1 //===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // Implement the Lexer for .ll files. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "llvm/AsmParser/LLLexer.h" 14 #include "llvm/ADT/APInt.h" 15 #include "llvm/ADT/STLExtras.h" 16 #include "llvm/ADT/StringExtras.h" 17 #include "llvm/ADT/Twine.h" 18 #include "llvm/IR/DerivedTypes.h" 19 #include "llvm/IR/Instruction.h" 20 #include "llvm/Support/ErrorHandling.h" 21 #include "llvm/Support/SourceMgr.h" 22 #include <cassert> 23 #include <cctype> 24 #include <cstdio> 25 26 using namespace llvm; 27 28 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const { 29 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg); 30 return true; 31 } 32 33 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const { 34 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg); 35 } 36 37 //===----------------------------------------------------------------------===// 38 // Helper functions. 39 //===----------------------------------------------------------------------===// 40 41 // atoull - Convert an ascii string of decimal digits into the unsigned long 42 // long representation... this does not have to do input error checking, 43 // because we know that the input will be matched by a suitable regex... 44 // 45 uint64_t LLLexer::atoull(const char *Buffer, const char *End) { 46 uint64_t Result = 0; 47 for (; Buffer != End; Buffer++) { 48 uint64_t OldRes = Result; 49 Result *= 10; 50 Result += *Buffer-'0'; 51 if (Result < OldRes) { // Uh, oh, overflow detected!!! 52 Error("constant bigger than 64 bits detected!"); 53 return 0; 54 } 55 } 56 return Result; 57 } 58 59 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) { 60 uint64_t Result = 0; 61 for (; Buffer != End; ++Buffer) { 62 uint64_t OldRes = Result; 63 Result *= 16; 64 Result += hexDigitValue(*Buffer); 65 66 if (Result < OldRes) { // Uh, oh, overflow detected!!! 67 Error("constant bigger than 64 bits detected!"); 68 return 0; 69 } 70 } 71 return Result; 72 } 73 74 void LLLexer::HexToIntPair(const char *Buffer, const char *End, 75 uint64_t Pair[2]) { 76 Pair[0] = 0; 77 if (End - Buffer >= 16) { 78 for (int i = 0; i < 16; i++, Buffer++) { 79 assert(Buffer != End); 80 Pair[0] *= 16; 81 Pair[0] += hexDigitValue(*Buffer); 82 } 83 } 84 Pair[1] = 0; 85 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 86 Pair[1] *= 16; 87 Pair[1] += hexDigitValue(*Buffer); 88 } 89 if (Buffer != End) 90 Error("constant bigger than 128 bits detected!"); 91 } 92 93 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into 94 /// { low64, high16 } as usual for an APInt. 95 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End, 96 uint64_t Pair[2]) { 97 Pair[1] = 0; 98 for (int i=0; i<4 && Buffer != End; i++, Buffer++) { 99 assert(Buffer != End); 100 Pair[1] *= 16; 101 Pair[1] += hexDigitValue(*Buffer); 102 } 103 Pair[0] = 0; 104 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 105 Pair[0] *= 16; 106 Pair[0] += hexDigitValue(*Buffer); 107 } 108 if (Buffer != End) 109 Error("constant bigger than 128 bits detected!"); 110 } 111 112 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the 113 // appropriate character. 114 static void UnEscapeLexed(std::string &Str) { 115 if (Str.empty()) return; 116 117 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size(); 118 char *BOut = Buffer; 119 for (char *BIn = Buffer; BIn != EndBuffer; ) { 120 if (BIn[0] == '\\') { 121 if (BIn < EndBuffer-1 && BIn[1] == '\\') { 122 *BOut++ = '\\'; // Two \ becomes one 123 BIn += 2; 124 } else if (BIn < EndBuffer-2 && 125 isxdigit(static_cast<unsigned char>(BIn[1])) && 126 isxdigit(static_cast<unsigned char>(BIn[2]))) { 127 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]); 128 BIn += 3; // Skip over handled chars 129 ++BOut; 130 } else { 131 *BOut++ = *BIn++; 132 } 133 } else { 134 *BOut++ = *BIn++; 135 } 136 } 137 Str.resize(BOut-Buffer); 138 } 139 140 /// isLabelChar - Return true for [-a-zA-Z$._0-9]. 141 static bool isLabelChar(char C) { 142 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' || 143 C == '.' || C == '_'; 144 } 145 146 /// isLabelTail - Return true if this pointer points to a valid end of a label. 