LLLexer.cpp revision 3f2d5f60b31fd057c10f77b2e607b23a8c94f6d3
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/DerivedTypes.h" 16#include "llvm/Instruction.h" 17#include "llvm/LLVMContext.h" 18#include "llvm/ADT/Twine.h" 19#include "llvm/Assembly/Parser.h" 20#include "llvm/Support/ErrorHandling.h" 21#include "llvm/Support/MemoryBuffer.h" 22#include "llvm/Support/MathExtras.h" 23#include "llvm/Support/SourceMgr.h" 24#include "llvm/Support/raw_ostream.h" 25#include <cctype> 26#include <cstdio> 27#include <cstdlib> 28#include <cstring> 29using namespace llvm; 30 31bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const { 32 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg); 33 return true; 34} 35 36//===----------------------------------------------------------------------===// 37// Helper functions. 38//===----------------------------------------------------------------------===// 39 40// atoull - Convert an ascii string of decimal digits into the unsigned long 41// long representation... this does not have to do input error checking, 42// because we know that the input will be matched by a suitable regex... 43// 44uint64_t LLLexer::atoull(const char *Buffer, const char *End) { 45 uint64_t Result = 0; 46 for (; Buffer != End; Buffer++) { 47 uint64_t OldRes = Result; 48 Result *= 10; 49 Result += *Buffer-'0'; 50 if (Result < OldRes) { // Uh, oh, overflow detected!!! 51 Error("constant bigger than 64 bits detected!"); 52 return 0; 53 } 54 } 55 return Result; 56} 57 58uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) { 59 uint64_t Result = 0; 60 for (; Buffer != End; ++Buffer) { 61 uint64_t OldRes = Result; 62 Result *= 16; 63 char C = *Buffer; 64 if (C >= '0' && C <= '9') 65 Result += C-'0'; 66 else if (C >= 'A' && C <= 'F') 67 Result += C-'A'+10; 68 else if (C >= 'a' && C <= 'f') 69 Result += C-'a'+10; 70 71 if (Result < OldRes) { // Uh, oh, overflow detected!!! 72 Error("constant bigger than 64 bits detected!"); 73 return 0; 74 } 75 } 76 return Result; 77} 78 79void LLLexer::HexToIntPair(const char *Buffer, const char *End, 80 uint64_t Pair[2]) { 81 Pair[0] = 0; 82 for (int i=0; i<16; i++, Buffer++) { 83 assert(Buffer != End); 84 Pair[0] *= 16; 85 char C = *Buffer; 86 if (C >= '0' && C <= '9') 87 Pair[0] += C-'0'; 88 else if (C >= 'A' && C <= 'F') 89 Pair[0] += C-'A'+10; 90 else if (C >= 'a' && C <= 'f') 91 Pair[0] += C-'a'+10; 92 } 93 Pair[1] = 0; 94 for (int i=0; i<16 && Buffer != End; i++, Buffer++) { 95 Pair[1] *= 16; 96 char C = *Buffer; 97 if (C >= '0' && C <= '9') 98 Pair[1] += C-'0'; 99 else if (C >= 'A' && C <= 'F') 100 Pair[1] += C-'A'+10; 101 else if (C >= 'a' && C <= 'f') 102 Pair[1] += C-'a'+10; 103 } 104 if (Buffer != End) 105 Error("constant bigger than 128 bits detected!"); 106} 107 108/// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into 109/// { low64, high16 } as usual for an APInt. 110void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End, 111 uint64_t Pair[2]) { 112 Pair[1] = 0; 113 for (int i=0; i<4 && Buffer != End; i++, Buffer++) { 114 assert(Buffer != End); 115 Pair[1] *= 16; 116 char C = *Buffer; 117 if (C >= '0' && C <= '9') 118 Pair[1] += C-'0'; 119 else if (C >= 'A' && C <= 'F') 120 Pair[1] += C-'A'+10; 121 else if (C >= 'a' && C <= 'f') 122 Pair[1] += C-'a'+10; 123 } 124 Pair[0] = 0; 125 for (int i=0; i<16; i++, Buffer++) { 126 Pair[0] *= 16; 127 char C = *Buffer; 128 if (C >= '0' && C <= '9') 129 Pair[0] += C-'0'; 130 else if (C >= 'A' && C <= 'F') 131 Pair[0] += C-'A'+10; 132 else if (C >= 'a' && C <= 'f') 133 Pair[0] += C-'a'+10; 134 } 135 if (Buffer != End) 136 Error("constant bigger than 128 bits detected!"); 137} 138 139// UnEscapeLexed - Run through the specified buffer and change \xx codes to the 140// appropriate character. 