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 if (End - Buffer >= 16) {
82 for (int i = 0; i < 16; i++, Buffer++) {
83 assert(Buffer != End);
84 Pair[0] *= 16;
85 Pair[0] += hexDigitValue(*Buffer);
86 }
87 }
88 Pair[1] = 0;
89 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
90 Pair[1] *= 16;
91 Pair[1] += hexDigitValue(*Buffer);
92 }
93 if (Buffer != End)
94 Error("constant bigger than 128 bits detected!");
95 }
96
97 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
98 /// { low64, high16 } as usual for an APInt.
FP80HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])99 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
100 uint64_t Pair[2]) {
101 Pair[1] = 0;
102 for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
103 assert(Buffer != End);
104 Pair[1] *= 16;
105 Pair[1] += hexDigitValue(*Buffer);
106 }
107 Pair[0] = 0;
108 for (int i=0; i<16; i++, Buffer++) {
109 Pair[0] *= 16;
110 Pair[0] += hexDigitValue(*Buffer);
111 }
112 if (Buffer != End)
113 Error("constant bigger than 128 bits detected!");
114 }
115
116 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
117 // appropriate character.
UnEscapeLexed(std::string & Str)118 static void UnEscapeLexed(std::string &Str) {
119 if (Str.empty()) return;
120
121 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
122 char *BOut = Buffer;
123 for (char *BIn = Buffer; BIn != EndBuffer; ) {
124 if (BIn[0] == '\\') {
125 if (BIn < EndBuffer-1 && BIn[1] == '\\') {
126 *BOut++ = '\\'; // Two \ becomes one
127 BIn += 2;
128 } else if (BIn < EndBuffer-2 &&
129 isxdigit(static_cast<unsigned char>(BIn[1])) &&
130 isxdigit(static_cast<unsigned char>(BIn[2]))) {
131 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
132 BIn += 3; // Skip over handled chars
133 ++BOut;
134 } else {
135 *BOut++ = *BIn++;
136 }
137 } else {
138 *BOut++ = *BIn++;
139 }
140 }
141 Str.resize(BOut-Buffer);
142 }
143
144 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
isLabelChar(char C)145 static bool isLabelChar(char C) {
146 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
147 C == '.' || C == '_';
148 }
149
150
151 /// isLabelTail - Return true if this pointer points to a valid end of a label.
isLabelTail(const char * CurPtr)152 static const char *isLabelTail(const char *CurPtr) {
153 while (1) {
154 if (CurPtr[0] == ':') return CurPtr+1;
155 if (!isLabelChar(CurPtr[0])) return nullptr;
156 ++CurPtr;
157 }
158 }
159
160
161
162 //===----------------------------------------------------------------------===//
163 // Lexer definition.
164 //===----------------------------------------------------------------------===//
165
LLLexer(StringRef StartBuf,SourceMgr & sm,SMDiagnostic & Err,LLVMContext & C)166 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &sm, SMDiagnostic &Err,
167 LLVMContext &C)
168 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
169 CurPtr = CurBuf.begin();
170 }
171
getNextChar()172 int LLLexer::getNextChar() {
173 char CurChar = *CurPtr++;
174 switch (CurChar) {
175 default: return (unsigned char)CurChar;
176 case 0:
177 // A nul character in the stream is either the end of the current buffer or
178 // a random nul in the file. Disambiguate that here.
179 if (CurPtr-1 != CurBuf.end())
180 return 0; // Just whitespace.
181
182 // Otherwise, return end of file.
183 --CurPtr; // Another call to lex will return EOF again.
184 return EOF;
185 }
186 }
187
188
LexToken()189 lltok::Kind LLLexer::LexToken() {
190 TokStart = CurPtr;
191
192 int CurChar = getNextChar();
193 switch (CurChar) {
194 default:
195 // Handle letters: [a-zA-Z_]
196 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
197 return LexIdentifier();
198
199 return lltok::Error;
200 case EOF: return lltok::Eof;
201 case 0:
202 case ' ':
203 case '\t':
204 case '\n':
205 case '\r':
206 // Ignore whitespace.
