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/StringExtras.h"
16 #include "llvm/ADT/Twine.h"
17 #include "llvm/AsmParser/Parser.h"
18 #include "llvm/IR/DerivedTypes.h"
19 #include "llvm/IR/Instruction.h"
20 #include "llvm/IR/LLVMContext.h"
21 #include "llvm/Support/ErrorHandling.h"
22 #include "llvm/Support/MathExtras.h"
23 #include "llvm/Support/MemoryBuffer.h"
24 #include "llvm/Support/SourceMgr.h"
25 #include "llvm/Support/raw_ostream.h"
26 #include <cctype>
27 #include <cstdio>
28 #include <cstdlib>
29 #include <cstring>
30 using namespace llvm;
31 
32 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
33   ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
34   return true;
35 }
36 
37 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
38   SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
39 }
40 
41 //===----------------------------------------------------------------------===//
42 // Helper functions.
43 //===----------------------------------------------------------------------===//
44 
45 // atoull - Convert an ascii string of decimal digits into the unsigned long
46 // long representation... this does not have to do input error checking,
47 // because we know that the input will be matched by a suitable regex...
48 //
49 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
50   uint64_t Result = 0;
51   for (; Buffer != End; Buffer++) {
52     uint64_t OldRes = Result;
53     Result *= 10;
54     Result += *Buffer-'0';
55     if (Result < OldRes) {  // Uh, oh, overflow detected!!!
56       Error("constant bigger than 64 bits detected!");
57       return 0;
58     }
59   }
60   return Result;
61 }
62 
63 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
64   uint64_t Result = 0;
65   for (; Buffer != End; ++Buffer) {
66     uint64_t OldRes = Result;
67     Result *= 16;
68     Result += hexDigitValue(*Buffer);
69 
70     if (Result < OldRes) {   // Uh, oh, overflow detected!!!
71       Error("constant bigger than 64 bits detected!");
72       return 0;
73     }
74   }
75   return Result;
76 }
77 
78 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
79                            uint64_t Pair[2]) {
80   Pair[0] = 0;
81   if (End - Buffer >= 16) {
82     for (int i = 0; i < 16; i++, Buffer++) {
83       assert(Buffer != End);
84       Pair[0] *= 16;
85       Pair[0] += hexDigitValue(*Buffer);
86     }
87   }
88   Pair[1] = 0;
89   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
90     Pair[1] *= 16;
91     Pair[1] += hexDigitValue(*Buffer);
92   }
93   if (Buffer != End)
94     Error("constant bigger than 128 bits detected!");
95 }
96 
97 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
98 /// { low64, high16 } as usual for an APInt.
99 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
100                            uint64_t Pair[2]) {
101   Pair[1] = 0;
102   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
103     assert(Buffer != End);
104     Pair[1] *= 16;
105     Pair[1] += hexDigitValue(*Buffer);
106   }
107   Pair[0] = 0;
108   for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
109     Pair[0] *= 16;
110     Pair[0] += hexDigitValue(*Buffer);
111   }
112   if (Buffer != End)
113     Error("constant bigger than 128 bits detected!");
114 }
115 
116 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
117 // appropriate character.
118 static void UnEscapeLexed(std::string &Str) {
119   if (Str.empty()) return;
120 
121   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
122   char *BOut = Buffer;
123   for (char *BIn = Buffer; BIn != EndBuffer; ) {
124     if (BIn[0] == '\\') {
125       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
126         *BOut++ = '\\'; // Two \ becomes one
127         BIn += 2;
128       } else if (BIn < EndBuffer-2 &&
129                  isxdigit(static_cast<unsigned char>(BIn[1])) &&
130                  isxdigit(static_cast<unsigned char>(BIn[2]))) {
131         *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
132         BIn += 3;                           // Skip over handled chars
133         ++BOut;
134       } else {
135         *BOut++ = *BIn++;
136       }
137     } else {
138       *BOut++ = *BIn++;
139     }
140   }
141   Str.resize(BOut-Buffer);
142 }
143 
144 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
145 static bool isLabelChar(char C) {
146   return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
147          C == '.' || C == '_';
148 }
149 
150 
151 /// isLabelTail - Return true if this pointer points to a valid end of a label.
