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"
48 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() {}
51 /// \brief Helper types for tracking macro definitions.
52 typedef std::vector<AsmToken> MCAsmMacroArgument;
53 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
55 struct MCAsmMacroParameter {
57 MCAsmMacroArgument Value;
61 MCAsmMacroParameter() : Required(false), Vararg(false) {}
64 typedef std::vector<MCAsmMacroParameter> MCAsmMacroParameters;
69 MCAsmMacroParameters Parameters;
72 MCAsmMacro(StringRef N, StringRef B, MCAsmMacroParameters P)
73 : Name(N), Body(B), Parameters(std::move(P)) {}
76 /// \brief Helper class for storing information about an active macro
78 struct MacroInstantiation {
79 /// The location of the instantiation.
80 SMLoc InstantiationLoc;
82 /// The buffer where parsing should resume upon instantiation completion.
85 /// The location where parsing should resume upon instantiation completion.
88 /// The depth of TheCondStack at the start of the instantiation.
89 size_t CondStackDepth;
92 MacroInstantiation(SMLoc IL, int EB, SMLoc EL, size_t CondStackDepth);
95 struct ParseStatementInfo {
96 /// \brief The parsed operands from the last parsed statement.
97 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands;
99 /// \brief The opcode from the last parsed instruction.
102 /// \brief Was there an error parsing the inline assembly?
105 SmallVectorImpl<AsmRewrite> *AsmRewrites;
107 ParseStatementInfo() : Opcode(~0U), ParseError(false), AsmRewrites(nullptr) {}
108 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
109 : Opcode(~0), ParseError(false), AsmRewrites(rewrites) {}
112 /// \brief The concrete assembly parser instance.
113 class AsmParser : public MCAsmParser {
114 AsmParser(const AsmParser &) = delete;
115 void operator=(const AsmParser &) = delete;
120 const MCAsmInfo &MAI;
122 SourceMgr::DiagHandlerTy SavedDiagHandler;
123 void *SavedDiagContext;
124 std::unique_ptr<MCAsmParserExtension> PlatformParser;
126 /// This is the current buffer index we're lexing from as managed by the
127 /// SourceMgr object.
130 AsmCond TheCondState;
131 std::vector<AsmCond> TheCondStack;
133 /// \brief maps directive names to handler methods in parser
134 /// extensions. Extensions register themselves in this map by calling
135 /// addDirectiveHandler.
136 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
138 /// \brief Map of currently defined macros.
139 StringMap<MCAsmMacro> MacroMap;
141 /// \brief Stack of active macro instantiations.
142 std::vector<MacroInstantiation*> ActiveMacros;
144 /// \brief List of bodies of anonymous macros.
145 std::deque<MCAsmMacro> MacroLikeBodies;
147 /// Boolean tracking whether macro substitution is enabled.
148 unsigned MacrosEnabledFlag : 1;
150 /// Flag tracking whether any errors have been encountered.
151 unsigned HadError : 1;
153 /// The values from the last parsed cpp hash file line comment if any.
154 StringRef CppHashFilename;
155 int64_t CppHashLineNumber;
158 /// When generating dwarf for assembly source files we need to calculate the
159 /// logical line number based on the last parsed cpp hash file line comment
160 /// and current line. Since this is slow and messes up the SourceMgr's
161 /// cache we save the last info we queried with SrcMgr.FindLineNumber().
162 SMLoc LastQueryIDLoc;
163 unsigned LastQueryBuffer;
164 unsigned LastQueryLine;
166 /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
167 unsigned AssemblerDialect;
169 /// \brief is Darwin compatibility enabled?
172 /// \brief Are we parsing ms-style inline assembly?
173 bool ParsingInlineAsm;
176 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
177 const MCAsmInfo &MAI);
178 ~AsmParser() override;
180 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
182 void addDirectiveHandler(StringRef Directive,
183 ExtensionDirectiveHandler Handler) override {
184 ExtensionDirectiveMap[Directive] = Handler;
188 /// @name MCAsmParser Interface
191 SourceMgr &getSourceManager() override { return SrcMgr; }
192 MCAsmLexer &getLexer() override { return Lexer; }
193 MCContext &getContext() override { return Ctx; }
194 MCStreamer &getStreamer() override { return Out; }
195 unsigned getAssemblerDialect() override {
196 if (AssemblerDialect == ~0U)
197 return MAI.getAssemblerDialect();
199 return AssemblerDialect;
201 void setAssemblerDialect(unsigned i) override {
202 AssemblerDialect = i;
205 void Note(SMLoc L, const Twine &Msg,
206 ArrayRef<SMRange> Ranges = None) override;
207 bool Warning(SMLoc L, const Twine &Msg,
208 ArrayRef<SMRange> Ranges = None) override;
209 bool Error(SMLoc L, const Twine &Msg,
210 ArrayRef<SMRange> Ranges = None) override;
212 const AsmToken &Lex() override;
214 void setParsingInlineAsm(bool V) override { ParsingInlineAsm = V; }
215 bool isParsingInlineAsm() override { return ParsingInlineAsm; }
217 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString,
218 unsigned &NumOutputs, unsigned &NumInputs,
219 SmallVectorImpl<std::pair<void *,bool> > &OpDecls,
220 SmallVectorImpl<std::string> &Constraints,
221 SmallVectorImpl<std::string> &Clobbers,
222 const MCInstrInfo *MII, const MCInstPrinter *IP,
223 MCAsmParserSemaCallback &SI) override;
225 bool parseExpression(const MCExpr *&Res);
226 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
227 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override;
228 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
229 bool parseAbsoluteExpression(int64_t &Res) override;
231 /// \brief Parse an identifier or string (as a quoted identifier)
232 /// and set \p Res to the identifier contents.
233 bool parseIdentifier(StringRef &Res) override;
234 void eatToEndOfStatement() override;
236 void checkForValidSection() override;
241 bool parseStatement(ParseStatementInfo &Info,
242 MCAsmParserSemaCallback *SI);
243 void eatToEndOfLine();
244 bool parseCppHashLineFilenameComment(const SMLoc &L);
246 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body,
247 ArrayRef<MCAsmMacroParameter> Parameters);
248 bool expandMacro(raw_svector_ostream &OS, StringRef Body,
249 ArrayRef<MCAsmMacroParameter> Parameters,
250 ArrayRef<MCAsmMacroArgument> A,
253 /// \brief Are macros enabled in the parser?
254 bool areMacrosEnabled() {return MacrosEnabledFlag;}
256 /// \brief Control a flag in the parser that enables or disables macros.
257 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;}
259 /// \brief Lookup a previously defined macro.
260 /// \param Name Macro name.
261 /// \returns Pointer to macro. NULL if no such macro was defined.
262 const MCAsmMacro* lookupMacro(StringRef Name);
264 /// \brief Define a new macro with the given name and information.
265 void defineMacro(StringRef Name, MCAsmMacro Macro);
267 /// \brief Undefine a macro. If no such macro was defined, it's a no-op.
268 void undefineMacro(StringRef Name);
270 /// \brief Are we inside a macro instantiation?
271 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();}
273 /// \brief Handle entry to macro instantiation.
275 /// \param M The macro.
276 /// \param NameLoc Instantiation location.
277 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc);
279 /// \brief Handle exit from macro instantiation.
280 void handleMacroExit();
282 /// \brief Extract AsmTokens for a macro argument.
283 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg);
285 /// \brief Parse all macro arguments for a given macro.
286 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A);
288 void printMacroInstantiations();
289 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
290 ArrayRef<SMRange> Ranges = None) const {
291 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
293 static void DiagHandler(const SMDiagnostic &Diag, void *Context);
295 /// \brief Enter the specified file. This returns true on failure.
296 bool enterIncludeFile(const std::string &Filename);
298 /// \brief Process the specified file for the .incbin directive.
299 /// This returns true on failure.
300 bool processIncbinFile(const std::string &Filename);
302 /// \brief Reset the current lexer position to that given by \p Loc. The
303 /// current token is not set; clients should ensure Lex() is called
306 /// \param InBuffer If not 0, should be the known buffer id that contains the
308 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0);
310 /// \brief Parse up to the end of statement and a return the contents from the
311 /// current token until the end of the statement; the current token on exit
312 /// will be either the EndOfStatement or EOF.
313 StringRef parseStringToEndOfStatement() override;
315 /// \brief Parse until the end of a statement or a comma is encountered,
316 /// return the contents from the current token up to the end or comma.
317 StringRef parseStringToComma();
319 bool parseAssignment(StringRef Name, bool allow_redef,
320 bool NoDeadStrip = false);
322 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
323 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
324 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
326 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
328 // Generic (target and platform independent) directive parsing.
330 DK_NO_DIRECTIVE, // Placeholder
331 DK_SET, DK_EQU, DK_EQUIV, DK_ASCII, DK_ASCIZ, DK_STRING, DK_BYTE, DK_SHORT,
332 DK_VALUE, DK_2BYTE, DK_LONG, DK_INT, DK_4BYTE, DK_QUAD, DK_8BYTE, DK_OCTA,
333 DK_SINGLE, DK_FLOAT, DK_DOUBLE, DK_ALIGN, DK_ALIGN32, DK_BALIGN, DK_BALIGNW,
334 DK_BALIGNL, DK_P2ALIGN, DK_P2ALIGNW, DK_P2ALIGNL, DK_ORG, DK_FILL, DK_ENDR,
335 DK_BUNDLE_ALIGN_MODE, DK_BUNDLE_LOCK, DK_BUNDLE_UNLOCK,
336 DK_ZERO, DK_EXTERN, DK_GLOBL, DK_GLOBAL,
337 DK_LAZY_REFERENCE, DK_NO_DEAD_STRIP, DK_SYMBOL_RESOLVER, DK_PRIVATE_EXTERN,
338 DK_REFERENCE, DK_WEAK_DEFINITION, DK_WEAK_REFERENCE,
339 DK_WEAK_DEF_CAN_BE_HIDDEN, DK_COMM, DK_COMMON, DK_LCOMM, DK_ABORT,
340 DK_INCLUDE, DK_INCBIN, DK_CODE16, DK_CODE16GCC, DK_REPT, DK_IRP, DK_IRPC,
341 DK_IF, DK_IFEQ, DK_IFGE, DK_IFGT, DK_IFLE, DK_IFLT, DK_IFNE, DK_IFB,
342 DK_IFNB, DK_IFC, DK_IFEQS, DK_IFNC, DK_IFNES, DK_IFDEF, DK_IFNDEF,
343 DK_IFNOTDEF, DK_ELSEIF, DK_ELSE, DK_ENDIF,
344 DK_SPACE, DK_SKIP, DK_FILE, DK_LINE, DK_LOC, DK_STABS,
345 DK_CFI_SECTIONS, DK_CFI_STARTPROC, DK_CFI_ENDPROC, DK_CFI_DEF_CFA,
346 DK_CFI_DEF_CFA_OFFSET, DK_CFI_ADJUST_CFA_OFFSET, DK_CFI_DEF_CFA_REGISTER,
347 DK_CFI_OFFSET, DK_CFI_REL_OFFSET, DK_CFI_PERSONALITY, DK_CFI_LSDA,
348 DK_CFI_REMEMBER_STATE, DK_CFI_RESTORE_STATE, DK_CFI_SAME_VALUE,
349 DK_CFI_RESTORE, DK_CFI_ESCAPE, DK_CFI_SIGNAL_FRAME, DK_CFI_UNDEFINED,
350 DK_CFI_REGISTER, DK_CFI_WINDOW_SAVE,
351 DK_MACROS_ON, DK_MACROS_OFF,
352 DK_MACRO, DK_EXITM, DK_ENDM, DK_ENDMACRO, DK_PURGEM,
353 DK_SLEB128, DK_ULEB128,
354 DK_ERR, DK_ERROR, DK_WARNING,
358 /// \brief Maps directive name --> DirectiveKind enum, for
359 /// directives parsed by this class.
360 StringMap<DirectiveKind> DirectiveKindMap;
362 // ".ascii", ".asciz", ".string"
363 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
364 bool parseDirectiveValue(unsigned Size); // ".byte", ".long", ...
365 bool parseDirectiveOctaValue(); // ".octa"
366 bool parseDirectiveRealValue(const fltSemantics &); // ".single", ...
367 bool parseDirectiveFill(); // ".fill"
368 bool parseDirectiveZero(); // ".zero"
369 // ".set", ".equ", ".equiv"
370 bool parseDirectiveSet(StringRef IDVal, bool allow_redef);
371 bool parseDirectiveOrg(); // ".org"
372 // ".align{,32}", ".p2align{,w,l}"
373 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
375 // ".file", ".line", ".loc", ".stabs"
376 bool parseDirectiveFile(SMLoc DirectiveLoc);
377 bool parseDirectiveLine();
378 bool parseDirectiveLoc();
379 bool parseDirectiveStabs();
382 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
383 bool parseDirectiveCFIWindowSave();
384 bool parseDirectiveCFISections();
385 bool parseDirectiveCFIStartProc();
386 bool parseDirectiveCFIEndProc();
387 bool parseDirectiveCFIDefCfaOffset();
388 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
389 bool parseDirectiveCFIAdjustCfaOffset();
390 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
391 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
392 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
393 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
394 bool parseDirectiveCFIRememberState();
395 bool parseDirectiveCFIRestoreState();
396 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
397 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
398 bool parseDirectiveCFIEscape();
399 bool parseDirectiveCFISignalFrame();
400 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
403 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
404 bool parseDirectiveExitMacro(StringRef Directive);
405 bool parseDirectiveEndMacro(StringRef Directive);
406 bool parseDirectiveMacro(SMLoc DirectiveLoc);
407 bool parseDirectiveMacrosOnOff(StringRef Directive);
409 // ".bundle_align_mode"
410 bool parseDirectiveBundleAlignMode();
412 bool parseDirectiveBundleLock();
414 bool parseDirectiveBundleUnlock();
417 bool parseDirectiveSpace(StringRef IDVal);
419 // .sleb128 (Signed=true) and .uleb128 (Signed=false)
420 bool parseDirectiveLEB128(bool Signed);
422 /// \brief Parse a directive like ".globl" which
423 /// accepts a single symbol (which should be a label or an external).
424 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
426 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
428 bool parseDirectiveAbort(); // ".abort"
429 bool parseDirectiveInclude(); // ".include"
430 bool parseDirectiveIncbin(); // ".incbin"
432 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne"
433 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
434 // ".ifb" or ".ifnb", depending on ExpectBlank.
435 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
436 // ".ifc" or ".ifnc", depending on ExpectEqual.
437 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
438 // ".ifeqs" or ".ifnes", depending on ExpectEqual.
439 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual);
440 // ".ifdef" or ".ifndef", depending on expect_defined
441 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
442 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
443 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else"
444 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
445 bool parseEscapedString(std::string &Data) override;
447 const MCExpr *applyModifierToExpr(const MCExpr *E,
448 MCSymbolRefExpr::VariantKind Variant);
450 // Macro-like directives
451 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
452 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
453 raw_svector_ostream &OS);
454 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive);
455 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp"
456 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
457 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
459 // "_emit" or "__emit"
460 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
464 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
467 bool parseDirectiveEnd(SMLoc DirectiveLoc);
469 // ".err" or ".error"
470 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage);
473 bool parseDirectiveWarning(SMLoc DirectiveLoc);
475 void initializeDirectiveKindMap();
481 extern MCAsmParserExtension *createDarwinAsmParser();
482 extern MCAsmParserExtension *createELFAsmParser();
483 extern MCAsmParserExtension *createCOFFAsmParser();
487 enum { DEFAULT_ADDRSPACE = 0 };
489 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
490 const MCAsmInfo &MAI)
491 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM),
492 PlatformParser(nullptr), CurBuffer(SM.getMainFileID()),
493 MacrosEnabledFlag(true), HadError(false), CppHashLineNumber(0),
494 AssemblerDialect(~0U), IsDarwin(false), ParsingInlineAsm(false) {
495 // Save the old handler.
496 SavedDiagHandler = SrcMgr.getDiagHandler();
497 SavedDiagContext = SrcMgr.getDiagContext();
498 // Set our own handler which calls the saved handler.
499 SrcMgr.setDiagHandler(DiagHandler, this);
500 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
502 // Initialize the platform / file format parser.
