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/APInt.h" 16 #include "llvm/ADT/STLExtras.h" 17 #include "llvm/ADT/StringExtras.h" 18 #include "llvm/ADT/Twine.h" 19 #include "llvm/IR/DerivedTypes.h" 20 #include "llvm/IR/Instruction.h" 21 #include "llvm/Support/ErrorHandling.h" 22 #include "llvm/Support/SourceMgr.h" 23 #include <cassert> 24 #include <cctype> 25 #include <cstdio> 26 27 using namespace llvm; 28 29 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const { 30 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg); 31 return true; 32 } 33 34 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const { 35 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg); 36 } 37 38 //===----------------------------------------------------------------------===// 39 // Helper functions. 40 //===----------------------------------------------------------------------===// 41 42 // atoull - Convert an ascii string of decimal digits into the unsigned long 43 // long representation... this does not have to do input error checking, 44 // because we know that the input will be matched by a suitable regex... 45 // 46 uint64_t LLLexer::atoull(const char *Buffer, const char *End) { 47 uint64_t Result = 0; 48 for (; Buffer != End; Buffer++) { 49 uint64_t OldRes = Result; 50 Result *= 10; 51 Result += *Buffer-'0'; 52 if (Result < OldRes) { // Uh, oh, overflow detected!!! 53 Error("constant bigger than 64 bits detected!"); 54 return 0; 55 } 56 } 57 return Result; 58 } 59 60 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) { 61 uint64_t Result = 0; 62 for (; Buffer != End; ++Buffer) { 63 uint64_t OldRes = Result; 64 Result *= 16; 65 Result += hexDigitValue(*Buffer); 66 67 if (Result < OldRes) { // Uh, oh, overflow detected!!! 68 Error("constant bigger than 64 bits detected!"); 69 return 0; 70 } 71 } 72 return Result; 73 } 74 75 void LLLexer::HexToIntPair(const char *Buffer, const char *End, 76 uint64_t Pair[2]) { 77 Pair[0] = 0; 78 if (End - Buffer >= 16) { 79 for (int i = 0; i < 16; i++, Buffer++) { 80 assert(Buffer != End); 81 Pair[0] *= 16; 82 Pair[0] += hexDigitValue(*Buffer); 83 } 84 } 85 Pair[1] = 0; 86 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 87 Pair[1] *= 16; 88 Pair[1] += hexDigitValue(*Buffer); 89 } 90 if (Buffer != End) 91 Error("constant bigger than 128 bits detected!"); 92 } 93 94 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into 95 /// { low64, high16 } as usual for an APInt. 96 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End, 97 uint64_t Pair[2]) { 98 Pair[1] = 0; 99 for (int i=0; i<4 && Buffer != End; i++, Buffer++) { 100 assert(Buffer != End); 101 Pair[1] *= 16; 102 Pair[1] += hexDigitValue(*Buffer); 103 } 104 Pair[0] = 0; 105 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 106 Pair[0] *= 16; 107 Pair[0] += hexDigitValue(*Buffer); 108 } 109 if (Buffer != End) 110 Error("constant bigger than 128 bits detected!"); 111 } 112 113 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the 114 // appropriate character. 115 static void UnEscapeLexed(std::string &Str) { 116 if (Str.empty()) return; 117 118 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size(); 119 char *BOut = Buffer; 120 for (char *BIn = Buffer; BIn != EndBuffer; ) { 121 if (BIn[0] == '\\') { 122 if (BIn < EndBuffer-1 && BIn[1] == '\\') { 123 *BOut++ = '\\'; // Two \ becomes one 124 BIn += 2; 125 } else if (BIn < EndBuffer-2 && 126 isxdigit(static_cast<unsigned char>(BIn[1])) && 127 isxdigit(static_cast<unsigned char>(BIn[2]))) { 128 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]); 129 BIn += 3; // Skip over handled chars 130 ++BOut; 131 } else { 132 *BOut++ = *BIn++; 133 } 134 } else { 135 *BOut++ = *BIn++; 136 } 137 } 138 Str.resize(BOut-Buffer); 139 } 140 141 /// isLabelChar - Return true for [-a-zA-Z$._0-9]. 