1 //===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // Implement the Lexer for .ll files.
11 //
12 //===----------------------------------------------------------------------===//
13
14 #include "LLLexer.h"
15 #include "llvm/ADT/StringExtras.h"
16 #include "llvm/ADT/Twine.h"
17 #include "llvm/AsmParser/Parser.h"
18 #include "llvm/IR/DerivedTypes.h"
19 #include "llvm/IR/Instruction.h"
20 #include "llvm/IR/LLVMContext.h"
21 #include "llvm/Support/ErrorHandling.h"
22 #include "llvm/Support/MathExtras.h"
23 #include "llvm/Support/MemoryBuffer.h"
24 #include "llvm/Support/SourceMgr.h"
25 #include "llvm/Support/raw_ostream.h"
26 #include <cctype>
27 #include <cstdio>
28 #include <cstdlib>
29 #include <cstring>
30 using namespace llvm;
31
Error(LocTy ErrorLoc,const Twine & Msg) const32 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
33 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
34 return true;
35 }
36
Warning(LocTy WarningLoc,const Twine & Msg) const37 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
38 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
39 }
40
41 //===----------------------------------------------------------------------===//
42 // Helper functions.
43 //===----------------------------------------------------------------------===//
44
45 // atoull - Convert an ascii string of decimal digits into the unsigned long
46 // long representation... this does not have to do input error checking,
47 // because we know that the input will be matched by a suitable regex...
48 //
atoull(const char * Buffer,const char * End)49 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
50 uint64_t Result = 0;
51 for (; Buffer != End; Buffer++) {
52 uint64_t OldRes = Result;
53 Result *= 10;
54 Result += *Buffer-'0';
55 if (Result < OldRes) { // Uh, oh, overflow detected!!!
56 Error("constant bigger than 64 bits detected!");
57 return 0;
58 }
59 }
60 return Result;
61 }
62
HexIntToVal(const char * Buffer,const char * End)63 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
64 uint64_t Result = 0;
65 for (; Buffer != End; ++Buffer) {
66 uint64_t OldRes = Result;
67 Result *= 16;
68 Result += hexDigitValue(*Buffer);
69
70 if (Result < OldRes) { // Uh, oh, overflow detected!!!
71 Error("constant bigger than 64 bits detected!");
72 return 0;
73 }
74 }
75 return Result;
76 }
77
HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])78 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
79 uint64_t Pair[2]) {
80 Pair[0] = 0;
81 for (int i=0; i<16; i++, Buffer++) {
82 assert(Buffer != End);
83 Pair[0] *= 16;
84 Pair[0] += hexDigitValue(*Buffer);
85 }
86 Pair[1] = 0;
87 for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
88 Pair[1] *= 16;
89 Pair[1] += hexDigitValue(*Buffer);
90 }
91 if (Buffer != End)
92 Error("constant bigger than 128 bits detected!");
93 }
94
95 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
96 /// { low64, high16 } as usual for an APInt.
FP80HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])97 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
98 uint64_t Pair[2]) {
99 Pair[1] = 0;
100 for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
101 assert(Buffer != End);
102 Pair[1] *= 16;
103 Pair[1] += hexDigitValue(*Buffer);
104 }
105 Pair[0] = 0;
106 for (int i=0; i<16; i++, Buffer++) {
107 Pair[0] *= 16;
108 Pair[0] += hexDigitValue(*Buffer);
109 }
110 if (Buffer != End)
111 Error("constant bigger than 128 bits detected!");
112 }
113
114 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
115 // appropriate character.