147 static const char *isLabelTail(const char *CurPtr) { 148 while (true) { 149 if (CurPtr[0] == ':') return CurPtr+1; 150 if (!isLabelChar(CurPtr[0])) return nullptr; 151 ++CurPtr; 152 } 153 } 154 155 //===----------------------------------------------------------------------===// 156 // Lexer definition. 157 //===----------------------------------------------------------------------===// 158 159 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &SM, SMDiagnostic &Err, 160 LLVMContext &C) 161 : CurBuf(StartBuf), ErrorInfo(Err), SM(SM), Context(C), APFloatVal(0.0), 162 IgnoreColonInIdentifiers(false) { 163 CurPtr = CurBuf.begin(); 164 } 165 166 int LLLexer::getNextChar() { 167 char CurChar = *CurPtr++; 168 switch (CurChar) { 169 default: return (unsigned char)CurChar; 170 case 0: 171 // A nul character in the stream is either the end of the current buffer or 172 // a random nul in the file. Disambiguate that here. 173 if (CurPtr-1 != CurBuf.end()) 174 return 0; // Just whitespace. 175 176 // Otherwise, return end of file. 177 --CurPtr; // Another call to lex will return EOF again. 178 return EOF; 179 } 180 } 181 182 lltok::Kind LLLexer::LexToken() { 183 while (true) { 184 TokStart = CurPtr; 185 186 int CurChar = getNextChar(); 187 switch (CurChar) { 188 default: 189 // Handle letters: [a-zA-Z_] 190 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_') 191 return LexIdentifier(); 192 193 return lltok::Error; 194 case EOF: return lltok::Eof; 195 case 0: 196 case ' ': 197 case '\t': 198 case '\n': 199 case '\r': 200 // Ignore whitespace. 201 continue; 202 case '+': return LexPositive(); 203 case '@': return LexAt(); 204 case '$': return LexDollar(); 205 case '%': return LexPercent(); 206 case '"': return LexQuote(); 207 case '.': 208 if (const char *Ptr = isLabelTail(CurPtr)) { 209 CurPtr = Ptr; 210 StrVal.assign(TokStart, CurPtr-1); 211 return lltok::LabelStr; 212 } 213 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 214 CurPtr += 2; 215 return lltok::dotdotdot; 216 } 217 return lltok::Error; 218 case ';': 219 SkipLineComment(); 220 continue; 221 case '!': return LexExclaim(); 222 case '^': 223 return LexCaret(); 224 case ':': 225 return lltok::colon; 226 case '#': return LexHash(); 227 case '0': case '1': case '2': case '3': case '4': 228 case '5': case '6': case '7': case '8': case '9': 229 case '-': 230 return LexDigitOrNegative(); 231 case '=': return lltok::equal; 232 case '[': return lltok::lsquare; 233 case ']': return lltok::rsquare; 234 case '{': return lltok::lbrace; 235 case '}': return lltok::rbrace; 236 case '<': return lltok::less; 237 case '>': return lltok::greater; 238 case '(': return lltok::lparen; 239 case ')': return lltok::rparen; 240 case ',': return lltok::comma; 241 case '*': return lltok::star; 242 case '|': return lltok::bar; 243 } 244 } 245 } 246 247 void LLLexer::SkipLineComment() { 248 while (true) { 249 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 250 return; 251 } 252 } 253 254 /// Lex all tokens that start with an @ character. 255 /// GlobalVar @\"[^\"]*\" 256 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 257 /// GlobalVarID @[0-9]+ 258 lltok::Kind LLLexer::LexAt() { 259 return LexVar(lltok::GlobalVar, lltok::GlobalID); 260 } 261 262 lltok::Kind LLLexer::LexDollar() { 263 if (const char *Ptr = isLabelTail(TokStart)) { 264 CurPtr = Ptr; 265 StrVal.assign(TokStart, CurPtr - 1); 266 return lltok::LabelStr; 267 } 268 269 // Handle DollarStringConstant: $\"[^\"]*\" 270 if (CurPtr[0] == '"') { 271 ++CurPtr; 272 273 while (true) { 274 int CurChar = getNextChar(); 275 276 if (CurChar == EOF) { 277 Error("end of file in COMDAT variable name"); 278 return lltok::Error; 279 } 280 if (CurChar == '"') { 281 StrVal.assign(TokStart + 2, CurPtr - 1); 282 UnEscapeLexed(StrVal); 283 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 284 Error("Null bytes are not allowed in names"); 285 return lltok::Error; 286 } 287 return lltok::ComdatVar; 288 } 289 } 290 } 291 292 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]* 293 if (ReadVarName()) 294 return lltok::ComdatVar; 295 296 return lltok::Error; 297 } 298 299 /// ReadString - Read a string until the closing quote. 