LLLexer.cpp revision 114baee1fa017daefad2339c77b45b9ca3d79a41
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 && isxdigit(BIn[1]) && isxdigit(BIn[2])) { 123 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]); 124 BIn += 3; // Skip over handled chars 125 ++BOut; 126 } else { 127 *BOut++ = *BIn++; 128 } 129 } else { 130 *BOut++ = *BIn++; 131 } 132 } 133 Str.resize(BOut-Buffer); 134} 135 136/// isLabelChar - Return true for [-a-zA-Z$._0-9]. 137static bool isLabelChar(char C) { 138 return isalnum(C) || C == '-' || C == '$' || C == '.' || C == '_'; 139} 140 141 142/// isLabelTail - Return true if this pointer points to a valid end of a label. 143static const char *isLabelTail(const char *CurPtr) { 144 while (1) { 145 if (CurPtr[0] == ':') return CurPtr+1; 146 if (!isLabelChar(CurPtr[0])) return 0; 147 ++CurPtr; 148 } 149} 150 151 152 153//===----------------------------------------------------------------------===// 154// Lexer definition. 155//===----------------------------------------------------------------------===// 156 157LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err, 158 LLVMContext &C) 159 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) { 160 CurPtr = CurBuf->getBufferStart(); 161} 162 163std::string LLLexer::getFilename() const { 164 return CurBuf->getBufferIdentifier(); 165} 166 167int LLLexer::getNextChar() { 168 char CurChar = *CurPtr++; 169 switch (CurChar) { 170 default: return (unsigned char)CurChar; 171 case 0: 172 // A nul character in the stream is either the end of the current buffer or 173 // a random nul in the file. Disambiguate that here. 174 if (CurPtr-1 != CurBuf->getBufferEnd()) 175 return 0; // Just whitespace. 176 177 // Otherwise, return end of file. 178 --CurPtr; // Another call to lex will return EOF again. 179 return EOF; 180 } 181} 182 183 184lltok::Kind LLLexer::LexToken() { 185 TokStart = CurPtr; 186 187 int CurChar = getNextChar(); 188 switch (CurChar) { 189 default: 190 // Handle letters: [a-zA-Z_] 191 if (isalpha(CurChar) || CurChar == '_') 192 return LexIdentifier(); 193 194 return lltok::Error; 195 case EOF: return lltok::Eof; 196 case 0: 197 case ' ': 198 case '\t': 199 case '\n': 200 case '\r': 201 // Ignore whitespace. 202 return LexToken(); 203 case '+': return LexPositive(); 204 case '@': return LexAt(); 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 '$': 219 if (const char *Ptr = isLabelTail(CurPtr)) { 220 CurPtr = Ptr; 221 StrVal.assign(TokStart, CurPtr-1); 222 return lltok::LabelStr; 223 } 224 return lltok::Error; 225 case ';': 226 SkipLineComment(); 227 return LexToken(); 228 case '!': return LexExclaim(); 229 case '0': case '1': case '2': case '3': case '4': 230 case '5': case '6': case '7': case '8': case '9': 231 case '-': 232 return LexDigitOrNegative(); 233 case '=': return lltok::equal; 234 case '[': return lltok::lsquare; 235 case ']': return lltok::rsquare; 236 case '{': return lltok::lbrace; 237 case '}': return lltok::rbrace; 238 case '<': return lltok::less; 239 case '>': return lltok::greater; 240 case '(': return lltok::lparen; 241 case ')': return lltok::rparen; 242 case ',': return lltok::comma; 243 case '*': return lltok::star; 244 case '\\': return lltok::backslash; 245 } 246} 247 248void LLLexer::SkipLineComment() { 249 while (1) { 250 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 251 return; 252 } 253} 254 255/// LexAt - Lex all tokens that start with an @ character: 256/// GlobalVar @\"[^\"]*\" 257/// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 258/// GlobalVarID @[0-9]+ 259lltok::Kind LLLexer::LexAt() { 260 // Handle AtStringConstant: @\"[^\"]*\" 261 if (CurPtr[0] == '"') { 262 ++CurPtr; 263 264 while (1) { 265 int CurChar = getNextChar(); 266 267 if (CurChar == EOF) { 268 Error("end of file in global variable name"); 269 return lltok::Error; 270 } 271 if (CurChar == '"') { 272 StrVal.assign(TokStart+2, CurPtr-1); 273 UnEscapeLexed(StrVal); 274 return lltok::GlobalVar; 275 } 276 } 277 } 278 279 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]* 280 if (ReadVarName()) 281 return lltok::GlobalVar; 282 283 // Handle GlobalVarID: @[0-9]+ 284 if (isdigit(CurPtr[0])) { 285 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 286 /*empty*/; 287 288 uint64_t Val = atoull(TokStart+1, CurPtr); 289 if ((unsigned)Val != Val) 290 Error("invalid value number (too large)!"); 291 UIntVal = unsigned(Val); 292 return lltok::GlobalID; 293 } 294 295 return lltok::Error; 296} 297 298/// ReadString - Read a string until the closing quote. 299lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 300 const char *Start = CurPtr; 301 while (1) { 302 int CurChar = getNextChar(); 303 304 if (CurChar == EOF) { 305 Error("end of file in string constant"); 306 return lltok::Error; 307 } 308 if (CurChar == '"') { 309 StrVal.assign(Start, CurPtr-1); 310 UnEscapeLexed(StrVal); 311 return kind; 312 } 313 } 314} 315 316/// ReadVarName - Read the rest of a token containing a variable name. 317bool LLLexer::ReadVarName() { 318 const char *NameStart = CurPtr; 319 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 320 CurPtr[0] == '.' || CurPtr[0] == '_') { 321 ++CurPtr; 322 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 323 CurPtr[0] == '.' || CurPtr[0] == '_') 324 ++CurPtr; 325 326 StrVal.assign(NameStart, CurPtr); 327 return true; 328 } 329 return false; 330} 331 332/// LexPercent - Lex all tokens that start with a % character: 333/// LocalVar ::= %\"[^\"]*\" 334/// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 335/// LocalVarID ::= %[0-9]+ 336lltok::Kind LLLexer::LexPercent() { 337 // Handle LocalVarName: %\"[^\"]*\" 338 if (CurPtr[0] == '"') { 339 ++CurPtr; 340 return ReadString(lltok::LocalVar); 341 } 342 343 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]* 344 if (ReadVarName()) 345 return lltok::LocalVar; 346 347 // Handle LocalVarID: %[0-9]+ 348 if (isdigit(CurPtr[0])) { 349 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 350 /*empty*/; 351 352 uint64_t Val = atoull(TokStart+1, CurPtr); 353 if ((unsigned)Val != Val) 354 Error("invalid value number (too large)!"); 355 UIntVal = unsigned(Val); 356 return lltok::LocalVarID; 357 } 358 359 return lltok::Error; 360} 361 362/// LexQuote - Lex all tokens that start with a " character: 363/// QuoteLabel "[^"]+": 364/// StringConstant "[^"]*" 365lltok::Kind LLLexer::LexQuote() { 366 lltok::Kind kind = ReadString(lltok::StringConstant); 367 if (kind == lltok::Error || kind == lltok::Eof) 368 return kind; 369 370 if (CurPtr[0] == ':') { 371 ++CurPtr; 372 kind = lltok::LabelStr; 373 } 374 375 return kind; 376} 377 378/// LexExclaim: 379/// !foo 380/// ! 381lltok::Kind LLLexer::LexExclaim() { 382 // Lex a metadata name as a MetadataVar. 383 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 384 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 385 ++CurPtr; 386 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 387 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 388 ++CurPtr; 389 390 StrVal.assign(TokStart+1, CurPtr); // Skip ! 