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/ADT/StringExtras.h" 16 #include "llvm/ADT/Twine.h" 17 #include "llvm/AsmParser/Parser.h" 18 #include "llvm/IR/DerivedTypes.h" 19 #include "llvm/IR/Instruction.h" 20 #include "llvm/IR/LLVMContext.h" 21 #include "llvm/Support/ErrorHandling.h" 22 #include "llvm/Support/MathExtras.h" 23 #include "llvm/Support/MemoryBuffer.h" 24 #include "llvm/Support/SourceMgr.h" 25 #include "llvm/Support/raw_ostream.h" 26 #include <cctype> 27 #include <cstdio> 28 #include <cstdlib> 29 #include <cstring> 30 using namespace llvm; 31 32 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const { 33 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg); 34 return true; 35 } 36 37 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const { 38 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg); 39 } 40 41 //===----------------------------------------------------------------------===// 42 // Helper functions. 43 //===----------------------------------------------------------------------===// 44 45 // atoull - Convert an ascii string of decimal digits into the unsigned long 46 // long representation... this does not have to do input error checking, 47 // because we know that the input will be matched by a suitable regex... 48 // 49 uint64_t LLLexer::atoull(const char *Buffer, const char *End) { 50 uint64_t Result = 0; 51 for (; Buffer != End; Buffer++) { 52 uint64_t OldRes = Result; 53 Result *= 10; 54 Result += *Buffer-'0'; 55 if (Result < OldRes) { // Uh, oh, overflow detected!!! 56 Error("constant bigger than 64 bits detected!"); 57 return 0; 58 } 59 } 60 return Result; 61 } 62 63 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) { 64 uint64_t Result = 0; 65 for (; Buffer != End; ++Buffer) { 66 uint64_t OldRes = Result; 67 Result *= 16; 68 Result += hexDigitValue(*Buffer); 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 if (End - Buffer >= 16) { 82 for (int i = 0; i < 16; i++, Buffer++) { 83 assert(Buffer != End); 84 Pair[0] *= 16; 85 Pair[0] += hexDigitValue(*Buffer); 86 } 87 } 88 Pair[1] = 0; 89 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 90 Pair[1] *= 16; 91 Pair[1] += hexDigitValue(*Buffer); 92 } 93 if (Buffer != End) 94 Error("constant bigger than 128 bits detected!"); 95 } 96 97 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into 98 /// { low64, high16 } as usual for an APInt. 99 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End, 100 uint64_t Pair[2]) { 101 Pair[1] = 0; 102 for (int i=0; i<4 && Buffer != End; i++, Buffer++) { 103 assert(Buffer != End); 104 Pair[1] *= 16; 105 Pair[1] += hexDigitValue(*Buffer); 106 } 107 Pair[0] = 0; 108 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 109 Pair[0] *= 16; 110 Pair[0] += hexDigitValue(*Buffer); 111 } 112 if (Buffer != End) 113 Error("constant bigger than 128 bits detected!"); 114 } 115 116 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the 117 // appropriate character. 118 static void UnEscapeLexed(std::string &Str) { 119 if (Str.empty()) return; 120 121 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size(); 122 char *BOut = Buffer; 123 for (char *BIn = Buffer; BIn != EndBuffer; ) { 124 if (BIn[0] == '\\') { 125 if (BIn < EndBuffer-1 && BIn[1] == '\\') { 126 *BOut++ = '\\'; // Two \ becomes one 127 BIn += 2; 128 } else if (BIn < EndBuffer-2 && 129 isxdigit(static_cast<unsigned char>(BIn[1])) && 130 isxdigit(static_cast<unsigned char>(BIn[2]))) { 131 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]); 132 BIn += 3; // Skip over handled chars 133 ++BOut; 134 } else { 135 *BOut++ = *BIn++; 136 } 137 } else { 138 *BOut++ = *BIn++; 139 } 140 } 141 Str.resize(BOut-Buffer); 142 } 143 144 /// isLabelChar - Return true for [-a-zA-Z$._0-9]. 