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 "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 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
29   ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
30   return true;
31 }
32 
33 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
34   SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
35 }
36 
37 //===----------------------------------------------------------------------===//
38 // Helper functions.
39 //===----------------------------------------------------------------------===//
40 
41 // atoull - Convert an ascii string of decimal digits into the unsigned long
42 // long representation... this does not have to do input error checking,
43 // because we know that the input will be matched by a suitable regex...
44 //
45 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
46   uint64_t Result = 0;
47   for (; Buffer != End; Buffer++) {
48     uint64_t OldRes = Result;
49     Result *= 10;
50     Result += *Buffer-'0';
51     if (Result < OldRes) {  // Uh, oh, overflow detected!!!
52       Error("constant bigger than 64 bits detected!");
53       return 0;
54     }
55   }
56   return Result;
57 }
58 
59 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
60   uint64_t Result = 0;
61   for (; Buffer != End; ++Buffer) {
62     uint64_t OldRes = Result;
63     Result *= 16;
64     Result += hexDigitValue(*Buffer);
65 
66     if (Result < OldRes) {   // Uh, oh, overflow detected!!!
67       Error("constant bigger than 64 bits detected!");
68       return 0;
69     }
70   }
71   return Result;
72 }
73 
74 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
75                            uint64_t Pair[2]) {
76   Pair[0] = 0;
77   if (End - Buffer >= 16) {
78     for (int i = 0; i < 16; i++, Buffer++) {
79       assert(Buffer != End);
80       Pair[0] *= 16;
81       Pair[0] += hexDigitValue(*Buffer);
82     }
83   }
84   Pair[1] = 0;
85   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
86     Pair[1] *= 16;
87     Pair[1] += hexDigitValue(*Buffer);
88   }
89   if (Buffer != End)
90     Error("constant bigger than 128 bits detected!");
91 }
92 
93 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
94 /// { low64, high16 } as usual for an APInt.
95 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
96                            uint64_t Pair[2]) {
97   Pair[1] = 0;
98   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
99     assert(Buffer != End);
100     Pair[1] *= 16;
101     Pair[1] += hexDigitValue(*Buffer);
102   }
103   Pair[0] = 0;
104   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
105     Pair[0] *= 16;
106     Pair[0] += hexDigitValue(*Buffer);
107   }
108   if (Buffer != End)
109     Error("constant bigger than 128 bits detected!");
110 }
111 
112 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
113 // appropriate character.
114 static void UnEscapeLexed(std::string &Str) {
115   if (Str.empty()) return;
116 
117   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
118   char *BOut = Buffer;
119   for (char *BIn = Buffer; BIn != EndBuffer; ) {
120     if (BIn[0] == '\\') {
121       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
122         *BOut++ = '\\'; // Two \ becomes one
123         BIn += 2;
124       } else if (BIn < EndBuffer-2 &&
125                  isxdigit(static_cast<unsigned char>(BIn[1])) &&
126                  isxdigit(static_cast<unsigned char>(BIn[2]))) {
127         *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
128         BIn += 3;                           // Skip over handled chars
129         ++BOut;
130       } else {
131         *BOut++ = *BIn++;
132       }
133     } else {
134       *BOut++ = *BIn++;
135     }
136   }
137   Str.resize(BOut-Buffer);
138 }
139 
140 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
141 static bool isLabelChar(char C) {
142   return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
143          C == '.' || C == '_';
144 }
145 
146 /// isLabelTail - Return true if this pointer points to a valid end of a label.
147 static const char *isLabelTail(const char *CurPtr) {
148   while (true) {
149     if (CurPtr[0] == ':') return CurPtr+1;
150     if (!isLabelChar(CurPtr[0])) return nullptr;
151     ++CurPtr;
152   }
153 }
154 
155 //===----------------------------------------------------------------------===//
156 // Lexer definition.
157 //===----------------------------------------------------------------------===//
158 
159 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &SM, SMDiagnostic &Err,
160                  LLVMContext &C)
161     : CurBuf(StartBuf), ErrorInfo(Err), SM(SM), Context(C), APFloatVal(0.0),
162       IgnoreColonInIdentifiers(false) {
163   CurPtr = CurBuf.begin();
164 }
165 
166 int LLLexer::getNextChar() {
167   char CurChar = *CurPtr++;
168   switch (CurChar) {
169   default: return (unsigned char)CurChar;
170   case 0:
171     // A nul character in the stream is either the end of the current buffer or
172     // a random nul in the file.  Disambiguate that here.
173     if (CurPtr-1 != CurBuf.end())
174       return 0;  // Just whitespace.
175 
176     // Otherwise, return end of file.
177     --CurPtr;  // Another call to lex will return EOF again.
