1 //===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // Implement the Lexer for .ll files.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "LLLexer.h"
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"
23 #include <cassert>
24 #include <cctype>
25 #include <cstdio>
26 
27 using namespace llvm;
28 
29 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
30   ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
31   return true;
32 }
33 
34 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
35   SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
36 }
37 
38 //===----------------------------------------------------------------------===//
39 // Helper functions.
40 //===----------------------------------------------------------------------===//
41 
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...
45 //
46 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
47   uint64_t Result = 0;
48   for (; Buffer != End; Buffer++) {
49     uint64_t OldRes = Result;
50     Result *= 10;
51     Result += *Buffer-'0';
52     if (Result < OldRes) {  // Uh, oh, overflow detected!!!
53       Error("constant bigger than 64 bits detected!");
54       return 0;
55     }
56   }
57   return Result;
58 }
59 
60 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
61   uint64_t Result = 0;
62   for (; Buffer != End; ++Buffer) {
63     uint64_t OldRes = Result;
64     Result *= 16;
65     Result += hexDigitValue(*Buffer);
66 
67     if (Result < OldRes) {   // Uh, oh, overflow detected!!!
68       Error("constant bigger than 64 bits detected!");
69       return 0;
70     }
71   }
72   return Result;
73 }
74 
75 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
76                            uint64_t Pair[2]) {
77   Pair[0] = 0;
78   if (End - Buffer >= 16) {
79     for (int i = 0; i < 16; i++, Buffer++) {
80       assert(Buffer != End);
81       Pair[0] *= 16;
82       Pair[0] += hexDigitValue(*Buffer);
83     }
84   }
85   Pair[1] = 0;
86   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
87     Pair[1] *= 16;
88     Pair[1] += hexDigitValue(*Buffer);
89   }
90   if (Buffer != End)
91     Error("constant bigger than 128 bits detected!");
92 }
93 
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,
97                            uint64_t Pair[2]) {
98   Pair[1] = 0;
99   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
100     assert(Buffer != End);
101     Pair[1] *= 16;
102     Pair[1] += hexDigitValue(*Buffer);
103   }
104   Pair[0] = 0;
105   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
106     Pair[0] *= 16;
107     Pair[0] += hexDigitValue(*Buffer);
108   }
109   if (Buffer != End)
110     Error("constant bigger than 128 bits detected!");
111 }
112 
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;
117 
118   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
119   char *BOut = Buffer;
120   for (char *BIn = Buffer; BIn != EndBuffer; ) {
121     if (BIn[0] == '\\') {
122       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
123         *BOut++ = '\\'; // Two \ becomes one
124         BIn += 2;
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
130         ++BOut;
131       } else {
132         *BOut++ = *BIn++;
133       }
134     } else {
135       *BOut++ = *BIn++;
136     }
137   }
138   Str.resize(BOut-Buffer);
139 }
140 
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 == '_';
145 }
146 
147 /// isLabelTail - Return true if this pointer points to a valid end of a label.
148 static const char *isLabelTail(const char *CurPtr) {
149   while (true) {
150     if (CurPtr[0] == ':') return CurPtr+1;
151     if (!isLabelChar(CurPtr[0])) return nullptr;
152     ++CurPtr;
153   }
154 }
155 
156 //===----------------------------------------------------------------------===//
157 // Lexer definition.