152 static const char *isLabelTail(const char *CurPtr) {
153   while (1) {
154     if (CurPtr[0] == ':') return CurPtr+1;
155     if (!isLabelChar(CurPtr[0])) return nullptr;
156     ++CurPtr;
157   }
158 }
159 
160 
161 
162 //===----------------------------------------------------------------------===//
163 // Lexer definition.
164 //===----------------------------------------------------------------------===//
165 
166 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &sm, SMDiagnostic &Err,
167                  LLVMContext &C)
168   : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
169   CurPtr = CurBuf.begin();
170 }
171 
172 int LLLexer::getNextChar() {
173   char CurChar = *CurPtr++;
174   switch (CurChar) {
175   default: return (unsigned char)CurChar;
176   case 0:
177     // A nul character in the stream is either the end of the current buffer or
178     // a random nul in the file.  Disambiguate that here.
179     if (CurPtr-1 != CurBuf.end())
180       return 0;  // Just whitespace.
181 
182     // Otherwise, return end of file.
183     --CurPtr;  // Another call to lex will return EOF again.
184     return EOF;
185   }
186 }
187 
188 
189 lltok::Kind LLLexer::LexToken() {
190   TokStart = CurPtr;
191 
192   int CurChar = getNextChar();
193   switch (CurChar) {
194   default:
195     // Handle letters: [a-zA-Z_]
196     if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
197       return LexIdentifier();
198 
199     return lltok::Error;
200   case EOF: return lltok::Eof;
201   case 0:
202   case ' ':
203   case '\t':
204   case '\n':
205   case '\r':
206     // Ignore whitespace.
207     return LexToken();
208   case '+': return LexPositive();
209   case '@': return LexAt();
210   case '$': return LexDollar();
211   case '%': return LexPercent();
212   case '"': return LexQuote();
213   case '.':
214     if (const char *Ptr = isLabelTail(CurPtr)) {
215       CurPtr = Ptr;
216       StrVal.assign(TokStart, CurPtr-1);
217       return lltok::LabelStr;
218     }
219     if (CurPtr[0] == '.' && CurPtr[1] == '.') {
220       CurPtr += 2;
221       return lltok::dotdotdot;
222     }
223     return lltok::Error;
224   case ';':
225     SkipLineComment();
226     return LexToken();
227   case '!': return LexExclaim();
228   case '#': return LexHash();
229   case '0': case '1': case '2': case '3': case '4':
230   case '5': case '6': case '7': case '8': case '9':
231   case '-':
232     return LexDigitOrNegative();
233   case '=': return lltok::equal;
234   case '[': return lltok::lsquare;
235   case ']': return lltok::rsquare;
236   case '{': return lltok::lbrace;
237   case '}': return lltok::rbrace;
238   case '<': return lltok::less;
239   case '>': return lltok::greater;
240   case '(': return lltok::lparen;
241   case ')': return lltok::rparen;
242   case ',': return lltok::comma;
243   case '*': return lltok::star;
244   case '|': return lltok::bar;
245   }
246 }
247 
248 void LLLexer::SkipLineComment() {
249   while (1) {
250     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
251       return;
252   }
253 }
254 
255 /// Lex all tokens that start with an @ character.
