LLLexer.cpp revision 476b242fe7a61e5f9ac6214b0bc5c680d24f152e
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, Msg, "error"); 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$._0-9]* 311 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 312 CurPtr[0] == '.' || CurPtr[0] == '_') { 313 ++CurPtr; 314 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 315 CurPtr[0] == '.' || CurPtr[0] == '_') 316 ++CurPtr; 317 318 StrVal.assign(TokStart+1, CurPtr); // Skip @ 319 return lltok::GlobalVar; 320 } 321 322 // Handle GlobalVarID: @[0-9]+ 323 if (isdigit(CurPtr[0])) { 324 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 325 /*empty*/; 326 327 uint64_t Val = atoull(TokStart+1, CurPtr); 328 if ((unsigned)Val != Val) 329 Error("invalid value number (too large)!"); 330 UIntVal = unsigned(Val); 331 return lltok::GlobalID; 332 } 333 334 return lltok::Error; 335} 336 337 338/// LexPercent - Lex all tokens that start with a % character: 339/// LocalVar ::= %\"[^\"]*\" 340/// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]* 341/// LocalVarID ::= %[0-9]+ 342lltok::Kind LLLexer::LexPercent() { 343 // Handle LocalVarName: %\"[^\"]*\" 344 if (CurPtr[0] == '"') { 345 ++CurPtr; 346 347 while (1) { 348 int CurChar = getNextChar(); 349 350 if (CurChar == EOF) { 351 Error("end of file in string constant"); 352 return lltok::Error; 353 } 354 if (CurChar == '"') { 355 StrVal.assign(TokStart+2, CurPtr-1); 356 UnEscapeLexed(StrVal); 357 return lltok::LocalVar; 358 } 359 } 360 } 361 362 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]* 363 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 364 CurPtr[0] == '.' || CurPtr[0] == '_') { 365 ++CurPtr; 366 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 367 CurPtr[0] == '.' || CurPtr[0] == '_') 368 ++CurPtr; 369 370 StrVal.assign(TokStart+1, CurPtr); // Skip % 371 return lltok::LocalVar; 372 } 373 374 // Handle LocalVarID: %[0-9]+ 375 if (isdigit(CurPtr[0])) { 376 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 377 /*empty*/; 378 379 uint64_t Val = atoull(TokStart+1, CurPtr); 380 if ((unsigned)Val != Val) 381 Error("invalid value number (too large)!"); 382 UIntVal = unsigned(Val); 383 return lltok::LocalVarID; 384 } 385 386 return lltok::Error; 387} 388 389/// LexQuote - Lex all tokens that start with a " character: 390/// QuoteLabel "[^"]+": 391/// StringConstant "[^"]*" 392lltok::Kind LLLexer::LexQuote() { 393 while (1) { 394 int CurChar = getNextChar(); 395 396 if (CurChar == EOF) { 397 Error("end of file in quoted string"); 398 return lltok::Error; 399 } 400 401 if (CurChar != '"') continue; 402 403 if (CurPtr[0] != ':') { 404 StrVal.assign(TokStart+1, CurPtr-1); 405 UnEscapeLexed(StrVal); 406 return lltok::StringConstant; 407 } 408 409 ++CurPtr; 410 StrVal.assign(TokStart+1, CurPtr-2); 411 UnEscapeLexed(StrVal); 412 return lltok::LabelStr; 413 } 414} 415 416static bool JustWhitespaceNewLine(const char *&Ptr) { 417 const char *ThisPtr = Ptr; 418 while (*ThisPtr == ' ' || *ThisPtr == '\t') 419 ++ThisPtr; 420 if (*ThisPtr == '\n' || *ThisPtr == '\r') { 421 Ptr = ThisPtr; 422 return true; 423 } 424 return false; 425} 426 427/// LexExclaim: 428/// !foo 429/// ! 430lltok::Kind LLLexer::LexExclaim() { 431 // Lex a metadata name as a MetadataVar. 432 if (isalpha(CurPtr[0])) { 433 ++CurPtr; 434 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' || 435 CurPtr[0] == '.' || CurPtr[0] == '_') 436 ++CurPtr; 437 438 StrVal.assign(TokStart+1, CurPtr); // Skip ! 