141static void UnEscapeLexed(std::string &Str) { 142 if (Str.empty()) return; 143 144 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size(); 145 char *BOut = Buffer; 146 for (char *BIn = Buffer; BIn != EndBuffer; ) { 147 if (BIn[0] == '\\') { 148 if (BIn < EndBuffer-1 && BIn[1] == '\\') { 149 *BOut++ = '\\'; // Two \ becomes one 150 BIn += 2; 151 } else if (BIn < EndBuffer-2 && isxdigit(BIn[1]) && isxdigit(BIn[2])) { 152 char Tmp = BIn[3]; BIn[3] = 0; // Terminate string 153 *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number 154 BIn[3] = Tmp; // Restore character 155 BIn += 3; // Skip over handled chars 156 ++BOut; 157 } else { 158 *BOut++ = *BIn++; 159 } 160 } else { 161 *BOut++ = *BIn++; 162 } 163 } 164 Str.resize(BOut-Buffer); 165} 166 167/// isLabelChar - Return true for [-a-zA-Z$._0-9]. 168static bool isLabelChar(char C) { 169 return isalnum(C) || C == '-' || C == '$' || C == '.' || C == '_'; 170} 171 172 173/// isLabelTail - Return true if this pointer points to a valid end of a label. 174static const char *isLabelTail(const char *CurPtr) { 175 while (1) { 176 if (CurPtr[0] == ':') return CurPtr+1; 177 if (!isLabelChar(CurPtr[0])) return 0; 178 ++CurPtr; 179 } 180} 181 182 183 184//===----------------------------------------------------------------------===// 185// Lexer definition. 186//===----------------------------------------------------------------------===// 187 188LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err, 189 LLVMContext &C) 190 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) { 191 CurPtr = CurBuf->getBufferStart(); 192} 193 194std::string LLLexer::getFilename() const { 195 return CurBuf->getBufferIdentifier(); 196} 197 198int LLLexer::getNextChar() { 199 char CurChar = *CurPtr++; 200 switch (CurChar) { 201 default: return (unsigned char)CurChar; 202 case 0: 203 // A nul character in the stream is either the end of the current buffer or 204 // a random nul in the file. Disambiguate that here. 205 if (CurPtr-1 != CurBuf->getBufferEnd()) 206 return 0; // Just whitespace. 207 208 // Otherwise, return end of file. 209 --CurPtr; // Another call to lex will return EOF again. 210 return EOF; 211 } 212} 213 214 215lltok::Kind LLLexer::LexToken() { 216 TokStart = CurPtr; 217 218 int CurChar = getNextChar(); 219 switch (CurChar) { 220 default: 221 // Handle letters: [a-zA-Z_] 222 if (isalpha(CurChar) || CurChar == '_') 223 return LexIdentifier(); 224 225 return lltok::Error; 226 case EOF: return lltok::Eof; 227 case 0: 228 case ' ': 229 case '\t': 230 case '\n': 231 case '\r': 232 // Ignore whitespace. 233 return LexToken(); 234 case '+': return LexPositive(); 235 case '@': return LexAt(); 236 case '%': return LexPercent(); 237 case '"': return LexQuote(); 238 case '.': 239 if (const char *Ptr = isLabelTail(CurPtr)) { 240 CurPtr = Ptr; 241 StrVal.assign(TokStart, CurPtr-1); 242 return lltok::LabelStr; 243 } 244 if (CurPtr[0] == '.' && CurPtr[1] == '.') { 245 CurPtr += 2; 246 return lltok::dotdotdot; 247 } 248 return lltok::Error; 249 case '$': 250 if (const char *Ptr = isLabelTail(CurPtr)) { 251 CurPtr = Ptr; 252 StrVal.assign(TokStart, CurPtr-1); 253 return