300 lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 301 const char *Start = CurPtr; 302 while (true) { 303 int CurChar = getNextChar(); 304 305 if (CurChar == EOF) { 306 Error("end of file in string constant"); 307 return lltok::Error; 308 } 309 if (CurChar == '"') { 310 StrVal.assign(Start, CurPtr-1); 311 UnEscapeLexed(StrVal); 312 return kind; 313 } 314 } 315 } 316 317 /// ReadVarName - Read the rest of a token containing a variable name. 318 bool LLLexer::ReadVarName() { 319 const char *NameStart = CurPtr; 320 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 321 CurPtr[0] == '-' || CurPtr[0] == '$' || 322 CurPtr[0] == '.' || CurPtr[0] == '_') { 323 ++CurPtr; 324 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 325 CurPtr[0] == '-' || CurPtr[0] == '$' || 326 CurPtr[0] == '.' || CurPtr[0] == '_') 327 ++CurPtr; 328 329 StrVal.assign(NameStart, CurPtr); 330 return true; 331 } 332 return false; 333 } 334 335 // Lex an ID: [0-9]+. On success, the ID is stored in UIntVal and Token is 336 // returned, otherwise the Error token is returned. 337 lltok::Kind LLLexer::LexUIntID(lltok::Kind Token) { 338 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 339 return lltok::Error; 340 341 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 342 /*empty*/; 343 344 uint64_t Val = atoull(TokStart + 1, CurPtr); 345 if ((unsigned)Val != Val) 346 Error("invalid value number (too large)!"); 347 UIntVal = unsigned(Val); 348 return Token; 349 } 350 351 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) { 352 // Handle StringConstant: \"[^\"]*\" 353 if (CurPtr[0] == '"') { 354 ++CurPtr; 355 356 while (true) { 357 int CurChar = getNextChar(); 358 359 if (CurChar == EOF) { 360 Error("end of file in global variable name"); 361 return lltok::Error; 362 } 363 if (CurChar == '"') { 364 StrVal.assign(TokStart+2, CurPtr-1); 365 UnEscapeLexed(StrVal); 366 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 367 Error("Null bytes are not allowed in names"); 368 return lltok::Error; 369 } 370 return Var; 371 } 372 } 373 } 374 375 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]* 376 if (ReadVarName()) 377 return Var; 378 379 // Handle VarID: [0-9]+ 380 return LexUIntID(VarID); 381 } 382 383 /// Lex all tokens that start with a % character. 384 /// LocalVar ::= %\"[^\"]*\" 385 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 386 /// LocalVarID ::= %[0-9]+ 387 lltok::Kind LLLexer::LexPercent() { 388 return LexVar(lltok::LocalVar, lltok::LocalVarID); 389 } 390 391 /// Lex all tokens that start with a " character. 392 /// QuoteLabel "[^"]+": 393 /// StringConstant "[^"]*" 394 lltok::Kind LLLexer::LexQuote() { 395 lltok::Kind kind = ReadString(lltok::StringConstant); 396 if (kind == lltok::Error || kind == lltok::Eof) 397 return kind; 398 399 if (CurPtr[0] == ':') { 400 ++CurPtr; 401 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 402 Error("Null bytes are not allowed in names"); 403 kind = lltok::Error; 404 } else { 405 kind = lltok::LabelStr; 406 } 407 } 408 409 return kind; 410 } 411 412 /// Lex all tokens that start with a ! character. 413 /// !foo 414 /// ! 415 lltok::Kind LLLexer::LexExclaim() { 416 // Lex a metadata name as a MetadataVar. 417 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 418 CurPtr[0] == '-' || CurPtr[0] == '$' || 419 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 420 ++CurPtr; 421 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 422 CurPtr[0] == '-' || CurPtr[0] == '$' || 423 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 424 ++CurPtr; 425 426 StrVal.assign(TokStart+1, CurPtr); // Skip ! 427 UnEscapeLexed(StrVal); 428 return lltok::MetadataVar; 429 } 430 return lltok::exclaim; 431 } 432 433 /// Lex all tokens that start with a ^ character. 434 /// SummaryID ::= ^[0-9]+ 435 lltok::Kind LLLexer::LexCaret() { 436 // Handle SummaryID: ^[0-9]+ 437 return LexUIntID(lltok::SummaryID); 438 } 439 440 /// Lex all tokens that start with a # character. 