xref: /freebsd/contrib/llvm-project/llvm/lib/AsmParser/LLLexer.cpp (revision 700637cbb5e582861067a11aaca4d053546871d2) 
1 //===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // Implement the Lexer for .ll files.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #include "llvm/AsmParser/LLLexer.h"
14 #include "llvm/ADT/APInt.h"
15 #include "llvm/ADT/STLExtras.h"
16 #include "llvm/ADT/StringExtras.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/IR/DerivedTypes.h"
19 #include "llvm/IR/Instruction.h"
20 #include "llvm/Support/ErrorHandling.h"
21 #include "llvm/Support/SourceMgr.h"
22 #include <cassert>
23 #include <cctype>
24 #include <cstdio>
25 
26 using namespace llvm;
27 
28 // Both the lexer and parser can issue error messages. If the lexer issues a
29 // lexer error, since we do not terminate execution immediately, usually that
30 // is followed by the parser issuing a parser error. However, the error issued
31 // by the lexer is more relevant in that case as opposed to potentially more
32 // generic parser error. So instead of always recording the last error message
33 // use the `Priority` to establish a priority, with Lexer > Parser > None. We
34 // record the issued message only if the message has same or higher priority
35 // than the existing one. This prevents lexer errors from being overwritten by
36 // parser errors.
Error(LocTy ErrorLoc,const Twine & Msg,LLLexer::ErrorPriority Priority)37 void LLLexer::Error(LocTy ErrorLoc, const Twine &Msg,
38                     LLLexer::ErrorPriority Priority) {
39   if (Priority < ErrorInfo.Priority)
40     return;
41   ErrorInfo.Error = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
42   ErrorInfo.Priority = Priority;
43 }
44 
Warning(LocTy WarningLoc,const Twine & Msg) const45 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
46   SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
47 }
48 
49 //===----------------------------------------------------------------------===//
50 // Helper functions.
51 //===----------------------------------------------------------------------===//
52 
53 // atoull - Convert an ascii string of decimal digits into the unsigned long
54 // long representation... this does not have to do input error checking,
55 // because we know that the input will be matched by a suitable regex...
56 //
atoull(const char * Buffer,const char * End)57 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
58   uint64_t Result = 0;
59   for (; Buffer != End; Buffer++) {
60     uint64_t OldRes = Result;
61     Result *= 10;
62     Result += *Buffer-'0';
63     if (Result < OldRes) { // overflow detected.
64       LexError("constant bigger than 64 bits detected");
65       return 0;
66     }
67   }
68   return Result;
69 }
70 
HexIntToVal(const char * Buffer,const char * End)71 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
72   uint64_t Result = 0;
73   for (; Buffer != End; ++Buffer) {
74     uint64_t OldRes = Result;
75     Result *= 16;
76     Result += hexDigitValue(*Buffer);
77 
78     if (Result < OldRes) { // overflow detected.
79       LexError("constant bigger than 64 bits detected");
80       return 0;
81     }
82   }
83   return Result;
84 }
85 
HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])86 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
87                            uint64_t Pair[2]) {
88   Pair[0] = 0;
89   if (End - Buffer >= 16) {
90     for (int i = 0; i < 16; i++, Buffer++) {
91       assert(Buffer != End);
92       Pair[0] *= 16;
93       Pair[0] += hexDigitValue(*Buffer);
94     }
95   }
96   Pair[1] = 0;
97   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
98     Pair[1] *= 16;
99     Pair[1] += hexDigitValue(*Buffer);
100   }
101   if (Buffer != End)
102     LexError("constant bigger than 128 bits detected");
103 }
104 
105 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
106 /// { low64, high16 } as usual for an APInt.
FP80HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])107 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
108                            uint64_t Pair[2]) {
109   Pair[1] = 0;
110   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
111     assert(Buffer != End);
112     Pair[1] *= 16;
113     Pair[1] += hexDigitValue(*Buffer);
114   }
115   Pair[0] = 0;
116   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
117     Pair[0] *= 16;
118     Pair[0] += hexDigitValue(*Buffer);
119   }
120   if (Buffer != End)
121     LexError("constant bigger than 128 bits detected");
122 }
123 
124 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
125 // appropriate character.
UnEscapeLexed(std::string & Str)126 static void UnEscapeLexed(std::string &Str) {
127   if (Str.empty()) return;
128 
129   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
130   char *BOut = Buffer;
131   for (char *BIn = Buffer; BIn != EndBuffer; ) {
132     if (BIn[0] == '\\') {
133       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
134         *BOut++ = '\\'; // Two \ becomes one
135         BIn += 2;
136       } else if (BIn < EndBuffer-2 &&
137                  isxdigit(static_cast<unsigned char>(BIn[1])) &&
138                  isxdigit(static_cast<unsigned char>(BIn[2]))) {
139         *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
140         BIn += 3;                           // Skip over handled chars
141         ++BOut;
142       } else {
143         *BOut++ = *BIn++;
144       }
145     } else {
146       *BOut++ = *BIn++;
147     }
148   }
149   Str.resize(BOut-Buffer);
150 }
151 
152 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
isLabelChar(char C)153 static bool isLabelChar(char C) {
154   return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
155          C == '.' || C == '_';
156 }
157 
158 /// isLabelTail - Return true if this pointer points to a valid end of a label.
isLabelTail(const char * CurPtr)159 static const char *isLabelTail(const char *CurPtr) {
160   while (true) {
161     if (CurPtr[0] == ':') return CurPtr+1;
162     if (!isLabelChar(CurPtr[0])) return nullptr;
163     ++CurPtr;
164   }
165 }
166 
167 //===----------------------------------------------------------------------===//
168 // Lexer definition.
