1 //===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // Implement the Lexer for .ll files. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "LLLexer.h" 15 #include "llvm/ADT/StringExtras.h" 16 #include "llvm/ADT/Twine.h" 17 #include "llvm/AsmParser/Parser.h" 18 #include "llvm/IR/DerivedTypes.h" 19 #include "llvm/IR/Instruction.h" 20 #include "llvm/IR/LLVMContext.h" 21 #include "llvm/Support/ErrorHandling.h" 22 #include "llvm/Support/MathExtras.h" 23 #include "llvm/Support/MemoryBuffer.h" 24 #include "llvm/Support/SourceMgr.h" 25 #include "llvm/Support/raw_ostream.h" 26 #include <cctype> 27 #include <cstdio> 28 #include <cstdlib> 29 #include <cstring> 30 using namespace llvm; 31 Error(LocTy ErrorLoc,const Twine & Msg) const32 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const { 33 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg); 34 return true; 35 } 36 Warning(LocTy WarningLoc,const Twine & Msg) const37 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const { 38 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg); 39 } 40 41 //===----------------------------------------------------------------------===// 42 // Helper functions. 43 //===----------------------------------------------------------------------===// 44 45 // atoull - Convert an ascii string of decimal digits into the unsigned long 46 // long representation... this does not have to do input error checking, 47 // because we know that the input will be matched by a suitable regex... 48 // atoull(const char * Buffer,const char * End)49 uint64_t LLLexer::atoull(const char *Buffer, const char *End) { 50 uint64_t Result = 0; 51 for (; Buffer != End; Buffer++) { 52 uint64_t OldRes = Result; 53 Result *= 10; 54 Result += *Buffer-'0'; 55 if (Result < OldRes) { // Uh, oh, overflow detected!!! 56 Error("constant bigger than 64 bits detected!"); 57 return 0; 58 } 59 } 60 return Result; 61 } 62 HexIntToVal(const char * Buffer,const char * End)63 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) { 64 uint64_t Result = 0; 65 for (; Buffer != End; ++Buffer) { 66 uint64_t OldRes = Result; 67 Result *= 16; 68 Result += hexDigitValue(*Buffer); 69 70 if (Result < OldRes) { // Uh, oh, overflow detected!!! 71 Error("constant bigger than 64 bits detected!"); 72 return 0; 73 } 74 } 75 return Result; 76 } 77 HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])78 void LLLexer::HexToIntPair(const char *Buffer, const char *End, 79 uint64_t Pair[2]) { 80 Pair[0] = 0; 81 if (End - Buffer >= 16) { 82 for (int i = 0; i < 16; i++, Buffer++) { 83 assert(Buffer != End); 84 Pair[0] *= 16; 85 Pair[0] += hexDigitValue(*Buffer); 86 } 87 } 88 Pair[1] = 0; 89 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 90 Pair[1] *= 16; 91 Pair[1] += hexDigitValue(*Buffer); 92 } 93 if (Buffer != End) 94 Error("constant bigger than 128 bits detected!"); 95 } 96 97 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into 98 /// { low64, high16 } as usual for an APInt. FP80HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])99 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End, 100 uint64_t Pair[2]) { 101 Pair[1] = 0; 102 for (int i=0; i<4 && Buffer != End; i++, Buffer++) { 103 assert(Buffer != End); 104 Pair[1] *= 16; 105 Pair[1] += hexDigitValue(*Buffer); 106 } 107 Pair[0] = 0; 108 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) { 109 Pair[0] *= 16; 110 Pair[0] += hexDigitValue(*Buffer); 111 } 112 if (Buffer != End) 113 Error("constant bigger than 128 bits detected!"); 114 } 115 116 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the 117 // appropriate character. UnEscapeLexed(std::string & Str)118 static void UnEscapeLexed(std::string &Str) { 119 if (Str.empty()) return; 120 121 