441 /// AttrGrpID ::= #[0-9]+ 442 lltok::Kind LLLexer::LexHash() { 443 // Handle AttrGrpID: #[0-9]+ 444 return LexUIntID(lltok::AttrGrpID); 445 } 446 447 /// Lex a label, integer type, keyword, or hexadecimal integer constant. 448 /// Label [-a-zA-Z$._0-9]+: 449 /// IntegerType i[0-9]+ 450 /// Keyword sdiv, float, ... 451 /// HexIntConstant [us]0x[0-9A-Fa-f]+ 452 lltok::Kind LLLexer::LexIdentifier() { 453 const char *StartChar = CurPtr; 454 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar; 455 const char *KeywordEnd = nullptr; 456 457 for (; isLabelChar(*CurPtr); ++CurPtr) { 458 // If we decide this is an integer, remember the end of the sequence. 459 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr))) 460 IntEnd = CurPtr; 461 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) && 462 *CurPtr != '_') 463 KeywordEnd = CurPtr; 464 } 465 466 // If we stopped due to a colon, unless we were directed to ignore it, 467 // this really is a label. 468 if (!IgnoreColonInIdentifiers && *CurPtr == ':') { 469 StrVal.assign(StartChar-1, CurPtr++); 470 return lltok::LabelStr; 471 } 472 473 // Otherwise, this wasn't a label. If this was valid as an integer type, 474 // return it. 475 if (!IntEnd) IntEnd = CurPtr; 476 if (IntEnd != StartChar) { 477 CurPtr = IntEnd; 478 uint64_t NumBits = atoull(StartChar, CurPtr); 479 if (NumBits < IntegerType::MIN_INT_BITS || 480 NumBits > IntegerType::MAX_INT_BITS) { 481 Error("bitwidth for integer type out of range!"); 482 return lltok::Error; 483 } 484 TyVal = IntegerType::get(Context, NumBits); 485 return lltok::Type; 486 } 487 488 // Otherwise, this was a letter sequence. See which keyword this is. 489 if (!KeywordEnd) KeywordEnd = CurPtr; 490 CurPtr = KeywordEnd; 491 --StartChar; 492 StringRef Keyword(StartChar, CurPtr - StartChar); 493 494 #define KEYWORD(STR) \ 495 do { \ 496 if (Keyword == #STR) \ 497 return lltok::kw_##STR; \ 498 } while (false) 499 500 KEYWORD(true); KEYWORD(false); 501 KEYWORD(declare); KEYWORD(define); 502 KEYWORD(global); KEYWORD(constant); 503 504 KEYWORD(dso_local); 505 KEYWORD(dso_preemptable); 506 507 KEYWORD(private); 508 KEYWORD(internal); 509 KEYWORD(available_externally); 510 KEYWORD(linkonce); 511 KEYWORD(linkonce_odr); 512 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg". 513 KEYWORD(weak_odr); 514 KEYWORD(appending); 515 KEYWORD(dllimport); 516 KEYWORD(dllexport); 517 KEYWORD(common); 518 KEYWORD(default); 519 KEYWORD(hidden); 520 KEYWORD(protected); 521 KEYWORD(unnamed_addr); 522 KEYWORD(local_unnamed_addr); 523 KEYWORD(externally_initialized); 524 KEYWORD(extern_weak); 525 KEYWORD(external); 526 KEYWORD(thread_local); 527 KEYWORD(localdynamic); 528 KEYWORD(initialexec); 529 KEYWORD(localexec); 530 KEYWORD(zeroinitializer); 531 KEYWORD(undef); 532 KEYWORD(null); 533 KEYWORD(none); 534 KEYWORD(poison); 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(datalayout); 547 KEYWORD(volatile); 548 KEYWORD(atomic); 549 KEYWORD(unordered); 550 KEYWORD(monotonic); 551 KEYWORD(acquire); 552 KEYWORD(release); 553 KEYWORD(acq_rel); 554 KEYWORD(seq_cst); 555 KEYWORD(syncscope); 556 557 KEYWORD(nnan); 558 KEYWORD(ninf); 559 KEYWORD(nsz); 560 KEYWORD(arcp); 561 KEYWORD(contract); 562 KEYWORD(reassoc); 563 KEYWORD(afn); 564 KEYWORD(fast); 565 KEYWORD(nuw); 566 KEYWORD(nsw); 567 KEYWORD(exact); 568 KEYWORD(inbounds); 569 KEYWORD(inrange); 570 KEYWORD(addrspace); 571 KEYWORD(section); 572 KEYWORD(partition); 573 KEYWORD(alias); 574 KEYWORD(ifunc); 575 KEYWORD(module); 576 KEYWORD(asm); 577 KEYWORD(sideeffect); 578 KEYWORD(inteldialect); 579 KEYWORD(gc); 580 KEYWORD(prefix); 581 KEYWORD(prologue); 582 583 KEYWORD(no_sanitize_address); 584 KEYWORD(no_sanitize_hwaddress); 585 KEYWORD(sanitize_address_dyninit); 586 587 KEYWORD(ccc); 588 KEYWORD(fastcc); 589 KEYWORD(coldcc); 590 KEYWORD(cfguard_checkcc); 591 KEYWORD(x86_stdcallcc); 592 KEYWORD(x86_fastcallcc); 593 KEYWORD(x86_thiscallcc); 594 KEYWORD(x86_vectorcallcc); 595 KEYWORD(arm_apcscc); 596 KEYWORD(arm_aapcscc); 597 KEYWORD(arm_aapcs_vfpcc); 598 KEYWORD(aarch64_vector_pcs); 599 KEYWORD(aarch64_sve_vector_pcs); 600 KEYWORD(aarch64_sme_preservemost_from_x0); 