LLLexer.cpp revision 69bd41dfe33f24414be281ba5e2204b7348c33ae
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/Assembly/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> 30using namespace llvm; 31 32bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const { 33 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg); 34 return true; 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// 45uint64_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 59uint64_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 74void LLLexer::HexToIntPair(const char *Buffer, const char *End, 75 uint64_t Pair[2]) { 76 Pair[0] = 0; 77 for (int i=0; i<16; i++, Buffer++) { 78 assert(Buffer != End); 79 Pair[0] *= 16; 80 Pair[0] += hexDigitValue(*Buffer); 81 } 82 Pair[1] = 0; 83 for (int i=0; i<16 && Buffer != End; i++, Buffer++) { 84 Pair[1] *= 16; 85 Pair[1] += hexDigitValue(*Buffer); 86 } 87 if (Buffer != End) 88 Error("constant bigger than 128 bits detected!"); 89} 90 91/// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into 92/// { low64, high16 } as usual for an APInt. 93void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End, 94 uint64_t Pair[2]) { 95 Pair[1] = 0; 96 for (int i=0; i<4 && Buffer != End; i++, Buffer++) { 97 assert(Buffer != End); 98 Pair[1] *= 16; 99 Pair[1] += hexDigitValue(*Buffer); 100 } 101 Pair[0] = 0; 102 for (int i=0; i<16; i++, Buffer++) { 103 Pair[0] *= 16; 104 Pair[0] += hexDigitValue(*Buffer); 105 } 106 if (Buffer != End) 107 Error("constant bigger than 128 bits detected!"); 108} 109 110// UnEscapeLexed - Run through the specified buffer and change \xx codes to the 111// appropriate character. 112static void UnEscapeLexed(std::string &Str) { 113 if (Str.empty()) return; 114 115 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size(); 116 char *BOut = Buffer; 117 for (char *BIn = Buffer; BIn != EndBuffer; ) { 118 if (BIn[0] == '\\') { 119 if (BIn < EndBuffer-1 && BIn[1] == '\\') { 120 *BOut++ = '\\'; // Two \ becomes one 121 BIn += 2; 122 } else if (BIn < EndBuffer-2 && 123 isxdigit(static_cast<unsigned char>(BIn[1])) && 124 isxdigit(static_cast<unsigned char>(BIn[2]))) { 125 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]); 126 BIn += 3; // Skip over handled chars 127 ++BOut; 128 } else { 129 *BOut++ = *BIn++; 130 } 131 } else { 132 *BOut++ = *BIn++; 133 } 134 } 135 Str.resize(BOut-Buffer); 136} 137 138/// isLabelChar - Return true for [-a-zA-Z$._0-9]. 139static bool isLabelChar(char C) { 140 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' || 141 C == '.' || C == '_'; 142} 143 144 145/// isLabelTail - Return true if this pointer points to a valid end of a label. 146static const char *isLabelTail(const char *CurPtr) { 147 while (1) { 148 if (CurPtr[0] == ':') return CurPtr+1; 149 if (!isLabelChar(CurPtr[0])) return 0; 150 ++CurPtr; 151 } 152} 153 154 155 156//===----------------------------------------------------------------------===// 157// Lexer definition. 158//===----------------------------------------------------------------------===// 159 160LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err, 161 LLVMContext &C) 162 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) { 163 CurPtr = CurBuf->getBufferStart(); 164} 165 166std::string LLLexer::getFilename() const { 167 return CurBuf->getBufferIdentifier(); 168} 169 170int LLLexer::getNextChar() { 171 char CurChar = *CurPtr++; 172 switch (CurChar) { 173 default: return (unsigned char)CurChar; 174 case 0: 175 // A nul character in the stream is either the end of the current buffer or 176 // a random nul in the file. Disambiguate that here. 177 if (CurPtr-1 != CurBuf->getBufferEnd()) 178 return 0; // Just whitespace. 