503 switch (Ctx.getObjectFileInfo()->getObjectFileType()) {
504 case MCObjectFileInfo::IsCOFF:
505 PlatformParser.reset(createCOFFAsmParser());
507 case MCObjectFileInfo::IsMachO:
508 PlatformParser.reset(createDarwinAsmParser());
511 case MCObjectFileInfo::IsELF:
512 PlatformParser.reset(createELFAsmParser());
516 PlatformParser->Initialize(*this);
517 initializeDirectiveKindMap();
520 AsmParser::~AsmParser() {
521 assert((HadError || ActiveMacros.empty()) &&
522 "Unexpected active macro instantiation!");
525 void AsmParser::printMacroInstantiations() {
526 // Print the active macro instantiation stack.
527 for (std::vector<MacroInstantiation *>::const_reverse_iterator
528 it = ActiveMacros.rbegin(),
529 ie = ActiveMacros.rend();
531 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
532 "while in macro instantiation");
535 void AsmParser::Note(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
536 printMessage(L, SourceMgr::DK_Note, Msg, Ranges);
537 printMacroInstantiations();
540 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
541 if (getTargetParser().getTargetOptions().MCFatalWarnings)
542 return Error(L, Msg, Ranges);
543 printMessage(L, SourceMgr::DK_Warning, Msg, Ranges);
544 printMacroInstantiations();
548 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
550 printMessage(L, SourceMgr::DK_Error, Msg, Ranges);
551 printMacroInstantiations();
555 bool AsmParser::enterIncludeFile(const std::string &Filename) {
556 std::string IncludedFile;
558 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
563 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
567 /// Process the specified .incbin file by searching for it in the include paths
568 /// then just emitting the byte contents of the file to the streamer. This
569 /// returns true on failure.
570 bool AsmParser::processIncbinFile(const std::string &Filename) {
571 std::string IncludedFile;
573 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
577 // Pick up the bytes from the file and emit them.
578 getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer());
582 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) {
583 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc);
584 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(),
588 const AsmToken &AsmParser::Lex() {
589 const AsmToken *tok = &Lexer.Lex();
591 if (tok->is(AsmToken::Eof)) {
592 // If this is the end of an included file, pop the parent file off the
594 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
595 if (ParentIncludeLoc != SMLoc()) {
596 jumpToLoc(ParentIncludeLoc);
601 if (tok->is(AsmToken::Error))
602 Error(Lexer.getErrLoc(), Lexer.getErr());
607 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
608 // Create the initial section, if requested.
609 if (!NoInitialTextSection)
610 Out.InitSections(false);
616 AsmCond StartingCondState = TheCondState;
618 // If we are generating dwarf for assembly source files save the initial text
619 // section and generate a .file directive.
620 if (getContext().getGenDwarfForAssembly()) {
621 MCSymbol *SectionStartSym = getContext().CreateTempSymbol();
622 getStreamer().EmitLabel(SectionStartSym);
623 auto InsertResult = getContext().addGenDwarfSection(
624 getStreamer().getCurrentSection().first);
625 assert(InsertResult.second && ".text section should not have debug info yet");
626 InsertResult.first->second.first = SectionStartSym;
627 getContext().setGenDwarfFileNumber(getStreamer().EmitDwarfFileDirective(
628 0, StringRef(), getContext().getMainFileName()));
631 // While we have input, parse each statement.
632 while (Lexer.isNot(AsmToken::Eof)) {
633 ParseStatementInfo Info;
634 if (!parseStatement(Info, nullptr))
637 // We had an error, validate that one was emitted and recover by skipping to
639 assert(HadError && "Parse statement returned an error, but none emitted!");
640 eatToEndOfStatement();
643 if (TheCondState.TheCond != StartingCondState.TheCond ||
644 TheCondState.Ignore != StartingCondState.Ignore)
645 return TokError("unmatched .ifs or .elses");
647 // Check to see there are no empty DwarfFile slots.
648 const auto &LineTables = getContext().getMCDwarfLineTables();
649 if (!LineTables.empty()) {
651 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
652 if (File.Name.empty() && Index != 0)
653 TokError("unassigned file number: " + Twine(Index) +
654 " for .file directives");
659 // Check to see that all assembler local symbols were actually defined.
660 // Targets that don't do subsections via symbols may not want this, though,
661 // so conservatively exclude them. Only do this if we're finalizing, though,
662 // as otherwise we won't necessarilly have seen everything yet.
663 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) {
664 const MCContext::SymbolTable &Symbols = getContext().getSymbols();
665 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(),
668 MCSymbol *Sym = i->getValue();
669 // Variable symbols may not be marked as defined, so check those
670 // explicitly. If we know it's a variable, we have a definition for
671 // the purposes of this check.
672 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
673 // FIXME: We would really like to refer back to where the symbol was
674 // first referenced for a source location. We need to add something
675 // to track that. Currently, we just point to the end of the file.
677 getLexer().getLoc(), SourceMgr::DK_Error,
678 "assembler local symbol '" + Sym->getName() + "' not defined");
682 // Finalize the output stream if there are no errors and if the client wants
684 if (!HadError && !NoFinalize)
690 void AsmParser::checkForValidSection() {
691 if (!ParsingInlineAsm && !getStreamer().getCurrentSection().first) {
692 TokError("expected section directive before assembly directive");
693 Out.InitSections(false);
697 /// \brief Throw away the rest of the line for testing purposes.
698 void AsmParser::eatToEndOfStatement() {
699 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
703 if (Lexer.is(AsmToken::EndOfStatement))
707 StringRef AsmParser::parseStringToEndOfStatement() {
708 const char *Start = getTok().getLoc().getPointer();
710 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
713 const char *End = getTok().getLoc().getPointer();
714 return StringRef(Start, End - Start);
717 StringRef AsmParser::parseStringToComma() {
718 const char *Start = getTok().getLoc().getPointer();
720 while (Lexer.isNot(AsmToken::EndOfStatement) &&
721 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof))
724 const char *End = getTok().getLoc().getPointer();
725 return StringRef(Start, End - Start);
728 /// \brief Parse a paren expression and return it.
729 /// NOTE: This assumes the leading '(' has already been consumed.
731 /// parenexpr ::= expr)
733 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
734 if (parseExpression(Res))
736 if (Lexer.isNot(AsmToken::RParen))
737 return TokError("expected ')' in parentheses expression");
738 EndLoc = Lexer.getTok().getEndLoc();
743 /// \brief Parse a bracket expression and return it.
744 /// NOTE: This assumes the leading '[' has already been consumed.
746 /// bracketexpr ::= expr]
748 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
749 if (parseExpression(Res))
751 if (Lexer.isNot(AsmToken::RBrac))
752 return TokError("expected ']' in brackets expression");
753 EndLoc = Lexer.getTok().getEndLoc();
758 /// \brief Parse a primary expression and return it.
759 /// primaryexpr ::= (parenexpr
760 /// primaryexpr ::= symbol
761 /// primaryexpr ::= number
762 /// primaryexpr ::= '.'
763 /// primaryexpr ::= ~,+,- primaryexpr
764 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
765 SMLoc FirstTokenLoc = getLexer().getLoc();
766 AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
767 switch (FirstTokenKind) {
769 return TokError("unknown token in expression");
770 // If we have an error assume that we've already handled it.
771 case AsmToken::Error:
773 case AsmToken::Exclaim:
774 Lex(); // Eat the operator.
775 if (parsePrimaryExpr(Res, EndLoc))
777 Res = MCUnaryExpr::CreateLNot(Res, getContext());
779 case AsmToken::Dollar:
781 case AsmToken::String:
782 case AsmToken::Identifier: {
783 StringRef Identifier;
784 if (parseIdentifier(Identifier)) {
785 if (FirstTokenKind == AsmToken::Dollar) {
786 if (Lexer.getMAI().getDollarIsPC()) {
787 // This is a '$' reference, which references the current PC. Emit a
788 // temporary label to the streamer and refer to it.
789 MCSymbol *Sym = Ctx.CreateTempSymbol();
791 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
793 EndLoc = FirstTokenLoc;
796 return Error(FirstTokenLoc, "invalid token in expression");
799 // Parse symbol variant
800 std::pair<StringRef, StringRef> Split;
801 if (!MAI.useParensForSymbolVariant()) {
802 if (FirstTokenKind == AsmToken::String) {
803 if (Lexer.is(AsmToken::At)) {
804 Lexer.Lex(); // eat @
805 SMLoc AtLoc = getLexer().getLoc();
807 if (parseIdentifier(VName))
808 return Error(AtLoc, "expected symbol variant after '@'");
810 Split = std::make_pair(Identifier, VName);
813 Split = Identifier.split('@');
815 } else if (Lexer.is(AsmToken::LParen)) {
816 Lexer.Lex(); // eat (
818 parseIdentifier(VName);
819 if (Lexer.isNot(AsmToken::RParen)) {
820 return Error(Lexer.getTok().getLoc(),
821 "unexpected token in variant, expected ')'");
823 Lexer.Lex(); // eat )
824 Split = std::make_pair(Identifier, VName);
827 EndLoc = SMLoc::getFromPointer(Identifier.end());
829 // This is a symbol reference.
830 StringRef SymbolName = Identifier;
831 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
833 // Lookup the symbol variant if used.
834 if (Split.second.size()) {
835 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
836 if (Variant != MCSymbolRefExpr::VK_Invalid) {
837 SymbolName = Split.first;
838 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
839 Variant = MCSymbolRefExpr::VK_None;
841 return Error(SMLoc::getFromPointer(Split.second.begin()),
842 "invalid variant '" + Split.second + "'");
846 MCSymbol *Sym = getContext().GetOrCreateSymbol(SymbolName);
848 // If this is an absolute variable reference, substitute it now to preserve
849 // semantics in the face of reassignment.
850 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
852 return Error(EndLoc, "unexpected modifier on variable reference");
854 Res = Sym->getVariableValue();
858 // Otherwise create a symbol ref.
859 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
862 case AsmToken::BigNum:
863 return TokError("literal value out of range for directive");
864 case AsmToken::Integer: {
865 SMLoc Loc = getTok().getLoc();
866 int64_t IntVal = getTok().getIntVal();
867 Res = MCConstantExpr::Create(IntVal, getContext());
868 EndLoc = Lexer.getTok().getEndLoc();
870 // Look for 'b' or 'f' following an Integer as a directional label
871 if (Lexer.getKind() == AsmToken::Identifier) {
872 StringRef IDVal = getTok().getString();
873 // Lookup the symbol variant if used.
874 std::pair<StringRef, StringRef> Split = IDVal.split('@');
875 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
876 if (Split.first.size() != IDVal.size()) {
877 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
878 if (Variant == MCSymbolRefExpr::VK_Invalid)
879 return TokError("invalid variant '" + Split.second + "'");
882 if (IDVal == "f" || IDVal == "b") {
884 Ctx.GetDirectionalLocalSymbol(IntVal, IDVal == "b");
885 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
886 if (IDVal == "b" && Sym->isUndefined())
887 return Error(Loc, "invalid reference to undefined symbol");
888 EndLoc = Lexer.getTok().getEndLoc();
889 Lex(); // Eat identifier.
894 case AsmToken::Real: {
895 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
896 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
897 Res = MCConstantExpr::Create(IntVal, getContext());
898 EndLoc = Lexer.getTok().getEndLoc();
902 case AsmToken::Dot: {
903 // This is a '.' reference, which references the current PC. Emit a
904 // temporary label to the streamer and refer to it.
905 MCSymbol *Sym = Ctx.CreateTempSymbol();
907 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
908 EndLoc = Lexer.getTok().getEndLoc();
909 Lex(); // Eat identifier.
912 case AsmToken::LParen:
913 Lex(); // Eat the '('.
914 return parseParenExpr(Res, EndLoc);
915 case AsmToken::LBrac:
916 if (!PlatformParser->HasBracketExpressions())
917 return TokError("brackets expression not supported on this target");
918 Lex(); // Eat the '['.
919 return parseBracketExpr(Res, EndLoc);
920 case AsmToken::Minus:
921 Lex(); // Eat the operator.
922 if (parsePrimaryExpr(Res, EndLoc))
924 Res = MCUnaryExpr::CreateMinus(Res, getContext());
927 Lex(); // Eat the operator.
928 if (parsePrimaryExpr(Res, EndLoc))
930 Res = MCUnaryExpr::CreatePlus(Res, getContext());
932 case AsmToken::Tilde:
933 Lex(); // Eat the operator.
934 if (parsePrimaryExpr(Res, EndLoc))
936 Res = MCUnaryExpr::CreateNot(Res, getContext());
941 bool AsmParser::parseExpression(const MCExpr *&Res) {
943 return parseExpression(Res, EndLoc);
947 AsmParser::applyModifierToExpr(const MCExpr *E,
948 MCSymbolRefExpr::VariantKind Variant) {
949 // Ask the target implementation about this expression first.
950 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx);
953 // Recurse over the given expression, rebuilding it to apply the given variant
954 // if there is exactly one symbol.
955 switch (E->getKind()) {
957 case MCExpr::Constant:
960 case MCExpr::SymbolRef: {
961 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
963 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
964 TokError("invalid variant on expression '" + getTok().getIdentifier() +
965 "' (already modified)");
969 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
972 case MCExpr::Unary: {
973 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
974 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant);
977 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
980 case MCExpr::Binary: {
981 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
982 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant);
983 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant);
993 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
997 llvm_unreachable("Invalid expression kind!");
1000 /// \brief Parse an expression and return it.
1002 /// expr ::= expr &&,|| expr -> lowest.
1003 /// expr ::= expr |,^,&,! expr
1004 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr
1005 /// expr ::= expr <<,>> expr
1006 /// expr ::= expr +,- expr
1007 /// expr ::= expr *,/,% expr -> highest.
1008 /// expr ::= primaryexpr
1010 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1011 // Parse the expression.
1013 if (parsePrimaryExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc))
1016 // As a special case, we support 'a op b @ modifier' by rewriting the
1017 // expression to include the modifier. This is inefficient, but in general we
1018 // expect users to use 'a@modifier op b'.
1019 if (Lexer.getKind() == AsmToken::At) {
1022 if (Lexer.isNot(AsmToken::Identifier))
1023 return TokError("unexpected symbol modifier following '@'");
1025 MCSymbolRefExpr::VariantKind Variant =
1026 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
1027 if (Variant == MCSymbolRefExpr::VK_Invalid)
1028 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
1030 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant);
1032 return TokError("invalid modifier '" + getTok().getIdentifier() +
1033 "' (no symbols present)");
1040 // Try to constant fold it up front, if possible.
1042 if (Res->EvaluateAsAbsolute(Value))
1043 Res = MCConstantExpr::Create(Value, getContext());
1048 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1050 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1053 bool AsmParser::parseAbsoluteExpression(int64_t &Res) {
1056 SMLoc StartLoc = Lexer.getLoc();
1057 if (parseExpression(Expr))
1060 if (!Expr->EvaluateAsAbsolute(Res))
1061 return Error(StartLoc, "expected absolute expression");
1066 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
1067 MCBinaryExpr::Opcode &Kind) {
1070 return 0; // not a binop.
1072 // Lowest Precedence: &&, ||
1073 case AsmToken::AmpAmp:
1074 Kind = MCBinaryExpr::LAnd;
1076 case AsmToken::PipePipe:
1077 Kind = MCBinaryExpr::LOr;
1080 // Low Precedence: |, &, ^
1082 // FIXME: gas seems to support '!' as an infix operator?
1083 case AsmToken::Pipe:
1084 Kind = MCBinaryExpr::Or;
1086 case AsmToken::Caret:
1087 Kind = MCBinaryExpr::Xor;
1090 Kind = MCBinaryExpr::And;
1093 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
1094 case AsmToken::EqualEqual:
1095 Kind = MCBinaryExpr::EQ;
1097 case AsmToken::ExclaimEqual:
1098 case AsmToken::LessGreater:
1099 Kind = MCBinaryExpr::NE;
1101 case AsmToken::Less:
1102 Kind = MCBinaryExpr::LT;
1104 case AsmToken::LessEqual:
1105 Kind = MCBinaryExpr::LTE;
1107 case AsmToken::Greater:
1108 Kind = MCBinaryExpr::GT;
1110 case AsmToken::GreaterEqual:
1111 Kind = MCBinaryExpr::GTE;
1114 // Intermediate Precedence: <<, >>
1115 case AsmToken::LessLess:
1116 Kind = MCBinaryExpr::Shl;
1118 case AsmToken::GreaterGreater:
1119 Kind = MCBinaryExpr::Shr;
1122 // High Intermediate Precedence: +, -
1123 case AsmToken::Plus:
1124 Kind = MCBinaryExpr::Add;
1126 case AsmToken::Minus:
1127 Kind = MCBinaryExpr::Sub;
1130 // Highest Precedence: *, /, %
1131 case AsmToken::Star:
1132 Kind = MCBinaryExpr::Mul;
1134 case AsmToken::Slash:
1135 Kind = MCBinaryExpr::Div;
1137 case AsmToken::Percent:
1138 Kind = MCBinaryExpr::Mod;
1143 /// \brief Parse all binary operators with precedence >= 'Precedence'.