601 KEYWORD(aarch64_sme_preservemost_from_x2); 602 KEYWORD(msp430_intrcc); 603 KEYWORD(avr_intrcc); 604 KEYWORD(avr_signalcc); 605 KEYWORD(ptx_kernel); 606 KEYWORD(ptx_device); 607 KEYWORD(spir_kernel); 608 KEYWORD(spir_func); 609 KEYWORD(intel_ocl_bicc); 610 KEYWORD(x86_64_sysvcc); 611 KEYWORD(win64cc); 612 KEYWORD(x86_regcallcc); 613 KEYWORD(webkit_jscc); 614 KEYWORD(swiftcc); 615 KEYWORD(swifttailcc); 616 KEYWORD(anyregcc); 617 KEYWORD(preserve_mostcc); 618 KEYWORD(preserve_allcc); 619 KEYWORD(ghccc); 620 KEYWORD(x86_intrcc); 621 KEYWORD(hhvmcc); 622 KEYWORD(hhvm_ccc); 623 KEYWORD(cxx_fast_tlscc); 624 KEYWORD(amdgpu_vs); 625 KEYWORD(amdgpu_ls); 626 KEYWORD(amdgpu_hs); 627 KEYWORD(amdgpu_es); 628 KEYWORD(amdgpu_gs); 629 KEYWORD(amdgpu_ps); 630 KEYWORD(amdgpu_cs); 631 KEYWORD(amdgpu_kernel); 632 KEYWORD(amdgpu_gfx); 633 KEYWORD(tailcc); 634 635 KEYWORD(cc); 636 KEYWORD(c); 637 638 KEYWORD(attributes); 639 KEYWORD(sync); 640 KEYWORD(async); 641 642 #define GET_ATTR_NAMES 643 #define ATTRIBUTE_ENUM(ENUM_NAME, DISPLAY_NAME) \ 644 KEYWORD(DISPLAY_NAME); 645 #include "llvm/IR/Attributes.inc" 646 647 KEYWORD(read); 648 KEYWORD(write); 649 KEYWORD(readwrite); 650 KEYWORD(argmem); 651 KEYWORD(inaccessiblemem); 652 KEYWORD(argmemonly); 653 KEYWORD(inaccessiblememonly); 654 KEYWORD(inaccessiblemem_or_argmemonly); 655 656 KEYWORD(type); 657 KEYWORD(opaque); 658 659 KEYWORD(comdat); 660 661 // Comdat types 662 KEYWORD(any); 663 KEYWORD(exactmatch); 664 KEYWORD(largest); 665 KEYWORD(nodeduplicate); 666 KEYWORD(samesize); 667 668 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 669 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 670 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 671 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 672 673 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 674 KEYWORD(umin); KEYWORD(fmax); KEYWORD(fmin); 675 KEYWORD(uinc_wrap); 676 KEYWORD(udec_wrap); 677 678 KEYWORD(vscale); 679 KEYWORD(x); 680 KEYWORD(blockaddress); 681 KEYWORD(dso_local_equivalent); 682 KEYWORD(no_cfi); 683 684 // Metadata types. 685 KEYWORD(distinct); 686 687 // Use-list order directives. 688 KEYWORD(uselistorder); 689 KEYWORD(uselistorder_bb); 690 691 KEYWORD(personality); 692 KEYWORD(cleanup); 693 KEYWORD(catch); 694 KEYWORD(filter); 695 696 // Summary index keywords. 697 KEYWORD(path); 698 KEYWORD(hash); 699 KEYWORD(gv); 700 KEYWORD(guid); 701 KEYWORD(name); 702 KEYWORD(summaries); 703 KEYWORD(flags); 704 KEYWORD(blockcount); 705 KEYWORD(linkage); 706 KEYWORD(visibility); 707 KEYWORD(notEligibleToImport); 708 KEYWORD(live); 709 KEYWORD(dsoLocal); 710 KEYWORD(canAutoHide); 711 KEYWORD(function); 712 KEYWORD(insts); 713 KEYWORD(funcFlags); 714 KEYWORD(readNone); 715 KEYWORD(readOnly); 716 KEYWORD(noRecurse); 717 KEYWORD(returnDoesNotAlias); 718 KEYWORD(noInline); 719 KEYWORD(alwaysInline); 720 KEYWORD(noUnwind); 721 KEYWORD(mayThrow); 722 KEYWORD(hasUnknownCall); 723 KEYWORD(mustBeUnreachable); 724 KEYWORD(calls); 725 KEYWORD(callee); 726 KEYWORD(params); 727 KEYWORD(param); 728 KEYWORD(hotness); 729 KEYWORD(unknown); 730 KEYWORD(critical); 731 KEYWORD(relbf); 732 KEYWORD(variable); 733 KEYWORD(vTableFuncs); 734 KEYWORD(virtFunc); 735 KEYWORD(aliasee); 736 KEYWORD(refs); 737 KEYWORD(typeIdInfo); 738 KEYWORD(typeTests); 739 KEYWORD(typeTestAssumeVCalls); 740 KEYWORD(typeCheckedLoadVCalls); 741 KEYWORD(typeTestAssumeConstVCalls); 742 KEYWORD(typeCheckedLoadConstVCalls); 743 KEYWORD(vFuncId); 744 KEYWORD(offset); 745 KEYWORD(args); 746 KEYWORD(typeid); 747 KEYWORD(typeidCompatibleVTable); 748 KEYWORD(summary); 749 KEYWORD(typeTestRes); 750 KEYWORD(kind); 751 KEYWORD(unsat); 752 KEYWORD(byteArray); 753 KEYWORD(inline); 754 KEYWORD(single); 755 KEYWORD(allOnes); 756 KEYWORD(sizeM1BitWidth); 757 KEYWORD(alignLog2); 758 KEYWORD(sizeM1); 759 KEYWORD(bitMask); 760 KEYWORD(inlineBits); 761 KEYWORD(vcall_visibility); 762 KEYWORD(wpdResolutions); 763 KEYWORD(wpdRes); 764 KEYWORD(indir); 765 KEYWORD(singleImpl); 766 KEYWORD(branchFunnel); 767 KEYWORD(singleImplName); 768 KEYWORD(resByArg); 769 KEYWORD(byArg); 770 KEYWORD(uniformRetVal); 771 KEYWORD(uniqueRetVal); 772 KEYWORD(virtualConstProp); 773 KEYWORD(info); 774 KEYWORD(byte); 775 KEYWORD(bit); 776 KEYWORD(varFlags); 777 KEYWORD(callsites); 778 KEYWORD(clones); 779 KEYWORD(stackIds); 780 KEYWORD(allocs); 781 KEYWORD(versions); 782 KEYWORD(memProf); 783 KEYWORD(notcold); 784 KEYWORD(notcoldandcold); 785 786 #undef KEYWORD 787 788 // Keywords for types. 