179 180 // Otherwise, return end of file. 181 --CurPtr; // Another call to lex will return EOF again. 182 return EOF; 183 } 184} 185 186 187lltok::Kind LLLexer::LexToken() { 188 TokStart = CurPtr; 189 190 int CurChar = getNextChar(); 191 switch (CurChar) { 192 default: 193 // Handle letters: [a-zA-Z_] 194 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_') 195 return LexIdentifier(); 196 197 return lltok::Error; 198 case EOF: return lltok::Eof; 199 case 0: 200 case ' ': 201 case '\t': 202 case '\n': 203 case '\r': 204 // Ignore whitespace. 205 return LexToken(); 206 case '+': return LexPositive(); 207 case '@': return LexAt(); 208 case '%': return LexPercent(); 209 case '"': return LexQuote(); 210 case '.': 211 if (const char *Ptr = isLabelTail(CurPtr)) { 212 CurPtr = Ptr; 213 StrVal.assign(TokStart, CurPtr-1); 214 return lltok::LabelStr; 215 } 216 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 217 CurPtr += 2; 218 return lltok::dotdotdot; 219 } 220 return lltok::Error; 221 case '$': 222 if (const char *Ptr = isLabelTail(CurPtr)) { 223 CurPtr = Ptr; 224 StrVal.assign(TokStart, CurPtr-1); 225 return lltok::LabelStr; 226 } 227 return lltok::Error; 228 case ';': 229 SkipLineComment(); 230 return LexToken(); 231 case '!': return LexExclaim(); 232 case '#': return LexHash(); 233 case '0': case '1': case '2': case '3': case '4': 234 case '5': case '6': case '7': case '8': case '9': 235 case '-': 236 return LexDigitOrNegative(); 237 case '=': return lltok::equal; 238 case '[': return lltok::lsquare; 239 case ']': return lltok::rsquare; 240 case '{': return lltok::lbrace; 241 case '}': return lltok::rbrace; 242 case '<': return lltok::less; 243 case '>': return lltok::greater; 244 case '(': return lltok::lparen; 245 case ')': return lltok::rparen; 246 case ',': return lltok::comma; 247 case '*': return lltok::star; 248 case '\\': return lltok::backslash; 249 } 250} 251 252void LLLexer::SkipLineComment() { 253 while (1) { 254 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 255 return; 256 } 257} 258 259/// LexAt - Lex all tokens that start with an @ character: 260/// GlobalVar @\"[^\"]*\" 261/// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 262/// GlobalVarID @[0-9]+ 263lltok::Kind LLLexer::LexAt() { 264 // Handle AtStringConstant: @\"[^\"]*\" 265 if (CurPtr[0] == '"') { 266 ++CurPtr; 267 268 while (1) { 269 int CurChar = getNextChar(); 270 271 if (CurChar == EOF) { 272 Error("end of file in global variable name"); 273 return lltok::Error; 274 } 275 if (CurChar == '"') { 276 StrVal.assign(TokStart+2, CurPtr-1); 277 UnEscapeLexed(StrVal); 278 return lltok::GlobalVar; 279 } 280 } 281 } 282 283 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]* 284 if (ReadVarName()) 285 return lltok::GlobalVar; 286 287 // Handle GlobalVarID: @[0-9]+ 288 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 289 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 290 /*empty*/; 291 292 uint64_t Val = atoull(TokStart+1, CurPtr); 293 if ((unsigned)Val != Val) 294 Error("invalid value number (too large)!"); 295 UIntVal = unsigned(Val); 296 return lltok::GlobalID; 297 } 298 299 return lltok::Error; 300} 301 302/// ReadString - Read a string until the closing quote. 303lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 304 const char *Start = CurPtr; 305 while (1) { 306 int CurChar = getNextChar(); 307 308 if (CurChar == EOF) { 309 Error("end of file in string constant"); 310 return lltok::Error; 311 } 312 if (CurChar == '"') { 313 StrVal.assign(Start, CurPtr-1); 314 UnEscapeLexed(StrVal); 315 return kind; 316 } 317 } 318} 319 320/// ReadVarName - Read the rest of a token containing a variable name. 321bool LLLexer::ReadVarName() { 322 const char *NameStart = CurPtr; 323 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 324 CurPtr[0] == '-' || CurPtr[0] == '$' || 325 CurPtr[0] == '.' || CurPtr[0] == '_') { 326 ++CurPtr; 327 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 328 CurPtr[0] == '-' || CurPtr[0] == '$' || 329 CurPtr[0] == '.' || CurPtr[0] == '_') 330 ++CurPtr; 331 332 StrVal.assign(NameStart, CurPtr); 333 return true; 334 } 335 return false; 336} 337 338/// LexPercent - Lex all tokens that start with a % character: 339/// LocalVar ::= %\"[^\"]*\" 340/// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 341/// LocalVarID ::= %[0-9]+ 342lltok::Kind LLLexer::LexPercent() { 343 // Handle LocalVarName: %\"[^\"]*\" 344 if (CurPtr[0] == '"') { 345 ++CurPtr; 346 return ReadString(lltok::LocalVar); 347 } 348 349 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]* 350 if (ReadVarName()) 351 return lltok::LocalVar; 352 353 // Handle LocalVarID: %[0-9]+ 354 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 355 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 356 /*empty*/; 357 358 uint64_t Val = atoull(TokStart+1, CurPtr); 359 if ((unsigned)Val != Val) 360 Error("invalid value number (too large)!"); 361 UIntVal = unsigned(Val); 362 return lltok::LocalVarID; 363 } 364 365 return lltok::Error; 366} 367 368/// LexQuote - Lex all tokens that start with a " character: 369/// QuoteLabel "[^"]+": 370/// StringConstant "[^"]*" 371lltok::Kind LLLexer::LexQuote() { 372 lltok::Kind kind = ReadString(lltok::StringConstant); 373 if (kind == lltok::Error || kind == lltok::Eof) 374 return kind; 375 376 if (CurPtr[0] == ':') { 377 ++CurPtr; 378 kind = lltok::LabelStr; 379 } 380 381 return kind; 382} 383 384/// LexExclaim: 385/// !foo 386/// ! 387lltok::Kind LLLexer::LexExclaim() { 388 // Lex a metadata name as a MetadataVar. 389 if (isalpha(static_cast<unsigned char>(CurPtr[0])) || 390 CurPtr[0] == '-' || CurPtr[0] == '$' || 391 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 392 ++CurPtr; 393 while (isalnum(static_cast<unsigned char>(CurPtr[0])) || 394 CurPtr[0] == '-' || CurPtr[0] == '$' || 395 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 396 ++CurPtr; 397 398 StrVal.assign(TokStart+1, CurPtr); // Skip ! 399 UnEscapeLexed(StrVal); 400 return lltok::MetadataVar; 401 } 402 return lltok::exclaim; 403} 404 405/// LexHash - Lex all tokens that start with a # character: 406/// AttrGrpID ::= #[0-9]+ 407lltok::Kind LLLexer::LexHash() { 408 // Handle AttrGrpID: #[0-9]+ 409 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) { 410 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 411 /*empty*/; 412 413 uint64_t Val = atoull(TokStart+1, CurPtr); 414 if ((unsigned)Val != Val) 415 Error("invalid value number (too large)!"); 416 UIntVal = unsigned(Val); 417 return lltok::AttrGrpID; 418 } 419 420 return lltok::Error; 421} 422 423/// LexIdentifier: Handle several related productions: 424/// Label [-a-zA-Z$._0-9]+: 425/// IntegerType i[0-9]+ 426/// Keyword sdiv, float, ... 427/// HexIntConstant [us]0x[0-9A-Fa-f]+ 428lltok::Kind LLLexer::LexIdentifier() { 429 const char *StartChar = CurPtr; 430 const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar; 431 const char *KeywordEnd = 0; 432 433 for (; isLabelChar(*CurPtr); ++CurPtr) { 434 // If we decide this is an integer, remember the end of the sequence. 