207 return LexToken();
208 case '+': return LexPositive();
209 case '@': return LexAt();
210 case '$': return LexDollar();
211 case '%': return LexPercent();
212 case '"': return LexQuote();
213 case '.':
214 if (const char *Ptr = isLabelTail(CurPtr)) {
215 CurPtr = Ptr;
216 StrVal.assign(TokStart, CurPtr-1);
217 return lltok::LabelStr;
218 }
219 if (CurPtr[0] == '.' && CurPtr[1] == '.') {
220 CurPtr += 2;
221 return lltok::dotdotdot;
222 }
223 return lltok::Error;
224 case ';':
225 SkipLineComment();
226 return LexToken();
227 case '!': return LexExclaim();
228 case '#': return LexHash();
229 case '0': case '1': case '2': case '3': case '4':
230 case '5': case '6': case '7': case '8': case '9':
231 case '-':
232 return LexDigitOrNegative();
233 case '=': return lltok::equal;
234 case '[': return lltok::lsquare;
235 case ']': return lltok::rsquare;
236 case '{': return lltok::lbrace;
237 case '}': return lltok::rbrace;
238 case '<': return lltok::less;
239 case '>': return lltok::greater;
240 case '(': return lltok::lparen;
241 case ')': return lltok::rparen;
242 case ',': return lltok::comma;
243 case '*': return lltok::star;
244 case '|': return lltok::bar;
245 }
246 }
247
SkipLineComment()248 void LLLexer::SkipLineComment() {
249 while (1) {
250 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
251 return;
252 }
253 }
254
255 /// LexAt - Lex all tokens that start with an @ character:
256 /// GlobalVar @\"[^\"]*\"
257 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
258 /// GlobalVarID @[0-9]+
LexAt()259 lltok::Kind LLLexer::LexAt() {
260 return LexVar(lltok::GlobalVar, lltok::GlobalID);
261 }
262
LexDollar()263 lltok::Kind LLLexer::LexDollar() {
264 if (const char *Ptr = isLabelTail(TokStart)) {
265 CurPtr = Ptr;
266 StrVal.assign(TokStart, CurPtr - 1);
267 return lltok::LabelStr;
268 }
269
270 // Handle DollarStringConstant: $\"[^\"]*\"
271 if (CurPtr[0] == '"') {
272 ++CurPtr;
273
274 while (1) {
275 int CurChar = getNextChar();
276
277 if (CurChar == EOF) {
278 Error("end of file in COMDAT variable name");
279 return lltok::Error;
280 }
281 if (CurChar == '"') {
282 StrVal.assign(TokStart + 2, CurPtr - 1);
283 UnEscapeLexed(StrVal);
284 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
285 Error("Null bytes are not allowed in names");
286 return lltok::Error;
287 }
288 return lltok::ComdatVar;
289 }
290 }
291 }
292
293 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
294 if (ReadVarName())
295 return lltok::ComdatVar;
296
297 return lltok::Error;
298 }
299
300 /// ReadString - Read a string until the closing quote.
ReadString(lltok::Kind kind)301 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
302 const char *Start = CurPtr;
303 while (1) {
304 int CurChar = getNextChar();
305
306 if (CurChar == EOF) {
307 Error("end of file in string constant");
308 return lltok::Error;
309 }
310 if (CurChar == '"') {
311 StrVal.assign(Start, CurPtr-1);
312 UnEscapeLexed(StrVal);
313 return kind;
314 }
315 }
316 }
317
318 /// ReadVarName - Read the rest of a token containing a variable name.