1144 /// Res contains the LHS of the expression on input.
1145 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1148 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1149 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1151 // If the next token is lower precedence than we are allowed to eat, return
1152 // successfully with what we ate already.
1153 if (TokPrec < Precedence)
1158 // Eat the next primary expression.
1160 if (parsePrimaryExpr(RHS, EndLoc))
1163 // If BinOp binds less tightly with RHS than the operator after RHS, let
1164 // the pending operator take RHS as its LHS.
1165 MCBinaryExpr::Opcode Dummy;
1166 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1167 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
1170 // Merge LHS and RHS according to operator.
1171 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
1176 /// ::= EndOfStatement
1177 /// ::= Label* Directive ...Operands... EndOfStatement
1178 /// ::= Label* Identifier OperandList* EndOfStatement
1179 bool AsmParser::parseStatement(ParseStatementInfo &Info,
1180 MCAsmParserSemaCallback *SI) {
1181 if (Lexer.is(AsmToken::EndOfStatement)) {
1187 // Statements always start with an identifier or are a full line comment.
1188 AsmToken ID = getTok();
1189 SMLoc IDLoc = ID.getLoc();
1191 int64_t LocalLabelVal = -1;
1192 // A full line comment is a '#' as the first token.
1193 if (Lexer.is(AsmToken::Hash))
1194 return parseCppHashLineFilenameComment(IDLoc);
1196 // Allow an integer followed by a ':' as a directional local label.
1197 if (Lexer.is(AsmToken::Integer)) {
1198 LocalLabelVal = getTok().getIntVal();
1199 if (LocalLabelVal < 0) {
1200 if (!TheCondState.Ignore)
1201 return TokError("unexpected token at start of statement");
1204 IDVal = getTok().getString();
1205 Lex(); // Consume the integer token to be used as an identifier token.
1206 if (Lexer.getKind() != AsmToken::Colon) {
1207 if (!TheCondState.Ignore)
1208 return TokError("unexpected token at start of statement");
1211 } else if (Lexer.is(AsmToken::Dot)) {
1212 // Treat '.' as a valid identifier in this context.
1215 } else if (parseIdentifier(IDVal)) {
1216 if (!TheCondState.Ignore)
1217 return TokError("unexpected token at start of statement");
1221 // Handle conditional assembly here before checking for skipping. We
1222 // have to do this so that .endif isn't skipped in a ".if 0" block for
1224 StringMap<DirectiveKind>::const_iterator DirKindIt =
1225 DirectiveKindMap.find(IDVal);
1226 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
1228 : DirKindIt->getValue();
1239 return parseDirectiveIf(IDLoc, DirKind);
1241 return parseDirectiveIfb(IDLoc, true);
1243 return parseDirectiveIfb(IDLoc, false);
1245 return parseDirectiveIfc(IDLoc, true);
1247 return parseDirectiveIfeqs(IDLoc, true);
1249 return parseDirectiveIfc(IDLoc, false);
1251 return parseDirectiveIfeqs(IDLoc, false);
1253 return parseDirectiveIfdef(IDLoc, true);
1256 return parseDirectiveIfdef(IDLoc, false);
1258 return parseDirectiveElseIf(IDLoc);
1260 return parseDirectiveElse(IDLoc);
1262 return parseDirectiveEndIf(IDLoc);
1265 // Ignore the statement if in the middle of inactive conditional
1267 if (TheCondState.Ignore) {
1268 eatToEndOfStatement();
1272 // FIXME: Recurse on local labels?
1274 // See what kind of statement we have.
1275 switch (Lexer.getKind()) {
1276 case AsmToken::Colon: {
1277 checkForValidSection();
1279 // identifier ':' -> Label.
1282 // Diagnose attempt to use '.' as a label.
1284 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1286 // Diagnose attempt to use a variable as a label.
1288 // FIXME: Diagnostics. Note the location of the definition as a label.
1289 // FIXME: This doesn't diagnose assignment to a symbol which has been
1290 // implicitly marked as external.
1292 if (LocalLabelVal == -1) {
1293 if (ParsingInlineAsm && SI) {
1294 StringRef RewrittenLabel = SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true);
1295 assert(RewrittenLabel.size() && "We should have an internal name here.");
1296 Info.AsmRewrites->push_back(AsmRewrite(AOK_Label, IDLoc,
1297 IDVal.size(), RewrittenLabel));
1298 IDVal = RewrittenLabel;
1300 Sym = getContext().GetOrCreateSymbol(IDVal);
1302 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
1304 Sym->redefineIfPossible();
1306 if (!Sym->isUndefined() || Sym->isVariable())
1307 return Error(IDLoc, "invalid symbol redefinition");
1310 if (!ParsingInlineAsm)
1313 // If we are generating dwarf for assembly source files then gather the
1314 // info to make a dwarf label entry for this label if needed.
1315 if (getContext().getGenDwarfForAssembly())
1316 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1319 getTargetParser().onLabelParsed(Sym);
1321 // Consume any end of statement token, if present, to avoid spurious
1322 // AddBlankLine calls().
1323 if (Lexer.is(AsmToken::EndOfStatement)) {
1325 if (Lexer.is(AsmToken::Eof))
1332 case AsmToken::Equal:
1333 // identifier '=' ... -> assignment statement
1336 return parseAssignment(IDVal, true);
1338 default: // Normal instruction or directive.
1342 // If macros are enabled, check to see if this is a macro instantiation.
1343 if (areMacrosEnabled())
1344 if (const MCAsmMacro *M = lookupMacro(IDVal)) {
1345 return handleMacroEntry(M, IDLoc);
1348 // Otherwise, we have a normal instruction or directive.
1350 // Directives start with "."
1351 if (IDVal[0] == '.' && IDVal != ".") {
1352 // There are several entities interested in parsing directives:
1354 // 1. The target-specific assembly parser. Some directives are target
1355 // specific or may potentially behave differently on certain targets.
1356 // 2. Asm parser extensions. For example, platform-specific parsers
1357 // (like the ELF parser) register themselves as extensions.
1358 // 3. The generic directive parser implemented by this class. These are
1359 // all the directives that behave in a target and platform independent
1360 // manner, or at least have a default behavior that's shared between
1361 // all targets and platforms.
1363 // First query the target-specific parser. It will return 'true' if it
1364 // isn't interested in this directive.
1365 if (!getTargetParser().ParseDirective(ID))
1368 // Next, check the extension directive map to see if any extension has
1369 // registered itself to parse this directive.
1370 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
1371 ExtensionDirectiveMap.lookup(IDVal);
1373 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1375 // Finally, if no one else is interested in this directive, it must be
1376 // generic and familiar to this class.
1382 return parseDirectiveSet(IDVal, true);
1384 return parseDirectiveSet(IDVal, false);
1386 return parseDirectiveAscii(IDVal, false);
1389 return parseDirectiveAscii(IDVal, true);
1391 return parseDirectiveValue(1);
1395 return parseDirectiveValue(2);
1399 return parseDirectiveValue(4);
1402 return parseDirectiveValue(8);
1404 return parseDirectiveOctaValue();
1407 return parseDirectiveRealValue(APFloat::IEEEsingle);
1409 return parseDirectiveRealValue(APFloat::IEEEdouble);
1411 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1412 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1415 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes();
1416 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1419 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1421 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1423 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1425 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1427 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1429 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1431 return parseDirectiveOrg();
1433 return parseDirectiveFill();
1435 return parseDirectiveZero();
1437 eatToEndOfStatement(); // .extern is the default, ignore it.
1441 return parseDirectiveSymbolAttribute(MCSA_Global);
1442 case DK_LAZY_REFERENCE:
1443 return parseDirectiveSymbolAttribute(MCSA_LazyReference);
1444 case DK_NO_DEAD_STRIP:
1445 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1446 case DK_SYMBOL_RESOLVER:
1447 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1448 case DK_PRIVATE_EXTERN:
1449 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1451 return parseDirectiveSymbolAttribute(MCSA_Reference);
1452 case DK_WEAK_DEFINITION:
1453 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1454 case DK_WEAK_REFERENCE:
1455 return parseDirectiveSymbolAttribute(MCSA_WeakReference);
1456 case DK_WEAK_DEF_CAN_BE_HIDDEN:
1457 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1460 return parseDirectiveComm(/*IsLocal=*/false);
1462 return parseDirectiveComm(/*IsLocal=*/true);
1464 return parseDirectiveAbort();
1466 return parseDirectiveInclude();
1468 return parseDirectiveIncbin();
1471 return TokError(Twine(IDVal) + " not supported yet");
1473 return parseDirectiveRept(IDLoc, IDVal);
1475 return parseDirectiveIrp(IDLoc);
1477 return parseDirectiveIrpc(IDLoc);
1479 return parseDirectiveEndr(IDLoc);
1480 case DK_BUNDLE_ALIGN_MODE:
1481 return parseDirectiveBundleAlignMode();
1482 case DK_BUNDLE_LOCK:
1483 return parseDirectiveBundleLock();
1484 case DK_BUNDLE_UNLOCK:
1485 return parseDirectiveBundleUnlock();
1487 return parseDirectiveLEB128(true);
1489 return parseDirectiveLEB128(false);
1492 return parseDirectiveSpace(IDVal);
1494 return parseDirectiveFile(IDLoc);
1496 return parseDirectiveLine();
1498 return parseDirectiveLoc();
1500 return parseDirectiveStabs();
1501 case DK_CFI_SECTIONS:
1502 return parseDirectiveCFISections();
1503 case DK_CFI_STARTPROC:
1504 return parseDirectiveCFIStartProc();
1505 case DK_CFI_ENDPROC:
1506 return parseDirectiveCFIEndProc();
1507 case DK_CFI_DEF_CFA:
1508 return parseDirectiveCFIDefCfa(IDLoc);
1509 case DK_CFI_DEF_CFA_OFFSET:
1510 return parseDirectiveCFIDefCfaOffset();
1511 case DK_CFI_ADJUST_CFA_OFFSET:
1512 return parseDirectiveCFIAdjustCfaOffset();
1513 case DK_CFI_DEF_CFA_REGISTER:
1514 return parseDirectiveCFIDefCfaRegister(IDLoc);
1516 return parseDirectiveCFIOffset(IDLoc);
1517 case DK_CFI_REL_OFFSET:
1518 return parseDirectiveCFIRelOffset(IDLoc);
1519 case DK_CFI_PERSONALITY:
1520 return parseDirectiveCFIPersonalityOrLsda(true);
1522 return parseDirectiveCFIPersonalityOrLsda(false);
1523 case DK_CFI_REMEMBER_STATE:
1524 return parseDirectiveCFIRememberState();
1525 case DK_CFI_RESTORE_STATE:
1526 return parseDirectiveCFIRestoreState();
1527 case DK_CFI_SAME_VALUE:
1528 return parseDirectiveCFISameValue(IDLoc);
1529 case DK_CFI_RESTORE:
1530 return parseDirectiveCFIRestore(IDLoc);
1532 return parseDirectiveCFIEscape();
1533 case DK_CFI_SIGNAL_FRAME:
1534 return parseDirectiveCFISignalFrame();
1535 case DK_CFI_UNDEFINED:
1536 return parseDirectiveCFIUndefined(IDLoc);
1537 case DK_CFI_REGISTER:
1538 return parseDirectiveCFIRegister(IDLoc);
1539 case DK_CFI_WINDOW_SAVE:
1540 return parseDirectiveCFIWindowSave();
1543 return parseDirectiveMacrosOnOff(IDVal);
1545 return parseDirectiveMacro(IDLoc);
1547 return parseDirectiveExitMacro(IDVal);
1550 return parseDirectiveEndMacro(IDVal);
1552 return parseDirectivePurgeMacro(IDLoc);
1554 return parseDirectiveEnd(IDLoc);
1556 return parseDirectiveError(IDLoc, false);
1558 return parseDirectiveError(IDLoc, true);
1560 return parseDirectiveWarning(IDLoc);
1563 return Error(IDLoc, "unknown directive");
1566 // __asm _emit or __asm __emit
1567 if (ParsingInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
1568 IDVal == "_EMIT" || IDVal == "__EMIT"))
1569 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
1572 if (ParsingInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
1573 return parseDirectiveMSAlign(IDLoc, Info);
1575 checkForValidSection();
1577 // Canonicalize the opcode to lower case.
1578 std::string OpcodeStr = IDVal.lower();
1579 ParseInstructionInfo IInfo(Info.AsmRewrites);
1580 bool HadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, IDLoc,
1581 Info.ParsedOperands);
1582 Info.ParseError = HadError;
1584 // Dump the parsed representation, if requested.
1585 if (getShowParsedOperands()) {
1586 SmallString<256> Str;
1587 raw_svector_ostream OS(Str);
1588 OS << "parsed instruction: [";
1589 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
1592 Info.ParsedOperands[i]->print(OS);
1596 printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1599 // If we are generating dwarf for the current section then generate a .loc
1600 // directive for the instruction.
1601 if (!HadError && getContext().getGenDwarfForAssembly() &&
1602 getContext().getGenDwarfSectionSyms().count(
1603 getStreamer().getCurrentSection().first)) {
1605 if (ActiveMacros.empty())
1606 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
1608 Line = SrcMgr.FindLineNumber(ActiveMacros.back()->InstantiationLoc,
1609 ActiveMacros.back()->ExitBuffer);
1611 // If we previously parsed a cpp hash file line comment then make sure the
1612 // current Dwarf File is for the CppHashFilename if not then emit the
1613 // Dwarf File table for it and adjust the line number for the .loc.
1614 if (CppHashFilename.size()) {
1615 unsigned FileNumber = getStreamer().EmitDwarfFileDirective(
1616 0, StringRef(), CppHashFilename);
1617 getContext().setGenDwarfFileNumber(FileNumber);
1619 // Since SrcMgr.FindLineNumber() is slow and messes up the SourceMgr's
1620 // cache with the different Loc from the call above we save the last
1621 // info we queried here with SrcMgr.FindLineNumber().
1622 unsigned CppHashLocLineNo;
1623 if (LastQueryIDLoc == CppHashLoc && LastQueryBuffer == CppHashBuf)
1624 CppHashLocLineNo = LastQueryLine;
1626 CppHashLocLineNo = SrcMgr.FindLineNumber(CppHashLoc, CppHashBuf);
1627 LastQueryLine = CppHashLocLineNo;
1628 LastQueryIDLoc = CppHashLoc;
1629 LastQueryBuffer = CppHashBuf;
1631 Line = CppHashLineNumber - 1 + (Line - CppHashLocLineNo);
1634 getStreamer().EmitDwarfLocDirective(
1635 getContext().getGenDwarfFileNumber(), Line, 0,
1636 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0,
1640 // If parsing succeeded, match the instruction.
1643 getTargetParser().MatchAndEmitInstruction(IDLoc, Info.Opcode,
1644 Info.ParsedOperands, Out,
1645 ErrorInfo, ParsingInlineAsm);
1648 // Don't skip the rest of the line, the instruction parser is responsible for
1653 /// eatToEndOfLine uses the Lexer to eat the characters to the end of the line
1654 /// since they may not be able to be tokenized to get to the end of line token.
1655 void AsmParser::eatToEndOfLine() {
1656 if (!Lexer.is(AsmToken::EndOfStatement))
1657 Lexer.LexUntilEndOfLine();
1662 /// parseCppHashLineFilenameComment as this:
1663 /// ::= # number "filename"
1664 /// or just as a full line comment if it doesn't have a number and a string.
1665 bool AsmParser::parseCppHashLineFilenameComment(const SMLoc &L) {
1666 Lex(); // Eat the hash token.
1668 if (getLexer().isNot(AsmToken::Integer)) {
1669 // Consume the line since in cases it is not a well-formed line directive,
1670 // as if were simply a full line comment.
1675 int64_t LineNumber = getTok().getIntVal();
1678 if (getLexer().isNot(AsmToken::String)) {
1683 StringRef Filename = getTok().getString();
1684 // Get rid of the enclosing quotes.
1685 Filename = Filename.substr(1, Filename.size() - 2);
1687 // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1689 CppHashFilename = Filename;
1690 CppHashLineNumber = LineNumber;
1691 CppHashBuf = CurBuffer;
1693 // Ignore any trailing characters, they're just comment.