169 //===----------------------------------------------------------------------===//
170 
LLLexer(StringRef StartBuf,SourceMgr & SM,SMDiagnostic & Err,LLVMContext & C)171 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &SM, SMDiagnostic &Err,
172                  LLVMContext &C)
173     : CurBuf(StartBuf), ErrorInfo(Err), SM(SM), Context(C) {
174   CurPtr = CurBuf.begin();
175 }
176 
getNextChar()177 int LLLexer::getNextChar() {
178   char CurChar = *CurPtr++;
179   switch (CurChar) {
180   default: return (unsigned char)CurChar;
181   case 0:
182     // A nul character in the stream is either the end of the current buffer or
183     // a random nul in the file.  Disambiguate that here.
184     if (CurPtr-1 != CurBuf.end())
185       return 0;  // Just whitespace.
186 
187     // Otherwise, return end of file.
188     --CurPtr;  // Another call to lex will return EOF again.
189     return EOF;
190   }
191 }
192 
LexToken()193 lltok::Kind LLLexer::LexToken() {
194   while (true) {
195     TokStart = CurPtr;
196 
197     int CurChar = getNextChar();
198     switch (CurChar) {
199     default:
200       // Handle letters: [a-zA-Z_]
201       if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
202         return LexIdentifier();
203       return lltok::Error;
204     case EOF: return lltok::Eof;
205     case 0:
206     case ' ':
207     case '\t':
208     case '\n':
209     case '\r':
210       // Ignore whitespace.
211       continue;
212     case '+': return LexPositive();
213     case '@': return LexAt();
214     case '$': return LexDollar();
215     case '%': return LexPercent();
216     case '"': return LexQuote();
217     case '.':
218       if (const char *Ptr = isLabelTail(CurPtr)) {
219         CurPtr = Ptr;
220         StrVal.assign(TokStart, CurPtr-1);
221         return lltok::LabelStr;
222       }
223       if (CurPtr[0] == '.' && CurPtr[1] == '.') {
224         CurPtr += 2;
225         return lltok::dotdotdot;
226       }
227       return lltok::Error;
228     case ';':
229       SkipLineComment();
230       continue;
231     case '!': return LexExclaim();
232     case '^':
233       return LexCaret();
234     case ':':
235       return lltok::colon;
236     case '#': return LexHash();
237     case '0': case '1': case '2': case '3': case '4':
238     case '5': case '6': case '7': case '8': case '9':
239     case '-':
240       return LexDigitOrNegative();
241     case '=': return lltok::equal;
242     case '[': return lltok::lsquare;
243     case ']': return lltok::rsquare;
244     case '{': return lltok::lbrace;
245     case '}': return lltok::rbrace;
246     case '<': return lltok::less;
247     case '>': return lltok::greater;
248     case '(': return lltok::lparen;
249     case ')': return lltok::rparen;
250     case ',': return lltok::comma;
251     case '*': return lltok::star;
252     case '|': return lltok::bar;
253     case '/':
254       if (getNextChar() != '*')
255         return lltok::Error;
256       if (SkipCComment())
257         return lltok::Error;
258       continue;
259     }
260   }
261 }
262 
SkipLineComment()263 void LLLexer::SkipLineComment() {
264   while (true) {
265     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
266       return;
267   }
268 }
269 
270 /// This skips C-style /**/ comments. Returns true if there
271 /// was an error.
SkipCComment()272 bool LLLexer::SkipCComment() {
273   while (true) {
274     int CurChar = getNextChar();
275     switch (CurChar) {
276     case EOF:
277       LexError("unterminated comment");
278       return true;
279     case '*':
280       // End of the comment?
281       CurChar = getNextChar();
282       if (CurChar == '/')
283         return false;
284       if (CurChar == EOF) {
285         LexError("unterminated comment");
286         return true;
287       }
288     }
289   }
290 }
291 
292 /// Lex all tokens that start with an @ character.
293 ///   GlobalVar   @\"[^\"]*\"
294 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
295 ///   GlobalVarID @[0-9]+
LexAt()296 lltok::Kind LLLexer::LexAt() {
297   return LexVar(lltok::GlobalVar, lltok::GlobalID);
298 }
299 
LexDollar()300 lltok::Kind LLLexer::LexDollar() {
301   if (const char *Ptr = isLabelTail(TokStart)) {
302     CurPtr = Ptr;
303     StrVal.assign(TokStart, CurPtr - 1);
304     return lltok::LabelStr;
305   }
306 
307   // Handle DollarStringConstant: $\"[^\"]*\"
308   if (CurPtr[0] == '"') {
309     ++CurPtr;
310 
311     while (true) {
312       int CurChar = getNextChar();
313 
314       if (CurChar == EOF) {
315         LexError("end of file in COMDAT variable name");
316         return lltok::Error;
317       }
318       if (CurChar == '"') {
319         StrVal.assign(TokStart + 2, CurPtr - 1);
320         UnEscapeLexed(StrVal);
321         if (StringRef(StrVal).contains(0)) {
322           LexError("NUL character is not allowed in names");
323           return lltok::Error;
324         }
325         return lltok::ComdatVar;
326       }
327     }
328   }
329 
330   // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
331   if (ReadVarName())
332     return lltok::ComdatVar;
333 
334   return lltok::Error;
335 }
336 
337 /// ReadString - Read a string until the closing quote.