UnEscapeLexed(std::string & Str)116 static void UnEscapeLexed(std::string &Str) {
117 if (Str.empty()) return;
118
119 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
120 char *BOut = Buffer;
121 for (char *BIn = Buffer; BIn != EndBuffer; ) {
122 if (BIn[0] == '\\') {
123 if (BIn < EndBuffer-1 && BIn[1] == '\\') {
124 *BOut++ = '\\'; // Two \ becomes one
125 BIn += 2;
126 } else if (BIn < EndBuffer-2 &&
127 isxdigit(static_cast<unsigned char>(BIn[1])) &&
128 isxdigit(static_cast<unsigned char>(BIn[2]))) {
129 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
130 BIn += 3; // Skip over handled chars
131 ++BOut;
132 } else {
133 *BOut++ = *BIn++;
134 }
135 } else {
136 *BOut++ = *BIn++;
137 }
138 }
139 Str.resize(BOut-Buffer);
140 }
141
142 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
isLabelChar(char C)143 static bool isLabelChar(char C) {
144 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
145 C == '.' || C == '_';
146 }
147
148
149 /// isLabelTail - Return true if this pointer points to a valid end of a label.
isLabelTail(const char * CurPtr)150 static const char *isLabelTail(const char *CurPtr) {
151 while (1) {
152 if (CurPtr[0] == ':') return CurPtr+1;
153 if (!isLabelChar(CurPtr[0])) return nullptr;
154 ++CurPtr;
155 }
156 }
157
158
159
160 //===----------------------------------------------------------------------===//
161 // Lexer definition.
162 //===----------------------------------------------------------------------===//
163
LLLexer(MemoryBuffer * StartBuf,SourceMgr & sm,SMDiagnostic & Err,LLVMContext & C)164 LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err,
165 LLVMContext &C)
166 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
167 CurPtr = CurBuf->getBufferStart();
168 }
169
getFilename() const170 std::string LLLexer::getFilename() const {
171 return CurBuf->getBufferIdentifier();
172 }
173
getNextChar()174 int LLLexer::getNextChar() {
175 char CurChar = *CurPtr++;
176 switch (CurChar) {
177 default: return (unsigned char)CurChar;
178 case 0:
179 // A nul character in the stream is either the end of the current buffer or
180 // a random nul in the file. Disambiguate that here.
181 if (CurPtr-1 != CurBuf->getBufferEnd())
182 return 0; // Just whitespace.
183
184 // Otherwise, return end of file.
185 --CurPtr; // Another call to lex will return EOF again.
186 return EOF;
187 }
188 }
189
190
LexToken()191 lltok::Kind LLLexer::LexToken() {
192 TokStart = CurPtr;
193
194 int CurChar = getNextChar();
195 switch (CurChar) {
196 default:
197 // Handle letters: [a-zA-Z_]
198 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
199 return LexIdentifier();
200
201 return lltok::Error;
202 case EOF: return lltok::Eof;
203 case 0:
204 case ' ':
205 case '\t':
206 case '\n':
207 case '\r':
208 // Ignore whitespace.
209 return LexToken();
210 case '+': return LexPositive();
211 case '@': return LexAt();
212 case '$': return LexDollar();
213 case '%': return LexPercent();
214 case '"': return LexQuote();
215 case '.':
216 if (const char *Ptr = isLabelTail(CurPtr)) {
217 CurPtr = Ptr;
218 StrVal.assign(TokStart, CurPtr-1);
219 return lltok::LabelStr;
220 }
221 if (CurPtr[0] == '.' && CurPtr[1] == '.') {
222 CurPtr += 2;
223 return lltok::dotdotdot;
224 }
225 return lltok::Error;
226 case ';':
227 SkipLineComment();
228 return LexToken();
229 case '!': return LexExclaim();
230 case '#': return LexHash();