1698 /// \brief will use the last parsed cpp hash line filename comment
1699 /// for the Filename and LineNo if any in the diagnostic.
1700 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1701 const AsmParser *Parser = static_cast<const AsmParser *>(Context);
1702 raw_ostream &OS = errs();
1704 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
1705 const SMLoc &DiagLoc = Diag.getLoc();
1706 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1707 unsigned CppHashBuf =
1708 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc);
1710 // Like SourceMgr::printMessage() we need to print the include stack if any
1711 // before printing the message.
1712 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1713 if (!Parser->SavedDiagHandler && DiagCurBuffer &&
1714 DiagCurBuffer != DiagSrcMgr.getMainFileID()) {
1715 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
1716 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
1719 // If we have not parsed a cpp hash line filename comment or the source
1720 // manager changed or buffer changed (like in a nested include) then just
1721 // print the normal diagnostic using its Filename and LineNo.
1722 if (!Parser->CppHashLineNumber || &DiagSrcMgr != &Parser->SrcMgr ||
1723 DiagBuf != CppHashBuf) {
1724 if (Parser->SavedDiagHandler)
1725 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
1727 Diag.print(nullptr, OS);
1731 // Use the CppHashFilename and calculate a line number based on the
1732 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for
1734 const std::string &Filename = Parser->CppHashFilename;
1736 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
1737 int CppHashLocLineNo =
1738 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf);
1740 Parser->CppHashLineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
1742 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
1743 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
1744 Diag.getLineContents(), Diag.getRanges());
1746 if (Parser->SavedDiagHandler)
1747 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
1749 NewDiag.print(nullptr, OS);
1752 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
1753 // difference being that that function accepts '@' as part of identifiers and
1754 // we can't do that. AsmLexer.cpp should probably be changed to handle
1755 // '@' as a special case when needed.
1756 static bool isIdentifierChar(char c) {
1757 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' ||
1761 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
1762 ArrayRef<MCAsmMacroParameter> Parameters,
1763 ArrayRef<MCAsmMacroArgument> A, const SMLoc &L) {
1764 unsigned NParameters = Parameters.size();
1765 bool HasVararg = NParameters ? Parameters.back().Vararg : false;
1766 if ((!IsDarwin || NParameters != 0) && NParameters != A.size())
1767 return Error(L, "Wrong number of arguments");
1769 // A macro without parameters is handled differently on Darwin:
1770 // gas accepts no arguments and does no substitutions
1771 while (!Body.empty()) {
1772 // Scan for the next substitution.
1773 std::size_t End = Body.size(), Pos = 0;
1774 for (; Pos != End; ++Pos) {
1775 // Check for a substitution or escape.
1776 if (IsDarwin && !NParameters) {
1777 // This macro has no parameters, look for $0, $1, etc.
1778 if (Body[Pos] != '$' || Pos + 1 == End)
1781 char Next = Body[Pos + 1];
1782 if (Next == '$' || Next == 'n' ||
1783 isdigit(static_cast<unsigned char>(Next)))
1786 // This macro has parameters, look for \foo, \bar, etc.
1787 if (Body[Pos] == '\\' && Pos + 1 != End)
1793 OS << Body.slice(0, Pos);
1795 // Check if we reached the end.
1799 if (IsDarwin && !NParameters) {
1800 switch (Body[Pos + 1]) {
1806 // $n => number of arguments
1811 // $[0-9] => argument
1813 // Missing arguments are ignored.
1814 unsigned Index = Body[Pos + 1] - '0';
1815 if (Index >= A.size())
1818 // Otherwise substitute with the token values, with spaces eliminated.
1819 for (MCAsmMacroArgument::const_iterator it = A[Index].begin(),
1820 ie = A[Index].end();
1822 OS << it->getString();
1828 unsigned I = Pos + 1;
1829 while (isIdentifierChar(Body[I]) && I + 1 != End)
1832 const char *Begin = Body.data() + Pos + 1;
1833 StringRef Argument(Begin, I - (Pos + 1));
1835 for (; Index < NParameters; ++Index)
1836 if (Parameters[Index].Name == Argument)
1839 if (Index == NParameters) {
1840 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
1843 OS << '\\' << Argument;
1847 bool VarargParameter = HasVararg && Index == (NParameters - 1);
1848 for (MCAsmMacroArgument::const_iterator it = A[Index].begin(),
1849 ie = A[Index].end();
1851 // We expect no quotes around the string's contents when
1852 // parsing for varargs.
1853 if (it->getKind() != AsmToken::String || VarargParameter)
1854 OS << it->getString();
1856 OS << it->getStringContents();
1858 Pos += 1 + Argument.size();
1861 // Update the scan point.
1862 Body = Body.substr(Pos);
1868 MacroInstantiation::MacroInstantiation(SMLoc IL, int EB, SMLoc EL,
1869 size_t CondStackDepth)
1870 : InstantiationLoc(IL), ExitBuffer(EB), ExitLoc(EL),
1871 CondStackDepth(CondStackDepth) {}
1873 static bool isOperator(AsmToken::TokenKind kind) {
1877 case AsmToken::Plus:
1878 case AsmToken::Minus:
1879 case AsmToken::Tilde:
1880 case AsmToken::Slash:
1881 case AsmToken::Star:
1883 case AsmToken::Equal:
1884 case AsmToken::EqualEqual:
1885 case AsmToken::Pipe:
1886 case AsmToken::PipePipe:
1887 case AsmToken::Caret:
1889 case AsmToken::AmpAmp:
1890 case AsmToken::Exclaim:
1891 case AsmToken::ExclaimEqual:
1892 case AsmToken::Percent:
1893 case AsmToken::Less:
1894 case AsmToken::LessEqual:
1895 case AsmToken::LessLess:
1896 case AsmToken::LessGreater:
1897 case AsmToken::Greater:
1898 case AsmToken::GreaterEqual:
1899 case AsmToken::GreaterGreater:
1905 class AsmLexerSkipSpaceRAII {
1907 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
1908 Lexer.setSkipSpace(SkipSpace);
1911 ~AsmLexerSkipSpaceRAII() {
1912 Lexer.setSkipSpace(true);
1920 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) {
1923 if (Lexer.isNot(AsmToken::EndOfStatement)) {
1924 StringRef Str = parseStringToEndOfStatement();
1925 MA.push_back(AsmToken(AsmToken::String, Str));
1930 unsigned ParenLevel = 0;
1931 unsigned AddTokens = 0;
1933 // Darwin doesn't use spaces to delmit arguments.
1934 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
1937 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
1938 return TokError("unexpected token in macro instantiation");
1940 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma))
1943 if (Lexer.is(AsmToken::Space)) {
1944 Lex(); // Eat spaces
1946 // Spaces can delimit parameters, but could also be part an expression.
1947 // If the token after a space is an operator, add the token and the next
1948 // one into this argument
1950 if (isOperator(Lexer.getKind())) {
1951 // Check to see whether the token is used as an operator,
1952 // or part of an identifier
1953 const char *NextChar = getTok().getEndLoc().getPointer();
1954 if (*NextChar == ' ')
1958 if (!AddTokens && ParenLevel == 0) {
1964 // handleMacroEntry relies on not advancing the lexer here
1965 // to be able to fill in the remaining default parameter values
1966 if (Lexer.is(AsmToken::EndOfStatement))
1969 // Adjust the current parentheses level.
1970 if (Lexer.is(AsmToken::LParen))
1972 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1975 // Append the token to the current argument list.
1976 MA.push_back(getTok());
1982 if (ParenLevel != 0)
1983 return TokError("unbalanced parentheses in macro argument");
1987 // Parse the macro instantiation arguments.
1988 bool AsmParser::parseMacroArguments(const MCAsmMacro *M,
1989 MCAsmMacroArguments &A) {
1990 const unsigned NParameters = M ? M->Parameters.size() : 0;
1991 bool NamedParametersFound = false;
1992 SmallVector<SMLoc, 4> FALocs;
1994 A.resize(NParameters);
1995 FALocs.resize(NParameters);
1997 // Parse two kinds of macro invocations:
1998 // - macros defined without any parameters accept an arbitrary number of them
1999 // - macros defined with parameters accept at most that many of them
2000 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false;
2001 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
2003 SMLoc IDLoc = Lexer.getLoc();
2004 MCAsmMacroParameter FA;
2006 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) {
2007 if (parseIdentifier(FA.Name)) {
2008 Error(IDLoc, "invalid argument identifier for formal argument");
2009 eatToEndOfStatement();
2013 if (!Lexer.is(AsmToken::Equal)) {
2014 TokError("expected '=' after formal parameter identifier");
2015 eatToEndOfStatement();
2020 NamedParametersFound = true;
2023 if (NamedParametersFound && FA.Name.empty()) {
2024 Error(IDLoc, "cannot mix positional and keyword arguments");
2025 eatToEndOfStatement();
2029 bool Vararg = HasVararg && Parameter == (NParameters - 1);
2030 if (parseMacroArgument(FA.Value, Vararg))
2033 unsigned PI = Parameter;
2034 if (!FA.Name.empty()) {
2036 for (FAI = 0; FAI < NParameters; ++FAI)
2037 if (M->Parameters[FAI].Name == FA.Name)
2040 if (FAI >= NParameters) {
2041 assert(M && "expected macro to be defined");
2043 "parameter named '" + FA.Name + "' does not exist for macro '" +
2050 if (!FA.Value.empty()) {
2055 if (FALocs.size() <= PI)
2056 FALocs.resize(PI + 1);
2058 FALocs[PI] = Lexer.getLoc();
2061 // At the end of the statement, fill in remaining arguments that have
2062 // default values. If there aren't any, then the next argument is
2063 // required but missing
2064 if (Lexer.is(AsmToken::EndOfStatement)) {
2065 bool Failure = false;
2066 for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
2067 if (A[FAI].empty()) {
2068 if (M->Parameters[FAI].Required) {
2069 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
2070 "missing value for required parameter "
2071 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
2075 if (!M->Parameters[FAI].Value.empty())
2076 A[FAI] = M->Parameters[FAI].Value;
2082 if (Lexer.is(AsmToken::Comma))
2086 return TokError("too many positional arguments");
2089 const MCAsmMacro *AsmParser::lookupMacro(StringRef Name) {
2090 StringMap<MCAsmMacro>::iterator I = MacroMap.find(Name);
2091 return (I == MacroMap.end()) ? nullptr : &I->getValue();
2094 void AsmParser::defineMacro(StringRef Name, MCAsmMacro Macro) {
2095 MacroMap.insert(std::make_pair(Name, std::move(Macro)));
2098 void AsmParser::undefineMacro(StringRef Name) { MacroMap.erase(Name); }
2100 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) {
2101 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
2102 // this, although we should protect against infinite loops.
2103 if (ActiveMacros.size() == 20)
2104 return TokError("macros cannot be nested more than 20 levels deep");
2106 MCAsmMacroArguments A;
2107 if (parseMacroArguments(M, A))
2110 // Macro instantiation is lexical, unfortunately. We construct a new buffer
2111 // to hold the macro body with substitutions.
2112 SmallString<256> Buf;
2113 StringRef Body = M->Body;
2114 raw_svector_ostream OS(Buf);
2116 if (expandMacro(OS, Body, M->Parameters, A, getTok().getLoc()))
2119 // We include the .endmacro in the buffer as our cue to exit the macro
2121 OS << ".endmacro\n";
2123 std::unique_ptr<MemoryBuffer> Instantiation =
2124 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
2126 // Create the macro instantiation object and add to the current macro
2127 // instantiation stack.
2128 MacroInstantiation *MI = new MacroInstantiation(
2129 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size());
2130 ActiveMacros.push_back(MI);
2132 // Jump to the macro instantiation and prime the lexer.
2133 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
2134 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
2140 void AsmParser::handleMacroExit() {
2141 // Jump to the EndOfStatement we should return to, and consume it.
2142 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer);
2145 // Pop the instantiation entry.
2146 delete ActiveMacros.back();
2147 ActiveMacros.pop_back();
2150 static bool isUsedIn(const MCSymbol *Sym, const MCExpr *Value) {
2151 switch (Value->getKind()) {
2152 case MCExpr::Binary: {
2153 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value);
2154 return isUsedIn(Sym, BE->getLHS()) || isUsedIn(Sym, BE->getRHS());
2156 case MCExpr::Target:
2157 case MCExpr::Constant:
2159 case MCExpr::SymbolRef: {
2161 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol();
2163 return isUsedIn(Sym, S.getVariableValue());
2167 return isUsedIn(Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr());
2170 llvm_unreachable("Unknown expr kind!");
2173 bool AsmParser::parseAssignment(StringRef Name, bool allow_redef,
2175 // FIXME: Use better location, we should use proper tokens.
2176 SMLoc EqualLoc = Lexer.getLoc();
2178 const MCExpr *Value;
2179 if (parseExpression(Value))
2182 // Note: we don't count b as used in "a = b". This is to allow
2186 if (Lexer.isNot(AsmToken::EndOfStatement))
2187 return TokError("unexpected token in assignment");
2189 // Eat the end of statement marker.
2192 // Validate that the LHS is allowed to be a variable (either it has not been
2193 // used as a symbol, or it is an absolute symbol).
2194 MCSymbol *Sym = getContext().LookupSymbol(Name);
2196 // Diagnose assignment to a label.
2198 // FIXME: Diagnostics. Note the location of the definition as a label.
2199 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
2200 if (isUsedIn(Sym, Value))
2201 return Error(EqualLoc, "Recursive use of '" + Name + "'");
2202 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
2203 ; // Allow redefinitions of undefined symbols only used in directives.
2204 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
2205 ; // Allow redefinitions of variables that haven't yet been used.
2206 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
2207 return Error(EqualLoc, "redefinition of '" + Name + "'");
2208 else if (!Sym->isVariable())
2209 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
2210 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
2211 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
2214 // Don't count these checks as uses.
2215 Sym->setUsed(false);
2216 } else if (Name == ".") {
2217 if (Out.EmitValueToOffset(Value, 0)) {
2218 Error(EqualLoc, "expected absolute expression");
2219 eatToEndOfStatement();
2223 Sym = getContext().GetOrCreateSymbol(Name);
2225 Sym->setRedefinable(allow_redef);
2227 // Do the assignment.
2228 Out.EmitAssignment(Sym, Value);
2230 Out.EmitSymbolAttribute(Sym, MCSA_NoDeadStrip);
2235 /// parseIdentifier:
2238 bool AsmParser::parseIdentifier(StringRef &Res) {
2239 // The assembler has relaxed rules for accepting identifiers, in particular we
2240 // allow things like '.globl $foo' and '.def @feat.00', which would normally be
2241 // separate tokens. At this level, we have already lexed so we cannot (currently)
2242 // handle this as a context dependent token, instead we detect adjacent tokens
2243 // and return the combined identifier.
2244 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
2245 SMLoc PrefixLoc = getLexer().getLoc();
2247 // Consume the prefix character, and check for a following identifier.
2249 if (Lexer.isNot(AsmToken::Identifier))
2252 // We have a '$' or '@' followed by an identifier, make sure they are adjacent.
2253 if (PrefixLoc.getPointer() + 1 != getTok().getLoc().getPointer())
2256 // Construct the joined identifier and consume the token.
2258 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1);
2263 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
2266 Res = getTok().getIdentifier();
2268 Lex(); // Consume the identifier token.
2273 /// parseDirectiveSet:
2274 /// ::= .equ identifier ',' expression
2275 /// ::= .equiv identifier ',' expression
2276 /// ::= .set identifier ',' expression
2277 bool AsmParser::parseDirectiveSet(StringRef IDVal, bool allow_redef) {
2280 if (parseIdentifier(Name))
2281 return TokError("expected identifier after '" + Twine(IDVal) + "'");
2283 if (getLexer().isNot(AsmToken::Comma))
2284 return TokError("unexpected token in '" + Twine(IDVal) + "'");
2287 return parseAssignment(Name, allow_redef, true);
2290 bool AsmParser::parseEscapedString(std::string &Data) {
2291 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
2294 StringRef Str = getTok().getStringContents();
2295 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
2296 if (Str[i] != '\\') {
2301 // Recognize escaped characters. Note that this escape semantics currently
2302 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
2305 return TokError("unexpected backslash at end of string");
2307 // Recognize octal sequences.
2308 if ((unsigned)(Str[i] - '0') <= 7) {
2309 // Consume up to three octal characters.
2310 unsigned Value = Str[i] - '0';
2312 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2314 Value = Value * 8 + (Str[i] - '0');
2316 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) {
2318 Value = Value * 8 + (Str[i] - '0');
2323 return TokError("invalid octal escape sequence (out of range)");
2325 Data += (unsigned char)Value;
2329 // Otherwise recognize individual escapes.