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size(); 122 char *BOut = Buffer; 123 for (char *BIn = Buffer; BIn != EndBuffer; ) { 124 if (BIn[0] == '\\') { 125 if (BIn < EndBuffer-1 && BIn[1] == '\\') { 126 *BOut++ = '\\'; // Two \ becomes one 127 BIn += 2; 128 } else if (BIn < EndBuffer-2 && 129 isxdigit(static_cast<unsigned char>(BIn[1])) && 130 isxdigit(static_cast<unsigned char>(BIn[2]))) { 131 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]); 132 BIn += 3; // Skip over handled chars 133 ++BOut; 134 } else { 135 *BOut++ = *BIn++; 136 } 137 } else { 138 *BOut++ = *BIn++; 139 } 140 } 141 Str.resize(BOut-Buffer); 142 } 143 144 /// isLabelChar - Return true for [-a-zA-Z$._0-9]. isLabelChar(char C)145 static bool isLabelChar(char C) { 146 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' || 147 C == '.' || C == '_'; 148 } 149 150 151 /// isLabelTail - Return true if this pointer points to a valid end of a label. isLabelTail(const char * CurPtr)152 static const char *isLabelTail(const char *CurPtr) { 153 while (1) { 154 if (CurPtr[0] == ':') return CurPtr+1; 155 if (!isLabelChar(CurPtr[0])) return nullptr; 156 ++CurPtr; 157 } 158 } 159 160 161 162 //===----------------------------------------------------------------------===// 163 // Lexer definition. 164 //===----------------------------------------------------------------------===// 165 LLLexer(StringRef StartBuf,SourceMgr & sm,SMDiagnostic & Err,LLVMContext & C)166 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &sm, SMDiagnostic &Err, 167 LLVMContext &C) 168 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) { 169 CurPtr = CurBuf.begin(); 170 } 171 getNextChar()172 int LLLexer::getNextChar() { 173 char CurChar = *CurPtr++; 174 switch (CurChar) { 175 default: return (unsigned char)CurChar; 176 case 0: 177 // A nul character in the stream is either the end of the current buffer or 178 // a random nul in the file. Disambiguate that here. 179 if (CurPtr-1 != CurBuf.end()) 180 return 0; // Just whitespace. 181 182 // Otherwise, return end of file. 183 --CurPtr; // Another call to lex will return EOF again. 184 return EOF; 185 } 186 } 187 188 LexToken()189 lltok::Kind LLLexer::LexToken() { 190 TokStart = CurPtr; 191 192 int CurChar = getNextChar(); 193 switch (CurChar) { 194 default: 195 // Handle letters: [a-zA-Z_] 196 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_') 197 return LexIdentifier(); 198 199 return lltok::Error; 200 case EOF: return lltok::Eof; 201 case 0: 202 case ' ': 203 case '\t': 204 case '\n': 205 case '\r': 206 // Ignore whitespace. 207 return LexToken(); 208 case '+': return LexPositive(); 209 case '@': return LexAt(); 210 case '$': return LexDollar(); 211 case '%': return LexPercent(); 212 case '"': return LexQuote(); 213 case '.': 214 if (const char *Ptr = isLabelTail(CurPtr)) { 215 CurPtr = Ptr; 216 StrVal.assign(TokStart, CurPtr-1); 217 return lltok::LabelStr; 218 } 219 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 220 CurPtr += 2; 221 return lltok::dotdotdot; 222 } 223 return lltok::Error; 224 case ';': 225 SkipLineComment(); 226 return LexToken(); 227 case '!': return LexExclaim(); 228 case '#': return LexHash(); 229 case '0': case '1': case '2': case '3': case '4': 230 case '5': case '6': case '7': case '8': case '9': 231 case '-': 232 return LexDigitOrNegative(); 233 case '=': return lltok::equal; 234 case '[': return lltok::lsquare; 235 case ']': return lltok::rsquare; 236 case '{': return lltok::lbrace; 237 case '}': return lltok::rbrace; 238 case '<': return lltok::less; 239 case '>': return lltok::greater; 240 case '(': return lltok::lparen; 241 case ')': return lltok::rparen; 242 case ',': return lltok::comma; 243 case '*': return lltok::star; 244 case '|': return lltok::bar; 245 } 246 } 247 SkipLineComment()248 void LLLexer::SkipLineComment() { 249 while (1) { 250 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 251 return; 252 } 253 } 254 255 /// Lex all tokens that start with an @ character. 