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 <cstdio>
  26 #include <cstdlib>
  27 #include <cstring>
  28 using namespace llvm;
  29
  30 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
  31   ErrorInfo = SM.GetMessage(ErrorLoc, Msg, "error");
  32   return true;
  33 }
  34
  35 //===----------------------------------------------------------------------===//
  36 // Helper functions.
  37 //===----------------------------------------------------------------------===//
  38
  39 // atoull - Convert an ascii string of decimal digits into the unsigned long
  40 // long representation... this does not have to do input error checking,
  41 // because we know that the input will be matched by a suitable regex...
  42 //
  43 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
  44   uint64_t Result = 0;
  45   for (; Buffer != End; Buffer++) {
  46     uint64_t OldRes = Result;
  47     Result *= 10;
  48     Result += *Buffer-'0';
  49     if (Result < OldRes) {  // Uh, oh, overflow detected!!!
  50       Error("constant bigger than 64 bits detected!");
  51       return 0;
  52     }
  53   }
  54   return Result;
  55 }
  56
  57 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
  58   uint64_t Result = 0;
  59   for (; Buffer != End; ++Buffer) {
  60     uint64_t OldRes = Result;
  61     Result *= 16;
  62     char C = *Buffer;
  63     if (C >= '0' && C <= '9')
  64       Result += C-'0';
  65     else if (C >= 'A' && C <= 'F')
  66       Result += C-'A'+10;
  67     else if (C >= 'a' && C <= 'f')
  68       Result += C-'a'+10;
  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     char C = *Buffer;
  85     if (C >= '0' && C <= '9')
  86       Pair[0] += C-'0';
  87     else if (C >= 'A' && C <= 'F')
  88       Pair[0] += C-'A'+10;
  89     else if (C >= 'a' && C <= 'f')
  90       Pair[0] += C-'a'+10;
  91   }
  92   Pair[1] = 0;
  93   for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
  94     Pair[1] *= 16;
  95     char C = *Buffer;
  96     if (C >= '0' && C <= '9')
  97       Pair[1] += C-'0';
  98     else if (C >= 'A' && C <= 'F')
  99       Pair[1] += C-'A'+10;
 100     else if (C >= 'a' && C <= 'f')
 101       Pair[1] += C-'a'+10;
 102   }
 103   if (Buffer != End)
 104     Error("constant bigger than 128 bits detected!");
 105 }
 106
 107 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
 108 /// { low64, high16 } as usual for an APInt.
 109 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
 110                            uint64_t Pair[2]) {
 111   Pair[1] = 0;
 112   for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
 113     assert(Buffer != End);
 114     Pair[1] *= 16;
 115     char C = *Buffer;
 116     if (C >= '0' && C <= '9')
 117       Pair[1] += C-'0';
 118     else if (C >= 'A' && C <= 'F')
 119       Pair[1] += C-'A'+10;
 120     else if (C >= 'a' && C <= 'f')
 121       Pair[1] += C-'a'+10;
 122   }
 123   Pair[0] = 0;
 124   for (int i=0; i<16; i++, Buffer++) {
 125     Pair[0] *= 16;
 126     char C = *Buffer;
 127     if (C >= '0' && C <= '9')
 128       Pair[0] += C-'0';
 129     else if (C >= 'A' && C <= 'F')
 130       Pair[0] += C-'A'+10;
 131     else if (C >= 'a' && C <= 'f')
 132       Pair[0] += C-'a'+10;
 133   }
 134   if (Buffer != End)
 135     Error("constant bigger than 128 bits detected!");
 136 }
 137
 138 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
 139 // appropriate character.
 140 static void UnEscapeLexed(std::string &Str) {
 141   if (Str.empty()) return;
 142
 143   char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
 144   char *BOut = Buffer;
 145   for (char *BIn = Buffer; BIn != EndBuffer; ) {
 146     if (BIn[0] == '\\') {
 147       if (BIn < EndBuffer-1 && BIn[1] == '\\') {
 148         *BOut++ = '\\'; // Two \ becomes one
 149         BIn += 2;
 150       } else if (BIn < EndBuffer-2 && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
 151         char Tmp = BIn[3]; BIn[3] = 0;      // Terminate string
 152         *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number
 153         BIn[3] = Tmp;                       // Restore character
 154         BIn += 3;                           // Skip over handled chars
 155         ++BOut;
 156       } else {
 157         *BOut++ = *BIn++;
 158       }
 159     } else {
 160       *BOut++ = *BIn++;
 161     }
 162   }
 163   Str.resize(BOut-Buffer);
 164 }
 165
 166 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
 167 static bool isLabelChar(char C) {
 168   return isalnum(C) || C == '-' || C == '$' || C == '.' || C == '_';
 169 }
 170
 171
 172 /// isLabelTail - Return true if this pointer points to a valid end of a label.