2332 // Just reject invalid escape sequences for now.
2333 return TokError("invalid escape sequence (unrecognized character)");
2335 case 'b': Data += '\b'; break;
2336 case 'f': Data += '\f'; break;
2337 case 'n': Data += '\n'; break;
2338 case 'r': Data += '\r'; break;
2339 case 't': Data += '\t'; break;
2340 case '"': Data += '"'; break;
2341 case '\\': Data += '\\'; break;
2348 /// parseDirectiveAscii:
2349 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
2350 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
2351 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2352 checkForValidSection();
2355 if (getLexer().isNot(AsmToken::String))
2356 return TokError("expected string in '" + Twine(IDVal) + "' directive");
2359 if (parseEscapedString(Data))
2362 getStreamer().EmitBytes(Data);
2364 getStreamer().EmitBytes(StringRef("\0", 1));
2368 if (getLexer().is(AsmToken::EndOfStatement))
2371 if (getLexer().isNot(AsmToken::Comma))
2372 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
2381 /// parseDirectiveValue
2382 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
2383 bool AsmParser::parseDirectiveValue(unsigned Size) {
2384 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2385 checkForValidSection();
2388 const MCExpr *Value;
2389 SMLoc ExprLoc = getLexer().getLoc();
2390 if (parseExpression(Value))
2393 // Special case constant expressions to match code generator.
2394 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2395 assert(Size <= 8 && "Invalid size");
2396 uint64_t IntValue = MCE->getValue();
2397 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
2398 return Error(ExprLoc, "literal value out of range for directive");
2399 getStreamer().EmitIntValue(IntValue, Size);
2401 getStreamer().EmitValue(Value, Size, ExprLoc);
2403 if (getLexer().is(AsmToken::EndOfStatement))
2406 // FIXME: Improve diagnostic.
2407 if (getLexer().isNot(AsmToken::Comma))
2408 return TokError("unexpected token in directive");
2417 /// ParseDirectiveOctaValue
2418 /// ::= .octa [ hexconstant (, hexconstant)* ]
2419 bool AsmParser::parseDirectiveOctaValue() {
2420 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2421 checkForValidSection();
2424 if (Lexer.getKind() == AsmToken::Error)
2426 if (Lexer.getKind() != AsmToken::Integer &&
2427 Lexer.getKind() != AsmToken::BigNum)
2428 return TokError("unknown token in expression");
2430 SMLoc ExprLoc = getLexer().getLoc();
2431 APInt IntValue = getTok().getAPIntVal();
2435 if (IntValue.isIntN(64)) {
2437 lo = IntValue.getZExtValue();
2438 } else if (IntValue.isIntN(128)) {
2439 // It might actually have more than 128 bits, but the top ones are zero.
2440 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
2441 lo = IntValue.getLoBits(64).getZExtValue();
2443 return Error(ExprLoc, "literal value out of range for directive");
2445 if (MAI.isLittleEndian()) {
2446 getStreamer().EmitIntValue(lo, 8);
2447 getStreamer().EmitIntValue(hi, 8);
2449 getStreamer().EmitIntValue(hi, 8);
2450 getStreamer().EmitIntValue(lo, 8);
2453 if (getLexer().is(AsmToken::EndOfStatement))
2456 // FIXME: Improve diagnostic.
2457 if (getLexer().isNot(AsmToken::Comma))
2458 return TokError("unexpected token in directive");
2467 /// parseDirectiveRealValue
2468 /// ::= (.single | .double) [ expression (, expression)* ]
2469 bool AsmParser::parseDirectiveRealValue(const fltSemantics &Semantics) {
2470 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2471 checkForValidSection();
2474 // We don't truly support arithmetic on floating point expressions, so we
2475 // have to manually parse unary prefixes.
2477 if (getLexer().is(AsmToken::Minus)) {
2480 } else if (getLexer().is(AsmToken::Plus))
2483 if (getLexer().isNot(AsmToken::Integer) &&
2484 getLexer().isNot(AsmToken::Real) &&
2485 getLexer().isNot(AsmToken::Identifier))
2486 return TokError("unexpected token in directive");
2488 // Convert to an APFloat.
2489 APFloat Value(Semantics);
2490 StringRef IDVal = getTok().getString();
2491 if (getLexer().is(AsmToken::Identifier)) {
2492 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
2493 Value = APFloat::getInf(Semantics);
2494 else if (!IDVal.compare_lower("nan"))
2495 Value = APFloat::getNaN(Semantics, false, ~0);
2497 return TokError("invalid floating point literal");
2498 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
2499 APFloat::opInvalidOp)
2500 return TokError("invalid floating point literal");
2504 // Consume the numeric token.
2507 // Emit the value as an integer.
2508 APInt AsInt = Value.bitcastToAPInt();
2509 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
2510 AsInt.getBitWidth() / 8);
2512 if (getLexer().is(AsmToken::EndOfStatement))
2515 if (getLexer().isNot(AsmToken::Comma))
2516 return TokError("unexpected token in directive");
2525 /// parseDirectiveZero
2526 /// ::= .zero expression
2527 bool AsmParser::parseDirectiveZero() {
2528 checkForValidSection();
2531 if (parseAbsoluteExpression(NumBytes))
2535 if (getLexer().is(AsmToken::Comma)) {
2537 if (parseAbsoluteExpression(Val))
2541 if (getLexer().isNot(AsmToken::EndOfStatement))
2542 return TokError("unexpected token in '.zero' directive");
2546 getStreamer().EmitFill(NumBytes, Val);
2551 /// parseDirectiveFill
2552 /// ::= .fill expression [ , expression [ , expression ] ]
2553 bool AsmParser::parseDirectiveFill() {
2554 checkForValidSection();
2556 SMLoc RepeatLoc = getLexer().getLoc();
2558 if (parseAbsoluteExpression(NumValues))
2561 if (NumValues < 0) {
2563 "'.fill' directive with negative repeat count has no effect");
2567 int64_t FillSize = 1;
2568 int64_t FillExpr = 0;
2570 SMLoc SizeLoc, ExprLoc;
2571 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2572 if (getLexer().isNot(AsmToken::Comma))
2573 return TokError("unexpected token in '.fill' directive");
2576 SizeLoc = getLexer().getLoc();
2577 if (parseAbsoluteExpression(FillSize))
2580 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2581 if (getLexer().isNot(AsmToken::Comma))
2582 return TokError("unexpected token in '.fill' directive");
2585 ExprLoc = getLexer().getLoc();
2586 if (parseAbsoluteExpression(FillExpr))
2589 if (getLexer().isNot(AsmToken::EndOfStatement))
2590 return TokError("unexpected token in '.fill' directive");
2597 Warning(SizeLoc, "'.fill' directive with negative size has no effect");
2601 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8");
2605 if (!isUInt<32>(FillExpr) && FillSize > 4)
2606 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits");
2608 if (NumValues > 0) {
2609 int64_t NonZeroFillSize = FillSize > 4 ? 4 : FillSize;
2610 FillExpr &= ~0ULL >> (64 - NonZeroFillSize * 8);
2611 for (uint64_t i = 0, e = NumValues; i != e; ++i) {
2612 getStreamer().EmitIntValue(FillExpr, NonZeroFillSize);
2613 if (NonZeroFillSize < FillSize)
2614 getStreamer().EmitIntValue(0, FillSize - NonZeroFillSize);
2621 /// parseDirectiveOrg
2622 /// ::= .org expression [ , expression ]
2623 bool AsmParser::parseDirectiveOrg() {
2624 checkForValidSection();
2626 const MCExpr *Offset;
2627 SMLoc Loc = getTok().getLoc();
2628 if (parseExpression(Offset))
2631 // Parse optional fill expression.
2632 int64_t FillExpr = 0;
2633 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2634 if (getLexer().isNot(AsmToken::Comma))
2635 return TokError("unexpected token in '.org' directive");
2638 if (parseAbsoluteExpression(FillExpr))
2641 if (getLexer().isNot(AsmToken::EndOfStatement))
2642 return TokError("unexpected token in '.org' directive");
2647 // Only limited forms of relocatable expressions are accepted here, it
2648 // has to be relative to the current section. The streamer will return
2649 // 'true' if the expression wasn't evaluatable.
2650 if (getStreamer().EmitValueToOffset(Offset, FillExpr))
2651 return Error(Loc, "expected assembly-time absolute expression");
2656 /// parseDirectiveAlign
2657 /// ::= {.align, ...} expression [ , expression [ , expression ]]
2658 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2659 checkForValidSection();
2661 SMLoc AlignmentLoc = getLexer().getLoc();
2663 if (parseAbsoluteExpression(Alignment))
2667 bool HasFillExpr = false;
2668 int64_t FillExpr = 0;
2669 int64_t MaxBytesToFill = 0;
2670 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2671 if (getLexer().isNot(AsmToken::Comma))
2672 return TokError("unexpected token in directive");
2675 // The fill expression can be omitted while specifying a maximum number of
2676 // alignment bytes, e.g:
2678 if (getLexer().isNot(AsmToken::Comma)) {
2680 if (parseAbsoluteExpression(FillExpr))
2684 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2685 if (getLexer().isNot(AsmToken::Comma))
2686 return TokError("unexpected token in directive");
2689 MaxBytesLoc = getLexer().getLoc();
2690 if (parseAbsoluteExpression(MaxBytesToFill))
2693 if (getLexer().isNot(AsmToken::EndOfStatement))
2694 return TokError("unexpected token in directive");
2703 // Compute alignment in bytes.
2705 // FIXME: Diagnose overflow.
2706 if (Alignment >= 32) {
2707 Error(AlignmentLoc, "invalid alignment value");
2711 Alignment = 1ULL << Alignment;
2713 // Reject alignments that aren't a power of two, for gas compatibility.
2714 if (!isPowerOf2_64(Alignment))
2715 Error(AlignmentLoc, "alignment must be a power of 2");
2718 // Diagnose non-sensical max bytes to align.
2719 if (MaxBytesLoc.isValid()) {
2720 if (MaxBytesToFill < 1) {
2721 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2722 "many bytes, ignoring maximum bytes expression");
2726 if (MaxBytesToFill >= Alignment) {
2727 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2733 // Check whether we should use optimal code alignment for this .align
2735 const MCSection *Section = getStreamer().getCurrentSection().first;
2736 assert(Section && "must have section to emit alignment");
2737 bool UseCodeAlign = Section->UseCodeAlign();
2738 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2739 ValueSize == 1 && UseCodeAlign) {
2740 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2742 // FIXME: Target specific behavior about how the "extra" bytes are filled.
2743 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2750 /// parseDirectiveFile
2751 /// ::= .file [number] filename
2752 /// ::= .file number directory filename
2753 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
2754 // FIXME: I'm not sure what this is.
2755 int64_t FileNumber = -1;
2756 SMLoc FileNumberLoc = getLexer().getLoc();
2757 if (getLexer().is(AsmToken::Integer)) {
2758 FileNumber = getTok().getIntVal();
2762 return TokError("file number less than one");
2765 if (getLexer().isNot(AsmToken::String))
2766 return TokError("unexpected token in '.file' directive");
2768 // Usually the directory and filename together, otherwise just the directory.
2769 // Allow the strings to have escaped octal character sequence.
2770 std::string Path = getTok().getString();
2771 if (parseEscapedString(Path))
2775 StringRef Directory;
2777 std::string FilenameData;
2778 if (getLexer().is(AsmToken::String)) {
2779 if (FileNumber == -1)
2780 return TokError("explicit path specified, but no file number");
2781 if (parseEscapedString(FilenameData))
2783 Filename = FilenameData;
2790 if (getLexer().isNot(AsmToken::EndOfStatement))
2791 return TokError("unexpected token in '.file' directive");
2793 if (FileNumber == -1)
2794 getStreamer().EmitFileDirective(Filename);
2796 if (getContext().getGenDwarfForAssembly())
2798 "input can't have .file dwarf directives when -g is "
2799 "used to generate dwarf debug info for assembly code");
2801 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename) ==
2803 Error(FileNumberLoc, "file number already allocated");
2809 /// parseDirectiveLine
2810 /// ::= .line [number]
2811 bool AsmParser::parseDirectiveLine() {
2812 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2813 if (getLexer().isNot(AsmToken::Integer))
2814 return TokError("unexpected token in '.line' directive");
2816 int64_t LineNumber = getTok().getIntVal();
2820 // FIXME: Do something with the .line.
2823 if (getLexer().isNot(AsmToken::EndOfStatement))
2824 return TokError("unexpected token in '.line' directive");
2829 /// parseDirectiveLoc
2830 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2831 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2832 /// The first number is a file number, must have been previously assigned with
2833 /// a .file directive, the second number is the line number and optionally the
2834 /// third number is a column position (zero if not specified). The remaining
2835 /// optional items are .loc sub-directives.
2836 bool AsmParser::parseDirectiveLoc() {
2837 if (getLexer().isNot(AsmToken::Integer))
2838 return TokError("unexpected token in '.loc' directive");
2839 int64_t FileNumber = getTok().getIntVal();
2841 return TokError("file number less than one in '.loc' directive");
2842 if (!getContext().isValidDwarfFileNumber(FileNumber))
2843 return TokError("unassigned file number in '.loc' directive");
2846 int64_t LineNumber = 0;
2847 if (getLexer().is(AsmToken::Integer)) {
2848 LineNumber = getTok().getIntVal();
2850 return TokError("line number less than zero in '.loc' directive");
2854 int64_t ColumnPos = 0;
2855 if (getLexer().is(AsmToken::Integer)) {
2856 ColumnPos = getTok().getIntVal();
2858 return TokError("column position less than zero in '.loc' directive");
2862 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2864 int64_t Discriminator = 0;
2865 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2867 if (getLexer().is(AsmToken::EndOfStatement))
2871 SMLoc Loc = getTok().getLoc();
2872 if (parseIdentifier(Name))
2873 return TokError("unexpected token in '.loc' directive");
2875 if (Name == "basic_block")
2876 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2877 else if (Name == "prologue_end")
2878 Flags |= DWARF2_FLAG_PROLOGUE_END;
2879 else if (Name == "epilogue_begin")
2880 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2881 else if (Name == "is_stmt") {
2882 Loc = getTok().getLoc();
2883 const MCExpr *Value;
2884 if (parseExpression(Value))
2886 // The expression must be the constant 0 or 1.
2887 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2888 int Value = MCE->getValue();
2890 Flags &= ~DWARF2_FLAG_IS_STMT;
2891 else if (Value == 1)
2892 Flags |= DWARF2_FLAG_IS_STMT;
2894 return Error(Loc, "is_stmt value not 0 or 1");
2896 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2898 } else if (Name == "isa") {
2899 Loc = getTok().getLoc();
2900 const MCExpr *Value;
2901 if (parseExpression(Value))
2903 // The expression must be a constant greater or equal to 0.
2904 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2905 int Value = MCE->getValue();
2907 return Error(Loc, "isa number less than zero");
2910 return Error(Loc, "isa number not a constant value");
2912 } else if (Name == "discriminator") {
2913 if (parseAbsoluteExpression(Discriminator))
2916 return Error(Loc, "unknown sub-directive in '.loc' directive");
2919 if (getLexer().is(AsmToken::EndOfStatement))
2924 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2925 Isa, Discriminator, StringRef());
2930 /// parseDirectiveStabs
2931 /// ::= .stabs string, number, number, number
2932 bool AsmParser::parseDirectiveStabs() {
2933 return TokError("unsupported directive '.stabs'");
2936 /// parseDirectiveCFISections
2937 /// ::= .cfi_sections section [, section]
2938 bool AsmParser::parseDirectiveCFISections() {
2943 if (parseIdentifier(Name))
2944 return TokError("Expected an identifier");
2946 if (Name == ".eh_frame")
2948 else if (Name == ".debug_frame")
2951 if (getLexer().is(AsmToken::Comma)) {
2954 if (parseIdentifier(Name))
2955 return TokError("Expected an identifier");
2957 if (Name == ".eh_frame")
2959 else if (Name == ".debug_frame")
2963 getStreamer().EmitCFISections(EH, Debug);
2967 /// parseDirectiveCFIStartProc
2968 /// ::= .cfi_startproc [simple]
2969 bool AsmParser::parseDirectiveCFIStartProc() {
2971 if (getLexer().isNot(AsmToken::EndOfStatement))
2972 if (parseIdentifier(Simple) || Simple != "simple")
2973 return TokError("unexpected token in .cfi_startproc directive");
2975 getStreamer().EmitCFIStartProc(!Simple.empty());
2979 /// parseDirectiveCFIEndProc
2980 /// ::= .cfi_endproc
2981 bool AsmParser::parseDirectiveCFIEndProc() {
2982 getStreamer().EmitCFIEndProc();
2986 /// \brief parse register name or number.
