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 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 '$': return LexDollar(); 205 case '%': return LexPercent(); 206 case '"': return LexQuote(); 207 case '.': 208 if (const char *Ptr = isLabelTail(CurPtr)) { 209 CurPtr = Ptr; 210 StrVal.assign(TokStart, CurPtr-1); 211 return lltok::LabelStr; 212 } 213 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 214 CurPtr += 2; 215 return lltok::dotdotdot; 216 } 217 return lltok::Error; 218 case ';': 219 SkipLineComment(); 220 continue; 221 case '!': return LexExclaim(); 222 case '#': return LexHash(); 223 case '0': case '1': case '2': case '3': case '4': 224 case '5': case '6': case '7': case '8': case '9': 225 case '-': 226 return LexDigitOrNegative(); 227 case '=': return lltok::equal; 228 case '[': return lltok::lsquare; 229 case ']': return lltok::rsquare; 230 case '{': return lltok::lbrace; 231 case '}': return lltok::rbrace; 232 case '<': return lltok::less; 233 case '>': return lltok::greater; 234 case '(': return lltok::lparen; 235 case ')': return lltok::rparen; 236 case ',': return lltok::comma; 237 case '*': return lltok::star; 238 case '|': return lltok::bar; 239 } 240 } 241 } 242 243 void LLLexer::SkipLineComment() { 244 while (true) { 245 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 246 return; 247 } 248 } 249 250 /// Lex all tokens that start with an @ character. 251 /// GlobalVar @\"[^\"]*\" 252 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 253 /// GlobalVarID @[0-9]+ 254 lltok::Kind LLLexer::LexAt() { 255 return LexVar(lltok::GlobalVar, lltok::GlobalID); 256 } 257 258 lltok::Kind LLLexer::LexDollar() { 259 if (const char *Ptr = isLabelTail(TokStart)) { 260 CurPtr = Ptr; 261 StrVal.assign(TokStart, CurPtr - 1); 262 return lltok::LabelStr; 263 } 264 265 // Handle DollarStringConstant: $\"[^\"]*\" 266 if (CurPtr[0] == '"') { 267 ++CurPtr; 268 269 while (true) { 270 int CurChar = getNextChar(); 271 272 if (CurChar == EOF) { 273 Error("end of file in COMDAT variable name"); 274 return lltok::Error; 275 } 276 if (CurChar == '"') { 277 StrVal.assign(TokStart + 2, CurPtr - 1); 278 UnEscapeLexed(StrVal); 279 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 280 Error("Null bytes are not allowed in names"); 281 return lltok::Error; 282 } 283 return lltok::ComdatVar; 284 } 285 } 286 } 287 288 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]* 289 if (ReadVarName()) 290 return lltok::ComdatVar; 291 292 return lltok::Error; 293 } 294 295 /// ReadString - Read a string until the closing quote. 296 lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 297 const char *Start = CurPtr; 298 while (true) { 299 int CurChar = getNextChar(); 300 301 if (CurChar == EOF) { 302 Error("end of file in string constant"); 303 return lltok::Error; 304 } 305 if (CurChar == '"') { 306 StrVal.assign(Start, CurPtr-1); 307 UnEscapeLexed(StrVal); 308 return kind; 309 } 310 } 311 } 312 313 /// ReadVarName - Read the rest of a token containing a variable name. 314 bool LLLexer::ReadVarName() { 315 const char *NameStart = CurPtr; 316 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 317 CurPtr[0] == '-' || CurPtr[0] == '$' || 318 CurPtr[0] == '.' || CurPtr[0] == '_') { 319 ++CurPtr; 320 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 321 CurPtr[0] == '-' || CurPtr[0] == '$' || 322 