi_adjust_cfa_offset adjustment
3053 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() {
3054 int64_t Adjustment = 0;
3055 if (parseAbsoluteExpression(Adjustment))
3058 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
3062 /// parseDirectiveCFIDefCfaRegister
3063 /// ::= .cfi_def_cfa_register register
3064 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) {
3065 int64_t Register = 0;
3066 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3069 getStreamer().EmitCFIDefCfaRegister(Register);
3073 /// parseDirectiveCFIOffset
3074 /// ::= .cfi_offset register, offset
3075 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) {
3076 int64_t Register = 0;
3079 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3082 if (getLexer().isNot(AsmToken::Comma))
3083 return TokError("unexpected token in directive");
3086 if (parseAbsoluteExpression(Offset))
3089 getStreamer().EmitCFIOffset(Register, Offset);
3093 /// parseDirectiveCFIRelOffset
3094 /// ::= .cfi_rel_offset register, offset
3095 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) {
3096 int64_t Register = 0;
3098 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3101 if (getLexer().isNot(AsmToken::Comma))
3102 return TokError("unexpected token in directive");
3106 if (parseAbsoluteExpression(Offset))
3109 getStreamer().EmitCFIRelOffset(Register, Offset);
3113 static bool isValidEncoding(int64_t Encoding) {
3114 if (Encoding & ~0xff)
3117 if (Encoding == dwarf::DW_EH_PE_omit)
3120 const unsigned Format = Encoding & 0xf;
3121 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
3122 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
3123 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
3124 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
3127 const unsigned Application = Encoding & 0x70;
3128 if (Application != dwarf::DW_EH_PE_absptr &&
3129 Application != dwarf::DW_EH_PE_pcrel)
3135 /// parseDirectiveCFIPersonalityOrLsda
3136 /// IsPersonality true for cfi_personality, false for cfi_lsda
3137 /// ::= .cfi_personality encoding, [symbol_name]
3138 /// ::= .cfi_lsda encoding, [symbol_name]
3139 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) {
3140 int64_t Encoding = 0;
3141 if (parseAbsoluteExpression(Encoding))
3143 if (Encoding == dwarf::DW_EH_PE_omit)
3146 if (!isValidEncoding(Encoding))
3147 return TokError("unsupported encoding.");
3149 if (getLexer().isNot(AsmToken::Comma))
3150 return TokError("unexpected token in directive");
3154 if (parseIdentifier(Name))
3155 return TokError("expected identifier in directive");
3157 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3160 getStreamer().EmitCFIPersonality(Sym, Encoding);
3162 getStreamer().EmitCFILsda(Sym, Encoding);
3166 /// parseDirectiveCFIRememberState
3167 /// ::= .cfi_remember_state
3168 bool AsmParser::parseDirectiveCFIRememberState() {
3169 getStreamer().EmitCFIRememberState();
3173 /// parseDirectiveCFIRestoreState
3174 /// ::= .cfi_remember_state
3175 bool AsmParser::parseDirectiveCFIRestoreState() {
3176 getStreamer().EmitCFIRestoreState();
3180 /// parseDirectiveCFISameValue
3181 /// ::= .cfi_same_value register
3182 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) {
3183 int64_t Register = 0;
3185 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3188 getStreamer().EmitCFISameValue(Register);
3192 /// parseDirectiveCFIRestore
3193 /// ::= .cfi_restore register
3194 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) {
3195 int64_t Register = 0;
3196 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3199 getStreamer().EmitCFIRestore(Register);
3203 /// parseDirectiveCFIEscape
3204 /// ::= .cfi_escape expression[,...]
3205 bool AsmParser::parseDirectiveCFIEscape() {
3208 if (parseAbsoluteExpression(CurrValue))
3211 Values.push_back((uint8_t)CurrValue);
3213 while (getLexer().is(AsmToken::Comma)) {
3216 if (parseAbsoluteExpression(CurrValue))
3219 Values.push_back((uint8_t)CurrValue);
3222 getStreamer().EmitCFIEscape(Values);
3226 /// parseDirectiveCFISignalFrame
3227 /// ::= .cfi_signal_frame
3228 bool AsmParser::parseDirectiveCFISignalFrame() {
3229 if (getLexer().isNot(AsmToken::EndOfStatement))
3230 return Error(getLexer().getLoc(),
3231 "unexpected token in '.cfi_signal_frame'");
3233 getStreamer().EmitCFISignalFrame();
3237 /// parseDirectiveCFIUndefined
3238 /// ::= .cfi_undefined register
3239 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) {
3240 int64_t Register = 0;
3242 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3245 getStreamer().EmitCFIUndefined(Register);
3249 /// parseDirectiveMacrosOnOff
3252 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) {
3253 if (getLexer().isNot(AsmToken::EndOfStatement))
3254 return Error(getLexer().getLoc(),
3255 "unexpected token in '" + Directive + "' directive");
3257 setMacrosEnabled(Directive == ".macros_on");
3261 /// parseDirectiveMacro
3262 /// ::= .macro name[,] [parameters]
3263 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) {
3265 if (parseIdentifier(Name))
3266 return TokError("expected identifier in '.macro' directive");
3268 if (getLexer().is(AsmToken::Comma))
3271 MCAsmMacroParameters Parameters;
3272 while (getLexer().isNot(AsmToken::EndOfStatement)) {
3274 if (Parameters.size() && Parameters.back().Vararg)
3275 return Error(Lexer.getLoc(),
3276 "Vararg parameter '" + Parameters.back().Name +
3277 "' should be last one in the list of parameters.");
3279 MCAsmMacroParameter Parameter;
3280 if (parseIdentifier(Parameter.Name))
3281 return TokError("expected identifier in '.macro' directive");
3283 if (Lexer.is(AsmToken::Colon)) {
3284 Lex(); // consume ':'
3287 StringRef Qualifier;
3289 QualLoc = Lexer.getLoc();
3290 if (parseIdentifier(Qualifier))
3291 return Error(QualLoc, "missing parameter qualifier for "
3292 "'" + Parameter.Name + "' in macro '" + Name + "'");
3294 if (Qualifier == "req")
3295 Parameter.Required = true;
3296 else if (Qualifier == "vararg" && !IsDarwin)
3297 Parameter.Vararg = true;
3299 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier "
3300 "for '" + Parameter.Name + "' in macro '" + Name + "'");
3303 if (getLexer().is(AsmToken::Equal)) {
3308 ParamLoc = Lexer.getLoc();
3309 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false ))
3312 if (Parameter.Required)
3313 Warning(ParamLoc, "pointless default value for required parameter "
3314 "'" + Parameter.Name + "' in macro '" + Name + "'");
3317 Parameters.push_back(Parameter);
3319 if (getLexer().is(AsmToken::Comma))
3323 // Eat the end of statement.