391 UnEscapeLexed(StrVal); 392 return lltok::MetadataVar; 393 } 394 return lltok::exclaim; 395} 396 397/// LexIdentifier: Handle several related productions: 398/// Label [-a-zA-Z$._0-9]+: 399/// IntegerType i[0-9]+ 400/// Keyword sdiv, float, ... 401/// HexIntConstant [us]0x[0-9A-Fa-f]+ 402lltok::Kind LLLexer::LexIdentifier() { 403 const char *StartChar = CurPtr; 404 const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar; 405 const char *KeywordEnd = 0; 406 407 for (; isLabelChar(*CurPtr); ++CurPtr) { 408 // If we decide this is an integer, remember the end of the sequence. 409 if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr; 410 if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr; 411 } 412 413 // If we stopped due to a colon, this really is a label. 414 if (*CurPtr == ':') { 415 StrVal.assign(StartChar-1, CurPtr++); 416 return lltok::LabelStr; 417 } 418 419 // Otherwise, this wasn't a label. If this was valid as an integer type, 420 // return it. 421 if (IntEnd == 0) IntEnd = CurPtr; 422 if (IntEnd != StartChar) { 423 CurPtr = IntEnd; 424 uint64_t NumBits = atoull(StartChar, CurPtr); 425 if (NumBits < IntegerType::MIN_INT_BITS || 426 NumBits > IntegerType::MAX_INT_BITS) { 427 Error("bitwidth for integer type out of range!"); 428 return lltok::Error; 429 } 430 TyVal = IntegerType::get(Context, NumBits); 431 return lltok::Type; 432 } 433 434 // Otherwise, this was a letter sequence. See which keyword this is. 435 if (KeywordEnd == 0) KeywordEnd = CurPtr; 436 CurPtr = KeywordEnd; 437 --StartChar; 438 unsigned Len = CurPtr-StartChar; 439#define KEYWORD(STR) \ 440 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \ 441 return lltok::kw_##STR; 442 443 KEYWORD(true); KEYWORD(false); 444 KEYWORD(declare); KEYWORD(define); 445 KEYWORD(global); KEYWORD(constant); 446 447 KEYWORD(private); 448 KEYWORD(linker_private); 449 KEYWORD(linker_private_weak); 450 KEYWORD(linker_private_weak_def_auto); // FIXME: For backwards compatibility. 451 KEYWORD(internal); 452 KEYWORD(available_externally); 453 KEYWORD(linkonce); 454 KEYWORD(linkonce_odr); 455 KEYWORD(linkonce_odr_auto_hide); 456 KEYWORD(weak); 457 KEYWORD(weak_odr); 458 KEYWORD(appending); 459 KEYWORD(dllimport); 460 KEYWORD(dllexport); 461 KEYWORD(common); 462 KEYWORD(default); 463 KEYWORD(hidden); 464 KEYWORD(protected); 465 KEYWORD(unnamed_addr); 466 KEYWORD(extern_weak); 467 KEYWORD(external); 468 KEYWORD(thread_local); 469 KEYWORD(localdynamic); 470 KEYWORD(initialexec); 471 KEYWORD(localexec); 472 KEYWORD(zeroinitializer); 473 KEYWORD(undef); 474 KEYWORD(null); 475 KEYWORD(to); 476 KEYWORD(tail); 477 KEYWORD(target); 478 KEYWORD(triple); 479 KEYWORD(unwind); 480 KEYWORD(deplibs); // FIXME: Remove in 4.0. 481 KEYWORD(datalayout); 482 KEYWORD(volatile); 483 KEYWORD(atomic); 484 KEYWORD(unordered); 485 KEYWORD(monotonic); 486 KEYWORD(acquire); 487 KEYWORD(release); 488 KEYWORD(acq_rel); 489 KEYWORD(seq_cst); 490 KEYWORD(singlethread); 491 492 KEYWORD(nnan) 493 KEYWORD(ninf) 494 KEYWORD(nsz) 495 KEYWORD(arcp) 496 KEYWORD(fast) 497 KEYWORD(nuw); 498 KEYWORD(nsw); 499 KEYWORD(exact); 500 KEYWORD(inbounds); 501 KEYWORD(align); 502 KEYWORD(addrspace); 503 KEYWORD(section); 504 KEYWORD(alias); 505 KEYWORD(module); 506 KEYWORD(asm); 507 KEYWORD(sideeffect); 508 KEYWORD(alignstack); 509 KEYWORD(inteldialect); 510 KEYWORD(gc); 