lib/AsmParser/LLLexer.cpp

   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/DerivedTypes.h"
  16 #include "llvm/Instruction.h"
  17 #include "llvm/LLVMContext.h"
  18 #include "llvm/Support/ErrorHandling.h"
  19 #include "llvm/Support/MemoryBuffer.h"
  20 #include "llvm/Support/MathExtras.h"
  21 #include "llvm/Support/SourceMgr.h"
  22 #include "llvm/Support/raw_ostream.h"
  23 #include "llvm/Assembly/Parser.h"
  24 #include <cstdio>
  25 #include <cstdlib>
  26 #include <cstring>
  27 using namespace llvm;
  28
  29 bool LLLexer::Error(LocTy ErrorLoc, const std::string &Msg) const {
  30   ErrorInfo = SM.GetMessage(ErrorLoc, Msg, "error");
  31   return true;
  32 }
  33
  34 //===----------------------------------------------------------------------===//
  35 // Helper functions.
  36 //===----------------------------------------------------------------------===//
  37
  38 // atoull - Convert an ascii string of decimal digits into the unsigned long
  39 // long representation... this does not have to do input error checking,
  40 // because we know that the input will be matched by a suitable regex...
  41 //
  42 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
  43   uint64_t Result = 0;
  44   for (; Buffer != End; Buffer++) {
  45     uint64_t OldRes = Result;
  46     Result *= 10;
  47     Result += *Buffer-'0';
  48     if (Result < OldRes) {  // Uh, oh, overflow detected!!!
  49       Error("constant bigger than 64 bits detected!");
  50       return 0;
  51     }
  52   }
  53   return Result;
  54 }
  55
  56 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
  57   uint64_t Result = 0;
  58   for (; Buffer != End; ++Buffer) {
  59     uint64_t OldRes = Result;
  60     Result *= 16;
  61     char C = *Buffer;
  62     if (C >= '0' && C <= '9')
  63       Result += C-'0';
  64     else if (C >= 'A' && C <= 'F')
  65       Result += C-'A'+10;
  66     else if (C >= 'a' && C <= 'f')
  67       Result += C-'a'+10;
  68
  69     if (Result < OldRes) {   // Uh, oh, overflow detected!!!
  70       Error("constant bigger than 64 bits detected!");
  71       return 0;
  72     }
  73   }
  74   return Result;
  75 }
  76
  77 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
  78                            uint64_t Pair[2]) {
  79   Pair[0] = 0;
  80   for (int i=0; i<16; i++, Buffer++) {
  81     assert(Buffer != End);
  82     Pair[0] *= 16;
  83     char C = *Buffer;
  84     if (C >= '0' && C <= '9')
  85       Pair[0] += C-'0';
  86     else if (C >= 'A' && C <= 'F')
  87       Pair[0] += C-'A'+10;
  88     else if (C >= 'a' && C <= 'f')
  89       Pair[0] += C-'a'+10;
  90   }
  91   Pair[1] = 0;
  92   for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
  93     Pair[1] *= 16;
  94     char C = *Buffer;
  95     if (C >= '0' && C <= '9')
  96       Pair[1] += C-'0';
  97     else if (C >= 'A' && C <= 'F')
  98       Pair[1] += C-'A'+10;
  99     else if (C >= 'a' && C <= 'f')
 100       Pair[1] += C-'a'+10;
 101   }
 102   if (Buffer != End)
 103     Error("constant bigger than 128 bits detected!");
 104 }
 105
 106 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
 107 /// { low64, high16 } as usual for an APInt.
