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>
29 using namespace llvm;
30
Error(LocTy ErrorLoc,const Twine & Msg) const31 bool 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 //
atoull(const char * Buffer,const char * End)44 uint64_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
parseHexChar(char C)58 static char parseHexChar(char C) {
59 if (C >= '0' && C <= '9')
60 return C-'0';
61 if (C >= 'A' && C <= 'F')
62 return C-'A'+10;
63 if (C >= 'a' && C <= 'f')
64 return C-'a'+10;
65 return 0;
66 }
67
HexIntToVal(const char * Buffer,const char * End)68 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
69 uint64_t Result = 0;
70 for (; Buffer != End; ++Buffer) {
71 uint64_t OldRes = Result;
72 Result *= 16;
73 Result += parseHexChar(*Buffer);
74
75 if (Result < OldRes) { // Uh, oh, overflow detected!!!
76 Error("constant bigger than 64 bits detected!");
77 return 0;
78 }
79 }
80 return Result;
81 }
82
HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])83 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
84 uint64_t Pair[2]) {
85 Pair[0] = 0;
86 for (int i=0; i<16; i++, Buffer++) {
87 assert(Buffer != End);
88 Pair[0] *= 16;
89 Pair[0] += parseHexChar(*Buffer);
90 }
91 Pair[1] = 0;
92 for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
93 Pair[1] *= 16;
94 Pair[1] += parseHexChar(*Buffer);
95 }
96 if (Buffer != End)
97 Error("constant bigger than 128 bits detected!");
98 }
99
100 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
101 /// { low64, high16 } as usual for an APInt.
FP80HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])102 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
103 uint64_t Pair[2]) {
104 Pair[1] = 0;
105 for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
106 assert(Buffer != End);
107 Pair[1] *= 16;
108 Pair[1] += parseHexChar(*Buffer);
109 }
110 Pair[0] = 0;
111 for (int i=0; i<16; i++, Buffer++) {
112 Pair[0] *= 16;
113 Pair[0] += parseHexChar(*Buffer);
114 }
115 if (Buffer != End)
116 Error("constant bigger than 128 bits detected!");
117 }
118
119 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
120 // appropriate character.
UnEscapeLexed(std::string & Str)121 static void UnEscapeLexed(std::string &Str) {
122 if (Str.empty()) return;
123
124 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
125 char *BOut = Buffer;
126 for (char *BIn = Buffer; BIn != EndBuffer; ) {
127 if (BIn[0] == '\\') {
128 if (BIn < EndBuffer-1 && BIn[1] == '\\') {
129 *BOut++ = '\\'; // Two \ becomes one
130 BIn += 2;
131 } else if (BIn < EndBuffer-2 && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
132 *BOut = parseHexChar(BIn[1]) * 16 + parseHexChar(BIn[2]);
133 BIn += 3; // Skip over handled chars
134 ++BOut;
135 } else {
136 *BOut++ = *BIn++;
137 }
138 } else {
139 *BOut++ = *BIn++;
140 }
141 }
142 Str.resize(BOut-Buffer);
143 }
144
145 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
isLabelChar(char C)146 static bool isLabelChar(char C) {
147 return isalnum(C) || C == '-' || C == '$' || 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 0;
156 ++CurPtr;
157 }
158 }
159
160
161
162 //===----------------------------------------------------------------------===//
163 // Lexer definition.
164 //===----------------------------------------------------------------------===//
165
LLLexer(MemoryBuffer * StartBuf,SourceMgr & sm,SMDiagnostic & Err,LLVMContext & C)166 LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err,
167 LLVMContext &C)
168 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
169 CurPtr = CurBuf->getBufferStart();
170 }
171
getFilename() const172 std::string LLLexer::getFilename() const {
173 return CurBuf->getBufferIdentifier();
174 }
175
getNextChar()176 int LLLexer::getNextChar() {
177 char CurChar = *CurPtr++;
178 switch (CurChar) {
179 default: return (unsigned char)CurChar;
180 case 0:
181 // A nul character in the stream is either the end of the current buffer or
182 // a random nul in the file. Disambiguate that here.
183 if (CurPtr-1 != CurBuf->getBufferEnd())
184 return 0; // Just whitespace.