256 /// GlobalVar @\"[^\"]*\" 257 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 258 /// GlobalVarID @[0-9]+ LexAt()259 lltok::Kind LLLexer::LexAt() { 260 return LexVar(lltok::GlobalVar, lltok::GlobalID); 261 } 262 LexDollar()263 lltok::Kind LLLexer::LexDollar() { 264 if (const char *Ptr = isLabelTail(TokStart)) { 265 CurPtr = Ptr; 266 StrVal.assign(TokStart, CurPtr - 1); 267 return lltok::LabelStr; 268 } 269 270 // Handle DollarStringConstant: $\"[^\"]*\" 271 if (CurPtr[0] == '"') { 272 ++CurPtr; 273 274 while (1) { 275 int CurChar = getNextChar(); 276 277 if (CurChar == EOF) { 278 Error("end of file in COMDAT variable name"); 279 return lltok::Error; 280 } 281 if (CurChar == '"') { 282 StrVal.assign(TokStart + 2, CurPtr - 1); 283 UnEscapeLexed(StrVal); 284 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 285 Error("Null bytes are not allowed in names"); 286 return lltok::Error; 287 } 288 return lltok::ComdatVar; 289 } 290 } 291 } 292 293 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]* 294 if (ReadVarName()) 295 return lltok::ComdatVar; 296 297 return lltok::Error; 298 } 299 300 /// ReadString - Read a string until the closing quote. ReadString(lltok::Kind kind)301 lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 302 const char *Start = CurPtr; 303 while (1) { 304 int CurChar = getNextChar(); 305 306 if (CurChar == EOF) { 307 Error("end of file in string constant"); 308 return lltok::Error; 309 } 310 if (CurChar == '"') { 311 StrVal.assign(Start, CurPtr-1); 312 UnEscapeLexed(StrVal); 313 return kind; 314 } 315 } 316 } 317 318 /// ReadVarName - Read the rest of a token containing a variable name. ReadVarName()319 bool LLLexer::ReadVarName() { 320 const char *NameStart = CurPtr; 321 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 322 CurPtr[0] == '-' || CurPtr[0] == '$' || 323 CurPtr[0] == '.' || CurPtr[0] == '_') { 324 ++CurPtr; 325 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 326 CurPtr[0] == '-' || CurPtr[0] == '$' || 327 CurPtr[0] == '.' || CurPtr[0] == '_') 328 ++CurPtr; 329 330 StrVal.assign(NameStart, CurPtr); 331 return true; 332 } 333 return false; 334 } 335 LexVar(lltok::Kind Var,lltok::Kind VarID)336 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) { 337 // Handle StringConstant: \"[^\"]*\" 338 if (CurPtr[0] == '"') { 339 ++CurPtr; 340 341 while (1) { 342 int CurChar = getNextChar(); 343 344 if (CurChar == EOF) { 345 Error("end of file in global variable name"); 346 return lltok::Error; 347 } 348 if (CurChar == '"') { 349 StrVal.assign(TokStart+2, CurPtr-1); 350 UnEscapeLexed(StrVal); 351 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 352 Error("Null bytes are not allowed in names"); 353 return lltok::Error; 354 } 355 return Var; 356 } 357 } 358 } 359 360 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]* 361 if (ReadVarName()) 362 return Var; 363 364 // Handle VarID: [0-9]+ 365 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 366 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 367 /*empty*/; 368 369 uint64_t Val = atoull(TokStart+1, CurPtr); 370 if ((unsigned)Val != Val) 371 Error("invalid value number (too large)!"); 372 UIntVal = unsigned(Val); 373 return VarID; 374 } 375 return lltok::Error; 376 } 377 378 /// Lex all tokens that start with a % character. 379 /// LocalVar ::= %\"[^\"]*\" 380 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 381 /// LocalVarID ::= %[0-9]+ LexPercent()382 lltok::Kind LLLexer::LexPercent() { 383 return LexVar(lltok::LocalVar, lltok::LocalVarID); 384 } 385 386 /// Lex all tokens that start with a " character. 