 108 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
 109                            uint64_t Pair[2]) {
 110   Pair[1] = 0;
 111   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
 112     assert(Buffer != End);
 113     Pair[1] *= 16;
 114     char C = *Buffer;
 115     if (C >= '0' && C <= '9')
 116       Pair[1] += C-'0';
 117     else if (C >= 'A' && C <= 'F')
 118       Pair[1] += C-'A'+10;
 119     else if (C >= 'a' && C <= 'f')
 120       Pair[1] += C-'a'+10;
 121   }
 122   Pair[0] = 0;
 123   for (int i=0; i<16; i++, Buffer++) {
 124     Pair[0] *= 16;
 125     char C = *Buffer;
 126     if (C >= '0' && C <= '9')
 127       Pair[0] += C-'0';
 128     else if (C >= 'A' && C <= 'F')
 129       Pair[0] += C-'A'+10;
 130     else if (C >= 'a' && C <= 'f')
 131       Pair[0] += C-'a'+10;
 132   }
 133   if (Buffer != End)
 134     Error("constant bigger than 128 bits detected!");
 135 }
 136
 137 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
 138 // appropriate character.
 139 static void UnEscapeLexed(std::string &Str) {
 140   if (Str.empty()) return;
 141
 142   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
 143   char *BOut = Buffer;
 144   for (char *BIn = Buffer; BIn != EndBuffer; ) {
 145     if (BIn[0] == '\\') {
 146       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
 147         *BOut++ = '\\'; // Two \ becomes one
 148         BIn += 2;
 149       } else if (BIn < EndBuffer-2 && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
 150         char Tmp = BIn[3]; BIn[3] = 0;      // Terminate string
 151         *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number
 152         BIn[3] = Tmp;                       // Restore character
 153         BIn += 3;                           // Skip over handled chars
 154         ++BOut;
 155       } else {
 156         *BOut++ = *BIn++;
 157       }
 158     } else {
 159       *BOut++ = *BIn++;
 160     }
 161   }
 162   Str.resize(BOut-Buffer);
 163 }
 164
 165 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
 166 static bool isLabelChar(char C) {
 167   return isalnum(C) || C == '-' || C == '$' || C == '.' || C == '_';
 168 }
 169
 170
 171 /// isLabelTail - Return true if this pointer points to a valid end of a label.
 172 static const char *isLabelTail(const char *CurPtr) {
 173   while (1) {
 174     if (CurPtr[0] == ':') return CurPtr+1;
 175     if (!isLabelChar(CurPtr[0])) return 0;
 176     ++CurPtr;
 177   }
 178 }
 179
 180
 181
 182 //===----------------------------------------------------------------------===//
 183 // Lexer definition.
 184 //===----------------------------------------------------------------------===//
 185
 186 LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err,
 187                  LLVMContext &C)
 188   : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
 189   CurPtr = CurBuf->getBufferStart();
 190 }
 191
 192 std::string LLLexer::getFilename() const {
 193   return CurBuf->getBufferIdentifier();
 194 }
 195
 196 int LLLexer::getNextChar() {
 197   char CurChar = *CurPtr++;
 198   switch (CurChar) {
 199   default: return (unsigned char)CurChar;
 200   case 0:
 201     // A nul character in the stream is either the end of the current buffer or
 202     // a random nul in the file.  Disambiguate that here.
 203     if (CurPtr-1 != CurBuf->getBufferEnd())
 204       return 0;  // Just whitespace.
 205
 206     // Otherwise, return end of file.
 207     --CurPtr;  // Another call to lex will return EOF again.
 208     return EOF;
 209   }
 210 }
 211
 212
 213 lltok::Kind LLLexer::LexToken() {
 214   TokStart = CurPtr;
 215
 216   int CurChar = getNextChar();
 217   switch (CurChar) {
 218   default:
 219     // Handle letters: [a-zA-Z_]
 220     if (isalpha(CurChar) || CurChar == '_')
 221       return LexIdentifier();
 222
 223     return lltok::Error;
 224   case EOF: return lltok::Eof;
 225   case 0:
 226   case ' ':
 227   case '\t':
 228   case '\n':
 229   case '\r':
 230     // Ignore whitespace.