439 return lltok::MetadataVar; 440 } 441 return lltok::exclaim; 442} 443 444/// LexIdentifier: Handle several related productions: 445/// Label [-a-zA-Z$._0-9]+: 446/// IntegerType i[0-9]+ 447/// Keyword sdiv, float, ... 448/// HexIntConstant [us]0x[0-9A-Fa-f]+ 449lltok::Kind LLLexer::LexIdentifier() { 450 const char *StartChar = CurPtr; 451 const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar; 452 const char *KeywordEnd = 0; 453 454 for (; isLabelChar(*CurPtr); ++CurPtr) { 455 // If we decide this is an integer, remember the end of the sequence. 456 if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr; 457 if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr; 458 } 459 460 // If we stopped due to a colon, this really is a label. 461 if (*CurPtr == ':') { 462 StrVal.assign(StartChar-1, CurPtr++); 463 return lltok::LabelStr; 464 } 465 466 // Otherwise, this wasn't a label. If this was valid as an integer type, 467 // return it. 468 if (IntEnd == 0) IntEnd = CurPtr; 469 if (IntEnd != StartChar) { 470 CurPtr = IntEnd; 471 uint64_t NumBits = atoull(StartChar, CurPtr); 472 if (NumBits < IntegerType::MIN_INT_BITS || 473 NumBits > IntegerType::MAX_INT_BITS) { 474 Error("bitwidth for integer type out of range!"); 475 return lltok::Error; 476 } 477 TyVal = IntegerType::get(Context, NumBits); 478 return lltok::Type; 479 } 480 481 // Otherwise, this was a letter sequence. See which keyword this is. 482 if (KeywordEnd == 0) KeywordEnd = CurPtr; 483 CurPtr = KeywordEnd; 484 --StartChar; 485 unsigned Len = CurPtr-StartChar; 486#define KEYWORD(STR) \ 487 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \ 488 return lltok::kw_##STR; 489 490 KEYWORD(begin); KEYWORD(end); 491 KEYWORD(true); KEYWORD(false); 492 KEYWORD(declare); KEYWORD(define); 493 KEYWORD(global); KEYWORD(constant); 494 495 KEYWORD(private); 496 KEYWORD(linker_private); 497 KEYWORD(linker_private_weak); 498 KEYWORD(linker_private_weak_def_auto); 499 KEYWORD(internal); 500 KEYWORD(available_externally); 501 KEYWORD(linkonce); 502 KEYWORD(linkonce_odr); 503 KEYWORD(weak); 504 KEYWORD(weak_odr); 505 KEYWORD(appending); 506 KEYWORD(dllimport); 507 KEYWORD(dllexport); 508 KEYWORD(common); 509 KEYWORD(default); 510 KEYWORD(hidden); 511 KEYWORD(protected); 512 KEYWORD(extern_weak); 513 KEYWORD(external); 514 KEYWORD(thread_local); 515 KEYWORD(zeroinitializer); 516 KEYWORD(undef); 517 KEYWORD(null); 518 KEYWORD(to); 519 KEYWORD(tail); 520 KEYWORD(target); 521 KEYWORD(triple); 522 KEYWORD(deplibs); 523 KEYWORD(datalayout); 524 KEYWORD(volatile); 525 KEYWORD(nuw); 526 KEYWORD(nsw); 527 KEYWORD(exact); 528 KEYWORD(inbounds); 529 KEYWORD(align); 530 KEYWORD(addrspace); 531 KEYWORD(section); 532 KEYWORD(alias); 533 KEYWORD(module); 534 KEYWORD(asm); 535 KEYWORD(sideeffect); 536 KEYWORD(alignstack); 537 KEYWORD(gc); 538 539 KEYWORD(ccc); 540 KEYWORD(fastcc); 541 KEYWORD(coldcc); 542 KEYWORD(x86_stdcallcc); 543 KEYWORD(x86_fastcallcc); 544 KEYWORD(x86_thiscallcc); 545 KEYWORD(arm_apcscc); 546 KEYWORD(arm_aapcscc); 547 KEYWORD(arm_aapcs_vfpcc); 548 KEYWORD(msp430_intrcc); 549 KEYWORD(ptx_kernel); 550 KEYWORD(ptx_device); 551 552 KEYWORD(cc); 553 KEYWORD(c); 554 555 KEYWORD(signext); 556 KEYWORD(zeroext); 557 KEYWORD(inreg); 558 KEYWORD(sret); 559 KEYWORD(nounwind); 560 KEYWORD(noreturn); 561 KEYWORD(noalias); 562 KEYWORD(nocapture); 563 KEYWORD(byval); 