ReadVarName()319 bool LLLexer::ReadVarName() {
320 const char *NameStart = CurPtr;
321 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
322 CurPtr[0] == '-' || CurPtr[0] == '$' ||
323 CurPtr[0] == '.' || CurPtr[0] == '_') {
324 ++CurPtr;
325 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
326 CurPtr[0] == '-' || CurPtr[0] == '$' ||
327 CurPtr[0] == '.' || CurPtr[0] == '_')
328 ++CurPtr;
329
330 StrVal.assign(NameStart, CurPtr);
331 return true;
332 }
333 return false;
334 }
335
LexVar(lltok::Kind Var,lltok::Kind VarID)336 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
337 // Handle StringConstant: \"[^\"]*\"
338 if (CurPtr[0] == '"') {
339 ++CurPtr;
340
341 while (1) {
342 int CurChar = getNextChar();
343
344 if (CurChar == EOF) {
345 Error("end of file in global variable name");
346 return lltok::Error;
347 }
348 if (CurChar == '"') {
349 StrVal.assign(TokStart+2, CurPtr-1);
350 UnEscapeLexed(StrVal);
351 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
352 Error("Null bytes are not allowed in names");
353 return lltok::Error;
354 }
355 return Var;
356 }
357 }
358 }
359
360 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
361 if (ReadVarName())
362 return Var;
363
364 // Handle VarID: [0-9]+
365 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
366 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
367 /*empty*/;
368
369 uint64_t Val = atoull(TokStart+1, CurPtr);
370 if ((unsigned)Val != Val)
371 Error("invalid value number (too large)!");
372 UIntVal = unsigned(Val);
373 return VarID;
374 }
375 return lltok::Error;
376 }
377
378 /// LexPercent - Lex all tokens that start with a % character:
379 /// LocalVar ::= %\"[^\"]*\"
380 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
381 /// LocalVarID ::= %[0-9]+
LexPercent()382 lltok::Kind LLLexer::LexPercent() {
383 return LexVar(lltok::LocalVar, lltok::LocalVarID);
384 }
385
386 /// LexQuote - Lex all tokens that start with a " character:
387 /// QuoteLabel "[^"]+":
388 /// StringConstant "[^"]*"
LexQuote()389 lltok::Kind LLLexer::LexQuote() {
390 lltok::Kind kind = ReadString(lltok::StringConstant);
391 if (kind == lltok::Error || kind == lltok::Eof)
392 return kind;
393
394 if (CurPtr[0] == ':') {
395 ++CurPtr;
396 if (StringRef(StrVal).find_first_of(0) != StringRef::npos) {
397 Error("Null bytes are not allowed in names");
398 kind = lltok::Error;
399 } else {
400 kind = lltok::LabelStr;
401 }
402 }
403
404 return kind;
405 }
406
407 /// LexExclaim:
408 /// !foo
409 /// !
LexExclaim()410 lltok::Kind LLLexer::LexExclaim() {
411 // Lex a metadata name as a MetadataVar.
412 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
413 CurPtr[0] == '-' || CurPtr[0] == '$' ||
414 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
415 ++CurPtr;
416 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
417 CurPtr[0] == '-' || CurPtr[0] == '$' ||
418 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
419 ++CurPtr;
420
421 StrVal.assign(TokStart+1, CurPtr); // Skip !
422 UnEscapeLexed(StrVal);
423 return lltok::MetadataVar;
424 }
425 return lltok::exclaim;
426 }
427
428 /// LexHash - Lex all tokens that start with a # character:
429 /// AttrGrpID ::= #[0-9]+
LexHash()430 lltok::Kind LLLexer::LexHash() {
431 // Handle AttrGrpID: #[0-9]+
432 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
433 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
434 /*empty*/;
435
436 uint64_t Val = atoull(TokStart+1, CurPtr);
437 if ((unsigned)Val != Val)
438 Error("invalid value number (too large)!");
439 UIntVal = unsigned(Val);
440 return lltok::AttrGrpID;
441 }
442
443 return lltok::Error;
444 }
445
446 /// LexIdentifier: Handle several related productions:
447 /// Label [-a-zA-Z$._0-9]+:
448 /// IntegerType i[0-9]+
449 /// Keyword sdiv, float, ...
450 /// HexIntConstant [us]0x[0-9A-Fa-f]+
LexIdentifier()451 lltok::Kind LLLexer::LexIdentifier() {
452 const char *StartChar = CurPtr;
453 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
454 const char *KeywordEnd = nullptr;
455
456 for (; isLabelChar(*CurPtr); ++CurPtr) {
457 // If we decide this is an integer, remember the end of the sequence.
458 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
459 IntEnd = CurPtr;
460 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
461 *CurPtr != '_')
462 KeywordEnd = CurPtr;
463 }
464
465 // If we stopped due to a colon, this really is a label.
466 if (*CurPtr == ':') {
467 StrVal.assign(StartChar-1, CurPtr++);
468 return lltok::LabelStr;
469 }
470
471 // Otherwise, this wasn't a label. If this was valid as an integer type,
472 // return it.