 231     return LexToken();
 232   case '+': return LexPositive();
 233   case '@': return LexAt();
 234   case '%': return LexPercent();
 235   case '"': return LexQuote();
 236   case '.':
 237     if (const char *Ptr = isLabelTail(CurPtr)) {
 238       CurPtr = Ptr;
 239       StrVal.assign(TokStart, CurPtr-1);
 240       return lltok::LabelStr;
 241     }
 242     if (CurPtr[0] == '.' && CurPtr[1] == '.') {
 243       CurPtr += 2;
 244       return lltok::dotdotdot;
 245     }
 246     return lltok::Error;
 247   case '$':
 248     if (const char *Ptr = isLabelTail(CurPtr)) {
 249       CurPtr = Ptr;
 250       StrVal.assign(TokStart, CurPtr-1);
 251       return lltok::LabelStr;
 252     }
 253     return lltok::Error;
 254   case ';':
 255     SkipLineComment();
 256     return LexToken();
 257   case '!': return LexExclaim();
 258   case '0': case '1': case '2': case '3': case '4':
 259   case '5': case '6': case '7': case '8': case '9':
 260   case '-':
 261     return LexDigitOrNegative();
 262   case '=': return lltok::equal;
 263   case '[': return lltok::lsquare;
 264   case ']': return lltok::rsquare;
 265   case '{': return lltok::lbrace;
 266   case '}': return lltok::rbrace;
 267   case '<': return lltok::less;
 268   case '>': return lltok::greater;
 269   case '(': return lltok::lparen;
 270   case ')': return lltok::rparen;
 271   case ',': return lltok::comma;
 272   case '*': return lltok::star;
 273   case '\\': return lltok::backslash;
 274   }
 275 }
 276
 277 void LLLexer::SkipLineComment() {
 278   while (1) {
 279     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
 280       return;
 281   }
 282 }
 283
 284 /// LexAt - Lex all tokens that start with an @ character:
 285 ///   GlobalVar   @\"[^\"]*\"
 286 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
 287 ///   GlobalVarID @[0-9]+
 288 lltok::Kind LLLexer::LexAt() {
 289   // Handle AtStringConstant: @\"[^\"]*\"
 290   if (CurPtr[0] == '"') {
 291     ++CurPtr;
 292
 293     while (1) {
 294       int CurChar = getNextChar();
 295
 296       if (CurChar == EOF) {
 297         Error("end of file in global variable name");
 298         return lltok::Error;
 299       }
 300       if (CurChar == '"') {
 301         StrVal.assign(TokStart+2, CurPtr-1);
 302         UnEscapeLexed(StrVal);
 303         return lltok::GlobalVar;
 304       }
 305     }
 306   }
 307
 308   // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
 309   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 310       CurPtr[0] == '.' || CurPtr[0] == '_') {
 311     ++CurPtr;
 312     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 313            CurPtr[0] == '.' || CurPtr[0] == '_')
 314       ++CurPtr;
 315
 316     StrVal.assign(TokStart+1, CurPtr);   // Skip @
 317     return lltok::GlobalVar;
 318   }
 319
 320   // Handle GlobalVarID: @[0-9]+
 321   if (isdigit(CurPtr[0])) {
 322     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 323       /*empty*/;
 324
 325     uint64_t Val = atoull(TokStart+1, CurPtr);
 326     if ((unsigned)Val != Val)
 327       Error("invalid value number (too large)!");
 328     UIntVal = unsigned(Val);
 329     return lltok::GlobalID;
 330   }
 331
 332   return lltok::Error;
 333 }
 334
 335
 336 /// LexPercent - Lex all tokens that start with a % character:
 337 ///   LocalVar   ::= %\"[^\"]*\"
 338 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
 339 ///   LocalVarID ::= %[0-9]+
