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/ADT/Twine.h"
  19 #include "llvm/Assembly/Parser.h"
  20 #include "llvm/Support/ErrorHandling.h"
  21 #include "llvm/Support/MemoryBuffer.h"
  22 #include "llvm/Support/MathExtras.h"
  23 #include "llvm/Support/SourceMgr.h"
  24 #include "llvm/Support/raw_ostream.h"
  25 #include <cctype>
  26 #include <cstdio>
  27 #include <cstdlib>
  28 #include <cstring>
  29 using namespace llvm;
  30
  31 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
  32   ErrorInfo = SM.GetMessage(ErrorLoc, Msg, "error");
  33   return true;
  34 }
  35
  36 //===----------------------------------------------------------------------===//
  37 // Helper functions.
  38 //===----------------------------------------------------------------------===//
  39
  40 // atoull - Convert an ascii string of decimal digits into the unsigned long
  41 // long representation... this does not have to do input error checking,
  42 // because we know that the input will be matched by a suitable regex...
  43 //
  44 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
  45   uint64_t Result = 0;
  46   for (; Buffer != End; Buffer++) {
  47     uint64_t OldRes = Result;
  48     Result *= 10;
  49     Result += *Buffer-'0';
  50     if (Result < OldRes) {  // Uh, oh, overflow detected!!!
  51       Error("constant bigger than 64 bits detected!");
  52       return 0;
  53     }
  54   }
  55   return Result;
  56 }
  57
  58 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
  59   uint64_t Result = 0;
  60   for (; Buffer != End; ++Buffer) {
  61     uint64_t OldRes = Result;
  62     Result *= 16;
  63     char C = *Buffer;
  64     if (C >= '0' && C <= '9')
  65       Result += C-'0';
  66     else if (C >= 'A' && C <= 'F')
  67       Result += C-'A'+10;
  68     else if (C >= 'a' && C <= 'f')
  69       Result += C-'a'+10;
  70
  71     if (Result < OldRes) {   // Uh, oh, overflow detected!!!
  72       Error("constant bigger than 64 bits detected!");
  73       return 0;
  74     }
  75   }
  76   return Result;
  77 }
  78
  79 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
  80                            uint64_t Pair[2]) {
  81   Pair[0] = 0;
  82   for (int i=0; i<16; i++, Buffer++) {
  83     assert(Buffer != End);
  84     Pair[0] *= 16;
  85     char C = *Buffer;
  86     if (C >= '0' && C <= '9')
  87       Pair[0] += C-'0';
  88     else if (C >= 'A' && C <= 'F')
  89       Pair[0] += C-'A'+10;
  90     else if (C >= 'a' && C <= 'f')
  91       Pair[0] += C-'a'+10;
  92   }
  93   Pair[1] = 0;
  94   for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
  95     Pair[1] *= 16;
  96     char C = *Buffer;
  97     if (C >= '0' && C <= '9')
  98       Pair[1] += C-'0';
  99     else if (C >= 'A' && C <= 'F')
 100       Pair[1] += C-'A'+10;
 101     else if (C >= 'a' && C <= 'f')
 102       Pair[1] += C-'a'+10;
 103   }
 104   if (Buffer != End)
 105     Error("constant bigger than 128 bits detected!");
 106 }
 107
 108 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
 109 /// { low64, high16 } as usual for an APInt.
 110 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
 111                            uint64_t Pair[2]) {
 112   Pair[1] = 0;
 113   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
 114     assert(Buffer != End);
 115     Pair[1] *= 16;
 116     char C = *Buffer;
 117     if (C >= '0' && C <= '9')
 118       Pair[1] += C-'0';
 119     else if (C >= 'A' && C <= 'F')
 120       Pair[1] += C-'A'+10;
 121     else if (C >= 'a' && C <= 'f')
 122       Pair[1] += C-'a'+10;
 123   }
 124   Pair[0] = 0;
 125   for (int i=0; i<16; i++, Buffer++) {
 126     Pair[0] *= 16;
 127     char C = *Buffer;
 128     if (C >= '0' && C <= '9')
 129       Pair[0] += C-'0';
 130     else if (C >= 'A' && C <= 'F')
 131       Pair[0] += C-'A'+10;
 132     else if (C >= 'a' && C <= 'f')
 133       Pair[0] += C-'a'+10;
 134   }
 135   if (Buffer != End)
 136     Error("constant bigger than 128 bits detected!");
 137 }
 138
 139 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
 140 // appropriate character.