145 static bool isLabelChar(char C) { 146 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' || 147 C == '.' || C == '_'; 148 } 149 150 151 /// isLabelTail - Return true if this pointer points to a valid end of a label. 152 static const char *isLabelTail(const char *CurPtr) { 153 while (1) { 154 if (CurPtr[0] == ':') return CurPtr+1; 155 if (!isLabelChar(CurPtr[0])) return nullptr; 156 ++CurPtr; 157 } 158 } 159 160 161 162 //===----------------------------------------------------------------------===// 163 // Lexer definition. 164 //===----------------------------------------------------------------------===// 165 166 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &sm, SMDiagnostic &Err, 167 LLVMContext &C) 168 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) { 169 CurPtr = CurBuf.begin(); 170 } 171 172 int LLLexer::getNextChar() { 173 char CurChar = *CurPtr++; 174 switch (CurChar) { 175 default: return (unsigned char)CurChar; 176 case 0: 177 // A nul character in the stream is either the end of the current buffer or 178 // a random nul in the file. Disambiguate that here. 179 if (CurPtr-1 != CurBuf.end()) 180 return 0; // Just whitespace. 181 182 // Otherwise, return end of file. 183 --CurPtr; // Another call to lex will return EOF again. 184 return EOF; 185 } 186 } 187 188 189 lltok::Kind LLLexer::LexToken() { 190 TokStart = CurPtr; 191 192 int CurChar = getNextChar(); 193 switch (CurChar) { 194 default: 195 // Handle letters: [a-zA-Z_] 196 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_') 197 return LexIdentifier(); 198 199 return lltok::Error; 200 case EOF: return lltok::Eof; 201 case 0: 202 case ' ': 203 case '\t': 204 case '\n': 205 case '\r': 206 // Ignore whitespace. 207 return LexToken(); 208 case '+': return LexPositive(); 209 case '@': return LexAt(); 210 case '$': return LexDollar(); 211 case '%': return LexPercent(); 212 case '"': return LexQuote(); 213 case '.': 214 if (const char *Ptr = isLabelTail(CurPtr)) { 215 CurPtr = Ptr; 216 StrVal.assign(TokStart, CurPtr-1); 217 return lltok::LabelStr; 218 } 219 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 220 CurPtr += 2; 221 return lltok::dotdotdot; 222 } 223 return lltok::Error; 224 case ';': 225 SkipLineComment(); 226 return LexToken(); 227 case '!': return LexExclaim(); 228 case '#': return LexHash(); 229 case '0': case '1': case '2': case '3': case '4': 230 case '5': case '6': case '7': case '8': case '9': 231 case '-': 232 return LexDigitOrNegative(); 233 case '=': return lltok::equal; 234 case '[': return lltok::lsquare; 235 case ']': return lltok::rsquare; 236 case '{': return lltok::lbrace; 237 case '}': return lltok::rbrace; 238 case '<': return lltok::less; 239 case '>': return lltok::greater; 240 case '(': return lltok::lparen; 241 case ')': return lltok::rparen; 242 case ',': return lltok::comma; 243 case '*': return lltok::star; 244 case '|': return lltok::bar; 245 } 246 } 247 248 void LLLexer::SkipLineComment() { 249 while (1) { 250 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 251 return; 252 } 253 } 254 255 /// Lex all tokens that start with an @ character. 