CurPtr[0] == '.' || CurPtr[0] == '_') 323 ++CurPtr; 324 325 StrVal.assign(NameStart, CurPtr); 326 return true; 327 } 328 return false; 329 } 330 331 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) { 332 // Handle StringConstant: \"[^\"]*\" 333 if (CurPtr[0] == '"') { 334 ++CurPtr; 335 336 while (true) { 337 int CurChar = getNextChar(); 338 339 if (CurChar == EOF) { 340 Error("end of file in global variable name"); 341 return lltok::Error; 342 } 343 if (CurChar == '"') { 344 StrVal.assign(TokStart+2, CurPtr-1); 345 UnEscapeLexed(StrVal); 346 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 347 Error("Null bytes are not allowed in names"); 348 return lltok::Error; 349 } 350 return Var; 351 } 352 } 353 } 354 355 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]* 356 if (ReadVarName()) 357 return Var; 358 359 // Handle VarID: [0-9]+ 360 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 361 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 362 /*empty*/; 363 364 uint64_t Val = atoull(TokStart+1, CurPtr); 365 if ((unsigned)Val != Val) 366 Error("invalid value number (too large)!"); 367 UIntVal = unsigned(Val); 368 return VarID; 369 } 370 return lltok::Error; 371 } 372 373 /// Lex all tokens that start with a % character. 374 /// LocalVar ::= %\"[^\"]*\" 375 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 376 /// LocalVarID ::= %[0-9]+ 377 lltok::Kind LLLexer::LexPercent() { 378 return LexVar(lltok::LocalVar, lltok::LocalVarID); 379 } 380 381 /// Lex all tokens that start with a " character. 382 /// QuoteLabel "[^"]+": 383 /// StringConstant "[^"]*" 384 lltok::Kind LLLexer::LexQuote() { 385 lltok::Kind kind = ReadString(lltok::StringConstant); 386 if (kind == lltok::Error || kind == lltok::Eof) 387 return kind; 388 389 if (CurPtr[0] == ':') { 390 ++CurPtr; 391 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 392 Error("Null bytes are not allowed in names"); 393 kind = lltok::Error; 394 } else { 395 kind = lltok::LabelStr; 396 } 397 } 398 399 return kind; 400 } 401 402 /// Lex all tokens that start with a ! character. 403 /// !foo 404 /// ! 405 lltok::Kind LLLexer::LexExclaim() { 406 // Lex a metadata name as a MetadataVar. 407 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 408 CurPtr[0] == '-' || CurPtr[0] == '$' || 409 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 410 ++CurPtr; 411 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 412 CurPtr[0] == '-' || CurPtr[0] == '$' || 413 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 414 ++CurPtr; 415 416 StrVal.assign(TokStart+1, CurPtr); // Skip ! 417 UnEscapeLexed(StrVal); 418 return lltok::MetadataVar; 419 } 420 return lltok::exclaim; 421 } 422 423 /// Lex all tokens that start with a # character. 424 /// AttrGrpID ::= #[0-9]+ 425 lltok::Kind LLLexer::LexHash() { 426 // Handle AttrGrpID: #[0-9]+ 427 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 428 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 429 /*empty*/; 430 431 uint64_t Val = atoull(TokStart+1, CurPtr); 432 if ((unsigned)Val != Val) 433 Error("invalid value number (too large)!"); 434 UIntVal = unsigned(Val); 435 return lltok::AttrGrpID; 436 } 437 438 return lltok::Error; 439 } 440 441 /// Lex a label, integer type, keyword, or hexadecimal integer constant. 442 /// Label [-a-zA-Z$._0-9]+: 443 /// IntegerType i[0-9]+ 444 /// Keyword sdiv, float, ... 