2987 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register,
2988 SMLoc DirectiveLoc) {
2991 if (getLexer().isNot(AsmToken::Integer)) {
2992 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc))
2994 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true);
2996 return parseAbsoluteExpression(Register);
3001 /// parseDirectiveCFIDefCfa
3002 /// ::= .cfi_def_cfa register, offset
3003 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) {
3004 int64_t Register = 0;
3005 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3008 if (getLexer().isNot(AsmToken::Comma))
3009 return TokError("unexpected token in directive");
3013 if (parseAbsoluteExpression(Offset))
3016 getStreamer().EmitCFIDefCfa(Register, Offset);
3020 /// parseDirectiveCFIDefCfaOffset
3021 /// ::= .cfi_def_cfa_offset offset
3022 bool AsmParser::parseDirectiveCFIDefCfaOffset() {
3024 if (parseAbsoluteExpression(Offset))
3027 getStreamer().EmitCFIDefCfaOffset(Offset);
3031 /// parseDirectiveCFIRegister
3032 /// ::= .cfi_register register, register
3033 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) {
3034 int64_t Register1 = 0;
3035 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc))
3038 if (getLexer().isNot(AsmToken::Comma))
3039 return TokError("unexpected token in directive");
3042 int64_t Register2 = 0;
3043 if (parseRegisterOrRegisterNumber(Register2, DirectiveLoc))
3046 getStreamer().EmitCFIRegister(Register1, Register2);
3050 /// parseDirectiveCFIWindowSave
3051 /// ::= .cfi_window_save
3052 bool AsmParser::parseDirectiveCFIWindowSave() {
3053 getStreamer().EmitCFIWindowSave();
3057 /// parseDirectiveCFIAdjustCfaOffset
3058 /// ::= .cfi_adjust_cfa_offset adjustment
3059 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() {
3060 int64_t Adjustment = 0;
3061 if (parseAbsoluteExpression(Adjustment))
3064 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
3068 /// parseDirectiveCFIDefCfaRegister
3069 /// ::= .cfi_def_cfa_register register
3070 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) {
3071 int64_t Register = 0;
3072 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3075 getStreamer().EmitCFIDefCfaRegister(Register);
3079 /// parseDirectiveCFIOffset
3080 /// ::= .cfi_offset register, offset
3081 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) {
3082 int64_t Register = 0;
3085 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3088 if (getLexer().isNot(AsmToken::Comma))
3089 return TokError("unexpected token in directive");
3092 if (parseAbsoluteExpression(Offset))
3095 getStreamer().EmitCFIOffset(Register, Offset);
3099 /// parseDirectiveCFIRelOffset
3100 /// ::= .cfi_rel_offset register, offset
3101 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) {
3102 int64_t Register = 0;
3104 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3107 if (getLexer().isNot(AsmToken::Comma))
3108 return TokError("unexpected token in directive");
3112 if (parseAbsoluteExpression(Offset))
3115 getStreamer().EmitCFIRelOffset(Register, Offset);
3119 static bool isValidEncoding(int64_t Encoding) {
3120 if (Encoding & ~0xff)
3123 if (Encoding == dwarf::DW_EH_PE_omit)
3126 const unsigned Format = Encoding & 0xf;
3127 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
3128 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
3129 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
3130 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
3133 const unsigned Application = Encoding & 0x70;
3134 if (Application != dwarf::DW_EH_PE_absptr &&
3135 Application != dwarf::DW_EH_PE_pcrel)
3141 /// parseDirectiveCFIPersonalityOrLsda
3142 /// IsPersonality true for cfi_personality, false for cfi_lsda
3143 /// ::= .cfi_personality encoding, [symbol_name]
3144 /// ::= .cfi_lsda encoding, [symbol_name]
3145 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) {
3146 int64_t Encoding = 0;
3147 if (parseAbsoluteExpression(Encoding))
3149 if (Encoding == dwarf::DW_EH_PE_omit)
3152 if (!isValidEncoding(Encoding))
3153 return TokError("unsupported encoding.");
3155 if (getLexer().isNot(AsmToken::Comma))
3156 return TokError("unexpected token in directive");
3160 if (parseIdentifier(Name))
3161 return TokError("expected identifier in directive");
3163 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3166 getStreamer().EmitCFIPersonality(Sym, Encoding);
3168 getStreamer().EmitCFILsda(Sym, Encoding);
3172 /// parseDirectiveCFIRememberState
3173 /// ::= .cfi_remember_state
3174 bool AsmParser::parseDirectiveCFIRememberState() {
3175 getStreamer().EmitCFIRememberState();
3179 /// parseDirectiveCFIRestoreState
3180 /// ::= .cfi_remember_state
3181 bool AsmParser::parseDirectiveCFIRestoreState() {
3182 getStreamer().EmitCFIRestoreState();
3186 /// parseDirectiveCFISameValue
3187 /// ::= .cfi_same_value register
3188 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) {
3189 int64_t Register = 0;
3191 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3194 getStreamer().EmitCFISameValue(Register);
3198 /// parseDirectiveCFIRestore
3199 /// ::= .cfi_restore register
3200 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) {
3201 int64_t Register = 0;
3202 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3205 getStreamer().EmitCFIRestore(Register);
3209 /// parseDirectiveCFIEscape
3210 /// ::= .cfi_escape expression[,...]
3211 bool AsmParser::parseDirectiveCFIEscape() {
3214 if (parseAbsoluteExpression(CurrValue))
3217 Values.push_back((uint8_t)CurrValue);
3219 while (getLexer().is(AsmToken::Comma)) {
3222 if (parseAbsoluteExpression(CurrValue))
3225 Values.push_back((uint8_t)CurrValue);
3228 getStreamer().EmitCFIEscape(Values);
3232 /// parseDirectiveCFISignalFrame
3233 /// ::= .cfi_signal_frame
3234 bool AsmParser::parseDirectiveCFISignalFrame() {
3235 if (getLexer().isNot(AsmToken::EndOfStatement))
3236 return Error(getLexer().getLoc(),
3237 "unexpected token in '.cfi_signal_frame'");
3239 getStreamer().EmitCFISignalFrame();
3243 /// parseDirectiveCFIUndefined
3244 /// ::= .cfi_undefined register
3245 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) {
3246 int64_t Register = 0;
3248 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc))
3251 getStreamer().EmitCFIUndefined(Register);
3255 /// parseDirectiveMacrosOnOff
3258 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) {
3259 if (getLexer().isNot(AsmToken::EndOfStatement))
3260 return Error(getLexer().getLoc(),
3261 "unexpected token in '" + Directive + "' directive");
3263 setMacrosEnabled(Directive == ".macros_on");
3267 /// parseDirectiveMacro
3268 /// ::= .macro name[,] [parameters]
3269 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) {
3271 if (parseIdentifier(Name))
3272 return TokError("expected identifier in '.macro' directive");
3274 if (getLexer().is(AsmToken::Comma))
3277 MCAsmMacroParameters Parameters;
3278 while (getLexer().isNot(AsmToken::EndOfStatement)) {
3280 if (!Parameters.empty() && Parameters.back().Vararg)
3281 return Error(Lexer.getLoc(),
3282 "Vararg parameter '" + Parameters.back().Name +
3283 "' should be last one in the list of parameters.");
3285 MCAsmMacroParameter Parameter;
3286 if (parseIdentifier(Parameter.Name))
3287 return TokError("expected identifier in '.macro' directive");
3289 if (Lexer.is(AsmToken::Colon)) {
3290 Lex(); // consume ':'
3293 StringRef Qualifier;
3295 QualLoc = Lexer.getLoc();
3296 if (parseIdentifier(Qualifier))
3297 return Error(QualLoc, "missing parameter qualifier for "
3298 "'" + Parameter.Name + "' in macro '" + Name + "'");
3300 if (Qualifier == "req")
3301 Parameter.Required = true;
3302 else if (Qualifier == "vararg")
3303 Parameter.Vararg = true;
3305 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier "
3306 "for '" + Parameter.Name + "' in macro '" + Name + "'");
3309 if (getLexer().is(AsmToken::Equal)) {
3314 ParamLoc = Lexer.getLoc();
3315 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false ))
3318 if (Parameter.Required)
3319 Warning(ParamLoc, "pointless default value for required parameter "
3320 "'" + Parameter.Name + "' in macro '" + Name + "'");
3323 Parameters.push_back(std::move(Parameter));
3325 if (getLexer().is(AsmToken::Comma))
3329 // Eat the end of statement.
3332 AsmToken EndToken, StartToken = getTok();
3333 unsigned MacroDepth = 0;
3335 // Lex the macro definition.
3337 // Check whether we have reached the end of the file.
3338 if (getLexer().is(AsmToken::Eof))
3339 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3341 // Otherwise, check whether we have reach the .endmacro.
3342 if (getLexer().is(AsmToken::Identifier)) {
3343 if (getTok().getIdentifier() == ".endm" ||
3344 getTok().getIdentifier() == ".endmacro") {
3345 if (MacroDepth == 0) { // Outermost macro.
3346 EndToken = getTok();
3348 if (getLexer().isNot(AsmToken::EndOfStatement))
3349 return TokError("unexpected token in '" + EndToken.getIdentifier() +
3353 // Otherwise we just found the end of an inner macro.
3356 } else if (getTok().getIdentifier() == ".macro") {
3357 // We allow nested macros. Those aren't instantiated until the outermost
3358 // macro is expanded so just ignore them for now.
3363 // Otherwise, scan til the end of the statement.
3364 eatToEndOfStatement();
3367 if (lookupMacro(Name)) {
3368 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3371 const char *BodyStart = StartToken.getLoc().getPointer();
3372 const char *BodyEnd = EndToken.getLoc().getPointer();
3373 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3374 checkForBadMacro(DirectiveLoc, Name, Body, Parameters);
3375 defineMacro(Name, MCAsmMacro(Name, Body, std::move(Parameters)));
3379 /// checkForBadMacro
3381 /// With the support added for named parameters there may be code out there that
3382 /// is transitioning from positional parameters. In versions of gas that did
3383 /// not support named parameters they would be ignored on the macro definition.
3384 /// But to support both styles of parameters this is not possible so if a macro
3385 /// definition has named parameters but does not use them and has what appears
3386 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a
3387 /// warning that the positional parameter found in body which have no effect.
3388 /// Hoping the developer will either remove the named parameters from the macro
3389 /// definition so the positional parameters get used if that was what was
3390 /// intended or change the macro to use the named parameters. It is possible
3391 /// this warning will trigger when the none of the named parameters are used
3392 /// and the strings like $1 are infact to simply to be passed trough unchanged.
3393 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name,
3395 ArrayRef<MCAsmMacroParameter> Parameters) {
3396 // If this macro is not defined with named parameters the warning we are
3397 // checking for here doesn't apply.
3398 unsigned NParameters = Parameters.size();
3399 if (NParameters == 0)
3402 bool NamedParametersFound = false;
3403 bool PositionalParametersFound = false;
3405 // Look at the body of the macro for use of both the named parameters and what
3406 // are likely to be positional parameters. This is what expandMacro() is
3407 // doing when it finds the parameters in the body.
3408 while (!Body.empty()) {
3409 // Scan for the next possible parameter.
3410 std::size_t End = Body.size(), Pos = 0;
3411 for (; Pos != End; ++Pos) {
3412 // Check for a substitution or escape.
3413 // This macro is defined with parameters, look for \foo, \bar, etc.
3414 if (Body[Pos] == '\\' && Pos + 1 != End)
3417 // This macro should have parameters, but look for $0, $1, ..., $n too.
3418 if (Body[Pos] != '$' || Pos + 1 == End)
3420 char Next = Body[Pos + 1];
3421 if (Next == '$' || Next == 'n' ||
3422 isdigit(static_cast<unsigned char>(Next)))
3426 // Check if we reached the end.
3430 if (Body[Pos] == '$') {
3431 switch (Body[Pos + 1]) {
3436 // $n => number of arguments
3438 PositionalParametersFound = true;
3441 // $[0-9] => argument
3443 PositionalParametersFound = true;
3449 unsigned I = Pos + 1;
3450 while (isIdentifierChar(Body[I]) && I + 1 != End)
3453 const char *Begin = Body.data() + Pos + 1;
3454 StringRef Argument(Begin, I - (Pos + 1));
3456 for (; Index < NParameters; ++Index)
3457 if (Parameters[Index].Name == Argument)
3460 if (Index == NParameters) {
3461 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')')
3467 NamedParametersFound = true;
3468 Pos += 1 + Argument.size();
3471 // Update the scan point.
3472 Body = Body.substr(Pos);
3475 if (!NamedParametersFound && PositionalParametersFound)
3476 Warning(DirectiveLoc, "macro defined with named parameters which are not "
3477 "used in macro body, possible positional parameter "
3478 "found in body which will have no effect");
3481 /// parseDirectiveExitMacro
3483 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) {
3484 if (getLexer().isNot(AsmToken::EndOfStatement))
3485 return TokError("unexpected token in '" + Directive + "' directive");
3487 if (!isInsideMacroInstantiation())
3488 return TokError("unexpected '" + Directive + "' in file, "
3489 "no current macro definition");
3491 // Exit all conditionals that are active in the current macro.
3492 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) {
3493 TheCondState = TheCondStack.back();
3494 TheCondStack.pop_back();
3501 /// parseDirectiveEndMacro
3504 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) {
3505 if (getLexer().isNot(AsmToken::EndOfStatement))
3506 return TokError("unexpected token in '" + Directive + "' directive");
3508 // If we are inside a macro instantiation, terminate the current
3510 if (isInsideMacroInstantiation()) {
3515 // Otherwise, this .endmacro is a stray entry in the file; well formed
3516 // .endmacro directives are handled during the macro definition parsing.
3517 return TokError("unexpected '" + Directive + "' in file, "
3518 "no current macro definition");
3521 /// parseDirectivePurgeMacro
3523 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) {
3525 if (parseIdentifier(Name))
3526 return TokError("expected identifier in '.purgem' directive");
3528 if (getLexer().isNot(AsmToken::EndOfStatement))
3529 return TokError("unexpected token in '.purgem' directive");
3531 if (!lookupMacro(Name))
3532 return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
3534 undefineMacro(Name);
3538 /// parseDirectiveBundleAlignMode
3539 /// ::= {.bundle_align_mode} expression
3540 bool AsmParser::parseDirectiveBundleAlignMode() {
3541 checkForValidSection();
3543 // Expect a single argument: an expression that evaluates to a constant
3544 // in the inclusive range 0-30.
3545 SMLoc ExprLoc = getLexer().getLoc();
3546 int64_t AlignSizePow2;
3547 if (parseAbsoluteExpression(AlignSizePow2))
3549 else if (getLexer().isNot(AsmToken::EndOfStatement))
3550 return TokError("unexpected token after expression in"
3551 " '.bundle_align_mode' directive");
3552 else if (AlignSizePow2 < 0 || AlignSizePow2 > 30)
3553 return Error(ExprLoc,
3554 "invalid bundle alignment size (expected between 0 and 30)");
3558 // Because of AlignSizePow2's verified range we can safely truncate it to
3560 getStreamer().EmitBundleAlignMode(static_cast<unsigned>(AlignSizePow2));
3564 /// parseDirectiveBundleLock
3565 /// ::= {.bundle_lock} [align_to_end]
3566 bool AsmParser::parseDirectiveBundleLock() {
3567 checkForValidSection();
3568 bool AlignToEnd = false;
3570 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3572 SMLoc Loc = getTok().getLoc();
3573 const char *kInvalidOptionError =
3574 "invalid option for '.bundle_lock' directive";
3576 if (parseIdentifier(Option))
3577 return Error(Loc, kInvalidOptionError);
3579 if (Option != "align_to_end")
3580 return Error(Loc, kInvalidOptionError);
3581 else if (getLexer().isNot(AsmToken::EndOfStatement))
3583 "unexpected token after '.bundle_lock' directive option");
3589 getStreamer().EmitBundleLock(AlignToEnd);
3593 /// parseDirectiveBundleLock
3594 /// ::= {.bundle_lock}
3595 bool AsmParser::parseDirectiveBundleUnlock() {
3596 checkForValidSection();
3598 if (getLexer().isNot(AsmToken::EndOfStatement))
3599 return TokError("unexpected token in '.bundle_unlock' directive");
3602 getStreamer().EmitBundleUnlock();
3606 /// parseDirectiveSpace
3607 /// ::= (.skip | .space) expression [ , expression ]
3608 bool AsmParser::parseDirectiveSpace(StringRef IDVal) {
3609 checkForValidSection();
3612 if (parseAbsoluteExpression(NumBytes))
3615 int64_t FillExpr = 0;
3616 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3617 if (getLexer().isNot(AsmToken::Comma))
3618 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
3621 if (parseAbsoluteExpression(FillExpr))
3624 if (getLexer().isNot(AsmToken::EndOfStatement))
3625 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
3631 return TokError("invalid number of bytes in '" + Twine(IDVal) +
3634 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
3635 getStreamer().EmitFill(NumBytes, FillExpr);
3640 /// parseDirectiveLEB128
3641 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ]
3642 bool AsmParser::parseDirectiveLEB128(bool Signed) {
3643 checkForValidSection();
3644 const MCExpr *Value;
3647 if (parseExpression(Value))
3651 getStreamer().EmitSLEB128Value(Value);
3653 getStreamer().EmitULEB128Value(Value);
3655 if (getLexer().is(AsmToken::EndOfStatement))
3658 if (getLexer().isNot(AsmToken::Comma))
3659 return TokError("unexpected token in directive");
3666 /// parseDirectiveSymbolAttribute
3667 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
3668 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
3669 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3672 SMLoc Loc = getTok().getLoc();
3674 if (parseIdentifier(Name))
3675 return Error(Loc, "expected identifier in directive");
3677 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3679 // Assembler local symbols don't make any sense here. Complain loudly.