256 /// GlobalVar @\"[^\"]*\" 257 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 258 /// GlobalVarID @[0-9]+ 259 lltok::Kind LLLexer::LexAt() { 260 return LexVar(lltok::GlobalVar, lltok::GlobalID); 261 } 262 263 lltok::Kind LLLexer::LexDollar() { 264 if (const char *Ptr = isLabelTail(TokStart)) { 265 CurPtr = Ptr; 266 StrVal.assign(TokStart, CurPtr - 1); 267 return lltok::LabelStr; 268 } 269 270 // Handle DollarStringConstant: $\"[^\"]*\" 271 if (CurPtr[0] == '"') { 272 ++CurPtr; 273 274 while (1) { 275 int CurChar = getNextChar(); 276 277 if (CurChar == EOF) { 278 Error("end of file in COMDAT variable name"); 279 return lltok::Error; 280 } 281 if (CurChar == '"') { 282 StrVal.assign(TokStart + 2, CurPtr - 1); 283 UnEscapeLexed(StrVal); 284 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 285 Error("Null bytes are not allowed in names"); 286 return lltok::Error; 287 } 288 return lltok::ComdatVar; 289 } 290 } 291 } 292 293 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]* 294 if (ReadVarName()) 295 return lltok::ComdatVar; 296 297 return lltok::Error; 298 } 299 300 /// ReadString - Read a string until the closing quote. 301 lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 302 const char *Start = CurPtr; 303 while (1) { 304 int CurChar = getNextChar(); 305 306 if (CurChar == EOF) { 307 Error("end of file in string constant"); 308 return lltok::Error; 309 } 310 if (CurChar == '"') { 311 StrVal.assign(Start, CurPtr-1); 312 UnEscapeLexed(StrVal); 313 return kind; 314 } 315 } 316 } 317 318 /// ReadVarName - Read the rest of a token containing a variable name. 319 bool LLLexer::ReadVarName() { 320 const char *NameStart = CurPtr; 321 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 322 CurPtr[0] == '-' || CurPtr[0] == '$' || 323 CurPtr[0] == '.' || CurPtr[0] == '_') { 324 ++CurPtr; 325 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 326 CurPtr[0] == '-' || CurPtr[0] == '$' || 327 CurPtr[0] == '.' || CurPtr[0] == '_') 328 ++CurPtr; 329 330 StrVal.assign(NameStart, CurPtr); 331 return true; 332 } 333 return false; 334 } 335 336 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) { 337 // Handle StringConstant: \"[^\"]*\" 338 if (CurPtr[0] == '"') { 339 ++CurPtr; 340 341 while (1) { 342 int CurChar = getNextChar(); 343 344 if (CurChar == EOF) { 345 Error("end of file in global variable name"); 346 return lltok::Error; 347 } 348 if (CurChar == '"') { 349 StrVal.assign(TokStart+2, CurPtr-1); 350 UnEscapeLexed(StrVal); 351 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 352 Error("Null bytes are not allowed in names"); 353 return lltok::Error; 354 } 355 return Var; 356 } 357 } 358 } 359 360 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]* 361 if (ReadVarName()) 362 return Var; 363 364 // Handle VarID: [0-9]+ 365 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 366 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 367 /*empty*/; 368 369 uint64_t Val = atoull(TokStart+1, CurPtr); 370 if ((unsigned)Val != Val) 371 Error("invalid value number (too large)!"); 372 UIntVal = unsigned(Val); 373 return VarID; 374 } 375 return lltok::Error; 376 } 377 378 /// Lex all tokens that start with a % character. 379 /// LocalVar ::= %\"[^\"]*\" 380 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 381 /// LocalVarID ::= %[0-9]+ 382 lltok::Kind LLLexer::LexPercent() { 383 return LexVar(lltok::LocalVar, lltok::LocalVarID); 384 } 385 386 /// Lex all tokens that start with a " character. 