473 if (!IntEnd) IntEnd = CurPtr;
474 if (IntEnd != StartChar) {
475 CurPtr = IntEnd;
476 uint64_t NumBits = atoull(StartChar, CurPtr);
477 if (NumBits < IntegerType::MIN_INT_BITS ||
478 NumBits > IntegerType::MAX_INT_BITS) {
479 Error("bitwidth for integer type out of range!");
480 return lltok::Error;
481 }
482 TyVal = IntegerType::get(Context, NumBits);
483 return lltok::Type;
484 }
485
486 // Otherwise, this was a letter sequence. See which keyword this is.
487 if (!KeywordEnd) KeywordEnd = CurPtr;
488 CurPtr = KeywordEnd;
489 --StartChar;
490 StringRef Keyword(StartChar, CurPtr - StartChar);
491 #define KEYWORD(STR) \
492 do { \
493 if (Keyword == #STR) \
494 return lltok::kw_##STR; \
495 } while (0)
496
497 KEYWORD(true); KEYWORD(false);
498 KEYWORD(declare); KEYWORD(define);
499 KEYWORD(global); KEYWORD(constant);
500
501 KEYWORD(private);
502 KEYWORD(internal);
503 KEYWORD(available_externally);
504 KEYWORD(linkonce);
505 KEYWORD(linkonce_odr);
506 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
507 KEYWORD(weak_odr);
508 KEYWORD(appending);
509 KEYWORD(dllimport);
510 KEYWORD(dllexport);
511 KEYWORD(common);
512 KEYWORD(default);
513 KEYWORD(hidden);
514 KEYWORD(protected);
515 KEYWORD(unnamed_addr);
516 KEYWORD(externally_initialized);
517 KEYWORD(extern_weak);
518 KEYWORD(external);
519 KEYWORD(thread_local);
520 KEYWORD(localdynamic);
521 KEYWORD(initialexec);
522 KEYWORD(localexec);
523 KEYWORD(zeroinitializer);
524 KEYWORD(undef);
525 KEYWORD(null);
526 KEYWORD(to);
527 KEYWORD(tail);
528 KEYWORD(musttail);
529 KEYWORD(target);
530 KEYWORD(triple);
531 KEYWORD(unwind);
532 KEYWORD(deplibs); // FIXME: Remove in 4.0.
533 KEYWORD(datalayout);
534 KEYWORD(volatile);
535 KEYWORD(atomic);
536 KEYWORD(unordered);
537 KEYWORD(monotonic);
538 KEYWORD(acquire);
539 KEYWORD(release);
540 KEYWORD(acq_rel);
541 KEYWORD(seq_cst);
542 KEYWORD(singlethread);
543
544 KEYWORD(nnan);
545 KEYWORD(ninf);
546 KEYWORD(nsz);
547 KEYWORD(arcp);
548 KEYWORD(fast);
549 KEYWORD(nuw);
550 KEYWORD(nsw);
551 KEYWORD(exact);
552 KEYWORD(inbounds);
553 KEYWORD(align);
554 KEYWORD(addrspace);
555 KEYWORD(section);
556 KEYWORD(alias);
557 KEYWORD(module);
558 KEYWORD(asm);
559 KEYWORD(sideeffect);
560 KEYWORD(alignstack);
561 KEYWORD(inteldialect);
562 KEYWORD(gc);
563 KEYWORD(prefix);
564 KEYWORD(prologue);
565
566 KEYWORD(ccc);
567 KEYWORD(fastcc);
568 KEYWORD(coldcc);
569 KEYWORD(x86_stdcallcc);
570 KEYWORD(x86_fastcallcc);
571 KEYWORD(x86_thiscallcc);
572 KEYWORD(x86_vectorcallcc);
573 KEYWORD(arm_apcscc);
574 KEYWORD(arm_aapcscc);
575 KEYWORD(arm_aapcs_vfpcc);
576 KEYWORD(msp430_intrcc);
577 KEYWORD(ptx_kernel);
578 KEYWORD(ptx_device);
579 KEYWORD(spir_kernel);
580 KEYWORD(spir_func);
581 KEYWORD(intel_ocl_bicc);
582 KEYWORD(x86_64_sysvcc);