256 ///   GlobalVar   @\"[^\"]*\"
257 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
258 ///   GlobalVarID @[0-9]+
259 lltok::Kind LLLexer::LexAt() {
260   return LexVar(lltok::GlobalVar, lltok::GlobalID);
261 }
262 
263 lltok::Kind LLLexer::LexDollar() {
264   if (const char *Ptr = isLabelTail(TokStart)) {
265     CurPtr = Ptr;
266     StrVal.assign(TokStart, CurPtr - 1);
267     return lltok::LabelStr;
268   }
269 
270   // Handle DollarStringConstant: $\"[^\"]*\"
271   if (CurPtr[0] == '"') {
272     ++CurPtr;
273 
274     while (1) {
275       int CurChar = getNextChar();
276 
277       if (CurChar == EOF) {
278         Error("end of file in COMDAT variable name");
279         return lltok::Error;
280       }
281       if (CurChar == '"') {
282         StrVal.assign(TokStart + 2, CurPtr - 1);
283         UnEscapeLexed(StrVal);
284         if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
285           Error("Null bytes are not allowed in names");
286           return lltok::Error;
287         }
288         return lltok::ComdatVar;
289       }
290     }
291   }
292 
293   // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
294   if (ReadVarName())
295     return lltok::ComdatVar;
296 
297   return lltok::Error;
298 }
299 
300 /// ReadString - Read a string until the closing quote.
301 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
302   const char *Start = CurPtr;
303   while (1) {
304     int CurChar = getNextChar();
305 
306     if (CurChar == EOF) {
307       Error("end of file in string constant");
308       return lltok::Error;
309     }
310     if (CurChar == '"') {
311       StrVal.assign(Start, CurPtr-1);
312       UnEscapeLexed(StrVal);
313       return kind;
314     }
315   }
316 }
317 
318 /// ReadVarName - Read the rest of a token containing a variable name.
319 bool LLLexer::ReadVarName() {
320   const char *NameStart = CurPtr;
321   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
322       CurPtr[0] == '-' || CurPtr[0] == '$' ||
323       CurPtr[0] == '.' || CurPtr[0] == '_') {
324     ++CurPtr;
325     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
326            CurPtr[0] == '-' || CurPtr[0] == '$' ||
327            CurPtr[0] == '.' || CurPtr[0] == '_')
328       ++CurPtr;
329 
330     StrVal.assign(NameStart, CurPtr);
331     return true;
332   }
333   return false;
334 }
335 
336 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
337   // Handle StringConstant: \"[^\"]*\"
338   if (CurPtr[0] == '"') {
339     ++CurPtr;
340 
341     while (1) {
342       int CurChar = getNextChar();
343 
344       if (CurChar == EOF) {
345         Error("end of file in global variable name");
346         return lltok::Error;
347       }
348       if (CurChar == '"') {
349         StrVal.assign(TokStart+2, CurPtr-1);
350         UnEscapeLexed(StrVal);
351         if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
352           Error("Null bytes are not allowed in names");
353           return lltok::Error;
354         }
355         return Var;
356       }
357     }
358   }
359 
360   // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
361   if (ReadVarName())
362     return Var;
363 
364   // Handle VarID: [0-9]+
365   if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
366     for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
367       /*empty*/;
368 
369     uint64_t Val = atoull(TokStart+1, CurPtr);
370     if ((unsigned)Val != Val)
371       Error("invalid value number (too large)!");
372     UIntVal = unsigned(Val);
373     return VarID;
374   }
375   return lltok::Error;
376 }
377 
378 /// Lex all tokens that start with a % character.
379 ///   LocalVar   ::= %\"[^\"]*\"
380 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
381 ///   LocalVarID ::= %[0-9]+
382 lltok::Kind LLLexer::LexPercent() {
383   return LexVar(lltok::LocalVar, lltok::LocalVarID);
384 }
385 
386 /// Lex all tokens that start with a " character.
387 ///   QuoteLabel        "[^"]+":
388 ///   StringConstant    "[^"]*"
389 lltok::Kind LLLexer::LexQuote() {
390   lltok::Kind kind = ReadString(lltok::StringConstant);
391   if (kind == lltok::Error || kind == lltok::Eof)
392     return kind;
393 
394   if (CurPtr[0] == ':') {
395     ++CurPtr;
396     if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
397       Error("Null bytes are not allowed in names");
398       kind = lltok::Error;
399     } else {
400       kind = lltok::LabelStr;
401     }
402   }
403 
404   return kind;
405 }
406 
407 /// Lex all tokens that start with a ! character.
408 ///    !foo
409 ///    !
410 lltok::Kind LLLexer::LexExclaim() {
411   // Lex a metadata name as a MetadataVar.