387 /// QuoteLabel "[^"]+": 388 /// StringConstant "[^"]*" 389 lltok::Kind LLLexer::LexQuote() { 390 lltok::Kind kind = ReadString(lltok::StringConstant); 391 if (kind == lltok::Error || kind == lltok::Eof) 392 return kind; 393 394 if (CurPtr[0] == ':') { 395 ++CurPtr; 396 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 397 Error("Null bytes are not allowed in names"); 398 kind = lltok::Error; 399 } else { 400 kind = lltok::LabelStr; 401 } 402 } 403 404 return kind; 405 } 406 407 /// Lex all tokens that start with a ! character. 408 /// !foo 409 /// ! 410 lltok::Kind LLLexer::LexExclaim() { 411 // Lex a metadata name as a MetadataVar. 412 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 413 CurPtr[0] == '-' || CurPtr[0] == '$' || 414 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 415 ++CurPtr; 416 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 417 CurPtr[0] == '-' || CurPtr[0] == '$' || 418 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 419 ++CurPtr; 420 421 StrVal.assign(TokStart+1, CurPtr); // Skip ! 422 UnEscapeLexed(StrVal); 423 return lltok::MetadataVar; 424 } 425 return lltok::exclaim; 426 } 427 428 /// Lex all tokens that start with a # character. 429 /// AttrGrpID ::= #[0-9]+ 430 lltok::Kind LLLexer::LexHash() { 431 // Handle AttrGrpID: #[0-9]+ 432 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 433 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 434 /*empty*/; 435 436 uint64_t Val = atoull(TokStart+1, CurPtr); 437 if ((unsigned)Val != Val) 438 Error("invalid value number (too large)!"); 439 UIntVal = unsigned(Val); 440 return lltok::AttrGrpID; 441 } 442 443 return lltok::Error; 444 } 445 446 /// Lex a label, integer type, keyword, or hexadecimal integer constant. 447 /// Label [-a-zA-Z$._0-9]+: 448 /// IntegerType i[0-9]+ 449 /// Keyword sdiv, float, ... 450 /// HexIntConstant [us]0x[0-9A-Fa-f]+ 451 lltok::Kind LLLexer::LexIdentifier() { 452 const char *StartChar = CurPtr; 453 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar; 454 const char *KeywordEnd = nullptr; 455 456 for (; isLabelChar(*CurPtr); ++CurPtr) { 457 // If we decide this is an integer, remember the end of the sequence. 458 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr))) 459 IntEnd = CurPtr; 460 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) && 461 *CurPtr != '_') 462 KeywordEnd = CurPtr; 463 } 464 465 // If we stopped due to a colon, this really is a label. 466 if (*CurPtr == ':') { 467 StrVal.assign(StartChar-1, CurPtr++); 468 return lltok::LabelStr; 469 } 470 471 // Otherwise, this wasn't a label. If this was valid as an integer type, 472 // return it. 473 if (!IntEnd) IntEnd = CurPtr; 474 if (IntEnd != StartChar) { 475 CurPtr = IntEnd; 476 uint64_t NumBits = atoull(StartChar, CurPtr); 477 if (NumBits < IntegerType::MIN_INT_BITS || 478 NumBits > IntegerType::MAX_INT_BITS) { 479 Error("bitwidth for integer type out of range!"); 480 return lltok::Error; 481 } 482 TyVal = IntegerType::get(Context, NumBits); 483 return lltok::Type; 484 } 485 486 // Otherwise, this was a letter sequence. See which keyword this is. 487 if (!KeywordEnd) KeywordEnd = CurPtr; 488 CurPtr = KeywordEnd; 489 --StartChar; 490 StringRef Keyword(StartChar, CurPtr - StartChar); 491 #define KEYWORD(STR) \ 492 do { \ 493 if (Keyword == #STR) \ 494 return lltok::kw_##STR; \ 495 } while (0) 496 497 KEYWORD(true); KEYWORD(false); 498 KEYWORD(declare); KEYWORD(define); 499 KEYWORD(global); KEYWORD(constant); 500 501 KEYWORD(private); 502 KEYWORD(internal); 503 KEYWORD(available_externally); 504 KEYWORD(linkonce); 505 KEYWORD(linkonce_odr); 506 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg". 