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