3326 AsmToken EndToken, StartToken = getTok();
3327 unsigned MacroDepth = 0;
3329 // Lex the macro definition.
3331 // Check whether we have reached the end of the file.
3332 if (getLexer().is(AsmToken::Eof))
3333 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3335 // Otherwise, check whether we have reach the .endmacro.
3336 if (getLexer().is(AsmToken::Identifier)) {
3337 if (getTok().getIdentifier() == ".endm" ||
3338 getTok().getIdentifier() == ".endmacro") {
3339 if (MacroDepth == 0) { // Outermost macro.
3340 EndToken = getTok();
3342 if (getLexer().isNot(AsmToken::EndOfStatement))
3343 return TokError("unexpected token in '" + EndToken.getIdentifier() +
3347 // Otherwise we just found the end of an inner macro.
3350 } else if (getTok().getIdentifier() == ".macro") {
3351 // We allow nested macros. Those aren't instantiated until the outermost
3352 // macro is expanded so just ignore them for now.
3357 // Otherwise, scan til the end of the statement.
3358 eatToEndOfStatement();
3361 if (lookupMacro(Name)) {
3362 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3365 const char *BodyStart = StartToken.getLoc().getPointer();
3366 const char *BodyEnd = EndToken.getLoc().getPointer();
3367 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3368 checkForBadMacro(DirectiveLoc, Name, Body, Parameters);
3369 defineMacro(Name, MCAsmMacro(Name, Body, Parameters));
3373 /// checkForBadMacro
3375 /// With the support added for named parameters there may be code out there that
3376 /// is transitioning from positional parameters. In versions of gas that did
3377 /// not support named parameters they would be ignored on the macro definition.
3378 /// But to support both styles of parameters this is not possible so if a macro
3379 /// definition has named parameters but does not use them and has what appears
3380 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a
3381 /// warning that the positional parameter found in body which have no effect.
3382 /// Hoping the developer will either remove the named parameters from the macro
3383 /// definition so the positional parameters get used if that was what was
3384 /// intended or change the macro to use the named parameters. It is possible
3385 /// this warning will trigger when the none of the named parameters are used
3386 /// and the strings like $1 are infact to simply to be passed trough unchanged.
3387 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name,
3389 ArrayRef<MCAsmMacroParameter> Parameters) {
3390 // If this macro is not defined with named parameters the warning we are
3391 // checking for here doesn't apply.
3392 unsigned NParameters = Parameters.size();
3393 if (NParameters == 0)
3396 bool NamedParametersFound = false;
3397 bool PositionalParametersFound = false;
3399 // Look at the body of the macro for use of both the named parameters and what
3400 // are likely to be positional parameters. This is what expandMacro() is
3401 // doing when it finds the parameters in the body.
3402 while (!Body.empty()) {
3403 // Scan for the next possible parameter.
3404 std::size_t End = Body.size(), Pos = 0;
3405 for (; Pos != End; ++Pos) {
3406 // Check for a substitution or escape.
3407 // This macro is defined with parameters, look for \foo, \bar, etc.
3408 if (Body[Pos] == '\\' && Pos + 1 != End)
3411 // This macro should have parameters, but look for $0, $1, ..., $n too.
3412 if (Body[Pos] != '$' || Pos + 1 == End)
3414 char Next = Body[Pos + 1];
3415 if (Next == '$' || Next == 'n' ||
3416 isdigit(static_cast<unsigned char>(Next)))
3420 // Check if we reached the end.
3424 if (Body[Pos] == '$') {
3425 switch (Body[Pos + 1]) {
3430 // $n => number of arguments
3432 PositionalParametersFound = true;
3435 // $[0-9] => argument
3437 PositionalParametersFound = true;
3443 unsigned I = Pos + 1;
3444 while (isIdentifierChar(Body[I]) && I + 1 != End)
3447 const char *Begin = Body.data() + Pos + 1;
3448 StringRef Argument(Begin, I - (Pos + 1));
3450 for (; Index < NParameters; ++Index)
3451 if (Parameters[Index].Name == Argument)
3454 if (Index == NParameters) {
3455 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
3461 NamedParametersFound = true;
3462 Pos += 1 + Argument.size();
3465 // Update the scan point.
3466 Body = Body.substr(Pos);
3469 if (!NamedParametersFound && PositionalParametersFound)
3470 Warning(DirectiveLoc, "macro defined with named parameters which are not "
3471 "used in macro body, possible positional parameter "
3472 "found in body which will have no effect");
3475 /// parseDirectiveEndMacro
3478 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) {
3479 if (getLexer().isNot(AsmToken::EndOfStatement))
3480 return TokError("unexpected token in '" + Directive + "' directive");
3482 // If we are inside a macro instantiation, terminate the current
3484 if (isInsideMacroInstantiation()) {
3489 // Otherwise, this .endmacro is a stray entry in the file; well formed
3490 // .endmacro directives are handled during the macro definition parsing.