178     return EOF;
179   }
180 }
181 
182 lltok::Kind LLLexer::LexToken() {
183   while (true) {
184     TokStart = CurPtr;
185 
186     int CurChar = getNextChar();
187     switch (CurChar) {
188     default:
189       // Handle letters: [a-zA-Z_]
190       if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
191         return LexIdentifier();
192 
193       return lltok::Error;
194     case EOF: return lltok::Eof;
195     case 0:
196     case ' ':
197     case '\t':
198     case '\n':
199     case '\r':
200       // Ignore whitespace.
201       continue;
202     case '+': return LexPositive();
203     case '@': return LexAt();
204     case '$': return LexDollar();
205     case '%': return LexPercent();
206     case '"': return LexQuote();
207     case '.':
208       if (const char *Ptr = isLabelTail(CurPtr)) {
209         CurPtr = Ptr;
210         StrVal.assign(TokStart, CurPtr-1);
211         return lltok::LabelStr;
212       }
213       if (CurPtr[0] == '.' && CurPtr[1] == '.') {
214         CurPtr += 2;
215         return lltok::dotdotdot;
216       }
217       return lltok::Error;
218     case ';':
219       SkipLineComment();
220       continue;
221     case '!': return LexExclaim();
222     case '^':
223       return LexCaret();
224     case ':':
225       return lltok::colon;
226     case '#': return LexHash();
227     case '0': case '1': case '2': case '3': case '4':
228     case '5': case '6': case '7': case '8': case '9':
229     case '-':
230       return LexDigitOrNegative();
231     case '=': return lltok::equal;
232     case '[': return lltok::lsquare;
233     case ']': return lltok::rsquare;
234     case '{': return lltok::lbrace;
235     case '}': return lltok::rbrace;
236     case '<': return lltok::less;
237     case '>': return lltok::greater;
238     case '(': return lltok::lparen;
239     case ')': return lltok::rparen;
240     case ',': return lltok::comma;
241     case '*': return lltok::star;
242     case '|': return lltok::bar;
243     }
244   }
245 }
246 
247 void LLLexer::SkipLineComment() {
248   while (true) {
249     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
250       return;
251   }
252 }
253 
254 /// Lex all tokens that start with an @ character.
255 ///   GlobalVar   @\"[^\"]*\"
256 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
257 ///   GlobalVarID @[0-9]+
258 lltok::Kind LLLexer::LexAt() {
259   return LexVar(lltok::GlobalVar, lltok::GlobalID);
260 }
261 
262 lltok::Kind LLLexer::LexDollar() {
263   if (const char *Ptr = isLabelTail(TokStart)) {
264     CurPtr = Ptr;
265     StrVal.assign(TokStart, CurPtr - 1);
266     return lltok::LabelStr;
267   }
268 
269   // Handle DollarStringConstant: $\"[^\"]*\"
270   if (CurPtr[0] == '"') {
271     ++CurPtr;
272 
273     while (true) {
274       int CurChar = getNextChar();
275 
276       if (CurChar == EOF) {
277         Error("end of file in COMDAT variable name");
278         return lltok::Error;
279       }
280       if (CurChar == '"') {
281         StrVal.assign(TokStart + 2, CurPtr - 1);
282         UnEscapeLexed(StrVal);
283         if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
284           Error("Null bytes are not allowed in names");
285           return lltok::Error;
286         }
287         return lltok::ComdatVar;
288       }
289     }
290   }
291 
292   // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
293   if (ReadVarName())
294     return lltok::ComdatVar;
295 
296   return lltok::Error;
297 }
298 
299 /// ReadString - Read a string until the closing quote.
300 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
301   const char *Start = CurPtr;
302   while (true) {
303     int CurChar = getNextChar();
304 
305     if (CurChar == EOF) {
306       Error("end of file in string constant");
307       return lltok::Error;
308     }
309     if (CurChar == '"') {
310       StrVal.assign(Start, CurPtr-1);
311       UnEscapeLexed(StrVal);
312       return kind;
313     }
314   }
315 }
316 
317 /// ReadVarName - Read the rest of a token containing a variable name.
318 bool LLLexer::ReadVarName() {
319   const char *NameStart = CurPtr;
320   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
321       CurPtr[0] == '-' || CurPtr[0] == '$' ||
322       CurPtr[0] == '.' || CurPtr[0] == '_') {
323     ++CurPtr;
324     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
325            CurPtr[0] == '-' || CurPtr[0] == '$' ||
326            CurPtr[0] == '.' || CurPtr[0] == '_')
327       ++CurPtr;
328 
329     StrVal.assign(NameStart, CurPtr);
330     return true;
331   }
332   return false;
333 }
334 
335 // Lex an ID: [0-9]+. On success, the ID is stored in UIntVal and Token is
336 // returned, otherwise the Error token is returned.