 173 static const char *isLabelTail(const char *CurPtr) {
 174   while (1) {
 175     if (CurPtr[0] == ':') return CurPtr+1;
 176     if (!isLabelChar(CurPtr[0])) return 0;
 177     ++CurPtr;
 178   }
 179 }
 180
 181
 182
 183 //===----------------------------------------------------------------------===//
 184 // Lexer definition.
 185 //===----------------------------------------------------------------------===//
 186
 187 LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err,
 188                  LLVMContext &C)
 189   : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
 190   CurPtr = CurBuf->getBufferStart();
 191 }
 192
 193 std::string LLLexer::getFilename() const {
 194   return CurBuf->getBufferIdentifier();
 195 }
 196
 197 int LLLexer::getNextChar() {
 198   char CurChar = *CurPtr++;
 199   switch (CurChar) {
 200   default: return (unsigned char)CurChar;
 201   case 0:
 202     // A nul character in the stream is either the end of the current buffer or
 203     // a random nul in the file.  Disambiguate that here.
 204     if (CurPtr-1 != CurBuf->getBufferEnd())
 205       return 0;  // Just whitespace.
 206
 207     // Otherwise, return end of file.
 208     --CurPtr;  // Another call to lex will return EOF again.
 209     return EOF;
 210   }
 211 }
 212
 213
 214 lltok::Kind LLLexer::LexToken() {
 215   TokStart = CurPtr;
 216
 217   int CurChar = getNextChar();
 218   switch (CurChar) {
 219   default:
 220     // Handle letters: [a-zA-Z_]
 221     if (isalpha(CurChar) || CurChar == '_')
 222       return LexIdentifier();
 223
 224     return lltok::Error;
 225   case EOF: return lltok::Eof;
 226   case 0:
 227   case ' ':
 228   case '\t':
 229   case '\n':
 230   case '\r':
 231     // Ignore whitespace.
 232     return LexToken();
 233   case '+': return LexPositive();
 234   case '@': return LexAt();
 235   case '%': return LexPercent();
 236   case '"': return LexQuote();
 237   case '.':
 238     if (const char *Ptr = isLabelTail(CurPtr)) {
 239       CurPtr = Ptr;
 240       StrVal.assign(TokStart, CurPtr-1);
 241       return lltok::LabelStr;
 242     }
 243     if (CurPtr[0] == '.' && CurPtr[1] == '.') {
 244       CurPtr += 2;
 245       return lltok::dotdotdot;
 246     }
 247     return lltok::Error;
 248   case '$':
 249     if (const char *Ptr = isLabelTail(CurPtr)) {
 250       CurPtr = Ptr;
 251       StrVal.assign(TokStart, CurPtr-1);
 252       return lltok::LabelStr;
 253     }
 254     return lltok::Error;
 255   case ';':
 256     SkipLineComment();
 257     return LexToken();
 258   case '!': return LexExclaim();
 259   case '0': case '1': case '2': case '3': case '4':
 260   case '5': case '6': case '7': case '8': case '9':
 261   case '-':
 262     return LexDigitOrNegative();
 263   case '=': return lltok::equal;
 264   case '[': return lltok::lsquare;
 265   case ']': return lltok::rsquare;
 266   case '{': return lltok::lbrace;
 267   case '}': return lltok::rbrace;
 268   case '<': return lltok::less;
 269   case '>': return lltok::greater;
 270   case '(': return lltok::lparen;
 271   case ')': return lltok::rparen;
 272   case ',': return lltok::comma;
 273   case '*': return lltok::star;
 274   case '\\': return lltok::backslash;
 275   }
 276 }
 277
 278 void LLLexer::SkipLineComment() {
 279   while (1) {
 280     if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
 281       return;
 282   }
 283 }
 284
 285 /// LexAt - Lex all tokens that start with an @ character:
 286 ///   GlobalVar   @\"[^\"]*\"
 287 ///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
 288 ///   GlobalVarID @[0-9]+
 289 lltok::Kind LLLexer::LexAt() {
 290   // Handle AtStringConstant: @\"[^\"]*\"
 291   if (CurPtr[0] == '"') {
 292     ++CurPtr;
 293
 294     while (1) {
 295       int CurChar = getNextChar();
 296
 297       if (CurChar == EOF) {