3680 if (Sym->isTemporary())
3681 return Error(Loc, "non-local symbol required in directive");
3683 if (!getStreamer().EmitSymbolAttribute(Sym, Attr))
3684 return Error(Loc, "unable to emit symbol attribute");
3686 if (getLexer().is(AsmToken::EndOfStatement))
3689 if (getLexer().isNot(AsmToken::Comma))
3690 return TokError("unexpected token in directive");
3699 /// parseDirectiveComm
3700 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
3701 bool AsmParser::parseDirectiveComm(bool IsLocal) {
3702 checkForValidSection();
3704 SMLoc IDLoc = getLexer().getLoc();
3706 if (parseIdentifier(Name))
3707 return TokError("expected identifier in directive");
3709 // Handle the identifier as the key symbol.
3710 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
3712 if (getLexer().isNot(AsmToken::Comma))
3713 return TokError("unexpected token in directive");
3717 SMLoc SizeLoc = getLexer().getLoc();
3718 if (parseAbsoluteExpression(Size))
3721 int64_t Pow2Alignment = 0;
3722 SMLoc Pow2AlignmentLoc;
3723 if (getLexer().is(AsmToken::Comma)) {
3725 Pow2AlignmentLoc = getLexer().getLoc();
3726 if (parseAbsoluteExpression(Pow2Alignment))
3729 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType();
3730 if (IsLocal && LCOMM == LCOMM::NoAlignment)
3731 return Error(Pow2AlignmentLoc, "alignment not supported on this target");
3733 // If this target takes alignments in bytes (not log) validate and convert.
3734 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) ||
3735 (IsLocal && LCOMM == LCOMM::ByteAlignment)) {
3736 if (!isPowerOf2_64(Pow2Alignment))
3737 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
3738 Pow2Alignment = Log2_64(Pow2Alignment);
3742 if (getLexer().isNot(AsmToken::EndOfStatement))
3743 return TokError("unexpected token in '.comm' or '.lcomm' directive");
3747 // NOTE: a size of zero for a .comm should create a undefined symbol
3748 // but a size of .lcomm creates a bss symbol of size zero.
3750 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
3751 "be less than zero");
3753 // NOTE: The alignment in the directive is a power of 2 value, the assembler
3754 // may internally end up wanting an alignment in bytes.
3755 // FIXME: Diagnose overflow.
3756 if (Pow2Alignment < 0)
3757 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
3758 "alignment, can't be less than zero");
3760 if (!Sym->isUndefined())
3761 return Error(IDLoc, "invalid symbol redefinition");
3763 // Create the Symbol as a common or local common with Size and Pow2Alignment
3765 getStreamer().EmitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment);
3769 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
3773 /// parseDirectiveAbort
3774 /// ::= .abort [... message ...]
3775 bool AsmParser::parseDirectiveAbort() {
3776 // FIXME: Use loc from directive.
3777 SMLoc Loc = getLexer().getLoc();
3779 StringRef Str = parseStringToEndOfStatement();
3780 if (getLexer().isNot(AsmToken::EndOfStatement))
3781 return TokError("unexpected token in '.abort' directive");
3786 Error(Loc, ".abort detected. Assembly stopping.");
3788 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
3789 // FIXME: Actually abort assembly here.
3794 /// parseDirectiveInclude
3795 /// ::= .include "filename"
3796 bool AsmParser::parseDirectiveInclude() {
3797 if (getLexer().isNot(AsmToken::String))
3798 return TokError("expected string in '.include' directive");
3800 // Allow the strings to have escaped octal character sequence.
3801 std::string Filename;
3802 if (parseEscapedString(Filename))
3804 SMLoc IncludeLoc = getLexer().getLoc();
3807 if (getLexer().isNot(AsmToken::EndOfStatement))
3808 return TokError("unexpected token in '.include' directive");
3810 // Attempt to switch the lexer to the included file before consuming the end
3811 // of statement to avoid losing it when we switch.
3812 if (enterIncludeFile(Filename)) {
3813 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
3820 /// parseDirectiveIncbin
3821 /// ::= .incbin "filename"
3822 bool AsmParser::parseDirectiveIncbin() {
3823 if (getLexer().isNot(AsmToken::String))
3824 return TokError("expected string in '.incbin' directive");
3826 // Allow the strings to have escaped octal character sequence.
3827 std::string Filename;
3828 if (parseEscapedString(Filename))
3830 SMLoc IncbinLoc = getLexer().getLoc();
3833 if (getLexer().isNot(AsmToken::EndOfStatement))
3834 return TokError("unexpected token in '.incbin' directive");
3836 // Attempt to process the included file.
3837 if (processIncbinFile(Filename)) {
3838 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
3845 /// parseDirectiveIf
3846 /// ::= .if{,eq,ge,gt,le,lt,ne} expression
3847 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) {
3848 TheCondStack.push_back(TheCondState);
3849 TheCondState.TheCond = AsmCond::IfCond;
3850 if (TheCondState.Ignore) {
3851 eatToEndOfStatement();
3854 if (parseAbsoluteExpression(ExprValue))
3857 if (getLexer().isNot(AsmToken::EndOfStatement))
3858 return TokError("unexpected token in '.if' directive");
3864 llvm_unreachable("unsupported directive");
3869 ExprValue = ExprValue == 0;
3872 ExprValue = ExprValue >= 0;
3875 ExprValue = ExprValue > 0;
3878 ExprValue = ExprValue <= 0;
3881 ExprValue = ExprValue < 0;
3885 TheCondState.CondMet = ExprValue;
3886 TheCondState.Ignore = !TheCondState.CondMet;
3892 /// parseDirectiveIfb
3894 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
3895 TheCondStack.push_back(TheCondState);
3896 TheCondState.TheCond = AsmCond::IfCond;
3898 if (TheCondState.Ignore) {
3899 eatToEndOfStatement();
3901 StringRef Str = parseStringToEndOfStatement();
3903 if (getLexer().isNot(AsmToken::EndOfStatement))
3904 return TokError("unexpected token in '.ifb' directive");
3908 TheCondState.CondMet = ExpectBlank == Str.empty();
3909 TheCondState.Ignore = !TheCondState.CondMet;
3915 /// parseDirectiveIfc
3916 /// ::= .ifc string1, string2
3917 /// ::= .ifnc string1, string2
3918 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
3919 TheCondStack.push_back(TheCondState);
3920 TheCondState.TheCond = AsmCond::IfCond;
3922 if (TheCondState.Ignore) {
3923 eatToEndOfStatement();
3925 StringRef Str1 = parseStringToComma();
3927 if (getLexer().isNot(AsmToken::Comma))
3928 return TokError("unexpected token in '.ifc' directive");
3932 StringRef Str2 = parseStringToEndOfStatement();
3934 if (getLexer().isNot(AsmToken::EndOfStatement))
3935 return TokError("unexpected token in '.ifc' directive");
3939 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim());
3940 TheCondState.Ignore = !TheCondState.CondMet;
3946 /// parseDirectiveIfeqs
3947 /// ::= .ifeqs string1, string2
3948 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) {
3949 if (Lexer.isNot(AsmToken::String)) {
3951 TokError("expected string parameter for '.ifeqs' directive");
3953 TokError("expected string parameter for '.ifnes' directive");
3954 eatToEndOfStatement();
3958 StringRef String1 = getTok().getStringContents();
3961 if (Lexer.isNot(AsmToken::Comma)) {
3963 TokError("expected comma after first string for '.ifeqs' directive");
3965 TokError("expected comma after first string for '.ifnes' directive");
3966 eatToEndOfStatement();
3972 if (Lexer.isNot(AsmToken::String)) {
3974 TokError("expected string parameter for '.ifeqs' directive");
3976 TokError("expected string parameter for '.ifnes' directive");
3977 eatToEndOfStatement();
3981 StringRef String2 = getTok().getStringContents();
3984 TheCondStack.push_back(TheCondState);
3985 TheCondState.TheCond = AsmCond::IfCond;
3986 TheCondState.CondMet = ExpectEqual == (String1 == String2);
3987 TheCondState.Ignore = !TheCondState.CondMet;
3992 /// parseDirectiveIfdef
3993 /// ::= .ifdef symbol
3994 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
3996 TheCondStack.push_back(TheCondState);
3997 TheCondState.TheCond = AsmCond::IfCond;
3999 if (TheCondState.Ignore) {
4000 eatToEndOfStatement();
4002 if (parseIdentifier(Name))
4003 return TokError("expected identifier after '.ifdef'");
4007 MCSymbol *Sym = getContext().LookupSymbol(Name);
4010 TheCondState.CondMet = (Sym && !Sym->isUndefined());
4012 TheCondState.CondMet = (!Sym || Sym->isUndefined());
4013 TheCondState.Ignore = !TheCondState.CondMet;
4019 /// parseDirectiveElseIf
4020 /// ::= .elseif expression
4021 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) {
4022 if (TheCondState.TheCond != AsmCond::IfCond &&
4023 TheCondState.TheCond != AsmCond::ElseIfCond)
4024 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
4026 TheCondState.TheCond = AsmCond::ElseIfCond;
4028 bool LastIgnoreState = false;
4029 if (!TheCondStack.empty())
4030 LastIgnoreState = TheCondStack.back().Ignore;
4031 if (LastIgnoreState || TheCondState.CondMet) {
4032 TheCondState.Ignore = true;
4033 eatToEndOfStatement();
4036 if (parseAbsoluteExpression(ExprValue))
4039 if (getLexer().isNot(AsmToken::EndOfStatement))
4040 return TokError("unexpected token in '.elseif' directive");
4043 TheCondState.CondMet = ExprValue;
4044 TheCondState.Ignore = !TheCondState.CondMet;
4050 /// parseDirectiveElse
4052 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) {
4053 if (getLexer().isNot(AsmToken::EndOfStatement))
4054 return TokError("unexpected token in '.else' directive");
4058 if (TheCondState.TheCond != AsmCond::IfCond &&
4059 TheCondState.TheCond != AsmCond::ElseIfCond)
4060 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
4062 TheCondState.TheCond = AsmCond::ElseCond;
4063 bool LastIgnoreState = false;
4064 if (!TheCondStack.empty())
4065 LastIgnoreState = TheCondStack.back().Ignore;
4066 if (LastIgnoreState || TheCondState.CondMet)
4067 TheCondState.Ignore = true;
4069 TheCondState.Ignore = false;
4074 /// parseDirectiveEnd
4076 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) {
4077 if (getLexer().isNot(AsmToken::EndOfStatement))
4078 return TokError("unexpected token in '.end' directive");
4082 while (Lexer.isNot(AsmToken::Eof))
4088 /// parseDirectiveError
4090 /// ::= .error [string]
4091 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) {
4092 if (!TheCondStack.empty()) {
4093 if (TheCondStack.back().Ignore) {
4094 eatToEndOfStatement();
4100 return Error(L, ".err encountered");
4102 StringRef Message = ".error directive invoked in source file";
4103 if (Lexer.isNot(AsmToken::EndOfStatement)) {
4104 if (Lexer.isNot(AsmToken::String)) {
4105 TokError(".error argument must be a string");
4106 eatToEndOfStatement();
4110 Message = getTok().getStringContents();
4118 /// parseDirectiveWarning
4119 /// ::= .warning [string]
4120 bool AsmParser::parseDirectiveWarning(SMLoc L) {
4121 if (!TheCondStack.empty()) {
4122 if (TheCondStack.back().Ignore) {
4123 eatToEndOfStatement();
4128 StringRef Message = ".warning directive invoked in source file";
4129 if (Lexer.isNot(AsmToken::EndOfStatement)) {
4130 if (Lexer.isNot(AsmToken::String)) {
4131 TokError(".warning argument must be a string");
4132 eatToEndOfStatement();
4136 Message = getTok().getStringContents();
4140 Warning(L, Message);
4144 /// parseDirectiveEndIf
4146 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) {
4147 if (getLexer().isNot(AsmToken::EndOfStatement))
4148 return TokError("unexpected token in '.endif' directive");
4152 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty())
4153 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
4155 if (!TheCondStack.empty()) {
4156 TheCondState = TheCondStack.back();
4157 TheCondStack.pop_back();
4163 void AsmParser::initializeDirectiveKindMap() {
4164 DirectiveKindMap[".set"] = DK_SET;
4165 DirectiveKindMap[".equ"] = DK_EQU;
4166 DirectiveKindMap[".equiv"] = DK_EQUIV;
4167 DirectiveKindMap[".ascii"] = DK_ASCII;
4168 DirectiveKindMap[".asciz"] = DK_ASCIZ;
4169 DirectiveKindMap[".string"] = DK_STRING;
4170 DirectiveKindMap[".byte"] = DK_BYTE;
4171 DirectiveKindMap[".short"] = DK_SHORT;
4172 DirectiveKindMap[".value"] = DK_VALUE;
4173 DirectiveKindMap[".2byte"] = DK_2BYTE;
4174 DirectiveKindMap[".long"] = DK_LONG;
4175 DirectiveKindMap[".int"] = DK_INT;
4176 DirectiveKindMap[".4byte"] = DK_4BYTE;
4177 DirectiveKindMap[".quad"] = DK_QUAD;
4178 DirectiveKindMap[".8byte"] = DK_8BYTE;
4179 DirectiveKindMap[".octa"] = DK_OCTA;
4180 DirectiveKindMap[".single"] = DK_SINGLE;
4181 DirectiveKindMap[".float"] = DK_FLOAT;
4182 DirectiveKindMap[".double"] = DK_DOUBLE;
4183 DirectiveKindMap[".align"] = DK_ALIGN;
4184 DirectiveKindMap[".align32"] = DK_ALIGN32;
4185 DirectiveKindMap[".balign"] = DK_BALIGN;
4186 DirectiveKindMap[".balignw"] = DK_BALIGNW;
4187 DirectiveKindMap[".balignl"] = DK_BALIGNL;
4188 DirectiveKindMap[".p2align"] = DK_P2ALIGN;
4189 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW;
4190 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL;
4191 DirectiveKindMap[".org"] = DK_ORG;
4192 DirectiveKindMap[".fill"] = DK_FILL;
4193 DirectiveKindMap[".zero"] = DK_ZERO;
4194 DirectiveKindMap[".extern"] = DK_EXTERN;
4195 DirectiveKindMap[".globl"] = DK_GLOBL;
4196 DirectiveKindMap[".global"] = DK_GLOBAL;
4197 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE;
4198 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP;
4199 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER;
4200 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN;
4201 DirectiveKindMap[".reference"] = DK_REFERENCE;
4202 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION;
4203 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE;
4204 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN;
4205 DirectiveKindMap[".comm"] = DK_COMM;
4206 DirectiveKindMap[".common"] = DK_COMMON;
4207 DirectiveKindMap[".lcomm"] = DK_LCOMM;
4208 DirectiveKindMap[".abort"] = DK_ABORT;
4209 DirectiveKindMap[".include"] = DK_INCLUDE;
4210 DirectiveKindMap[".incbin"] = DK_INCBIN;
4211 DirectiveKindMap[".code16"] = DK_CODE16;
4212 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC;
4213 DirectiveKindMap[".rept"] = DK_REPT;
4214 DirectiveKindMap[".rep"] = DK_REPT;
4215 DirectiveKindMap[".irp"] = DK_IRP;
4216 DirectiveKindMap[".irpc"] = DK_IRPC;
4217 DirectiveKindMap[".endr"] = DK_ENDR;
4218 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE;
4219 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK;
4220 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK;
4221 DirectiveKindMap[".if"] = DK_IF;
4222 DirectiveKindMap[".ifeq"] = DK_IFEQ;
4223 DirectiveKindMap[".ifge"] = DK_IFGE;
4224 DirectiveKindMap[".ifgt"] = DK_IFGT;
4225 DirectiveKindMap[".ifle"] = DK_IFLE;
4226 DirectiveKindMap[".iflt"] = DK_IFLT;
4227 DirectiveKindMap[".ifne"] = DK_IFNE;
4228 DirectiveKindMap[".ifb"] = DK_IFB;
4229 DirectiveKindMap[".ifnb"] = DK_IFNB;
4230 DirectiveKindMap[".ifc"] = DK_IFC;
4231 DirectiveKindMap[".ifeqs"] = DK_IFEQS;
4232 DirectiveKindMap[".ifnc"] = DK_IFNC;
4233 DirectiveKindMap[".ifnes"] = DK_IFNES;
4234 DirectiveKindMap[".ifdef"] = DK_IFDEF;
4235 DirectiveKindMap[".ifndef"] = DK_IFNDEF;
4236 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF;
4237 DirectiveKindMap[".elseif"] = DK_ELSEIF;
4238 DirectiveKindMap[".else"] = DK_ELSE;
4239 DirectiveKindMap[".end"] = DK_END;
4240 DirectiveKindMap[".endif"] = DK_ENDIF;
4241 DirectiveKindMap[".skip"] = DK_SKIP;
4242 DirectiveKindMap[".space"] = DK_SPACE;
4243 DirectiveKindMap[".file"] = DK_FILE;
4244 DirectiveKindMap[".line"] = DK_LINE;
4245 DirectiveKindMap[".loc"] = DK_LOC;
4246 DirectiveKindMap[".stabs"] = DK_STABS;
4247 DirectiveKindMap[".sleb128"] = DK_SLEB128;
4248 DirectiveKindMap[".uleb128"] = DK_ULEB128;
4249 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS;
4250 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC;
4251 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC;
4252 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA;
4253 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET;
4254 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET;
4255 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER;
4256 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET;
4257 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET;
4258 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY;
4259 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA;
4260 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE;
4261 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE;
4262 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE;
4263 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE;
4264 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE;
4265 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME;
4266 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED;
4267 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER;
4268 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE;
4269 DirectiveKindMap[".macros_on"] = DK_MACROS_ON;
4270 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF;
4271 DirectiveKindMap[".macro"] = DK_MACRO;
4272 DirectiveKindMap[".exitm"] = DK_EXITM;
4273 DirectiveKindMap[".endm"] = DK_ENDM;
4274 DirectiveKindMap[".endmacro"] = DK_ENDMACRO;
4275 DirectiveKindMap[".purgem"] = DK_PURGEM;
4276 DirectiveKindMap[".err"] = DK_ERR;
4277 DirectiveKindMap[".error"] = DK_ERROR;
4278 DirectiveKindMap[".warning"] = DK_WARNING;
4281 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) {
4282 AsmToken EndToken, StartToken = getTok();
4284 unsigned NestLevel = 0;
4286 // Check whether we have reached the end of the file.