158 //===----------------------------------------------------------------------===//
159 
160 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &sm, SMDiagnostic &Err,
161                  LLVMContext &C)
162   : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
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 '$': return LexDollar();
205     case '%': return LexPercent();
206     case '"': return LexQuote();
207     case '.':
208       if (const char *Ptr = isLabelTail(CurPtr)) {
209         CurPtr = Ptr;
210         StrVal.assign(TokStart, CurPtr-1);
211         return lltok::LabelStr;
212       }
213       if (CurPtr[0] == '.' && CurPtr[1] == '.') {
214         CurPtr += 2;
215         return lltok::dotdotdot;
216       }
217       return lltok::Error;
218     case ';':
219       SkipLineComment();
220       continue;
221     case '!': return LexExclaim();
222     case '#': return LexHash();
223     case '0': case '1': case '2': case '3': case '4':
224     case '5': case '6': case '7': case '8': case '9':
225     case '-':
226       return LexDigitOrNegative();
227     case '=': return lltok::equal;
228     case '[': return lltok::lsquare;
229     case ']': return lltok::rsquare;
230     case '{': return lltok::lbrace;
231     case '}': return lltok::rbrace;
232     case '<': return lltok::less;
233     case '>': return lltok::greater;
234     case '(': return lltok::lparen;
235     case ')': return lltok::rparen;
236     case ',': return lltok::comma;
237     case '*': return lltok::star;
238     case '|': return lltok::bar;
239     }
240   }
241 }
242 
243 void LLLexer::SkipLineComment() {
244   while (true) {
245     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
246       return;
247   }
248 }
249 
250 /// Lex all tokens that start with an @ character.
251 ///   GlobalVar   @\"[^\"]*\"
252 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
253 ///   GlobalVarID @[0-9]+
254 lltok::Kind LLLexer::LexAt() {
255   return LexVar(lltok::GlobalVar, lltok::GlobalID);
256 }
257 
258 lltok::Kind LLLexer::LexDollar() {
259   if (const char *Ptr = isLabelTail(TokStart)) {
260     CurPtr = Ptr;
261     StrVal.assign(TokStart, CurPtr - 1);
262     return lltok::LabelStr;
263   }
264 
265   // Handle DollarStringConstant: $\"[^\"]*\"
266   if (CurPtr[0] == '"') {
267     ++CurPtr;
268 
269     while (true) {
270       int CurChar = getNextChar();
271 
272       if (CurChar == EOF) {
273         Error("end of file in COMDAT variable name");
274         return lltok::Error;
275       }
276       if (CurChar == '"') {
277         StrVal.assign(TokStart + 2, CurPtr - 1);
278         UnEscapeLexed(StrVal);
279         if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
280           Error("Null bytes are not allowed in names");
281           return lltok::Error;
282         }
283         return lltok::ComdatVar;
284       }
285     }
286   }
287 
288   // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
289   if (ReadVarName())
290     return lltok::ComdatVar;
291 
292   return lltok::Error;
293 }
294 
295 /// ReadString - Read a string until the closing quote.
296 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
297   const char *Start = CurPtr;
298   while (true) {
299     int CurChar = getNextChar();
300 
301     if (CurChar == EOF) {
302       Error("end of file in string constant");
303       return lltok::Error;
304     }
305     if (CurChar == '"') {
306       StrVal.assign(Start, CurPtr-1);
307       UnEscapeLexed(StrVal);
308       return kind;
309     }
310   }
311 }
312 
313 /// ReadVarName - Read the rest of a token containing a variable name.
314 bool LLLexer::ReadVarName() {
315   const char *NameStart = CurPtr;
316   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
317       CurPtr[0] == '-' || CurPtr[0] == '$' ||
318       CurPtr[0] == '.' || CurPtr[0] == '_') {
319     ++CurPtr;
320     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
321            CurPtr[0] == '-' || CurPtr[0] == '$' ||
322            CurPtr[0] == '.' || CurPtr[0] == '_')
323       ++CurPtr;
324 
325     StrVal.assign(NameStart, CurPtr);
326     return true;
327   }
328   return false;
329 }
330 
331 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
332   // Handle StringConstant: \"[^\"]*\"
333   if (CurPtr[0] == '"') {
334     ++CurPtr;
335 
336     while (true) {
337       int CurChar = getNextChar();
338 
339       if (CurChar == EOF) {
340         Error("end of file in global variable name");
341         return lltok::Error;
342       }
343       if (CurChar == '"') {
344         StrVal.assign(TokStart+2, CurPtr-1);
345         UnEscapeLexed(StrVal);
346         if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
347           Error("Null bytes are not allowed in names");
348           return lltok::Error;
349         }
350         return Var;
351       }
352     }
353   }
354 
355   // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
356   if (ReadVarName())
357     return Var;
358 
359   // Handle VarID: [0-9]+
360   if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
361     for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
362       /*empty*/;
363 
364     uint64_t Val = atoull(TokStart+1, CurPtr);
365     if ((unsigned)Val != Val)
366       Error("invalid value number (too large)!");
367     UIntVal = unsigned(Val);
368     return VarID;
369   }
370   return lltok::Error;
371 }
372 
373 /// Lex all tokens that start with a % character.