 298         Error("end of file in global variable name");
 299         return lltok::Error;
 300       }
 301       if (CurChar == '"') {
 302         StrVal.assign(TokStart+2, CurPtr-1);
 303         UnEscapeLexed(StrVal);
 304         return lltok::GlobalVar;
 305       }
 306     }
 307   }
 308
 309   // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
 310   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 311       CurPtr[0] == '.' || CurPtr[0] == '_') {
 312     ++CurPtr;
 313     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 314            CurPtr[0] == '.' || CurPtr[0] == '_')
 315       ++CurPtr;
 316
 317     StrVal.assign(TokStart+1, CurPtr);   // Skip @
 318     return lltok::GlobalVar;
 319   }
 320
 321   // Handle GlobalVarID: @[0-9]+
 322   if (isdigit(CurPtr[0])) {
 323     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 324       /*empty*/;
 325
 326     uint64_t Val = atoull(TokStart+1, CurPtr);
 327     if ((unsigned)Val != Val)
 328       Error("invalid value number (too large)!");
 329     UIntVal = unsigned(Val);
 330     return lltok::GlobalID;
 331   }
 332
 333   return lltok::Error;
 334 }
 335
 336
 337 /// LexPercent - Lex all tokens that start with a % character:
 338 ///   LocalVar   ::= %\"[^\"]*\"
 339 ///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
 340 ///   LocalVarID ::= %[0-9]+
 341 lltok::Kind LLLexer::LexPercent() {
 342   // Handle LocalVarName: %\"[^\"]*\"
 343   if (CurPtr[0] == '"') {
 344     ++CurPtr;
 345
 346     while (1) {
 347       int CurChar = getNextChar();
 348
 349       if (CurChar == EOF) {
 350         Error("end of file in string constant");
 351         return lltok::Error;
 352       }
 353       if (CurChar == '"') {
 354         StrVal.assign(TokStart+2, CurPtr-1);
 355         UnEscapeLexed(StrVal);
 356         return lltok::LocalVar;
 357       }
 358     }
 359   }
 360
 361   // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
 362   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 363       CurPtr[0] == '.' || CurPtr[0] == '_') {
 364     ++CurPtr;
 365     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 366            CurPtr[0] == '.' || CurPtr[0] == '_')
 367       ++CurPtr;
 368
 369     StrVal.assign(TokStart+1, CurPtr);   // Skip %
 370     return lltok::LocalVar;
 371   }
 372
 373   // Handle LocalVarID: %[0-9]+
 374   if (isdigit(CurPtr[0])) {
 375     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 376       /*empty*/;
 377
 378     uint64_t Val = atoull(TokStart+1, CurPtr);
 379     if ((unsigned)Val != Val)
 380       Error("invalid value number (too large)!");
 381     UIntVal = unsigned(Val);
 382     return lltok::LocalVarID;
 383   }
 384
 385   return lltok::Error;
 386 }
 387
 388 /// LexQuote - Lex all tokens that start with a " character:
 389 ///   QuoteLabel        "[^"]+":
 390 ///   StringConstant    "[^"]*"
 391 lltok::Kind LLLexer::LexQuote() {
 392   while (1) {
 393     int CurChar = getNextChar();
 394
 395     if (CurChar == EOF) {
 396       Error("end of file in quoted string");
 397       return lltok::Error;
 398     }
 399
 400     if (CurChar != '"') continue;
 401
 402     if (CurPtr[0] != ':') {
 403       StrVal.assign(TokStart+1, CurPtr-1);
 404       UnEscapeLexed(StrVal);
 405       return lltok::StringConstant;
 406     }
 407
 408     ++CurPtr;
 409     StrVal.assign(TokStart+1, CurPtr-2);
 410     UnEscapeLexed(StrVal);
 411     return lltok::LabelStr;
 412   }
 413 }
 414
 415 static bool JustWhitespaceNewLine(const char *&Ptr) {
 416   const char *ThisPtr = Ptr;
 417   while (*ThisPtr == ' ' || *ThisPtr == '\t')
 418     ++ThisPtr;
 419   if (*ThisPtr == '\n' || *ThisPtr == '\r') {
 420     Ptr = ThisPtr;
 421     return true;
 422   }
 423   return false;
 424 }
 425
 426 /// LexExclaim:
 427 ///    !foo
 428 ///    !