387 /// QuoteLabel "[^"]+": 388 /// StringConstant "[^"]*" LexQuote()389 lltok::Kind LLLexer::LexQuote() { 390 lltok::Kind kind = ReadString(lltok::StringConstant); 391 if (kind == lltok::Error || kind == lltok::Eof) 392 return kind; 393 394 if (CurPtr[0] == ':') { 395 ++CurPtr; 396 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) { 397 Error("Null bytes are not allowed in names"); 398 kind = lltok::Error; 399 } else { 400 kind = lltok::LabelStr; 401 } 402 } 403 404 return kind; 405 } 406 407 /// Lex all tokens that start with a ! character. 408 /// !foo 409 /// ! LexExclaim()410 lltok::Kind LLLexer::LexExclaim() { 411 // Lex a metadata name as a MetadataVar. 412 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 413 CurPtr[0] == '-' || CurPtr[0] == '$' || 414 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 415 ++CurPtr; 416 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 417 CurPtr[0] == '-' || CurPtr[0] == '$' || 418 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 419 ++CurPtr; 420 421 StrVal.assign(TokStart+1, CurPtr); // Skip ! 422 UnEscapeLexed(StrVal); 423 return lltok::MetadataVar; 424 } 425 return lltok::exclaim; 426 } 427 428 /// Lex all tokens that start with a # character. 429 /// AttrGrpID ::= #[0-9]+ LexHash()430 lltok::Kind LLLexer::LexHash() { 431 // Handle AttrGrpID: #[0-9]+ 432 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 433 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 434 /*empty*/; 435 436 uint64_t Val = atoull(TokStart+1, CurPtr); 437 if ((unsigned)Val != Val) 438 Error("invalid value number (too large)!"); 439 UIntVal = unsigned(Val); 440 return lltok::AttrGrpID; 441 } 442 443 return lltok::Error; 444 } 445 446 /// Lex a label, integer type, keyword, or hexadecimal integer constant. 447 /// Label [-a-zA-Z$._0-9]+: 448 /// IntegerType i[0-9]+ 449 /// Keyword sdiv, float, ... 450 /// HexIntConstant [us]0x[0-9A-Fa-f]+ LexIdentifier()451 lltok::Kind LLLexer::LexIdentifier() { 452 const char *StartChar = CurPtr; 453 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar; 454 const char *KeywordEnd = nullptr; 455 456 for (; isLabelChar(*CurPtr); ++CurPtr) { 457 // If we decide this is an integer, remember the end of the sequence. 458 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr))) 459 IntEnd = CurPtr; 460 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) && 461 *CurPtr != '_') 462 KeywordEnd = CurPtr; 463 } 464 465 // If we stopped due to a colon, this really is a label. 466 if (*CurPtr == ':') { 467 StrVal.assign(StartChar-1, CurPtr++); 468 return lltok::LabelStr; 469 } 470 471 // Otherwise, this wasn't a label. If this was valid as an integer type, 472 // return it. 473 if (!IntEnd) IntEnd = CurPtr; 474 if (IntEnd != StartChar) { 475 CurPtr = IntEnd; 476 uint64_t NumBits = atoull(StartChar, CurPtr); 477 if (NumBits < IntegerType::MIN_INT_BITS || 478 NumBits > IntegerType::MAX_INT_BITS) { 479 Error("bitwidth for integer type out of range!"); 480 return lltok::Error; 481 } 482 TyVal = IntegerType::get(Context, NumBits); 483 return lltok::Type; 484 } 485 486 // Otherwise, this was a letter sequence. See which keyword this is. 487 if (!KeywordEnd) KeywordEnd = CurPtr; 488 CurPtr = KeywordEnd; 489 --StartChar; 490 StringRef Keyword(StartChar, CurPtr - StartChar); 491 #define KEYWORD(STR) \ 492 do { \ 493 if (Keyword == #STR) \ 494 return lltok::kw_##STR; \ 495 } while (0) 496 497 KEYWORD(true); KEYWORD(false); 498 KEYWORD(declare); KEYWORD(define); 499 KEYWORD(global); KEYWORD(constant); 500 501 KEYWORD(private); 502 KEYWORD(internal); 503 KEYWORD(available_externally); 504 KEYWORD(linkonce); 505 KEYWORD(linkonce_odr); 506 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg". 