507 KEYWORD(weak_odr); 508 KEYWORD(appending); 509 KEYWORD(dllimport); 510 KEYWORD(dllexport); 511 KEYWORD(common); 512 KEYWORD(default); 513 KEYWORD(hidden); 514 KEYWORD(protected); 515 KEYWORD(unnamed_addr); 516 KEYWORD(externally_initialized); 517 KEYWORD(extern_weak); 518 KEYWORD(external); 519 KEYWORD(thread_local); 520 KEYWORD(localdynamic); 521 KEYWORD(initialexec); 522 KEYWORD(localexec); 523 KEYWORD(zeroinitializer); 524 KEYWORD(undef); 525 KEYWORD(null); 526 KEYWORD(none); 527 KEYWORD(to); 528 KEYWORD(caller); 529 KEYWORD(within); 530 KEYWORD(from); 531 KEYWORD(tail); 532 KEYWORD(musttail); 533 KEYWORD(notail); 534 KEYWORD(target); 535 KEYWORD(triple); 536 KEYWORD(unwind); 537 KEYWORD(deplibs); // FIXME: Remove in 4.0. 538 KEYWORD(datalayout); 539 KEYWORD(volatile); 540 KEYWORD(atomic); 541 KEYWORD(unordered); 542 KEYWORD(monotonic); 543 KEYWORD(acquire); 544 KEYWORD(release); 545 KEYWORD(acq_rel); 546 KEYWORD(seq_cst); 547 KEYWORD(singlethread); 548 549 KEYWORD(nnan); 550 KEYWORD(ninf); 551 KEYWORD(nsz); 552 KEYWORD(arcp); 553 KEYWORD(fast); 554 KEYWORD(nuw); 555 KEYWORD(nsw); 556 KEYWORD(exact); 557 KEYWORD(inbounds); 558 KEYWORD(align); 559 KEYWORD(addrspace); 560 KEYWORD(section); 561 KEYWORD(alias); 562 KEYWORD(module); 563 KEYWORD(asm); 564 KEYWORD(sideeffect); 565 KEYWORD(alignstack); 566 KEYWORD(inteldialect); 567 KEYWORD(gc); 568 KEYWORD(prefix); 569 KEYWORD(prologue); 570 571 KEYWORD(ccc); 572 KEYWORD(fastcc); 573 KEYWORD(coldcc); 574 KEYWORD(x86_stdcallcc); 575 KEYWORD(x86_fastcallcc); 576 KEYWORD(x86_thiscallcc); 577 KEYWORD(x86_vectorcallcc); 578 KEYWORD(arm_apcscc); 579 KEYWORD(arm_aapcscc); 580 KEYWORD(arm_aapcs_vfpcc); 581 KEYWORD(msp430_intrcc); 582 KEYWORD(avr_intrcc); 583 KEYWORD(avr_signalcc); 584 KEYWORD(ptx_kernel); 585 KEYWORD(ptx_device); 586 KEYWORD(spir_kernel); 587 KEYWORD(spir_func); 588 KEYWORD(intel_ocl_bicc); 589 KEYWORD(x86_64_sysvcc); 590 KEYWORD(x86_64_win64cc); 591 KEYWORD(webkit_jscc); 592 KEYWORD(anyregcc); 593 KEYWORD(preserve_mostcc); 594 KEYWORD(preserve_allcc); 595 KEYWORD(ghccc); 596 KEYWORD(x86_intrcc); 597 KEYWORD(hhvmcc); 598 KEYWORD(hhvm_ccc); 599 KEYWORD(cxx_fast_tlscc); 600 601 KEYWORD(cc); 602 KEYWORD(c); 603 604 KEYWORD(attributes); 605 606 KEYWORD(alwaysinline); 607 KEYWORD(argmemonly); 608 KEYWORD(builtin); 609 KEYWORD(byval); 610 KEYWORD(inalloca); 611 KEYWORD(cold); 612 KEYWORD(convergent); 613 KEYWORD(dereferenceable); 614 KEYWORD(dereferenceable_or_null); 615 KEYWORD(inaccessiblememonly); 616 KEYWORD(inaccessiblemem_or_argmemonly); 617 KEYWORD(inlinehint); 618 KEYWORD(inreg); 619 KEYWORD(jumptable); 620 KEYWORD(minsize); 621 KEYWORD(naked); 622 KEYWORD(nest); 623 KEYWORD(noalias); 624 KEYWORD(nobuiltin); 625 KEYWORD(nocapture); 626 KEYWORD(noduplicate); 627 KEYWORD(noimplicitfloat); 628 KEYWORD(noinline); 629 KEYWORD(norecurse); 630 KEYWORD(nonlazybind); 631 KEYWORD(nonnull); 632 KEYWORD(noredzone); 633 KEYWORD(noreturn); 634 KEYWORD(nounwind); 635 KEYWORD(optnone); 636 KEYWORD(optsize); 637 KEYWORD(readnone); 638 KEYWORD(readonly); 639 KEYWORD(returned); 640 KEYWORD(returns_twice); 641 KEYWORD(signext); 642 KEYWORD(sret); 643 KEYWORD(ssp); 644 KEYWORD(sspreq); 645 KEYWORD(sspstrong); 646 KEYWORD(safestack); 647 KEYWORD(sanitize_address); 648 KEYWORD(sanitize_thread); 649 KEYWORD(sanitize_memory); 650 KEYWORD(uwtable); 651 KEYWORD(zeroext); 652 653 KEYWORD(type); 654 KEYWORD(opaque); 655 656 KEYWORD(comdat); 657 658 // Comdat types 659 KEYWORD(any); 660 KEYWORD(exactmatch); 661 KEYWORD(largest); 662 KEYWORD(noduplicates); 663 KEYWORD(samesize); 664 665 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 666 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 667 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 668 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 669 670 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 671 KEYWORD(umin); 672 673 KEYWORD(x); 674 KEYWORD(blockaddress); 675 676 // Metadata types. 