142 static bool isLabelChar(char C) { 143 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' || 144 C == '.' || C == '_'; 145 } 146 147 /// isLabelTail - Return true if this pointer points to a valid end of a label. 148 static const char *isLabelTail(const char *CurPtr) { 149 while (true) { 150 if (CurPtr[0] == ':') return CurPtr+1; 151 if (!isLabelChar(CurPtr[0])) return nullptr; 152 ++CurPtr; 153 } 154 } 155 156 //===----------------------------------------------------------------------===// 157 // Lexer definition. 158 //===----------------------------------------------------------------------===// 159 160 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &SM, SMDiagnostic &Err, 161 LLVMContext &C) 162 : CurBuf(StartBuf), ErrorInfo(Err), SM(SM), Context(C), APFloatVal(0.0), 163 IgnoreColonInIdentifiers(false) { 164 CurPtr = CurBuf.begin(); 165 } 166 167 int LLLexer::getNextChar() { 168 char CurChar = *CurPtr++; 169 switch (CurChar) { 170 default: return (unsigned char)CurChar; 171 case 0: 172 // A nul character in the stream is either the end of the current buffer or 173 // a random nul in the file. Disambiguate that here. 174 if (CurPtr-1 != CurBuf.end()) 175 return 0; // Just whitespace. 176 177 // Otherwise, return end of file. 178 --CurPtr; // Another call to lex will return EOF again. 179 return EOF; 180 } 181 } 182 183 lltok::Kind LLLexer::LexToken() { 184 while (true) { 185 TokStart = CurPtr; 186 187 int CurChar = getNextChar(); 188 switch (CurChar) { 189 default: 190 // Handle letters: [a-zA-Z_] 191 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_') 192 return LexIdentifier(); 193 194 return lltok::Error; 195 case EOF: return lltok::Eof; 196 case 0: 197 case ' ': 198 case '\t': 199 case '\n': 200 case '\r': 201 // Ignore whitespace. 202 continue; 203 case '+': return LexPositive(); 204 case '@': return LexAt(); 205 case '$': return LexDollar(); 206 case '%': return LexPercent(); 207 case '"': return LexQuote(); 208 case '.': 209 if (const char *Ptr = isLabelTail(CurPtr)) { 210 CurPtr = Ptr; 211 StrVal.assign(TokStart, CurPtr-1); 212 return lltok::LabelStr; 213 } 214 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 215 CurPtr += 2; 216 return lltok::dotdotdot; 217 } 218 return lltok::Error; 219 case ';': 220 SkipLineComment(); 221 continue; 222 case '!': return LexExclaim(); 223 case '^': 224 return LexCaret(); 225 case ':': 226 return lltok::colon; 227 case '#': return LexHash(); 228 case '0': case '1': case '2': case '3': case '4': 229 case '5': case '6': case '7': case '8': case '9': 230 case '-': 231 return LexDigitOrNegative(); 232 case '=': return lltok::equal; 233 case '[': return lltok::lsquare; 234 case ']': return lltok::rsquare; 235 case '{': return lltok::lbrace; 236 case '}': return lltok::rbrace; 237 case '<': return lltok::less; 238 case '>': return lltok::greater; 239 case '(': return lltok::lparen; 240 case ')': return lltok::rparen; 241 case ',': return lltok::comma; 242 case '*': return lltok::star; 243 case '|': return lltok::bar; 244 } 245 } 246 } 247 248 void LLLexer::SkipLineComment() { 249 while (true) { 250 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 251 return; 252 } 253 } 254 255 /// Lex all tokens that start with an @ character. 256 /// GlobalVar @\"[^\"]*\" 257 /// GlobalVar @[-a-zA-Z$._][-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 (true) { 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$._0-9]* 294 if (ReadVarName()) 295 return lltok::ComdatVar; 296 297 return lltok::Error; 298 } 299 300 /// ReadString - Read a string until the closing quote. 