lltok::LabelStr; 254 } 255 return lltok::Error; 256 case ';': 257 SkipLineComment(); 258 return LexToken(); 259 case '!': return LexExclaim(); 260 case '0': case '1': case '2': case '3': case '4': 261 case '5': case '6': case '7': case '8': case '9': 262 case '-': 263 return LexDigitOrNegative(); 264 case '=': return lltok::equal; 265 case '[': return lltok::lsquare; 266 case ']': return lltok::rsquare; 267 case '{': return lltok::lbrace; 268 case '}': return lltok::rbrace; 269 case '<': return lltok::less; 270 case '>': return lltok::greater; 271 case '(': return lltok::lparen; 272 case ')': return lltok::rparen; 273 case ',': return lltok::comma; 274 case '*': return lltok::star; 275 case '\\': return lltok::backslash; 276 } 277} 278 279void LLLexer::SkipLineComment() { 280 while (1) { 281 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF) 282 return; 283 } 284} 285 286/// LexAt - Lex all tokens that start with an @ character: 287/// GlobalVar @\"[^\"]*\" 288/// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]* 289/// GlobalVarID @[0-9]+ 290lltok::Kind LLLexer::LexAt() { 291 // Handle AtStringConstant: @\"[^\"]*\" 292 if (CurPtr[0] == '"') { 293 ++CurPtr; 294 295 while (1) { 296 int CurChar = getNextChar(); 297 298 if (CurChar == EOF) { 299 Error("end of file in global variable name"); 300 return lltok::Error; 301 } 302 if (CurChar == '"') { 303 StrVal.assign(TokStart+2, CurPtr-1); 304 UnEscapeLexed(StrVal); 305 return lltok::GlobalVar; 306 } 307 } 308 } 309 310 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]* 311 if (ReadVarName()) 312 return lltok::GlobalVar; 313 314 // Handle GlobalVarID: @[0-9]+ 315 if (isdigit(CurPtr[0])) { 316 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 317 /*empty*/; 318 319 uint64_t Val = atoull(TokStart+1, CurPtr); 320 if ((unsigned)Val != Val) 321 Error("invalid value number (too large)!"); 322 UIntVal = unsigned(Val); 323 return lltok::GlobalID; 324 } 325 326 return lltok::Error; 327} 328 329/// ReadString - Read a string until the closing quote. 330lltok::Kind LLLexer::ReadString(lltok::Kind kind) { 331 const char *Start = CurPtr; 332 while (1) { 333 int CurChar = getNextChar(); 334 335 if (CurChar == EOF) { 336 Error("end of file in string constant"); 337 return lltok::Error; 338 } 339 if (CurChar == '"') { 340 StrVal.assign(Start, CurPtr-1); 341 UnEscapeLexed(StrVal); 342 return kind; 343 } 344 } 345} 346 347/// ReadVarName - Read the rest of a token containing a variable name. 348bool LLLexer::ReadVarName() { 349 const char *NameStart = CurPtr; 350 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 351 CurPtr[0] == '.' || CurPtr[0] == '_') { 352 ++CurPtr; 353 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 354 CurPtr[0] == '.' || CurPtr[0] == '_') 355 ++CurPtr; 356 357 StrVal.assign(NameStart, CurPtr); 358 return true; 359 } 360 return false; 361} 362 363/// LexPercent - Lex all tokens that start with a % character: 364/// LocalVar ::= %\"[^\"]*\" 365/// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 366/// LocalVarID ::= %[0-9]+ 367lltok::Kind LLLexer::LexPercent() { 368 // Handle LocalVarName: %\"[^\"]*\" 369 if (CurPtr[0] == '"') { 370 ++CurPtr; 371 return ReadString(lltok::LocalVar); 372 } 373 374 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]* 375 if (ReadVarName()) 376 return lltok::LocalVar; 377 378 // Handle LocalVarID: %[0-9]+ 379 if (isdigit(CurPtr[0])) { 380 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 