337 lltok::Kind LLLexer::LexUIntID(lltok::Kind Token) {
338   if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
339     return lltok::Error;
340 
341   for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
342     /*empty*/;
343 
344   uint64_t Val = atoull(TokStart + 1, CurPtr);
345   if ((unsigned)Val != Val)
346     Error("invalid value number (too large)!");
347   UIntVal = unsigned(Val);
348   return Token;
349 }
350 
351 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
352   // Handle StringConstant: \"[^\"]*\"
353   if (CurPtr[0] == '"') {
354     ++CurPtr;
355 
356     while (true) {
357       int CurChar = getNextChar();
358 
359       if (CurChar == EOF) {
360         Error("end of file in global variable name");
361         return lltok::Error;
362       }
363       if (CurChar == '"') {
364         StrVal.assign(TokStart+2, CurPtr-1);
365         UnEscapeLexed(StrVal);
366         if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
367           Error("Null bytes are not allowed in names");
368           return lltok::Error;
369         }
370         return Var;
371       }
372     }
373   }
374 
375   // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
376   if (ReadVarName())
377     return Var;
378 
379   // Handle VarID: [0-9]+
380   return LexUIntID(VarID);
381 }
382 
383 /// Lex all tokens that start with a % character.
384 ///   LocalVar   ::= %\"[^\"]*\"
385 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
386 ///   LocalVarID ::= %[0-9]+
387 lltok::Kind LLLexer::LexPercent() {
388   return LexVar(lltok::LocalVar, lltok::LocalVarID);
389 }
390 
391 /// Lex all tokens that start with a " character.
392 ///   QuoteLabel        "[^"]+":
393 ///   StringConstant    "[^"]*"
394 lltok::Kind LLLexer::LexQuote() {
395   lltok::Kind kind = ReadString(lltok::StringConstant);
396   if (kind == lltok::Error || kind == lltok::Eof)
397     return kind;
398 
399   if (CurPtr[0] == ':') {
400     ++CurPtr;
401     if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
402       Error("Null bytes are not allowed in names");
403       kind = lltok::Error;
404     } else {
405       kind = lltok::LabelStr;
406     }
407   }
408 
409   return kind;
410 }
411 
412 /// Lex all tokens that start with a ! character.
413 ///    !foo
414 ///    !
415 lltok::Kind LLLexer::LexExclaim() {
416   // Lex a metadata name as a MetadataVar.
417   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
418       CurPtr[0] == '-' || CurPtr[0] == '$' ||
419       CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
420     ++CurPtr;
421     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
422            CurPtr[0] == '-' || CurPtr[0] == '$' ||
423            CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
424       ++CurPtr;
425 
426     StrVal.assign(TokStart+1, CurPtr);   // Skip !
427     UnEscapeLexed(StrVal);
428     return lltok::MetadataVar;
429   }
430   return lltok::exclaim;
431 }
432 
433 /// Lex all tokens that start with a ^ character.
434 ///    SummaryID ::= ^[0-9]+
435 lltok::Kind LLLexer::LexCaret() {
436   // Handle SummaryID: ^[0-9]+
437   return LexUIntID(lltok::SummaryID);
438 }
439 
440 /// Lex all tokens that start with a # character.
441 ///    AttrGrpID ::= #[0-9]+
442 lltok::Kind LLLexer::LexHash() {
443   // Handle AttrGrpID: #[0-9]+
444   return LexUIntID(lltok::AttrGrpID);
445 }
446 
447 /// Lex a label, integer type, keyword, or hexadecimal integer constant.
448 ///    Label           [-a-zA-Z$._0-9]+:
449 ///    IntegerType     i[0-9]+
450 ///    Keyword         sdiv, float, ...
451 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
452 lltok::Kind LLLexer::LexIdentifier() {
453   const char *StartChar = CurPtr;
454   const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
455   const char *KeywordEnd = nullptr;
456 
457   for (; isLabelChar(*CurPtr); ++CurPtr) {
458     // If we decide this is an integer, remember the end of the sequence.
459     if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
460       IntEnd = CurPtr;
461     if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
462         *CurPtr != '_')
463       KeywordEnd = CurPtr;
464   }
465 
466   // If we stopped due to a colon, unless we were directed to ignore it,
467   // this really is a label.
468   if (!IgnoreColonInIdentifiers && *CurPtr == ':') {
469     StrVal.assign(StartChar-1, CurPtr++);
470     return lltok::LabelStr;
471   }
472 
473   // Otherwise, this wasn't a label.  If this was valid as an integer type,
474   // return it.
475   if (!IntEnd) IntEnd = CurPtr;
476   if (IntEnd != StartChar) {
477     CurPtr = IntEnd;
478     uint64_t NumBits = atoull(StartChar, CurPtr);
479     if (NumBits < IntegerType::MIN_INT_BITS ||
480         NumBits > IntegerType::MAX_INT_BITS) {
481       Error("bitwidth for integer type out of range!");
482       return lltok::Error;
483     }
484     TyVal = IntegerType::get(Context, NumBits);
485     return lltok::Type;
486   }
487 
488   // Otherwise, this was a letter sequence.  See which keyword this is.