374 ///   LocalVar   ::= %\"[^\"]*\"
375 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
376 ///   LocalVarID ::= %[0-9]+
377 lltok::Kind LLLexer::LexPercent() {
378   return LexVar(lltok::LocalVar, lltok::LocalVarID);
379 }
380 
381 /// Lex all tokens that start with a " character.
382 ///   QuoteLabel        "[^"]+":
383 ///   StringConstant    "[^"]*"
384 lltok::Kind LLLexer::LexQuote() {
385   lltok::Kind kind = ReadString(lltok::StringConstant);
386   if (kind == lltok::Error || kind == lltok::Eof)
387     return kind;
388 
389   if (CurPtr[0] == ':') {
390     ++CurPtr;
391     if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
392       Error("Null bytes are not allowed in names");
393       kind = lltok::Error;
394     } else {
395       kind = lltok::LabelStr;
396     }
397   }
398 
399   return kind;
400 }
401 
402 /// Lex all tokens that start with a ! character.
403 ///    !foo
404 ///    !
405 lltok::Kind LLLexer::LexExclaim() {
406   // Lex a metadata name as a MetadataVar.
407   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
408       CurPtr[0] == '-' || CurPtr[0] == '$' ||
409       CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
410     ++CurPtr;
411     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
412            CurPtr[0] == '-' || CurPtr[0] == '$' ||
413            CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
414       ++CurPtr;
415 
416     StrVal.assign(TokStart+1, CurPtr);   // Skip !
417     UnEscapeLexed(StrVal);
418     return lltok::MetadataVar;
419   }
420   return lltok::exclaim;
421 }
422 
423 /// Lex all tokens that start with a # character.
424 ///    AttrGrpID ::= #[0-9]+
425 lltok::Kind LLLexer::LexHash() {
426   // Handle AttrGrpID: #[0-9]+
427   if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
428     for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
429       /*empty*/;
430 
431     uint64_t Val = atoull(TokStart+1, CurPtr);
432     if ((unsigned)Val != Val)
433       Error("invalid value number (too large)!");
434     UIntVal = unsigned(Val);
435     return lltok::AttrGrpID;
436   }
437 
438   return lltok::Error;
439 }
440 
441 /// Lex a label, integer type, keyword, or hexadecimal integer constant.
442 ///    Label           [-a-zA-Z$._0-9]+:
443 ///    IntegerType     i[0-9]+
444 ///    Keyword         sdiv, float, ...
445 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
446 lltok::Kind LLLexer::LexIdentifier() {
447   const char *StartChar = CurPtr;
448   const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
449   const char *KeywordEnd = nullptr;
450 
451   for (; isLabelChar(*CurPtr); ++CurPtr) {
452     // If we decide this is an integer, remember the end of the sequence.
453     if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
454       IntEnd = CurPtr;
455     if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
456         *CurPtr != '_')
457       KeywordEnd = CurPtr;
458   }
459 
460   // If we stopped due to a colon, this really is a label.
461   if (*CurPtr == ':') {
462     StrVal.assign(StartChar-1, CurPtr++);
463     return lltok::LabelStr;
464   }
465 
466   // Otherwise, this wasn't a label.  If this was valid as an integer type,
467   // return it.
