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$._0-9]*
 375   if (ReadVarName())
 376     return lltok::LocalVar;
 377
 378   // Handle LocalVarID: %[0-9]+
 379   if (isdigit(CurPtr[0])) {
 380     for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 381       /*empty*/;
 382
 383     uint64_t Val = atoull(TokStart+1, CurPtr);
 384     if ((unsigned)Val != Val)
 385       Error("invalid value number (too large)!");
 386     UIntVal = unsigned(Val);
 387     return lltok::LocalVarID;
 388   }
 389
 390   return lltok::Error;
 391 }
 392
 393 /// LexQuote - Lex all tokens that start with a " character:
 394 ///   QuoteLabel        "[^"]+":
 395 ///   StringConstant    "[^"]*"
 396 lltok::Kind LLLexer::LexQuote() {
 397   lltok::Kind kind = ReadString(lltok::StringConstant);
 398   if (kind == lltok::Error || kind == lltok::Eof)
 399     return kind;
 400
 401   if (CurPtr[0] == ':') {
 402     ++CurPtr;
 403     kind = lltok::LabelStr;
 404   }
 405
 406   return kind;
 407 }
 408
 409 /// LexExclaim:
 410 ///    !foo
 411 ///    !
 412 lltok::Kind LLLexer::LexExclaim() {
 413   // Lex a metadata name as a MetadataVar.
 414   if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 415       CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
 416     ++CurPtr;
 417     while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
 418            CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
 419       ++CurPtr;
 420
 421     StrVal.assign(TokStart+1, CurPtr);   // Skip !
 422     UnEscapeLexed(StrVal);
 423     return lltok::MetadataVar;
 424   }
 425   return lltok::exclaim;
 426 }
 427
 428 /// LexIdentifier: Handle several related productions:
 429 ///    Label           [-a-zA-Z$._0-9]+:
 430 ///    IntegerType     i[0-9]+
 431 ///    Keyword         sdiv, float, ...
 432 ///    HexIntConstant  [us]0x[0-9A-Fa-f]+
 433 lltok::Kind LLLexer::LexIdentifier() {
 434   const char *StartChar = CurPtr;
 435   const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
 436   const char *KeywordEnd = 0;
 437
 438   for (; isLabelChar(*CurPtr); ++CurPtr) {
 439     // If we decide this is an integer, remember the end of the sequence.
 440     if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr;
 441     if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr;
 442   }
 443
 444   // If we stopped due to a colon, this really is a label.
 445   if (*CurPtr == ':') {
 446     StrVal.assign(StartChar-1, CurPtr++);
 447     return lltok::LabelStr;
 448   }
 449
 450   // Otherwise, this wasn't a label.  If this was valid as an integer type,
 451   // return it.
 452   if (IntEnd == 0) IntEnd = CurPtr;
 453   if (IntEnd != StartChar) {
 454     CurPtr = IntEnd;
 455     uint64_t NumBits = atoull(StartChar, CurPtr);
 456     if (NumBits < IntegerType::MIN_INT_BITS ||
 457         NumBits > IntegerType::MAX_INT_BITS) {
 458       Error("bitwidth for integer type out of range!");
 459       return lltok::Error;
 460     }
 461     TyVal = IntegerType::get(Context, NumBits);
 462     return lltok::Type;
 463   }
 464
 465   // Otherwise, this was a letter sequence.  See which keyword this is.