511 512 KEYWORD(ccc); 513 KEYWORD(fastcc); 514 KEYWORD(coldcc); 515 KEYWORD(x86_stdcallcc); 516 KEYWORD(x86_fastcallcc); 517 KEYWORD(x86_thiscallcc); 518 KEYWORD(arm_apcscc); 519 KEYWORD(arm_aapcscc); 520 KEYWORD(arm_aapcs_vfpcc); 521 KEYWORD(msp430_intrcc); 522 KEYWORD(ptx_kernel); 523 KEYWORD(ptx_device); 524 KEYWORD(spir_kernel); 525 KEYWORD(spir_func); 526 KEYWORD(intel_ocl_bicc); 527 528 KEYWORD(cc); 529 KEYWORD(c); 530 531 KEYWORD(signext); 532 KEYWORD(zeroext); 533 KEYWORD(inreg); 534 KEYWORD(sret); 535 KEYWORD(nounwind); 536 KEYWORD(noreturn); 537 KEYWORD(noalias); 538 KEYWORD(nocapture); 539 KEYWORD(byval); 540 KEYWORD(nest); 541 KEYWORD(readnone); 542 KEYWORD(readonly); 543 KEYWORD(uwtable); 544 KEYWORD(returns_twice); 545 546 KEYWORD(inlinehint); 547 KEYWORD(noinline); 548 KEYWORD(alwaysinline); 549 KEYWORD(optsize); 550 KEYWORD(ssp); 551 KEYWORD(sspreq); 552 KEYWORD(sspstrong); 553 KEYWORD(noredzone); 554 KEYWORD(noimplicitfloat); 555 KEYWORD(naked); 556 KEYWORD(nonlazybind); 557 KEYWORD(address_safety); 558 KEYWORD(minsize); 559 KEYWORD(noduplicate); 560 561 KEYWORD(type); 562 KEYWORD(opaque); 563 564 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 565 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 566 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 567 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 568 569 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 570 KEYWORD(umin); 571 572 KEYWORD(x); 573 KEYWORD(blockaddress); 574 575 KEYWORD(personality); 576 KEYWORD(cleanup); 577 KEYWORD(catch); 578 KEYWORD(filter); 579#undef KEYWORD 580 581 // Keywords for types. 582#define TYPEKEYWORD(STR, LLVMTY) \ 583 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \ 584 TyVal = LLVMTY; return lltok::Type; } 585 TYPEKEYWORD("void", Type::getVoidTy(Context)); 586 TYPEKEYWORD("half", Type::getHalfTy(Context)); 587 TYPEKEYWORD("float", Type::getFloatTy(Context)); 588 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 589 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 590 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 591 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 592 TYPEKEYWORD("label", Type::getLabelTy(Context)); 593 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 594 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 595#undef TYPEKEYWORD 596 597 // Keywords for instructions. 598#define INSTKEYWORD(STR, Enum) \ 599 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \ 600 UIntVal = Instruction::Enum; return lltok::kw_##STR; } 601 602 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 603 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 604 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 605 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 606 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 607 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 608 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 609 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 610 611 INSTKEYWORD(phi, PHI); 612 INSTKEYWORD(call, Call); 613 INSTKEYWORD(trunc, Trunc); 614 INSTKEYWORD(zext, ZExt); 615 INSTKEYWORD(sext, SExt); 616 INSTKEYWORD(fptrunc, FPTrunc); 617 INSTKEYWORD(fpext, FPExt); 618 INSTKEYWORD(uitofp, UIToFP); 619 INSTKEYWORD(sitofp, SIToFP); 620 INSTKEYWORD(fptoui, FPToUI); 621 INSTKEYWORD(fptosi, FPToSI); 622 INSTKEYWORD(inttoptr, IntToPtr); 623 INSTKEYWORD(ptrtoint, PtrToInt); 624 INSTKEYWORD(bitcast, BitCast); 