789 #define TYPEKEYWORD(STR, LLVMTY) \ 790 do { \ 791 if (Keyword == STR) { \ 792 TyVal = LLVMTY; \ 793 return lltok::Type; \ 794 } \ 795 } while (false) 796 797 TYPEKEYWORD("void", Type::getVoidTy(Context)); 798 TYPEKEYWORD("half", Type::getHalfTy(Context)); 799 TYPEKEYWORD("bfloat", Type::getBFloatTy(Context)); 800 TYPEKEYWORD("float", Type::getFloatTy(Context)); 801 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 802 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 803 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 804 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 805 TYPEKEYWORD("label", Type::getLabelTy(Context)); 806 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 807 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 808 TYPEKEYWORD("x86_amx", Type::getX86_AMXTy(Context)); 809 TYPEKEYWORD("token", Type::getTokenTy(Context)); 810 811 if (Keyword == "ptr") { 812 if (Context.supportsTypedPointers()) { 813 Warning("ptr type is only supported in -opaque-pointers mode"); 814 return lltok::Error; 815 } 816 TyVal = PointerType::getUnqual(Context); 817 return lltok::Type; 818 } 819 820 #undef TYPEKEYWORD 821 822 // Keywords for instructions. 823 #define INSTKEYWORD(STR, Enum) \ 824 do { \ 825 if (Keyword == #STR) { \ 826 UIntVal = Instruction::Enum; \ 827 return lltok::kw_##STR; \ 828 } \ 829 } while (false) 830 831 INSTKEYWORD(fneg, FNeg); 832 833 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 834 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 835 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 836 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 837 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 838 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 839 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 840 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 841 842 INSTKEYWORD(phi, PHI); 843 INSTKEYWORD(call, Call); 844 INSTKEYWORD(trunc, Trunc); 845 INSTKEYWORD(zext, ZExt); 846 INSTKEYWORD(sext, SExt); 847 INSTKEYWORD(fptrunc, FPTrunc); 848 INSTKEYWORD(fpext, FPExt); 849 INSTKEYWORD(uitofp, UIToFP); 850 INSTKEYWORD(sitofp, SIToFP); 851 INSTKEYWORD(fptoui, FPToUI); 852 INSTKEYWORD(fptosi, FPToSI); 853 INSTKEYWORD(inttoptr, IntToPtr); 854 INSTKEYWORD(ptrtoint, PtrToInt); 855 INSTKEYWORD(bitcast, BitCast); 856 INSTKEYWORD(addrspacecast, AddrSpaceCast); 857 INSTKEYWORD(select, Select); 858 INSTKEYWORD(va_arg, VAArg); 859 INSTKEYWORD(ret, Ret); 860 INSTKEYWORD(br, Br); 861 INSTKEYWORD(switch, Switch); 862 INSTKEYWORD(indirectbr, IndirectBr); 863 INSTKEYWORD(invoke, Invoke); 864 INSTKEYWORD(resume, Resume); 865 INSTKEYWORD(unreachable, Unreachable); 866 INSTKEYWORD(callbr, CallBr); 867 868 INSTKEYWORD(alloca, Alloca); 869 INSTKEYWORD(load, Load); 870 INSTKEYWORD(store, Store); 871 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 872 INSTKEYWORD(atomicrmw, AtomicRMW); 873 INSTKEYWORD(fence, Fence); 874 INSTKEYWORD(getelementptr, GetElementPtr); 875 876 INSTKEYWORD(extractelement, ExtractElement); 877 INSTKEYWORD(insertelement, InsertElement); 878 INSTKEYWORD(shufflevector, ShuffleVector); 879 INSTKEYWORD(extractvalue, ExtractValue); 880 INSTKEYWORD(insertvalue, InsertValue); 881 INSTKEYWORD(landingpad, LandingPad); 882 INSTKEYWORD(cleanupret, CleanupRet); 883 INSTKEYWORD(catchret, CatchRet); 884 INSTKEYWORD(catchswitch, CatchSwitch); 885 INSTKEYWORD(catchpad, CatchPad); 886 INSTKEYWORD(cleanuppad, CleanupPad); 887 888 INSTKEYWORD(freeze, Freeze); 889 890 #undef INSTKEYWORD 891 892 #define DWKEYWORD(TYPE, TOKEN) \ 893 do { \ 894 if (Keyword.startswith("DW_" #TYPE "_")) { \ 895 StrVal.assign(Keyword.begin(), Keyword.end()); \ 896 return lltok::TOKEN; \ 897 } \ 898 } while (false) 