489   if (!KeywordEnd) KeywordEnd = CurPtr;
490   CurPtr = KeywordEnd;
491   --StartChar;
492   StringRef Keyword(StartChar, CurPtr - StartChar);
493 
494 #define KEYWORD(STR)                                                           \
495   do {                                                                         \
496     if (Keyword == #STR)                                                       \
497       return lltok::kw_##STR;                                                  \
498   } while (false)
499 
500   KEYWORD(true);    KEYWORD(false);
501   KEYWORD(declare); KEYWORD(define);
502   KEYWORD(global);  KEYWORD(constant);
503 
504   KEYWORD(dso_local);
505   KEYWORD(dso_preemptable);
506 
507   KEYWORD(private);
508   KEYWORD(internal);
509   KEYWORD(available_externally);
510   KEYWORD(linkonce);
511   KEYWORD(linkonce_odr);
512   KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
513   KEYWORD(weak_odr);
514   KEYWORD(appending);
515   KEYWORD(dllimport);
516   KEYWORD(dllexport);
517   KEYWORD(common);
518   KEYWORD(default);
519   KEYWORD(hidden);
520   KEYWORD(protected);
521   KEYWORD(unnamed_addr);
522   KEYWORD(local_unnamed_addr);
523   KEYWORD(externally_initialized);
524   KEYWORD(extern_weak);
525   KEYWORD(external);
526   KEYWORD(thread_local);
527   KEYWORD(localdynamic);
528   KEYWORD(initialexec);
529   KEYWORD(localexec);
530   KEYWORD(zeroinitializer);
531   KEYWORD(undef);
532   KEYWORD(null);
533   KEYWORD(none);
534   KEYWORD(to);
535   KEYWORD(caller);
536   KEYWORD(within);
537   KEYWORD(from);
538   KEYWORD(tail);
539   KEYWORD(musttail);
540   KEYWORD(notail);
541   KEYWORD(target);
542   KEYWORD(triple);
543   KEYWORD(source_filename);
544   KEYWORD(unwind);
545   KEYWORD(deplibs);             // FIXME: Remove in 4.0.
546   KEYWORD(datalayout);
547   KEYWORD(volatile);
548   KEYWORD(atomic);
549   KEYWORD(unordered);
550   KEYWORD(monotonic);
551   KEYWORD(acquire);
552   KEYWORD(release);
553   KEYWORD(acq_rel);
554   KEYWORD(seq_cst);
555   KEYWORD(syncscope);
556 
557   KEYWORD(nnan);
558   KEYWORD(ninf);
559   KEYWORD(nsz);
560   KEYWORD(arcp);
561   KEYWORD(contract);
562   KEYWORD(reassoc);
563   KEYWORD(afn);
564   KEYWORD(fast);
565   KEYWORD(nuw);
566   KEYWORD(nsw);
567   KEYWORD(exact);
568   KEYWORD(inbounds);
569   KEYWORD(inrange);
570   KEYWORD(align);
571   KEYWORD(addrspace);
572   KEYWORD(section);
573   KEYWORD(partition);
574   KEYWORD(alias);
575   KEYWORD(ifunc);
576   KEYWORD(module);
577   KEYWORD(asm);
578   KEYWORD(sideeffect);
579   KEYWORD(alignstack);
580   KEYWORD(inteldialect);
581   KEYWORD(gc);
582   KEYWORD(prefix);
583   KEYWORD(prologue);
584 
585   KEYWORD(ccc);
586   KEYWORD(fastcc);
587   KEYWORD(coldcc);
588   KEYWORD(cfguard_checkcc);
589   KEYWORD(x86_stdcallcc);
590   KEYWORD(x86_fastcallcc);
591   KEYWORD(x86_thiscallcc);
592   KEYWORD(x86_vectorcallcc);
593   KEYWORD(arm_apcscc);
594   KEYWORD(arm_aapcscc);
595   KEYWORD(arm_aapcs_vfpcc);
596   KEYWORD(aarch64_vector_pcs);
597   KEYWORD(aarch64_sve_vector_pcs);
598   KEYWORD(msp430_intrcc);
599   KEYWORD(avr_intrcc);
600   KEYWORD(avr_signalcc);
601   KEYWORD(ptx_kernel);
602   KEYWORD(ptx_device);
603   KEYWORD(spir_kernel);
604   KEYWORD(spir_func);
605   KEYWORD(intel_ocl_bicc);