231 case '0': case '1': case '2': case '3': case '4':
232 case '5': case '6': case '7': case '8': case '9':
233 case '-':
234 return LexDigitOrNegative();
235 case '=': return lltok::equal;
236 case '[': return lltok::lsquare;
237 case ']': return lltok::rsquare;
238 case '{': return lltok::lbrace;
239 case '}': return lltok::rbrace;
240 case '<': return lltok::less;
241 case '>': return lltok::greater;
242 case '(': return lltok::lparen;
243 case ')': return lltok::rparen;
244 case ',': return lltok::comma;
245 case '*': return lltok::star;
246 case '\\': return lltok::backslash;
247 }
248 }
249
SkipLineComment()250 void LLLexer::SkipLineComment() {
251 while (1) {
252 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
253 return;
254 }
255 }
256
257 /// LexAt - Lex all tokens that start with an @ character:
258 /// GlobalVar @\"[^\"]*\"
259 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
260 /// GlobalVarID @[0-9]+
LexAt()261 lltok::Kind LLLexer::LexAt() {
262 // Handle AtStringConstant: @\"[^\"]*\"
263 if (CurPtr[0] == '"') {
264 ++CurPtr;
265
266 while (1) {
267 int CurChar = getNextChar();
268
269 if (CurChar == EOF) {
270 Error("end of file in global variable name");
271 return lltok::Error;
272 }
273 if (CurChar == '"') {
274 StrVal.assign(TokStart+2, CurPtr-1);
275 UnEscapeLexed(StrVal);
276 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
277 Error("Null bytes are not allowed in names");
278 return lltok::Error;
279 }
280 return lltok::GlobalVar;
281 }
282 }
283 }
284
285 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
286 if (ReadVarName())
287 return lltok::GlobalVar;
288
289 // Handle GlobalVarID: @[0-9]+
290 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
291 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
292 /*empty*/;
293
294 uint64_t Val = atoull(TokStart+1, CurPtr);
295 if ((unsigned)Val != Val)
296 Error("invalid value number (too large)!");
297 UIntVal = unsigned(Val);
298 return lltok::GlobalID;
299 }
300
301 return lltok::Error;
302 }
303
LexDollar()304 lltok::Kind LLLexer::LexDollar() {
305 if (const char *Ptr = isLabelTail(TokStart)) {
306 CurPtr = Ptr;
307 StrVal.assign(TokStart, CurPtr - 1);
308 return lltok::LabelStr;
309 }
310
311 // Handle DollarStringConstant: $\"[^\"]*\"
312 if (CurPtr[0] == '"') {
313 ++CurPtr;
314
315 while (1) {
316 int CurChar = getNextChar();
317
318 if (CurChar == EOF) {
319 Error("end of file in COMDAT variable name");
320 return lltok::Error;
321 }
322 if (CurChar == '"') {
323 StrVal.assign(TokStart + 2, CurPtr - 1);
324 UnEscapeLexed(StrVal);
325 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
326 Error("Null bytes are not allowed in names");
327 return lltok::Error;
328 }
329 return lltok::ComdatVar;
330 }
331 }
332 }
333
334 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
335 if (ReadVarName())
336 return lltok::ComdatVar;
337
338 return lltok::Error;
339 }
340
341 /// ReadString - Read a string until the closing quote.
ReadString(lltok::Kind kind)342 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
343 const char *Start = CurPtr;
344 while (1) {
345 int CurChar = getNextChar();
346
347 if (CurChar == EOF) {
348 Error("end of file in string constant");
349 return lltok::Error;
350 }
351 if (CurChar == '"') {
352 StrVal.assign(Start, CurPtr-1);
353 UnEscapeLexed(StrVal);
354 return kind;
355 }
356 }
357 }
358
359 /// ReadVarName - Read the rest of a token containing a variable name.