445 /// HexIntConstant [us]0x[0-9A-Fa-f]+ 446 lltok::Kind LLLexer::LexIdentifier() { 447 const char *StartChar = CurPtr; 448 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar; 449 const char *KeywordEnd = nullptr; 450 451 for (; isLabelChar(*CurPtr); ++CurPtr) { 452 // If we decide this is an integer, remember the end of the sequence. 453 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr))) 454 IntEnd = CurPtr; 455 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) && 456 *CurPtr != '_') 457 KeywordEnd = CurPtr; 458 } 459 460 // If we stopped due to a colon, this really is a label. 461 if (*CurPtr == ':') { 462 StrVal.assign(StartChar-1, CurPtr++); 463 return lltok::LabelStr; 464 } 465 466 // Otherwise, this wasn't a label. If this was valid as an integer type, 467 // return it. 468 if (!IntEnd) IntEnd = CurPtr; 469 if (IntEnd != StartChar) { 470 CurPtr = IntEnd; 471 uint64_t NumBits = atoull(StartChar, CurPtr); 472 if (NumBits < IntegerType::MIN_INT_BITS || 473 NumBits > IntegerType::MAX_INT_BITS) { 474 Error("bitwidth for integer type out of range!"); 475 return lltok::Error; 476 } 477 TyVal = IntegerType::get(Context, NumBits); 478 return lltok::Type; 479 } 480 481 // Otherwise, this was a letter sequence. See which keyword this is. 482 if (!KeywordEnd) KeywordEnd = CurPtr; 483 CurPtr = KeywordEnd; 484 --StartChar; 485 StringRef Keyword(StartChar, CurPtr - StartChar); 486 487 #define KEYWORD(STR) \ 488 do { \ 489 if (Keyword == #STR) \ 490 return lltok::kw_##STR; \ 491 } while (false) 492 493 KEYWORD(true); KEYWORD(false); 494 KEYWORD(declare); KEYWORD(define); 495 KEYWORD(global); KEYWORD(constant); 496 497 KEYWORD(dso_local); 498 KEYWORD(dso_preemptable); 499 500 KEYWORD(private); 501 KEYWORD(internal); 502 KEYWORD(available_externally); 503 KEYWORD(linkonce); 504 KEYWORD(linkonce_odr); 505 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg". 506 KEYWORD(weak_odr); 507 KEYWORD(appending); 508 KEYWORD(dllimport); 509 KEYWORD(dllexport); 510 KEYWORD(common); 511 KEYWORD(default); 512 KEYWORD(hidden); 513 KEYWORD(protected); 514 KEYWORD(unnamed_addr); 515 KEYWORD(local_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(source_filename); 537 KEYWORD(unwind); 538 KEYWORD(deplibs); // FIXME: Remove in 4.0. 539 KEYWORD(datalayout); 540 KEYWORD(volatile); 541 KEYWORD(atomic); 542 KEYWORD(unordered); 543 KEYWORD(monotonic); 544 KEYWORD(acquire); 545 KEYWORD(release); 546 KEYWORD(acq_rel); 547 KEYWORD(seq_cst); 548 KEYWORD(syncscope); 549 550 KEYWORD(nnan); 551 KEYWORD(ninf); 552 KEYWORD(nsz); 553 KEYWORD(arcp); 554 KEYWORD(contract); 555 KEYWORD(reassoc); 556 KEYWORD(afn); 557 KEYWORD(fast); 558 KEYWORD(nuw); 559 KEYWORD(nsw); 560 KEYWORD(exact); 561 KEYWORD(inbounds); 562 KEYWORD(inrange); 563 KEYWORD(align); 564 KEYWORD(addrspace); 565 KEYWORD(section); 566 KEYWORD(alias); 567 KEYWORD(ifunc); 568 KEYWORD(module); 569 KEYWORD(asm); 570 KEYWORD(sideeffect); 571 KEYWORD(alignstack); 572 KEYWORD(inteldialect); 573 KEYWORD(gc); 574 KEYWORD(prefix); 575 KEYWORD(prologue); 576 577 KEYWORD(ccc); 578 KEYWORD(fastcc); 579 KEYWORD(coldcc); 580 KEYWORD(x86_stdcallcc); 581 KEYWORD(x86_fastcallcc); 582 KEYWORD(x86_thiscallcc); 583 KEYWORD(x86_vectorcallcc); 584 KEYWORD(arm_apcscc); 585 KEYWORD(arm_aapcscc); 586 KEYWORD(arm_aapcs_vfpcc); 587 KEYWORD(msp430_intrcc); 588 KEYWORD(avr_intrcc); 589 