 429 lltok::Kind LLLexer::LexExclaim() {
 430   // Lex a metadata name as a MetadataVar.
 431   if (isalpha(CurPtr[0])) {
 432     ++CurPtr;
 433     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 434            CurPtr[0] == '.' || CurPtr[0] == '_')
 435       ++CurPtr;
 436
 437     StrVal.assign(TokStart+1, CurPtr);   // Skip !
 438     return lltok::MetadataVar;
 439   }
 440   return lltok::exclaim;
 441 }
 442
 443 /// LexIdentifier: Handle several related productions:
 444 ///    Label           [-a-zA-Z$._0-9]+:
 445 ///    IntegerType     i[0-9]+
 446 ///    Keyword         sdiv, float, ...
 447 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
 448 lltok::Kind LLLexer::LexIdentifier() {
 449   const char *StartChar = CurPtr;
 450   const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
 451   const char *KeywordEnd = 0;
 452
 453   for (; isLabelChar(*CurPtr); ++CurPtr) {
 454     // If we decide this is an integer, remember the end of the sequence.
 455     if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr;
 456     if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr;
 457   }
 458
 459   // If we stopped due to a colon, this really is a label.
 460   if (*CurPtr == ':') {
 461     StrVal.assign(StartChar-1, CurPtr++);
 462     return lltok::LabelStr;
 463   }
 464
 465   // Otherwise, this wasn't a label.  If this was valid as an integer type,
 466   // return it.
 467   if (IntEnd == 0) IntEnd = CurPtr;
 468   if (IntEnd != StartChar) {
 469     CurPtr = IntEnd;
 470     uint64_t NumBits = atoull(StartChar, CurPtr);
 471     if (NumBits < IntegerType::MIN_INT_BITS ||
 472         NumBits > IntegerType::MAX_INT_BITS) {
 473       Error("bitwidth for integer type out of range!");
 474       return lltok::Error;
 475     }
 476     TyVal = IntegerType::get(Context, NumBits);
 477     return lltok::Type;
 478   }
 479
 480   // Otherwise, this was a letter sequence.  See which keyword this is.
 481   if (KeywordEnd == 0) KeywordEnd = CurPtr;
 482   CurPtr = KeywordEnd;
 483   --StartChar;
 484   unsigned Len = CurPtr-StartChar;
 485 #define KEYWORD(STR) \
 486   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
 487     return lltok::kw_##STR;
 488
 489   KEYWORD(begin);   KEYWORD(end);
 490   KEYWORD(true);    KEYWORD(false);
 491   KEYWORD(declare); KEYWORD(define);
 492   KEYWORD(global);  KEYWORD(constant);
 493
 494   KEYWORD(private);
 495   KEYWORD(linker_private);
 496   KEYWORD(linker_private_weak);
 497   KEYWORD(linker_private_weak_def_auto);
 498   KEYWORD(internal);
 499   KEYWORD(available_externally);
 500   KEYWORD(linkonce);
 501   KEYWORD(linkonce_odr);
 502   KEYWORD(weak);
 503   KEYWORD(weak_odr);
 504   KEYWORD(appending);
 505   KEYWORD(dllimport);
 506   KEYWORD(dllexport);
 507   KEYWORD(common);
 508   KEYWORD(default);
 509   KEYWORD(hidden);
 510   KEYWORD(protected);
 511   KEYWORD(extern_weak);
 512   KEYWORD(external);
 513   KEYWORD(thread_local);
 514   KEYWORD(zeroinitializer);
 515   KEYWORD(undef);
 516   KEYWORD(null);
 517   KEYWORD(to);
 518   KEYWORD(tail);
 519   KEYWORD(target);
 520   KEYWORD(triple);
 521   KEYWORD(deplibs);
 522   KEYWORD(datalayout);
 523   KEYWORD(volatile);
 524   KEYWORD(nuw);
 525   KEYWORD(nsw);
 526   KEYWORD(exact);
 527   KEYWORD(inbounds);