 466   if (KeywordEnd == 0) KeywordEnd = CurPtr;
 467   CurPtr = KeywordEnd;
 468   --StartChar;
 469   unsigned Len = CurPtr-StartChar;
 470 #define KEYWORD(STR) \
 471   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
 472     return lltok::kw_##STR;
 473
 474   KEYWORD(true);    KEYWORD(false);
 475   KEYWORD(declare); KEYWORD(define);
 476   KEYWORD(global);  KEYWORD(constant);
 477
 478   KEYWORD(private);
 479   KEYWORD(linker_private);
 480   KEYWORD(linker_private_weak);
 481   KEYWORD(linker_private_weak_def_auto);
 482   KEYWORD(internal);
 483   KEYWORD(available_externally);
 484   KEYWORD(linkonce);
 485   KEYWORD(linkonce_odr);
 486   KEYWORD(weak);
 487   KEYWORD(weak_odr);
 488   KEYWORD(appending);
 489   KEYWORD(dllimport);
 490   KEYWORD(dllexport);
 491   KEYWORD(common);
 492   KEYWORD(default);
 493   KEYWORD(hidden);
 494   KEYWORD(protected);
 495   KEYWORD(unnamed_addr);
 496   KEYWORD(extern_weak);
 497   KEYWORD(external);
 498   KEYWORD(thread_local);
 499   KEYWORD(zeroinitializer);
 500   KEYWORD(undef);
 501   KEYWORD(null);
 502   KEYWORD(to);
 503   KEYWORD(tail);
 504   KEYWORD(target);
 505   KEYWORD(triple);
 506   KEYWORD(deplibs);
 507   KEYWORD(datalayout);
 508   KEYWORD(volatile);
 509   KEYWORD(nuw);
 510   KEYWORD(nsw);
 511   KEYWORD(exact);
 512   KEYWORD(inbounds);
 513   KEYWORD(align);
 514   KEYWORD(addrspace);
 515   KEYWORD(section);
 516   KEYWORD(alias);
 517   KEYWORD(module);
 518   KEYWORD(asm);
 519   KEYWORD(sideeffect);
 520   KEYWORD(alignstack);
 521   KEYWORD(gc);
 522
 523   KEYWORD(ccc);
 524   KEYWORD(fastcc);
 525   KEYWORD(coldcc);
 526   KEYWORD(x86_stdcallcc);
 527   KEYWORD(x86_fastcallcc);
 528   KEYWORD(x86_thiscallcc);
 529   KEYWORD(arm_apcscc);
 530   KEYWORD(arm_aapcscc);
 531   KEYWORD(arm_aapcs_vfpcc);
 532   KEYWORD(msp430_intrcc);
 533   KEYWORD(ptx_kernel);
 534   KEYWORD(ptx_device);
 535
 536   KEYWORD(cc);
 537   KEYWORD(c);
 538
 539   KEYWORD(signext);
 540   KEYWORD(zeroext);
 541   KEYWORD(inreg);
 542   KEYWORD(sret);
 543   KEYWORD(nounwind);
 544   KEYWORD(noreturn);
 545   KEYWORD(noalias);
 546   KEYWORD(nocapture);
 547   KEYWORD(byval);
 548   KEYWORD(nest);
 549   KEYWORD(readnone);
 550   KEYWORD(readonly);
 551   KEYWORD(uwtable);
 552
 553   KEYWORD(inlinehint);
 554   KEYWORD(noinline);
 555   KEYWORD(alwaysinline);
 556   KEYWORD(optsize);
 557   KEYWORD(ssp);
 558   KEYWORD(sspreq);
 559   KEYWORD(noredzone);
 560   KEYWORD(noimplicitfloat);
 561   KEYWORD(naked);
 562   KEYWORD(hotpatch);
 563   KEYWORD(nonlazybind);
 564
 565   KEYWORD(type);
 566   KEYWORD(opaque);
 567
 568   KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
 569   KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
 570   KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
 571   KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
 572
 573   KEYWORD(x);
 574   KEYWORD(blockaddress);
 575 #undef KEYWORD
 576
 577   // Keywords for types.
 578 #define TYPEKEYWORD(STR, LLVMTY) \
 579   if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
 580     TyVal = LLVMTY; return lltok::Type; }
 581   TYPEKEYWORD("void",      Type::getVoidTy(Context));
 582   TYPEKEYWORD("float",     Type::getFloatTy(Context));
 583   TYPEKEYWORD("double",    Type::getDoubleTy(Context));
 584   TYPEKEYWORD("x86_fp80",  Type::getX86_FP80Ty(Context));
 585   TYPEKEYWORD("fp128",     Type::getFP128Ty(Context));
 586   TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
 587   TYPEKEYWORD("label",     Type::getLabelTy(Context));
 588   TYPEKEYWORD("metadata",  Type::getMetadataTy(Context));
 589   TYPEKEYWORD("x86_mmx",   Type::getX86_MMXTy(Context));
 590 #undef TYPEKEYWORD
 591
 592   // Keywords for instructions.