468   if (!IntEnd) IntEnd = CurPtr;
469   if (IntEnd != StartChar) {
470     CurPtr = IntEnd;
471     uint64_t NumBits = atoull(StartChar, CurPtr);
472     if (NumBits < IntegerType::MIN_INT_BITS ||
473         NumBits > IntegerType::MAX_INT_BITS) {
474       Error("bitwidth for integer type out of range!");
475       return lltok::Error;
476     }
477     TyVal = IntegerType::get(Context, NumBits);
478     return lltok::Type;
479   }
480 
481   // Otherwise, this was a letter sequence.  See which keyword this is.
482   if (!KeywordEnd) KeywordEnd = CurPtr;
483   CurPtr = KeywordEnd;
484   --StartChar;
485   StringRef Keyword(StartChar, CurPtr - StartChar);
486 
487 #define KEYWORD(STR)                                                           \
488   do {                                                                         \
489     if (Keyword == #STR)                                                       \
490       return lltok::kw_##STR;                                                  \
491   } while (false)
492 
493   KEYWORD(true);    KEYWORD(false);
494   KEYWORD(declare); KEYWORD(define);
495   KEYWORD(global);  KEYWORD(constant);
496 
497   KEYWORD(private);
498   KEYWORD(internal);
499   KEYWORD(available_externally);
500   KEYWORD(linkonce);
501   KEYWORD(linkonce_odr);
502   KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
503   KEYWORD(weak_odr);
504   KEYWORD(appending);
505   KEYWORD(dllimport);
506   KEYWORD(dllexport);
507   KEYWORD(common);
508   KEYWORD(default);
509   KEYWORD(hidden);
510   KEYWORD(protected);
511   KEYWORD(unnamed_addr);
512   KEYWORD(local_unnamed_addr);
513   KEYWORD(externally_initialized);
514   KEYWORD(extern_weak);
515   KEYWORD(external);
516   KEYWORD(thread_local);
517   KEYWORD(localdynamic);
518   KEYWORD(initialexec);
519   KEYWORD(localexec);
520   KEYWORD(zeroinitializer);
521   KEYWORD(undef);
522   KEYWORD(null);
523   KEYWORD(none);
524   KEYWORD(to);
525   KEYWORD(caller);
526   KEYWORD(within);
527   KEYWORD(from);
528   KEYWORD(tail);
529   KEYWORD(musttail);
530   KEYWORD(notail);
531   KEYWORD(target);
532   KEYWORD(triple);
533   KEYWORD(source_filename);
534   KEYWORD(unwind);
535   KEYWORD(deplibs);             // FIXME: Remove in 4.0.
536   KEYWORD(datalayout);
537   KEYWORD(volatile);
538   KEYWORD(atomic);
539   KEYWORD(unordered);
540   KEYWORD(monotonic);
541   KEYWORD(acquire);
542   KEYWORD(release);
543   KEYWORD(acq_rel);
544   KEYWORD(seq_cst);
545   KEYWORD(singlethread);
546 
547   KEYWORD(nnan);
548   KEYWORD(ninf);
549   KEYWORD(nsz);
550   KEYWORD(arcp);
551   KEYWORD(fast);
552   KEYWORD(nuw);
553   KEYWORD(nsw);
554   KEYWORD(exact);
555   KEYWORD(inbounds);
556   KEYWORD(inrange);
557   KEYWORD(align);
558   KEYWORD(addrspace);
559   KEYWORD(section);
560   KEYWORD(alias);
561   KEYWORD(ifunc);
562   KEYWORD(module);
563   KEYWORD(asm);
564   KEYWORD(sideeffect);
565   KEYWORD(alignstack);
566   KEYWORD(inteldialect);
567   KEYWORD(gc);
568   KEYWORD(prefix);
569   KEYWORD(prologue);
570 