KEYWORD(avr_signalcc); 590 KEYWORD(ptx_kernel); 591 KEYWORD(ptx_device); 592 KEYWORD(spir_kernel); 593 KEYWORD(spir_func); 594 KEYWORD(intel_ocl_bicc); 595 KEYWORD(x86_64_sysvcc); 596 KEYWORD(win64cc); 597 KEYWORD(x86_regcallcc); 598 KEYWORD(webkit_jscc); 599 KEYWORD(swiftcc); 600 KEYWORD(anyregcc); 601 KEYWORD(preserve_mostcc); 602 KEYWORD(preserve_allcc); 603 KEYWORD(ghccc); 604 KEYWORD(x86_intrcc); 605 KEYWORD(hhvmcc); 606 KEYWORD(hhvm_ccc); 607 KEYWORD(cxx_fast_tlscc); 608 KEYWORD(amdgpu_vs); 609 KEYWORD(amdgpu_ls); 610 KEYWORD(amdgpu_hs); 611 KEYWORD(amdgpu_es); 612 KEYWORD(amdgpu_gs); 613 KEYWORD(amdgpu_ps); 614 KEYWORD(amdgpu_cs); 615 KEYWORD(amdgpu_kernel); 616 617 KEYWORD(cc); 618 KEYWORD(c); 619 620 KEYWORD(attributes); 621 622 KEYWORD(alwaysinline); 623 KEYWORD(allocsize); 624 KEYWORD(argmemonly); 625 KEYWORD(builtin); 626 KEYWORD(byval); 627 KEYWORD(inalloca); 628 KEYWORD(cold); 629 KEYWORD(convergent); 630 KEYWORD(dereferenceable); 631 KEYWORD(dereferenceable_or_null); 632 KEYWORD(inaccessiblememonly); 633 KEYWORD(inaccessiblemem_or_argmemonly); 634 KEYWORD(inlinehint); 635 KEYWORD(inreg); 636 KEYWORD(jumptable); 637 KEYWORD(minsize); 638 KEYWORD(naked); 639 KEYWORD(nest); 640 KEYWORD(noalias); 641 KEYWORD(nobuiltin); 642 KEYWORD(nocapture); 643 KEYWORD(noduplicate); 644 KEYWORD(noimplicitfloat); 645 KEYWORD(noinline); 646 KEYWORD(norecurse); 647 KEYWORD(nonlazybind); 648 KEYWORD(nonnull); 649 KEYWORD(noredzone); 650 KEYWORD(noreturn); 651 KEYWORD(nocf_check); 652 KEYWORD(nounwind); 653 KEYWORD(optforfuzzing); 654 KEYWORD(optnone); 655 KEYWORD(optsize); 656 KEYWORD(readnone); 657 KEYWORD(readonly); 658 KEYWORD(returned); 659 KEYWORD(returns_twice); 660 KEYWORD(signext); 661 KEYWORD(speculatable); 662 KEYWORD(sret); 663 KEYWORD(ssp); 664 KEYWORD(sspreq); 665 KEYWORD(sspstrong); 666 KEYWORD(strictfp); 667 KEYWORD(safestack); 668 KEYWORD(shadowcallstack); 669 KEYWORD(sanitize_address); 670 KEYWORD(sanitize_hwaddress); 671 KEYWORD(sanitize_thread); 672 KEYWORD(sanitize_memory); 673 KEYWORD(swifterror); 674 KEYWORD(swiftself); 675 KEYWORD(uwtable); 676 KEYWORD(writeonly); 677 KEYWORD(zeroext); 678 679 KEYWORD(type); 680 KEYWORD(opaque); 681 682 KEYWORD(comdat); 683 684 // Comdat types 685 KEYWORD(any); 686 KEYWORD(exactmatch); 687 KEYWORD(largest); 688 KEYWORD(noduplicates); 689 KEYWORD(samesize); 690 691 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 692 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 693 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 694 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 695 696 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 697 KEYWORD(umin); 698 699 KEYWORD(x); 700 KEYWORD(blockaddress); 701 702 // Metadata types. 703 KEYWORD(distinct); 704 705 // Use-list order directives. 706 KEYWORD(uselistorder); 707 KEYWORD(uselistorder_bb); 708 709 KEYWORD(personality); 710 KEYWORD(cleanup); 711 KEYWORD(catch); 712 KEYWORD(filter); 713 714 #undef KEYWORD 715 716 // Keywords for types. 