412   if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
413       CurPtr[0] == '-' || CurPtr[0] == '$' ||
414       CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
415     ++CurPtr;
416     while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
417            CurPtr[0] == '-' || CurPtr[0] == '$' ||
418            CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
419       ++CurPtr;
420 
421     StrVal.assign(TokStart+1, CurPtr);   // Skip !
422     UnEscapeLexed(StrVal);
423     return lltok::MetadataVar;
424   }
425   return lltok::exclaim;
426 }
427 
428 /// Lex all tokens that start with a # character.
429 ///    AttrGrpID ::= #[0-9]+
430 lltok::Kind LLLexer::LexHash() {
431   // Handle AttrGrpID: #[0-9]+
432   if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
433     for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
434       /*empty*/;
435 
436     uint64_t Val = atoull(TokStart+1, CurPtr);
437     if ((unsigned)Val != Val)
438       Error("invalid value number (too large)!");
439     UIntVal = unsigned(Val);
440     return lltok::AttrGrpID;
441   }
442 
443   return lltok::Error;
444 }
445 
446 /// Lex a label, integer type, keyword, or hexadecimal integer constant.
447 ///    Label           [-a-zA-Z$._0-9]+:
448 ///    IntegerType     i[0-9]+
449 ///    Keyword         sdiv, float, ...
450 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
451 lltok::Kind LLLexer::LexIdentifier() {
452   const char *StartChar = CurPtr;
453   const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
454   const char *KeywordEnd = nullptr;
455 
456   for (; isLabelChar(*CurPtr); ++CurPtr) {
457     // If we decide this is an integer, remember the end of the sequence.
458     if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
459       IntEnd = CurPtr;
460     if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
461         *CurPtr != '_')
462       KeywordEnd = CurPtr;
463   }
464 
465   // If we stopped due to a colon, this really is a label.
466   if (*CurPtr == ':') {
467     StrVal.assign(StartChar-1, CurPtr++);
468     return lltok::LabelStr;
469   }
470 
471   // Otherwise, this wasn't a label.  If this was valid as an integer type,
472   // return it.
473   if (!IntEnd) IntEnd = CurPtr;
474   if (IntEnd != StartChar) {
475     CurPtr = IntEnd;
476     uint64_t NumBits = atoull(StartChar, CurPtr);
477     if (NumBits < IntegerType::MIN_INT_BITS ||
478         NumBits > IntegerType::MAX_INT_BITS) {
479       Error("bitwidth for integer type out of range!");
480       return lltok::Error;
481     }
482     TyVal = IntegerType::get(Context, NumBits);
483     return lltok::Type;
484   }
485 
486   // Otherwise, this was a letter sequence.  See which keyword this is.
487   if (!KeywordEnd) KeywordEnd = CurPtr;
488   CurPtr = KeywordEnd;
489   --StartChar;
490   StringRef Keyword(StartChar, CurPtr - StartChar);
491 #define KEYWORD(STR)                                                           \
492   do {                                                                         \
493     if (Keyword == #STR)                                                       \
494       return lltok::kw_##STR;                                                  \
495   } while (0)