381 /*empty*/; 382 383 uint64_t Val = atoull(TokStart+1, CurPtr); 384 if ((unsigned)Val != Val) 385 Error("invalid value number (too large)!"); 386 UIntVal = unsigned(Val); 387 return lltok::LocalVarID; 388 } 389 390 return lltok::Error; 391} 392 393/// LexQuote - Lex all tokens that start with a " character: 394/// QuoteLabel "[^"]+": 395/// StringConstant "[^"]*" 396lltok::Kind LLLexer::LexQuote() { 397 lltok::Kind kind = ReadString(lltok::StringConstant); 398 if (kind == lltok::Error || kind == lltok::Eof) 399 return kind; 400 401 if (CurPtr[0] == ':') { 402 ++CurPtr; 403 kind = lltok::LabelStr; 404 } 405 406 return kind; 407} 408 409/// LexExclaim: 410/// !foo 411/// ! 412lltok::Kind LLLexer::LexExclaim() { 413 // Lex a metadata name as a MetadataVar. 414 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 415 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') { 416 ++CurPtr; 417 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 418 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') 419 ++CurPtr; 420 421 StrVal.assign(TokStart+1, CurPtr); // Skip ! 422 UnEscapeLexed(StrVal); 423 return lltok::MetadataVar; 424 } 425 return lltok::exclaim; 426} 427 428/// LexIdentifier: Handle several related productions: 429/// Label [-a-zA-Z$._0-9]+: 430/// IntegerType i[0-9]+ 431/// Keyword sdiv, float, ... 432/// HexIntConstant [us]0x[0-9A-Fa-f]+ 433lltok::Kind LLLexer::LexIdentifier() { 434 const char *StartChar = CurPtr; 435 const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar; 436 const char *KeywordEnd = 0; 437 438 for (; isLabelChar(*CurPtr); ++CurPtr) { 439 // If we decide this is an integer, remember the end of the sequence. 440 if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr; 441 if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr; 442 } 443 444 // If we stopped due to a colon, this really is a label. 445 if (*CurPtr == ':') { 446 StrVal.assign(StartChar-1, CurPtr++); 447 return lltok::LabelStr; 448 } 449 450 // Otherwise, this wasn't a label. If this was valid as an integer type, 451 // return it. 452 if (IntEnd == 0) IntEnd = CurPtr; 453 if (IntEnd != StartChar) { 454 CurPtr = IntEnd; 455 uint64_t NumBits = atoull(StartChar, CurPtr); 456 if (NumBits < IntegerType::MIN_INT_BITS || 457 NumBits > IntegerType::MAX_INT_BITS) { 458 Error("bitwidth for integer type out of range!"); 459 return lltok::Error; 460 } 461 TyVal = IntegerType::get(Context, NumBits); 462 return lltok::Type; 463 } 464 465 // Otherwise, this was a letter sequence. See which keyword this is. 466 if (KeywordEnd == 0) KeywordEnd = CurPtr; 467 CurPtr = KeywordEnd; 468 --StartChar; 469 unsigned Len = CurPtr-StartChar; 470#define KEYWORD(STR) \ 471 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \ 472 return lltok::kw_##STR; 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); 482 KEYWORD(internal); 483 KEYWORD(available_externally); 484 KEYWORD(linkonce); 485 KEYWORD(linkonce_odr); 486 KEYWORD(weak); 487 KEYWORD(weak_odr); 488 KEYWORD(appending); 489 KEYWORD(dllimport); 490 KEYWORD(dllexport); 491 KEYWORD(common); 492 KEYWORD(default); 493 KEYWORD(hidden); 494 KEYWORD(protected); 495 KEYWORD(unnamed_addr); 496 KEYWORD(extern_weak); 497 KEYWORD(external); 498 KEYWORD(thread_local); 499 KEYWORD(zeroinitializer); 500 KEYWORD(undef); 501 KEYWORD(null); 502 KEYWORD(to); 503 KEYWORD(tail); 504 KEYWORD(target); 505 KEYWORD(triple); 506 KEYWORD(deplibs); 507 KEYWORD(datalayout); 508 KEYWORD(volatile); 