677 KEYWORD(distinct); 678 679 // Use-list order directives. 680 KEYWORD(uselistorder); 681 KEYWORD(uselistorder_bb); 682 683 KEYWORD(personality); 684 KEYWORD(cleanup); 685 KEYWORD(catch); 686 KEYWORD(filter); 687 #undef KEYWORD 688 689 // Keywords for types. 690 #define TYPEKEYWORD(STR, LLVMTY) \ 691 do { \ 692 if (Keyword == STR) { \ 693 TyVal = LLVMTY; \ 694 return lltok::Type; \ 695 } \ 696 } while (false) 697 TYPEKEYWORD("void", Type::getVoidTy(Context)); 698 TYPEKEYWORD("half", Type::getHalfTy(Context)); 699 TYPEKEYWORD("float", Type::getFloatTy(Context)); 700 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 701 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 702 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 703 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 704 TYPEKEYWORD("label", Type::getLabelTy(Context)); 705 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 706 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 707 TYPEKEYWORD("token", Type::getTokenTy(Context)); 708 #undef TYPEKEYWORD 709 710 // Keywords for instructions. 711 #define INSTKEYWORD(STR, Enum) \ 712 do { \ 713 if (Keyword == #STR) { \ 714 UIntVal = Instruction::Enum; \ 715 return lltok::kw_##STR; \ 716 } \ 717 } while (false) 718 719 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 720 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 721 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 722 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 723 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 724 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 725 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 726 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 727 728 INSTKEYWORD(phi, PHI); 729 INSTKEYWORD(call, Call); 730 INSTKEYWORD(trunc, Trunc); 731 INSTKEYWORD(zext, ZExt); 732 INSTKEYWORD(sext, SExt); 733 INSTKEYWORD(fptrunc, FPTrunc); 734 INSTKEYWORD(fpext, FPExt); 735 INSTKEYWORD(uitofp, UIToFP); 736 INSTKEYWORD(sitofp, SIToFP); 737 INSTKEYWORD(fptoui, FPToUI); 738 INSTKEYWORD(fptosi, FPToSI); 739 INSTKEYWORD(inttoptr, IntToPtr); 740 INSTKEYWORD(ptrtoint, PtrToInt); 741 INSTKEYWORD(bitcast, BitCast); 742 INSTKEYWORD(addrspacecast, AddrSpaceCast); 743 INSTKEYWORD(select, Select); 744 INSTKEYWORD(va_arg, VAArg); 745 INSTKEYWORD(ret, Ret); 746 INSTKEYWORD(br, Br); 747 INSTKEYWORD(switch, Switch); 748 INSTKEYWORD(indirectbr, IndirectBr); 749 INSTKEYWORD(invoke, Invoke); 750 INSTKEYWORD(resume, Resume); 751 INSTKEYWORD(unreachable, Unreachable); 752 753 INSTKEYWORD(alloca, Alloca); 754 INSTKEYWORD(load, Load); 755 INSTKEYWORD(store, Store); 756 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 757 INSTKEYWORD(atomicrmw, AtomicRMW); 758 INSTKEYWORD(fence, Fence); 759 INSTKEYWORD(getelementptr, GetElementPtr); 760 761 INSTKEYWORD(extractelement, ExtractElement); 762 INSTKEYWORD(insertelement, InsertElement); 763 INSTKEYWORD(shufflevector, ShuffleVector); 764 INSTKEYWORD(extractvalue, ExtractValue); 765 INSTKEYWORD(insertvalue, InsertValue); 766 INSTKEYWORD(landingpad, LandingPad); 767 INSTKEYWORD(cleanupret, CleanupRet); 768 INSTKEYWORD(catchret, CatchRet); 769 INSTKEYWORD(catchswitch, CatchSwitch); 