496 
497   KEYWORD(true);    KEYWORD(false);
498   KEYWORD(declare); KEYWORD(define);
499   KEYWORD(global);  KEYWORD(constant);
500 
501   KEYWORD(private);
502   KEYWORD(internal);
503   KEYWORD(available_externally);
504   KEYWORD(linkonce);
505   KEYWORD(linkonce_odr);
506   KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
507   KEYWORD(weak_odr);
508   KEYWORD(appending);
509   KEYWORD(dllimport);
510   KEYWORD(dllexport);
511   KEYWORD(common);
512   KEYWORD(default);
513   KEYWORD(hidden);
514   KEYWORD(protected);
515   KEYWORD(unnamed_addr);
516   KEYWORD(externally_initialized);
517   KEYWORD(extern_weak);
518   KEYWORD(external);
519   KEYWORD(thread_local);
520   KEYWORD(localdynamic);
521   KEYWORD(initialexec);
522   KEYWORD(localexec);
523   KEYWORD(zeroinitializer);
524   KEYWORD(undef);
525   KEYWORD(null);
526   KEYWORD(to);
527   KEYWORD(caller);
528   KEYWORD(tail);
529   KEYWORD(musttail);
530   KEYWORD(target);
531   KEYWORD(triple);
532   KEYWORD(unwind);
533   KEYWORD(deplibs);             // FIXME: Remove in 4.0.
534   KEYWORD(datalayout);
535   KEYWORD(volatile);
536   KEYWORD(atomic);
537   KEYWORD(unordered);
538   KEYWORD(monotonic);
539   KEYWORD(acquire);
540   KEYWORD(release);
541   KEYWORD(acq_rel);
542   KEYWORD(seq_cst);
543   KEYWORD(singlethread);
544 
545   KEYWORD(nnan);
546   KEYWORD(ninf);
547   KEYWORD(nsz);
548   KEYWORD(arcp);
549   KEYWORD(fast);
550   KEYWORD(nuw);
551   KEYWORD(nsw);
552   KEYWORD(exact);
553   KEYWORD(inbounds);
554   KEYWORD(align);
555   KEYWORD(addrspace);
556   KEYWORD(section);
557   KEYWORD(alias);
558   KEYWORD(module);
559   KEYWORD(asm);
560   KEYWORD(sideeffect);
561   KEYWORD(alignstack);
562   KEYWORD(inteldialect);
563   KEYWORD(gc);
564   KEYWORD(prefix);
565   KEYWORD(prologue);
566 
567   KEYWORD(ccc);
568   KEYWORD(fastcc);
569   KEYWORD(coldcc);
570   KEYWORD(x86_stdcallcc);
571   KEYWORD(x86_fastcallcc);
572   KEYWORD(x86_thiscallcc);
573   KEYWORD(x86_vectorcallcc);
574   KEYWORD(arm_apcscc);
575   KEYWORD(arm_aapcscc);
576   KEYWORD(arm_aapcs_vfpcc);
577   KEYWORD(msp430_intrcc);
578   KEYWORD(ptx_kernel);
579   KEYWORD(ptx_device);
580   KEYWORD(spir_kernel);
581   KEYWORD(spir_func);
582   KEYWORD(intel_ocl_bicc);
583   KEYWORD(x86_64_sysvcc);
584   KEYWORD(x86_64_win64cc);
585   KEYWORD(webkit_jscc);
586   KEYWORD(anyregcc);
587   KEYWORD(preserve_mostcc);
588   KEYWORD(preserve_allcc);
589   KEYWORD(ghccc);
590   KEYWORD(hhvmcc);
591   KEYWORD(hhvm_ccc);
592 
593   KEYWORD(cc);
594   KEYWORD(c);
595 
596   KEYWORD(attributes);
597 
598   KEYWORD(alwaysinline);
599   KEYWORD(argmemonly);
600   KEYWORD(builtin);
601   KEYWORD(byval);
602   KEYWORD(inalloca);
603   KEYWORD(cold);
604   KEYWORD(convergent);
605   KEYWORD(dereferenceable);
606   KEYWORD(dereferenceable_or_null);
607   KEYWORD(inlinehint);
608   KEYWORD(inreg);
609   KEYWORD(jumptable);
610   KEYWORD(minsize);
611   KEYWORD(naked);
612   KEYWORD(nest);
613   KEYWORD(noalias);
614   KEYWORD(nobuiltin);
615   KEYWORD(nocapture);
616   KEYWORD(noduplicate);
617   KEYWORD(noimplicitfloat);
618   KEYWORD(noinline);
619   KEYWORD(nonlazybind);
620   KEYWORD(nonnull);
621   KEYWORD(noredzone);
622   KEYWORD(noreturn);
623   KEYWORD(nounwind);
624   KEYWORD(optnone);
625   KEYWORD(optsize);
626   KEYWORD(readnone);
627   KEYWORD(readonly);
628   KEYWORD(returned);
629   KEYWORD(returns_twice);
630   KEYWORD(signext);
631   KEYWORD(sret);
632   KEYWORD(ssp);
633   KEYWORD(sspreq);
634   KEYWORD(sspstrong);
635   KEYWORD(safestack);
636   KEYWORD(sanitize_address);
637   KEYWORD(sanitize_thread);
638   KEYWORD(sanitize_memory);
639   KEYWORD(uwtable);
640   KEYWORD(zeroext);
641 
642   KEYWORD(type);
643   KEYWORD(opaque);
644 
645   KEYWORD(comdat);
646 
647   // Comdat types
648   KEYWORD(any);
649   KEYWORD(exactmatch);
650   KEYWORD(largest);
651   KEYWORD(noduplicates);
652   KEYWORD(samesize);
653 
654   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
655   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
656   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
657   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
658 
659   KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
660   KEYWORD(umin);
661 
662   KEYWORD(x);
663   KEYWORD(blockaddress);
664 
665   // Metadata types.
