1 //===- LLLexer.cpp - Lexer for .ll 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 // Implement the Lexer for .ll files.
12 //===----------------------------------------------------------------------===//
15 #include "llvm/ADT/APInt.h"
16 #include "llvm/ADT/STLExtras.h"
17 #include "llvm/ADT/StringExtras.h"
18 #include "llvm/ADT/Twine.h"
19 #include "llvm/IR/DerivedTypes.h"
20 #include "llvm/IR/Instruction.h"
21 #include "llvm/Support/ErrorHandling.h"
22 #include "llvm/Support/SourceMgr.h"
29 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
30 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
34 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
35 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
38 //===----------------------------------------------------------------------===//
40 //===----------------------------------------------------------------------===//
42 // atoull - Convert an ascii string of decimal digits into the unsigned long
43 // long representation... this does not have to do input error checking,
44 // because we know that the input will be matched by a suitable regex...
46 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
48 for (; Buffer != End; Buffer++) {
49 uint64_t OldRes = Result;
51 Result += *Buffer-'0';
52 if (Result < OldRes) { // Uh, oh, overflow detected!!!
53 Error("constant bigger than 64 bits detected!");
60 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
62 for (; Buffer != End; ++Buffer) {
63 uint64_t OldRes = Result;
65 Result += hexDigitValue(*Buffer);
67 if (Result < OldRes) { // Uh, oh, overflow detected!!!
68 Error("constant bigger than 64 bits detected!");
75 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
78 if (End - Buffer >= 16) {
79 for (int i = 0; i < 16; i++, Buffer++) {
80 assert(Buffer != End);
82 Pair[0] += hexDigitValue(*Buffer);
86 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
88 Pair[1] += hexDigitValue(*Buffer);
91 Error("constant bigger than 128 bits detected!");
94 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
95 /// { low64, high16 } as usual for an APInt.
96 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
99 for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
100 assert(Buffer != End);
102 Pair[1] += hexDigitValue(*Buffer);
105 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
107 Pair[0] += hexDigitValue(*Buffer);
110 Error("constant bigger than 128 bits detected!");
113 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
114 // appropriate character.
115 static void UnEscapeLexed(std::string &Str) {
116 if (Str.empty()) return;
118 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
120 for (char *BIn = Buffer; BIn != EndBuffer; ) {
121 if (BIn[0] == '\\') {
122 if (BIn < EndBuffer-1 && BIn[1] == '\\') {
123 *BOut++ = '\\'; // Two \ becomes one
125 } else if (BIn < EndBuffer-2 &&
126 isxdigit(static_cast<unsigned char>(BIn[1])) &&
127 isxdigit(static_cast<unsigned char>(BIn[2]))) {
128 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
129 BIn += 3; // Skip over handled chars
138 Str.resize(BOut-Buffer);
141 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
142 static bool isLabelChar(char C) {
143 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
144 C == '.' || C == '_';
147 /// isLabelTail - Return true if this pointer points to a valid end of a label.
148 static const char *isLabelTail(const char *CurPtr) {
150 if (CurPtr[0] == ':') return CurPtr+1;
151 if (!isLabelChar(CurPtr[0])) return nullptr;
156 //===----------------------------------------------------------------------===//
158 //===----------------------------------------------------------------------===//
160 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &SM, SMDiagnostic &Err,
162 : CurBuf(StartBuf), ErrorInfo(Err), SM(SM), Context(C), APFloatVal(0.0),
163 IgnoreColonInIdentifiers(false) {
164 CurPtr = CurBuf.begin();
167 int LLLexer::getNextChar() {
168 char CurChar = *CurPtr++;
170 default: return (unsigned char)CurChar;
172 // A nul character in the stream is either the end of the current buffer or
173 // a random nul in the file. Disambiguate that here.
174 if (CurPtr-1 != CurBuf.end())
175 return 0; // Just whitespace.
177 // Otherwise, return end of file.
178 --CurPtr; // Another call to lex will return EOF again.