185
186 // Otherwise, return end of file.
187 --CurPtr; // Another call to lex will return EOF again.
188 return EOF;
189 }
190 }
191
192
LexToken()193 lltok::Kind LLLexer::LexToken() {
194 TokStart = CurPtr;
195
196 int CurChar = getNextChar();
197 switch (CurChar) {
198 default:
199 // Handle letters: [a-zA-Z_]
200 if (isalpha(CurChar) || CurChar == '_')
201 return LexIdentifier();
202
203 return lltok::Error;
204 case EOF: return lltok::Eof;
205 case 0:
206 case ' ':
207 case '\t':
208 case '\n':
209 case '\r':
210 // Ignore whitespace.
211 return LexToken();
212 case '+': return LexPositive();
213 case '@': return LexAt();
214 case '%': return LexPercent();
215 case '"': return LexQuote();
216 case '.':
217 if (const char *Ptr = isLabelTail(CurPtr)) {
218 CurPtr = Ptr;
219 StrVal.assign(TokStart, CurPtr-1);
220 return lltok::LabelStr;
221 }
222 if (CurPtr[0] == '.' && CurPtr[1] == '.') {
223 CurPtr += 2;
224 return lltok::dotdotdot;
225 }
226 return lltok::Error;
227 case '$':
228 if (const char *Ptr = isLabelTail(CurPtr)) {
229 CurPtr = Ptr;
230 StrVal.assign(TokStart, CurPtr-1);
231 return lltok::LabelStr;
232 }
233 return lltok::Error;
234 case ';':
235 SkipLineComment();
236 return LexToken();
237 case '!': return LexExclaim();
238 case '0': case '1': case '2': case '3': case '4':
239 case '5': case '6': case '7': case '8': case '9':
240 case '-':
241 return LexDigitOrNegative();
242 case '=': return lltok::equal;
243 case '[': return lltok::lsquare;
244 case ']': return lltok::rsquare;
245 case '{': return lltok::lbrace;
246 case '}': return lltok::rbrace;
247 case '<': return lltok::less;
248 case '>': return lltok::greater;
249 case '(': return lltok::lparen;
250 case ')': return lltok::rparen;
251 case ',': return lltok::comma;
252 case '*': return lltok::star;
253 case '\\': return lltok::backslash;
254 }
255 }
256
SkipLineComment()257 void LLLexer::SkipLineComment() {
258 while (1) {
259 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
260 return;
261 }
262 }
263
264 /// LexAt - Lex all tokens that start with an @ character:
265 /// GlobalVar @\"[^\"]*\"
266 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
267 /// GlobalVarID @[0-9]+
LexAt()268 lltok::Kind LLLexer::LexAt() {
269 // Handle AtStringConstant: @\"[^\"]*\"
270 if (CurPtr[0] == '"') {
271 ++CurPtr;
272
273 while (1) {
274 int CurChar = getNextChar();
275
276 if (CurChar == EOF) {
277 Error("end of file in global variable name");
278 return lltok::Error;
279 }
280 if (CurChar == '"') {
281 StrVal.assign(TokStart+2, CurPtr-1);
282 UnEscapeLexed(StrVal);
283 return lltok::GlobalVar;
284 }
285 }
286 }
287
288 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
289 if (ReadVarName())
290 return lltok::GlobalVar;
291
292 // Handle GlobalVarID: @[0-9]+
293 if (isdigit(CurPtr[0])) {
294 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
295 /*empty*/;
296
297 uint64_t Val = atoull(TokStart+1, CurPtr);
298 if ((unsigned)Val != Val)
299 Error("invalid value number (too large)!");
300 UIntVal = unsigned(Val);
301 return lltok::GlobalID;
302 }
303
304 return lltok::Error;
305 }
306
307 /// ReadString - Read a string until the closing quote.
ReadString(lltok::Kind kind)308 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
309 const char *Start = CurPtr;
310 while (1) {
311 int CurChar = getNextChar();
312
313 if (CurChar == EOF) {
314 Error("end of file in string constant");
315 return lltok::Error;
316 }
317 if (CurChar == '"') {
318 StrVal.assign(Start, CurPtr-1);
319 UnEscapeLexed(StrVal);
320 return kind;
321 }
322 }
323 }
324
325 /// ReadVarName - Read the rest of a token containing a variable name.