 141 static void UnEscapeLexed(std::string &Str) {
 142   if (Str.empty()) return;
 143
 144   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
 145   char *BOut = Buffer;
 146   for (char *BIn = Buffer; BIn != EndBuffer; ) {
 147     if (BIn[0] == '\\') {
 148       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
 149         *BOut++ = '\\'; // Two \ becomes one
 150         BIn += 2;
 151       } else if (BIn < EndBuffer-2 && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
 152         char Tmp = BIn[3]; BIn[3] = 0;      // Terminate string
 153         *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number
 154         BIn[3] = Tmp;                       // Restore character
 155         BIn += 3;                           // Skip over handled chars
 156         ++BOut;
 157       } else {
 158         *BOut++ = *BIn++;
 159       }
 160     } else {
 161       *BOut++ = *BIn++;
 162     }
 163   }
 164   Str.resize(BOut-Buffer);
 165 }
 166
 167 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
 168 static bool isLabelChar(char C) {
 169   return isalnum(C) || C == '-' || C == '$' || C == '.' || C == '_';
 170 }
 171
 172
 173 /// isLabelTail - Return true if this pointer points to a valid end of a label.
 174 static const char *isLabelTail(const char *CurPtr) {
 175   while (1) {
 176     if (CurPtr[0] == ':') return CurPtr+1;
 177     if (!isLabelChar(CurPtr[0])) return 0;
 178     ++CurPtr;
 179   }
 180 }
 181
 182
 183
 184 //===----------------------------------------------------------------------===//
 185 // Lexer definition.
 186 //===----------------------------------------------------------------------===//
 187
 188 LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err,
 189                  LLVMContext &C)
 190   : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
 191   CurPtr = CurBuf->getBufferStart();
 192 }
 193
 194 std::string LLLexer::getFilename() const {
 195   return CurBuf->getBufferIdentifier();
 196 }
 197
 198 int LLLexer::getNextChar() {
 199   char CurChar = *CurPtr++;
 200   switch (CurChar) {
 201   default: return (unsigned char)CurChar;
 202   case 0:
 203     // A nul character in the stream is either the end of the current buffer or
 204     // a random nul in the file.  Disambiguate that here.
 205     if (CurPtr-1 != CurBuf->getBufferEnd())
 206       return 0;  // Just whitespace.
 207
 208     // Otherwise, return end of file.
 209     --CurPtr;  // Another call to lex will return EOF again.
 210     return EOF;
 211   }
 212 }
 213
 214
 215 lltok::Kind LLLexer::LexToken() {
 216   TokStart = CurPtr;
 217
 218   int CurChar = getNextChar();
 219   switch (CurChar) {
 220   default:
 221     // Handle letters: [a-zA-Z_]
 222     if (isalpha(CurChar) || CurChar == '_')
 223       return LexIdentifier();
 224
 225     return lltok::Error;
 226   case EOF: return lltok::Eof;
 227   case 0:
 228   case ' ':
 229   case '\t':
 230   case '\n':
 231   case '\r':
 232     // Ignore whitespace.