 593 #define INSTKEYWORD(STR, Enum) \
 594   if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
 595     UIntVal = Instruction::Enum; return lltok::kw_##STR; }
 596
 597   INSTKEYWORD(add,   Add);  INSTKEYWORD(fadd,   FAdd);
 598   INSTKEYWORD(sub,   Sub);  INSTKEYWORD(fsub,   FSub);
 599   INSTKEYWORD(mul,   Mul);  INSTKEYWORD(fmul,   FMul);
 600   INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
 601   INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
 602   INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
 603   INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
 604   INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
 605
 606   INSTKEYWORD(phi,         PHI);
 607   INSTKEYWORD(call,        Call);
 608   INSTKEYWORD(trunc,       Trunc);
 609   INSTKEYWORD(zext,        ZExt);
 610   INSTKEYWORD(sext,        SExt);
 611   INSTKEYWORD(fptrunc,     FPTrunc);
 612   INSTKEYWORD(fpext,       FPExt);
 613   INSTKEYWORD(uitofp,      UIToFP);
 614   INSTKEYWORD(sitofp,      SIToFP);
 615   INSTKEYWORD(fptoui,      FPToUI);
 616   INSTKEYWORD(fptosi,      FPToSI);
 617   INSTKEYWORD(inttoptr,    IntToPtr);
 618   INSTKEYWORD(ptrtoint,    PtrToInt);
 619   INSTKEYWORD(bitcast,     BitCast);
 620   INSTKEYWORD(select,      Select);
 621   INSTKEYWORD(va_arg,      VAArg);
 622   INSTKEYWORD(ret,         Ret);
 623   INSTKEYWORD(br,          Br);
 624   INSTKEYWORD(switch,      Switch);
 625   INSTKEYWORD(indirectbr,  IndirectBr);
 626   INSTKEYWORD(invoke,      Invoke);
 627   INSTKEYWORD(unwind,      Unwind);
 628   INSTKEYWORD(unreachable, Unreachable);
 629
 630   INSTKEYWORD(alloca,      Alloca);
 631   INSTKEYWORD(load,        Load);
 632   INSTKEYWORD(store,       Store);
 633   INSTKEYWORD(getelementptr, GetElementPtr);
 634
 635   INSTKEYWORD(extractelement, ExtractElement);
 636   INSTKEYWORD(insertelement,  InsertElement);
 637   INSTKEYWORD(shufflevector,  ShuffleVector);
 638   INSTKEYWORD(extractvalue,   ExtractValue);
 639   INSTKEYWORD(insertvalue,    InsertValue);
 640 #undef INSTKEYWORD
 641
 642   // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
 643   // the CFE to avoid forcing it to deal with 64-bit numbers.
 644   if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
 645       TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) {
 646     int len = CurPtr-TokStart-3;
 647     uint32_t bits = len * 4;
 648     APInt Tmp(bits, StringRef(TokStart+3, len), 16);
 649     uint32_t activeBits = Tmp.getActiveBits();
 650     if (activeBits > 0 && activeBits < bits)
 651       Tmp = Tmp.trunc(activeBits);
 652     APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
 653     return lltok::APSInt;
 654   }
 655
 656   // If this is "cc1234", return this as just "cc".
 657   if (TokStart[0] == 'c' && TokStart[1] == 'c') {
 658     CurPtr = TokStart+2;
 659     return lltok::kw_cc;
 660   }
 661
 662   // Finally, if this isn't known, return an error.
 663   CurPtr = TokStart+1;
 664   return lltok::Error;
 665 }
 666
 667
 668 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
 669 /// that this is not a label:
 670 ///    HexFPConstant     0x[0-9A-Fa-f]+
 671 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 672 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 673 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 674 lltok::Kind LLLexer::Lex0x() {
 675   CurPtr = TokStart + 2;
 676
 677   char Kind;
 678   if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') {
 679     Kind = *CurPtr++;
 680   } else {
 681     Kind = 'J';
 682   }
 683
 684   if (!isxdigit(CurPtr[0])) {
 685     // Bad token, return it as an error.
 686     CurPtr = TokStart+1;
 687     return lltok::Error;
 688   }
 689
 690   while (isxdigit(CurPtr[0]))
 691     ++CurPtr;
 692
 693   if (Kind == 'J') {
 694     // HexFPConstant - Floating point constant represented in IEEE format as a
 695     // hexadecimal number for when exponential notation is not precise enough.
 696     // Float and double only.