 340 lltok::Kind LLLexer::LexPercent() {
 341   // Handle LocalVarName: %\"[^\"]*\"
 342   if (CurPtr[0] == '"') {
 343     ++CurPtr;
 344
 345     while (1) {
 346       int CurChar = getNextChar();
 347
 348       if (CurChar == EOF) {
 349         Error("end of file in string constant");
 350         return lltok::Error;
 351       }
 352       if (CurChar == '"') {
 353         StrVal.assign(TokStart+2, CurPtr-1);
 354         UnEscapeLexed(StrVal);
 355         return lltok::LocalVar;
 356       }
 357     }
 358   }
 359
 360   // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
 361   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 362       CurPtr[0] == '.' || CurPtr[0] == '_') {
 363     ++CurPtr;
 364     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 365            CurPtr[0] == '.' || CurPtr[0] == '_')
 366       ++CurPtr;
 367
 368     StrVal.assign(TokStart+1, CurPtr);   // Skip %
 369     return lltok::LocalVar;
 370   }
 371
 372   // Handle LocalVarID: %[0-9]+
 373   if (isdigit(CurPtr[0])) {
 374     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 375       /*empty*/;
 376
 377     uint64_t Val = atoull(TokStart+1, CurPtr);
 378     if ((unsigned)Val != Val)
 379       Error("invalid value number (too large)!");
 380     UIntVal = unsigned(Val);
 381     return lltok::LocalVarID;
 382   }
 383
 384   return lltok::Error;
 385 }
 386
 387 /// LexQuote - Lex all tokens that start with a " character:
 388 ///   QuoteLabel        "[^"]+":
 389 ///   StringConstant    "[^"]*"
 390 lltok::Kind LLLexer::LexQuote() {
 391   while (1) {
 392     int CurChar = getNextChar();
 393
 394     if (CurChar == EOF) {
 395       Error("end of file in quoted string");
 396       return lltok::Error;
 397     }
 398
 399     if (CurChar != '"') continue;
 400
 401     if (CurPtr[0] != ':') {
 402       StrVal.assign(TokStart+1, CurPtr-1);
 403       UnEscapeLexed(StrVal);
 404       return lltok::StringConstant;
 405     }
 406
 407     ++CurPtr;
 408     StrVal.assign(TokStart+1, CurPtr-2);
 409     UnEscapeLexed(StrVal);
 410     return lltok::LabelStr;
 411   }
 412 }
 413
 414 static bool JustWhitespaceNewLine(const char *&Ptr) {
 415   const char *ThisPtr = Ptr;
 416   while (*ThisPtr == ' ' || *ThisPtr == '\t')
 417     ++ThisPtr;
 418   if (*ThisPtr == '\n' || *ThisPtr == '\r') {
 419     Ptr = ThisPtr;
 420     return true;
 421   }
 422   return false;
 423 }
 424
 425 /// LexExclaim:
 426 ///    !foo
 427 ///    !
 428 lltok::Kind LLLexer::LexExclaim() {
 429   // Lex a metadata name as a MetadataVar.
 430   if (isalpha(CurPtr[0])) {
 431     ++CurPtr;
 432     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 433            CurPtr[0] == '.' || CurPtr[0] == '_')
 434       ++CurPtr;
 435
 436     StrVal.assign(TokStart+1, CurPtr);   // Skip !
 437     return lltok::MetadataVar;
 438   }
 439   return lltok::exclaim;
 440 }
 441
 442 /// LexIdentifier: Handle several related productions:
 443 ///    Label           [-a-zA-Z$._0-9]+:
 444 ///    IntegerType     i[0-9]+
 445 ///    Keyword         sdiv, float, ...
 446 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
 447 lltok::Kind LLLexer::LexIdentifier() {
 448   const char *StartChar = CurPtr;
 449   const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
 450   const char *KeywordEnd = 0;
 451
 452   for (; isLabelChar(*CurPtr); ++CurPtr) {
 453     // If we decide this is an integer, remember the end of the sequence.