301 lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 302 const char *Start = CurPtr; 303 while (true) { 304 int CurChar = getNextChar(); 305 306 if (CurChar == EOF) { 307 Error("end of file in string constant"); 308 return lltok::Error; 309 } 310 if (CurChar == '"') { 311 StrVal.assign(Start, CurPtr-1); 312 UnEscapeLexed(StrVal); 313 return kind; 314 } 315 } 316 } 317 318 /// ReadVarName - Read the rest of a token containing a variable name. 319 bool LLLexer::ReadVarName() { 320 const char *NameStart = CurPtr; 321 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 322 CurPtr[0] == '-' || CurPtr[0] == '$' || 323 CurPtr[0] == '.' || CurPtr[0] == '_') { 324 ++CurPtr; 325 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 326 CurPtr[0] == '-' || CurPtr[0] == '$' || 327 CurPtr[0] == '.' || CurPtr[0] == '_') 328 ++CurPtr; 329 330 StrVal.assign(NameStart, CurPtr); 331 return true; 332 } 333 return false; 334 } 335 336 // Lex an ID: [0-9]+. On success, the ID is stored in UIntVal and Token is 337 // returned, otherwise the Error token is returned. 338 lltok::Kind LLLexer::LexUIntID(lltok::Kind Token) { 339 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 340 return lltok::Error; 341 342 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 343 /*empty*/; 344 345 uint64_t Val = atoull(TokStart + 1, CurPtr); 346 if ((unsigned)Val != Val) 347 Error("invalid value number (too large)!"); 348 UIntVal = unsigned(Val); 349 return Token; 350 } 351 352 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) { 353 // Handle StringConstant: \"[^\"]*\" 354 if (CurPtr[0] == '"') { 355 ++CurPtr; 356 357 while (true) { 358 int CurChar = getNextChar(); 359 360 if (CurChar == EOF) { 361 Error("end of file in global variable name"); 362 return lltok::Error; 363 } 364 if (CurChar == '"') { 365 StrVal.assign(TokStart+2, CurPtr-1); 366 UnEscapeLexed(StrVal); 367 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 368 Error("Null bytes are not allowed in names"); 369 return lltok::Error; 370 } 371 return Var; 372 } 373 } 374 } 375 376 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]* 377 if (ReadVarName()) 378 return Var; 379 380 // Handle VarID: [0-9]+ 381 return LexUIntID(VarID); 382 } 383 384 /// Lex all tokens that start with a % character. 385 /// LocalVar ::= %\"[^\"]*\" 386 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 387 /// LocalVarID ::= %[0-9]+ 388 lltok::Kind LLLexer::LexPercent() { 389 return LexVar(lltok::LocalVar, lltok::LocalVarID); 390 } 391 392 /// Lex all tokens that start with a " character. 393 /// QuoteLabel "[^"]+": 394 /// StringConstant "[^"]*" 395 lltok::Kind LLLexer::LexQuote() { 396 lltok::Kind kind = ReadString(lltok::StringConstant); 397 if (kind == lltok::Error || kind == lltok::Eof) 398 return kind; 399 400 if (CurPtr[0] == ':') { 401 ++CurPtr; 402 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 403 Error("Null bytes are not allowed in names"); 404 kind = lltok::Error; 405 } else { 406 kind = lltok::LabelStr; 407 } 408 } 409 410 return kind; 411 } 412 413 /// Lex all tokens that start with a ! character. 414 /// !foo 415 /// ! 416 lltok::Kind LLLexer::LexExclaim() { 417 // Lex a metadata name as a MetadataVar. 418 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 419 CurPtr[0] == '-' || CurPtr[0] == '$' || 420 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 421 ++CurPtr; 422 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 423 CurPtr[0] == '-' || CurPtr[0] == '$' || 424 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 425 ++CurPtr; 426 427 StrVal.assign(TokStart+1, CurPtr); // Skip ! 428 UnEscapeLexed(StrVal); 429 return lltok::MetadataVar; 430 } 431 return lltok::exclaim; 432 } 433 434 /// Lex all tokens that start with a ^ character. 435 /// SummaryID ::= ^[0-9]+ 436 lltok::Kind LLLexer::LexCaret() { 437 // Handle SummaryID: ^[0-9]+ 438 return LexUIntID(lltok::SummaryID); 439 } 440 441 /// Lex all tokens that start with a # character. 442 /// AttrGrpID ::= #[0-9]+ 443 lltok::Kind LLLexer::LexHash() { 444 // Handle AttrGrpID: #[0-9]+ 445 return LexUIntID(lltok::AttrGrpID); 446 } 447 448 /// Lex a label, integer type, keyword, or hexadecimal integer constant. 