183 lltok::Kind LLLexer::LexToken() {
187 int CurChar = getNextChar();
190 // Handle letters: [a-zA-Z_]
191 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
192 return LexIdentifier();
195 case EOF: return lltok::Eof;
201 // Ignore whitespace.
203 case '+': return LexPositive();
204 case '@': return LexAt();
205 case '$': return LexDollar();
206 case '%': return LexPercent();
207 case '"': return LexQuote();
209 if (const char *Ptr = isLabelTail(CurPtr)) {
211 StrVal.assign(TokStart, CurPtr-1);
212 return lltok::LabelStr;
214 if (CurPtr[0] == '.' && CurPtr[1] == '.') {
216 return lltok::dotdotdot;
222 case '!': return LexExclaim();
227 case '#': return LexHash();
228 case '0': case '1': case '2': case '3': case '4':
229 case '5': case '6': case '7': case '8': case '9':
231 return LexDigitOrNegative();
232 case '=': return lltok::equal;
233 case '[': return lltok::lsquare;
234 case ']': return lltok::rsquare;
235 case '{': return lltok::lbrace;
236 case '}': return lltok::rbrace;
237 case '<': return lltok::less;
238 case '>': return lltok::greater;
239 case '(': return lltok::lparen;
240 case ')': return lltok::rparen;
241 case ',': return lltok::comma;
242 case '*': return lltok::star;
243 case '|': return lltok::bar;
248 void LLLexer::SkipLineComment() {
250 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
255 /// Lex all tokens that start with an @ character.
256 /// GlobalVar @\"[^\"]*\"
257 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
258 /// GlobalVarID @[0-9]+
259 lltok::Kind LLLexer::LexAt() {
260 return LexVar(lltok::GlobalVar, lltok::GlobalID);
263 lltok::Kind LLLexer::LexDollar() {
264 if (const char *Ptr = isLabelTail(TokStart)) {
266 StrVal.assign(TokStart, CurPtr - 1);
267 return lltok::LabelStr;
270 // Handle DollarStringConstant: $\"[^\"]*\"
271 if (CurPtr[0] == '"') {
275 int CurChar = getNextChar();
277 if (CurChar == EOF) {
278 Error("end of file in COMDAT variable name");
281 if (CurChar == '"') {
282 StrVal.assign(TokStart + 2, CurPtr - 1);
283 UnEscapeLexed(StrVal);
284 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
285 Error("Null bytes are not allowed in names");
288 return lltok::ComdatVar;
293 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
295 return lltok::ComdatVar;
300 /// ReadString - Read a string until the closing quote.
301 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
302 const char *Start = CurPtr;
304 int CurChar = getNextChar();
306 if (CurChar == EOF) {
307 Error("end of file in string constant");
310 if (CurChar == '"') {
311 StrVal.assign(Start, CurPtr-1);
312 UnEscapeLexed(StrVal);
318 /// ReadVarName - Read the rest of a token containing a variable name.
319 bool LLLexer::ReadVarName() {
320 const char *NameStart = CurPtr;
321 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
322 CurPtr[0] == '-' || CurPtr[0] == '$' ||
323 CurPtr[0] == '.' || CurPtr[0] == '_') {
325 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
326 CurPtr[0] == '-' || CurPtr[0] == '$' ||
327 CurPtr[0] == '.' || CurPtr[0] == '_')
330 StrVal.assign(NameStart, CurPtr);
336 // Lex an ID: [0-9]+. On success, the ID is stored in UIntVal and Token is
337 // returned, otherwise the Error token is returned.
338 lltok::Kind LLLexer::LexUIntID(lltok::Kind Token) {
339 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
342 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
345 uint64_t Val = atoull(TokStart + 1, CurPtr);
346 if ((unsigned)Val != Val)
347 Error("invalid value number (too large)!");
348 UIntVal = unsigned(Val);
352 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
353 // Handle StringConstant: \"[^\"]*\"
354 if (CurPtr[0] == '"') {
358 int CurChar = getNextChar();
360 if (CurChar == EOF) {
361 Error("end of file in global variable name");
364 if (CurChar == '"') {
365 StrVal.assign(TokStart+2, CurPtr-1);
366 UnEscapeLexed(StrVal);
367 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
368 Error("Null bytes are not allowed in names");
376 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
380 // Handle VarID: [0-9]+
381 return LexUIntID(VarID);
384 /// Lex all tokens that start with a % character.