 233     return LexToken();
 234   case '+': return LexPositive();
 235   case '@': return LexAt();
 236   case '%': return LexPercent();
 237   case '"': return LexQuote();
 238   case '.':
 239     if (const char *Ptr = isLabelTail(CurPtr)) {
 240       CurPtr = Ptr;
 241       StrVal.assign(TokStart, CurPtr-1);
 242       return lltok::LabelStr;
 243     }
 244     if (CurPtr[0] == '.' && CurPtr[1] == '.') {
 245       CurPtr += 2;
 246       return lltok::dotdotdot;
 247     }
 248     return lltok::Error;
 249   case '$':
 250     if (const char *Ptr = isLabelTail(CurPtr)) {
 251       CurPtr = Ptr;
 252       StrVal.assign(TokStart, CurPtr-1);
 253       return lltok::LabelStr;
 254     }
 255     return lltok::Error;
 256   case ';':
 257     SkipLineComment();
 258     return LexToken();
 259   case '!': return LexExclaim();
 260   case '0': case '1': case '2': case '3': case '4':
 261   case '5': case '6': case '7': case '8': case '9':
 262   case '-':
 263     return LexDigitOrNegative();
 264   case '=': return lltok::equal;
 265   case '[': return lltok::lsquare;
 266   case ']': return lltok::rsquare;
 267   case '{': return lltok::lbrace;
 268   case '}': return lltok::rbrace;
 269   case '<': return lltok::less;
 270   case '>': return lltok::greater;
 271   case '(': return lltok::lparen;
 272   case ')': return lltok::rparen;
 273   case ',': return lltok::comma;
 274   case '*': return lltok::star;
 275   case '\\': return lltok::backslash;
 276   }
 277 }
 278
 279 void LLLexer::SkipLineComment() {
 280   while (1) {
 281     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
 282       return;
 283   }
 284 }
 285
 286 /// LexAt - Lex all tokens that start with an @ character:
 287 ///   GlobalVar   @\"[^\"]*\"
 288 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
 289 ///   GlobalVarID @[0-9]+
 290 lltok::Kind LLLexer::LexAt() {
 291   // Handle AtStringConstant: @\"[^\"]*\"
 292   if (CurPtr[0] == '"') {
 293     ++CurPtr;
 294
 295     while (1) {
 296       int CurChar = getNextChar();
 297
 298       if (CurChar == EOF) {
 299         Error("end of file in global variable name");
 300         return lltok::Error;
 301       }
 302       if (CurChar == '"') {
 303         StrVal.assign(TokStart+2, CurPtr-1);
 304         UnEscapeLexed(StrVal);
 305         return lltok::GlobalVar;
 306       }
 307     }
 308   }
 309
 310   // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
 311   if (ReadVarName())
 312     return lltok::GlobalVar;
 313
 314   // Handle GlobalVarID: @[0-9]+
 315   if (isdigit(CurPtr[0])) {
 316     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 317       /*empty*/;
 318
 319     uint64_t Val = atoull(TokStart+1, CurPtr);
 320     if ((unsigned)Val != Val)
 321       Error("invalid value number (too large)!");
 322     UIntVal = unsigned(Val);
 323     return lltok::GlobalID;
 324   }
 325
 326   return lltok::Error;
 327 }
 328
 329 /// ReadString - Read a string until the closing quote.
 330 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
 331   const char *Start = CurPtr;
 332   while (1) {
 333     int CurChar = getNextChar();
 334
 335     if (CurChar == EOF) {
 336       Error("end of file in string constant");
 337       return lltok::Error;
 338     }
 339     if (CurChar == '"') {
 340       StrVal.assign(Start, CurPtr-1);
 341       UnEscapeLexed(StrVal);
 342       return kind;
 343     }
 344   }
 345 }
 346
 347 /// ReadVarName - Read the rest of a token containing a variable name.
 348 bool LLLexer::ReadVarName() {
 349   const char *NameStart = CurPtr;
 350   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 351       CurPtr[0] == '.' || CurPtr[0] == '_') {
 352     ++CurPtr;
 353     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 354            CurPtr[0] == '.' || CurPtr[0] == '_')
 355       ++CurPtr;
 356
 357     StrVal.assign(NameStart, CurPtr);
 358     return true;
 359   }
 360   return false;
 361 }
 362
 363 /// LexPercent - Lex all tokens that start with a % character:
 364 ///   LocalVar   ::= %\"[^\"]*\"
 365 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
 366 ///   LocalVarID ::= %[0-9]+
 367 lltok::Kind LLLexer::LexPercent() {
 368   // Handle LocalVarName: %\"[^\"]*\"
 369   if (CurPtr[0] == '"') {
 370     ++CurPtr;
 371     return ReadString(lltok::LocalVar);
 372   }
 373
 374   // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
 375   if (ReadVarName())
 376     return lltok::LocalVar;
 377
 378   // Handle LocalVarID: %[0-9]+
 379   if (isdigit(CurPtr[0])) {
 380     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 381       /*empty*/;
 382
 383     uint64_t Val = atoull(TokStart+1, CurPtr);
 384     if ((unsigned)Val != Val)
 385       Error("invalid value number (too large)!");
 386     UIntVal = unsigned(Val);
 387     return lltok::LocalVarID;
 388   }
 389
 390   return lltok::Error;
 391 }
 392
 393 /// LexQuote - Lex all tokens that start with a " character:
 394 ///   QuoteLabel        "[^"]+":
 395 ///   StringConstant    "[^"]*"
 396 lltok::Kind LLLexer::LexQuote() {
 397   lltok::Kind kind = ReadString(lltok::StringConstant);
 398   if (kind == lltok::Error || kind == lltok::Eof)
 399     return kind;
 400
 401   if (CurPtr[0] == ':') {
 402     ++CurPtr;
 403     kind = lltok::LabelStr;
 404   }
 405
 406   return kind;
 407 }
 408
 409 static bool JustWhitespaceNewLine(const char *&Ptr) {
 410   const char *ThisPtr = Ptr;
 411   while (*ThisPtr == ' ' || *ThisPtr == '\t')
 412     ++ThisPtr;
 413   if (*ThisPtr == '\n' || *ThisPtr == '\r') {
 414     Ptr = ThisPtr;
 415     return true;
 416   }
 417   return false;
 418 }
 419
 420 /// LexExclaim:
 421 ///    !foo
 422 ///    !