 454     if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr;
 455     if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr;
 456   }
 457
 458   // If we stopped due to a colon, this really is a label.
 459   if (*CurPtr == ':') {
 460     StrVal.assign(StartChar-1, CurPtr++);
 461     return lltok::LabelStr;
 462   }
 463
 464   // Otherwise, this wasn't a label.  If this was valid as an integer type,
 465   // return it.
 466   if (IntEnd == 0) IntEnd = CurPtr;
 467   if (IntEnd != StartChar) {
 468     CurPtr = IntEnd;
 469     uint64_t NumBits = atoull(StartChar, CurPtr);
 470     if (NumBits < IntegerType::MIN_INT_BITS ||
 471         NumBits > IntegerType::MAX_INT_BITS) {
 472       Error("bitwidth for integer type out of range!");
 473       return lltok::Error;
 474     }
 475     TyVal = IntegerType::get(Context, NumBits);
 476     return lltok::Type;
 477   }
 478
 479   // Otherwise, this was a letter sequence.  See which keyword this is.
 480   if (KeywordEnd == 0) KeywordEnd = CurPtr;
 481   CurPtr = KeywordEnd;
 482   --StartChar;
 483   unsigned Len = CurPtr-StartChar;
 484 #define KEYWORD(STR) \
 485   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
 486     return lltok::kw_##STR;
 487
 488   KEYWORD(begin);   KEYWORD(end);
 489   KEYWORD(true);    KEYWORD(false);
 490   KEYWORD(declare); KEYWORD(define);
 491   KEYWORD(global);  KEYWORD(constant);
 492
 493   KEYWORD(private);
 494   KEYWORD(linker_private);
 495   KEYWORD(linker_private_weak);
 496   KEYWORD(internal);
 497   KEYWORD(available_externally);
 498   KEYWORD(linkonce);
 499   KEYWORD(linkonce_odr);
 500   KEYWORD(weak);
 501   KEYWORD(weak_odr);
 502   KEYWORD(appending);
 503   KEYWORD(dllimport);
 504   KEYWORD(dllexport);
 505   KEYWORD(common);
 506   KEYWORD(default);
 507   KEYWORD(hidden);
 508   KEYWORD(protected);
 509   KEYWORD(extern_weak);
 510   KEYWORD(external);
 511   KEYWORD(thread_local);
 512   KEYWORD(zeroinitializer);
 513   KEYWORD(undef);
 514   KEYWORD(null);
 515   KEYWORD(to);
 516   KEYWORD(tail);
 517   KEYWORD(target);
 518   KEYWORD(triple);
 519   KEYWORD(deplibs);
 520   KEYWORD(datalayout);
 521   KEYWORD(volatile);
 522   KEYWORD(nuw);
 523   KEYWORD(nsw);
 524   KEYWORD(exact);
 525   KEYWORD(inbounds);
 526   KEYWORD(align);
 527   KEYWORD(addrspace);
 528   KEYWORD(section);
 529   KEYWORD(alias);
 530   KEYWORD(module);
 531   KEYWORD(asm);
 532   KEYWORD(sideeffect);
 533   KEYWORD(alignstack);
 534   KEYWORD(gc);
 535
 536   KEYWORD(ccc);
 537   KEYWORD(fastcc);
 538   KEYWORD(coldcc);
 539   KEYWORD(x86_stdcallcc);
 540   KEYWORD(x86_fastcallcc);
 541   KEYWORD(x86_thiscallcc);
 542   KEYWORD(arm_apcscc);
 543   KEYWORD(arm_aapcscc);
 544   KEYWORD(arm_aapcs_vfpcc);
 545   KEYWORD(msp430_intrcc);
 546
 547   KEYWORD(cc);
 548   KEYWORD(c);
 549
 550   KEYWORD(signext);
 551   KEYWORD(zeroext);
 552   KEYWORD(inreg);
 553   KEYWORD(sret);
 554   KEYWORD(nounwind);
 555   KEYWORD(noreturn);
 556   KEYWORD(noalias);
 557   KEYWORD(nocapture);
 558   KEYWORD(byval);
 559   KEYWORD(nest);
 560   KEYWORD(readnone);
 561   KEYWORD(readonly);
 562
 563   KEYWORD(inlinehint);
 564   KEYWORD(noinline);
 565   KEYWORD(alwaysinline);
 566   KEYWORD(optsize);
 567   KEYWORD(ssp);
 568   KEYWORD(sspreq);
 569   KEYWORD(noredzone);
 570   KEYWORD(noimplicitfloat);
 571   KEYWORD(naked);
 572
 573   KEYWORD(type);
 574   KEYWORD(opaque);
 575   KEYWORD(union);
 576
 577   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
 578   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
 579   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
 580   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
 581
 582   KEYWORD(x);
 583   KEYWORD(blockaddress);
 584 #undef KEYWORD
 585
 586   // Keywords for types.