ReadString(lltok::Kind kind)338 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
339   const char *Start = CurPtr;
340   while (true) {
341     int CurChar = getNextChar();
342 
343     if (CurChar == EOF) {
344       LexError("end of file in string constant");
345       return lltok::Error;
346     }
347     if (CurChar == '"') {
348       StrVal.assign(Start, CurPtr-1);
349       UnEscapeLexed(StrVal);
350       return kind;
351     }
352   }
353 }
354 
355 /// ReadVarName - Read the rest of a token containing a variable name.
ReadVarName()356 bool LLLexer::ReadVarName() {
357   const char *NameStart = CurPtr;
358   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
359       CurPtr[0] == '-' || CurPtr[0] == '$' ||
360       CurPtr[0] == '.' || CurPtr[0] == '_') {
361     ++CurPtr;
362     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
363            CurPtr[0] == '-' || CurPtr[0] == '$' ||
364            CurPtr[0] == '.' || CurPtr[0] == '_')
365       ++CurPtr;
366 
367     StrVal.assign(NameStart, CurPtr);
368     return true;
369   }
370   return false;
371 }
372 
373 // Lex an ID: [0-9]+. On success, the ID is stored in UIntVal and Token is
374 // returned, otherwise the Error token is returned.
LexUIntID(lltok::Kind Token)375 lltok::Kind LLLexer::LexUIntID(lltok::Kind Token) {
376   if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
377     return lltok::Error;
378 
379   for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
380     /*empty*/;
381 
382   uint64_t Val = atoull(TokStart + 1, CurPtr);
383   if ((unsigned)Val != Val)
384     LexError("invalid value number (too large)");
385   UIntVal = unsigned(Val);
386   return Token;
387 }
388 
LexVar(lltok::Kind Var,lltok::Kind VarID)389 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
390   // Handle StringConstant: \"[^\"]*\"
391   if (CurPtr[0] == '"') {
392     ++CurPtr;
393 
394     while (true) {
395       int CurChar = getNextChar();
396 
397       if (CurChar == EOF) {
398         LexError("end of file in global variable name");
399         return lltok::Error;
400       }
401       if (CurChar == '"') {
402         StrVal.assign(TokStart+2, CurPtr-1);
403         UnEscapeLexed(StrVal);
404         if (StringRef(StrVal).contains(0)) {
405           LexError("NUL character is not allowed in names");
406           return lltok::Error;
407         }
408         return Var;
409       }
410     }
411   }
412 
413   // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
414   if (ReadVarName())
415     return Var;
416 
417   // Handle VarID: [0-9]+
418   return LexUIntID(VarID);
419 }
420 
421 /// Lex all tokens that start with a % character.
422 ///   LocalVar   ::= %\"[^\"]*\"
423 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
424 ///   LocalVarID ::= %[0-9]+
LexPercent()425 lltok::Kind LLLexer::LexPercent() {
426   return LexVar(lltok::LocalVar, lltok::LocalVarID);
427 }
428 
429 /// Lex all tokens that start with a " character.
430 ///   QuoteLabel        "[^"]+":
431 ///   StringConstant    "[^"]*"
LexQuote()432 lltok::Kind LLLexer::LexQuote() {
433   lltok::Kind kind = ReadString(lltok::StringConstant);
434   if (kind == lltok::Error || kind == lltok::Eof)
435     return kind;
436 
437   if (CurPtr[0] == ':') {
438     ++CurPtr;
439     if (StringRef(StrVal).contains(0)) {
440       LexError("NUL character is not allowed in names");
441       kind = lltok::Error;
442     } else {
443       kind = lltok::LabelStr;
444     }
445   }
446 
447   return kind;
448 }
449 
450 /// Lex all tokens that start with a ! character.
451 ///    !foo
452 ///    !
LexExclaim()453 lltok::Kind LLLexer::LexExclaim() {
454   // Lex a metadata name as a MetadataVar.
455   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
456       CurPtr[0] == '-' || CurPtr[0] == '$' ||
457       CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
458     ++CurPtr;
459     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
460            CurPtr[0] == '-' || CurPtr[0] == '$' ||
461            CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
462       ++CurPtr;
463 
464     StrVal.assign(TokStart+1, CurPtr);   // Skip !
465     UnEscapeLexed(StrVal);
466     return lltok::MetadataVar;
467   }
468   return lltok::exclaim;
469 }
470 
471 /// Lex all tokens that start with a ^ character.
472 ///    SummaryID ::= ^[0-9]+
LexCaret()473 lltok::Kind LLLexer::LexCaret() {
474   // Handle SummaryID: ^[0-9]+
475   return LexUIntID(lltok::SummaryID);
476 }
477 
478 /// Lex all tokens that start with a # character.
479 ///    AttrGrpID ::= #[0-9]+
480 ///    Hash ::= #
LexHash()481 lltok::Kind LLLexer::LexHash() {
482   // Handle AttrGrpID: #[0-9]+
483   if (isdigit(static_cast<unsigned char>(CurPtr[0])))
484     return LexUIntID(lltok::AttrGrpID);
485   return lltok::hash;
486 }
487 
488 /// Lex a label, integer type, keyword, or hexadecimal integer constant.
489 ///    Label           [-a-zA-Z$._0-9]+:
490 ///    IntegerType     i[0-9]+
491 ///    Keyword         sdiv, float, ...
492 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
LexIdentifier()493 lltok::Kind LLLexer::LexIdentifier() {
494   const char *StartChar = CurPtr;
495   const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
496   const char *KeywordEnd = nullptr;
497 
498   for (; isLabelChar(*CurPtr); ++CurPtr) {
499     // If we decide this is an integer, remember the end of the sequence.
500     if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
501       IntEnd = CurPtr;
502     if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
503         *CurPtr != '_')
504       KeywordEnd = CurPtr;
505   }
506 
507   // If we stopped due to a colon, unless we were directed to ignore it,
508   // this really is a label.