 423 lltok::Kind LLLexer::LexExclaim() {
 424   // Lex a metadata name as a MetadataVar.
 425   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 426       CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
 427     ++CurPtr;
 428     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 429            CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
 430       ++CurPtr;
 431
 432     StrVal.assign(TokStart+1, CurPtr);   // Skip !
 433     UnEscapeLexed(StrVal);
 434     return lltok::MetadataVar;
 435   }
 436   return lltok::exclaim;
 437 }
 438
 439 /// LexIdentifier: Handle several related productions:
 440 ///    Label           [-a-zA-Z$._0-9]+:
 441 ///    IntegerType     i[0-9]+
 442 ///    Keyword         sdiv, float, ...
 443 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
 444 lltok::Kind LLLexer::LexIdentifier() {
 445   const char *StartChar = CurPtr;
 446   const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
 447   const char *KeywordEnd = 0;
 448
 449   for (; isLabelChar(*CurPtr); ++CurPtr) {
 450     // If we decide this is an integer, remember the end of the sequence.
 451     if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr;
 452     if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr;
 453   }
 454
 455   // If we stopped due to a colon, this really is a label.
 456   if (*CurPtr == ':') {
 457     StrVal.assign(StartChar-1, CurPtr++);
 458     return lltok::LabelStr;
 459   }
 460
 461   // Otherwise, this wasn't a label.  If this was valid as an integer type,
 462   // return it.
 463   if (IntEnd == 0) IntEnd = CurPtr;
 464   if (IntEnd != StartChar) {
 465     CurPtr = IntEnd;
 466     uint64_t NumBits = atoull(StartChar, CurPtr);
 467     if (NumBits < IntegerType::MIN_INT_BITS ||
 468         NumBits > IntegerType::MAX_INT_BITS) {
 469       Error("bitwidth for integer type out of range!");
 470       return lltok::Error;
 471     }
 472     TyVal = IntegerType::get(Context, NumBits);
 473     return lltok::Type;
 474   }
 475
 476   // Otherwise, this was a letter sequence.  See which keyword this is.
 477   if (KeywordEnd == 0) KeywordEnd = CurPtr;
 478   CurPtr = KeywordEnd;
 479   --StartChar;
 480   unsigned Len = CurPtr-StartChar;
 481 #define KEYWORD(STR) \
 482   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
 483     return lltok::kw_##STR;
 484
 485   KEYWORD(begin);   KEYWORD(end);
 486   KEYWORD(true);    KEYWORD(false);
 487   KEYWORD(declare); KEYWORD(define);
 488   KEYWORD(global);  KEYWORD(constant);
 489
 490   KEYWORD(private);
 491   KEYWORD(linker_private);
 492   KEYWORD(linker_private_weak);
 493   KEYWORD(linker_private_weak_def_auto);
 494   KEYWORD(internal);
 495   KEYWORD(available_externally);
 496   KEYWORD(linkonce);
 497   KEYWORD(linkonce_odr);
 498   KEYWORD(weak);
 499   KEYWORD(weak_odr);
 500   KEYWORD(appending);
 501   KEYWORD(dllimport);
 502   KEYWORD(dllexport);
 503   KEYWORD(common);
 504   KEYWORD(default);
 505   KEYWORD(hidden);
 506   KEYWORD(protected);
 507   KEYWORD(unnamed_addr);
 508   KEYWORD(extern_weak);
 509   KEYWORD(external);
 510   KEYWORD(thread_local);
 511   KEYWORD(zeroinitializer);
 512   KEYWORD(undef);
 513   KEYWORD(null);
 514   KEYWORD(to);
 515   KEYWORD(tail);
 516   KEYWORD(target);
 517   KEYWORD(triple);
 518   KEYWORD(deplibs);
 519   KEYWORD(datalayout);
 520   KEYWORD(volatile);