449 /// Label [-a-zA-Z$._0-9]+: 450 /// IntegerType i[0-9]+ 451 /// Keyword sdiv, float, ... 452 /// HexIntConstant [us]0x[0-9A-Fa-f]+ 453 lltok::Kind LLLexer::LexIdentifier() { 454 const char *StartChar = CurPtr; 455 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar; 456 const char *KeywordEnd = nullptr; 457 458 for (; isLabelChar(*CurPtr); ++CurPtr) { 459 // If we decide this is an integer, remember the end of the sequence. 460 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr))) 461 IntEnd = CurPtr; 462 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) && 463 *CurPtr != '_') 464 KeywordEnd = CurPtr; 465 } 466 467 // If we stopped due to a colon, unless we were directed to ignore it, 468 // this really is a label. 469 if (!IgnoreColonInIdentifiers && *CurPtr == ':') { 470 StrVal.assign(StartChar-1, CurPtr++); 471 return lltok::LabelStr; 472 } 473 474 // Otherwise, this wasn't a label. If this was valid as an integer type, 475 // return it. 476 if (!IntEnd) IntEnd = CurPtr; 477 if (IntEnd != StartChar) { 478 CurPtr = IntEnd; 479 uint64_t NumBits = atoull(StartChar, CurPtr); 480 if (NumBits < IntegerType::MIN_INT_BITS || 481 NumBits > IntegerType::MAX_INT_BITS) { 482 Error("bitwidth for integer type out of range!"); 483 return lltok::Error; 484 } 485 TyVal = IntegerType::get(Context, NumBits); 486 return lltok::Type; 487 } 488 489 // Otherwise, this was a letter sequence. See which keyword this is. 490 if (!KeywordEnd) KeywordEnd = CurPtr; 491 CurPtr = KeywordEnd; 492 --StartChar; 493 StringRef Keyword(StartChar, CurPtr - StartChar); 494 495 #define KEYWORD(STR) \ 496 do { \ 497 if (Keyword == #STR) \ 498 return lltok::kw_##STR; \ 499 } while (false) 500 501 KEYWORD(true); KEYWORD(false); 502 KEYWORD(declare); KEYWORD(define); 503 KEYWORD(global); KEYWORD(constant); 504 505 KEYWORD(dso_local); 506 KEYWORD(dso_preemptable); 507 508 KEYWORD(private); 509 KEYWORD(internal); 510 KEYWORD(available_externally); 511 KEYWORD(linkonce); 512 KEYWORD(linkonce_odr); 513 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg". 514 KEYWORD(weak_odr); 515 KEYWORD(appending); 516 KEYWORD(dllimport); 517 KEYWORD(dllexport); 518 KEYWORD(common); 519 KEYWORD(default); 520 KEYWORD(hidden); 521 KEYWORD(protected); 522 KEYWORD(unnamed_addr); 523 KEYWORD(local_unnamed_addr); 524 KEYWORD(externally_initialized); 525 KEYWORD(extern_weak); 526 KEYWORD(external); 527 KEYWORD(thread_local); 528 KEYWORD(localdynamic); 529 KEYWORD(initialexec); 530 KEYWORD(localexec); 531 KEYWORD(zeroinitializer); 532 KEYWORD(undef); 533 KEYWORD(null); 534 KEYWORD(none); 535 KEYWORD(to); 536 KEYWORD(caller); 537 KEYWORD(within); 538 KEYWORD(from); 539 KEYWORD(tail); 540 KEYWORD(musttail); 541 KEYWORD(notail); 542 KEYWORD(target); 543 KEYWORD(triple); 544 KEYWORD(source_filename); 545 KEYWORD(unwind); 546 KEYWORD(deplibs); // FIXME: Remove in 4.0. 547 KEYWORD(datalayout); 548 KEYWORD(volatile); 549 KEYWORD(atomic); 550 KEYWORD(unordered); 551 KEYWORD(monotonic); 552 KEYWORD(acquire); 553 KEYWORD(release); 554 KEYWORD(acq_rel); 555 KEYWORD(seq_cst); 556 KEYWORD(syncscope); 557 558 KEYWORD(nnan); 559 KEYWORD(ninf); 560 KEYWORD(nsz); 561 KEYWORD(arcp); 562 KEYWORD(contract); 563 KEYWORD(reassoc); 564 KEYWORD(afn); 565 KEYWORD(fast); 566 KEYWORD(nuw); 567 KEYWORD(nsw); 568 KEYWORD(exact); 569 KEYWORD(inbounds); 570 KEYWORD(inrange); 571 KEYWORD(align); 572 KEYWORD(addrspace); 573 KEYWORD(section); 574 KEYWORD(alias); 575 KEYWORD(ifunc); 576 KEYWORD(module); 577 KEYWORD(asm); 578 KEYWORD(sideeffect); 579 KEYWORD(alignstack); 580 KEYWORD(inteldialect); 581 KEYWORD(gc); 582 KEYWORD(prefix); 583 KEYWORD(prologue); 584 585 KEYWORD(ccc); 586 KEYWORD(fastcc); 587 KEYWORD(coldcc); 588 KEYWORD(x86_stdcallcc); 589 KEYWORD(x86_fastcallcc); 590 KEYWORD(x86_thiscallcc); 591 KEYWORD(x86_vectorcallcc); 592 KEYWORD(arm_apcscc); 