666   KEYWORD(distinct);
667 
668   // Use-list order directives.
669   KEYWORD(uselistorder);
670   KEYWORD(uselistorder_bb);
671 
672   KEYWORD(personality);
673   KEYWORD(cleanup);
674   KEYWORD(catch);
675   KEYWORD(filter);
676 #undef KEYWORD
677 
678   // Keywords for types.
679 #define TYPEKEYWORD(STR, LLVMTY)                                               \
680   do {                                                                         \
681     if (Keyword == STR) {                                                      \
682       TyVal = LLVMTY;                                                          \
683       return lltok::Type;                                                      \
684     }                                                                          \
685   } while (false)
686   TYPEKEYWORD("void",      Type::getVoidTy(Context));
687   TYPEKEYWORD("half",      Type::getHalfTy(Context));
688   TYPEKEYWORD("float",     Type::getFloatTy(Context));
689   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
690   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
691   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
692   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
693   TYPEKEYWORD("label",     Type::getLabelTy(Context));
694   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
695   TYPEKEYWORD("x86_mmx",   Type::getX86_MMXTy(Context));
696   TYPEKEYWORD("token",     Type::getTokenTy(Context));
697 #undef TYPEKEYWORD
698 
699   // Keywords for instructions.
700 #define INSTKEYWORD(STR, Enum)                                                 \
701   do {                                                                         \
702     if (Keyword == #STR) {                                                     \
703       UIntVal = Instruction::Enum;                                             \
704       return lltok::kw_##STR;                                                  \
705     }                                                                          \
706   } while (false)
707 
708   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
709   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
710   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
711   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
712   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
713   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
714   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
715   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
716 
717   INSTKEYWORD(phi,         PHI);
718   INSTKEYWORD(call,        Call);
719   INSTKEYWORD(trunc,       Trunc);
720   INSTKEYWORD(zext,        ZExt);
721   INSTKEYWORD(sext,        SExt);
722   INSTKEYWORD(fptrunc,     FPTrunc);
723   INSTKEYWORD(fpext,       FPExt);
724   INSTKEYWORD(uitofp,      UIToFP);
725   INSTKEYWORD(sitofp,      SIToFP);
726   INSTKEYWORD(fptoui,      FPToUI);
727   INSTKEYWORD(fptosi,      FPToSI);
728   INSTKEYWORD(inttoptr,    IntToPtr);
729   INSTKEYWORD(ptrtoint,    PtrToInt);
730   INSTKEYWORD(bitcast,     BitCast);
731   INSTKEYWORD(addrspacecast, AddrSpaceCast);
732   INSTKEYWORD(select,      Select);
733   INSTKEYWORD(va_arg,      VAArg);
734   INSTKEYWORD(ret,         Ret);
735   INSTKEYWORD(br,          Br);
736   INSTKEYWORD(switch,      Switch);
737   INSTKEYWORD(indirectbr,  IndirectBr);
738   INSTKEYWORD(invoke,      Invoke);
739   INSTKEYWORD(resume,      Resume);
740   INSTKEYWORD(unreachable, Unreachable);
741 
742   INSTKEYWORD(alloca,      Alloca);
743   INSTKEYWORD(load,        Load);
