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