435 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr))) 436 IntEnd = CurPtr; 437 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) && 438 *CurPtr != '_') 439 KeywordEnd = CurPtr; 440 } 441 442 // If we stopped due to a colon, this really is a label. 443 if (*CurPtr == ':') { 444 StrVal.assign(StartChar-1, CurPtr++); 445 return lltok::LabelStr; 446 } 447 448 // Otherwise, this wasn't a label. If this was valid as an integer type, 449 // return it. 450 if (IntEnd == 0) IntEnd = CurPtr; 451 if (IntEnd != StartChar) { 452 CurPtr = IntEnd; 453 uint64_t NumBits = atoull(StartChar, CurPtr); 454 if (NumBits < IntegerType::MIN_INT_BITS || 455 NumBits > IntegerType::MAX_INT_BITS) { 456 Error("bitwidth for integer type out of range!"); 457 return lltok::Error; 458 } 459 TyVal = IntegerType::get(Context, NumBits); 460 return lltok::Type; 461 } 462 463 // Otherwise, this was a letter sequence. See which keyword this is. 464 if (KeywordEnd == 0) KeywordEnd = CurPtr; 465 CurPtr = KeywordEnd; 466 --StartChar; 467 unsigned Len = CurPtr-StartChar; 468#define KEYWORD(STR) \ 469 do { \ 470 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \ 471 return lltok::kw_##STR; \ 472 } while (0) 473 474 KEYWORD(true); KEYWORD(false); 475 KEYWORD(declare); KEYWORD(define); 476 KEYWORD(global); KEYWORD(constant); 477 478 KEYWORD(private); 479 KEYWORD(linker_private); 480 KEYWORD(linker_private_weak); 481 KEYWORD(linker_private_weak_def_auto); // FIXME: For backwards compatibility. 482 KEYWORD(internal); 483 KEYWORD(available_externally); 484 KEYWORD(linkonce); 485 KEYWORD(linkonce_odr); 486 KEYWORD(linkonce_odr_auto_hide); 487 KEYWORD(weak); 488 KEYWORD(weak_odr); 489 KEYWORD(appending); 490 KEYWORD(dllimport); 491 KEYWORD(dllexport); 492 KEYWORD(common); 493 KEYWORD(default); 494 KEYWORD(hidden); 495 KEYWORD(protected); 496 KEYWORD(unnamed_addr); 497 KEYWORD(externally_initialized); 498 KEYWORD(extern_weak); 499 KEYWORD(external); 500 KEYWORD(thread_local); 501 KEYWORD(localdynamic); 502 KEYWORD(initialexec); 503 KEYWORD(localexec); 504 KEYWORD(zeroinitializer); 505 KEYWORD(undef); 506 KEYWORD(null); 507 KEYWORD(to); 508 KEYWORD(tail); 509 KEYWORD(target); 510 KEYWORD(triple); 511 KEYWORD(unwind); 512 KEYWORD(deplibs); // FIXME: Remove in 4.0. 513 KEYWORD(datalayout); 514 KEYWORD(volatile); 515 KEYWORD(atomic); 516 KEYWORD(unordered); 517 KEYWORD(monotonic); 518 KEYWORD(acquire); 519 KEYWORD(release); 520 KEYWORD(acq_rel); 521 KEYWORD(seq_cst); 522 KEYWORD(singlethread); 523 524 KEYWORD(nnan); 525 KEYWORD(ninf); 526 KEYWORD(nsz); 527 KEYWORD(arcp); 528 KEYWORD(fast); 529 KEYWORD(nuw); 530 KEYWORD(nsw); 531 KEYWORD(exact); 532 KEYWORD(inbounds); 533 KEYWORD(align); 534 KEYWORD(addrspace); 535 KEYWORD(section); 536 KEYWORD(alias); 537 KEYWORD(module); 538 KEYWORD(asm); 539 KEYWORD(sideeffect); 540 KEYWORD(alignstack); 541 KEYWORD(inteldialect); 542 KEYWORD(gc); 543 KEYWORD(prefix); 544 545 KEYWORD(ccc); 546 KEYWORD(fastcc); 547 KEYWORD(coldcc); 548 KEYWORD(x86_stdcallcc); 549 KEYWORD(x86_fastcallcc); 550 KEYWORD(x86_thiscallcc); 551 KEYWORD(arm_apcscc); 552 KEYWORD(arm_aapcscc); 553 KEYWORD(arm_aapcs_vfpcc); 554 KEYWORD(msp430_intrcc); 555 KEYWORD(ptx_kernel); 556 KEYWORD(ptx_device); 557 KEYWORD(spir_kernel); 558 KEYWORD(spir_func); 559 KEYWORD(intel_ocl_bicc); 560 KEYWORD(x86_64_sysvcc); 561 KEYWORD(x86_64_win64cc); 562 563 KEYWORD(cc); 