 528   KEYWORD(align);
 529   KEYWORD(addrspace);
 530   KEYWORD(section);
 531   KEYWORD(alias);
 532   KEYWORD(module);
 533   KEYWORD(asm);
 534   KEYWORD(sideeffect);
 535   KEYWORD(alignstack);
 536   KEYWORD(gc);
 537
 538   KEYWORD(ccc);
 539   KEYWORD(fastcc);
 540   KEYWORD(coldcc);
 541   KEYWORD(x86_stdcallcc);
 542   KEYWORD(x86_fastcallcc);
 543   KEYWORD(x86_thiscallcc);
 544   KEYWORD(arm_apcscc);
 545   KEYWORD(arm_aapcscc);
 546   KEYWORD(arm_aapcs_vfpcc);
 547   KEYWORD(msp430_intrcc);
 548   KEYWORD(ptx_kernel);
 549   KEYWORD(ptx_device);
 550
 551   KEYWORD(cc);
 552   KEYWORD(c);
 553
 554   KEYWORD(signext);
 555   KEYWORD(zeroext);
 556   KEYWORD(inreg);
 557   KEYWORD(sret);
 558   KEYWORD(nounwind);
 559   KEYWORD(noreturn);
 560   KEYWORD(noalias);
 561   KEYWORD(nocapture);
 562   KEYWORD(byval);
 563   KEYWORD(nest);
 564   KEYWORD(readnone);
 565   KEYWORD(readonly);
 566
 567   KEYWORD(inlinehint);
 568   KEYWORD(noinline);
 569   KEYWORD(alwaysinline);
 570   KEYWORD(optsize);
 571   KEYWORD(ssp);
 572   KEYWORD(sspreq);
 573   KEYWORD(noredzone);
 574   KEYWORD(noimplicitfloat);
 575   KEYWORD(naked);
 576   KEYWORD(hotpatch);
 577
 578   KEYWORD(type);
 579   KEYWORD(opaque);
 580
 581   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
 582   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
 583   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
 584   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
 585
 586   KEYWORD(x);
 587   KEYWORD(blockaddress);
 588 #undef KEYWORD
 589
 590   // Keywords for types.
 591 #define TYPEKEYWORD(STR, LLVMTY) \
 592   if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
 593     TyVal = LLVMTY; return lltok::Type; }
 594   TYPEKEYWORD("void",      Type::getVoidTy(Context));
 595   TYPEKEYWORD("float",     Type::getFloatTy(Context));
 596   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
 597   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
 598   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
 599   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
 600   TYPEKEYWORD("label",     Type::getLabelTy(Context));
 601   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
 602   TYPEKEYWORD("x86_mmx",   Type::getX86_MMXTy(Context));
 603 #undef TYPEKEYWORD
 604
 605   // Handle special forms for autoupgrading.  Drop these in LLVM 3.0.  This is
 606   // to avoid conflicting with the sext/zext instructions, below.
 607   if (Len == 4 && !memcmp(StartChar, "sext", 4)) {
 608     // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
 609     if (JustWhitespaceNewLine(CurPtr))
 610       return lltok::kw_signext;
 611   } else if (Len == 4 && !memcmp(StartChar, "zext", 4)) {
 612     // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
 613     if (JustWhitespaceNewLine(CurPtr))
 614       return lltok::kw_zeroext;
 615   } else if (Len == 6 && !memcmp(StartChar, "malloc", 6)) {
 616     // FIXME: Remove in LLVM 3.0.
 617     // Autoupgrade malloc instruction.
 618     return lltok::kw_malloc;
 619   } else if (Len == 4 && !memcmp(StartChar, "free", 4)) {
 620     // FIXME: Remove in LLVM 3.0.
 621     // Autoupgrade malloc instruction.
 622     return lltok::kw_free;
 623   }
 624
 625   // Keywords for instructions.