 587 #define TYPEKEYWORD(STR, LLVMTY) \
 588   if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
 589     TyVal = LLVMTY; return lltok::Type; }
 590   TYPEKEYWORD("void",      Type::getVoidTy(Context));
 591   TYPEKEYWORD("float",     Type::getFloatTy(Context));
 592   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
 593   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
 594   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
 595   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
 596   TYPEKEYWORD("label",     Type::getLabelTy(Context));
 597   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
 598 #undef TYPEKEYWORD
 599
 600   // Handle special forms for autoupgrading.  Drop these in LLVM 3.0.  This is
 601   // to avoid conflicting with the sext/zext instructions, below.
 602   if (Len == 4 && !memcmp(StartChar, "sext", 4)) {
 603     // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
 604     if (JustWhitespaceNewLine(CurPtr))
 605       return lltok::kw_signext;
 606   } else if (Len == 4 && !memcmp(StartChar, "zext", 4)) {
 607     // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
 608     if (JustWhitespaceNewLine(CurPtr))
 609       return lltok::kw_zeroext;
 610   } else if (Len == 6 && !memcmp(StartChar, "malloc", 6)) {
 611     // FIXME: Remove in LLVM 3.0.
 612     // Autoupgrade malloc instruction.
 613     return lltok::kw_malloc;
 614   } else if (Len == 4 && !memcmp(StartChar, "free", 4)) {
 615     // FIXME: Remove in LLVM 3.0.
 616     // Autoupgrade malloc instruction.
 617     return lltok::kw_free;
 618   }
 619
 620   // Keywords for instructions.
 621 #define INSTKEYWORD(STR, Enum) \
 622   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
 623     UIntVal = Instruction::Enum; return lltok::kw_##STR; }
 624
 625   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
 626   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
 627   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
 628   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
 629   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
 630   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
 631   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
 632   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
 633
 634   INSTKEYWORD(phi,         PHI);
 635   INSTKEYWORD(call,        Call);
 636   INSTKEYWORD(trunc,       Trunc);
 637   INSTKEYWORD(zext,        ZExt);
 638   INSTKEYWORD(sext,        SExt);
 639   INSTKEYWORD(fptrunc,     FPTrunc);
 640   INSTKEYWORD(fpext,       FPExt);
 641   INSTKEYWORD(uitofp,      UIToFP);
 642   INSTKEYWORD(sitofp,      SIToFP);
 643   INSTKEYWORD(fptoui,      FPToUI);
 644   INSTKEYWORD(fptosi,      FPToSI);
 645   INSTKEYWORD(inttoptr,    IntToPtr);
 646   INSTKEYWORD(ptrtoint,    PtrToInt);
 647   INSTKEYWORD(bitcast,     BitCast);
 648   INSTKEYWORD(select,      Select);
 649   INSTKEYWORD(va_arg,      VAArg);
 650   INSTKEYWORD(ret,         Ret);
 651   INSTKEYWORD(br,          Br);
 652   INSTKEYWORD(switch,      Switch);
 653   INSTKEYWORD(indirectbr,  IndirectBr);
 654   INSTKEYWORD(invoke,      Invoke);
 655   INSTKEYWORD(unwind,      Unwind);
 656   INSTKEYWORD(unreachable, Unreachable);
 657
 658   INSTKEYWORD(alloca,      Alloca);
 659   INSTKEYWORD(load,        Load);
 660   INSTKEYWORD(store,       Store);
 661   INSTKEYWORD(getelementptr, GetElementPtr);
 662
 663   INSTKEYWORD(extractelement, ExtractElement);
 664   INSTKEYWORD(insertelement,  InsertElement);
 665   INSTKEYWORD(shufflevector,  ShuffleVector);
 666   INSTKEYWORD(getresult,      ExtractValue);
 667   INSTKEYWORD(extractvalue,   ExtractValue);
 668   INSTKEYWORD(insertvalue,    InsertValue);
 669 #undef INSTKEYWORD
 670
 671   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
 672   // the CFE to avoid forcing it to deal with 64-bit numbers.