ReadVarName()360 bool LLLexer::ReadVarName() {
361 const char *NameStart = CurPtr;
362 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
363 CurPtr[0] == '-' || CurPtr[0] == '$' ||
364 CurPtr[0] == '.' || CurPtr[0] == '_') {
365 ++CurPtr;
366 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
367 CurPtr[0] == '-' || CurPtr[0] == '$' ||
368 CurPtr[0] == '.' || CurPtr[0] == '_')
369 ++CurPtr;
370
371 StrVal.assign(NameStart, CurPtr);
372 return true;
373 }
374 return false;
375 }
376
377 /// LexPercent - Lex all tokens that start with a % character:
378 /// LocalVar ::= %\"[^\"]*\"
379 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
380 /// LocalVarID ::= %[0-9]+
LexPercent()381 lltok::Kind LLLexer::LexPercent() {
382 // Handle LocalVarName: %\"[^\"]*\"
383 if (CurPtr[0] == '"') {
384 ++CurPtr;
385 return ReadString(lltok::LocalVar);
386 }
387
388 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
389 if (ReadVarName())
390 return lltok::LocalVar;
391
392 // Handle LocalVarID: %[0-9]+
393 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
394 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
395 /*empty*/;
396
397 uint64_t Val = atoull(TokStart+1, CurPtr);
398 if ((unsigned)Val != Val)
399 Error("invalid value number (too large)!");
400 UIntVal = unsigned(Val);
401 return lltok::LocalVarID;
402 }
403
404 return lltok::Error;
405 }
406
407 /// LexQuote - Lex all tokens that start with a " character:
408 /// QuoteLabel "[^"]+":
409 /// StringConstant "[^"]*"
LexQuote()410 lltok::Kind LLLexer::LexQuote() {
411 lltok::Kind kind = ReadString(lltok::StringConstant);
412 if (kind == lltok::Error || kind == lltok::Eof)
413 return kind;
414
415 if (CurPtr[0] == ':') {
416 ++CurPtr;
417 kind = lltok::LabelStr;
418 }
419
420 return kind;
421 }
422
423 /// LexExclaim:
424 /// !foo
425 /// !
LexExclaim()426 lltok::Kind LLLexer::LexExclaim() {
427 // Lex a metadata name as a MetadataVar.
428 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
429 CurPtr[0] == '-' || CurPtr[0] == '$' ||
430 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
431 ++CurPtr;
432 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
433 CurPtr[0] == '-' || CurPtr[0] == '$' ||
434 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
435 ++CurPtr;
436
437 StrVal.assign(TokStart+1, CurPtr); // Skip !
438 UnEscapeLexed(StrVal);
439 return lltok::MetadataVar;
440 }
441 return lltok::exclaim;
442 }
443
444 /// LexHash - Lex all tokens that start with a # character:
445 /// AttrGrpID ::= #[0-9]+
LexHash()446 lltok::Kind LLLexer::LexHash() {
447 // Handle AttrGrpID: #[0-9]+
448 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
449 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
450 /*empty*/;
451
452 uint64_t Val = atoull(TokStart+1, CurPtr);
453 if ((unsigned)Val != Val)
454 Error("invalid value number (too large)!");
455 UIntVal = unsigned(Val);
456 return lltok::AttrGrpID;
457 }
458
459 return lltok::Error;
460 }
461
462 /// LexIdentifier: Handle several related productions:
463 /// Label [-a-zA-Z$._0-9]+:
464 /// IntegerType i[0-9]+
465 /// Keyword sdiv, float, ...
466 /// HexIntConstant [us]0x[0-9A-Fa-f]+
LexIdentifier()467 lltok::Kind LLLexer::LexIdentifier() {
468 const char *StartChar = CurPtr;
469 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
470 const char *KeywordEnd = nullptr;
471
472 for (; isLabelChar(*CurPtr); ++CurPtr) {
473 // If we decide this is an integer, remember the end of the sequence.
474 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
475 IntEnd = CurPtr;
476 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
477 *CurPtr != '_')
478 KeywordEnd = CurPtr;
479 }
480
481 // If we stopped due to a colon, this really is a label.
482 if (*CurPtr == ':') {
483 StrVal.assign(StartChar-1, CurPtr++);
484 return lltok::LabelStr;
485 }
486
487 // Otherwise, this wasn't a label. If this was valid as an integer type,
488 // return it.