 697     APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
 698     return lltok::APFloat;
 699   }
 700
 701   uint64_t Pair[2];
 702   switch (Kind) {
 703   default: llvm_unreachable("Unknown kind!");
 704   case 'K':
 705     // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
 706     FP80HexToIntPair(TokStart+3, CurPtr, Pair);
 707     APFloatVal = APFloat(APInt(80, 2, Pair));
 708     return lltok::APFloat;
 709   case 'L':
 710     // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
 711     HexToIntPair(TokStart+3, CurPtr, Pair);
 712     APFloatVal = APFloat(APInt(128, 2, Pair), true);
 713     return lltok::APFloat;
 714   case 'M':
 715     // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
 716     HexToIntPair(TokStart+3, CurPtr, Pair);
 717     APFloatVal = APFloat(APInt(128, 2, Pair));
 718     return lltok::APFloat;
 719   }
 720 }
 721
 722 /// LexIdentifier: Handle several related productions:
 723 ///    Label             [-a-zA-Z$._0-9]+:
 724 ///    NInteger          -[0-9]+
 725 ///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 726 ///    PInteger          [0-9]+
 727 ///    HexFPConstant     0x[0-9A-Fa-f]+
 728 ///    HexFP80Constant   0xK[0-9A-Fa-f]+
 729 ///    HexFP128Constant  0xL[0-9A-Fa-f]+
 730 ///    HexPPC128Constant 0xM[0-9A-Fa-f]+
 731 lltok::Kind LLLexer::LexDigitOrNegative() {
 732   // If the letter after the negative is a number, this is probably a label.
 733   if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) {
 734     // Okay, this is not a number after the -, it's probably a label.
 735     if (const char *End = isLabelTail(CurPtr)) {
 736       StrVal.assign(TokStart, End-1);
 737       CurPtr = End;
 738       return lltok::LabelStr;
 739     }
 740
 741     return lltok::Error;
 742   }
 743
 744   // At this point, it is either a label, int or fp constant.
 745
 746   // Skip digits, we have at least one.
 747   for (; isdigit(CurPtr[0]); ++CurPtr)
 748     /*empty*/;
 749
 750   // Check to see if this really is a label afterall, e.g. "-1:".
 751   if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
 752     if (const char *End = isLabelTail(CurPtr)) {
 753       StrVal.assign(TokStart, End-1);
 754       CurPtr = End;
 755       return lltok::LabelStr;
 756     }
 757   }
 758
 759   // If the next character is a '.', then it is a fp value, otherwise its
 760   // integer.
 761   if (CurPtr[0] != '.') {
 762     if (TokStart[0] == '0' && TokStart[1] == 'x')
 763       return Lex0x();
 764     unsigned Len = CurPtr-TokStart;
 765     uint32_t numBits = ((Len * 64) / 19) + 2;
 766     APInt Tmp(numBits, StringRef(TokStart, Len), 10);
 767     if (TokStart[0] == '-') {
 768       uint32_t minBits = Tmp.getMinSignedBits();
 769       if (minBits > 0 && minBits < numBits)
 770         Tmp = Tmp.trunc(minBits);
 771       APSIntVal = APSInt(Tmp, false);
 772     } else {
 773       uint32_t activeBits = Tmp.getActiveBits();
 774       if (activeBits > 0 && activeBits < numBits)
 775         Tmp = Tmp.trunc(activeBits);
 776       APSIntVal = APSInt(Tmp, true);
 777     }
 778     return lltok::APSInt;
 779   }
 780
 781   ++CurPtr;
 782
 783   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 784   while (isdigit(CurPtr[0])) ++CurPtr;
 785
 786   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 787     if (isdigit(CurPtr[1]) ||
 788         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 789       CurPtr += 2;
 790       while (isdigit(CurPtr[0])) ++CurPtr;
 791     }
 792   }
 793
 794   APFloatVal = APFloat(std::atof(TokStart));
 795   return lltok::APFloat;
 796 }
 797
 798 ///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
 799 lltok::Kind LLLexer::LexPositive() {
 800   // If the letter after the negative is a number, this is probably not a
 801   // label.
 802   if (!isdigit(CurPtr[0]))
 803     return lltok::Error;
 804
 805   // Skip digits.
 806   for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
 807     /*empty*/;
 808
 809   // At this point, we need a '.'.
 810   if (CurPtr[0] != '.') {
 811     CurPtr = TokStart+1;
 812     return lltok::Error;
 813   }
 814
 815   ++CurPtr;
 816
 817   // Skip over [0-9]*([eE][-+]?[0-9]+)?
 818   while (isdigit(CurPtr[0])) ++CurPtr;
 819
 820   if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
 821     if (isdigit(CurPtr[1]) ||
 822         ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
 823       CurPtr += 2;
 824       while (isdigit(CurPtr[0])) ++CurPtr;
 825     }
 826   }
 827
 828   APFloatVal = APFloat(std::atof(TokStart));
 829   return lltok::APFloat;
 830 }