 673   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
 674       TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) {
 675     int len = CurPtr-TokStart-3;
 676     uint32_t bits = len * 4;
 677     APInt Tmp(bits, StringRef(TokStart+3, len), 16);
 678     uint32_t activeBits = Tmp.getActiveBits();
 679     if (activeBits > 0 && activeBits < bits)
 680       Tmp.trunc(activeBits);
 681     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
 682     return lltok::APSInt;
 683   }
 684
 685   // If this is "cc1234", return this as just "cc".
 686   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
 687     CurPtr = TokStart+2;
 688     return lltok::kw_cc;
 689   }
 690
 691   // If this starts with "call", return it as CALL.  This is to support old
 692   // broken .ll files.  FIXME: remove this with LLVM 3.0.
 693   if (CurPtr-TokStart > 4 && !memcmp(TokStart, "call", 4)) {
 694     CurPtr = TokStart+4;
 695     UIntVal = Instruction::Call;
 696     return lltok::kw_call;
 697   }
 698
 699   // Finally, if this isn't known, return an error.
 700   CurPtr = TokStart+1;
 701   return lltok::Error;
 702 }
 703
 704
 705 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
 706 /// that this is not a label:
 707 ///    HexFPConstant     0x[0-9A-Fa-f]+
 708 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 709 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 710 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 711 lltok::Kind LLLexer::Lex0x() {
 712   CurPtr = TokStart + 2;
 713
 714   char Kind;
 715   if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') {
 716     Kind = *CurPtr++;
 717   } else {
 718     Kind = 'J';
 719   }
 720
 721   if (!isxdigit(CurPtr[0])) {
 722     // Bad token, return it as an error.
 723     CurPtr = TokStart+1;
 724     return lltok::Error;
 725   }
 726
 727   while (isxdigit(CurPtr[0]))
 728     ++CurPtr;
 729
 730   if (Kind == 'J') {
 731     // HexFPConstant - Floating point constant represented in IEEE format as a
 732     // hexadecimal number for when exponential notation is not precise enough.
 733     // Float and double only.
 734     APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
 735     return lltok::APFloat;
 736   }
 737
 738   uint64_t Pair[2];
 739   switch (Kind) {
 740   default: llvm_unreachable("Unknown kind!");
 741   case 'K':
 742     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
 743     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
 744     APFloatVal = APFloat(APInt(80, 2, Pair));
 745     return lltok::APFloat;
 746   case 'L':
 747     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
 748     HexToIntPair(TokStart+3, CurPtr, Pair);
 749     APFloatVal = APFloat(APInt(128, 2, Pair), true);
 750     return lltok::APFloat;
 751   case 'M':
 752     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
 753     HexToIntPair(TokStart+3, CurPtr, Pair);
 754     APFloatVal = APFloat(APInt(128, 2, Pair));
 755     return lltok::APFloat;
 756   }
 757 }
 758
 759 /// LexIdentifier: Handle several related productions:
 760 ///    Label             [-a-zA-Z$._0-9]+:
 761 ///    NInteger          -[0-9]+
 762 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 763 ///    PInteger          [0-9]+
 764 ///    HexFPConstant     0x[0-9A-Fa-f]+
 765 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 766 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 767 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 768 lltok::Kind LLLexer::LexDigitOrNegative() {
 769   // If the letter after the negative is a number, this is probably a label.