606   KEYWORD(x86_64_sysvcc);
607   KEYWORD(win64cc);
608   KEYWORD(x86_regcallcc);
609   KEYWORD(webkit_jscc);
610   KEYWORD(swiftcc);
611   KEYWORD(anyregcc);
612   KEYWORD(preserve_mostcc);
613   KEYWORD(preserve_allcc);
614   KEYWORD(ghccc);
615   KEYWORD(x86_intrcc);
616   KEYWORD(hhvmcc);
617   KEYWORD(hhvm_ccc);
618   KEYWORD(cxx_fast_tlscc);
619   KEYWORD(amdgpu_vs);
620   KEYWORD(amdgpu_ls);
621   KEYWORD(amdgpu_hs);
622   KEYWORD(amdgpu_es);
623   KEYWORD(amdgpu_gs);
624   KEYWORD(amdgpu_ps);
625   KEYWORD(amdgpu_cs);
626   KEYWORD(amdgpu_kernel);
627   KEYWORD(tailcc);
628 
629   KEYWORD(cc);
630   KEYWORD(c);
631 
632   KEYWORD(attributes);
633 
634   KEYWORD(alwaysinline);
635   KEYWORD(allocsize);
636   KEYWORD(argmemonly);
637   KEYWORD(builtin);
638   KEYWORD(byval);
639   KEYWORD(inalloca);
640   KEYWORD(cold);
641   KEYWORD(convergent);
642   KEYWORD(dereferenceable);
643   KEYWORD(dereferenceable_or_null);
644   KEYWORD(inaccessiblememonly);
645   KEYWORD(inaccessiblemem_or_argmemonly);
646   KEYWORD(inlinehint);
647   KEYWORD(inreg);
648   KEYWORD(jumptable);
649   KEYWORD(minsize);
650   KEYWORD(naked);
651   KEYWORD(nest);
652   KEYWORD(noalias);
653   KEYWORD(nobuiltin);
654   KEYWORD(nocapture);
655   KEYWORD(noduplicate);
656   KEYWORD(nofree);
657   KEYWORD(noimplicitfloat);
658   KEYWORD(noinline);
659   KEYWORD(norecurse);
660   KEYWORD(nonlazybind);
661   KEYWORD(nonnull);
662   KEYWORD(noredzone);
663   KEYWORD(noreturn);
664   KEYWORD(nosync);
665   KEYWORD(nocf_check);
666   KEYWORD(nounwind);
667   KEYWORD(optforfuzzing);
668   KEYWORD(optnone);
669   KEYWORD(optsize);
670   KEYWORD(preallocated);
671   KEYWORD(readnone);
672   KEYWORD(readonly);
673   KEYWORD(returned);
674   KEYWORD(returns_twice);
675   KEYWORD(signext);
676   KEYWORD(speculatable);
677   KEYWORD(sret);
678   KEYWORD(ssp);
679   KEYWORD(sspreq);
680   KEYWORD(sspstrong);
681   KEYWORD(strictfp);
682   KEYWORD(safestack);
683   KEYWORD(shadowcallstack);
684   KEYWORD(sanitize_address);
685   KEYWORD(sanitize_hwaddress);
686   KEYWORD(sanitize_memtag);
687   KEYWORD(sanitize_thread);
688   KEYWORD(sanitize_memory);
689   KEYWORD(speculative_load_hardening);
690   KEYWORD(swifterror);
691   KEYWORD(swiftself);
692   KEYWORD(uwtable);
693   KEYWORD(willreturn);
694   KEYWORD(writeonly);
695   KEYWORD(zeroext);
696   KEYWORD(immarg);
697 
698   KEYWORD(type);
699   KEYWORD(opaque);
700 
701   KEYWORD(comdat);
702 
703   // Comdat types
704   KEYWORD(any);
705   KEYWORD(exactmatch);
706   KEYWORD(largest);
707   KEYWORD(noduplicates);
708   KEYWORD(samesize);
709 
710   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
711   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
712   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
713   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
714 
715   KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
716   KEYWORD(umin);
717 
718   KEYWORD(vscale);
719   KEYWORD(x);
720   KEYWORD(blockaddress);
721 
722   // Metadata types.
723   KEYWORD(distinct);
724 
725   // Use-list order directives.
726   KEYWORD(uselistorder);
727   KEYWORD(uselistorder_bb);
728 
729   KEYWORD(personality);
730   KEYWORD(cleanup);
731   KEYWORD(catch);
732   KEYWORD(filter);
733 
734   // Summary index keywords.