ReadVarName()326 bool LLLexer::ReadVarName() {
327 const char *NameStart = CurPtr;
328 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
329 CurPtr[0] == '.' || CurPtr[0] == '_') {
330 ++CurPtr;
331 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
332 CurPtr[0] == '.' || CurPtr[0] == '_')
333 ++CurPtr;
334
335 StrVal.assign(NameStart, CurPtr);
336 return true;
337 }
338 return false;
339 }
340
341 /// LexPercent - Lex all tokens that start with a % character:
342 /// LocalVar ::= %\"[^\"]*\"
343 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
344 /// LocalVarID ::= %[0-9]+
LexPercent()345 lltok::Kind LLLexer::LexPercent() {
346 // Handle LocalVarName: %\"[^\"]*\"
347 if (CurPtr[0] == '"') {
348 ++CurPtr;
349 return ReadString(lltok::LocalVar);
350 }
351
352 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
353 if (ReadVarName())
354 return lltok::LocalVar;
355
356 // Handle LocalVarID: %[0-9]+
357 if (isdigit(CurPtr[0])) {
358 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
359 /*empty*/;
360
361 uint64_t Val = atoull(TokStart+1, CurPtr);
362 if ((unsigned)Val != Val)
363 Error("invalid value number (too large)!");
364 UIntVal = unsigned(Val);
365 return lltok::LocalVarID;
366 }
367
368 return lltok::Error;
369 }
370
371 /// LexQuote - Lex all tokens that start with a " character:
372 /// QuoteLabel "[^"]+":
373 /// StringConstant "[^"]*"
LexQuote()374 lltok::Kind LLLexer::LexQuote() {
375 lltok::Kind kind = ReadString(lltok::StringConstant);
376 if (kind == lltok::Error || kind == lltok::Eof)
377 return kind;
378
379 if (CurPtr[0] == ':') {
380 ++CurPtr;
381 kind = lltok::LabelStr;
382 }
383
384 return kind;
385 }
386
387 /// LexExclaim:
388 /// !foo
389 /// !
LexExclaim()390 lltok::Kind LLLexer::LexExclaim() {
391 // Lex a metadata name as a MetadataVar.
392 if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
393 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
394 ++CurPtr;
395 while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
396 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
397 ++CurPtr;
398
399 StrVal.assign(TokStart+1, CurPtr); // Skip !
400 UnEscapeLexed(StrVal);
401 return lltok::MetadataVar;
402 }
403 return lltok::exclaim;
404 }
405
406 /// LexIdentifier: Handle several related productions:
407 /// Label [-a-zA-Z$._0-9]+:
408 /// IntegerType i[0-9]+
409 /// Keyword sdiv, float, ...
410 /// HexIntConstant [us]0x[0-9A-Fa-f]+
LexIdentifier()411 lltok::Kind LLLexer::LexIdentifier() {
412 const char *StartChar = CurPtr;
413 const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
414 const char *KeywordEnd = 0;
415
416 for (; isLabelChar(*CurPtr); ++CurPtr) {
417 // If we decide this is an integer, remember the end of the sequence.
418 if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr;
419 if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr;
420 }
421
422 // If we stopped due to a colon, this really is a label.
423 if (*CurPtr == ':') {
424 StrVal.assign(StartChar-1, CurPtr++);
425 return lltok::LabelStr;
426 }
427
428 // Otherwise, this wasn't a label. If this was valid as an integer type,
429 // return it.
430 if (IntEnd == 0) IntEnd = CurPtr;
431 if (IntEnd != StartChar) {
432 CurPtr = IntEnd;
433 uint64_t NumBits = atoull(StartChar, CurPtr);
434 if (NumBits < IntegerType::MIN_INT_BITS ||
435 NumBits > IntegerType::MAX_INT_BITS) {
436 Error("bitwidth for integer type out of range!");
437 return lltok::Error;
438 }
439 TyVal = IntegerType::get(Context, NumBits);
440 return lltok::Type;
441 }
442
443 // Otherwise, this was a letter sequence. See which keyword this is.