507 KEYWORD(weak_odr); 508 KEYWORD(appending); 509 KEYWORD(dllimport); 510 KEYWORD(dllexport); 511 KEYWORD(common); 512 KEYWORD(default); 513 KEYWORD(hidden); 514 KEYWORD(protected); 515 KEYWORD(unnamed_addr); 516 KEYWORD(local_unnamed_addr); 517 KEYWORD(externally_initialized); 518 KEYWORD(extern_weak); 519 KEYWORD(external); 520 KEYWORD(thread_local); 521 KEYWORD(localdynamic); 522 KEYWORD(initialexec); 523 KEYWORD(localexec); 524 KEYWORD(zeroinitializer); 525 KEYWORD(undef); 526 KEYWORD(null); 527 KEYWORD(none); 528 KEYWORD(to); 529 KEYWORD(caller); 530 KEYWORD(within); 531 KEYWORD(from); 532 KEYWORD(tail); 533 KEYWORD(musttail); 534 KEYWORD(notail); 535 KEYWORD(target); 536 KEYWORD(triple); 537 KEYWORD(source_filename); 538 KEYWORD(unwind); 539 KEYWORD(deplibs); // FIXME: Remove in 4.0. 540 KEYWORD(datalayout); 541 KEYWORD(volatile); 542 KEYWORD(atomic); 543 KEYWORD(unordered); 544 KEYWORD(monotonic); 545 KEYWORD(acquire); 546 KEYWORD(release); 547 KEYWORD(acq_rel); 548 KEYWORD(seq_cst); 549 KEYWORD(singlethread); 550 551 KEYWORD(nnan); 552 KEYWORD(ninf); 553 KEYWORD(nsz); 554 KEYWORD(arcp); 555 KEYWORD(fast); 556 KEYWORD(nuw); 557 KEYWORD(nsw); 558 KEYWORD(exact); 559 KEYWORD(inbounds); 560 KEYWORD(align); 561 KEYWORD(addrspace); 562 KEYWORD(section); 563 KEYWORD(alias); 564 KEYWORD(ifunc); 565 KEYWORD(module); 566 KEYWORD(asm); 567 KEYWORD(sideeffect); 568 KEYWORD(alignstack); 569 KEYWORD(inteldialect); 570 KEYWORD(gc); 571 KEYWORD(prefix); 572 KEYWORD(prologue); 573 574 KEYWORD(ccc); 575 KEYWORD(fastcc); 576 KEYWORD(coldcc); 577 KEYWORD(x86_stdcallcc); 578 KEYWORD(x86_fastcallcc); 579 KEYWORD(x86_thiscallcc); 580 KEYWORD(x86_vectorcallcc); 581 KEYWORD(arm_apcscc); 582 KEYWORD(arm_aapcscc); 583 KEYWORD(arm_aapcs_vfpcc); 584 KEYWORD(msp430_intrcc); 585 KEYWORD(avr_intrcc); 586 KEYWORD(avr_signalcc); 587 KEYWORD(ptx_kernel); 588 KEYWORD(ptx_device); 589 KEYWORD(spir_kernel); 590 KEYWORD(spir_func); 591 KEYWORD(intel_ocl_bicc); 592 KEYWORD(x86_64_sysvcc); 593 KEYWORD(x86_64_win64cc); 594 KEYWORD(webkit_jscc); 595 KEYWORD(swiftcc); 596 KEYWORD(anyregcc); 597 KEYWORD(preserve_mostcc); 598 KEYWORD(preserve_allcc); 599 KEYWORD(ghccc); 600 KEYWORD(x86_intrcc); 601 KEYWORD(hhvmcc); 602 KEYWORD(hhvm_ccc); 603 KEYWORD(cxx_fast_tlscc); 604 KEYWORD(amdgpu_vs); 605 KEYWORD(amdgpu_gs); 606 KEYWORD(amdgpu_ps); 607 KEYWORD(amdgpu_cs); 608 KEYWORD(amdgpu_kernel); 609 610 KEYWORD(cc); 611 KEYWORD(c); 612 613 KEYWORD(attributes); 614 615 KEYWORD(alwaysinline); 616 KEYWORD(allocsize); 617 KEYWORD(argmemonly); 618 KEYWORD(builtin); 619 KEYWORD(byval); 620 KEYWORD(inalloca); 621 KEYWORD(cold); 622 KEYWORD(convergent); 623 KEYWORD(dereferenceable); 624 KEYWORD(dereferenceable_or_null); 625 KEYWORD(inaccessiblememonly); 626 KEYWORD(inaccessiblemem_or_argmemonly); 627 KEYWORD(inlinehint); 628 KEYWORD(inreg); 629 KEYWORD(jumptable); 630 KEYWORD(minsize); 631 KEYWORD(naked); 632 KEYWORD(nest); 633 KEYWORD(noalias); 634 KEYWORD(nobuiltin); 635 KEYWORD(nocapture); 636 KEYWORD(noduplicate); 637 KEYWORD(noimplicitfloat); 638 KEYWORD(noinline); 639 KEYWORD(norecurse); 640 KEYWORD(nonlazybind); 641 KEYWORD(nonnull); 642 KEYWORD(noredzone); 643 KEYWORD(noreturn); 644 KEYWORD(nounwind); 645 KEYWORD(optnone); 646 KEYWORD(optsize); 647 KEYWORD(readnone); 648 KEYWORD(readonly); 649 KEYWORD(returned); 650 KEYWORD(returns_twice); 651 KEYWORD(signext); 652 KEYWORD(sret); 653 KEYWORD(ssp); 654 KEYWORD(sspreq); 655 KEYWORD(sspstrong); 656 KEYWORD(safestack); 657 KEYWORD(sanitize_address); 658 KEYWORD(sanitize_thread); 659 KEYWORD(sanitize_memory); 660 KEYWORD(swifterror); 661 KEYWORD(swiftself); 