3491 return TokError("unexpected '" + Directive + "' in file, "
3492 "no current macro definition");
3495 /// parseDirectivePurgeMacro
3497 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) {
3499 if (parseIdentifier(Name))
3500 return TokError("expected identifier in '.purgem' directive");
3502 if (getLexer().isNot(AsmToken::EndOfStatement))
3503 return TokError("unexpected token in '.purgem' directive");
3505 if (!lookupMacro(Name))
3506 return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
3508 undefineMacro(Name);
3512 /// parseDirectiveBundleAlignMode
3513 /// ::= {.bundle_align_mode} expression
3514 bool AsmParser::parseDirectiveBundleAlignMode() {
3515 checkForValidSection();
3517 // Expect a single argument: an expression that evaluates to a constant
3518 // in the inclusive range 0-30.
3519 SMLoc ExprLoc = getLexer().getLoc();
3520 int64_t AlignSizePow2;
3521 if (parseAbsoluteExpression(AlignSizePow2))
3523 else if (getLexer().isNot(AsmToken::EndOfStatement))
3524 return TokError("unexpected token after expression in"
3525 " '.bundle_align_mode' directive");
3526 else if (AlignSizePow2 < 0 || AlignSizePow2 > 30)
3527 return Error(ExprLoc,
3528 "invalid bundle alignment size (expected between 0 and 30)");
3532 // Because of AlignSizePow2's verified range we can safely truncate it to
3534 getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2));
3538 /// parseDirectiveBundleLock
3539 /// ::= {.bundle_lock} [align_to_end]
3540 bool AsmParser::parseDirectiveBundleLock() {
3541 checkForValidSection();
3542 bool AlignToEnd = false;
3544 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3546 SMLoc Loc = getTok().getLoc();
3547 const char *kInvalidOptionError =
3548 "invalid option for '.bundle_lock' directive";
3550 if (parseIdentifier(Option))
3551 return Error(Loc, kInvalidOptionError);
3553 if (Option != "align_to_end")
3554 return Error(Loc, kInvalidOptionError);
3555 else if (getLexer().isNot(AsmToken::EndOfStatement))
3557 "unexpected token after '.bundle_lock' directive option");
3563 getStreamer().EmitBundleLock(AlignToEnd);
3567 /// parseDirectiveBundleLock
3568 /// ::= {.bundle_lock}
3569 bool AsmParser::parseDirectiveBundleUnlock() {
3570 checkForValidSection();
3572 if (getLexer().isNot(AsmToken::EndOfStatement))
3573 return TokError("unexpected token in '.bundle_unlock' directive");
3576 getStreamer().EmitBundleUnlock();
3580 /// parseDirectiveSpace
3581 /// ::= (.skip | .space) expression [ , expression ]
3582 bool AsmParser::parseDirectiveSpace(StringRef IDVal) {
3583 checkForValidSection();
3586 if (parseAbsoluteExpression(NumBytes))
3589 int64_t FillExpr = 0;
3590 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3591 if (getLexer().isNot(AsmToken::Comma))
3592 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
3595 if (parseAbsoluteExpression(FillExpr))
3598 if (getLexer().isNot(AsmToken::EndOfStatement))
3599 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
3605 return TokError("invalid number of bytes in '" + Twine(IDVal) +
3608 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
3609 getStreamer().EmitFill(NumBytes, FillExpr);
3614 /// parseDirectiveLEB128
3615 /// ::= (.sleb128 | .uleb128) expression
3616 bool AsmParser::parseDirectiveLEB128(bool Signed) {
3617 checkForValidSection();
3618 const MCExpr *Value;
3620 if (parseExpression(Value))
3623 if (getLexer().isNot(AsmToken::EndOfStatement))
3624 return TokError("unexpected token in directive");
3627 getStreamer().EmitSLEB128Value(Value);
3629 getStreamer().EmitULEB128Value(Value);
3634 /// parseDirectiveSymbolAttribute
3635 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
3636 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
3637 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3640 SMLoc Loc = getTok().getLoc();
3642 if (parseIdentifier(Name))
3643 return Error(Loc, "expected identifier in directive");
3645 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3647 // Assembler local symbols don't make any sense here. Complain loudly.
3648 if (Sym->isTemporary())
3649 return Error(Loc, "non-local symbol required in directive");
3651 if (!getStreamer().EmitSymbolAttribute(Sym, Attr))
3652 return Error(Loc, "unable to emit symbol attribute");
3654 if (getLexer().is(AsmToken::EndOfStatement))
3657 if (getLexer().isNot(AsmToken::Comma))
3658 return TokError("unexpected token in directive");
3667 /// parseDirectiveComm
3668 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
3669 bool AsmParser::parseDirectiveComm(bool IsLocal) {
3670 checkForValidSection();
3672 SMLoc IDLoc = getLexer().getLoc();
3674 if (parseIdentifier(Name))
3675 return TokError("expected identifier in directive");
3677 // Handle the identifier as the key symbol.
3678 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3680 if (getLexer().isNot(AsmToken::Comma))
3681 return TokError("unexpected token in directive");
3685 SMLoc SizeLoc = getLexer().getLoc();
3686 if (parseAbsoluteExpression(Size))
3689 int64_t Pow2Alignment = 0;
3690 SMLoc Pow2AlignmentLoc;
3691 if (getLexer().is(AsmToken::Comma)) {
3693 Pow2AlignmentLoc = getLexer().getLoc();
3694 if (parseAbsoluteExpression(Pow2Alignment))
3697 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
3698 if (IsLocal && LCOMM == LCOMM::NoAlignment)
3699 return Error(Pow2AlignmentLoc, "alignment not supported on this target");
3701 // If this target takes alignments in bytes (not log) validate and convert.