564 KEYWORD(c); 565 566 KEYWORD(attributes); 567 568 KEYWORD(alwaysinline); 569 KEYWORD(builtin); 570 KEYWORD(byval); 571 KEYWORD(cold); 572 KEYWORD(inlinehint); 573 KEYWORD(inreg); 574 KEYWORD(minsize); 575 KEYWORD(naked); 576 KEYWORD(nest); 577 KEYWORD(noalias); 578 KEYWORD(nobuiltin); 579 KEYWORD(nocapture); 580 KEYWORD(noduplicate); 581 KEYWORD(noimplicitfloat); 582 KEYWORD(noinline); 583 KEYWORD(nonlazybind); 584 KEYWORD(noredzone); 585 KEYWORD(noreturn); 586 KEYWORD(nounwind); 587 KEYWORD(optnone); 588 KEYWORD(optsize); 589 KEYWORD(readnone); 590 KEYWORD(readonly); 591 KEYWORD(returned); 592 KEYWORD(returns_twice); 593 KEYWORD(signext); 594 KEYWORD(sret); 595 KEYWORD(ssp); 596 KEYWORD(sspreq); 597 KEYWORD(sspstrong); 598 KEYWORD(sanitize_address); 599 KEYWORD(sanitize_thread); 600 KEYWORD(sanitize_memory); 601 KEYWORD(uwtable); 602 KEYWORD(zeroext); 603 604 KEYWORD(type); 605 KEYWORD(opaque); 606 607 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 608 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 609 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 610 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 611 612 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 613 KEYWORD(umin); 614 615 KEYWORD(x); 616 KEYWORD(blockaddress); 617 618 KEYWORD(personality); 619 KEYWORD(cleanup); 620 KEYWORD(catch); 621 KEYWORD(filter); 622#undef KEYWORD 623 624 // Keywords for types. 625#define TYPEKEYWORD(STR, LLVMTY) \ 626 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \ 627 TyVal = LLVMTY; return lltok::Type; } 628 TYPEKEYWORD("void", Type::getVoidTy(Context)); 629 TYPEKEYWORD("half", Type::getHalfTy(Context)); 630 TYPEKEYWORD("float", Type::getFloatTy(Context)); 631 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 632 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 633 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 634 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 635 TYPEKEYWORD("label", Type::getLabelTy(Context)); 636 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 637 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 638#undef TYPEKEYWORD 639 640 // Keywords for instructions. 641#define INSTKEYWORD(STR, Enum) \ 642 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \ 643 UIntVal = Instruction::Enum; return lltok::kw_##STR; } 644 645 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 646 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 647 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 648 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 649 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 650 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 651 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 652 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 653 654 INSTKEYWORD(phi, PHI); 655 INSTKEYWORD(call, Call); 656 INSTKEYWORD(trunc, Trunc); 657 INSTKEYWORD(zext, ZExt); 658 INSTKEYWORD(sext, SExt); 659 INSTKEYWORD(fptrunc, FPTrunc); 660 INSTKEYWORD(fpext, FPExt); 661 INSTKEYWORD(uitofp, UIToFP); 662 INSTKEYWORD(sitofp, SIToFP); 663 INSTKEYWORD(fptoui, FPToUI); 664 INSTKEYWORD(fptosi, FPToSI); 665 INSTKEYWORD(inttoptr, IntToPtr); 666 INSTKEYWORD(ptrtoint, PtrToInt); 667 INSTKEYWORD(bitcast, BitCast); 668 INSTKEYWORD(select, Select); 669 INSTKEYWORD(va_arg, VAArg); 670 INSTKEYWORD(ret, Ret); 671 INSTKEYWORD(br, Br); 672 INSTKEYWORD(switch, Switch); 673 INSTKEYWORD(indirectbr, IndirectBr); 674 INSTKEYWORD(invoke, Invoke); 675 INSTKEYWORD(resume, Resume); 676 