583 KEYWORD(x86_64_win64cc);
584 KEYWORD(webkit_jscc);
585 KEYWORD(anyregcc);
586 KEYWORD(preserve_mostcc);
587 KEYWORD(preserve_allcc);
588 KEYWORD(ghccc);
589
590 KEYWORD(cc);
591 KEYWORD(c);
592
593 KEYWORD(attributes);
594
595 KEYWORD(alwaysinline);
596 KEYWORD(builtin);
597 KEYWORD(byval);
598 KEYWORD(inalloca);
599 KEYWORD(cold);
600 KEYWORD(dereferenceable);
601 KEYWORD(dereferenceable_or_null);
602 KEYWORD(inlinehint);
603 KEYWORD(inreg);
604 KEYWORD(jumptable);
605 KEYWORD(minsize);
606 KEYWORD(naked);
607 KEYWORD(nest);
608 KEYWORD(noalias);
609 KEYWORD(nobuiltin);
610 KEYWORD(nocapture);
611 KEYWORD(noduplicate);
612 KEYWORD(noimplicitfloat);
613 KEYWORD(noinline);
614 KEYWORD(nonlazybind);
615 KEYWORD(nonnull);
616 KEYWORD(noredzone);
617 KEYWORD(noreturn);
618 KEYWORD(nounwind);
619 KEYWORD(optnone);
620 KEYWORD(optsize);
621 KEYWORD(readnone);
622 KEYWORD(readonly);
623 KEYWORD(returned);
624 KEYWORD(returns_twice);
625 KEYWORD(signext);
626 KEYWORD(sret);
627 KEYWORD(ssp);
628 KEYWORD(sspreq);
629 KEYWORD(sspstrong);
630 KEYWORD(sanitize_address);
631 KEYWORD(sanitize_thread);
632 KEYWORD(sanitize_memory);
633 KEYWORD(uwtable);
634 KEYWORD(zeroext);
635
636 KEYWORD(type);
637 KEYWORD(opaque);
638
639 KEYWORD(comdat);
640
641 // Comdat types
642 KEYWORD(any);
643 KEYWORD(exactmatch);
644 KEYWORD(largest);
645 KEYWORD(noduplicates);
646 KEYWORD(samesize);
647
648 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
649 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
650 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
651 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
652
653 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
654 KEYWORD(umin);
655
656 KEYWORD(x);
657 KEYWORD(blockaddress);
658
659 // Metadata types.
660 KEYWORD(distinct);
661
662 // Use-list order directives.
663 KEYWORD(uselistorder);
664 KEYWORD(uselistorder_bb);
665
666 KEYWORD(personality);
667 KEYWORD(cleanup);
668 KEYWORD(catch);
669 KEYWORD(filter);
670 #undef KEYWORD
671
672 // Keywords for types.
673 #define TYPEKEYWORD(STR, LLVMTY) \
674 do { \
675 if (Keyword == STR) { \
676 TyVal = LLVMTY; \
677 return lltok::Type; \
678 } \
679 } while (false)
680 TYPEKEYWORD("void", Type::getVoidTy(Context));
681 TYPEKEYWORD("half", Type::getHalfTy(Context));
682 TYPEKEYWORD("float", Type::getFloatTy(Context));
683 TYPEKEYWORD("double", Type::getDoubleTy(Context));
684 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
685 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
686 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
687 TYPEKEYWORD("label", Type::getLabelTy(Context));
688 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
689 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
690 #undef TYPEKEYWORD
691
692 // Keywords for instructions.