3702 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
3703 (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
3704 if (!isPowerOf2_64(Pow2Alignment))
3705 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
3706 Pow2Alignment = Log2_64(Pow2Alignment);
3710 if (getLexer().isNot(AsmToken::EndOfStatement))
3711 return TokError("unexpected token in '.comm' or '.lcomm' directive");
3715 // NOTE: a size of zero for a .comm should create a undefined symbol
3716 // but a size of .lcomm creates a bss symbol of size zero.
3718 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
3719 "be less than zero");
3721 // NOTE: The alignment in the directive is a power of 2 value, the assembler
3722 // may internally end up wanting an alignment in bytes.
3723 // FIXME: Diagnose overflow.
3724 if (Pow2Alignment < 0)
3725 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
3726 "alignment, can't be less than zero");
3728 if (!Sym->isUndefined())
3729 return Error(IDLoc, "invalid symbol redefinition");
3731 // Create the Symbol as a common or local common with Size and Pow2Alignment
3733 getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment);
3737 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
3741 /// parseDirectiveAbort
3742 /// ::= .abort [... message ...]
3743 bool AsmParser::parseDirectiveAbort() {
3744 // FIXME: Use loc from directive.
3745 SMLoc Loc = getLexer().getLoc();
3747 StringRef Str = parseStringToEndOfStatement();
3748 if (getLexer().isNot(AsmToken::EndOfStatement))
3749 return TokError("unexpected token in '.abort' directive");
3754 Error(Loc, ".abort detected. Assembly stopping.");
3756 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
3757 // FIXME: Actually abort assembly here.
3762 /// parseDirectiveInclude
3763 /// ::= .include "filename"
3764 bool AsmParser::parseDirectiveInclude() {
3765 if (getLexer().isNot(AsmToken::String))
3766 return TokError("expected string in '.include' directive");
3768 // Allow the strings to have escaped octal character sequence.
3769 std::string Filename;
3770 if (parseEscapedString(Filename))
3772 SMLoc IncludeLoc = getLexer().getLoc();
3775 if (getLexer().isNot(AsmToken::EndOfStatement))
3776 return TokError("unexpected token in '.include' directive");
3778 // Attempt to switch the lexer to the included file before consuming the end
3779 // of statement to avoid losing it when we switch.
3780 if (enterIncludeFile(Filename)) {
3781 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
3788 /// parseDirectiveIncbin
3789 /// ::= .incbin "filename"
3790 bool AsmParser::parseDirectiveIncbin() {
3791 if (getLexer().isNot(AsmToken::String))
3792 return TokError("expected string in '.incbin' directive");
3794 // Allow the strings to have escaped octal character sequence.
3795 std::string Filename;
3796 if (parseEscapedString(Filename))
3798 SMLoc IncbinLoc = getLexer().getLoc();
3801 if (getLexer().isNot(AsmToken::EndOfStatement))
3802 return TokError("unexpected token in '.incbin' directive");
3804 // Attempt to process the included file.
3805 if (processIncbinFile(Filename)) {
3806 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
3813 /// parseDirectiveIf
3814 /// ::= .if{,eq,ge,gt,le,lt,ne} expression
3815 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) {
3816 TheCondStack.push_back(TheCondState);
3817 TheCondState.TheCond = AsmCond::IfCond;
3818 if (TheCondState.Ignore) {
3819 eatToEndOfStatement();
3822 if (parseAbsoluteExpression(ExprValue))
3825 if (getLexer().isNot(AsmToken::EndOfStatement))
3826 return TokError("unexpected token in '.if' directive");
3832 llvm_unreachable("unsupported directive");
3837 ExprValue = ExprValue == 0;
3840 ExprValue = ExprValue >= 0;
3843 ExprValue = ExprValue > 0;
3846 ExprValue = ExprValue <= 0;
3849 ExprValue = ExprValue < 0;
3853 TheCondState.CondMet = ExprValue;
3854 TheCondState.Ignore = !TheCondState.CondMet;
3860 /// parseDirectiveIfb
3862 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
3863 TheCondStack.push_back(TheCondState);
3864 TheCondState.TheCond = AsmCond::IfCond;
3866 if (TheCondState.Ignore) {
3867 eatToEndOfStatement();
3869 StringRef Str = parseStringToEndOfStatement();
3871 if (getLexer().isNot(AsmToken::EndOfStatement))
3872 return TokError("unexpected token in '.ifb' directive");
3876 TheCondState.CondMet = ExpectBlank == Str.empty();
3877 TheCondState.Ignore = !TheCondState.CondMet;
3883 /// parseDirectiveIfc
3884 /// ::= .ifc string1, string2
3885 /// ::= .ifnc string1, string2
3886 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
3887 TheCondStack.push_back(TheCondState);
3888 TheCondState.TheCond = AsmCond::IfCond;
3890 if (TheCondState.Ignore) {
3891 eatToEndOfStatement();
3893 StringRef Str1 = parseStringToComma();
3895 if (getLexer().isNot(AsmToken::Comma))
3896 return TokError("unexpected token in '.ifc' directive");
3900 StringRef Str2 = parseStringToEndOfStatement();
3902 if (getLexer().isNot(AsmToken::EndOfStatement))
3903 return TokError("unexpected token in '.ifc' directive");
3907 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim());
3908 TheCondState.Ignore = !TheCondState.CondMet;
3914 /// parseDirectiveIfeqs
3915 /// ::= .ifeqs string1, string2
3916 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc) {
3917 if (Lexer.isNot(AsmToken::String)) {
3918 TokError("expected string parameter for '.ifeqs' directive");
3919 eatToEndOfStatement();
3923 StringRef String1 = getTok().getStringContents();