571   KEYWORD(ccc);
572   KEYWORD(fastcc);
573   KEYWORD(coldcc);
574   KEYWORD(x86_stdcallcc);
575   KEYWORD(x86_fastcallcc);
576   KEYWORD(x86_thiscallcc);
577   KEYWORD(x86_vectorcallcc);
578   KEYWORD(arm_apcscc);
579   KEYWORD(arm_aapcscc);
580   KEYWORD(arm_aapcs_vfpcc);
581   KEYWORD(msp430_intrcc);
582   KEYWORD(avr_intrcc);
583   KEYWORD(avr_signalcc);
584   KEYWORD(ptx_kernel);
585   KEYWORD(ptx_device);
586   KEYWORD(spir_kernel);
587   KEYWORD(spir_func);
588   KEYWORD(intel_ocl_bicc);
589   KEYWORD(x86_64_sysvcc);
590   KEYWORD(x86_64_win64cc);
591   KEYWORD(x86_regcallcc);
592   KEYWORD(webkit_jscc);
593   KEYWORD(swiftcc);
594   KEYWORD(anyregcc);
595   KEYWORD(preserve_mostcc);
596   KEYWORD(preserve_allcc);
597   KEYWORD(ghccc);
598   KEYWORD(x86_intrcc);
599   KEYWORD(hhvmcc);
600   KEYWORD(hhvm_ccc);
601   KEYWORD(cxx_fast_tlscc);
602   KEYWORD(amdgpu_vs);
603   KEYWORD(amdgpu_gs);
604   KEYWORD(amdgpu_ps);
605   KEYWORD(amdgpu_cs);
606   KEYWORD(amdgpu_kernel);
607 
608   KEYWORD(cc);
609   KEYWORD(c);
610 
611   KEYWORD(attributes);
612 
613   KEYWORD(alwaysinline);
614   KEYWORD(allocsize);
615   KEYWORD(argmemonly);
616   KEYWORD(builtin);
617   KEYWORD(byval);
618   KEYWORD(inalloca);
619   KEYWORD(cold);
620   KEYWORD(convergent);
621   KEYWORD(dereferenceable);
622   KEYWORD(dereferenceable_or_null);
623   KEYWORD(inaccessiblememonly);
624   KEYWORD(inaccessiblemem_or_argmemonly);
625   KEYWORD(inlinehint);
626   KEYWORD(inreg);
627   KEYWORD(jumptable);
628   KEYWORD(minsize);
629   KEYWORD(naked);
630   KEYWORD(nest);
631   KEYWORD(noalias);
632   KEYWORD(nobuiltin);
633   KEYWORD(nocapture);
634   KEYWORD(noduplicate);
635   KEYWORD(noimplicitfloat);
636   KEYWORD(noinline);
637   KEYWORD(norecurse);
638   KEYWORD(nonlazybind);
639   KEYWORD(nonnull);
640   KEYWORD(noredzone);
641   KEYWORD(noreturn);
642   KEYWORD(nounwind);
643   KEYWORD(optnone);
644   KEYWORD(optsize);
645   KEYWORD(readnone);
646   KEYWORD(readonly);
647   KEYWORD(returned);
648   KEYWORD(returns_twice);
649   KEYWORD(signext);
650   KEYWORD(sret);
651   KEYWORD(ssp);
652   KEYWORD(sspreq);
653   KEYWORD(sspstrong);
654   KEYWORD(safestack);
655   KEYWORD(sanitize_address);
656   KEYWORD(sanitize_thread);
657   KEYWORD(sanitize_memory);
658   KEYWORD(swifterror);
659   KEYWORD(swiftself);
660   KEYWORD(uwtable);
661   KEYWORD(writeonly);
662   KEYWORD(zeroext);
663 
664   KEYWORD(type);
665   KEYWORD(opaque);
666 
667   KEYWORD(comdat);
668 
669   // Comdat types
670   KEYWORD(any);
671   KEYWORD(exactmatch);
672   KEYWORD(largest);
673   KEYWORD(noduplicates);
674   KEYWORD(samesize);
675 
676   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
677   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
678   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
679   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
680 
681   KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
682   KEYWORD(umin);
683 
684   KEYWORD(x);
685   KEYWORD(blockaddress);
686 
687   // Metadata types.