INSTKEYWORD(unreachable, Unreachable); 677 678 INSTKEYWORD(alloca, Alloca); 679 INSTKEYWORD(load, Load); 680 INSTKEYWORD(store, Store); 681 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 682 INSTKEYWORD(atomicrmw, AtomicRMW); 683 INSTKEYWORD(fence, Fence); 684 INSTKEYWORD(getelementptr, GetElementPtr); 685 686 INSTKEYWORD(extractelement, ExtractElement); 687 INSTKEYWORD(insertelement, InsertElement); 688 INSTKEYWORD(shufflevector, ShuffleVector); 689 INSTKEYWORD(extractvalue, ExtractValue); 690 INSTKEYWORD(insertvalue, InsertValue); 691 INSTKEYWORD(landingpad, LandingPad); 692#undef INSTKEYWORD 693 694 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 695 // the CFE to avoid forcing it to deal with 64-bit numbers. 696 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 697 TokStart[1] == '0' && TokStart[2] == 'x' && 698 isxdigit(static_cast<unsigned char>(TokStart[3]))) { 699 int len = CurPtr-TokStart-3; 700 uint32_t bits = len * 4; 701 APInt Tmp(bits, StringRef(TokStart+3, len), 16); 702 uint32_t activeBits = Tmp.getActiveBits(); 703 if (activeBits > 0 && activeBits < bits) 704 Tmp = Tmp.trunc(activeBits); 705 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 706 return lltok::APSInt; 707 } 708 709 // If this is "cc1234", return this as just "cc". 710 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 711 CurPtr = TokStart+2; 712 return lltok::kw_cc; 713 } 714 715 // Finally, if this isn't known, return an error. 716 CurPtr = TokStart+1; 717 return lltok::Error; 718} 719 720 721/// Lex0x: Handle productions that start with 0x, knowing that it matches and 722/// that this is not a label: 723/// HexFPConstant 0x[0-9A-Fa-f]+ 724/// HexFP80Constant 0xK[0-9A-Fa-f]+ 725/// HexFP128Constant 0xL[0-9A-Fa-f]+ 726/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 727/// HexHalfConstant 0xH[0-9A-Fa-f]+ 728lltok::Kind LLLexer::Lex0x() { 729 CurPtr = TokStart + 2; 730 731 char Kind; 732 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') { 733 Kind = *CurPtr++; 734 } else { 735 Kind = 'J'; 736 } 737 738 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) { 739 // Bad token, return it as an error. 740 CurPtr = TokStart+1; 741 return lltok::Error; 742 } 743 744 while (isxdigit(static_cast<unsigned char>(CurPtr[0]))) 745 ++CurPtr; 746 747 if (Kind == 'J') { 748 // HexFPConstant - Floating point constant represented in IEEE format as a 749 // hexadecimal number for when exponential notation is not precise enough. 750 // Half, Float, and double only. 751 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr))); 752 return lltok::APFloat; 753 } 754 755 uint64_t Pair[2]; 756 switch (Kind) { 757 default: llvm_unreachable("Unknown kind!"); 758 case 'K': 759 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 760 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 761 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair)); 762 return lltok::APFloat; 763 case 'L': 764 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 765 HexToIntPair(TokStart+3, CurPtr, Pair); 766 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair)); 767 return lltok::APFloat; 768 case 'M': 769 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 770 HexToIntPair(TokStart+3, CurPtr, Pair); 771 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair)); 772 return lltok::APFloat; 773 case 'H': 774 APFloatVal = APFloat(APFloat::IEEEhalf, 775 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 776 return lltok::APFloat; 777 } 778} 779 780/// LexIdentifier: Handle several related productions: 781/// Label [-a-zA-Z$._0-9]+: 782/// NInteger -[0-9]+ 783/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 784/// PInteger [0-9]+ 785/// HexFPConstant 0x[0-9A-Fa-f]+ 786/// HexFP80Constant 0xK[0-9A-Fa-f]+ 787/// HexFP128Constant 0xL[0-9A-Fa-f]+ 788/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 789lltok::Kind LLLexer::LexDigitOrNegative() { 790 // If the letter after the negative is not a number, this is probably a label. 