593 KEYWORD(arm_aapcscc); 594 KEYWORD(arm_aapcs_vfpcc); 595 KEYWORD(msp430_intrcc); 596 KEYWORD(avr_intrcc); 597 KEYWORD(avr_signalcc); 598 KEYWORD(ptx_kernel); 599 KEYWORD(ptx_device); 600 KEYWORD(spir_kernel); 601 KEYWORD(spir_func); 602 KEYWORD(intel_ocl_bicc); 603 KEYWORD(x86_64_sysvcc); 604 KEYWORD(win64cc); 605 KEYWORD(x86_regcallcc); 606 KEYWORD(webkit_jscc); 607 KEYWORD(swiftcc); 608 KEYWORD(anyregcc); 609 KEYWORD(preserve_mostcc); 610 KEYWORD(preserve_allcc); 611 KEYWORD(ghccc); 612 KEYWORD(x86_intrcc); 613 KEYWORD(hhvmcc); 614 KEYWORD(hhvm_ccc); 615 KEYWORD(cxx_fast_tlscc); 616 KEYWORD(amdgpu_vs); 617 KEYWORD(amdgpu_ls); 618 KEYWORD(amdgpu_hs); 619 KEYWORD(amdgpu_es); 620 KEYWORD(amdgpu_gs); 621 KEYWORD(amdgpu_ps); 622 KEYWORD(amdgpu_cs); 623 KEYWORD(amdgpu_kernel); 624 625 KEYWORD(cc); 626 KEYWORD(c); 627 628 KEYWORD(attributes); 629 630 KEYWORD(alwaysinline); 631 KEYWORD(allocsize); 632 KEYWORD(argmemonly); 633 KEYWORD(builtin); 634 KEYWORD(byval); 635 KEYWORD(inalloca); 636 KEYWORD(cold); 637 KEYWORD(convergent); 638 KEYWORD(dereferenceable); 639 KEYWORD(dereferenceable_or_null); 640 KEYWORD(inaccessiblememonly); 641 KEYWORD(inaccessiblemem_or_argmemonly); 642 KEYWORD(inlinehint); 643 KEYWORD(inreg); 644 KEYWORD(jumptable); 645 KEYWORD(minsize); 646 KEYWORD(naked); 647 KEYWORD(nest); 648 KEYWORD(noalias); 649 KEYWORD(nobuiltin); 650 KEYWORD(nocapture); 651 KEYWORD(noduplicate); 652 KEYWORD(noimplicitfloat); 653 KEYWORD(noinline); 654 KEYWORD(norecurse); 655 KEYWORD(nonlazybind); 656 KEYWORD(nonnull); 657 KEYWORD(noredzone); 658 KEYWORD(noreturn); 659 KEYWORD(nocf_check); 660 KEYWORD(nounwind); 661 KEYWORD(optforfuzzing); 662 KEYWORD(optnone); 663 KEYWORD(optsize); 664 KEYWORD(readnone); 665 KEYWORD(readonly); 666 KEYWORD(returned); 667 KEYWORD(returns_twice); 668 KEYWORD(signext); 669 KEYWORD(speculatable); 670 KEYWORD(sret); 671 KEYWORD(ssp); 672 KEYWORD(sspreq); 673 KEYWORD(sspstrong); 674 KEYWORD(strictfp); 675 KEYWORD(safestack); 676 KEYWORD(shadowcallstack); 677 KEYWORD(sanitize_address); 678 KEYWORD(sanitize_hwaddress); 679 KEYWORD(sanitize_thread); 680 KEYWORD(sanitize_memory); 681 KEYWORD(swifterror); 682 KEYWORD(swiftself); 683 KEYWORD(uwtable); 684 KEYWORD(writeonly); 685 KEYWORD(zeroext); 686 687 KEYWORD(type); 688 KEYWORD(opaque); 689 690 KEYWORD(comdat); 691 692 // Comdat types 693 KEYWORD(any); 694 KEYWORD(exactmatch); 695 KEYWORD(largest); 696 KEYWORD(noduplicates); 697 KEYWORD(samesize); 698 699 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 700 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 701 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 702 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 703 704 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 705 KEYWORD(umin); 706 707 KEYWORD(x); 708 KEYWORD(blockaddress); 709 710 // Metadata types. 711 KEYWORD(distinct); 712 713 // Use-list order directives. 714 KEYWORD(uselistorder); 715 KEYWORD(uselistorder_bb); 716 717 KEYWORD(personality); 718 KEYWORD(cleanup); 719 KEYWORD(catch); 720 KEYWORD(filter); 721 722 // Summary index keywords. 723 KEYWORD(path); 724 KEYWORD(hash); 725 KEYWORD(gv); 726 KEYWORD(guid); 727 KEYWORD(name); 728 KEYWORD(summaries); 729 KEYWORD(flags); 730 KEYWORD(linkage); 731 KEYWORD(notEligibleToImport); 732 KEYWORD(live); 733 KEYWORD(dsoLocal); 734 KEYWORD(function); 735 KEYWORD(insts); 736 KEYWORD(funcFlags); 737 KEYWORD(readNone); 738 KEYWORD(readOnly); 739 KEYWORD(noRecurse); 740 KEYWORD(returnDoesNotAlias); 741 KEYWORD(calls); 742 KEYWORD(callee); 743 KEYWORD(hotness); 744 KEYWORD(unknown); 745 KEYWORD(hot); 746 KEYWORD(critical); 747 KEYWORD(relbf); 748 KEYWORD(variable); 749 KEYWORD(aliasee); 750 KEYWORD(refs); 751 KEYWORD(typeIdInfo); 752 KEYWORD(typeTests); 753 KEYWORD(typeTestAssumeVCalls); 754 KEYWORD(typeCheckedLoadVCalls); 755 KEYWORD(typeTestAssumeConstVCalls); 