688   KEYWORD(distinct);
689 
690   // Use-list order directives.
691   KEYWORD(uselistorder);
692   KEYWORD(uselistorder_bb);
693 
694   KEYWORD(personality);
695   KEYWORD(cleanup);
696   KEYWORD(catch);
697   KEYWORD(filter);
698 
699 #undef KEYWORD
700 
701   // Keywords for types.
702 #define TYPEKEYWORD(STR, LLVMTY)                                               \
703   do {                                                                         \
704     if (Keyword == STR) {                                                      \
705       TyVal = LLVMTY;                                                          \
706       return lltok::Type;                                                      \
707     }                                                                          \
708   } while (false)
709 
710   TYPEKEYWORD("void",      Type::getVoidTy(Context));
711   TYPEKEYWORD("half",      Type::getHalfTy(Context));
712   TYPEKEYWORD("float",     Type::getFloatTy(Context));
713   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
714   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
715   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
716   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
717   TYPEKEYWORD("label",     Type::getLabelTy(Context));
718   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
719   TYPEKEYWORD("x86_mmx",   Type::getX86_MMXTy(Context));
720   TYPEKEYWORD("token",     Type::getTokenTy(Context));
721 
722 #undef TYPEKEYWORD
723 
724   // Keywords for instructions.
725 #define INSTKEYWORD(STR, Enum)                                                 \
726   do {                                                                         \
727     if (Keyword == #STR) {                                                     \
728       UIntVal = Instruction::Enum;                                             \
729       return lltok::kw_##STR;                                                  \
730     }                                                                          \
731   } while (false)
732 
733   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
734   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
735   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
736   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
737   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
738   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
739   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
740   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
741 
742   INSTKEYWORD(phi,         PHI);
743   INSTKEYWORD(call,        Call);
744   INSTKEYWORD(trunc,       Trunc);
745   INSTKEYWORD(zext,        ZExt);
746   INSTKEYWORD(sext,        SExt);
747   INSTKEYWORD(fptrunc,     FPTrunc);
748   INSTKEYWORD(fpext,       FPExt);
749   INSTKEYWORD(uitofp,      UIToFP);
750   INSTKEYWORD(sitofp,      SIToFP);
751   INSTKEYWORD(fptoui,      FPToUI);
752   INSTKEYWORD(fptosi,      FPToSI);
753   INSTKEYWORD(inttoptr,    IntToPtr);
754   INSTKEYWORD(ptrtoint,    PtrToInt);
755   INSTKEYWORD(bitcast,     BitCast);
756   INSTKEYWORD(addrspacecast, AddrSpaceCast);
757   INSTKEYWORD(select,      Select);
758   INSTKEYWORD(va_arg,      VAArg);
759   INSTKEYWORD(ret,         Ret);
760   INSTKEYWORD(br,          Br);
761   INSTKEYWORD(switch,      Switch);
762   INSTKEYWORD(indirectbr,  IndirectBr);
763   INSTKEYWORD(invoke,      Invoke);
764   INSTKEYWORD(resume,      Resume);
765   INSTKEYWORD(unreachable, Unreachable);
766 
767   INSTKEYWORD(alloca,      Alloca);
768   INSTKEYWORD(load,        Load);
769   INSTKEYWORD(store,       Store);
770   INSTKEYWORD(cmpxchg,     AtomicCmpXchg);
771   INSTKEYWORD(atomicrmw,   AtomicRMW);
772   INSTKEYWORD(fence,       Fence);
773   INSTKEYWORD(getelementptr, GetElementPtr);
774 
775   INSTKEYWORD(extractelement, ExtractElement);
776   INSTKEYWORD(insertelement,  InsertElement);