770 INSTKEYWORD(catchpad, CatchPad); 771 INSTKEYWORD(cleanuppad, CleanupPad); 772 #undef INSTKEYWORD 773 774 #define DWKEYWORD(TYPE, TOKEN) \ 775 do { \ 776 if (Keyword.startswith("DW_" #TYPE "_")) { \ 777 StrVal.assign(Keyword.begin(), Keyword.end()); \ 778 return lltok::TOKEN; \ 779 } \ 780 } while (false) 781 DWKEYWORD(TAG, DwarfTag); 782 DWKEYWORD(ATE, DwarfAttEncoding); 783 DWKEYWORD(VIRTUALITY, DwarfVirtuality); 784 DWKEYWORD(LANG, DwarfLang); 785 DWKEYWORD(OP, DwarfOp); 786 DWKEYWORD(MACINFO, DwarfMacinfo); 787 #undef DWKEYWORD 788 789 if (Keyword.startswith("DIFlag")) { 790 StrVal.assign(Keyword.begin(), Keyword.end()); 791 return lltok::DIFlag; 792 } 793 794 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 795 // the CFE to avoid forcing it to deal with 64-bit numbers. 796 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 797 TokStart[1] == '0' && TokStart[2] == 'x' && 798 isxdigit(static_cast<unsigned char>(TokStart[3]))) { 799 int len = CurPtr-TokStart-3; 800 uint32_t bits = len * 4; 801 StringRef HexStr(TokStart + 3, len); 802 if (!std::all_of(HexStr.begin(), HexStr.end(), isxdigit)) { 803 // Bad token, return it as an error. 804 CurPtr = TokStart+3; 805 return lltok::Error; 806 } 807 APInt Tmp(bits, HexStr, 16); 808 uint32_t activeBits = Tmp.getActiveBits(); 809 if (activeBits > 0 && activeBits < bits) 810 Tmp = Tmp.trunc(activeBits); 811 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 812 return lltok::APSInt; 813 } 814 815 // If this is "cc1234", return this as just "cc". 816 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 817 CurPtr = TokStart+2; 818 return lltok::kw_cc; 819 } 820 821 // Finally, if this isn't known, return an error. 822 CurPtr = TokStart+1; 823 return lltok::Error; 824 } 825 826 /// Lex all tokens that start with a 0x prefix, knowing they match and are not 827 /// labels. 828 /// HexFPConstant 0x[0-9A-Fa-f]+ 829 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 830 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 831 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 832 /// HexHalfConstant 0xH[0-9A-Fa-f]+ 833 lltok::Kind LLLexer::Lex0x() { 834 CurPtr = TokStart + 2; 835 836 char Kind; 837 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') { 838 Kind = *CurPtr++; 839 } else { 840 Kind = 'J'; 841 } 842 843 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) { 844 // Bad token, return it as an error. 845 CurPtr = TokStart+1; 846 return lltok::Error; 847 } 848 849 while (isxdigit(static_cast<unsigned char>(CurPtr[0]))) 850 ++CurPtr; 851 852 if (Kind == 'J') { 853 // HexFPConstant - Floating point constant represented in IEEE format as a 854 // hexadecimal number for when exponential notation is not precise enough. 855 // Half, Float, and double only. 856 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr))); 857 return lltok::APFloat; 858 } 859 860 uint64_t Pair[2]; 861 switch (Kind) { 862 default: llvm_unreachable("Unknown kind!"); 863 case 'K': 864 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 865 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 866 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair)); 867 return lltok::APFloat; 868 case 'L': 869 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 870 HexToIntPair(TokStart+3, CurPtr, Pair); 871 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair)); 872 return lltok::APFloat; 873 case 'M': 874 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 875 HexToIntPair(TokStart+3, CurPtr, Pair); 876 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair)); 877 return lltok::APFloat; 878 case 'H': 879 APFloatVal = APFloat(APFloat::IEEEhalf, 880 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 881 return lltok::APFloat; 882 } 883 } 884 885 /// Lex tokens for a label or a numeric constant, possibly starting with -. 