385 /// LocalVar ::= %\"[^\"]*\"
386 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
387 /// LocalVarID ::= %[0-9]+
388 lltok::Kind LLLexer::LexPercent() {
389 return LexVar(lltok::LocalVar, lltok::LocalVarID);
392 /// Lex all tokens that start with a " character.
393 /// QuoteLabel "[^"]+":
394 /// StringConstant "[^"]*"
395 lltok::Kind LLLexer::LexQuote() {
396 lltok::Kind kind = ReadString(lltok::StringConstant);
397 if (kind == lltok::Error || kind == lltok::Eof)
400 if (CurPtr[0] == ':') {
402 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
403 Error("Null bytes are not allowed in names");
406 kind = lltok::LabelStr;
413 /// Lex all tokens that start with a ! character.
416 lltok::Kind LLLexer::LexExclaim() {
417 // Lex a metadata name as a MetadataVar.
418 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
419 CurPtr[0] == '-' || CurPtr[0] == '$' ||
420 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
422 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
423 CurPtr[0] == '-' || CurPtr[0] == '$' ||
424 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
427 StrVal.assign(TokStart+1, CurPtr); // Skip !
428 UnEscapeLexed(StrVal);
429 return lltok::MetadataVar;
431 return lltok::exclaim;
434 /// Lex all tokens that start with a ^ character.
435 /// SummaryID ::= ^[0-9]+
436 lltok::Kind LLLexer::LexCaret() {
437 // Handle SummaryID: ^[0-9]+
438 return LexUIntID(lltok::SummaryID);
441 /// Lex all tokens that start with a # character.
442 /// AttrGrpID ::= #[0-9]+
443 lltok::Kind LLLexer::LexHash() {
444 // Handle AttrGrpID: #[0-9]+
445 return LexUIntID(lltok::AttrGrpID);
448 /// Lex a label, integer type, keyword, or hexadecimal integer constant.
449 /// Label [-a-zA-Z$._0-9]+:
450 /// IntegerType i[0-9]+
451 /// Keyword sdiv, float, ...
452 /// HexIntConstant [us]0x[0-9A-Fa-f]+
453 lltok::Kind LLLexer::LexIdentifier() {
454 const char *StartChar = CurPtr;
455 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
456 const char *KeywordEnd = nullptr;
458 for (; isLabelChar(*CurPtr); ++CurPtr) {
459 // If we decide this is an integer, remember the end of the sequence.
460 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
462 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
467 // If we stopped due to a colon, unless we were directed to ignore it,
468 // this really is a label.
469 if (!IgnoreColonInIdentifiers && *CurPtr == ':') {
470 StrVal.assign(StartChar-1, CurPtr++);
471 return lltok::LabelStr;
474 // Otherwise, this wasn't a label. If this was valid as an integer type,
476 if (!IntEnd) IntEnd = CurPtr;
477 if (IntEnd != StartChar) {
479 uint64_t NumBits = atoull(StartChar, CurPtr);
480 if (NumBits < IntegerType::MIN_INT_BITS ||
481 NumBits > IntegerType::MAX_INT_BITS) {
482 Error("bitwidth for integer type out of range!");
485 TyVal = IntegerType::get(Context, NumBits);
489 // Otherwise, this was a letter sequence. See which keyword this is.
490 if (!KeywordEnd) KeywordEnd = CurPtr;