735   KEYWORD(path);
736   KEYWORD(hash);
737   KEYWORD(gv);
738   KEYWORD(guid);
739   KEYWORD(name);
740   KEYWORD(summaries);
741   KEYWORD(flags);
742   KEYWORD(linkage);
743   KEYWORD(notEligibleToImport);
744   KEYWORD(live);
745   KEYWORD(dsoLocal);
746   KEYWORD(canAutoHide);
747   KEYWORD(function);
748   KEYWORD(insts);
749   KEYWORD(funcFlags);
750   KEYWORD(readNone);
751   KEYWORD(readOnly);
752   KEYWORD(noRecurse);
753   KEYWORD(returnDoesNotAlias);
754   KEYWORD(noInline);
755   KEYWORD(alwaysInline);
756   KEYWORD(calls);
757   KEYWORD(callee);
758   KEYWORD(hotness);
759   KEYWORD(unknown);
760   KEYWORD(hot);
761   KEYWORD(critical);
762   KEYWORD(relbf);
763   KEYWORD(variable);
764   KEYWORD(vTableFuncs);
765   KEYWORD(virtFunc);
766   KEYWORD(aliasee);
767   KEYWORD(refs);
768   KEYWORD(typeIdInfo);
769   KEYWORD(typeTests);
770   KEYWORD(typeTestAssumeVCalls);
771   KEYWORD(typeCheckedLoadVCalls);
772   KEYWORD(typeTestAssumeConstVCalls);
773   KEYWORD(typeCheckedLoadConstVCalls);
774   KEYWORD(vFuncId);
775   KEYWORD(offset);
776   KEYWORD(args);
777   KEYWORD(typeid);
778   KEYWORD(typeidCompatibleVTable);
779   KEYWORD(summary);
780   KEYWORD(typeTestRes);
781   KEYWORD(kind);
782   KEYWORD(unsat);
783   KEYWORD(byteArray);
784   KEYWORD(inline);
785   KEYWORD(single);
786   KEYWORD(allOnes);
787   KEYWORD(sizeM1BitWidth);
788   KEYWORD(alignLog2);
789   KEYWORD(sizeM1);
790   KEYWORD(bitMask);
791   KEYWORD(inlineBits);
792   KEYWORD(vcall_visibility);
793   KEYWORD(wpdResolutions);
794   KEYWORD(wpdRes);
795   KEYWORD(indir);
796   KEYWORD(singleImpl);
797   KEYWORD(branchFunnel);
798   KEYWORD(singleImplName);
799   KEYWORD(resByArg);
800   KEYWORD(byArg);
801   KEYWORD(uniformRetVal);
802   KEYWORD(uniqueRetVal);
803   KEYWORD(virtualConstProp);
804   KEYWORD(info);
805   KEYWORD(byte);
806   KEYWORD(bit);
807   KEYWORD(varFlags);
808 
809 #undef KEYWORD
810 
811   // Keywords for types.
812 #define TYPEKEYWORD(STR, LLVMTY)                                               \
813   do {                                                                         \
814     if (Keyword == STR) {                                                      \
815       TyVal = LLVMTY;                                                          \
816       return lltok::Type;                                                      \
817     }                                                                          \
818   } while (false)
819 
820   TYPEKEYWORD("void",      Type::getVoidTy(Context));
821   TYPEKEYWORD("half",      Type::getHalfTy(Context));
822   TYPEKEYWORD("float",     Type::getFloatTy(Context));
823   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
824   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
825   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
826   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
827   TYPEKEYWORD("label",     Type::getLabelTy(Context));
828   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
829   TYPEKEYWORD("x86_mmx",   Type::getX86_MMXTy(Context));
830   TYPEKEYWORD("token",     Type::getTokenTy(Context));
831 
832 #undef TYPEKEYWORD
833 
834   // Keywords for instructions.
835 #define INSTKEYWORD(STR, Enum)                                                 \
836   do {                                                                         \
837     if (Keyword == #STR) {                                                     \
838       UIntVal = Instruction::Enum;                                             \
839       return lltok::kw_##STR;                                                  \
840     }                                                                          \
841   } while (false)
842 
843   INSTKEYWORD(fneg,  FNeg);
844 
845   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
846   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
847   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
848   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
849   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
850   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
851   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
852   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
853 
854   INSTKEYWORD(phi,         PHI);
855   INSTKEYWORD(call,        Call);
856   INSTKEYWORD(trunc,       Trunc);
857   INSTKEYWORD(zext,        ZExt);
858   INSTKEYWORD(sext,        SExt);
859   INSTKEYWORD(fptrunc,     FPTrunc);
860   INSTKEYWORD(fpext,       FPExt);
861   INSTKEYWORD(uitofp,      UIToFP);
862   INSTKEYWORD(sitofp,      SIToFP);
863   INSTKEYWORD(fptoui,      FPToUI);
864   INSTKEYWORD(fptosi,      FPToSI);
865   INSTKEYWORD(inttoptr,    IntToPtr);
866   INSTKEYWORD(ptrtoint,    PtrToInt);
867   INSTKEYWORD(bitcast,     BitCast);
868   INSTKEYWORD(addrspacecast, AddrSpaceCast);