509   if (!IgnoreColonInIdentifiers && *CurPtr == ':') {
510     StrVal.assign(StartChar-1, CurPtr++);
511     return lltok::LabelStr;
512   }
513 
514   // Otherwise, this wasn't a label.  If this was valid as an integer type,
515   // return it.
516   if (!IntEnd) IntEnd = CurPtr;
517   if (IntEnd != StartChar) {
518     CurPtr = IntEnd;
519     uint64_t NumBits = atoull(StartChar, CurPtr);
520     if (NumBits < IntegerType::MIN_INT_BITS ||
521         NumBits > IntegerType::MAX_INT_BITS) {
522       LexError("bitwidth for integer type out of range");
523       return lltok::Error;
524     }
525     TyVal = IntegerType::get(Context, NumBits);
526     return lltok::Type;
527   }
528 
529   // Otherwise, this was a letter sequence.  See which keyword this is.
530   if (!KeywordEnd) KeywordEnd = CurPtr;
531   CurPtr = KeywordEnd;
532   --StartChar;
533   StringRef Keyword(StartChar, CurPtr - StartChar);
534 
535 #define KEYWORD(STR)                                                           \
536   do {                                                                         \
537     if (Keyword == #STR)                                                       \
538       return lltok::kw_##STR;                                                  \
539   } while (false)
540 
541   KEYWORD(true);    KEYWORD(false);
542   KEYWORD(declare); KEYWORD(define);
543   KEYWORD(global);  KEYWORD(constant);
544 
545   KEYWORD(dso_local);
546   KEYWORD(dso_preemptable);
547 
548   KEYWORD(private);
549   KEYWORD(internal);
550   KEYWORD(available_externally);
551   KEYWORD(linkonce);
552   KEYWORD(linkonce_odr);
553   KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
554   KEYWORD(weak_odr);
555   KEYWORD(appending);
556   KEYWORD(dllimport);
557   KEYWORD(dllexport);
558   KEYWORD(common);
559   KEYWORD(default);
560   KEYWORD(hidden);
561   KEYWORD(protected);
562   KEYWORD(unnamed_addr);
563   KEYWORD(local_unnamed_addr);
564   KEYWORD(externally_initialized);
565   KEYWORD(extern_weak);
566   KEYWORD(external);
567   KEYWORD(thread_local);
568   KEYWORD(localdynamic);
569   KEYWORD(initialexec);
570   KEYWORD(localexec);
571   KEYWORD(zeroinitializer);
572   KEYWORD(undef);
573   KEYWORD(null);
574   KEYWORD(none);
575   KEYWORD(poison);
576   KEYWORD(to);
577   KEYWORD(caller);
578   KEYWORD(within);
579   KEYWORD(from);
580   KEYWORD(tail);
581   KEYWORD(musttail);
582   KEYWORD(notail);
583   KEYWORD(target);
584   KEYWORD(triple);
585   KEYWORD(source_filename);
586   KEYWORD(unwind);
587   KEYWORD(datalayout);
588   KEYWORD(volatile);
589   KEYWORD(atomic);
590   KEYWORD(unordered);
591   KEYWORD(monotonic);
592   KEYWORD(acquire);
593   KEYWORD(release);
594   KEYWORD(acq_rel);
595   KEYWORD(seq_cst);
596   KEYWORD(syncscope);
597 
598   KEYWORD(nnan);
599   KEYWORD(ninf);
600   KEYWORD(nsz);
601   KEYWORD(arcp);
602   KEYWORD(contract);
603   KEYWORD(reassoc);
604   KEYWORD(afn);
605   KEYWORD(fast);
606   KEYWORD(nuw);
607   KEYWORD(nsw);
608   KEYWORD(nusw);
609   KEYWORD(exact);
610   KEYWORD(disjoint);
611   KEYWORD(inbounds);
612   KEYWORD(nneg);
613   KEYWORD(samesign);
614   KEYWORD(inrange);
615   KEYWORD(addrspace);
616   KEYWORD(section);
617   KEYWORD(partition);
618   KEYWORD(code_model);
619   KEYWORD(alias);
620   KEYWORD(ifunc);
621   KEYWORD(module);
622   KEYWORD(asm);
623   KEYWORD(sideeffect);
624   KEYWORD(inteldialect);
625   KEYWORD(gc);
626   KEYWORD(prefix);
627   KEYWORD(prologue);
628 
629   KEYWORD(no_sanitize_address);
630   KEYWORD(no_sanitize_hwaddress);
631   KEYWORD(sanitize_address_dyninit);
632 
633   KEYWORD(ccc);
634   KEYWORD(fastcc);
635   KEYWORD(coldcc);
636   KEYWORD(cfguard_checkcc);
637   KEYWORD(x86_stdcallcc);