564 KEYWORD(nest); 565 KEYWORD(readnone); 566 KEYWORD(readonly); 567 568 KEYWORD(inlinehint); 569 KEYWORD(noinline); 570 KEYWORD(alwaysinline); 571 KEYWORD(optsize); 572 KEYWORD(ssp); 573 KEYWORD(sspreq); 574 KEYWORD(noredzone); 575 KEYWORD(noimplicitfloat); 576 KEYWORD(naked); 577 KEYWORD(hotpatch); 578 579 KEYWORD(type); 580 KEYWORD(opaque); 581 582 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle); 583 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge); 584 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole); 585 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une); 586 587 KEYWORD(x); 588 KEYWORD(blockaddress); 589#undef KEYWORD 590 591 // Keywords for types. 592#define TYPEKEYWORD(STR, LLVMTY) \ 593 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \ 594 TyVal = LLVMTY; return lltok::Type; } 595 TYPEKEYWORD("void", Type::getVoidTy(Context)); 596 TYPEKEYWORD("float", Type::getFloatTy(Context)); 597 TYPEKEYWORD("double", Type::getDoubleTy(Context)); 598 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context)); 599 TYPEKEYWORD("fp128", Type::getFP128Ty(Context)); 600 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context)); 601 TYPEKEYWORD("label", Type::getLabelTy(Context)); 602 TYPEKEYWORD("metadata", Type::getMetadataTy(Context)); 603 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context)); 604#undef TYPEKEYWORD 605 606 // Handle special forms for autoupgrading. Drop these in LLVM 3.0. This is 607 // to avoid conflicting with the sext/zext instructions, below. 608 if (Len == 4 && !memcmp(StartChar, "sext", 4)) { 609 // Scan CurPtr ahead, seeing if there is just whitespace before the newline. 610 if (JustWhitespaceNewLine(CurPtr)) 611 return lltok::kw_signext; 612 } else if (Len == 4 && !memcmp(StartChar, "zext", 4)) { 613 // Scan CurPtr ahead, seeing if there is just whitespace before the newline. 614 if (JustWhitespaceNewLine(CurPtr)) 615 return lltok::kw_zeroext; 616 } else if (Len == 6 && !memcmp(StartChar, "malloc", 6)) { 617 // FIXME: Remove in LLVM 3.0. 618 // Autoupgrade malloc instruction. 619 return lltok::kw_malloc; 620 } else if (Len == 4 && !memcmp(StartChar, "free", 4)) { 621 // FIXME: Remove in LLVM 3.0. 622 // Autoupgrade malloc instruction. 623 return lltok::kw_free; 624 } 625 626 // Keywords for instructions. 627#define INSTKEYWORD(STR, Enum) \ 628 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \ 629 UIntVal = Instruction::Enum; return lltok::kw_##STR; } 630 631 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd); 632 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub); 633 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul); 634 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv); 635 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem); 636 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr); 637 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor); 638 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp); 639 640 INSTKEYWORD(phi, PHI); 641 INSTKEYWORD(call, Call); 642 INSTKEYWORD(trunc, Trunc); 643 INSTKEYWORD(zext, ZExt); 644 INSTKEYWORD(sext, SExt); 645 INSTKEYWORD(fptrunc, FPTrunc); 646 INSTKEYWORD(fpext, FPExt); 647 INSTKEYWORD(uitofp, UIToFP); 648 INSTKEYWORD(sitofp, SIToFP); 649 INSTKEYWORD(fptoui, FPToUI); 650 INSTKEYWORD(fptosi, FPToSI); 651 INSTKEYWORD(inttoptr, IntToPtr); 652 INSTKEYWORD(ptrtoint, PtrToInt); 653 INSTKEYWORD(bitcast, BitCast); 