777   INSTKEYWORD(shufflevector,  ShuffleVector);
778   INSTKEYWORD(extractvalue,   ExtractValue);
779   INSTKEYWORD(insertvalue,    InsertValue);
780   INSTKEYWORD(landingpad,     LandingPad);
781   INSTKEYWORD(cleanupret,     CleanupRet);
782   INSTKEYWORD(catchret,       CatchRet);
783   INSTKEYWORD(catchswitch,  CatchSwitch);
784   INSTKEYWORD(catchpad,     CatchPad);
785   INSTKEYWORD(cleanuppad,   CleanupPad);
786 
787 #undef INSTKEYWORD
788 
789 #define DWKEYWORD(TYPE, TOKEN)                                                 \
790   do {                                                                         \
791     if (Keyword.startswith("DW_" #TYPE "_")) {                                 \
792       StrVal.assign(Keyword.begin(), Keyword.end());                           \
793       return lltok::TOKEN;                                                     \
794     }                                                                          \
795   } while (false)
796 
797   DWKEYWORD(TAG, DwarfTag);
798   DWKEYWORD(ATE, DwarfAttEncoding);
799   DWKEYWORD(VIRTUALITY, DwarfVirtuality);
800   DWKEYWORD(LANG, DwarfLang);
801   DWKEYWORD(CC, DwarfCC);
802   DWKEYWORD(OP, DwarfOp);
803   DWKEYWORD(MACINFO, DwarfMacinfo);
804 
805 #undef DWKEYWORD
806 
807   if (Keyword.startswith("DIFlag")) {
808     StrVal.assign(Keyword.begin(), Keyword.end());
809     return lltok::DIFlag;
810   }
811 
812   if (Keyword.startswith("CSK_")) {
813     StrVal.assign(Keyword.begin(), Keyword.end());
814     return lltok::ChecksumKind;
815   }
816 
817   if (Keyword == "NoDebug" || Keyword == "FullDebug" ||
818       Keyword == "LineTablesOnly") {
819     StrVal.assign(Keyword.begin(), Keyword.end());
820     return lltok::EmissionKind;
821   }
822 
823   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
824   // the CFE to avoid forcing it to deal with 64-bit numbers.
825   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
826       TokStart[1] == '0' && TokStart[2] == 'x' &&
827       isxdigit(static_cast<unsigned char>(TokStart[3]))) {
828     int len = CurPtr-TokStart-3;
829     uint32_t bits = len * 4;
830     StringRef HexStr(TokStart + 3, len);
831     if (!all_of(HexStr, isxdigit)) {
832       // Bad token, return it as an error.
833       CurPtr = TokStart+3;
834       return lltok::Error;
835     }
836     APInt Tmp(bits, HexStr, 16);
837     uint32_t activeBits = Tmp.getActiveBits();
838     if (activeBits > 0 && activeBits < bits)
839       Tmp = Tmp.trunc(activeBits);
840     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
841     return lltok::APSInt;
842   }
843 
844   // If this is "cc1234", return this as just "cc".
845   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
846     CurPtr = TokStart+2;
847     return lltok::kw_cc;
848   }
849 
850   // Finally, if this isn't known, return an error.
851   CurPtr = TokStart+1;
852   return lltok::Error;
853 }
854 
855 /// Lex all tokens that start with a 0x prefix, knowing they match and are not
856 /// labels.
857 ///    HexFPConstant     0x[0-9A-Fa-f]+
858 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
859 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
860 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
861 ///    HexHalfConstant   0xH[0-9A-Fa-f]+
862 lltok::Kind LLLexer::Lex0x() {
863   CurPtr = TokStart + 2;
864 
865   char Kind;
866   if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
867     Kind = *CurPtr++;
868   } else {
869     Kind = 'J';
870   }
871 
872   if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
873     // Bad token, return it as an error.
874     CurPtr = TokStart+1;
875     return lltok::Error;
876   }
877 
878   while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
879     ++CurPtr;
880 
881   if (Kind == 'J') {
882     // HexFPConstant - Floating point constant represented in IEEE format as a
883     // hexadecimal number for when exponential notation is not precise enough.