 626 #define INSTKEYWORD(STR, Enum) \
 627   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
 628     UIntVal = Instruction::Enum; return lltok::kw_##STR; }
 629
 630   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
 631   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
 632   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
 633   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
 634   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
 635   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
 636   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
 637   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
 638
 639   INSTKEYWORD(phi,         PHI);
 640   INSTKEYWORD(call,        Call);
 641   INSTKEYWORD(trunc,       Trunc);
 642   INSTKEYWORD(zext,        ZExt);
 643   INSTKEYWORD(sext,        SExt);
 644   INSTKEYWORD(fptrunc,     FPTrunc);
 645   INSTKEYWORD(fpext,       FPExt);
 646   INSTKEYWORD(uitofp,      UIToFP);
 647   INSTKEYWORD(sitofp,      SIToFP);
 648   INSTKEYWORD(fptoui,      FPToUI);
 649   INSTKEYWORD(fptosi,      FPToSI);
 650   INSTKEYWORD(inttoptr,    IntToPtr);
 651   INSTKEYWORD(ptrtoint,    PtrToInt);
 652   INSTKEYWORD(bitcast,     BitCast);
 653   INSTKEYWORD(select,      Select);
 654   INSTKEYWORD(va_arg,      VAArg);
 655   INSTKEYWORD(ret,         Ret);
 656   INSTKEYWORD(br,          Br);
 657   INSTKEYWORD(switch,      Switch);
 658   INSTKEYWORD(indirectbr,  IndirectBr);
 659   INSTKEYWORD(invoke,      Invoke);
 660   INSTKEYWORD(unwind,      Unwind);
 661   INSTKEYWORD(unreachable, Unreachable);
 662
 663   INSTKEYWORD(alloca,      Alloca);
 664   INSTKEYWORD(load,        Load);
 665   INSTKEYWORD(store,       Store);
 666   INSTKEYWORD(getelementptr, GetElementPtr);
 667
 668   INSTKEYWORD(extractelement, ExtractElement);
 669   INSTKEYWORD(insertelement,  InsertElement);
 670   INSTKEYWORD(shufflevector,  ShuffleVector);
 671   INSTKEYWORD(getresult,      ExtractValue);
 672   INSTKEYWORD(extractvalue,   ExtractValue);
 673   INSTKEYWORD(insertvalue,    InsertValue);
 674 #undef INSTKEYWORD
 675
 676   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
 677   // the CFE to avoid forcing it to deal with 64-bit numbers.
 678   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
 679       TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) {
 680     int len = CurPtr-TokStart-3;
 681     uint32_t bits = len * 4;
 682     APInt Tmp(bits, StringRef(TokStart+3, len), 16);
 683     uint32_t activeBits = Tmp.getActiveBits();
 684     if (activeBits > 0 && activeBits < bits)
 685       Tmp.trunc(activeBits);
 686     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
 687     return lltok::APSInt;
 688   }
 689
 690   // If this is "cc1234", return this as just "cc".
 691   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
 692     CurPtr = TokStart+2;
 693     return lltok::kw_cc;
 694   }
 695
 696   // If this starts with "call", return it as CALL.  This is to support old
 697   // broken .ll files.  FIXME: remove this with LLVM 3.0.
 698   if (CurPtr-TokStart > 4 && !memcmp(TokStart, "call", 4)) {
 699     CurPtr = TokStart+4;
 700     UIntVal = Instruction::Call;
 701     return lltok::kw_call;
 702   }
 703
 704   // Finally, if this isn't known, return an error.
 705   CurPtr = TokStart+1;
 706   return lltok::Error;
 707 }
 708
 709
 710 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
 711 /// that this is not a label:
 712 ///    HexFPConstant     0x[0-9A-Fa-f]+
 713 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 714 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 715 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 716 lltok::Kind LLLexer::Lex0x() {
 717   CurPtr = TokStart + 2;
 718
 719   char Kind;
 720   if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') {
 721     Kind = *CurPtr++;
 722   } else {
 723     Kind = 'J';
 724   }
 725
 726   if (!isxdigit(CurPtr[0])) {
 727     // Bad token, return it as an error.
 728     CurPtr = TokStart+1;
 729     return lltok::Error;
 730   }
 731
 732   while (isxdigit(CurPtr[0]))
 733     ++CurPtr;
 734
 735   if (Kind == 'J') {
 736     // HexFPConstant - Floating point constant represented in IEEE format as a
 737     // hexadecimal number for when exponential notation is not precise enough.
 738     // Float and double only.