654 INSTKEYWORD(select, Select); 655 INSTKEYWORD(va_arg, VAArg); 656 INSTKEYWORD(ret, Ret); 657 INSTKEYWORD(br, Br); 658 INSTKEYWORD(switch, Switch); 659 INSTKEYWORD(indirectbr, IndirectBr); 660 INSTKEYWORD(invoke, Invoke); 661 INSTKEYWORD(unwind, Unwind); 662 INSTKEYWORD(unreachable, Unreachable); 663 664 INSTKEYWORD(alloca, Alloca); 665 INSTKEYWORD(load, Load); 666 INSTKEYWORD(store, Store); 667 INSTKEYWORD(getelementptr, GetElementPtr); 668 669 INSTKEYWORD(extractelement, ExtractElement); 670 INSTKEYWORD(insertelement, InsertElement); 671 INSTKEYWORD(shufflevector, ShuffleVector); 672 INSTKEYWORD(getresult, ExtractValue); 673 INSTKEYWORD(extractvalue, ExtractValue); 674 INSTKEYWORD(insertvalue, InsertValue); 675#undef INSTKEYWORD 676 677 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by 678 // the CFE to avoid forcing it to deal with 64-bit numbers. 679 if ((TokStart[0] == 'u' || TokStart[0] == 's') && 680 TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) { 681 int len = CurPtr-TokStart-3; 682 uint32_t bits = len * 4; 683 APInt Tmp(bits, StringRef(TokStart+3, len), 16); 684 uint32_t activeBits = Tmp.getActiveBits(); 685 if (activeBits > 0 && activeBits < bits) 686 Tmp = Tmp.trunc(activeBits); 687 APSIntVal = APSInt(Tmp, TokStart[0] == 'u'); 688 return lltok::APSInt; 689 } 690 691 // If this is "cc1234", return this as just "cc". 692 if (TokStart[0] == 'c' && TokStart[1] == 'c') { 693 CurPtr = TokStart+2; 694 return lltok::kw_cc; 695 } 696 697 // If this starts with "call", return it as CALL. This is to support old 698 // broken .ll files. FIXME: remove this with LLVM 3.0. 699 if (CurPtr-TokStart > 4 && !memcmp(TokStart, "call", 4)) { 700 CurPtr = TokStart+4; 701 UIntVal = Instruction::Call; 702 return lltok::kw_call; 703 } 704 705 // Finally, if this isn't known, return an error. 706 CurPtr = TokStart+1; 707 return lltok::Error; 708} 709 710 711/// Lex0x: Handle productions that start with 0x, knowing that it matches and 712/// that this is not a label: 713/// HexFPConstant 0x[0-9A-Fa-f]+ 714/// HexFP80Constant 0xK[0-9A-Fa-f]+ 715/// HexFP128Constant 0xL[0-9A-Fa-f]+ 716/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 717lltok::Kind LLLexer::Lex0x() { 718 CurPtr = TokStart + 2; 719 720 char Kind; 721 if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') { 722 Kind = *CurPtr++; 723 } else { 724 Kind = 'J'; 725 } 726 727 if (!isxdigit(CurPtr[0])) { 728 // Bad token, return it as an error. 729 CurPtr = TokStart+1; 730 return lltok::Error; 731 } 732 733 while (isxdigit(CurPtr[0])) 734 ++CurPtr; 735 736 if (Kind == 'J') { 737 // HexFPConstant - Floating point constant represented in IEEE format as a 738 // hexadecimal number for when exponential notation is not precise enough. 739 // Float and double only. 740 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr))); 741 return lltok::APFloat; 742 } 743 744 uint64_t Pair[2]; 745 switch (Kind) { 746 default: llvm_unreachable("Unknown kind!"); 747 case 'K': 748 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes) 749 FP80HexToIntPair(TokStart+3, CurPtr, Pair); 750 APFloatVal = APFloat(APInt(80, 2, Pair)); 751 return lltok::APFloat; 752 case 'L': 753 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes) 754 HexToIntPair(TokStart+3, CurPtr, Pair); 755 APFloatVal = APFloat(APInt(128, 2, Pair), true); 756 return lltok::APFloat; 757 case 'M': 758 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes) 