489 if (!IntEnd) IntEnd = CurPtr;
490 if (IntEnd != StartChar) {
491 CurPtr = IntEnd;
492 uint64_t NumBits = atoull(StartChar, CurPtr);
493 if (NumBits < IntegerType::MIN_INT_BITS ||
494 NumBits > IntegerType::MAX_INT_BITS) {
495 Error("bitwidth for integer type out of range!");
496 return lltok::Error;
497 }
498 TyVal = IntegerType::get(Context, NumBits);
499 return lltok::Type;
500 }
501
502 // Otherwise, this was a letter sequence. See which keyword this is.
503 if (!KeywordEnd) KeywordEnd = CurPtr;
504 CurPtr = KeywordEnd;
505 --StartChar;
506 unsigned Len = CurPtr-StartChar;
507 #define KEYWORD(STR) \
508 do { \
509 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
510 return lltok::kw_##STR; \
511 } while (0)
512
513 KEYWORD(true); KEYWORD(false);
514 KEYWORD(declare); KEYWORD(define);
515 KEYWORD(global); KEYWORD(constant);
516
517 KEYWORD(private);
518 KEYWORD(internal);
519 KEYWORD(linker_private); // NOTE: deprecated, for parser compatibility
520 KEYWORD(linker_private_weak); // NOTE: deprecated, for parser compatibility
521 KEYWORD(available_externally);
522 KEYWORD(linkonce);
523 KEYWORD(linkonce_odr);
524 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
525 KEYWORD(weak_odr);
526 KEYWORD(appending);
527 KEYWORD(dllimport);
528 KEYWORD(dllexport);
529 KEYWORD(common);
530 KEYWORD(default);
531 KEYWORD(hidden);
532 KEYWORD(protected);
533 KEYWORD(unnamed_addr);
534 KEYWORD(externally_initialized);
535 KEYWORD(extern_weak);
536 KEYWORD(external);
537 KEYWORD(thread_local);
538 KEYWORD(localdynamic);
539 KEYWORD(initialexec);
540 KEYWORD(localexec);
541 KEYWORD(zeroinitializer);
542 KEYWORD(undef);
543 KEYWORD(null);
544 KEYWORD(to);
545 KEYWORD(tail);
546 KEYWORD(musttail);
547 KEYWORD(target);
548 KEYWORD(triple);
549 KEYWORD(unwind);
550 KEYWORD(deplibs); // FIXME: Remove in 4.0.
551 KEYWORD(datalayout);
552 KEYWORD(volatile);
553 KEYWORD(atomic);
554 KEYWORD(unordered);
555 KEYWORD(monotonic);
556 KEYWORD(acquire);
557 KEYWORD(release);
558 KEYWORD(acq_rel);
559 KEYWORD(seq_cst);
560 KEYWORD(singlethread);
561
562 KEYWORD(nnan);
563 KEYWORD(ninf);
564 KEYWORD(nsz);
565 KEYWORD(arcp);
566 KEYWORD(fast);
567 KEYWORD(nuw);
568 KEYWORD(nsw);
569 KEYWORD(exact);
570 KEYWORD(inbounds);
571 KEYWORD(align);
572 KEYWORD(addrspace);
573 KEYWORD(section);
574 KEYWORD(alias);
575 KEYWORD(module);
576 KEYWORD(asm);
577 KEYWORD(sideeffect);
578 KEYWORD(alignstack);
579 KEYWORD(inteldialect);
580 KEYWORD(gc);
581 KEYWORD(prefix);
582
583 KEYWORD(ccc);
584 KEYWORD(fastcc);
585 KEYWORD(coldcc);
586 KEYWORD(x86_stdcallcc);
587 KEYWORD(x86_fastcallcc);
588 KEYWORD(x86_thiscallcc);
589 KEYWORD(arm_apcscc);
590 KEYWORD(arm_aapcscc);
591 KEYWORD(arm_aapcs_vfpcc);
592 KEYWORD(msp430_intrcc);