744   INSTKEYWORD(store,       Store);
745   INSTKEYWORD(cmpxchg,     AtomicCmpXchg);
746   INSTKEYWORD(atomicrmw,   AtomicRMW);
747   INSTKEYWORD(fence,       Fence);
748   INSTKEYWORD(getelementptr, GetElementPtr);
749 
750   INSTKEYWORD(extractelement, ExtractElement);
751   INSTKEYWORD(insertelement,  InsertElement);
752   INSTKEYWORD(shufflevector,  ShuffleVector);
753   INSTKEYWORD(extractvalue,   ExtractValue);
754   INSTKEYWORD(insertvalue,    InsertValue);
755   INSTKEYWORD(landingpad,     LandingPad);
756   INSTKEYWORD(cleanupret,     CleanupRet);
757   INSTKEYWORD(catchret,       CatchRet);
758   INSTKEYWORD(catchpad,     CatchPad);
759   INSTKEYWORD(terminatepad, TerminatePad);
760   INSTKEYWORD(cleanuppad,   CleanupPad);
761   INSTKEYWORD(catchendpad,  CatchEndPad);
762   INSTKEYWORD(cleanupendpad, CleanupEndPad);
763 #undef INSTKEYWORD
764 
765 #define DWKEYWORD(TYPE, TOKEN)                                                 \
766   do {                                                                         \
767     if (Keyword.startswith("DW_" #TYPE "_")) {                                 \
768       StrVal.assign(Keyword.begin(), Keyword.end());                           \
769       return lltok::TOKEN;                                                     \
770     }                                                                          \
771   } while (false)
772   DWKEYWORD(TAG, DwarfTag);
773   DWKEYWORD(ATE, DwarfAttEncoding);
774   DWKEYWORD(VIRTUALITY, DwarfVirtuality);
775   DWKEYWORD(LANG, DwarfLang);
776   DWKEYWORD(OP, DwarfOp);
777 #undef DWKEYWORD
778 
779   if (Keyword.startswith("DIFlag")) {
780     StrVal.assign(Keyword.begin(), Keyword.end());
781     return lltok::DIFlag;
782   }
783 
784   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
785   // the CFE to avoid forcing it to deal with 64-bit numbers.
786   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
787       TokStart[1] == '0' && TokStart[2] == 'x' &&
788       isxdigit(static_cast<unsigned char>(TokStart[3]))) {
789     int len = CurPtr-TokStart-3;
790     uint32_t bits = len * 4;
791     StringRef HexStr(TokStart + 3, len);
792     if (!std::all_of(HexStr.begin(), HexStr.end(), isxdigit)) {
793       // Bad token, return it as an error.
794       CurPtr = TokStart+3;
795       return lltok::Error;
796     }
797     APInt Tmp(bits, HexStr, 16);
798     uint32_t activeBits = Tmp.getActiveBits();
799     if (activeBits > 0 && activeBits < bits)
800       Tmp = Tmp.trunc(activeBits);
801     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
802     return lltok::APSInt;
803   }
804 
805   // If this is "cc1234", return this as just "cc".
806   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
807     CurPtr = TokStart+2;
808     return lltok::kw_cc;
809   }
810 
811   // Finally, if this isn't known, return an error.
812   CurPtr = TokStart+1;
813   return lltok::Error;
814 }
815 
816 /// Lex all tokens that start with a 0x prefix, knowing they match and are not
817 /// labels.
818 ///    HexFPConstant     0x[0-9A-Fa-f]+
819 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
820 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
821 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
822 ///    HexHalfConstant   0xH[0-9A-Fa-f]+
823 lltok::Kind LLLexer::Lex0x() {
824   CurPtr = TokStart + 2;
825 
826   char Kind;
827   if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
828     Kind = *CurPtr++;
829   } else {
830     Kind = 'J';
831   }
832 
833   if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
834     // Bad token, return it as an error.