756 KEYWORD(typeCheckedLoadConstVCalls); 757 KEYWORD(vFuncId); 758 KEYWORD(offset); 759 KEYWORD(args); 760 KEYWORD(typeid); 761 KEYWORD(summary); 762 KEYWORD(typeTestRes); 763 KEYWORD(kind); 764 KEYWORD(unsat); 765 KEYWORD(byteArray); 766 KEYWORD(inline); 767 KEYWORD(single); 768 KEYWORD(allOnes); 769 KEYWORD(sizeM1BitWidth); 770 KEYWORD(alignLog2); 771 KEYWORD(sizeM1); 772 KEYWORD(bitMask); 773 KEYWORD(inlineBits); 774 KEYWORD(wpdResolutions); 775 KEYWORD(wpdRes); 776 KEYWORD(indir); 777 KEYWORD(singleImpl); 778 KEYWORD(branchFunnel); 779 KEYWORD(singleImplName); 780 KEYWORD(resByArg); 781 KEYWORD(byArg); 782 KEYWORD(uniformRetVal); 783 KEYWORD(uniqueRetVal); 784 KEYWORD(virtualConstProp); 785 KEYWORD(info); 786 KEYWORD(byte); 787 KEYWORD(bit); 788 789 #undef KEYWORD 790 791 // Keywords for types. 792 #define TYPEKEYWORD(STR, LLVMTY) \ 793 do { \ 794 if (Keyword == STR) { \ 795 TyVal = LLVMTY; \ 796 return lltok::Type; \ 797 } \ 798 } while (false) 799 800 TYPEKEYWORD("void", Type::getVoidTy(Context)); 801 TYPEKEYWORD("half", Type::getHalfTy(Context)); 802 TYPEKEYWORD("float", Type::getFloatTy(Context)); 803 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 804 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 805 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 806 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 807 TYPEKEYWORD("label", Type::getLabelTy(Context)); 808 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 809 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 810 TYPEKEYWORD("token", Type::getTokenTy(Context)); 811 812 #undef TYPEKEYWORD 813 814 // Keywords for instructions. 815 #define INSTKEYWORD(STR, Enum) \ 816 do { \ 817 if (Keyword == #STR) { \ 818 UIntVal = Instruction::Enum; \ 819 return lltok::kw_##STR; \ 820 } \ 821 } while (false) 822 823 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 824 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 825 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 826 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 827 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 828 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 829 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 830 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 831 832 INSTKEYWORD(phi, PHI); 833 INSTKEYWORD(call, Call); 834 INSTKEYWORD(trunc, Trunc); 835 INSTKEYWORD(zext, ZExt); 836 INSTKEYWORD(sext, SExt); 837 INSTKEYWORD(fptrunc, FPTrunc); 838 INSTKEYWORD(fpext, FPExt); 839 INSTKEYWORD(uitofp, UIToFP); 840 INSTKEYWORD(sitofp, SIToFP); 841 INSTKEYWORD(fptoui, FPToUI); 842 INSTKEYWORD(fptosi, FPToSI); 843 INSTKEYWORD(inttoptr, IntToPtr); 844 INSTKEYWORD(ptrtoint, PtrToInt); 845 INSTKEYWORD(bitcast, BitCast); 846 INSTKEYWORD(addrspacecast, AddrSpaceCast); 847 INSTKEYWORD(select, Select); 848 INSTKEYWORD(va_arg, VAArg); 849 INSTKEYWORD(ret, Ret); 850 INSTKEYWORD(br, Br); 851 INSTKEYWORD(switch, Switch); 852 INSTKEYWORD(indirectbr, IndirectBr); 853 INSTKEYWORD(invoke, Invoke); 854 INSTKEYWORD(resume, Resume); 855 INSTKEYWORD(unreachable, Unreachable); 856 857 INSTKEYWORD(alloca, Alloca); 858 INSTKEYWORD(load, Load); 859 INSTKEYWORD(store, Store); 860 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 861 INSTKEYWORD(atomicrmw, AtomicRMW); 862 INSTKEYWORD(fence, Fence); 863 INSTKEYWORD(getelementptr, GetElementPtr); 864 865 INSTKEYWORD(extractelement, ExtractElement); 866 INSTKEYWORD(insertelement, InsertElement); 867 INSTKEYWORD(shufflevector, ShuffleVector); 868 INSTKEYWORD(extractvalue, ExtractValue); 869 INSTKEYWORD(insertvalue, InsertValue); 870 INSTKEYWORD(landingpad, LandingPad); 871 INSTKEYWORD(cleanupret, CleanupRet); 872 INSTKEYWORD(catchret, CatchRet); 873 INSTKEYWORD(catchswitch, CatchSwitch); 874 INSTKEYWORD(catchpad, CatchPad); 875 INSTKEYWORD(cleanuppad, CleanupPad); 876 877 #undef