493 StringRef Keyword(StartChar, CurPtr - StartChar);
495 #define KEYWORD(STR) \
497 if (Keyword == #STR) \
498 return lltok::kw_##STR; \
501 KEYWORD(true); KEYWORD(false);
502 KEYWORD(declare); KEYWORD(define);
503 KEYWORD(global); KEYWORD(constant);
506 KEYWORD(dso_preemptable);
510 KEYWORD(available_externally);
512 KEYWORD(linkonce_odr);
513 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
522 KEYWORD(unnamed_addr);
523 KEYWORD(local_unnamed_addr);
524 KEYWORD(externally_initialized);
525 KEYWORD(extern_weak);
527 KEYWORD(thread_local);
528 KEYWORD(localdynamic);
529 KEYWORD(initialexec);
531 KEYWORD(zeroinitializer);
544 KEYWORD(source_filename);
546 KEYWORD(deplibs); // FIXME: Remove in 4.0.
580 KEYWORD(inteldialect);
588 KEYWORD(x86_stdcallcc);
589 KEYWORD(x86_fastcallcc);
590 KEYWORD(x86_thiscallcc);
591 KEYWORD(x86_vectorcallcc);
593 KEYWORD(arm_aapcscc);
594 KEYWORD(arm_aapcs_vfpcc);
595 KEYWORD(msp430_intrcc);
597 KEYWORD(avr_signalcc);
600 KEYWORD(spir_kernel);
602 KEYWORD(intel_ocl_bicc);
603 KEYWORD(x86_64_sysvcc);
605 KEYWORD(x86_regcallcc);
606 KEYWORD(webkit_jscc);
609 KEYWORD(preserve_mostcc);
610 KEYWORD(preserve_allcc);
615 KEYWORD(cxx_fast_tlscc);
623 KEYWORD(amdgpu_kernel);
630 KEYWORD(alwaysinline);
638 KEYWORD(dereferenceable);
639 KEYWORD(dereferenceable_or_null);
640 KEYWORD(inaccessiblememonly);
641 KEYWORD(inaccessiblemem_or_argmemonly);
651 KEYWORD(noduplicate);
652 KEYWORD(noimplicitfloat);
655 KEYWORD(nonlazybind);
661 KEYWORD(optforfuzzing);
667 KEYWORD(returns_twice);
669 KEYWORD(speculatable);
676 KEYWORD(shadowcallstack);
677 KEYWORD(sanitize_address);
678 KEYWORD(sanitize_hwaddress);
679 KEYWORD(sanitize_thread);
680 KEYWORD(sanitize_memory);
696 KEYWORD(noduplicates);
699 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
700 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
701 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
702 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
704 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
708 KEYWORD(blockaddress);
713 // Use-list order directives.
714 KEYWORD(uselistorder);
715 KEYWORD(uselistorder_bb);
717 KEYWORD(personality);
722 // Summary index keywords.
731 KEYWORD(notEligibleToImport);
740 KEYWORD(returnDoesNotAlias);
753 KEYWORD(typeTestAssumeVCalls);
754 KEYWORD(typeCheckedLoadVCalls);
755 KEYWORD(typeTestAssumeConstVCalls);
756 KEYWORD(typeCheckedLoadConstVCalls);
762 KEYWORD(typeTestRes);
769 KEYWORD(sizeM1BitWidth);
774 KEYWORD(wpdResolutions);
778 KEYWORD(branchFunnel);
779 KEYWORD(singleImplName);
782 KEYWORD(uniformRetVal);
783 KEYWORD(uniqueRetVal);
784 KEYWORD(virtualConstProp);
791 // Keywords for types.
792 #define TYPEKEYWORD(STR, LLVMTY) \
794 if (Keyword == STR) { \
796 return lltok::Type; \
800 TYPEKEYWORD("void", Type::getVoidTy(Context));
801 TYPEKEYWORD("half", Type::getHalfTy(Context));
802 TYPEKEYWORD("float", Type::getFloatTy(Context));
803 TYPEKEYWORD("double", Type::getDoubleTy(Context));
804 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
805 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
806 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
807 TYPEKEYWORD("label", Type::getLabelTy(Context));
808 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
809 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
810 TYPEKEYWORD("token", Type::getTokenTy(Context));
814 // Keywords for instructions.