662 KEYWORD(uwtable); 663 KEYWORD(writeonly); 664 KEYWORD(zeroext); 665 666 KEYWORD(type); 667 KEYWORD(opaque); 668 669 KEYWORD(comdat); 670 671 // Comdat types 672 KEYWORD(any); 673 KEYWORD(exactmatch); 674 KEYWORD(largest); 675 KEYWORD(noduplicates); 676 KEYWORD(samesize); 677 678 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 679 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 680 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 681 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 682 683 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 684 KEYWORD(umin); 685 686 KEYWORD(x); 687 KEYWORD(blockaddress); 688 689 // Metadata types. 690 KEYWORD(distinct); 691 692 // Use-list order directives. 693 KEYWORD(uselistorder); 694 KEYWORD(uselistorder_bb); 695 696 KEYWORD(personality); 697 KEYWORD(cleanup); 698 KEYWORD(catch); 699 KEYWORD(filter); 700 #undef KEYWORD 701 702 // Keywords for types. 703 #define TYPEKEYWORD(STR, LLVMTY) \ 704 do { \ 705 if (Keyword == STR) { \ 706 TyVal = LLVMTY; \ 707 return lltok::Type; \ 708 } \ 709 } while (false) 710 TYPEKEYWORD("void", Type::getVoidTy(Context)); 711 TYPEKEYWORD("half", Type::getHalfTy(Context)); 712 TYPEKEYWORD("float", Type::getFloatTy(Context)); 713 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 714 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 715 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 716 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 717 TYPEKEYWORD("label", Type::getLabelTy(Context)); 718 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 719 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 720 TYPEKEYWORD("token", Type::getTokenTy(Context)); 721 #undef TYPEKEYWORD 722 723 // Keywords for instructions. 724 #define INSTKEYWORD(STR, Enum) \ 725 do { \ 726 if (Keyword == #STR) { \ 727 UIntVal = Instruction::Enum; \ 728 return lltok::kw_##STR; \ 729 } \ 730 } while (false) 731 732 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 733 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 734 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 735 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 736 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 737 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 738 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 739 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 740 741 INSTKEYWORD(phi, PHI); 742 INSTKEYWORD(call, Call); 743 INSTKEYWORD(trunc, Trunc); 744 INSTKEYWORD(zext, ZExt); 745 INSTKEYWORD(sext, SExt); 746 INSTKEYWORD(fptrunc, FPTrunc); 747 INSTKEYWORD(fpext, FPExt); 748 INSTKEYWORD(uitofp, UIToFP); 749 INSTKEYWORD(sitofp, SIToFP); 750 INSTKEYWORD(fptoui, FPToUI); 751 INSTKEYWORD(fptosi, FPToSI); 752 INSTKEYWORD(inttoptr, IntToPtr); 753 INSTKEYWORD(ptrtoint, PtrToInt); 754 INSTKEYWORD(bitcast, BitCast); 755 INSTKEYWORD(addrspacecast, AddrSpaceCast); 756 INSTKEYWORD(select, Select); 757 INSTKEYWORD(va_arg, VAArg); 758 INSTKEYWORD(ret, Ret); 759 INSTKEYWORD(br, Br); 760 INSTKEYWORD(switch, Switch); 761 INSTKEYWORD(indirectbr, IndirectBr); 762 INSTKEYWORD(invoke, Invoke); 763 INSTKEYWORD(resume, Resume); 764 INSTKEYWORD(unreachable, Unreachable); 765 766 INSTKEYWORD(alloca, Alloca); 767 INSTKEYWORD(load, Load); 768 INSTKEYWORD(store, Store); 769 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 770 INSTKEYWORD(atomicrmw, AtomicRMW); 771 INSTKEYWORD(fence, Fence); 772 INSTKEYWORD(getelementptr, GetElementPtr); 773 774 INSTKEYWORD(extractelement, ExtractElement); 775 INSTKEYWORD(insertelement, InsertElement); 776 INSTKEYWORD(shufflevector, ShuffleVector); 777 INSTKEYWORD(extractvalue, ExtractValue); 778 INSTKEYWORD(insertvalue, InsertValue); 779 INSTKEYWORD(landingpad, LandingPad); 