638   KEYWORD(x86_fastcallcc);
639   KEYWORD(x86_thiscallcc);
640   KEYWORD(x86_vectorcallcc);
641   KEYWORD(arm_apcscc);
642   KEYWORD(arm_aapcscc);
643   KEYWORD(arm_aapcs_vfpcc);
644   KEYWORD(aarch64_vector_pcs);
645   KEYWORD(aarch64_sve_vector_pcs);
646   KEYWORD(aarch64_sme_preservemost_from_x0);
647   KEYWORD(aarch64_sme_preservemost_from_x1);
648   KEYWORD(aarch64_sme_preservemost_from_x2);
649   KEYWORD(msp430_intrcc);
650   KEYWORD(avr_intrcc);
651   KEYWORD(avr_signalcc);
652   KEYWORD(ptx_kernel);
653   KEYWORD(ptx_device);
654   KEYWORD(spir_kernel);
655   KEYWORD(spir_func);
656   KEYWORD(intel_ocl_bicc);
657   KEYWORD(x86_64_sysvcc);
658   KEYWORD(win64cc);
659   KEYWORD(x86_regcallcc);
660   KEYWORD(swiftcc);
661   KEYWORD(swifttailcc);
662   KEYWORD(anyregcc);
663   KEYWORD(preserve_mostcc);
664   KEYWORD(preserve_allcc);
665   KEYWORD(preserve_nonecc);
666   KEYWORD(ghccc);
667   KEYWORD(x86_intrcc);
668   KEYWORD(hhvmcc);
669   KEYWORD(hhvm_ccc);
670   KEYWORD(cxx_fast_tlscc);
671   KEYWORD(amdgpu_vs);
672   KEYWORD(amdgpu_ls);
673   KEYWORD(amdgpu_hs);
674   KEYWORD(amdgpu_es);
675   KEYWORD(amdgpu_gs);
676   KEYWORD(amdgpu_ps);
677   KEYWORD(amdgpu_cs);
678   KEYWORD(amdgpu_cs_chain);
679   KEYWORD(amdgpu_cs_chain_preserve);
680   KEYWORD(amdgpu_kernel);
681   KEYWORD(amdgpu_gfx);
682   KEYWORD(tailcc);
683   KEYWORD(m68k_rtdcc);
684   KEYWORD(graalcc);
685   KEYWORD(riscv_vector_cc);
686   KEYWORD(riscv_vls_cc);
687 
688   KEYWORD(cc);
689   KEYWORD(c);
690 
691   KEYWORD(attributes);
692   KEYWORD(sync);
693   KEYWORD(async);
694 
695 #define GET_ATTR_NAMES
696 #define ATTRIBUTE_ENUM(ENUM_NAME, DISPLAY_NAME) \
697   KEYWORD(DISPLAY_NAME);
698 #include "llvm/IR/Attributes.inc"
699 
700   KEYWORD(read);
701   KEYWORD(write);
702   KEYWORD(readwrite);
703   KEYWORD(argmem);
704   KEYWORD(inaccessiblemem);
705   KEYWORD(errnomem);
706   KEYWORD(argmemonly);
707   KEYWORD(inaccessiblememonly);
708   KEYWORD(inaccessiblemem_or_argmemonly);
709   KEYWORD(nocapture);
710   KEYWORD(address_is_null);
711   KEYWORD(address);
712   KEYWORD(provenance);
713   KEYWORD(read_provenance);
714 
715   // nofpclass attribute
716   KEYWORD(all);
717   KEYWORD(nan);
718   KEYWORD(snan);
719   KEYWORD(qnan);
720   KEYWORD(inf);
721   // ninf already a keyword
722   KEYWORD(pinf);
723   KEYWORD(norm);
724   KEYWORD(nnorm);
725   KEYWORD(pnorm);
726   // sub already a keyword
727   KEYWORD(nsub);
728   KEYWORD(psub);
729   KEYWORD(zero);
730   KEYWORD(nzero);
731   KEYWORD(pzero);
732 
733   KEYWORD(type);
734   KEYWORD(opaque);
735 
736   KEYWORD(comdat);
737 
738   // Comdat types
739   KEYWORD(any);
740   KEYWORD(exactmatch);
741   KEYWORD(largest);
742   KEYWORD(nodeduplicate);
743   KEYWORD(samesize);
744 
745   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
746   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
747   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
748   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
749 
750   KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
751   KEYWORD(umin); KEYWORD(fmax); KEYWORD(fmin);
752   KEYWORD(fmaximum);
753   KEYWORD(fminimum);
754   KEYWORD(uinc_wrap);
755   KEYWORD(udec_wrap);
756   KEYWORD(usub_cond);
757   KEYWORD(usub_sat);
758 
759   KEYWORD(splat);
760   KEYWORD(vscale);
761   KEYWORD(x);
762   KEYWORD(blockaddress);
763   KEYWORD(dso_local_equivalent);
764   KEYWORD(no_cfi);
765   KEYWORD(ptrauth);
766 
767   // Metadata types.
768   KEYWORD(distinct);
769 
770   // Use-list order directives.
771   KEYWORD(uselistorder);
772   KEYWORD(uselistorder_bb);
773 
774   KEYWORD(personality);
775   KEYWORD(cleanup);
776   KEYWORD(catch);
777   KEYWORD(filter);
778 
779   // Summary index keywords.