3926 if (Lexer.isNot(AsmToken::Comma)) {
3927 TokError("expected comma after first string for '.ifeqs' directive");
3928 eatToEndOfStatement();
3934 if (Lexer.isNot(AsmToken::String)) {
3935 TokError("expected string parameter for '.ifeqs' directive");
3936 eatToEndOfStatement();
3940 StringRef String2 = getTok().getStringContents();
3943 TheCondStack.push_back(TheCondState);
3944 TheCondState.TheCond = AsmCond::IfCond;
3945 TheCondState.CondMet = String1 == String2;
3946 TheCondState.Ignore = !TheCondState.CondMet;
3951 /// parseDirectiveIfdef
3952 /// ::= .ifdef symbol
3953 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
3955 TheCondStack.push_back(TheCondState);
3956 TheCondState.TheCond = AsmCond::IfCond;
3958 if (TheCondState.Ignore) {
3959 eatToEndOfStatement();
3961 if (parseIdentifier(Name))
3962 return TokError("expected identifier after '.ifdef'");
3966 MCSymbol *Sym = getContext().LookupSymbol(Name);
3969 TheCondState.CondMet = (Sym && !Sym->isUndefined());
3971 TheCondState.CondMet = (!Sym || Sym->isUndefined());
3972 TheCondState.Ignore = !TheCondState.CondMet;
3978 /// parseDirectiveElseIf
3979 /// ::= .elseif expression
3980 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) {
3981 if (TheCondState.TheCond != AsmCond::IfCond &&
3982 TheCondState.TheCond != AsmCond::ElseIfCond)
3983 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
3985 TheCondState.TheCond = AsmCond::ElseIfCond;
3987 bool LastIgnoreState = false;
3988 if (!TheCondStack.empty())
3989 LastIgnoreState = TheCondStack.back().Ignore;
3990 if (LastIgnoreState || TheCondState.CondMet) {
3991 TheCondState.Ignore = true;
3992 eatToEndOfStatement();
3995 if (parseAbsoluteExpression(ExprValue))
3998 if (getLexer().isNot(AsmToken::EndOfStatement))
3999 return TokError("unexpected token in '.elseif' directive");
4002 TheCondState.CondMet = ExprValue;
4003 TheCondState.Ignore = !TheCondState.CondMet;
4009 /// parseDirectiveElse
4011 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) {
4012 if (getLexer().isNot(AsmToken::EndOfStatement))
4013 return TokError("unexpected token in '.else' directive");
4017 if (TheCondState.TheCond != AsmCond::IfCond &&
4018 TheCondState.TheCond != AsmCond::ElseIfCond)
4019 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
4021 TheCondState.TheCond = AsmCond::ElseCond;
4022 bool LastIgnoreState = false;
4023 if (!TheCondStack.empty())
4024 LastIgnoreState = TheCondStack.back().Ignore;
4025 if (LastIgnoreState || TheCondState.CondMet)
4026 TheCondState.Ignore = true;
4028 TheCondState.Ignore = false;
4033 /// parseDirectiveEnd
4035 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) {
4036 if (getLexer().isNot(AsmToken::EndOfStatement))
4037 return TokError("unexpected token in '.end' directive");
4041 while (Lexer.isNot(AsmToken::Eof))
4047 /// parseDirectiveError
4049 /// ::= .error [string]
4050 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) {
4051 if (!TheCondStack.empty()) {
4052 if (TheCondStack.back().Ignore) {
4053 eatToEndOfStatement();
4059 return Error(L, ".err encountered");
4061 StringRef Message = ".error directive invoked in source file";
4062 if (Lexer.isNot(AsmToken::EndOfStatement)) {
4063 if (Lexer.isNot(AsmToken::String)) {
4064 TokError(".error argument must be a string");
4065 eatToEndOfStatement();
4069 Message = getTok().getStringContents();
4077 /// parseDirectiveEndIf
4079 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) {
4080 if (getLexer().isNot(AsmToken::EndOfStatement))
4081 return TokError("unexpected token in '.endif' directive");
4085 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty())
4086 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
4088 if (!TheCondStack.empty()) {
4089 TheCondState = TheCondStack.back();
4090 TheCondStack.pop_back();
4096 void AsmParser::initializeDirectiveKindMap() {
4097 DirectiveKindMap[".set"] = DK_SET;
4098 DirectiveKindMap[".equ"] = DK_EQU;
4099 DirectiveKindMap[".equiv"] = DK_EQUIV;
4100 DirectiveKindMap[".ascii"] = DK_ASCII;
4101 DirectiveKindMap[".asciz"] = DK_ASCIZ;
4102 DirectiveKindMap[".string"] = DK_STRING;
4103 DirectiveKindMap[".byte"] = DK_BYTE;
4104 DirectiveKindMap[".short"] = DK_SHORT;
4105 DirectiveKindMap[".value"] = DK_VALUE;
4106 DirectiveKindMap[".2byte"] = DK_2BYTE;
4107 DirectiveKindMap[".long"] = DK_LONG;
4108 DirectiveKindMap[".int"] = DK_INT;
4109 DirectiveKindMap[".4byte"] = DK_4BYTE;
4110 DirectiveKindMap[".quad"] = DK_QUAD;
4111 DirectiveKindMap[".8byte"] = DK_8BYTE;
4112 DirectiveKindMap[".octa"] = DK_OCTA;
4113 DirectiveKindMap[".single"] = DK_SINGLE;
4114 DirectiveKindMap[".float"] = DK_FLOAT;
4115 DirectiveKindMap[".double"] = DK_DOUBLE;
4116 DirectiveKindMap[".align"] = DK_ALIGN;
4117 DirectiveKindMap[".align32"] = DK_ALIGN32;
4118 DirectiveKindMap[".balign"] = DK_BALIGN;
4119 DirectiveKindMap[".balignw"] = DK_BALIGNW;
4120 DirectiveKindMap[".balignl"] = DK_BALIGNL;
4121 DirectiveKindMap[".p2align"] = DK_P2ALIGN;
4122 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW;
4123 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL;
4124 DirectiveKindMap[".org"] = DK_ORG;
4125 DirectiveKindMap[".fill"] = DK_FILL;
4126 DirectiveKindMap[".zero"] = DK_ZERO;
4127 DirectiveKindMap[".extern"] = DK_EXTERN;
4128 DirectiveKindMap[".globl"] = DK_GLOBL;
4129 DirectiveKindMap[".global"] = DK_GLOBAL;
4130 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE;
4131 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP;