780 INSTKEYWORD(cleanupret, CleanupRet); 781 INSTKEYWORD(catchret, CatchRet); 782 INSTKEYWORD(catchswitch, CatchSwitch); 783 INSTKEYWORD(catchpad, CatchPad); 784 INSTKEYWORD(cleanuppad, CleanupPad); 785 #undef INSTKEYWORD 786 787 #define DWKEYWORD(TYPE, TOKEN) \ 788 do { \ 789 if (Keyword.startswith("DW_" #TYPE "_")) { \ 790 StrVal.assign(Keyword.begin(), Keyword.end()); \ 791 return lltok::TOKEN; \ 792 } \ 793 } while (false) 794 DWKEYWORD(TAG, DwarfTag); 795 DWKEYWORD(ATE, DwarfAttEncoding); 796 DWKEYWORD(VIRTUALITY, DwarfVirtuality); 797 DWKEYWORD(LANG, DwarfLang); 798 DWKEYWORD(CC, DwarfCC); 799 DWKEYWORD(OP, DwarfOp); 800 DWKEYWORD(MACINFO, DwarfMacinfo); 801 #undef DWKEYWORD 802 if (Keyword.startswith("DIFlag")) { 803 StrVal.assign(Keyword.begin(), Keyword.end()); 804 return lltok::DIFlag; 805 } 806 if (Keyword == "NoDebug" || Keyword == "FullDebug" || 807 Keyword == "LineTablesOnly") { 808 StrVal.assign(Keyword.begin(), Keyword.end()); 809 return lltok::EmissionKind; 810 } 811 812 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 813 // the CFE to avoid forcing it to deal with 64-bit numbers. 814 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 815 TokStart[1] == '0' && TokStart[2] == 'x' && 816 isxdigit(static_cast<unsigned char>(TokStart[3]))) { 817 int len = CurPtr-TokStart-3; 818 uint32_t bits = len * 4; 819 StringRef HexStr(TokStart + 3, len); 820 if (!std::all_of(HexStr.begin(), HexStr.end(), isxdigit)) { 821 // Bad token, return it as an error. 822 CurPtr = TokStart+3; 823 return lltok::Error; 824 } 825 APInt Tmp(bits, HexStr, 16); 826 uint32_t activeBits = Tmp.getActiveBits(); 827 if (activeBits > 0 && activeBits < bits) 828 Tmp = Tmp.trunc(activeBits); 829 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 830 return lltok::APSInt; 831 } 832 833 // If this is "cc1234", return this as just "cc". 834 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 835 CurPtr = TokStart+2; 836 return lltok::kw_cc; 837 } 838 839 // Finally, if this isn't known, return an error. 840 CurPtr = TokStart+1; 841 return lltok::Error; 842 } 843 844 /// Lex all tokens that start with a 0x prefix, knowing they match and are not 845 /// labels. 846 /// HexFPConstant 0x[0-9A-Fa-f]+ 847 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 848 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 849 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ 850 /// HexHalfConstant 0xH[0-9A-Fa-f]+ Lex0x()851 lltok::Kind LLLexer::Lex0x() { 852 CurPtr = TokStart + 2; 853 854 char Kind; 855 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') { 856 Kind = *CurPtr++; 857 } else { 858 Kind = 'J'; 859 } 860 861 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) { 862 // Bad token, return it as an error. 863 CurPtr = TokStart+1; 864 return lltok::Error; 865 } 866 867 while (isxdigit(static_cast<unsigned char>(CurPtr[0]))) 868 ++CurPtr; 869 870 if (Kind == 'J') { 871 // HexFPConstant - Floating point constant represented in IEEE format as a 872 // hexadecimal number for when exponential notation is not precise enough. 873 // Half, Float, and double only. 874 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr))); 875 return lltok::APFloat; 876 } 877 878 uint64_t Pair[2]; 879 switch (Kind) { 880 default: llvm_unreachable("Unknown kind!"); 881 case 'K': 882 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 883 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 884 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair)); 885 return lltok::APFloat; 886 case 'L': 887 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 888 HexToIntPair(TokStart+3, CurPtr, Pair); 889 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair)); 890 return lltok::APFloat; 891 case 'M': 892 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 893 HexToIntPair(TokStart+3, CurPtr, Pair); 894 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair)); 895 return lltok::APFloat; 896 case 'H': 897 APFloatVal = APFloat(APFloat::IEEEhalf, 898 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 899 return lltok::APFloat; 900 } 901 } 902 903 /// Lex tokens for a label or a numeric constant, possibly starting with -. 