693 #define INSTKEYWORD(STR, Enum) \
694 do { \
695 if (Keyword == #STR) { \
696 UIntVal = Instruction::Enum; \
697 return lltok::kw_##STR; \
698 } \
699 } while (false)
700
701 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
702 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
703 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
704 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
705 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
706 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
707 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
708 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
709
710 INSTKEYWORD(phi, PHI);
711 INSTKEYWORD(call, Call);
712 INSTKEYWORD(trunc, Trunc);
713 INSTKEYWORD(zext, ZExt);
714 INSTKEYWORD(sext, SExt);
715 INSTKEYWORD(fptrunc, FPTrunc);
716 INSTKEYWORD(fpext, FPExt);
717 INSTKEYWORD(uitofp, UIToFP);
718 INSTKEYWORD(sitofp, SIToFP);
719 INSTKEYWORD(fptoui, FPToUI);
720 INSTKEYWORD(fptosi, FPToSI);
721 INSTKEYWORD(inttoptr, IntToPtr);
722 INSTKEYWORD(ptrtoint, PtrToInt);
723 INSTKEYWORD(bitcast, BitCast);
724 INSTKEYWORD(addrspacecast, AddrSpaceCast);
725 INSTKEYWORD(select, Select);
726 INSTKEYWORD(va_arg, VAArg);
727 INSTKEYWORD(ret, Ret);
728 INSTKEYWORD(br, Br);
729 INSTKEYWORD(switch, Switch);
730 INSTKEYWORD(indirectbr, IndirectBr);
731 INSTKEYWORD(invoke, Invoke);
732 INSTKEYWORD(resume, Resume);
733 INSTKEYWORD(unreachable, Unreachable);
734
735 INSTKEYWORD(alloca, Alloca);
736 INSTKEYWORD(load, Load);
737 INSTKEYWORD(store, Store);
738 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
739 INSTKEYWORD(atomicrmw, AtomicRMW);
740 INSTKEYWORD(fence, Fence);
741 INSTKEYWORD(getelementptr, GetElementPtr);
742
743 INSTKEYWORD(extractelement, ExtractElement);
744 INSTKEYWORD(insertelement, InsertElement);
745 INSTKEYWORD(shufflevector, ShuffleVector);
746 INSTKEYWORD(extractvalue, ExtractValue);
747 INSTKEYWORD(insertvalue, InsertValue);
748 INSTKEYWORD(landingpad, LandingPad);
749 #undef INSTKEYWORD
750
751 #define DWKEYWORD(TYPE, TOKEN) \
752 do { \
753 if (Keyword.startswith("DW_" #TYPE "_")) { \
754 StrVal.assign(Keyword.begin(), Keyword.end()); \
755 return lltok::TOKEN; \
756 } \
757 } while (false)
758 DWKEYWORD(TAG, DwarfTag);
759 DWKEYWORD(ATE, DwarfAttEncoding);
760 DWKEYWORD(VIRTUALITY, DwarfVirtuality);
761 DWKEYWORD(LANG, DwarfLang);
762 DWKEYWORD(OP, DwarfOp);
763 #undef DWKEYWORD
764
765 if (Keyword.startswith("DIFlag")) {
766 StrVal.assign(Keyword.begin(), Keyword.end());
767 return lltok::DIFlag;
768 }
769
770 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
771 // the CFE to avoid forcing it to deal with 64-bit numbers.
772 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
773 TokStart[1] == '0' && TokStart[2] == 'x' &&
774 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
775 int len = CurPtr-TokStart-3;
776 uint32_t bits = len * 4;
777 StringRef HexStr(TokStart + 3, len);
778 if (!std::all_of(HexStr.begin(), HexStr.end(), isxdigit)) {
779 // Bad token, return it as an error.
780 CurPtr = TokStart+3;
781 return lltok::Error;
782 }
783 APInt Tmp(bits, HexStr, 16);
784 uint32_t activeBits = Tmp.getActiveBits();
785 if (activeBits > 0 && activeBits < bits)
786 Tmp = Tmp.trunc(activeBits);
787 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
788 return lltok::APSInt;
789 }
790
791 // If this is "cc1234", return this as just "cc".
792 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
793 CurPtr = TokStart+2;
794 return lltok::kw_cc;
795 }
796
797 // Finally, if this isn't known, return an error.
798 CurPtr = TokStart+1;
799 return lltok::Error;
800 }
801
802
803 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
804 /// that this is not a label:
805 /// HexFPConstant 0x[0-9A-Fa-f]+
806 /// HexFP80Constant 0xK[0-9A-Fa-f]+
807 /// HexFP128Constant 0xL[0-9A-Fa-f]+
808 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
809 /// HexHalfConstant 0xH[0-9A-Fa-f]+
Lex0x()810 lltok::Kind LLLexer::Lex0x() {
811 CurPtr = TokStart + 2;
812
813 char Kind;
814 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
815 Kind = *CurPtr++;
816 } else {
817 Kind = 'J';
818 }
819
820 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
821 // Bad token, return it as an error.
822 CurPtr = TokStart+1;
823 return lltok::Error;
824 }
825
826 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
827 ++CurPtr;
828
829 if (Kind == 'J') {
830 // HexFPConstant - Floating point constant represented in IEEE format as a
831 // hexadecimal number for when exponential notation is not precise enough.