886 /// Label [-a-zA-Z$._0-9]+: 887 /// NInteger -[0-9]+ 888 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 889 /// PInteger [0-9]+ 890 /// HexFPConstant 0x[0-9A-Fa-f]+ 891 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 892 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 893 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 894 lltok::Kind LLLexer::LexDigitOrNegative() { 895 // If the letter after the negative is not a number, this is probably a label. 896 if (!isdigit(static_cast<unsigned char>(TokStart[0])) && 897 !isdigit(static_cast<unsigned char>(CurPtr[0]))) { 898 // Okay, this is not a number after the -, it's probably a label. 899 if (const char *End = isLabelTail(CurPtr)) { 900 StrVal.assign(TokStart, End-1); 901 CurPtr = End; 902 return lltok::LabelStr; 903 } 904 905 return lltok::Error; 906 } 907 908 // At this point, it is either a label, int or fp constant. 909 910 // Skip digits, we have at least one. 911 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 912 /*empty*/; 913 914 // Check to see if this really is a label afterall, e.g. "-1:". 915 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 916 if (const char *End = isLabelTail(CurPtr)) { 917 StrVal.assign(TokStart, End-1); 918 CurPtr = End; 919 return lltok::LabelStr; 920 } 921 } 922 923 // If the next character is a '.', then it is a fp value, otherwise its 924 // integer. 925 if (CurPtr[0] != '.') { 926 if (TokStart[0] == '0' && TokStart[1] == 'x') 927 return Lex0x(); 928 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart)); 929 return lltok::APSInt; 930 } 931 932 ++CurPtr; 933 934 // Skip over [0-9]*([eE][-+]?[0-9]+)? 935 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 936 937 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 938 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 939 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 940 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 941 CurPtr += 2; 942 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 943 } 944 } 945 946 APFloatVal = APFloat(std::atof(TokStart)); 947 return lltok::APFloat; 948 } 949 950 /// Lex a floating point constant starting with +. 951 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 952 lltok::Kind LLLexer::LexPositive() { 953 // If the letter after the negative is a number, this is probably not a 954 // label. 955 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 956 return lltok::Error; 957 958 // Skip digits. 959 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 960 /*empty*/; 961 962 // At this point, we need a '.'. 963 if (CurPtr[0] != '.') { 964 CurPtr = TokStart+1; 965 return lltok::Error; 966 } 967 968 ++CurPtr; 969 970 // Skip over [0-9]*([eE][-+]?[0-9]+)? 971 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 972 973 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 974 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 975 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 976 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 977 CurPtr += 2; 978 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 979 } 980 } 981 982 APFloatVal = APFloat(std::atof(TokStart)); 983 return lltok::APFloat; 984 } 985