593 KEYWORD(ptx_kernel);
594 KEYWORD(ptx_device);
595 KEYWORD(spir_kernel);
596 KEYWORD(spir_func);
597 KEYWORD(intel_ocl_bicc);
598 KEYWORD(x86_64_sysvcc);
599 KEYWORD(x86_64_win64cc);
600 KEYWORD(webkit_jscc);
601 KEYWORD(anyregcc);
602 KEYWORD(preserve_mostcc);
603 KEYWORD(preserve_allcc);
604
605 KEYWORD(cc);
606 KEYWORD(c);
607
608 KEYWORD(attributes);
609
610 KEYWORD(alwaysinline);
611 KEYWORD(builtin);
612 KEYWORD(byval);
613 KEYWORD(inalloca);
614 KEYWORD(cold);
615 KEYWORD(inlinehint);
616 KEYWORD(inreg);
617 KEYWORD(jumptable);
618 KEYWORD(minsize);
619 KEYWORD(naked);
620 KEYWORD(nest);
621 KEYWORD(noalias);
622 KEYWORD(nobuiltin);
623 KEYWORD(nocapture);
624 KEYWORD(noduplicate);
625 KEYWORD(noimplicitfloat);
626 KEYWORD(noinline);
627 KEYWORD(nonlazybind);
628 KEYWORD(nonnull);
629 KEYWORD(noredzone);
630 KEYWORD(noreturn);
631 KEYWORD(nounwind);
632 KEYWORD(optnone);
633 KEYWORD(optsize);
634 KEYWORD(readnone);
635 KEYWORD(readonly);
636 KEYWORD(returned);
637 KEYWORD(returns_twice);
638 KEYWORD(signext);
639 KEYWORD(sret);
640 KEYWORD(ssp);
641 KEYWORD(sspreq);
642 KEYWORD(sspstrong);
643 KEYWORD(sanitize_address);
644 KEYWORD(sanitize_thread);
645 KEYWORD(sanitize_memory);
646 KEYWORD(uwtable);
647 KEYWORD(zeroext);
648
649 KEYWORD(type);
650 KEYWORD(opaque);
651
652 KEYWORD(comdat);
653
654 // Comdat types
655 KEYWORD(any);
656 KEYWORD(exactmatch);
657 KEYWORD(largest);
658 KEYWORD(noduplicates);
659 KEYWORD(samesize);
660
661 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
662 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
663 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
664 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
665
666 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
667 KEYWORD(umin);
668
669 KEYWORD(x);
670 KEYWORD(blockaddress);
671
672 KEYWORD(personality);
673 KEYWORD(cleanup);
674 KEYWORD(catch);
675 KEYWORD(filter);
676 #undef KEYWORD
677
678 // Keywords for types.
679 #define TYPEKEYWORD(STR, LLVMTY) \
680 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
681 TyVal = LLVMTY; return lltok::Type; }
682 TYPEKEYWORD("void", Type::getVoidTy(Context));
683 TYPEKEYWORD("half", Type::getHalfTy(Context));
684 TYPEKEYWORD("float", Type::getFloatTy(Context));
685 TYPEKEYWORD("double", Type::getDoubleTy(Context));
686 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
687 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
688 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
689 TYPEKEYWORD("label", Type::getLabelTy(Context));
690 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
691 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
692 #undef TYPEKEYWORD
693
694 // Keywords for instructions.