717 #define TYPEKEYWORD(STR, LLVMTY) \ 718 do { \ 719 if (Keyword == STR) { \ 720 TyVal = LLVMTY; \ 721 return lltok::Type; \ 722 } \ 723 } while (false) 724 725 TYPEKEYWORD("void", Type::getVoidTy(Context)); 726 TYPEKEYWORD("half", Type::getHalfTy(Context)); 727 TYPEKEYWORD("float", Type::getFloatTy(Context)); 728 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 729 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 730 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 731 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 732 TYPEKEYWORD("label", Type::getLabelTy(Context)); 733 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 734 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 735 TYPEKEYWORD("token", Type::getTokenTy(Context)); 736 737 #undef TYPEKEYWORD 738 739 // Keywords for instructions. 740 #define INSTKEYWORD(STR, Enum) \ 741 do { \ 742 if (Keyword == #STR) { \ 743 UIntVal = Instruction::Enum; \ 744 return lltok::kw_##STR; \ 745 } \ 746 } while (false) 747 748 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 749 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 750 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 751 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 752 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 753 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 754 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 755 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 756 757 INSTKEYWORD(phi, PHI); 758 INSTKEYWORD(call, Call); 759 INSTKEYWORD(trunc, Trunc); 760 INSTKEYWORD(zext, ZExt); 761 INSTKEYWORD(sext, SExt); 762 INSTKEYWORD(fptrunc, FPTrunc); 763 INSTKEYWORD(fpext, FPExt); 764 INSTKEYWORD(uitofp, UIToFP); 765 INSTKEYWORD(sitofp, SIToFP); 766 INSTKEYWORD(fptoui, FPToUI); 767 INSTKEYWORD(fptosi, FPToSI); 768 INSTKEYWORD(inttoptr, IntToPtr); 769 INSTKEYWORD(ptrtoint, PtrToInt); 770 INSTKEYWORD(bitcast, BitCast); 771 INSTKEYWORD(addrspacecast, AddrSpaceCast); 772 INSTKEYWORD(select, Select); 773 INSTKEYWORD(va_arg, VAArg); 774 INSTKEYWORD(ret, Ret); 775 INSTKEYWORD(br, Br); 776 INSTKEYWORD(switch, Switch); 777 INSTKEYWORD(indirectbr, IndirectBr); 778 INSTKEYWORD(invoke, Invoke); 779 INSTKEYWORD(resume, Resume); 780 INSTKEYWORD(unreachable, Unreachable); 781 782 INSTKEYWORD(alloca, Alloca); 783 INSTKEYWORD(load, Load); 784 INSTKEYWORD(store, Store); 785 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 786 INSTKEYWORD(atomicrmw, AtomicRMW); 787 INSTKEYWORD(fence, Fence); 788 INSTKEYWORD(getelementptr, GetElementPtr); 789 790 INSTKEYWORD(extractelement, ExtractElement); 791 INSTKEYWORD(insertelement, InsertElement); 792 INSTKEYWORD(shufflevector, ShuffleVector); 793 INSTKEYWORD(extractvalue, ExtractValue); 794 INSTKEYWORD(insertvalue, InsertValue); 795 INSTKEYWORD(landingpad, LandingPad); 796 INSTKEYWORD(cleanupret, CleanupRet); 797 INSTKEYWORD(catchret, CatchRet); 798 INSTKEYWORD(catchswitch, CatchSwitch); 799 INSTKEYWORD(catchpad, CatchPad); 800 INSTKEYWORD(cleanuppad, CleanupPad); 801 802 #undef INSTKEYWORD 803 804 #define DWKEYWORD(TYPE, TOKEN) \ 805 do { \ 806 if (Keyword.startswith("DW_" #TYPE "_")) { \ 807 StrVal.assign(Keyword.begin(), Keyword.end()); \ 808 return lltok::TOKEN; \ 809 } \ 810 } while (false) 811 812 DWKEYWORD(TAG, DwarfTag); 813 DWKEYWORD(ATE, DwarfAttEncoding); 814 DWKEYWORD(VIRTUALITY, DwarfVirtuality); 815 DWKEYWORD(LANG, DwarfLang); 816 DWKEYWORD(CC, DwarfCC); 817 DWKEYWORD(OP, DwarfOp); 818 DWKEYWORD(MACINFO, DwarfMacinfo); 819 820 #undef DWKEYWORD 821 822 if (Keyword.startswith("DIFlag")) { 823 StrVal.assign(Keyword.begin(), Keyword.end()); 824 return lltok::DIFlag; 825 } 826 827 if (Keyword.startswith("CSK_")) { 828 StrVal.assign(Keyword.begin(), Keyword.end()); 829 return lltok::ChecksumKind; 830 } 831 832 if (Keyword == "NoDebug" || Keyword == "FullDebug" || 833 Keyword == "LineTablesOnly") { 834 StrVal.assign(Keyword.begin(), Keyword.end()); 835 return lltok::EmissionKind; 836 } 837 838 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 839 // the CFE to avoid forcing it to deal with 64-bit numbers. 