4287 if (getLexer().is(AsmToken::Eof)) {
4288 Error(DirectiveLoc, "no matching '.endr' in definition");
4292 if (Lexer.is(AsmToken::Identifier) &&
4293 (getTok().getIdentifier() == ".rept")) {
4297 // Otherwise, check whether we have reached the .endr.
4298 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") {
4299 if (NestLevel == 0) {
4300 EndToken = getTok();
4302 if (Lexer.isNot(AsmToken::EndOfStatement)) {
4303 TokError("unexpected token in '.endr' directive");
4311 // Otherwise, scan till the end of the statement.
4312 eatToEndOfStatement();
4315 const char *BodyStart = StartToken.getLoc().getPointer();
4316 const char *BodyEnd = EndToken.getLoc().getPointer();
4317 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
4319 // We Are Anonymous.
4320 MacroLikeBodies.push_back(
4321 MCAsmMacro(StringRef(), Body, MCAsmMacroParameters()));
4322 return &MacroLikeBodies.back();
4325 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
4326 raw_svector_ostream &OS) {
4329 std::unique_ptr<MemoryBuffer> Instantiation =
4330 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
4332 // Create the macro instantiation object and add to the current macro
4333 // instantiation stack.
4334 MacroInstantiation *MI = new MacroInstantiation(
4335 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size());
4336 ActiveMacros.push_back(MI);
4338 // Jump to the macro instantiation and prime the lexer.
4339 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
4340 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
4344 /// parseDirectiveRept
4345 /// ::= .rep | .rept count
4346 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) {
4347 const MCExpr *CountExpr;
4348 SMLoc CountLoc = getTok().getLoc();
4349 if (parseExpression(CountExpr))
4353 if (!CountExpr->EvaluateAsAbsolute(Count)) {
4354 eatToEndOfStatement();
4355 return Error(CountLoc, "unexpected token in '" + Dir + "' directive");
4359 return Error(CountLoc, "Count is negative");
4361 if (Lexer.isNot(AsmToken::EndOfStatement))
4362 return TokError("unexpected token in '" + Dir + "' directive");
4364 // Eat the end of statement.
4367 // Lex the rept definition.
4368 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4372 // Macro instantiation is lexical, unfortunately. We construct a new buffer
4373 // to hold the macro body with substitutions.
4374 SmallString<256> Buf;
4375 raw_svector_ostream OS(Buf);
4377 if (expandMacro(OS, M->Body, None, None, getTok().getLoc()))
4380 instantiateMacroLikeBody(M, DirectiveLoc, OS);
4385 /// parseDirectiveIrp
4386 /// ::= .irp symbol,values
4387 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) {
4388 MCAsmMacroParameter Parameter;
4390 if (parseIdentifier(Parameter.Name))
4391 return TokError("expected identifier in '.irp' directive");
4393 if (Lexer.isNot(AsmToken::Comma))
4394 return TokError("expected comma in '.irp' directive");
4398 MCAsmMacroArguments A;
4399 if (parseMacroArguments(nullptr, A))
4402 // Eat the end of statement.
4405 // Lex the irp definition.
4406 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4410 // Macro instantiation is lexical, unfortunately. We construct a new buffer
4411 // to hold the macro body with substitutions.
4412 SmallString<256> Buf;
4413 raw_svector_ostream OS(Buf);
4415 for (MCAsmMacroArguments::iterator i = A.begin(), e = A.end(); i != e; ++i) {
4416 if (expandMacro(OS, M->Body, Parameter, *i, getTok().getLoc()))
4420 instantiateMacroLikeBody(M, DirectiveLoc, OS);
4425 /// parseDirectiveIrpc
4426 /// ::= .irpc symbol,values
4427 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) {
4428 MCAsmMacroParameter Parameter;
4430 if (parseIdentifier(Parameter.Name))
4431 return TokError("expected identifier in '.irpc' directive");
4433 if (Lexer.isNot(AsmToken::Comma))
4434 return TokError("expected comma in '.irpc' directive");
4438 MCAsmMacroArguments A;
4439 if (parseMacroArguments(nullptr, A))
4442 if (A.size() != 1 || A.front().size() != 1)
4443 return TokError("unexpected token in '.irpc' directive");
4445 // Eat the end of statement.
4448 // Lex the irpc definition.
4449 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc);
4453 // Macro instantiation is lexical, unfortunately. We construct a new buffer
4454 // to hold the macro body with substitutions.
4455 SmallString<256> Buf;
4456 raw_svector_ostream OS(Buf);
4458 StringRef Values = A.front().front().getString();
4459 for (std::size_t I = 0, End = Values.size(); I != End; ++I) {
4460 MCAsmMacroArgument Arg;
4461 Arg.push_back(AsmToken(AsmToken::Identifier, Values.slice(I, I + 1)));
4463 if (expandMacro(OS, M->Body, Parameter, Arg, getTok().getLoc()))
4467 instantiateMacroLikeBody(M, DirectiveLoc, OS);
4472 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) {
4473 if (ActiveMacros.empty())
4474 return TokError("unmatched '.endr' directive");
4476 // The only .repl that should get here are the ones created by
4477 // instantiateMacroLikeBody.
4478 assert(getLexer().is(AsmToken::EndOfStatement));
4484 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info,
4486 const MCExpr *Value;
4487 SMLoc ExprLoc = getLexer().getLoc();
4488 if (parseExpression(Value))
4490 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4492 return Error(ExprLoc, "unexpected expression in _emit");
4493 uint64_t IntValue = MCE->getValue();
4494 if (!isUIntN(8, IntValue) && !isIntN(8, IntValue))
4495 return Error(ExprLoc, "literal value out of range for directive");
4497 Info.AsmRewrites->push_back(AsmRewrite(AOK_Emit, IDLoc, Len));
4501 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) {
4502 const MCExpr *Value;
4503 SMLoc ExprLoc = getLexer().getLoc();
4504 if (parseExpression(Value))
4506 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4508 return Error(ExprLoc, "unexpected expression in align");
4509 uint64_t IntValue = MCE->getValue();
4510 if (!isPowerOf2_64(IntValue))
4511 return Error(ExprLoc, "literal value not a power of two greater then zero");
4513 Info.AsmRewrites->push_back(
4514 AsmRewrite(AOK_Align, IDLoc, 5, Log2_64(IntValue)));
4518 // We are comparing pointers, but the pointers are relative to a single string.
4519 // Thus, this should always be deterministic.
4520 static int rewritesSort(const AsmRewrite *AsmRewriteA,
4521 const AsmRewrite *AsmRewriteB) {
4522 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer())
4524 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer())
4527 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output
4528 // rewrite to the same location. Make sure the SizeDirective rewrite is
4529 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This
4530 // ensures the sort algorithm is stable.
4531 if (AsmRewritePrecedence[AsmRewriteA->Kind] >
4532 AsmRewritePrecedence[AsmRewriteB->Kind])
4535 if (AsmRewritePrecedence[AsmRewriteA->Kind] <
4536 AsmRewritePrecedence[AsmRewriteB->Kind])
4538 llvm_unreachable("Unstable rewrite sort.");
4541 bool AsmParser::parseMSInlineAsm(
4542 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs,
4543 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool> > &OpDecls,
4544 SmallVectorImpl<std::string> &Constraints,
4545 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII,
4546 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) {
4547 SmallVector<void *, 4> InputDecls;
4548 SmallVector<void *, 4> OutputDecls;
4549 SmallVector<bool, 4> InputDeclsAddressOf;
4550 SmallVector<bool, 4> OutputDeclsAddressOf;
4551 SmallVector<std::string, 4> InputConstraints;
4552 SmallVector<std::string, 4> OutputConstraints;
4553 SmallVector<unsigned, 4> ClobberRegs;
4555 SmallVector<AsmRewrite, 4> AsmStrRewrites;
4560 // While we have input, parse each statement.
4561 unsigned InputIdx = 0;
4562 unsigned OutputIdx = 0;
4563 while (getLexer().isNot(AsmToken::Eof)) {
4564 ParseStatementInfo Info(&AsmStrRewrites);
4565 if (parseStatement(Info, &SI))
4568 if (Info.ParseError)
4571 if (Info.Opcode == ~0U)
4574 const MCInstrDesc &Desc = MII->get(Info.Opcode);
4576 // Build the list of clobbers, outputs and inputs.
4577 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) {
4578 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i];
4581 if (Operand.isImm())
4584 // Register operand.
4585 if (Operand.isReg() && !Operand.needAddressOf() &&
4586 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) {
4587 unsigned NumDefs = Desc.getNumDefs();
4589 if (NumDefs && Operand.getMCOperandNum() < NumDefs)
4590 ClobberRegs.push_back(Operand.getReg());
4594 // Expr/Input or Output.
4595 StringRef SymName = Operand.getSymName();
4596 if (SymName.empty())
4599 void *OpDecl = Operand.getOpDecl();
4603 bool isOutput = (i == 1) && Desc.mayStore();
4604 SMLoc Start = SMLoc::getFromPointer(SymName.data());
4607 OutputDecls.push_back(OpDecl);
4608 OutputDeclsAddressOf.push_back(Operand.needAddressOf());
4609 OutputConstraints.push_back(("=" + Operand.getConstraint()).str());
4610 AsmStrRewrites.push_back(AsmRewrite(AOK_Output, Start, SymName.size()));
4612 InputDecls.push_back(OpDecl);
4613 InputDeclsAddressOf.push_back(Operand.needAddressOf());
4614 InputConstraints.push_back(Operand.getConstraint().str());
4615 AsmStrRewrites.push_back(AsmRewrite(AOK_Input, Start, SymName.size()));
4619 // Consider implicit defs to be clobbers. Think of cpuid and push.
4620 ArrayRef<uint16_t> ImpDefs(Desc.getImplicitDefs(),
4621 Desc.getNumImplicitDefs());
4622 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end());
4625 // Set the number of Outputs and Inputs.
4626 NumOutputs = OutputDecls.size();
4627 NumInputs = InputDecls.size();
4629 // Set the unique clobbers.
4630 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end());
4631 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()),
4633 Clobbers.assign(ClobberRegs.size(), std::string());
4634 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) {
4635 raw_string_ostream OS(Clobbers[I]);
4636 IP->printRegName(OS, ClobberRegs[I]);
4639 // Merge the various outputs and inputs. Output are expected first.
4640 if (NumOutputs || NumInputs) {
4641 unsigned NumExprs = NumOutputs + NumInputs;
4642 OpDecls.resize(NumExprs);
4643 Constraints.resize(NumExprs);
4644 for (unsigned i = 0; i < NumOutputs; ++i) {
4645 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]);
4646 Constraints[i] = OutputConstraints[i];
4648 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) {
4649 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]);
4650 Constraints[j] = InputConstraints[i];
4654 // Build the IR assembly string.
4655 std::string AsmStringIR;
4656 raw_string_ostream OS(AsmStringIR);
4657 StringRef ASMString =
4658 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer();
4659 const char *AsmStart = ASMString.begin();
4660 const char *AsmEnd = ASMString.end();
4661 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort);
4662 for (const AsmRewrite &AR : AsmStrRewrites) {
4663 AsmRewriteKind Kind = AR.Kind;
4664 if (Kind == AOK_Delete)
4667 const char *Loc = AR.Loc.getPointer();
4668 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!");
4670 // Emit everything up to the immediate/expression.
4671 if (unsigned Len = Loc - AsmStart)
4672 OS << StringRef(AsmStart, Len);
4674 // Skip the original expression.
4675 if (Kind == AOK_Skip) {
4676 AsmStart = Loc + AR.Len;
4680 unsigned AdditionalSkip = 0;
4681 // Rewrite expressions in $N notation.
4686 OS << "$$" << AR.Val;
4692 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label;
4695 OS << '$' << InputIdx++;
4698 OS << '$' << OutputIdx++;
4700 case AOK_SizeDirective:
4703 case 8: OS << "byte ptr "; break;
4704 case 16: OS << "word ptr "; break;
4705 case 32: OS << "dword ptr "; break;
4706 case 64: OS << "qword ptr "; break;
4707 case 80: OS << "xword ptr "; break;
4708 case 128: OS << "xmmword ptr "; break;
4709 case 256: OS << "ymmword ptr "; break;
4716 unsigned Val = AR.Val;
4717 OS << ".align " << Val;
4719 // Skip the original immediate.
4720 assert(Val < 10 && "Expected alignment less then 2^10.");
4721 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4;
4724 case AOK_DotOperator:
4725 // Insert the dot if the user omitted it.
4727 if (AsmStringIR.back() != '.')
4733 // Skip the original expression.
4734 AsmStart = Loc + AR.Len + AdditionalSkip;
4737 // Emit the remainder of the asm string.
4738 if (AsmStart != AsmEnd)
4739 OS << StringRef(AsmStart, AsmEnd - AsmStart);
4741 AsmString = OS.str();
4745 /// \brief Create an MCAsmParser instance.
4746 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C,
4747 MCStreamer &Out, const MCAsmInfo &MAI) {
4748 return new AsmParser(SM, C, Out, MAI);