4132 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER;
4133 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN;
4134 DirectiveKindMap[".reference"] = DK_REFERENCE;
4135 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION;
4136 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE;
4137 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN;
4138 DirectiveKindMap[".comm"] = DK_COMM;
4139 DirectiveKindMap[".common"] = DK_COMMON;
4140 DirectiveKindMap[".lcomm"] = DK_LCOMM;
4141 DirectiveKindMap[".abort"] = DK_ABORT;
4142 DirectiveKindMap[".include"] = DK_INCLUDE;
4143 DirectiveKindMap[".incbin"] = DK_INCBIN;
4144 DirectiveKindMap[".code16"] = DK_CODE16;
4145 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC;
4146 DirectiveKindMap[".rept"] = DK_REPT;
4147 DirectiveKindMap[".rep"] = DK_REPT;
4148 DirectiveKindMap[".irp"] = DK_IRP;
4149 DirectiveKindMap[".irpc"] = DK_IRPC;
4150 DirectiveKindMap[".endr"] = DK_ENDR;
4151 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE;
4152 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK;
4153 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK;
4154 DirectiveKindMap[".if"] = DK_IF;
4155 DirectiveKindMap[".ifeq"] = DK_IFEQ;
4156 DirectiveKindMap[".ifge"] = DK_IFGE;
4157 DirectiveKindMap[".ifgt"] = DK_IFGT;
4158 DirectiveKindMap[".ifle"] = DK_IFLE;
4159 DirectiveKindMap[".iflt"] = DK_IFLT;
4160 DirectiveKindMap[".ifne"] = DK_IFNE;
4161 DirectiveKindMap[".ifb"] = DK_IFB;
4162 DirectiveKindMap[".ifnb"] = DK_IFNB;
4163 DirectiveKindMap[".ifc"] = DK_IFC;
4164 DirectiveKindMap[".ifeqs"] = DK_IFEQS;
4165 DirectiveKindMap[".ifnc"] = DK_IFNC;
4166 DirectiveKindMap[".ifdef"] = DK_IFDEF;
4167 DirectiveKindMap[".ifndef"] = DK_IFNDEF;
4168 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF;
4169 DirectiveKindMap[".elseif"] = DK_ELSEIF;
4170 DirectiveKindMap[".else"] = DK_ELSE;
4171 DirectiveKindMap[".end"] = DK_END;
4172 DirectiveKindMap[".endif"] = DK_ENDIF;
4173 DirectiveKindMap[".skip"] = DK_SKIP;
4174 DirectiveKindMap[".space"] = DK_SPACE;
4175 DirectiveKindMap[".file"] = DK_FILE;
4176 DirectiveKindMap[".line"] = DK_LINE;
4177 DirectiveKindMap[".loc"] = DK_LOC;
4178 DirectiveKindMap[".stabs"] = DK_STABS;
4179 DirectiveKindMap[".sleb128"] = DK_SLEB128;
4180 DirectiveKindMap[".uleb128"] = DK_ULEB128;
4181 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS;
4182 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC;
4183 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC;
4184 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA;
4185 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET;
4186 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET;
4187 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER;
4188 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET;
4189 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET;
4190 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY;
4191 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA;
4192 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE;
4193 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE;
4194 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE;
4195 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE;
4196 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE;
4197 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME;
4198 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED;
4199 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER;
4200 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE;
4201 DirectiveKindMap[".macros_on"] = DK_MACROS_ON;
4202 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF;
4203 DirectiveKindMap[".macro"] = DK_MACRO;
4204 DirectiveKindMap[".endm"] = DK_ENDM;
4205 DirectiveKindMap[".endmacro"] = DK_ENDMACRO;
4206 DirectiveKindMap[".purgem"] = DK_PURGEM;
4207 DirectiveKindMap[".err"] = DK_ERR;
4208 DirectiveKindMap[".error"] = DK_ERROR;
4211 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) {
4212 AsmToken EndToken, StartToken = getTok();
4214 unsigned NestLevel = 0;
4216 // Check whether we have reached the end of the file.
4217 if (getLexer().is(AsmToken::Eof)) {
4218 Error(DirectiveLoc, "no matching '.endr' in definition");
4222 if (Lexer.is(AsmToken::Identifier) &&
4223 (getTok().getIdentifier() == ".rept")) {
4227 // Otherwise, check whether we have reached the .endr.
4228 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") {
4229 if (NestLevel == 0) {
4230 EndToken = getTok();
4232 if (Lexer.isNot(AsmToken::EndOfStatement)) {
4233 TokError("unexpected token in '.endr' directive");
4241 // Otherwise, scan till the end of the statement.
4242 eatToEndOfStatement();
4245 const char *BodyStart = StartToken.getLoc().getPointer();
4246 const char *BodyEnd = EndToken.getLoc().getPointer();
4247 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
4249 // We Are Anonymous.
4250 MacroLikeBodies.push_back(MCAsmMacro(StringRef(), Body, None));
4251 return &MacroLikeBodies.back();
4254 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
4255 raw_svector_ostream &OS) {
4258 MemoryBuffer *Instantiation =
4259 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
4261 // Create the macro instantiation object and add to the current macro
4262 // instantiation stack.
4263 MacroInstantiation *MI = new MacroInstantiation(
4264 M, DirectiveLoc, CurBuffer, getTok().getLoc(), Instantiation);
4265 ActiveMacros.push_back(MI);
4267 // Jump to the macro instantiation and prime the lexer.