840 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 841 TokStart[1] == '0' && TokStart[2] == 'x' && 842 isxdigit(static_cast<unsigned char>(TokStart[3]))) { 843 int len = CurPtr-TokStart-3; 844 uint32_t bits = len * 4; 845 StringRef HexStr(TokStart + 3, len); 846 if (!all_of(HexStr, isxdigit)) { 847 // Bad token, return it as an error. 848 CurPtr = TokStart+3; 849 return lltok::Error; 850 } 851 APInt Tmp(bits, HexStr, 16); 852 uint32_t activeBits = Tmp.getActiveBits(); 853 if (activeBits > 0 && activeBits < bits) 854 Tmp = Tmp.trunc(activeBits); 855 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 856 return lltok::APSInt; 857 } 858 859 // If this is "cc1234", return this as just "cc". 860 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 861 CurPtr = TokStart+2; 862 return lltok::kw_cc; 863 } 864 865 // Finally, if this isn't known, return an error. 866 CurPtr = TokStart+1; 867 return lltok::Error; 868 } 869 870 /// Lex all tokens that start with a 0x prefix, knowing they match and are not 871 /// labels. 872 /// HexFPConstant 0x[0-9A-Fa-f]+ 873 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 874 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 875 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 876 /// HexHalfConstant 0xH[0-9A-Fa-f]+ 877 lltok::Kind LLLexer::Lex0x() { 878 CurPtr = TokStart + 2; 879 880 char Kind; 881 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') { 882 Kind = *CurPtr++; 883 } else { 884 Kind = 'J'; 885 } 886 887 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) { 888 // Bad token, return it as an error. 889 CurPtr = TokStart+1; 890 return lltok::Error; 891 } 892 893 while (isxdigit(static_cast<unsigned char>(CurPtr[0]))) 894 ++CurPtr; 895 896 if (Kind == 'J') { 897 // HexFPConstant - Floating point constant represented in IEEE format as a 898 // hexadecimal number for when exponential notation is not precise enough. 899 // Half, Float, and double only. 900 APFloatVal = APFloat(APFloat::IEEEdouble(), 901 APInt(64, HexIntToVal(TokStart + 2, CurPtr))); 902 return lltok::APFloat; 903 } 904 905 uint64_t Pair[2]; 906 switch (Kind) { 907 default: llvm_unreachable("Unknown kind!"); 908 case 'K': 909 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 910 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 911 APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair)); 912 return lltok::APFloat; 913 case 'L': 914 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 915 HexToIntPair(TokStart+3, CurPtr, Pair); 916 APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair)); 917 return lltok::APFloat; 918 case 'M': 919 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 920 HexToIntPair(TokStart+3, CurPtr, Pair); 921 APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair)); 922 return lltok::APFloat; 923 case 'H': 924 APFloatVal = APFloat(APFloat::IEEEhalf(), 925 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 926 return lltok::APFloat; 927 } 928 } 929 930 /// Lex tokens for a label or a numeric constant, possibly starting with -. 