509 KEYWORD(atomic); 510 KEYWORD(unordered); 511 KEYWORD(monotonic); 512 KEYWORD(acquire); 513 KEYWORD(release); 514 KEYWORD(acq_rel); 515 KEYWORD(seq_cst); 516 KEYWORD(singlethread); 517 518 KEYWORD(nuw); 519 KEYWORD(nsw); 520 KEYWORD(exact); 521 KEYWORD(inbounds); 522 KEYWORD(align); 523 KEYWORD(addrspace); 524 KEYWORD(section); 525 KEYWORD(alias); 526 KEYWORD(module); 527 KEYWORD(asm); 528 KEYWORD(sideeffect); 529 KEYWORD(alignstack); 530 KEYWORD(gc); 531 532 KEYWORD(ccc); 533 KEYWORD(fastcc); 534 KEYWORD(coldcc); 535 KEYWORD(x86_stdcallcc); 536 KEYWORD(x86_fastcallcc); 537 KEYWORD(x86_thiscallcc); 538 KEYWORD(arm_apcscc); 539 KEYWORD(arm_aapcscc); 540 KEYWORD(arm_aapcs_vfpcc); 541 KEYWORD(msp430_intrcc); 542 KEYWORD(ptx_kernel); 543 KEYWORD(ptx_device); 544 545 KEYWORD(cc); 546 KEYWORD(c); 547 548 KEYWORD(signext); 549 KEYWORD(zeroext); 550 KEYWORD(inreg); 551 KEYWORD(sret); 552 KEYWORD(nounwind); 553 KEYWORD(noreturn); 554 KEYWORD(noalias); 555 KEYWORD(nocapture); 556 KEYWORD(byval); 557 KEYWORD(nest); 558 KEYWORD(readnone); 559 KEYWORD(readonly); 560 KEYWORD(uwtable); 561 KEYWORD(returns_twice); 562 563 KEYWORD(inlinehint); 564 KEYWORD(noinline); 565 KEYWORD(alwaysinline); 566 KEYWORD(optsize); 567 KEYWORD(ssp); 568 KEYWORD(sspreq); 569 KEYWORD(noredzone); 570 KEYWORD(noimplicitfloat); 571 KEYWORD(naked); 572 KEYWORD(nonlazybind); 573 574 KEYWORD(type); 575 KEYWORD(opaque); 576 577 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 578 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 579 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 580 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 581 582 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax); 583 KEYWORD(umin); 584 585 KEYWORD(x); 586 KEYWORD(blockaddress); 587 588 KEYWORD(personality); 589 KEYWORD(cleanup); 590 KEYWORD(catch); 591 KEYWORD(filter); 592#undef KEYWORD 593 594 // Keywords for types. 595#define TYPEKEYWORD(STR, LLVMTY) \ 596 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \ 597 TyVal = LLVMTY; return lltok::Type; } 598 TYPEKEYWORD("void", Type::getVoidTy(Context)); 599 TYPEKEYWORD("float", Type::getFloatTy(Context)); 600 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 601 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 602 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 603 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 604 TYPEKEYWORD("label", Type::getLabelTy(Context)); 605 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 606 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 607#undef TYPEKEYWORD 608 609 // Keywords for instructions. 610#define INSTKEYWORD(STR, Enum) \ 611 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \ 612 UIntVal = Instruction::Enum; return lltok::kw_##STR; } 613 614 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 615 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 616 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 617 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 618 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 619 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 620 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 621 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 622 623 INSTKEYWORD(phi, PHI); 624 INSTKEYWORD(call, Call); 625 INSTKEYWORD(trunc, Trunc); 626 INSTKEYWORD(zext, ZExt); 627 INSTKEYWORD(sext, SExt); 628 INSTKEYWORD(fptrunc, FPTrunc); 629 INSTKEYWORD(fpext, FPExt); 630 