4268 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
4269 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
4273 /// parseDirectiveRept
4274 /// ::= .rep | .rept count
4275 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) {
4276 const MCExpr *CountExpr;
4277 SMLoc CountLoc = getTok().getLoc();
4278 if (parseExpression(CountExpr))
4282 if (!CountExpr->EvaluateAsAbsolute(Count)) {
4283 eatToEndOfStatement();
4284 return Error(CountLoc, "unexpected token in '" + Dir + "' directive");
4288 return Error(CountLoc, "Count is negative");
4290 if (Lexer.isNot(AsmToken::EndOfStatement))
4291 return TokError("unexpected token in '" + Dir + "' directive");
4293 // Eat the end of statement.
4296 // Lex the rept definition.
4297 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4301 // Macro instantiation is lexical, unfortunately. We construct a new buffer
4302 // to hold the macro body with substitutions.
4303 SmallString<256> Buf;
4304 raw_svector_ostream OS(Buf);
4306 if (expandMacro(OS, M->Body, None, None, getTok().getLoc()))
4309 instantiateMacroLikeBody(M, DirectiveLoc, OS);
4314 /// parseDirectiveIrp
4315 /// ::= .irp symbol,values
4316 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) {
4317 MCAsmMacroParameter Parameter;
4319 if (parseIdentifier(Parameter.Name))
4320 return TokError("expected identifier in '.irp' directive");
4322 if (Lexer.isNot(AsmToken::Comma))
4323 return TokError("expected comma in '.irp' directive");
4327 MCAsmMacroArguments A;
4328 if (parseMacroArguments(nullptr, A))
4331 // Eat the end of statement.
4334 // Lex the irp definition.
4335 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4339 // Macro instantiation is lexical, unfortunately. We construct a new buffer
4340 // to hold the macro body with substitutions.
4341 SmallString<256> Buf;
4342 raw_svector_ostream OS(Buf);
4344 for (MCAsmMacroArguments::iterator i = A.begin(), e = A.end(); i != e; ++i) {
4345 if (expandMacro(OS, M->Body, Parameter, *i, getTok().getLoc()))
4349 instantiateMacroLikeBody(M, DirectiveLoc, OS);
4354 /// parseDirectiveIrpc
4355 /// ::= .irpc symbol,values
4356 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) {
4357 MCAsmMacroParameter Parameter;
4359 if (parseIdentifier(Parameter.Name))
4360 return TokError("expected identifier in '.irpc' directive");
4362 if (Lexer.isNot(AsmToken::Comma))
4363 return TokError("expected comma in '.irpc' directive");
4367 MCAsmMacroArguments A;
4368 if (parseMacroArguments(nullptr, A))
4371 if (A.size() != 1 || A.front().size() != 1)
4372 return TokError("unexpected token in '.irpc' directive");
4374 // Eat the end of statement.
4377 // Lex the irpc definition.
4378 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4382 // Macro instantiation is lexical, unfortunately. We construct a new buffer
4383 // to hold the macro body with substitutions.
4384 SmallString<256> Buf;
4385 raw_svector_ostream OS(Buf);
4387 StringRef Values = A.front().front().getString();
4388 for (std::size_t I = 0, End = Values.size(); I != End; ++I) {
4389 MCAsmMacroArgument Arg;
4390 Arg.push_back(AsmToken(AsmToken::Identifier, Values.slice(I, I + 1)));
4392 if (expandMacro(OS, M->Body, Parameter, Arg, getTok().getLoc()))
4396 instantiateMacroLikeBody(M, DirectiveLoc, OS);
4401 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) {
4402 if (ActiveMacros.empty())
4403 return TokError("unmatched '.endr' directive");
4405 // The only .repl that should get here are the ones created by
4406 // instantiateMacroLikeBody.
4407 assert(getLexer().is(AsmToken::EndOfStatement));
4413 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info,
4415 const MCExpr *Value;
4416 SMLoc ExprLoc = getLexer().getLoc();
4417 if (parseExpression(Value))
4419 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4421 return Error(ExprLoc, "unexpected expression in _emit");
4422 uint64_t IntValue = MCE->getValue();
4423 if (!isUIntN(8, IntValue) && !isIntN(8, IntValue))
4424 return Error(ExprLoc, "literal value out of range for directive");
4426 Info.AsmRewrites->push_back(AsmRewrite(AOK_Emit, IDLoc, Len));
4430 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) {
4431 const MCExpr *Value;
4432 SMLoc ExprLoc = getLexer().getLoc();
4433 if (parseExpression(Value))
4435 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4437 return Error(ExprLoc, "unexpected expression in align");
4438 uint64_t IntValue = MCE->getValue();
4439 if (!isPowerOf2_64(IntValue))
4440 return Error(ExprLoc, "literal value not a power of two greater then zero");
4442 Info.AsmRewrites->push_back(
4443 AsmRewrite(AOK_Align, IDLoc, 5, Log2_64(IntValue)));
4447 // We are comparing pointers, but the pointers are relative to a single string.
4448 // Thus, this should always be deterministic.
4449 static int rewritesSort(const AsmRewrite *AsmRewriteA,
4450 const AsmRewrite *AsmRewriteB) {
4451 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer())
4453 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer())
4456 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output
4457 // rewrite to the same location. Make sure the SizeDirective rewrite is
4458 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This
4459 // ensures the sort algorithm is stable.