815 #define INSTKEYWORD(STR, Enum) \
817 if (Keyword == #STR) { \
818 UIntVal = Instruction::Enum; \
819 return lltok::kw_##STR; \
823 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
824 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
825 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
826 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
827 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
828 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
829 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
830 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
832 INSTKEYWORD(phi, PHI);
833 INSTKEYWORD(call, Call);
834 INSTKEYWORD(trunc, Trunc);
835 INSTKEYWORD(zext, ZExt);
836 INSTKEYWORD(sext, SExt);
837 INSTKEYWORD(fptrunc, FPTrunc);
838 INSTKEYWORD(fpext, FPExt);
839 INSTKEYWORD(uitofp, UIToFP);
840 INSTKEYWORD(sitofp, SIToFP);
841 INSTKEYWORD(fptoui, FPToUI);
842 INSTKEYWORD(fptosi, FPToSI);
843 INSTKEYWORD(inttoptr, IntToPtr);
844 INSTKEYWORD(ptrtoint, PtrToInt);
845 INSTKEYWORD(bitcast, BitCast);
846 INSTKEYWORD(addrspacecast, AddrSpaceCast);
847 INSTKEYWORD(select, Select);
848 INSTKEYWORD(va_arg, VAArg);
849 INSTKEYWORD(ret, Ret);
851 INSTKEYWORD(switch, Switch);
852 INSTKEYWORD(indirectbr, IndirectBr);
853 INSTKEYWORD(invoke, Invoke);
854 INSTKEYWORD(resume, Resume);
855 INSTKEYWORD(unreachable, Unreachable);
857 INSTKEYWORD(alloca, Alloca);
858 INSTKEYWORD(load, Load);
859 INSTKEYWORD(store, Store);
860 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
861 INSTKEYWORD(atomicrmw, AtomicRMW);
862 INSTKEYWORD(fence, Fence);
863 INSTKEYWORD(getelementptr, GetElementPtr);
865 INSTKEYWORD(extractelement, ExtractElement);
866 INSTKEYWORD(insertelement, InsertElement);
867 INSTKEYWORD(shufflevector, ShuffleVector);
868 INSTKEYWORD(extractvalue, ExtractValue);
869 INSTKEYWORD(insertvalue, InsertValue);
870 INSTKEYWORD(landingpad, LandingPad);
871 INSTKEYWORD(cleanupret, CleanupRet);
872 INSTKEYWORD(catchret, CatchRet);
873 INSTKEYWORD(catchswitch, CatchSwitch);
874 INSTKEYWORD(catchpad, CatchPad);
875 INSTKEYWORD(cleanuppad, CleanupPad);
879 #define DWKEYWORD(TYPE, TOKEN) \
881 if (Keyword.startswith("DW_" #TYPE "_")) { \
882 StrVal.assign(Keyword.begin(), Keyword.end()); \
883 return lltok::TOKEN; \
887 DWKEYWORD(TAG, DwarfTag);
888 DWKEYWORD(ATE, DwarfAttEncoding);
889 DWKEYWORD(VIRTUALITY, DwarfVirtuality);
890 DWKEYWORD(LANG, DwarfLang);
891 DWKEYWORD(CC, DwarfCC);
892 DWKEYWORD(OP, DwarfOp);
893 DWKEYWORD(MACINFO, DwarfMacinfo);
897 if (Keyword.startswith("DIFlag")) {
898 StrVal.assign(Keyword.begin(), Keyword.end());
899 return lltok::DIFlag;
902 if (Keyword.startswith("CSK_")) {
903 StrVal.assign(Keyword.begin(), Keyword.end());
904 return lltok::ChecksumKind;
907 if (Keyword == "NoDebug" || Keyword == "FullDebug" ||
908 Keyword == "LineTablesOnly") {
909 StrVal.assign(Keyword.begin(), Keyword.end());
910 return lltok::EmissionKind;
913 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
914 // the CFE to avoid forcing it to deal with 64-bit numbers.
915 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
916 TokStart[1] == '0' && TokStart[2] == 'x' &&
917 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
918 int len = CurPtr-TokStart-3;
919 uint32_t bits = len * 4;
920 StringRef HexStr(TokStart + 3, len);
921 if (!all_of(HexStr, isxdigit)) {
922 // Bad token, return it as an error.
926 APInt Tmp(bits, HexStr, 16);
927 uint32_t activeBits = Tmp.getActiveBits();
928 if (activeBits > 0 && activeBits < bits)
929 Tmp = Tmp.trunc(activeBits);
930 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
931 return lltok::APSInt;
934 // If this is "cc1234", return this as just "cc".