444 if (KeywordEnd == 0) KeywordEnd = CurPtr;
445 CurPtr = KeywordEnd;
446 --StartChar;
447 unsigned Len = CurPtr-StartChar;
448 #define KEYWORD(STR) \
449 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
450 return lltok::kw_##STR;
451
452 KEYWORD(true); KEYWORD(false);
453 KEYWORD(declare); KEYWORD(define);
454 KEYWORD(global); KEYWORD(constant);
455
456 KEYWORD(private);
457 KEYWORD(linker_private);
458 KEYWORD(linker_private_weak);
459 KEYWORD(linker_private_weak_def_auto);
460 KEYWORD(internal);
461 KEYWORD(available_externally);
462 KEYWORD(linkonce);
463 KEYWORD(linkonce_odr);
464 KEYWORD(weak);
465 KEYWORD(weak_odr);
466 KEYWORD(appending);
467 KEYWORD(dllimport);
468 KEYWORD(dllexport);
469 KEYWORD(common);
470 KEYWORD(default);
471 KEYWORD(hidden);
472 KEYWORD(protected);
473 KEYWORD(unnamed_addr);
474 KEYWORD(extern_weak);
475 KEYWORD(external);
476 KEYWORD(thread_local);
477 KEYWORD(zeroinitializer);
478 KEYWORD(undef);
479 KEYWORD(null);
480 KEYWORD(to);
481 KEYWORD(tail);
482 KEYWORD(target);
483 KEYWORD(triple);
484 KEYWORD(unwind);
485 KEYWORD(deplibs);
486 KEYWORD(datalayout);
487 KEYWORD(volatile);
488 KEYWORD(atomic);
489 KEYWORD(unordered);
490 KEYWORD(monotonic);
491 KEYWORD(acquire);
492 KEYWORD(release);
493 KEYWORD(acq_rel);
494 KEYWORD(seq_cst);
495 KEYWORD(singlethread);
496
497 KEYWORD(nuw);
498 KEYWORD(nsw);
499 KEYWORD(exact);
500 KEYWORD(inbounds);
501 KEYWORD(align);
502 KEYWORD(addrspace);
503 KEYWORD(section);
504 KEYWORD(alias);
505 KEYWORD(module);
506 KEYWORD(asm);
507 KEYWORD(sideeffect);
508 KEYWORD(alignstack);
509 KEYWORD(gc);
510
511 KEYWORD(ccc);
512 KEYWORD(fastcc);
513 KEYWORD(coldcc);
514 KEYWORD(x86_stdcallcc);
515 KEYWORD(x86_fastcallcc);
516 KEYWORD(x86_thiscallcc);
517 KEYWORD(arm_apcscc);
518 KEYWORD(arm_aapcscc);
519 KEYWORD(arm_aapcs_vfpcc);
520 KEYWORD(msp430_intrcc);
521 KEYWORD(ptx_kernel);
522 KEYWORD(ptx_device);
523
524 KEYWORD(cc);
525 KEYWORD(c);
526
527 KEYWORD(signext);
528 KEYWORD(zeroext);
529 KEYWORD(inreg);
530 KEYWORD(sret);
531 KEYWORD(nounwind);
532 KEYWORD(noreturn);
533 KEYWORD(noalias);
534 KEYWORD(nocapture);
535 KEYWORD(byval);
536 KEYWORD(nest);
537 KEYWORD(readnone);
538 KEYWORD(readonly);
539 KEYWORD(uwtable);
540 KEYWORD(returns_twice);
541
542 KEYWORD(inlinehint);
543 KEYWORD(noinline);
544 KEYWORD(alwaysinline);
545 KEYWORD(optsize);
546 KEYWORD(ssp);
547 KEYWORD(sspreq);
548 KEYWORD(noredzone);
549 KEYWORD(noimplicitfloat);
550 KEYWORD(naked);
551 KEYWORD(nonlazybind);
552 KEYWORD(address_safety);
553
554 KEYWORD(type);
555 KEYWORD(opaque);
556
557 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
558 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
559 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
560 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
561
562 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
563 KEYWORD(umin);
564
565 KEYWORD(x);
566 KEYWORD(blockaddress);
567
568 KEYWORD(personality);
569 KEYWORD(cleanup);
570 KEYWORD(catch);
571 KEYWORD(filter);
572 #undef KEYWORD
573
574 // Keywords for types.