791 if (!isdigit(static_cast<unsigned char>(TokStart[0])) && 792 !isdigit(static_cast<unsigned char>(CurPtr[0]))) { 793 // Okay, this is not a number after the -, it's probably a label. 794 if (const char *End = isLabelTail(CurPtr)) { 795 StrVal.assign(TokStart, End-1); 796 CurPtr = End; 797 return lltok::LabelStr; 798 } 799 800 return lltok::Error; 801 } 802 803 // At this point, it is either a label, int or fp constant. 804 805 // Skip digits, we have at least one. 806 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 807 /*empty*/; 808 809 // Check to see if this really is a label afterall, e.g. "-1:". 810 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 811 if (const char *End = isLabelTail(CurPtr)) { 812 StrVal.assign(TokStart, End-1); 813 CurPtr = End; 814 return lltok::LabelStr; 815 } 816 } 817 818 // If the next character is a '.', then it is a fp value, otherwise its 819 // integer. 820 if (CurPtr[0] != '.') { 821 if (TokStart[0] == '0' && TokStart[1] == 'x') 822 return Lex0x(); 823 unsigned Len = CurPtr-TokStart; 824 uint32_t numBits = ((Len * 64) / 19) + 2; 825 APInt Tmp(numBits, StringRef(TokStart, Len), 10); 826 if (TokStart[0] == '-') { 827 uint32_t minBits = Tmp.getMinSignedBits(); 828 if (minBits > 0 && minBits < numBits) 829 Tmp = Tmp.trunc(minBits); 830 APSIntVal = APSInt(Tmp, false); 831 } else { 832 uint32_t activeBits = Tmp.getActiveBits(); 833 if (activeBits > 0 && activeBits < numBits) 834 Tmp = Tmp.trunc(activeBits); 835 APSIntVal = APSInt(Tmp, true); 836 } 837 return lltok::APSInt; 838 } 839 840 ++CurPtr; 841 842 // Skip over [0-9]*([eE][-+]?[0-9]+)? 843 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 844 845 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 846 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 847 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 848 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 849 CurPtr += 2; 850 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 851 } 852 } 853 854 APFloatVal = APFloat(std::atof(TokStart)); 855 return lltok::APFloat; 856} 857 858/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 859lltok::Kind LLLexer::LexPositive() { 860 // If the letter after the negative is a number, this is probably not a 861 // label. 862 if (!isdigit(static_cast<unsigned char>(CurPtr[0]))) 863 return lltok::Error; 864 865 // Skip digits. 866 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr) 867 /*empty*/; 868 869 // At this point, we need a '.'. 870 if (CurPtr[0] != '.') { 871 CurPtr = TokStart+1; 872 return lltok::Error; 873 } 874 875 ++CurPtr; 876 877 // Skip over [0-9]*([eE][-+]?[0-9]+)? 878 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 879 880 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 881 if (isdigit(static_cast<unsigned char>(CurPtr[1])) || 882 ((CurPtr[1] == '-' || CurPtr[1] == '+') && 883 isdigit(static_cast<unsigned char>(CurPtr[2])))) { 884 CurPtr += 2; 885 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr; 886 } 887 } 888 889 APFloatVal = APFloat(std::atof(TokStart)); 890 return lltok::APFloat; 891} 892