INSTKEYWORD(uitofp, UIToFP); 631 INSTKEYWORD(sitofp, SIToFP); 632 INSTKEYWORD(fptoui, FPToUI); 633 INSTKEYWORD(fptosi, FPToSI); 634 INSTKEYWORD(inttoptr, IntToPtr); 635 INSTKEYWORD(ptrtoint, PtrToInt); 636 INSTKEYWORD(bitcast, BitCast); 637 INSTKEYWORD(select, Select); 638 INSTKEYWORD(va_arg, VAArg); 639 INSTKEYWORD(ret, Ret); 640 INSTKEYWORD(br, Br); 641 INSTKEYWORD(switch, Switch); 642 INSTKEYWORD(indirectbr, IndirectBr); 643 INSTKEYWORD(invoke, Invoke); 644 INSTKEYWORD(resume, Resume); 645 INSTKEYWORD(unwind, Unwind); 646 INSTKEYWORD(unreachable, Unreachable); 647 648 INSTKEYWORD(alloca, Alloca); 649 INSTKEYWORD(load, Load); 650 INSTKEYWORD(store, Store); 651 INSTKEYWORD(cmpxchg, AtomicCmpXchg); 652 INSTKEYWORD(atomicrmw, AtomicRMW); 653 INSTKEYWORD(fence, Fence); 654 INSTKEYWORD(getelementptr, GetElementPtr); 655 656 INSTKEYWORD(extractelement, ExtractElement); 657 INSTKEYWORD(insertelement, InsertElement); 658 INSTKEYWORD(shufflevector, ShuffleVector); 659 INSTKEYWORD(extractvalue, ExtractValue); 660 INSTKEYWORD(insertvalue, InsertValue); 661 INSTKEYWORD(landingpad, LandingPad); 662#undef INSTKEYWORD 663 664 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 665 // the CFE to avoid forcing it to deal with 64-bit numbers. 666 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 667 TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) { 668 int len = CurPtr-TokStart-3; 669 uint32_t bits = len * 4; 670 APInt Tmp(bits, StringRef(TokStart+3, len), 16); 671 uint32_t activeBits = Tmp.getActiveBits(); 672 if (activeBits > 0 && activeBits < bits) 673 Tmp = Tmp.trunc(activeBits); 674 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 675 return lltok::APSInt; 676 } 677 678 // If this is "cc1234", return this as just "cc". 679 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 680 CurPtr = TokStart+2; 681 return lltok::kw_cc; 682 } 683 684 // Finally, if this isn't known, return an error. 685 CurPtr = TokStart+1; 686 return lltok::Error; 687} 688 689 690/// Lex0x: Handle productions that start with 0x, knowing that it matches and 691/// that this is not a label: 692/// HexFPConstant 0x[0-9A-Fa-f]+ 693/// HexFP80Constant 0xK[0-9A-Fa-f]+ 694/// HexFP128Constant 0xL[0-9A-Fa-f]+ 695/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 696lltok::Kind LLLexer::Lex0x() { 697 CurPtr = TokStart + 2; 698 699 char Kind; 700 if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') { 701 Kind = *CurPtr++; 702 } else { 703 Kind = 'J'; 704 } 705 706 if (!isxdigit(CurPtr[0])) { 707 // Bad token, return it as an error. 708 CurPtr = TokStart+1; 709 return lltok::Error; 710 } 711 712 while (isxdigit(CurPtr[0])) 713 ++CurPtr; 714 715 if (Kind == 'J') { 716 // HexFPConstant - Floating point constant represented in IEEE format as a 717 // hexadecimal number for when exponential notation is not precise enough. 718 // Float and double only. 719 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr))); 720 return lltok::APFloat; 721 } 722 723 uint64_t Pair[2]; 724 switch (Kind) { 725 default: llvm_unreachable("Unknown kind!"); 726 case 'K': 727 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 728 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 729 APFloatVal = APFloat(APInt(80, Pair)); 730 return lltok::APFloat; 731 case 'L': 732 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 733 HexToIntPair(TokStart+3, CurPtr, Pair); 734 APFloatVal = APFloat(APInt(128, Pair), true); 735 return lltok::APFloat; 736 