575 #define TYPEKEYWORD(STR, LLVMTY) \
576 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
577 TyVal = LLVMTY; return lltok::Type; }
578 TYPEKEYWORD("void", Type::getVoidTy(Context));
579 TYPEKEYWORD("half", Type::getHalfTy(Context));
580 TYPEKEYWORD("float", Type::getFloatTy(Context));
581 TYPEKEYWORD("double", Type::getDoubleTy(Context));
582 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
583 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
584 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
585 TYPEKEYWORD("label", Type::getLabelTy(Context));
586 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
587 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
588 #undef TYPEKEYWORD
589
590 // Keywords for instructions.
591 #define INSTKEYWORD(STR, Enum) \
592 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
593 UIntVal = Instruction::Enum; return lltok::kw_##STR; }
594
595 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
596 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
597 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
598 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
599 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
600 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
601 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
602 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
603
604 INSTKEYWORD(phi, PHI);
605 INSTKEYWORD(call, Call);
606 INSTKEYWORD(trunc, Trunc);
607 INSTKEYWORD(zext, ZExt);
608 INSTKEYWORD(sext, SExt);
609 INSTKEYWORD(fptrunc, FPTrunc);
610 INSTKEYWORD(fpext, FPExt);
611 INSTKEYWORD(uitofp, UIToFP);
612 INSTKEYWORD(sitofp, SIToFP);
613 INSTKEYWORD(fptoui, FPToUI);
614 INSTKEYWORD(fptosi, FPToSI);
615 INSTKEYWORD(inttoptr, IntToPtr);
616 INSTKEYWORD(ptrtoint, PtrToInt);
617 INSTKEYWORD(bitcast, BitCast);
618 INSTKEYWORD(select, Select);
619 INSTKEYWORD(va_arg, VAArg);
620 INSTKEYWORD(ret, Ret);
621 INSTKEYWORD(br, Br);
622 INSTKEYWORD(switch, Switch);
623 INSTKEYWORD(indirectbr, IndirectBr);
624 INSTKEYWORD(invoke, Invoke);
625 INSTKEYWORD(resume, Resume);
626 INSTKEYWORD(unreachable, Unreachable);
627
628 INSTKEYWORD(alloca, Alloca);
629 INSTKEYWORD(load, Load);
630 INSTKEYWORD(store, Store);
631 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
632 INSTKEYWORD(atomicrmw, AtomicRMW);
633 INSTKEYWORD(fence, Fence);
634 INSTKEYWORD(getelementptr, GetElementPtr);
635
636 INSTKEYWORD(extractelement, ExtractElement);
637 INSTKEYWORD(insertelement, InsertElement);
638 INSTKEYWORD(shufflevector, ShuffleVector);
639 INSTKEYWORD(extractvalue, ExtractValue);
640 INSTKEYWORD(insertvalue, InsertValue);
641 INSTKEYWORD(landingpad, LandingPad);
642 #undef INSTKEYWORD
643
644 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
645 // the CFE to avoid forcing it to deal with 64-bit numbers.
646 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
647 TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) {
648 int len = CurPtr-TokStart-3;
649 uint32_t bits = len * 4;
650 APInt Tmp(bits, StringRef(TokStart+3, len), 16);
651 uint32_t activeBits = Tmp.getActiveBits();
652 if (activeBits > 0 && activeBits < bits)
653 Tmp = Tmp.trunc(activeBits);
654 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
655 return lltok::APSInt;
656 }
657
658 // If this is "cc1234", return this as just "cc".
659 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
660 CurPtr = TokStart+2;
661 return lltok::kw_cc;
662 }
663
664 // Finally, if this isn't known, return an error.
665 CurPtr = TokStart+1;
666 return lltok::Error;
667 }
668
669
670 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
671 /// that this is not a label:
672 /// HexFPConstant 0x[0-9A-Fa-f]+
673 /// HexFP80Constant 0xK[0-9A-Fa-f]+
674 /// HexFP128Constant 0xL[0-9A-Fa-f]+
675 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
Lex0x()676 lltok::Kind LLLexer::Lex0x() {
677 CurPtr = TokStart + 2;
678
679 char Kind;
680 if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') {
681 Kind = *CurPtr++;
682 } else {
683 Kind = 'J';
684 }
685
686 if (!isxdigit(CurPtr[0])) {
687 // Bad token, return it as an error.