899 900 DWKEYWORD(TAG, DwarfTag); 901 DWKEYWORD(ATE, DwarfAttEncoding); 902 DWKEYWORD(VIRTUALITY, DwarfVirtuality); 903 DWKEYWORD(LANG, DwarfLang); 904 DWKEYWORD(CC, DwarfCC); 905 DWKEYWORD(OP, DwarfOp); 906 DWKEYWORD(MACINFO, DwarfMacinfo); 907 908 #undef DWKEYWORD 909 910 if (Keyword.startswith("DIFlag")) { 911 StrVal.assign(Keyword.begin(), Keyword.end()); 912 return lltok::DIFlag; 913 } 914 915 if (Keyword.startswith("DISPFlag")) { 916 StrVal.assign(Keyword.begin(), Keyword.end()); 917 return lltok::DISPFlag; 918 } 919 920 if (Keyword.startswith("CSK_")) { 921 StrVal.assign(Keyword.begin(), Keyword.end()); 922 return lltok::ChecksumKind; 923 } 924 925 if (Keyword == "NoDebug" || Keyword == "FullDebug" || 926 Keyword == "LineTablesOnly" || Keyword == "DebugDirectivesOnly") { 927 StrVal.assign(Keyword.begin(), Keyword.end()); 928 return lltok::EmissionKind; 929 } 930 931 if (Keyword == "GNU" || Keyword == "None" || Keyword == "Default") { 932 StrVal.assign(Keyword.begin(), Keyword.end()); 933 return lltok::NameTableKind; 934 } 935 936 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 937 // the CFE to avoid forcing it to deal with 64-bit numbers. 938 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 939 TokStart[1] == '0' && TokStart[2] == 'x' && 940 isxdigit(static_cast<unsigned char>(TokStart[3]))) { 941 int len = CurPtr-TokStart-3; 942 uint32_t bits = len * 4; 943 StringRef HexStr(TokStart + 3, len); 944 if (!all_of(HexStr, isxdigit)) { 945 // Bad token, return it as an error. 946 CurPtr = TokStart+3; 947 return lltok::Error; 948 } 949 APInt Tmp(bits, HexStr, 16); 950 uint32_t activeBits = Tmp.getActiveBits(); 951 if (activeBits > 0 && activeBits < bits) 952 Tmp = Tmp.trunc(activeBits); 953 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 954 return lltok::APSInt; 955 } 956 957 // If this is "cc1234", return this as just "cc". 958 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 959 CurPtr = TokStart+2; 960 return lltok::kw_cc; 961 } 962 963 // Finally, if this isn't known, return an error. 964 CurPtr = TokStart+1; 965 return lltok::Error; 966 } 967 968 /// Lex all tokens that start with a 0x prefix, knowing they match and are not 969 /// labels. 970 /// HexFPConstant 0x[0-9A-Fa-f]+ 971 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 972 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 973 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 974 /// HexHalfConstant 0xH[0-9A-Fa-f]+ 975 /// HexBFloatConstant 0xR[0-9A-Fa-f]+ 976 lltok::Kind LLLexer::Lex0x() { 977 CurPtr = TokStart + 2; 978 979 char Kind; 980 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H' || 981 CurPtr[0] == 'R') { 982 Kind = *CurPtr++; 983 } else { 984 Kind = 'J'; 985 } 986 987 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) { 988 // Bad token, return it as an error. 989 CurPtr = TokStart+1; 990 return lltok::Error; 991 } 992 993 while (isxdigit(static_cast<unsigned char>(CurPtr[0]))) 994 ++CurPtr; 995 996 if (Kind == 'J') { 997 // HexFPConstant - Floating point constant represented in IEEE format as a 998 // hexadecimal number for when exponential notation is not precise enough. 999 // Half, BFloat, Float, and double only. 1000 APFloatVal = APFloat(APFloat::IEEEdouble(), 1001 APInt(64, HexIntToVal(TokStart + 2, CurPtr))); 1002 return lltok::APFloat; 1003 } 1004 1005 uint64_t Pair[2]; 1006 switch (Kind) { 1007 default: llvm_unreachable("Unknown kind!"); 1008 case 'K': 1009 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 1010 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 1011 APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair)); 1012 return lltok::APFloat; 1013 case 'L': 1014 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 1015 HexToIntPair(TokStart+3, CurPtr, Pair); 1016 APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair)); 1017 return lltok::APFloat; 1018 case 'M': 1019 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 1020 HexToIntPair(TokStart+3, CurPtr, Pair); 1021 APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair)); 1022 return lltok::APFloat; 1023 case 'H': 1024 APFloatVal = APFloat(APFloat::IEEEhalf(), 1025 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 1026 return lltok::APFloat; 1027 case 'R': 1028 // Brain floating point 1029 APFloatVal = APFloat(APFloat::BFloat(), 1030 APInt(16, HexIntToVal(TokStart + 3, CurPtr))); 1031 return lltok::APFloat; 1032 } 1033 } 1034 1035 /// Lex tokens for a label or a numeric constant, possibly starting with -. 