625 INSTKEYWORD(select, Select); 626 INSTKEYWORD(va_arg, VAArg); 627 INSTKEYWORD(ret, Ret); 628 INSTKEYWORD(br, Br); 629 INSTKEYWORD(switch, Switch); 630 INSTKEYWORD(indirectbr, IndirectBr); 631 INSTKEYWORD(invoke, Invoke); 632 INSTKEYWORD(resume, Resume); 633 INSTKEYWORD(unreachable, Unreachable); 634 635 INSTKEYWORD(alloca, Alloca); 636 INSTKEYWORD(load, Load); 637 INSTKEYWORD(store, Store); 638 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 639 INSTKEYWORD(atomicrmw, AtomicRMW); 640 INSTKEYWORD(fence, Fence); 641 INSTKEYWORD(getelementptr, GetElementPtr); 642 643 INSTKEYWORD(extractelement, ExtractElement); 644 INSTKEYWORD(insertelement, InsertElement); 645 INSTKEYWORD(shufflevector, ShuffleVector); 646 INSTKEYWORD(extractvalue, ExtractValue); 647 INSTKEYWORD(insertvalue, InsertValue); 648 INSTKEYWORD(landingpad, LandingPad); 649#undef INSTKEYWORD 650 651 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 652 // the CFE to avoid forcing it to deal with 64-bit numbers. 653 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 654 TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) { 655 int len = CurPtr-TokStart-3; 656 uint32_t bits = len * 4; 657 APInt Tmp(bits, StringRef(TokStart+3, len), 16); 658 uint32_t activeBits = Tmp.getActiveBits(); 659 if (activeBits > 0 && activeBits < bits) 660 Tmp = Tmp.trunc(activeBits); 661 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 662 return lltok::APSInt; 663 } 664 665 // If this is "cc1234", return this as just "cc". 666 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 667 CurPtr = TokStart+2; 668 return lltok::kw_cc; 669 } 670 671 // Finally, if this isn't known, return an error. 672 CurPtr = TokStart+1; 673 return lltok::Error; 674} 675 676 677/// Lex0x: Handle productions that start with 0x, knowing that it matches and 678/// that this is not a label: 679/// HexFPConstant 0x[0-9A-Fa-f]+ 680/// HexFP80Constant 0xK[0-9A-Fa-f]+ 681/// HexFP128Constant 0xL[0-9A-Fa-f]+ 682/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 683/// HexHalfConstant 0xH[0-9A-Fa-f]+ 684lltok::Kind LLLexer::Lex0x() { 685 CurPtr = TokStart + 2; 686 687 char Kind; 688 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') { 689 Kind = *CurPtr++; 690 } else { 691 Kind = 'J'; 692 } 693 694 if (!isxdigit(CurPtr[0])) { 695 // Bad token, return it as an error. 696 CurPtr = TokStart+1; 697 return lltok::Error; 698 } 699 700 while (isxdigit(CurPtr[0])) 701 ++CurPtr; 702 703 if (Kind == 'J') { 704 // HexFPConstant - Floating point constant represented in IEEE format as a 705 // hexadecimal number for when exponential notation is not precise enough. 706 // Half, Float, and double only. 707 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr))); 708 return lltok::APFloat; 709 } 710 711 uint64_t Pair[2]; 712 switch (Kind) { 713 default: llvm_unreachable("Unknown kind!"); 714 case 'K': 715 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 716 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 717 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair)); 718 return lltok::APFloat; 719 case 'L': 720 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 721 HexToIntPair(TokStart+3, CurPtr, Pair); 722 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair)); 723 return lltok::APFloat; 724 case 'M': 725 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 726 HexToIntPair(TokStart+3, CurPtr, Pair); 727 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair)); 