835     CurPtr = TokStart+1;
836     return lltok::Error;
837   }
838 
839   while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
840     ++CurPtr;
841 
842   if (Kind == 'J') {
843     // HexFPConstant - Floating point constant represented in IEEE format as a
844     // hexadecimal number for when exponential notation is not precise enough.
845     // Half, Float, and double only.
846     APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
847     return lltok::APFloat;
848   }
849 
850   uint64_t Pair[2];
851   switch (Kind) {
852   default: llvm_unreachable("Unknown kind!");
853   case 'K':
854     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
855     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
856     APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair));
857     return lltok::APFloat;
858   case 'L':
859     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
860     HexToIntPair(TokStart+3, CurPtr, Pair);
861     APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair));
862     return lltok::APFloat;
863   case 'M':
864     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
865     HexToIntPair(TokStart+3, CurPtr, Pair);
866     APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair));
867     return lltok::APFloat;
868   case 'H':
869     APFloatVal = APFloat(APFloat::IEEEhalf,
870                          APInt(16,HexIntToVal(TokStart+3, CurPtr)));
871     return lltok::APFloat;
872   }
873 }
874 
875 /// Lex tokens for a label or a numeric constant, possibly starting with -.
876 ///    Label             [-a-zA-Z$._0-9]+:
877 ///    NInteger          -[0-9]+
878 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
879 ///    PInteger          [0-9]+
880 ///    HexFPConstant     0x[0-9A-Fa-f]+
881 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
882 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
883 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
884 lltok::Kind LLLexer::LexDigitOrNegative() {
885   // If the letter after the negative is not a number, this is probably a label.
886   if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
887       !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
888     // Okay, this is not a number after the -, it's probably a label.
889     if (const char *End = isLabelTail(CurPtr)) {
890       StrVal.assign(TokStart, End-1);
891       CurPtr = End;
892       return lltok::LabelStr;
893     }
894 
895     return lltok::Error;
896   }
897 
898   // At this point, it is either a label, int or fp constant.
899 
900   // Skip digits, we have at least one.
901   for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
902     /*empty*/;
903 
904   // Check to see if this really is a label afterall, e.g. "-1:".
905   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
906     if (const char *End = isLabelTail(CurPtr)) {
907       StrVal.assign(TokStart, End-1);
908       CurPtr = End;
909       return lltok::LabelStr;
910     }
911   }
912 
913   // If the next character is a '.', then it is a fp value, otherwise its
914   // integer.
915   if (CurPtr[0] != '.') {
916     if (TokStart[0] == '0' && TokStart[1] == 'x')
917       return Lex0x();
918     APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart));
919     return lltok::APSInt;
920   }
921 
922   ++CurPtr;
923 
924   // Skip over [0-9]*([eE][-+]?[0-9]+)?
925   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
926 
927   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
928     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
929         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
930           isdigit(static_cast<unsigned char>(CurPtr[2])))) {
931       CurPtr += 2;
932       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
933     }
934   }
935 
936   APFloatVal = APFloat(std::atof(TokStart));
937   return lltok::APFloat;
938 }
939 
940 /// Lex a floating point constant starting with +.
941 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
942 lltok::Kind LLLexer::LexPositive() {
943   // If the letter after the negative is a number, this is probably not a
944   // label.
945   if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
946     return lltok::Error;
947 
948   // Skip digits.
949   for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
950     /*empty*/;
951 
952   // At this point, we need a '.'.
953   if (CurPtr[0] != '.') {
954     CurPtr = TokStart+1;
955     return lltok::Error;
956   }
957 
958   ++CurPtr;
959 
960   // Skip over [0-9]*([eE][-+]?[0-9]+)?
961   while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
962 
963   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
964     if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
965         ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
966         isdigit(static_cast<unsigned char>(CurPtr[2])))) {
967       CurPtr += 2;
968       while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
969     }
970   }
971 
972   APFloatVal = APFloat(std::atof(TokStart));
973   return lltok::APFloat;
974 }
975