780   KEYWORD(path);
781   KEYWORD(hash);
782   KEYWORD(gv);
783   KEYWORD(guid);
784   KEYWORD(name);
785   KEYWORD(summaries);
786   KEYWORD(flags);
787   KEYWORD(blockcount);
788   KEYWORD(linkage);
789   KEYWORD(visibility);
790   KEYWORD(notEligibleToImport);
791   KEYWORD(live);
792   KEYWORD(dsoLocal);
793   KEYWORD(canAutoHide);
794   KEYWORD(importType);
795   KEYWORD(definition);
796   KEYWORD(declaration);
797   KEYWORD(function);
798   KEYWORD(insts);
799   KEYWORD(funcFlags);
800   KEYWORD(readNone);
801   KEYWORD(readOnly);
802   KEYWORD(noRecurse);
803   KEYWORD(returnDoesNotAlias);
804   KEYWORD(noInline);
805   KEYWORD(alwaysInline);
806   KEYWORD(noUnwind);
807   KEYWORD(mayThrow);
808   KEYWORD(hasUnknownCall);
809   KEYWORD(mustBeUnreachable);
810   KEYWORD(calls);
811   KEYWORD(callee);
812   KEYWORD(params);
813   KEYWORD(param);
814   KEYWORD(hotness);
815   KEYWORD(unknown);
816   KEYWORD(critical);
817   KEYWORD(relbf);
818   KEYWORD(variable);
819   KEYWORD(vTableFuncs);
820   KEYWORD(virtFunc);
821   KEYWORD(aliasee);
822   KEYWORD(refs);
823   KEYWORD(typeIdInfo);
824   KEYWORD(typeTests);
825   KEYWORD(typeTestAssumeVCalls);
826   KEYWORD(typeCheckedLoadVCalls);
827   KEYWORD(typeTestAssumeConstVCalls);
828   KEYWORD(typeCheckedLoadConstVCalls);
829   KEYWORD(vFuncId);
830   KEYWORD(offset);
831   KEYWORD(args);
832   KEYWORD(typeid);
833   KEYWORD(typeidCompatibleVTable);
834   KEYWORD(summary);
835   KEYWORD(typeTestRes);
836   KEYWORD(kind);
837   KEYWORD(unsat);
838   KEYWORD(byteArray);
839   KEYWORD(inline);
840   KEYWORD(single);
841   KEYWORD(allOnes);
842   KEYWORD(sizeM1BitWidth);
843   KEYWORD(alignLog2);
844   KEYWORD(sizeM1);
845   KEYWORD(bitMask);
846   KEYWORD(inlineBits);
847   KEYWORD(vcall_visibility);
848   KEYWORD(wpdResolutions);
849   KEYWORD(wpdRes);
850   KEYWORD(indir);
851   KEYWORD(singleImpl);
852   KEYWORD(branchFunnel);
853   KEYWORD(singleImplName);
854   KEYWORD(resByArg);
855   KEYWORD(byArg);
856   KEYWORD(uniformRetVal);
857   KEYWORD(uniqueRetVal);
858   KEYWORD(virtualConstProp);
859   KEYWORD(info);
860   KEYWORD(byte);
861   KEYWORD(bit);
862   KEYWORD(varFlags);
863   KEYWORD(callsites);
864   KEYWORD(clones);
865   KEYWORD(stackIds);
866   KEYWORD(allocs);
867   KEYWORD(versions);
868   KEYWORD(memProf);
869   KEYWORD(notcold);
870 
871 #undef KEYWORD
872 
873   // Keywords for types.
874 #define TYPEKEYWORD(STR, LLVMTY)                                               \
875   do {                                                                         \
876     if (Keyword == STR) {                                                      \
877       TyVal = LLVMTY;                                                          \
878       return lltok::Type;                                                      \
879     }                                                                          \
880   } while (false)
881 
882   TYPEKEYWORD("void",      Type::getVoidTy(Context));
883   TYPEKEYWORD("half",      Type::getHalfTy(Context));
884   TYPEKEYWORD("bfloat",    Type::getBFloatTy(Context));
885   TYPEKEYWORD("float",     Type::getFloatTy(Context));
886   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
887   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
888   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
889   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
890   TYPEKEYWORD("label",     Type::getLabelTy(Context));
891   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
892   TYPEKEYWORD("x86_amx",   Type::getX86_AMXTy(Context));
893   TYPEKEYWORD("token",     Type::getTokenTy(Context));
894   TYPEKEYWORD("ptr",       PointerType::getUnqual(Context));
895 
896 #undef TYPEKEYWORD
897 
898   // Keywords for instructions.
899 #define INSTKEYWORD(STR, Enum)                                                 \
900   do {                                                                         \
901     if (Keyword == #STR) {                                                     \
902       UIntVal = Instruction::Enum;                                             \
903       return lltok::kw_##STR;                                                  \
904     }                                                                          \
905   } while (false)
906 
907   INSTKEYWORD(fneg,  FNeg);
908 
909   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
910   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
911   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
912   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
913   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
914   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
915   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
916   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
917 
918   INSTKEYWORD(phi,         PHI);
919   INSTKEYWORD(call,        Call);
920   INSTKEYWORD(trunc,       Trunc);
921   INSTKEYWORD(zext,        ZExt);
922   INSTKEYWORD(sext,        SExt);
923   INSTKEYWORD(fptrunc,     FPTrunc);
924   INSTKEYWORD(fpext,       FPExt);
925   INSTKEYWORD(uitofp,      UIToFP);
926   INSTKEYWORD(sitofp,      SIToFP);
927   INSTKEYWORD(fptoui,      FPToUI);
928   INSTKEYWORD(fptosi,      FPToSI);
929   INSTKEYWORD(inttoptr,    IntToPtr);
930   INSTKEYWORD(ptrtoint,    PtrToInt);
931   INSTKEYWORD(bitcast,     BitCast);