INSTKEYWORD 878 879 #define DWKEYWORD(TYPE, TOKEN) \ 880 do { \ 881 if (Keyword.startswith("DW_" #TYPE "_")) { \ 882 StrVal.assign(Keyword.begin(), Keyword.end()); \ 883 return lltok::TOKEN; \ 884 } \ 885 } while (false) 886 887 DWKEYWORD(TAG, DwarfTag); 888 DWKEYWORD(ATE, DwarfAttEncoding); 889 DWKEYWORD(VIRTUALITY, DwarfVirtuality); 890 DWKEYWORD(LANG, DwarfLang); 891 DWKEYWORD(CC, DwarfCC); 892 DWKEYWORD(OP, DwarfOp); 893 DWKEYWORD(MACINFO, DwarfMacinfo); 894 895 #undef DWKEYWORD 896 897 if (Keyword.startswith("DIFlag")) { 898 StrVal.assign(Keyword.begin(), Keyword.end()); 899 return lltok::DIFlag; 900 } 901 902 if (Keyword.startswith("CSK_")) { 903 StrVal.assign(Keyword.begin(), Keyword.end()); 904 return lltok::ChecksumKind; 905 } 906 907 if (Keyword == "NoDebug" || Keyword == "FullDebug" || 908 Keyword == "LineTablesOnly" || Keyword == "DebugDirectivesOnly") { 909 StrVal.assign(Keyword.begin(), Keyword.end()); 910 return lltok::EmissionKind; 911 } 912 913 if (Keyword == "GNU" || Keyword == "None" || Keyword == "Default") { 914 StrVal.assign(Keyword.begin(), Keyword.end()); 915 return lltok::NameTableKind; 916 } 917 918 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 919 // the CFE to avoid forcing it to deal with 64-bit numbers. 920 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 921 TokStart[1] == '0' && TokStart[2] == 'x' && 922 isxdigit(static_cast<unsigned char>(TokStart[3]))) { 923 int len = CurPtr-TokStart-3; 924 uint32_t bits = len * 4; 925 StringRef HexStr(TokStart + 3, len); 926 if (!all_of(HexStr, isxdigit)) { 927 // Bad token, return it as an error. 928 CurPtr = TokStart+3; 929 return lltok::Error; 930 } 931 APInt Tmp(bits, HexStr, 16); 932 uint32_t activeBits = Tmp.getActiveBits(); 933 if (activeBits > 0 && activeBits < bits) 934 Tmp = Tmp.trunc(activeBits); 935 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 936 return lltok::APSInt; 937 } 938 939 // If this is "cc1234", return this as just "cc". 940 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 941 CurPtr = TokStart+2; 942 return lltok::kw_cc; 943 } 944 945 // Finally, if this isn't known, return an error. 946 CurPtr = TokStart+1; 947 return lltok::Error; 948 } 949 950 /// Lex all tokens that start with a 0x prefix, knowing they match and are not 951 /// labels. 952 /// HexFPConstant 0x[0-9A-Fa-f]+ 953 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 954 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 955 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 956 /// HexHalfConstant 0xH[0-9A-Fa-f]+ 957 lltok::Kind LLLexer::Lex0x() { 958 CurPtr = TokStart + 2; 959 960 char Kind; 961 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') { 962 Kind = *CurPtr++; 963 } else { 964 Kind = 'J'; 965 } 966 967 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) { 968 // Bad token, return it as an error. 969 CurPtr = TokStart+1; 970 return lltok::Error; 971 } 972 973 while (isxdigit(static_cast<unsigned char>(CurPtr[0]))) 974 ++CurPtr; 975 976 if (Kind == 'J') { 977 // HexFPConstant - Floating point constant represented in IEEE format as a 978 // hexadecimal number for when exponential notation is not precise enough. 979 // Half, Float, and double only. 980 APFloatVal = APFloat(APFloat::IEEEdouble(), 981 APInt(64, HexIntToVal(TokStart + 2, CurPtr))); 982 return lltok::APFloat; 983 } 984 985 uint64_t Pair[2]; 986 switch (Kind) { 987 default: llvm_unreachable("Unknown kind!"); 988 case 'K': 989 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 990 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 991 APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair)); 992 return lltok::APFloat; 993 case 'L': 994 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 995 HexToIntPair(TokStart+3, CurPtr, Pair); 996 APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair)); 997 return lltok::APFloat; 998 case 'M': 999 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 1000 HexToIntPair(TokStart+3, CurPtr, Pair); 1001 APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair)); 1002 return lltok::APFloat; 1003 case 'H': 1004 APFloatVal = APFloat(APFloat::IEEEhalf(), 1005 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 1006 return lltok::APFloat; 1007 } 1008 } 1009 1010 /// Lex tokens for a label or a numeric constant, possibly starting with -. 