935 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
940 // Finally, if this isn't known, return an error.
945 /// Lex all tokens that start with a 0x prefix, knowing they match and are not
947 /// HexFPConstant 0x[0-9A-Fa-f]+
948 /// HexFP80Constant 0xK[0-9A-Fa-f]+
949 /// HexFP128Constant 0xL[0-9A-Fa-f]+
950 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
951 /// HexHalfConstant 0xH[0-9A-Fa-f]+
952 lltok::Kind LLLexer::Lex0x() {
953 CurPtr = TokStart + 2;
956 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
962 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
963 // Bad token, return it as an error.
968 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
972 // HexFPConstant - Floating point constant represented in IEEE format as a
973 // hexadecimal number for when exponential notation is not precise enough.
974 // Half, Float, and double only.
975 APFloatVal = APFloat(APFloat::IEEEdouble(),
976 APInt(64, HexIntToVal(TokStart + 2, CurPtr)));
977 return lltok::APFloat;
982 default: llvm_unreachable("Unknown kind!");
984 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
985 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
986 APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair));
987 return lltok::APFloat;
989 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
990 HexToIntPair(TokStart+3, CurPtr, Pair);
991 APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair));
992 return lltok::APFloat;
994 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
995 HexToIntPair(TokStart+3, CurPtr, Pair);
996 APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair));
997 return lltok::APFloat;
999 APFloatVal = APFloat(APFloat::IEEEhalf(),
1000 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
1001 return lltok::APFloat;
1005 /// Lex tokens for a label or a numeric constant, possibly starting with -.
1006 /// Label [-a-zA-Z$._0-9]+:
1007 /// NInteger -[0-9]+
1008 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1010 /// HexFPConstant 0x[0-9A-Fa-f]+
1011 /// HexFP80Constant 0xK[0-9A-Fa-f]+
1012 /// HexFP128Constant 0xL[0-9A-Fa-f]+
1013 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
1014 lltok::Kind LLLexer::LexDigitOrNegative() {
1015 // If the letter after the negative is not a number, this is probably a label.
1016 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
1017 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
1018 // Okay, this is not a number after the -, it's probably a label.
1019 if (const char *End = isLabelTail(CurPtr)) {
1020 StrVal.assign(TokStart, End-1);
1022 return lltok::LabelStr;
1025 return lltok::Error;
1028 // At this point, it is either a label, int or fp constant.
1030 // Skip digits, we have at least one.
1031 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1034 // Check to see if this really is a label afterall, e.g. "-1:".
1035 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
1036 if (const char *End = isLabelTail(CurPtr)) {
1037 StrVal.assign(TokStart, End-1);
1039 return lltok::LabelStr;
1043 // If the next character is a '.', then it is a fp value, otherwise its
1045 if (CurPtr[0] != '.') {
1046 if (TokStart[0] == '0' && TokStart[1] == 'x')
1048 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart));
1049 return lltok::APSInt;
1054 // Skip over [0-9]*([eE][-+]?[0-9]+)?
1055 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1057 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1058 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1059 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1060 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1062 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1066 APFloatVal = APFloat(APFloat::IEEEdouble(),
1067 StringRef(TokStart, CurPtr - TokStart));
1068 return lltok::APFloat;
1071 /// Lex a floating point constant starting with +.
1072 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1073 lltok::Kind LLLexer::LexPositive() {
1074 // If the letter after the negative is a number, this is probably not a
1076 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
1077 return lltok::Error;
1080 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1083 // At this point, we need a '.'.
1084 if (CurPtr[0] != '.') {
1085 CurPtr = TokStart+1;
1086 return lltok::Error;
1091 // Skip over [0-9]*([eE][-+]?[0-9]+)?
1092 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1094 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1095 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1096 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1097 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1099 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1103 APFloatVal = APFloat(APFloat::IEEEdouble(),
1104 StringRef(TokStart, CurPtr - TokStart));
1105 return lltok::APFloat;