869   INSTKEYWORD(select,      Select);
870   INSTKEYWORD(va_arg,      VAArg);
871   INSTKEYWORD(ret,         Ret);
872   INSTKEYWORD(br,          Br);
873   INSTKEYWORD(switch,      Switch);
874   INSTKEYWORD(indirectbr,  IndirectBr);
875   INSTKEYWORD(invoke,      Invoke);
876   INSTKEYWORD(resume,      Resume);
877   INSTKEYWORD(unreachable, Unreachable);
878   INSTKEYWORD(callbr,      CallBr);
879 
880   INSTKEYWORD(alloca,      Alloca);
881   INSTKEYWORD(load,        Load);
882   INSTKEYWORD(store,       Store);
883   INSTKEYWORD(cmpxchg,     AtomicCmpXchg);
884   INSTKEYWORD(atomicrmw,   AtomicRMW);
885   INSTKEYWORD(fence,       Fence);
886   INSTKEYWORD(getelementptr, GetElementPtr);
887 
888   INSTKEYWORD(extractelement, ExtractElement);
889   INSTKEYWORD(insertelement,  InsertElement);
890   INSTKEYWORD(shufflevector,  ShuffleVector);
891   INSTKEYWORD(extractvalue,   ExtractValue);
892   INSTKEYWORD(insertvalue,    InsertValue);
893   INSTKEYWORD(landingpad,     LandingPad);
894   INSTKEYWORD(cleanupret,     CleanupRet);
895   INSTKEYWORD(catchret,       CatchRet);
896   INSTKEYWORD(catchswitch,  CatchSwitch);
897   INSTKEYWORD(catchpad,     CatchPad);
898   INSTKEYWORD(cleanuppad,   CleanupPad);
899 
900   INSTKEYWORD(freeze,       Freeze);
901 
902 #undef INSTKEYWORD
903 
904 #define DWKEYWORD(TYPE, TOKEN)                                                 \
905   do {                                                                         \
906     if (Keyword.startswith("DW_" #TYPE "_")) {                                 \
907       StrVal.assign(Keyword.begin(), Keyword.end());                           \
908       return lltok::TOKEN;                                                     \
909     }                                                                          \
910   } while (false)
911 
912   DWKEYWORD(TAG, DwarfTag);
913   DWKEYWORD(ATE, DwarfAttEncoding);
914   DWKEYWORD(VIRTUALITY, DwarfVirtuality);
915   DWKEYWORD(LANG, DwarfLang);
916   DWKEYWORD(CC, DwarfCC);
917   DWKEYWORD(OP, DwarfOp);
918   DWKEYWORD(MACINFO, DwarfMacinfo);
919 
920 #undef DWKEYWORD
921 
922   if (Keyword.startswith("DIFlag")) {
923     StrVal.assign(Keyword.begin(), Keyword.end());
924     return lltok::DIFlag;
925   }
926 
927   if (Keyword.startswith("DISPFlag")) {
928     StrVal.assign(Keyword.begin(), Keyword.end());
929     return lltok::DISPFlag;
930   }
931 
932   if (Keyword.startswith("CSK_")) {
933     StrVal.assign(Keyword.begin(), Keyword.end());
934     return lltok::ChecksumKind;
935   }
936 
937   if (Keyword == "NoDebug" || Keyword == "FullDebug" ||
938       Keyword == "LineTablesOnly" || Keyword == "DebugDirectivesOnly") {
939     StrVal.assign(Keyword.begin(), Keyword.end());
940     return lltok::EmissionKind;
941   }
942 
943   if (Keyword == "GNU" || Keyword == "None" || Keyword == "Default") {
944     StrVal.assign(Keyword.begin(), Keyword.end());
945     return lltok::NameTableKind;
946   }
947 
948   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
949   // the CFE to avoid forcing it to deal with 64-bit numbers.
950   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
951       TokStart[1] == '0' && TokStart[2] == 'x' &&
952       isxdigit(static_cast<unsigned char>(TokStart[3]))) {
953     int len = CurPtr-TokStart-3;
954     uint32_t bits = len * 4;
955     StringRef HexStr(TokStart + 3, len);
956     if (!all_of(HexStr, isxdigit)) {
957       // Bad token, return it as an error.
958       CurPtr = TokStart+3;
959       return lltok::Error;
960     }
961     APInt Tmp(bits, HexStr, 16);
962     uint32_t activeBits = Tmp.getActiveBits();
963     if (activeBits > 0 && activeBits < bits)
964       Tmp = Tmp.trunc(activeBits);
965     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
966     return lltok::APSInt;
967   }
968 
969   // If this is "cc1234", return this as just "cc".
970   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
971     CurPtr = TokStart+2;
972     return lltok::kw_cc;
973   }
974 
975   // Finally, if this isn't known, return an error.
976   CurPtr = TokStart+1;
977   return lltok::Error;
978 }
979 
980 /// Lex all tokens that start with a 0x prefix, knowing they match and are not
981 /// labels.
982 ///    HexFPConstant     0x[0-9A-Fa-f]+
983 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
984 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
985 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
986 ///    HexHalfConstant   0xH[0-9A-Fa-f]+
987 lltok::Kind LLLexer::Lex0x() {
988   CurPtr = TokStart + 2;
989 
990   char Kind;
991   if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
992     Kind = *CurPtr++;
993   } else {
994     Kind = 'J';
995   }
996 
997   if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
998     // Bad token, return it as an error.