 521   KEYWORD(nuw);
 522   KEYWORD(nsw);
 523   KEYWORD(exact);
 524   KEYWORD(inbounds);
 525   KEYWORD(align);
 526   KEYWORD(addrspace);
 527   KEYWORD(section);
 528   KEYWORD(alias);
 529   KEYWORD(module);
 530   KEYWORD(asm);
 531   KEYWORD(sideeffect);
 532   KEYWORD(alignstack);
 533   KEYWORD(gc);
 534
 535   KEYWORD(ccc);
 536   KEYWORD(fastcc);
 537   KEYWORD(coldcc);
 538   KEYWORD(x86_stdcallcc);
 539   KEYWORD(x86_fastcallcc);
 540   KEYWORD(x86_thiscallcc);
 541   KEYWORD(arm_apcscc);
 542   KEYWORD(arm_aapcscc);
 543   KEYWORD(arm_aapcs_vfpcc);
 544   KEYWORD(msp430_intrcc);
 545   KEYWORD(ptx_kernel);
 546   KEYWORD(ptx_device);
 547
 548   KEYWORD(cc);
 549   KEYWORD(c);
 550
 551   KEYWORD(signext);
 552   KEYWORD(zeroext);
 553   KEYWORD(inreg);
 554   KEYWORD(sret);
 555   KEYWORD(nounwind);
 556   KEYWORD(noreturn);
 557   KEYWORD(noalias);
 558   KEYWORD(nocapture);
 559   KEYWORD(byval);
 560   KEYWORD(nest);
 561   KEYWORD(readnone);
 562   KEYWORD(readonly);
 563   KEYWORD(uwtable);
 564
 565   KEYWORD(inlinehint);
 566   KEYWORD(noinline);
 567   KEYWORD(alwaysinline);
 568   KEYWORD(optsize);
 569   KEYWORD(ssp);
 570   KEYWORD(sspreq);
 571   KEYWORD(noredzone);
 572   KEYWORD(noimplicitfloat);
 573   KEYWORD(naked);
 574   KEYWORD(hotpatch);
 575
 576   KEYWORD(type);
 577   KEYWORD(opaque);
 578
 579   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
 580   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
 581   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
 582   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
 583
 584   KEYWORD(x);
 585   KEYWORD(blockaddress);
 586 #undef KEYWORD
 587
 588   // Keywords for types.
 589 #define TYPEKEYWORD(STR, LLVMTY) \
 590   if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
 591     TyVal = LLVMTY; return lltok::Type; }
 592   TYPEKEYWORD("void",      Type::getVoidTy(Context));
 593   TYPEKEYWORD("float",     Type::getFloatTy(Context));
 594   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
 595   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
 596   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
 597   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
 598   TYPEKEYWORD("label",     Type::getLabelTy(Context));
 599   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
 600   TYPEKEYWORD("x86_mmx",   Type::getX86_MMXTy(Context));
 601 #undef TYPEKEYWORD
 602
 603   // Handle special forms for autoupgrading.  Drop these in LLVM 3.0.  This is
 604   // to avoid conflicting with the sext/zext instructions, below.
 605   if (Len == 4 && !memcmp(StartChar, "sext", 4)) {
 606     // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
 607     if (JustWhitespaceNewLine(CurPtr))
 608       return lltok::kw_signext;
 609   } else if (Len == 4 && !memcmp(StartChar, "zext", 4)) {
 610     // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
 611     if (JustWhitespaceNewLine(CurPtr))
 612       return lltok::kw_zeroext;
 613   } else if (Len == 6 && !memcmp(StartChar, "malloc", 6)) {
 614     // FIXME: Remove in LLVM 3.0.
 615     // Autoupgrade malloc instruction.
 616     return lltok::kw_malloc;
 617   } else if (Len == 4 && !memcmp(StartChar, "free", 4)) {
 618     // FIXME: Remove in LLVM 3.0.
 619     // Autoupgrade malloc instruction.
 620     return lltok::kw_free;
 621   }
 622
 623   // Keywords for instructions.