832 // Half, Float, and double only.
833 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
834 return lltok::APFloat;
835 }
836
837 uint64_t Pair[2];
838 switch (Kind) {
839 default: llvm_unreachable("Unknown kind!");
840 case 'K':
841 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
842 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
843 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair));
844 return lltok::APFloat;
845 case 'L':
846 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
847 HexToIntPair(TokStart+3, CurPtr, Pair);
848 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair));
849 return lltok::APFloat;
850 case 'M':
851 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
852 HexToIntPair(TokStart+3, CurPtr, Pair);
853 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair));
854 return lltok::APFloat;
855 case 'H':
856 APFloatVal = APFloat(APFloat::IEEEhalf,
857 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
858 return lltok::APFloat;
859 }
860 }
861
862 /// LexIdentifier: Handle several related productions:
863 /// Label [-a-zA-Z$._0-9]+:
864 /// NInteger -[0-9]+
865 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
866 /// PInteger [0-9]+
867 /// HexFPConstant 0x[0-9A-Fa-f]+
868 /// HexFP80Constant 0xK[0-9A-Fa-f]+
869 /// HexFP128Constant 0xL[0-9A-Fa-f]+
870 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
LexDigitOrNegative()871 lltok::Kind LLLexer::LexDigitOrNegative() {
872 // If the letter after the negative is not a number, this is probably a label.
873 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
874 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
875 // Okay, this is not a number after the -, it's probably a label.
876 if (const char *End = isLabelTail(CurPtr)) {
877 StrVal.assign(TokStart, End-1);
878 CurPtr = End;
879 return lltok::LabelStr;
880 }
881
882 return lltok::Error;
883 }
884
885 // At this point, it is either a label, int or fp constant.
886
887 // Skip digits, we have at least one.
888 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
889 /*empty*/;
890
891 // Check to see if this really is a label afterall, e.g. "-1:".
892 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
893 if (const char *End = isLabelTail(CurPtr)) {
894 StrVal.assign(TokStart, End-1);
895 CurPtr = End;
896 return lltok::LabelStr;
897 }
898 }
899
900 // If the next character is a '.', then it is a fp value, otherwise its
901 // integer.
902 if (CurPtr[0] != '.') {
903 if (TokStart[0] == '0' && TokStart[1] == 'x')
904 return Lex0x();
905 unsigned Len = CurPtr-TokStart;
906 uint32_t numBits = ((Len * 64) / 19) + 2;
907 APInt Tmp(numBits, StringRef(TokStart, Len), 10);
908 if (TokStart[0] == '-') {
909 uint32_t minBits = Tmp.getMinSignedBits();
910 if (minBits > 0 && minBits < numBits)
911 Tmp = Tmp.trunc(minBits);
912 APSIntVal = APSInt(Tmp, false);
913 } else {
914 uint32_t activeBits = Tmp.getActiveBits();
915 if (activeBits > 0 && activeBits < numBits)
916 Tmp = Tmp.trunc(activeBits);
917 APSIntVal = APSInt(Tmp, true);
918 }
919 return lltok::APSInt;
920 }
921
922 ++CurPtr;
923
924 // Skip over [0-9]*([eE][-+]?[0-9]+)?
925 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
926
927 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
928 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
929 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
930 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
931 CurPtr += 2;
932 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
933 }
934 }
935
936 APFloatVal = APFloat(std::atof(TokStart));
937 return lltok::APFloat;
938 }
939
940 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
LexPositive()941 lltok::Kind LLLexer::LexPositive() {
942 // If the letter after the negative is a number, this is probably not a
943 // label.
944 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
945 return lltok::Error;
946
947 // Skip digits.
948 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
949 /*empty*/;
950
951 // At this point, we need a '.'.
952 if (CurPtr[0] != '.') {
953 CurPtr = TokStart+1;
954 return lltok::Error;
955 }
956
957 ++CurPtr;
958
959 // Skip over [0-9]*([eE][-+]?[0-9]+)?
960 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
961
962 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
963 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
964 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
965 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
966 CurPtr += 2;
967 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
968 }
969 }
970
971 APFloatVal = APFloat(std::atof(TokStart));
972 return lltok::APFloat;
973 }
974