999     CurPtr = TokStart+1;
1000     return lltok::Error;
1001   }
1002 
1003   while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
1004     ++CurPtr;
1005 
1006   if (Kind == 'J') {
1007     // HexFPConstant - Floating point constant represented in IEEE format as a
1008     // hexadecimal number for when exponential notation is not precise enough.
1009     // Half, Float, and double only.
1010     APFloatVal = APFloat(APFloat::IEEEdouble(),
1011                          APInt(64, HexIntToVal(TokStart + 2, CurPtr)));
1012     return lltok::APFloat;
1013   }
1014 
1015   uint64_t Pair[2];
1016   switch (Kind) {
1017   default: llvm_unreachable("Unknown kind!");
1018   case 'K':
1019     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
1020     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
1021     APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair));
1022     return lltok::APFloat;
1023   case 'L':
1024     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
1025     HexToIntPair(TokStart+3, CurPtr, Pair);
1026     APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair));
1027     return lltok::APFloat;
1028   case 'M':
1029     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
1030     HexToIntPair(TokStart+3, CurPtr, Pair);
1031     APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair));
1032     return lltok::APFloat;
1033   case 'H':
1034     APFloatVal = APFloat(APFloat::IEEEhalf(),
1035                          APInt(16,HexIntToVal(TokStart+3, CurPtr)));
1036     return lltok::APFloat;
1037   }
1038 }
1039 
1040 /// Lex tokens for a label or a numeric constant, possibly starting with -.
1041 ///    Label             [-a-zA-Z$._0-9]+:
1042 ///    NInteger          -[0-9]+
1043 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1044 ///    PInteger          [0-9]+
1045 ///    HexFPConstant     0x[0-9A-Fa-f]+
1046 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
1047 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
1048 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
1049 lltok::Kind LLLexer::LexDigitOrNegative() {
1050   // If the letter after the negative is not a number, this is probably a label.
1051   if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
1052       !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
1053     // Okay, this is not a number after the -, it's probably a label.
1054     if (const char *End = isLabelTail(CurPtr)) {
1055       StrVal.assign(TokStart, End-1);
1056       CurPtr = End;
1057       return lltok::LabelStr;
1058     }
1059 
1060     return lltok::Error;
1061   }
1062 
1063   // At this point, it is either a label, int or fp constant.
1064 
1065   // Skip digits, we have at least one.
1066   for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1067     /*empty*/;
1068 
1069   // Check if this is a fully-numeric label:
1070   if (isdigit(TokStart[0]) && CurPtr[0] == ':') {
1071     uint64_t Val = atoull(TokStart, CurPtr);
1072     ++CurPtr; // Skip the colon.
1073     if ((unsigned)Val != Val)
1074       Error("invalid value number (too large)!");
1075     UIntVal = unsigned(Val);
1076     return lltok::LabelID;
1077   }
1078 
1079   // Check to see if this really is a string label, e.g. "-1:".
1080   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
1081     if (const char *End = isLabelTail(CurPtr)) {
1082       StrVal.assign(TokStart, End-1);
1083       CurPtr = End;
1084       return lltok::LabelStr;
1085     }
1086   }
1087 
1088   // If the next character is a '.', then it is a fp value, otherwise its
1089   // integer.
1090   if (CurPtr[0] != '.') {
1091     if (TokStart[0] == '0' && TokStart[1] == 'x')
1092       return Lex0x();
1093     APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart));
1094     return lltok::APSInt;
1095   }
1096 
1097   ++CurPtr;
1098 
1099   // Skip over [0-9]*([eE][-+]?[0-9]+)?
1100   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1101 
1102   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1103     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1104         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1105           isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1106       CurPtr += 2;
1107       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1108     }
1109   }
1110 
1111   APFloatVal = APFloat(APFloat::IEEEdouble(),
1112                        StringRef(TokStart, CurPtr - TokStart));
1113   return lltok::APFloat;
1114 }
1115 
1116 /// Lex a floating point constant starting with +.
1117 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1118 lltok::Kind LLLexer::LexPositive() {
1119   // If the letter after the negative is a number, this is probably not a
1120   // label.
1121   if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
1122     return lltok::Error;
1123 
1124   // Skip digits.
1125   for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1126     /*empty*/;
1127 
1128   // At this point, we need a '.'.
1129   if (CurPtr[0] != '.') {
1130     CurPtr = TokStart+1;
1131     return lltok::Error;
1132   }
1133 
1134   ++CurPtr;
1135 
1136   // Skip over [0-9]*([eE][-+]?[0-9]+)?
1137   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1138 
1139   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1140     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1141         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1142         isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1143       CurPtr += 2;
1144       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1145     }
1146   }
1147 
1148   APFloatVal = APFloat(APFloat::IEEEdouble(),
1149                        StringRef(TokStart, CurPtr - TokStart));
1150   return lltok::APFloat;
1151 }
1152