695 #define INSTKEYWORD(STR, Enum) \
696 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
697 UIntVal = Instruction::Enum; return lltok::kw_##STR; }
698
699 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
700 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
701 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
702 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
703 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
704 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
705 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
706 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
707
708 INSTKEYWORD(phi, PHI);
709 INSTKEYWORD(call, Call);
710 INSTKEYWORD(trunc, Trunc);
711 INSTKEYWORD(zext, ZExt);
712 INSTKEYWORD(sext, SExt);
713 INSTKEYWORD(fptrunc, FPTrunc);
714 INSTKEYWORD(fpext, FPExt);
715 INSTKEYWORD(uitofp, UIToFP);
716 INSTKEYWORD(sitofp, SIToFP);
717 INSTKEYWORD(fptoui, FPToUI);
718 INSTKEYWORD(fptosi, FPToSI);
719 INSTKEYWORD(inttoptr, IntToPtr);
720 INSTKEYWORD(ptrtoint, PtrToInt);
721 INSTKEYWORD(bitcast, BitCast);
722 INSTKEYWORD(addrspacecast, AddrSpaceCast);
723 INSTKEYWORD(select, Select);
724 INSTKEYWORD(va_arg, VAArg);
725 INSTKEYWORD(ret, Ret);
726 INSTKEYWORD(br, Br);
727 INSTKEYWORD(switch, Switch);
728 INSTKEYWORD(indirectbr, IndirectBr);
729 INSTKEYWORD(invoke, Invoke);
730 INSTKEYWORD(resume, Resume);
731 INSTKEYWORD(unreachable, Unreachable);
732
733 INSTKEYWORD(alloca, Alloca);
734 INSTKEYWORD(load, Load);
735 INSTKEYWORD(store, Store);
736 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
737 INSTKEYWORD(atomicrmw, AtomicRMW);
738 INSTKEYWORD(fence, Fence);
739 INSTKEYWORD(getelementptr, GetElementPtr);
740
741 INSTKEYWORD(extractelement, ExtractElement);
742 INSTKEYWORD(insertelement, InsertElement);
743 INSTKEYWORD(shufflevector, ShuffleVector);
744 INSTKEYWORD(extractvalue, ExtractValue);
745 INSTKEYWORD(insertvalue, InsertValue);
746 INSTKEYWORD(landingpad, LandingPad);
747 #undef INSTKEYWORD
748
749 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
750 // the CFE to avoid forcing it to deal with 64-bit numbers.
751 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
752 TokStart[1] == '0' && TokStart[2] == 'x' &&
753 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
754 int len = CurPtr-TokStart-3;
755 uint32_t bits = len * 4;
756 APInt Tmp(bits, StringRef(TokStart+3, len), 16);
757 uint32_t activeBits = Tmp.getActiveBits();
758 if (activeBits > 0 && activeBits < bits)
759 Tmp = Tmp.trunc(activeBits);
760 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
761 return lltok::APSInt;
762 }
763
764 // If this is "cc1234", return this as just "cc".
765 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
766 CurPtr = TokStart+2;
767 return lltok::kw_cc;
768 }
769
770 // Finally, if this isn't known, return an error.
771 CurPtr = TokStart+1;
772 return lltok::Error;
773 }
774
775
776 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
777 /// that this is not a label:
778 /// HexFPConstant 0x[0-9A-Fa-f]+
779 /// HexFP80Constant 0xK[0-9A-Fa-f]+
780 /// HexFP128Constant 0xL[0-9A-Fa-f]+
781 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
782 /// HexHalfConstant 0xH[0-9A-Fa-f]+
Lex0x()783 lltok::Kind LLLexer::Lex0x() {
784 CurPtr = TokStart + 2;
785
786 char Kind;
787 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
788 Kind = *CurPtr++;
789 } else {
790 Kind = 'J';
791 }
792
793 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
794 // Bad token, return it as an error.
795 CurPtr = TokStart+1;
796 return lltok::Error;
797 }
798
799 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
800 ++CurPtr;
801
802 if (Kind == 'J') {
803 // HexFPConstant - Floating point constant represented in IEEE format as a
804 // hexadecimal number for when exponential notation is not precise enough.
805 // Half, Float, and double only.