759 HexToIntPair(TokStart+3, CurPtr, Pair); 760 APFloatVal = APFloat(APInt(128, 2, Pair)); 761 return lltok::APFloat; 762 } 763} 764 765/// LexIdentifier: Handle several related productions: 766/// Label [-a-zA-Z$._0-9]+: 767/// NInteger -[0-9]+ 768/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 769/// PInteger [0-9]+ 770/// HexFPConstant 0x[0-9A-Fa-f]+ 771/// HexFP80Constant 0xK[0-9A-Fa-f]+ 772/// HexFP128Constant 0xL[0-9A-Fa-f]+ 773/// HexPPC128Constant 0xM[0-9A-Fa-f]+ 774lltok::Kind LLLexer::LexDigitOrNegative() { 775 // If the letter after the negative is a number, this is probably a label. 776 if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) { 777 // Okay, this is not a number after the -, it's probably a label. 778 if (const char *End = isLabelTail(CurPtr)) { 779 StrVal.assign(TokStart, End-1); 780 CurPtr = End; 781 return lltok::LabelStr; 782 } 783 784 return lltok::Error; 785 } 786 787 // At this point, it is either a label, int or fp constant. 788 789 // Skip digits, we have at least one. 790 for (; isdigit(CurPtr[0]); ++CurPtr) 791 /*empty*/; 792 793 // Check to see if this really is a label afterall, e.g. "-1:". 794 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') { 795 if (const char *End = isLabelTail(CurPtr)) { 796 StrVal.assign(TokStart, End-1); 797 CurPtr = End; 798 return lltok::LabelStr; 799 } 800 } 801 802 // If the next character is a '.', then it is a fp value, otherwise its 803 // integer. 804 if (CurPtr[0] != '.') { 805 if (TokStart[0] == '0' && TokStart[1] == 'x') 806 return Lex0x(); 807 unsigned Len = CurPtr-TokStart; 808 uint32_t numBits = ((Len * 64) / 19) + 2; 809 APInt Tmp(numBits, StringRef(TokStart, Len), 10); 810 if (TokStart[0] == '-') { 811 uint32_t minBits = Tmp.getMinSignedBits(); 812 if (minBits > 0 && minBits < numBits) 813 Tmp = Tmp.trunc(minBits); 814 APSIntVal = APSInt(Tmp, false); 815 } else { 816 uint32_t activeBits = Tmp.getActiveBits(); 817 if (activeBits > 0 && activeBits < numBits) 818 Tmp = Tmp.trunc(activeBits); 819 APSIntVal = APSInt(Tmp, true); 820 } 821 return lltok::APSInt; 822 } 823 824 ++CurPtr; 825 826 // Skip over [0-9]*([eE][-+]?[0-9]+)? 827 while (isdigit(CurPtr[0])) ++CurPtr; 828 829 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 830 if (isdigit(CurPtr[1]) || 831 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) { 832 CurPtr += 2; 833 while (isdigit(CurPtr[0])) ++CurPtr; 834 } 835 } 836 837 APFloatVal = APFloat(std::atof(TokStart)); 838 return lltok::APFloat; 839} 840 841/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)? 842lltok::Kind LLLexer::LexPositive() { 843 // If the letter after the negative is a number, this is probably not a 844 // label. 845 if (!isdigit(CurPtr[0])) 846 return lltok::Error; 847 848 // Skip digits. 849 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr) 850 /*empty*/; 851 852 // At this point, we need a '.'. 853 if (CurPtr[0] != '.') { 854 CurPtr = TokStart+1; 855 return lltok::Error; 856 } 857 858 ++CurPtr; 859 860 // Skip over [0-9]*([eE][-+]?[0-9]+)? 861 while (isdigit(CurPtr[0])) ++CurPtr; 862 863 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') { 864 if (isdigit(CurPtr[1]) || 865 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) { 866 CurPtr += 2; 867 while (isdigit(CurPtr[0])) ++CurPtr; 868 } 869 } 870 871 APFloatVal = APFloat(std::atof(TokStart)); 872 return lltok::APFloat; 873} 874