932   INSTKEYWORD(addrspacecast, AddrSpaceCast);
933   INSTKEYWORD(select,      Select);
934   INSTKEYWORD(va_arg,      VAArg);
935   INSTKEYWORD(ret,         Ret);
936   INSTKEYWORD(br,          Br);
937   INSTKEYWORD(switch,      Switch);
938   INSTKEYWORD(indirectbr,  IndirectBr);
939   INSTKEYWORD(invoke,      Invoke);
940   INSTKEYWORD(resume,      Resume);
941   INSTKEYWORD(unreachable, Unreachable);
942   INSTKEYWORD(callbr,      CallBr);
943 
944   INSTKEYWORD(alloca,      Alloca);
945   INSTKEYWORD(load,        Load);
946   INSTKEYWORD(store,       Store);
947   INSTKEYWORD(cmpxchg,     AtomicCmpXchg);
948   INSTKEYWORD(atomicrmw,   AtomicRMW);
949   INSTKEYWORD(fence,       Fence);
950   INSTKEYWORD(getelementptr, GetElementPtr);
951 
952   INSTKEYWORD(extractelement, ExtractElement);
953   INSTKEYWORD(insertelement,  InsertElement);
954   INSTKEYWORD(shufflevector,  ShuffleVector);
955   INSTKEYWORD(extractvalue,   ExtractValue);
956   INSTKEYWORD(insertvalue,    InsertValue);
957   INSTKEYWORD(landingpad,     LandingPad);
958   INSTKEYWORD(cleanupret,     CleanupRet);
959   INSTKEYWORD(catchret,       CatchRet);
960   INSTKEYWORD(catchswitch,  CatchSwitch);
961   INSTKEYWORD(catchpad,     CatchPad);
962   INSTKEYWORD(cleanuppad,   CleanupPad);
963 
964   INSTKEYWORD(freeze,       Freeze);
965 
966 #undef INSTKEYWORD
967 
968 #define DWKEYWORD(TYPE, TOKEN)                                                 \
969   do {                                                                         \
970     if (Keyword.starts_with("DW_" #TYPE "_")) {                                \
971       StrVal.assign(Keyword.begin(), Keyword.end());                           \
972       return lltok::TOKEN;                                                     \
973     }                                                                          \
974   } while (false)
975 
976   DWKEYWORD(TAG, DwarfTag);
977   DWKEYWORD(ATE, DwarfAttEncoding);
978   DWKEYWORD(VIRTUALITY, DwarfVirtuality);
979   DWKEYWORD(LANG, DwarfLang);
980   DWKEYWORD(CC, DwarfCC);
981   DWKEYWORD(OP, DwarfOp);
982   DWKEYWORD(MACINFO, DwarfMacinfo);
983   DWKEYWORD(APPLE_ENUM_KIND, DwarfEnumKind);
984 
985 #undef DWKEYWORD
986 
987 // Keywords for debug record types.
988 #define DBGRECORDTYPEKEYWORD(STR)                                              \
989   do {                                                                         \
990     if (Keyword == "dbg_" #STR) {                                              \
991       StrVal = #STR;                                                           \
992       return lltok::DbgRecordType;                                             \
993     }                                                                          \
994   } while (false)
995 
996   DBGRECORDTYPEKEYWORD(value);
997   DBGRECORDTYPEKEYWORD(declare);
998   DBGRECORDTYPEKEYWORD(assign);
999   DBGRECORDTYPEKEYWORD(label);
1000 #undef DBGRECORDTYPEKEYWORD
1001 
1002   if (Keyword.starts_with("DIFlag")) {
1003     StrVal.assign(Keyword.begin(), Keyword.end());
1004     return lltok::DIFlag;
1005   }
1006 
1007   if (Keyword.starts_with("DISPFlag")) {
1008     StrVal.assign(Keyword.begin(), Keyword.end());
1009     return lltok::DISPFlag;
1010   }
1011 
1012   if (Keyword.starts_with("CSK_")) {
1013     StrVal.assign(Keyword.begin(), Keyword.end());
1014     return lltok::ChecksumKind;
1015   }
1016 
1017   if (Keyword == "NoDebug" || Keyword == "FullDebug" ||
1018       Keyword == "LineTablesOnly" || Keyword == "DebugDirectivesOnly") {
1019     StrVal.assign(Keyword.begin(), Keyword.end());
1020     return lltok::EmissionKind;
1021   }
1022 
1023   if (Keyword == "GNU" || Keyword == "Apple" || Keyword == "None" ||
1024       Keyword == "Default") {
1025     StrVal.assign(Keyword.begin(), Keyword.end());
1026     return lltok::NameTableKind;
1027   }
1028 
1029   if (Keyword == "Binary" || Keyword == "Decimal" || Keyword == "Rational") {
1030     StrVal.assign(Keyword.begin(), Keyword.end());
1031     return lltok::FixedPointKind;
1032   }
1033 
1034   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
1035   // the CFE to avoid forcing it to deal with 64-bit numbers.
1036   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
1037       TokStart[1] == '0' && TokStart[2] == 'x' &&
1038       isxdigit(static_cast<unsigned char>(TokStart[3]))) {
1039     int len = CurPtr-TokStart-3;
1040     uint32_t bits = len * 4;
1041     StringRef HexStr(TokStart + 3, len);
1042     if (!all_of(HexStr, isxdigit)) {
1043       // Bad token, return it as an error.
1044       CurPtr = TokStart+3;
1045       return lltok::Error;
1046     }
1047     APInt Tmp(bits, HexStr, 16);
1048     uint32_t activeBits = Tmp.getActiveBits();
1049     if (activeBits > 0 && activeBits < bits)
1050       Tmp = Tmp.trunc(activeBits);
1051     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
1052     return lltok::APSInt;
1053   }
1054 
1055   // If this is "cc1234", return this as just "cc".
1056   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
1057     CurPtr = TokStart+2;
1058     return lltok::kw_cc;
1059   }
1060 
1061   // Finally, if this isn't known, return an error.
1062   CurPtr = TokStart+1;
1063   return lltok::Error;
1064 }
1065 
1066 /// Lex all tokens that start with a 0x prefix, knowing they match and are not
1067 /// labels.