 739     APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
 740     return lltok::APFloat;
 741   }
 742
 743   uint64_t Pair[2];
 744   switch (Kind) {
 745   default: llvm_unreachable("Unknown kind!");
 746   case 'K':
 747     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
 748     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
 749     APFloatVal = APFloat(APInt(80, 2, Pair));
 750     return lltok::APFloat;
 751   case 'L':
 752     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
 753     HexToIntPair(TokStart+3, CurPtr, Pair);
 754     APFloatVal = APFloat(APInt(128, 2, Pair), true);
 755     return lltok::APFloat;
 756   case 'M':
 757     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
 758     HexToIntPair(TokStart+3, CurPtr, Pair);
 759     APFloatVal = APFloat(APInt(128, 2, Pair));
 760     return lltok::APFloat;
 761   }
 762 }
 763
 764 /// LexIdentifier: Handle several related productions:
 765 ///    Label             [-a-zA-Z$._0-9]+:
 766 ///    NInteger          -[0-9]+
 767 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 768 ///    PInteger          [0-9]+
 769 ///    HexFPConstant     0x[0-9A-Fa-f]+
 770 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 771 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 772 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 773 lltok::Kind LLLexer::LexDigitOrNegative() {
 774   // If the letter after the negative is a number, this is probably a label.
 775   if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) {
 776     // Okay, this is not a number after the -, it's probably a label.
 777     if (const char *End = isLabelTail(CurPtr)) {
 778       StrVal.assign(TokStart, End-1);
 779       CurPtr = End;
 780       return lltok::LabelStr;
 781     }
 782
 783     return lltok::Error;
 784   }
 785
 786   // At this point, it is either a label, int or fp constant.
 787
 788   // Skip digits, we have at least one.
 789   for (; isdigit(CurPtr[0]); ++CurPtr)
 790     /*empty*/;
 791
 792   // Check to see if this really is a label afterall, e.g. "-1:".
 793   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
 794     if (const char *End = isLabelTail(CurPtr)) {
 795       StrVal.assign(TokStart, End-1);
 796       CurPtr = End;
 797       return lltok::LabelStr;
 798     }
 799   }
 800
 801   // If the next character is a '.', then it is a fp value, otherwise its
 802   // integer.
 803   if (CurPtr[0] != '.') {
 804     if (TokStart[0] == '0' && TokStart[1] == 'x')
 805       return Lex0x();
 806     unsigned Len = CurPtr-TokStart;
 807     uint32_t numBits = ((Len * 64) / 19) + 2;
 808     APInt Tmp(numBits, StringRef(TokStart, Len), 10);
 809     if (TokStart[0] == '-') {
 810       uint32_t minBits = Tmp.getMinSignedBits();
 811       if (minBits > 0 && minBits < numBits)
 812         Tmp.trunc(minBits);
 813       APSIntVal = APSInt(Tmp, false);
 814     } else {
 815       uint32_t activeBits = Tmp.getActiveBits();
 816       if (activeBits > 0 && activeBits < numBits)
 817         Tmp.trunc(activeBits);
 818       APSIntVal = APSInt(Tmp, true);
 819     }
 820     return lltok::APSInt;
 821   }
 822
 823   ++CurPtr;
 824
 825   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 826   while (isdigit(CurPtr[0])) ++CurPtr;
 827
 828   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 829     if (isdigit(CurPtr[1]) ||
 830         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 831       CurPtr += 2;
 832       while (isdigit(CurPtr[0])) ++CurPtr;
 833     }
 834   }
 835
 836   APFloatVal = APFloat(atof(TokStart));
 837   return lltok::APFloat;
 838 }
 839
 840 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 841 lltok::Kind LLLexer::LexPositive() {
 842   // If the letter after the negative is a number, this is probably not a
 843   // label.
 844   if (!isdigit(CurPtr[0]))
 845     return lltok::Error;
 846
 847   // Skip digits.
 848   for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 849     /*empty*/;
 850
 851   // At this point, we need a '.'.
 852   if (CurPtr[0] != '.') {
 853     CurPtr = TokStart+1;
 854     return lltok::Error;
 855   }
 856
 857   ++CurPtr;
 858
 859   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 860   while (isdigit(CurPtr[0])) ++CurPtr;
 861
 862   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 863     if (isdigit(CurPtr[1]) ||
 864         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 865       CurPtr += 2;
 866       while (isdigit(CurPtr[0])) ++CurPtr;
 867     }
 868   }
 869
 870   APFloatVal = APFloat(atof(TokStart));
 871   return lltok::APFloat;
 872 }