 624 #define INSTKEYWORD(STR, Enum) \
 625   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
 626     UIntVal = Instruction::Enum; return lltok::kw_##STR; }
 627
 628   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
 629   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
 630   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
 631   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
 632   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
 633   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
 634   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
 635   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
 636
 637   INSTKEYWORD(phi,         PHI);
 638   INSTKEYWORD(call,        Call);
 639   INSTKEYWORD(trunc,       Trunc);
 640   INSTKEYWORD(zext,        ZExt);
 641   INSTKEYWORD(sext,        SExt);
 642   INSTKEYWORD(fptrunc,     FPTrunc);
 643   INSTKEYWORD(fpext,       FPExt);
 644   INSTKEYWORD(uitofp,      UIToFP);
 645   INSTKEYWORD(sitofp,      SIToFP);
 646   INSTKEYWORD(fptoui,      FPToUI);
 647   INSTKEYWORD(fptosi,      FPToSI);
 648   INSTKEYWORD(inttoptr,    IntToPtr);
 649   INSTKEYWORD(ptrtoint,    PtrToInt);
 650   INSTKEYWORD(bitcast,     BitCast);
 651   INSTKEYWORD(select,      Select);
 652   INSTKEYWORD(va_arg,      VAArg);
 653   INSTKEYWORD(ret,         Ret);
 654   INSTKEYWORD(br,          Br);
 655   INSTKEYWORD(switch,      Switch);
 656   INSTKEYWORD(indirectbr,  IndirectBr);
 657   INSTKEYWORD(invoke,      Invoke);
 658   INSTKEYWORD(unwind,      Unwind);
 659   INSTKEYWORD(unreachable, Unreachable);
 660
 661   INSTKEYWORD(alloca,      Alloca);
 662   INSTKEYWORD(load,        Load);
 663   INSTKEYWORD(store,       Store);
 664   INSTKEYWORD(getelementptr, GetElementPtr);
 665
 666   INSTKEYWORD(extractelement, ExtractElement);
 667   INSTKEYWORD(insertelement,  InsertElement);
 668   INSTKEYWORD(shufflevector,  ShuffleVector);
 669   INSTKEYWORD(getresult,      ExtractValue);
 670   INSTKEYWORD(extractvalue,   ExtractValue);
 671   INSTKEYWORD(insertvalue,    InsertValue);
 672 #undef INSTKEYWORD
 673
 674   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
 675   // the CFE to avoid forcing it to deal with 64-bit numbers.
 676   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
 677       TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) {
 678     int len = CurPtr-TokStart-3;
 679     uint32_t bits = len * 4;
 680     APInt Tmp(bits, StringRef(TokStart+3, len), 16);
 681     uint32_t activeBits = Tmp.getActiveBits();
 682     if (activeBits > 0 && activeBits < bits)
 683       Tmp = Tmp.trunc(activeBits);
 684     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
 685     return lltok::APSInt;
 686   }
 687
 688   // If this is "cc1234", return this as just "cc".
 689   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
 690     CurPtr = TokStart+2;
 691     return lltok::kw_cc;
 692   }
 693
 694   // If this starts with "call", return it as CALL.  This is to support old
 695   // broken .ll files.  FIXME: remove this with LLVM 3.0.
 696   if (CurPtr-TokStart > 4 && !memcmp(TokStart, "call", 4)) {
 697     CurPtr = TokStart+4;
 698     UIntVal = Instruction::Call;
 699     return lltok::kw_call;
 700   }
 701
 702   // Finally, if this isn't known, return an error.
 703   CurPtr = TokStart+1;
 704   return lltok::Error;
 705 }
 706
 707
 708 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
 709 /// that this is not a label:
 710 ///    HexFPConstant     0x[0-9A-Fa-f]+
 711 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 712 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 713 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 714 lltok::Kind LLLexer::Lex0x() {
 715   CurPtr = TokStart + 2;
 716
 717   char Kind;
 718   if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') {
 719     Kind = *CurPtr++;
 720   } else {
 721     Kind = 'J';
 722   }
 723
 724   if (!isxdigit(CurPtr[0])) {
 725     // Bad token, return it as an error.
 726     CurPtr = TokStart+1;
 727     return lltok::Error;
 728   }
 729
 730   while (isxdigit(CurPtr[0]))
 731     ++CurPtr;
 732
 733   if (Kind == 'J') {
 734     // HexFPConstant - Floating point constant represented in IEEE format as a
 735     // hexadecimal number for when exponential notation is not precise enough.
 736     // Float and double only.