1011 /// Label [-a-zA-Z$._0-9]+: 1012 /// NInteger -[0-9]+ 1013 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 1014 /// PInteger [0-9]+ 1015 /// HexFPConstant 0x[0-9A-Fa-f]+ 1016 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 1017 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 1018 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 1019 lltok::Kind LLLexer::LexDigitOrNegative() { 1020 // If the letter after the negative is not a number, this is probably a label. 1021 if (!isdigit(static_cast<unsigned char>(TokStart[0])) && 1022 !isdigit(static_cast<unsigned char>(CurPtr[0]))) { 1023 // Okay, this is not a number after the -, it's probably a label. 1024 if (const char *End = isLabelTail(CurPtr)) { 1025 StrVal.assign(TokStart, End-1); 1026 CurPtr = End; 1027 return lltok::LabelStr; 1028 } 1029 1030 return lltok::Error; 1031 } 1032 1033 // At this point, it is either a label, int or fp constant. 1034 1035 // Skip digits, we have at least one. 1036 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 1037 /*empty*/; 1038 1039 // Check to see if this really is a label afterall, e.g. "-1:". 1040 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 1041 if (const char *End = isLabelTail(CurPtr)) { 1042 StrVal.assign(TokStart, End-1); 1043 CurPtr = End; 1044 return lltok::LabelStr; 1045 } 1046 } 1047 1048 // If the next character is a '.', then it is a fp value, otherwise its 1049 // integer. 1050 if (CurPtr[0] != '.') { 1051 if (TokStart[0] == '0' && TokStart[1] == 'x') 1052 return Lex0x(); 1053 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart)); 1054 return lltok::APSInt; 1055 } 1056 1057 ++CurPtr; 1058 1059 // Skip over [0-9]*([eE][-+]?[0-9]+)? 1060 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1061 1062 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 1063 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 1064 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 1065 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 1066 CurPtr += 2; 1067 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1068 } 1069 } 1070 1071 APFloatVal = APFloat(APFloat::IEEEdouble(), 1072 StringRef(TokStart, CurPtr - TokStart)); 1073 return lltok::APFloat; 1074 } 1075 1076 /// Lex a floating point constant starting with +. 1077 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 1078 lltok::Kind LLLexer::LexPositive() { 1079 // If the letter after the negative is a number, this is probably not a 1080 // label. 1081 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 1082 return lltok::Error; 1083 1084 // Skip digits. 1085 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 1086 /*empty*/; 1087 1088 // At this point, we need a '.'. 1089 if (CurPtr[0] != '.') { 1090 CurPtr = TokStart+1; 1091 return lltok::Error; 1092 } 1093 1094 ++CurPtr; 1095 1096 // Skip over [0-9]*([eE][-+]?[0-9]+)? 1097 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1098 1099 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 1100 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 1101 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 1102 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 1103 CurPtr += 2; 1104 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1105 } 1106 } 1107 1108 APFloatVal = APFloat(APFloat::IEEEdouble(), 1109 StringRef(TokStart, CurPtr - TokStart)); 1110 return lltok::APFloat; 1111 } 1112