904 /// Label [-a-zA-Z$._0-9]+: 905 /// NInteger -[0-9]+ 906 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 907 /// PInteger [0-9]+ 908 /// HexFPConstant 0x[0-9A-Fa-f]+ 909 /// HexFP80Constant 0xK[0-9A-Fa-f]+ 910 /// HexFP128Constant 0xL[0-9A-Fa-f]+ 911 /// HexPPC128Constant 0xM[0-9A-Fa-f]+ LexDigitOrNegative()912 lltok::Kind LLLexer::LexDigitOrNegative() { 913 // If the letter after the negative is not a number, this is probably a label. 914 if (!isdigit(static_cast<unsigned char>(TokStart[0])) && 915 !isdigit(static_cast<unsigned char>(CurPtr[0]))) { 916 // Okay, this is not a number after the -, it's probably a label. 917 if (const char *End = isLabelTail(CurPtr)) { 918 StrVal.assign(TokStart, End-1); 919 CurPtr = End; 920 return lltok::LabelStr; 921 } 922 923 return lltok::Error; 924 } 925 926 // At this point, it is either a label, int or fp constant. 927 928 // Skip digits, we have at least one. 929 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 930 /*empty*/; 931 932 // Check to see if this really is a label afterall, e.g. "-1:". 933 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 934 if (const char *End = isLabelTail(CurPtr)) { 935 StrVal.assign(TokStart, End-1); 936 CurPtr = End; 937 return lltok::LabelStr; 938 } 939 } 940 941 // If the next character is a '.', then it is a fp value, otherwise its 942 // integer. 943 if (CurPtr[0] != '.') { 944 if (TokStart[0] == '0' && TokStart[1] == 'x') 945 return Lex0x(); 946 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart)); 947 return lltok::APSInt; 948 } 949 950 ++CurPtr; 951 952 // Skip over [0-9]*([eE][-+]?[0-9]+)? 953 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 954 955 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 956 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 957 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 958 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 959 CurPtr += 2; 960 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 961 } 962 } 963 964 APFloatVal = APFloat(APFloat::IEEEdouble, 965 StringRef(TokStart, CurPtr - TokStart)); 966 return lltok::APFloat; 967 } 968 969 /// Lex a floating point constant starting with +. 970 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? LexPositive()971 lltok::Kind LLLexer::LexPositive() { 972 // If the letter after the negative is a number, this is probably not a 973 // label. 974 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 975 return lltok::Error; 976 977 // Skip digits. 978 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 979 /*empty*/; 980 981 // At this point, we need a '.'. 982 if (CurPtr[0] != '.') { 983 CurPtr = TokStart+1; 984 return lltok::Error; 985 } 986 987 ++CurPtr; 988 989 // Skip over [0-9]*([eE][-+]?[0-9]+)? 990 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 991 992 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 993 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 994 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 995 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 996 CurPtr += 2; 997 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 998 } 999 } 1000 1001 APFloatVal = APFloat(APFloat::IEEEdouble, 1002 StringRef(TokStart, CurPtr - TokStart)); 1003 return lltok::APFloat; 1004 } 1005