4460 if (AsmRewritePrecedence[AsmRewriteA->Kind] >
4461 AsmRewritePrecedence[AsmRewriteB->Kind])
4464 if (AsmRewritePrecedence[AsmRewriteA->Kind] <
4465 AsmRewritePrecedence[AsmRewriteB->Kind])
4467 llvm_unreachable("Unstable rewrite sort.");
4470 bool AsmParser::parseMSInlineAsm(
4471 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs,
4472 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
4473 SmallVectorImpl<std::string> &Constraints,
4474 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII,
4475 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) {
4476 SmallVector<void *, 4> InputDecls;
4477 SmallVector<void *, 4> OutputDecls;
4478 SmallVector<bool, 4> InputDeclsAddressOf;
4479 SmallVector<bool, 4> OutputDeclsAddressOf;
4480 SmallVector<std::string, 4> InputConstraints;
4481 SmallVector<std::string, 4> OutputConstraints;
4482 SmallVector<unsigned, 4> ClobberRegs;
4484 SmallVector<AsmRewrite, 4> AsmStrRewrites;
4489 // While we have input, parse each statement.
4490 unsigned InputIdx = 0;
4491 unsigned OutputIdx = 0;
4492 while (getLexer().isNot(AsmToken::Eof)) {
4493 ParseStatementInfo Info(&AsmStrRewrites);
4494 if (parseStatement(Info))
4497 if (Info.ParseError)
4500 if (Info.Opcode == ~0U)
4503 const MCInstrDesc &Desc = MII->get(Info.Opcode);
4505 // Build the list of clobbers, outputs and inputs.
4506 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
4507 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i];
4510 if (Operand.isImm())
4513 // Register operand.
4514 if (Operand.isReg() && !Operand.needAddressOf()) {
4515 unsigned NumDefs = Desc.getNumDefs();
4517 if (NumDefs && Operand.getMCOperandNum() < NumDefs)
4518 ClobberRegs.push_back(Operand.getReg());
4522 // Expr/Input or Output.
4523 StringRef SymName = Operand.getSymName();
4524 if (SymName.empty())
4527 void *OpDecl = Operand.getOpDecl();
4531 bool isOutput = (i == 1) && Desc.mayStore();
4532 SMLoc Start = SMLoc::getFromPointer(SymName.data());
4535 OutputDecls.push_back(OpDecl);
4536 OutputDeclsAddressOf.push_back(Operand.needAddressOf());
4537 OutputConstraints.push_back('=' + Operand.getConstraint().str());
4538 AsmStrRewrites.push_back(AsmRewrite(AOK_Output, Start, SymName.size()));
4540 InputDecls.push_back(OpDecl);
4541 InputDeclsAddressOf.push_back(Operand.needAddressOf());
4542 InputConstraints.push_back(Operand.getConstraint().str());
4543 AsmStrRewrites.push_back(AsmRewrite(AOK_Input, Start, SymName.size()));
4547 // Consider implicit defs to be clobbers. Think of cpuid and push.
4548 ArrayRef<uint16_t> ImpDefs(Desc.getImplicitDefs(),
4549 Desc.getNumImplicitDefs());
4550 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end());
4553 // Set the number of Outputs and Inputs.
4554 NumOutputs = OutputDecls.size();
4555 NumInputs = InputDecls.size();
4557 // Set the unique clobbers.
4558 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end());
4559 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()),
4561 Clobbers.assign(ClobberRegs.size(), std::string());
4562 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) {
4563 raw_string_ostream OS(Clobbers[I]);
4564 IP->printRegName(OS, ClobberRegs[I]);
4567 // Merge the various outputs and inputs. Output are expected first.
4568 if (NumOutputs || NumInputs) {
4569 unsigned NumExprs = NumOutputs + NumInputs;
4570 OpDecls.resize(NumExprs);
4571 Constraints.resize(NumExprs);
4572 for (unsigned i = 0; i < NumOutputs; ++i) {
4573 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]);
4574 Constraints[i] = OutputConstraints[i];
4576 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
4577 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]);
4578 Constraints[j] = InputConstraints[i];
4582 // Build the IR assembly string.
4584 const char *AsmStart = SrcMgr.getMemoryBuffer(0)->getBufferStart();
4585 const char *AsmEnd = SrcMgr.getMemoryBuffer(0)->getBufferEnd();
4586 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort);
4587 for (const AsmRewrite &AR : AsmStrRewrites) {
4588 AsmRewriteKind Kind = AR.Kind;
4589 if (Kind == AOK_Delete)
4592 const char *Loc = AR.Loc.getPointer();
4593 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!");
4595 // Emit everything up to the immediate/expression.
4596 if (unsigned Len = Loc - AsmStart)
4597 OS << StringRef(AsmStart, Len);
4599 // Skip the original expression.
4600 if (Kind == AOK_Skip) {
4601 AsmStart = Loc + AR.Len;
4605 unsigned AdditionalSkip = 0;
4606 // Rewrite expressions in $N notation.
4611 OS << "$$" << AR.Val;
4617 OS << '$' << InputIdx++;
4620 OS << '$' << OutputIdx++;
4622 case AOK_SizeDirective:
4625 case 8: OS << "byte ptr "; break;
4626 case 16: OS << "word ptr "; break;
4627 case 32: OS << "dword ptr "; break;
4628 case 64: OS << "qword ptr "; break;
4629 case 80: OS << "xword ptr "; break;
4630 case 128: OS << "xmmword ptr "; break;
4631 case 256: OS << "ymmword ptr "; break;
4638 unsigned Val = AR.Val;
4639 OS << ".align " << Val;
4641 // Skip the original immediate.
4642 assert(Val < 10 && "Expected alignment less then 2^10.");
4643 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4;
4646 case AOK_DotOperator:
4647 // Insert the dot if the user omitted it.
4648 if (OS.str().back() != '.')
4654 // Skip the original expression.
4655 AsmStart = Loc + AR.Len + AdditionalSkip;
4658 // Emit the remainder of the asm string.
4659 if (AsmStart != AsmEnd)
4660 OS << StringRef(AsmStart, AsmEnd - AsmStart);
4662 AsmString = OS.str();
4666 /// \brief Create an MCAsmParser instance.
4667 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C,
4668 MCStreamer &Out, const MCAsmInfo &MAI) {
4669 return new AsmParser(SM, C, Out, MAI);