 737     APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
 738     return lltok::APFloat;
 739   }
 740
 741   uint64_t Pair[2];
 742   switch (Kind) {
 743   default: llvm_unreachable("Unknown kind!");
 744   case 'K':
 745     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
 746     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
 747     APFloatVal = APFloat(APInt(80, 2, Pair));
 748     return lltok::APFloat;
 749   case 'L':
 750     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
 751     HexToIntPair(TokStart+3, CurPtr, Pair);
 752     APFloatVal = APFloat(APInt(128, 2, Pair), true);
 753     return lltok::APFloat;
 754   case 'M':
 755     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
 756     HexToIntPair(TokStart+3, CurPtr, Pair);
 757     APFloatVal = APFloat(APInt(128, 2, Pair));
 758     return lltok::APFloat;
 759   }
 760 }
 761
 762 /// LexIdentifier: Handle several related productions:
 763 ///    Label             [-a-zA-Z$._0-9]+:
 764 ///    NInteger          -[0-9]+
 765 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 766 ///    PInteger          [0-9]+
 767 ///    HexFPConstant     0x[0-9A-Fa-f]+
 768 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 769 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 770 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 771 lltok::Kind LLLexer::LexDigitOrNegative() {
 772   // If the letter after the negative is a number, this is probably a label.
 773   if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) {
 774     // Okay, this is not a number after the -, it's probably a label.
 775     if (const char *End = isLabelTail(CurPtr)) {
 776       StrVal.assign(TokStart, End-1);
 777       CurPtr = End;
 778       return lltok::LabelStr;
 779     }
 780
 781     return lltok::Error;
 782   }
 783
 784   // At this point, it is either a label, int or fp constant.
 785
 786   // Skip digits, we have at least one.
 787   for (; isdigit(CurPtr[0]); ++CurPtr)
 788     /*empty*/;
 789
 790   // Check to see if this really is a label afterall, e.g. "-1:".
 791   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
 792     if (const char *End = isLabelTail(CurPtr)) {
 793       StrVal.assign(TokStart, End-1);
 794       CurPtr = End;
 795       return lltok::LabelStr;
 796     }
 797   }
 798
 799   // If the next character is a '.', then it is a fp value, otherwise its
 800   // integer.
 801   if (CurPtr[0] != '.') {
 802     if (TokStart[0] == '0' && TokStart[1] == 'x')
 803       return Lex0x();
 804     unsigned Len = CurPtr-TokStart;
 805     uint32_t numBits = ((Len * 64) / 19) + 2;
 806     APInt Tmp(numBits, StringRef(TokStart, Len), 10);
 807     if (TokStart[0] == '-') {
 808       uint32_t minBits = Tmp.getMinSignedBits();
 809       if (minBits > 0 && minBits < numBits)
 810         Tmp = Tmp.trunc(minBits);
 811       APSIntVal = APSInt(Tmp, false);
 812     } else {
 813       uint32_t activeBits = Tmp.getActiveBits();
 814       if (activeBits > 0 && activeBits < numBits)
 815         Tmp = Tmp.trunc(activeBits);
 816       APSIntVal = APSInt(Tmp, true);
 817     }
 818     return lltok::APSInt;
 819   }
 820
 821   ++CurPtr;
 822
 823   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 824   while (isdigit(CurPtr[0])) ++CurPtr;
 825
 826   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 827     if (isdigit(CurPtr[1]) ||
 828         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 829       CurPtr += 2;
 830       while (isdigit(CurPtr[0])) ++CurPtr;
 831     }
 832   }
 833
 834   APFloatVal = APFloat(std::atof(TokStart));
 835   return lltok::APFloat;
 836 }
 837
 838 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 839 lltok::Kind LLLexer::LexPositive() {
 840   // If the letter after the negative is a number, this is probably not a
 841   // label.
 842   if (!isdigit(CurPtr[0]))
 843     return lltok::Error;
 844
 845   // Skip digits.
 846   for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 847     /*empty*/;
 848
 849   // At this point, we need a '.'.
 850   if (CurPtr[0] != '.') {
 851     CurPtr = TokStart+1;
 852     return lltok::Error;
 853   }
 854
 855   ++CurPtr;
 856
 857   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 858   while (isdigit(CurPtr[0])) ++CurPtr;
 859
 860   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 861     if (isdigit(CurPtr[1]) ||
 862         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 863       CurPtr += 2;
 864       while (isdigit(CurPtr[0])) ++CurPtr;
 865     }
 866   }
 867
 868   APFloatVal = APFloat(std::atof(TokStart));
 869   return lltok::APFloat;
 870 }