806 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
807 return lltok::APFloat;
808 }
809
810 uint64_t Pair[2];
811 switch (Kind) {
812 default: llvm_unreachable("Unknown kind!");
813 case 'K':
814 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
815 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
816 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair));
817 return lltok::APFloat;
818 case 'L':
819 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
820 HexToIntPair(TokStart+3, CurPtr, Pair);
821 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair));
822 return lltok::APFloat;
823 case 'M':
824 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
825 HexToIntPair(TokStart+3, CurPtr, Pair);
826 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair));
827 return lltok::APFloat;
828 case 'H':
829 APFloatVal = APFloat(APFloat::IEEEhalf,
830 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
831 return lltok::APFloat;
832 }
833 }
834
835 /// LexIdentifier: Handle several related productions:
836 /// Label [-a-zA-Z$._0-9]+:
837 /// NInteger -[0-9]+
838 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
839 /// PInteger [0-9]+
840 /// HexFPConstant 0x[0-9A-Fa-f]+
841 /// HexFP80Constant 0xK[0-9A-Fa-f]+
842 /// HexFP128Constant 0xL[0-9A-Fa-f]+
843 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
LexDigitOrNegative()844 lltok::Kind LLLexer::LexDigitOrNegative() {
845 // If the letter after the negative is not a number, this is probably a label.
846 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
847 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
848 // Okay, this is not a number after the -, it's probably a label.
849 if (const char *End = isLabelTail(CurPtr)) {
850 StrVal.assign(TokStart, End-1);
851 CurPtr = End;
852 return lltok::LabelStr;
853 }
854
855 return lltok::Error;
856 }
857
858 // At this point, it is either a label, int or fp constant.
859
860 // Skip digits, we have at least one.
861 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
862 /*empty*/;
863
864 // Check to see if this really is a label afterall, e.g. "-1:".
865 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
866 if (const char *End = isLabelTail(CurPtr)) {
867 StrVal.assign(TokStart, End-1);
868 CurPtr = End;
869 return lltok::LabelStr;
870 }
871 }
872
873 // If the next character is a '.', then it is a fp value, otherwise its
874 // integer.
875 if (CurPtr[0] != '.') {
876 if (TokStart[0] == '0' && TokStart[1] == 'x')
877 return Lex0x();
878 unsigned Len = CurPtr-TokStart;
879 uint32_t numBits = ((Len * 64) / 19) + 2;
880 APInt Tmp(numBits, StringRef(TokStart, Len), 10);
881 if (TokStart[0] == '-') {
882 uint32_t minBits = Tmp.getMinSignedBits();
883 if (minBits > 0 && minBits < numBits)
884 Tmp = Tmp.trunc(minBits);
885 APSIntVal = APSInt(Tmp, false);
886 } else {
887 uint32_t activeBits = Tmp.getActiveBits();
888 if (activeBits > 0 && activeBits < numBits)
889 Tmp = Tmp.trunc(activeBits);
890 APSIntVal = APSInt(Tmp, true);
891 }
892 return lltok::APSInt;
893 }
894
895 ++CurPtr;
896
897 // Skip over [0-9]*([eE][-+]?[0-9]+)?
898 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
899
900 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
901 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
902 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
903 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
904 CurPtr += 2;
905 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
906 }
907 }
908
909 APFloatVal = APFloat(std::atof(TokStart));
910 return lltok::APFloat;
911 }
912
913 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
LexPositive()914 lltok::Kind LLLexer::LexPositive() {
915 // If the letter after the negative is a number, this is probably not a
916 // label.
917 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
918 return lltok::Error;
919
920 // Skip digits.
921 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
922 /*empty*/;
923
924 // At this point, we need a '.'.
925 if (CurPtr[0] != '.') {
926 CurPtr = TokStart+1;
927 return lltok::Error;
928 }
929
930 ++CurPtr;
931
932 // Skip over [0-9]*([eE][-+]?[0-9]+)?
933 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
934
935 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
936 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
937 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
938 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
939 CurPtr += 2;
940 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
941 }
942 }
943
944 APFloatVal = APFloat(std::atof(TokStart));
945 return lltok::APFloat;
946 }
947