1068 ///    HexFPConstant     0x[0-9A-Fa-f]+
1069 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
1070 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
1071 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
1072 ///    HexHalfConstant   0xH[0-9A-Fa-f]+
1073 ///    HexBFloatConstant 0xR[0-9A-Fa-f]+
Lex0x()1074 lltok::Kind LLLexer::Lex0x() {
1075   CurPtr = TokStart + 2;
1076 
1077   char Kind;
1078   if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H' ||
1079       CurPtr[0] == 'R') {
1080     Kind = *CurPtr++;
1081   } else {
1082     Kind = 'J';
1083   }
1084 
1085   if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
1086     // Bad token, return it as an error.
1087     CurPtr = TokStart+1;
1088     return lltok::Error;
1089   }
1090 
1091   while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
1092     ++CurPtr;
1093 
1094   if (Kind == 'J') {
1095     // HexFPConstant - Floating point constant represented in IEEE format as a
1096     // hexadecimal number for when exponential notation is not precise enough.
1097     // Half, BFloat, Float, and double only.
1098     APFloatVal = APFloat(APFloat::IEEEdouble(),
1099                          APInt(64, HexIntToVal(TokStart + 2, CurPtr)));
1100     return lltok::APFloat;
1101   }
1102 
1103   uint64_t Pair[2];
1104   switch (Kind) {
1105   default: llvm_unreachable("Unknown kind!");
1106   case 'K':
1107     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
1108     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
1109     APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair));
1110     return lltok::APFloat;
1111   case 'L':
1112     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
1113     HexToIntPair(TokStart+3, CurPtr, Pair);
1114     APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair));
1115     return lltok::APFloat;
1116   case 'M':
1117     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
1118     HexToIntPair(TokStart+3, CurPtr, Pair);
1119     APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair));
1120     return lltok::APFloat;
1121   case 'H':
1122     APFloatVal = APFloat(APFloat::IEEEhalf(),
1123                          APInt(16,HexIntToVal(TokStart+3, CurPtr)));
1124     return lltok::APFloat;
1125   case 'R':
1126     // Brain floating point
1127     APFloatVal = APFloat(APFloat::BFloat(),
1128                          APInt(16, HexIntToVal(TokStart + 3, CurPtr)));
1129     return lltok::APFloat;
1130   }
1131 }
1132 
1133 /// Lex tokens for a label or a numeric constant, possibly starting with -.
1134 ///    Label             [-a-zA-Z$._0-9]+:
1135 ///    NInteger          -[0-9]+
1136 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1137 ///    PInteger          [0-9]+
1138 ///    HexFPConstant     0x[0-9A-Fa-f]+
1139 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
1140 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
1141 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
LexDigitOrNegative()1142 lltok::Kind LLLexer::LexDigitOrNegative() {
1143   // If the letter after the negative is not a number, this is probably a label.
1144   if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
1145       !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
1146     // Okay, this is not a number after the -, it's probably a label.
1147     if (const char *End = isLabelTail(CurPtr)) {
1148       StrVal.assign(TokStart, End-1);
1149       CurPtr = End;
1150       return lltok::LabelStr;
1151     }
1152 
1153     return lltok::Error;
1154   }
1155 
1156   // At this point, it is either a label, int or fp constant.
1157 
1158   // Skip digits, we have at least one.
1159   for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1160     /*empty*/;
1161 
1162   // Check if this is a fully-numeric label:
1163   if (isdigit(TokStart[0]) && CurPtr[0] == ':') {
1164     uint64_t Val = atoull(TokStart, CurPtr);
1165     ++CurPtr; // Skip the colon.
1166     if ((unsigned)Val != Val)
1167       LexError("invalid value number (too large)");
1168     UIntVal = unsigned(Val);
1169     return lltok::LabelID;
1170   }
1171 
1172   // Check to see if this really is a string label, e.g. "-1:".
1173   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
1174     if (const char *End = isLabelTail(CurPtr)) {
1175       StrVal.assign(TokStart, End-1);
1176       CurPtr = End;
1177       return lltok::LabelStr;
1178     }
1179   }
1180 
1181   // If the next character is a '.', then it is a fp value, otherwise its
1182   // integer.
1183   if (CurPtr[0] != '.') {
1184     if (TokStart[0] == '0' && TokStart[1] == 'x')
1185       return Lex0x();
1186     APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart));
1187     return lltok::APSInt;
1188   }
1189 
1190   ++CurPtr;
1191 
1192   // Skip over [0-9]*([eE][-+]?[0-9]+)?
1193   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1194 
1195   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1196     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1197         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1198           isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1199       CurPtr += 2;
1200       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1201     }
1202   }
1203 
1204   APFloatVal = APFloat(APFloat::IEEEdouble(),
1205                        StringRef(TokStart, CurPtr - TokStart));
1206   return lltok::APFloat;
1207 }
1208 
1209 /// Lex a floating point constant starting with +.
1210 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
LexPositive()1211 lltok::Kind LLLexer::LexPositive() {
1212   // If the letter after the negative is a number, this is probably not a
1213   // label.
1214   if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
1215     return lltok::Error;
1216 
1217   // Skip digits.
1218   for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1219     /*empty*/;
1220 
1221   // At this point, we need a '.'.
1222   if (CurPtr[0] != '.') {
1223     CurPtr = TokStart+1;
1224     return lltok::Error;
1225   }
1226 
1227   ++CurPtr;
1228 
1229   // Skip over [0-9]*([eE][-+]?[0-9]+)?
1230   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1231 
1232   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1233     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1234         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1235         isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1236       CurPtr += 2;
1237       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1238     }
1239   }
1240 
1241   APFloatVal = APFloat(APFloat::IEEEdouble(),
1242                        StringRef(TokStart, CurPtr - TokStart));
1243   return lltok::APFloat;
1244 }
1245