931 /// Label [-a-zA-Z$._0-9]+: 932 /// NInteger -[0-9]+ 933 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 934 /// PInteger [0-9]+ 935 /// HexFPConstant 0x[0-9A-Fa-f]+ 936 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 937 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 938 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 939 lltok::Kind LLLexer::LexDigitOrNegative() { 940 // If the letter after the negative is not a number, this is probably a label. 941 if (!isdigit(static_cast<unsigned char>(TokStart[0])) && 942 !isdigit(static_cast<unsigned char>(CurPtr[0]))) { 943 // Okay, this is not a number after the -, it's probably a label. 944 if (const char *End = isLabelTail(CurPtr)) { 945 StrVal.assign(TokStart, End-1); 946 CurPtr = End; 947 return lltok::LabelStr; 948 } 949 950 return lltok::Error; 951 } 952 953 // At this point, it is either a label, int or fp constant. 954 955 // Skip digits, we have at least one. 956 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 957 /*empty*/; 958 959 // Check to see if this really is a label afterall, e.g. "-1:". 960 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 961 if (const char *End = isLabelTail(CurPtr)) { 962 StrVal.assign(TokStart, End-1); 963 CurPtr = End; 964 return lltok::LabelStr; 965 } 966 } 967 968 // If the next character is a '.', then it is a fp value, otherwise its 969 // integer. 970 if (CurPtr[0] != '.') { 971 if (TokStart[0] == '0' && TokStart[1] == 'x') 972 return Lex0x(); 973 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart)); 974 return lltok::APSInt; 975 } 976 977 ++CurPtr; 978 979 // Skip over [0-9]*([eE][-+]?[0-9]+)? 980 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 981 982 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 983 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 984 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 985 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 986 CurPtr += 2; 987 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 988 } 989 } 990 991 APFloatVal = APFloat(APFloat::IEEEdouble(), 992 StringRef(TokStart, CurPtr - TokStart)); 993 return lltok::APFloat; 994 } 995 996 /// Lex a floating point constant starting with +. 997 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 998 lltok::Kind LLLexer::LexPositive() { 999 // If the letter after the negative is a number, this is probably not a 1000 // label. 1001 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 1002 return lltok::Error; 1003 1004 // Skip digits. 1005 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 1006 /*empty*/; 1007 1008 // At this point, we need a '.'. 1009 if (CurPtr[0] != '.') { 1010 CurPtr = TokStart+1; 1011 return lltok::Error; 1012 } 1013 1014 ++CurPtr; 1015 1016 // Skip over [0-9]*([eE][-+]?[0-9]+)? 1017 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1018 1019 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 1020 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 1021 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 1022 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 1023 CurPtr += 2; 1024 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 1025 } 1026 } 1027 1028 APFloatVal = APFloat(APFloat::IEEEdouble(), 1029 StringRef(TokStart, CurPtr - TokStart)); 1030 return lltok::APFloat; 1031 } 1032