728 return lltok::APFloat; 729 case 'H': 730 APFloatVal = APFloat(APFloat::IEEEhalf, 731 APInt(16,HexIntToVal(TokStart+3, CurPtr))); 732 return lltok::APFloat; 733 } 734} 735 736/// LexIdentifier: Handle several related productions: 737/// Label [-a-zA-Z$._0-9]+: 738/// NInteger -[0-9]+ 739/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 740/// PInteger [0-9]+ 741/// HexFPConstant 0x[0-9A-Fa-f]+ 742/// HexFP80Constant 0xK[0-9A-Fa-f]+ 743/// HexFP128Constant 0xL[0-9A-Fa-f]+ 744/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 745lltok::Kind LLLexer::LexDigitOrNegative() { 746 // If the letter after the negative is not a number, this is probably a label. 747 if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) { 748 // Okay, this is not a number after the -, it's probably a label. 749 if (const char *End = isLabelTail(CurPtr)) { 750 StrVal.assign(TokStart, End-1); 751 CurPtr = End; 752 return lltok::LabelStr; 753 } 754 755 return lltok::Error; 756 } 757 758 // At this point, it is either a label, int or fp constant. 759 760 // Skip digits, we have at least one. 761 for (; isdigit(CurPtr[0]); ++CurPtr) 762 /*empty*/; 763 764 // Check to see if this really is a label afterall, e.g. "-1:". 765 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 766 if (const char *End = isLabelTail(CurPtr)) { 767 StrVal.assign(TokStart, End-1); 768 CurPtr = End; 769 return lltok::LabelStr; 770 } 771 } 772 773 // If the next character is a '.', then it is a fp value, otherwise its 774 // integer. 775 if (CurPtr[0] != '.') { 776 if (TokStart[0] == '0' && TokStart[1] == 'x') 777 return Lex0x(); 778 unsigned Len = CurPtr-TokStart; 779 uint32_t numBits = ((Len * 64) / 19) + 2; 780 APInt Tmp(numBits, StringRef(TokStart, Len), 10); 781 if (TokStart[0] == '-') { 782 uint32_t minBits = Tmp.getMinSignedBits(); 783 if (minBits > 0 && minBits < numBits) 784 Tmp = Tmp.trunc(minBits); 785 APSIntVal = APSInt(Tmp, false); 786 } else { 787 uint32_t activeBits = Tmp.getActiveBits(); 788 if (activeBits > 0 && activeBits < numBits) 789 Tmp = Tmp.trunc(activeBits); 790 APSIntVal = APSInt(Tmp, true); 791 } 792 return lltok::APSInt; 793 } 794 795 ++CurPtr; 796 797 // Skip over [0-9]*([eE][-+]?[0-9]+)? 798 while (isdigit(CurPtr[0])) ++CurPtr; 799 800 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 801 if (isdigit(CurPtr[1]) || 802 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) { 803 CurPtr += 2; 804 while (isdigit(CurPtr[0])) ++CurPtr; 805 } 806 } 807 808 APFloatVal = APFloat(std::atof(TokStart)); 809 return lltok::APFloat; 810} 811 812/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 813lltok::Kind LLLexer::LexPositive() { 814 // If the letter after the negative is a number, this is probably not a 815 // label. 816 if (!isdigit(CurPtr[0])) 817 return lltok::Error; 818 819 // Skip digits. 820 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 821 /*empty*/; 822 823 // At this point, we need a '.'. 824 if (CurPtr[0] != '.') { 825 CurPtr = TokStart+1; 826 return lltok::Error; 827 } 828 829 ++CurPtr; 830 831 // Skip over [0-9]*([eE][-+]?[0-9]+)? 832 while (isdigit(CurPtr[0])) ++CurPtr; 833 834 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 835 if (isdigit(CurPtr[1]) || 836 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) { 837 CurPtr += 2; 838 while (isdigit(CurPtr[0])) ++CurPtr; 839 } 840 } 841 842 APFloatVal = APFloat(std::atof(TokStart)); 843 return lltok::APFloat; 844} 845