884     // Half, Float, and double only.
885     APFloatVal = APFloat(APFloat::IEEEdouble(),
886                          APInt(64, HexIntToVal(TokStart + 2, CurPtr)));
887     return lltok::APFloat;
888   }
889 
890   uint64_t Pair[2];
891   switch (Kind) {
892   default: llvm_unreachable("Unknown kind!");
893   case 'K':
894     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
895     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
896     APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair));
897     return lltok::APFloat;
898   case 'L':
899     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
900     HexToIntPair(TokStart+3, CurPtr, Pair);
901     APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair));
902     return lltok::APFloat;
903   case 'M':
904     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
905     HexToIntPair(TokStart+3, CurPtr, Pair);
906     APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair));
907     return lltok::APFloat;
908   case 'H':
909     APFloatVal = APFloat(APFloat::IEEEhalf(),
910                          APInt(16,HexIntToVal(TokStart+3, CurPtr)));
911     return lltok::APFloat;
912   }
913 }
914 
915 /// Lex tokens for a label or a numeric constant, possibly starting with -.
916 ///    Label             [-a-zA-Z$._0-9]+:
917 ///    NInteger          -[0-9]+
918 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
919 ///    PInteger          [0-9]+
920 ///    HexFPConstant     0x[0-9A-Fa-f]+
921 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
922 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
923 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
924 lltok::Kind LLLexer::LexDigitOrNegative() {
925   // If the letter after the negative is not a number, this is probably a label.
926   if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
927       !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
928     // Okay, this is not a number after the -, it's probably a label.
929     if (const char *End = isLabelTail(CurPtr)) {
930       StrVal.assign(TokStart, End-1);
931       CurPtr = End;
932       return lltok::LabelStr;
933     }
934 
935     return lltok::Error;
936   }
937 
938   // At this point, it is either a label, int or fp constant.
939 
940   // Skip digits, we have at least one.
941   for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
942     /*empty*/;
943 
944   // Check to see if this really is a label afterall, e.g. "-1:".
945   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
946     if (const char *End = isLabelTail(CurPtr)) {
947       StrVal.assign(TokStart, End-1);
948       CurPtr = End;
949       return lltok::LabelStr;
950     }
951   }
952 
953   // If the next character is a '.', then it is a fp value, otherwise its
954   // integer.
955   if (CurPtr[0] != '.') {
956     if (TokStart[0] == '0' && TokStart[1] == 'x')
957       return Lex0x();
958     APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart));
959     return lltok::APSInt;
960   }
961 
962   ++CurPtr;
963 
964   // Skip over [0-9]*([eE][-+]?[0-9]+)?
965   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
966 
967   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
968     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
969         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
970           isdigit(static_cast<unsigned char>(CurPtr[2])))) {
971       CurPtr += 2;
972       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
973     }
974   }
975 
976   APFloatVal = APFloat(APFloat::IEEEdouble(),
977                        StringRef(TokStart, CurPtr - TokStart));
978   return lltok::APFloat;
979 }
980 
981 /// Lex a floating point constant starting with +.
982 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
983 lltok::Kind LLLexer::LexPositive() {
984   // If the letter after the negative is a number, this is probably not a
985   // label.
986   if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
987     return lltok::Error;
988 
989   // Skip digits.
990   for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
991     /*empty*/;
992 
993   // At this point, we need a '.'.
994   if (CurPtr[0] != '.') {
995     CurPtr = TokStart+1;
996     return lltok::Error;
997   }
998 
999   ++CurPtr;
1000 
1001   // Skip over [0-9]*([eE][-+]?[0-9]+)?
1002   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1003 
1004   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1005     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1006         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1007         isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1008       CurPtr += 2;
1009       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1010     }
1011   }
1012 
1013   APFloatVal = APFloat(APFloat::IEEEdouble(),
1014                        StringRef(TokStart, CurPtr - TokStart));
1015   return lltok::APFloat;
1016 }
1017