case 'M': 737 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 738 HexToIntPair(TokStart+3, CurPtr, Pair); 739 APFloatVal = APFloat(APInt(128, Pair)); 740 return lltok::APFloat; 741 } 742} 743 744/// LexIdentifier: Handle several related productions: 745/// Label [-a-zA-Z$._0-9]+: 746/// NInteger -[0-9]+ 747/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 748/// PInteger [0-9]+ 749/// HexFPConstant 0x[0-9A-Fa-f]+ 750/// HexFP80Constant 0xK[0-9A-Fa-f]+ 751/// HexFP128Constant 0xL[0-9A-Fa-f]+ 752/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 753lltok::Kind LLLexer::LexDigitOrNegative() { 754 // If the letter after the negative is a number, this is probably a label. 755 if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) { 756 // Okay, this is not a number after the -, it's probably a label. 757 if (const char *End = isLabelTail(CurPtr)) { 758 StrVal.assign(TokStart, End-1); 759 CurPtr = End; 760 return lltok::LabelStr; 761 } 762 763 return lltok::Error; 764 } 765 766 // At this point, it is either a label, int or fp constant. 767 768 // Skip digits, we have at least one. 769 for (; isdigit(CurPtr[0]); ++CurPtr) 770 /*empty*/; 771 772 // Check to see if this really is a label afterall, e.g. "-1:". 773 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 774 if (const char *End = isLabelTail(CurPtr)) { 775 StrVal.assign(TokStart, End-1); 776 CurPtr = End; 777 return lltok::LabelStr; 778 } 779 } 780 781 // If the next character is a '.', then it is a fp value, otherwise its 782 // integer. 783 if (CurPtr[0] != '.') { 784 if (TokStart[0] == '0' && TokStart[1] == 'x') 785 return Lex0x(); 786 unsigned Len = CurPtr-TokStart; 787 uint32_t numBits = ((Len * 64) / 19) + 2; 788 APInt Tmp(numBits, StringRef(TokStart, Len), 10); 789 if (TokStart[0] == '-') { 790 uint32_t minBits = Tmp.getMinSignedBits(); 791 if (minBits > 0 && minBits < numBits) 792 Tmp = Tmp.trunc(minBits); 793 APSIntVal = APSInt(Tmp, false); 794 } else { 795 uint32_t activeBits = Tmp.getActiveBits(); 796 if (activeBits > 0 && activeBits < numBits) 797 Tmp = Tmp.trunc(activeBits); 798 APSIntVal = APSInt(Tmp, true); 799 } 800 return lltok::APSInt; 801 } 802 803 ++CurPtr; 804 805 // Skip over [0-9]*([eE][-+]?[0-9]+)? 806 while (isdigit(CurPtr[0])) ++CurPtr; 807 808 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 809 if (isdigit(CurPtr[1]) || 810 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) { 811 CurPtr += 2; 812 while (isdigit(CurPtr[0])) ++CurPtr; 813 } 814 } 815 816 APFloatVal = APFloat(std::atof(TokStart)); 817 return lltok::APFloat; 818} 819 820/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 821lltok::Kind LLLexer::LexPositive() { 822 // If the letter after the negative is a number, this is probably not a 823 // label. 824 if (!isdigit(CurPtr[0])) 825 return lltok::Error; 826 827 // Skip digits. 828 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 829 /*empty*/; 830 831 // At this point, we need a '.'. 832 if (CurPtr[0] != '.') { 833 CurPtr = TokStart+1; 834 return lltok::Error; 835 } 836 837 ++CurPtr; 838 839 // Skip over [0-9]*([eE][-+]?[0-9]+)? 840 while (isdigit(CurPtr[0])) ++CurPtr; 841 842 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 843 if (isdigit(CurPtr[1]) || 844 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) { 845 CurPtr += 2; 846 while (isdigit(CurPtr[0])) ++CurPtr; 847 } 848 } 849 850 APFloatVal = APFloat(std::atof(TokStart)); 851 return lltok::APFloat; 852} 853