 770   if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) {
 771     // Okay, this is not a number after the -, it's probably a label.
 772     if (const char *End = isLabelTail(CurPtr)) {
 773       StrVal.assign(TokStart, End-1);
 774       CurPtr = End;
 775       return lltok::LabelStr;
 776     }
 777
 778     return lltok::Error;
 779   }
 780
 781   // At this point, it is either a label, int or fp constant.
 782
 783   // Skip digits, we have at least one.
 784   for (; isdigit(CurPtr[0]); ++CurPtr)
 785     /*empty*/;
 786
 787   // Check to see if this really is a label afterall, e.g. "-1:".
 788   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
 789     if (const char *End = isLabelTail(CurPtr)) {
 790       StrVal.assign(TokStart, End-1);
 791       CurPtr = End;
 792       return lltok::LabelStr;
 793     }
 794   }
 795
 796   // If the next character is a '.', then it is a fp value, otherwise its
 797   // integer.
 798   if (CurPtr[0] != '.') {
 799     if (TokStart[0] == '0' && TokStart[1] == 'x')
 800       return Lex0x();
 801     unsigned Len = CurPtr-TokStart;
 802     uint32_t numBits = ((Len * 64) / 19) + 2;
 803     APInt Tmp(numBits, StringRef(TokStart, Len), 10);
 804     if (TokStart[0] == '-') {
 805       uint32_t minBits = Tmp.getMinSignedBits();
 806       if (minBits > 0 && minBits < numBits)
 807         Tmp.trunc(minBits);
 808       APSIntVal = APSInt(Tmp, false);
 809     } else {
 810       uint32_t activeBits = Tmp.getActiveBits();
 811       if (activeBits > 0 && activeBits < numBits)
 812         Tmp.trunc(activeBits);
 813       APSIntVal = APSInt(Tmp, true);
 814     }
 815     return lltok::APSInt;
 816   }
 817
 818   ++CurPtr;
 819
 820   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 821   while (isdigit(CurPtr[0])) ++CurPtr;
 822
 823   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 824     if (isdigit(CurPtr[1]) ||
 825         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 826       CurPtr += 2;
 827       while (isdigit(CurPtr[0])) ++CurPtr;
 828     }
 829   }
 830
 831   APFloatVal = APFloat(atof(TokStart));
 832   return lltok::APFloat;
 833 }
 834
 835 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 836 lltok::Kind LLLexer::LexPositive() {
 837   // If the letter after the negative is a number, this is probably not a
 838   // label.
 839   if (!isdigit(CurPtr[0]))
 840     return lltok::Error;
 841
 842   // Skip digits.
 843   for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 844     /*empty*/;
 845
 846   // At this point, we need a '.'.
 847   if (CurPtr[0] != '.') {
 848     CurPtr = TokStart+1;
 849     return lltok::Error;
 850   }
 851
 852   ++CurPtr;
 853
 854   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 855   while (isdigit(CurPtr[0])) ++CurPtr;
 856
 857   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 858     if (isdigit(CurPtr[1]) ||
 859         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 860       CurPtr += 2;
 861       while (isdigit(CurPtr[0])) ++CurPtr;
 862     }
 863   }
 864
 865   APFloatVal = APFloat(atof(TokStart));
 866   return lltok::APFloat;
 867 }