1036 /// Label [-a-zA-Z$._0-9]+: 1037 /// NInteger -[0-9]+ 1038 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 1039 /// PInteger [0-9]+ 1040 /// HexFPConstant 0x[0-9A-Fa-f]+ 1041 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 1042 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 1043 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 1044 lltok::Kind LLLexer::LexDigitOrNegative() { 1045 // If the letter after the negative is not a number, this is probably a label. 1046 if (!isdigit(static_cast<unsigned char>(TokStart[0])) && 1047 !isdigit(static_cast<unsigned char>(CurPtr[0]))) { 1048 // Okay, this is not a number after the -, it's probably a label. 1049 if (const char *End = isLabelTail(CurPtr)) { 1050 StrVal.assign(TokStart, End-1); 1051 CurPtr = End; 1052 return lltok::LabelStr; 1053 } 1054 1055 return lltok::Error; 1056 } 1057 1058 // At this point, it is either a label, int or fp constant. 1059 1060 // Skip digits, we have at least one. 1061 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 1062 /*empty*/; 1063 1064 // Check if this is a fully-numeric label: 1065 if (isdigit(TokStart[0]) && CurPtr[0] == ':') { 1066 uint64_t Val = atoull(TokStart, CurPtr); 1067 ++CurPtr; // Skip the colon. 1068 if ((unsigned)Val != Val) 1069 Error("invalid value number (too large)!"); 1070 UIntVal = unsigned(Val); 1071 return lltok::LabelID; 1072 } 1073 1074 // Check to see if this really is a string label, e.g. "-1:". 1075 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 1076 if (const char *End = isLabelTail(CurPtr)) { 1077 StrVal.assign(TokStart, End-1); 1078 CurPtr = End; 1079 return lltok::LabelStr; 1080 } 1081 } 1082 1083 // If the next character is a '.', then it is a fp value, otherwise its 1084 // integer. 1085 if (CurPtr[0] != '.') { 1086 if (TokStart[0] == '0' && TokStart[1] == 'x') 1087 return Lex0x(); 1088 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart)); 1089 return lltok::APSInt; 1090 } 1091 1092 ++CurPtr; 1093 1094 // Skip over [0-9]*([eE][-+]?[0-9]+)? 1095 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1096 1097 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 1098 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 1099 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 1100 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 1101 CurPtr += 2; 1102 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1103 } 1104 } 1105 1106 APFloatVal = APFloat(APFloat::IEEEdouble(), 1107 StringRef(TokStart, CurPtr - TokStart)); 1108 return lltok::APFloat; 1109 } 1110 1111 /// Lex a floating point constant starting with +. 1112 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 1113 lltok::Kind LLLexer::LexPositive() { 1114 // If the letter after the negative is a number, this is probably not a 1115 // label. 1116 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 1117 return lltok::Error; 1118 1119 // Skip digits. 1120 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 1121 /*empty*/; 1122 1123 // At this point, we need a '.'. 1124 if (CurPtr[0] != '.') { 1125 CurPtr = TokStart+1; 1126 return lltok::Error; 1127 } 1128 1129 ++CurPtr; 1130 1131 // Skip over [0-9]*([eE][-+]?[0-9]+)? 1132 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1133 1134 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 1135 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 1136 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 1137 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 1138 CurPtr += 2; 1139 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1140 } 1141 } 1142 1143 APFloatVal = APFloat(APFloat::IEEEdouble(), 1144 StringRef(TokStart, CurPtr - TokStart)); 1145 return lltok::APFloat; 1146 } 1147