688 CurPtr = TokStart+1;
689 return lltok::Error;
690 }
691
692 while (isxdigit(CurPtr[0]))
693 ++CurPtr;
694
695 if (Kind == 'J') {
696 // HexFPConstant - Floating point constant represented in IEEE format as a
697 // hexadecimal number for when exponential notation is not precise enough.
698 // Half, Float, and double only.
699 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
700 return lltok::APFloat;
701 }
702
703 uint64_t Pair[2];
704 switch (Kind) {
705 default: llvm_unreachable("Unknown kind!");
706 case 'K':
707 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
708 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
709 APFloatVal = APFloat(APInt(80, Pair));
710 return lltok::APFloat;
711 case 'L':
712 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
713 HexToIntPair(TokStart+3, CurPtr, Pair);
714 APFloatVal = APFloat(APInt(128, Pair), true);
715 return lltok::APFloat;
716 case 'M':
717 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
718 HexToIntPair(TokStart+3, CurPtr, Pair);
719 APFloatVal = APFloat(APInt(128, Pair));
720 return lltok::APFloat;
721 }
722 }
723
724 /// LexIdentifier: Handle several related productions:
725 /// Label [-a-zA-Z$._0-9]+:
726 /// NInteger -[0-9]+
727 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
728 /// PInteger [0-9]+
729 /// HexFPConstant 0x[0-9A-Fa-f]+
730 /// HexFP80Constant 0xK[0-9A-Fa-f]+
731 /// HexFP128Constant 0xL[0-9A-Fa-f]+
732 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
LexDigitOrNegative()733 lltok::Kind LLLexer::LexDigitOrNegative() {
734 // If the letter after the negative is a number, this is probably a label.
735 if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) {
736 // Okay, this is not a number after the -, it's probably a label.
737 if (const char *End = isLabelTail(CurPtr)) {
738 StrVal.assign(TokStart, End-1);
739 CurPtr = End;
740 return lltok::LabelStr;
741 }
742
743 return lltok::Error;
744 }
745
746 // At this point, it is either a label, int or fp constant.
747
748 // Skip digits, we have at least one.
749 for (; isdigit(CurPtr[0]); ++CurPtr)
750 /*empty*/;
751
752 // Check to see if this really is a label afterall, e.g. "-1:".
753 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
754 if (const char *End = isLabelTail(CurPtr)) {
755 StrVal.assign(TokStart, End-1);
756 CurPtr = End;
757 return lltok::LabelStr;
758 }
759 }
760
761 // If the next character is a '.', then it is a fp value, otherwise its
762 // integer.
763 if (CurPtr[0] != '.') {
764 if (TokStart[0] == '0' && TokStart[1] == 'x')
765 return Lex0x();
766 unsigned Len = CurPtr-TokStart;
767 uint32_t numBits = ((Len * 64) / 19) + 2;
768 APInt Tmp(numBits, StringRef(TokStart, Len), 10);
769 if (TokStart[0] == '-') {
770 uint32_t minBits = Tmp.getMinSignedBits();
771 if (minBits > 0 && minBits < numBits)
772 Tmp = Tmp.trunc(minBits);
773 APSIntVal = APSInt(Tmp, false);
774 } else {
775 uint32_t activeBits = Tmp.getActiveBits();
776 if (activeBits > 0 && activeBits < numBits)
777 Tmp = Tmp.trunc(activeBits);
778 APSIntVal = APSInt(Tmp, true);
779 }
780 return lltok::APSInt;
781 }
782
783 ++CurPtr;
784
785 // Skip over [0-9]*([eE][-+]?[0-9]+)?
786 while (isdigit(CurPtr[0])) ++CurPtr;
787
788 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
789 if (isdigit(CurPtr[1]) ||
790 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
791 CurPtr += 2;
792 while (isdigit(CurPtr[0])) ++CurPtr;
793 }
794 }
795
796 APFloatVal = APFloat(std::atof(TokStart));
797 return lltok::APFloat;
798 }
799
800 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
LexPositive()801 lltok::Kind LLLexer::LexPositive() {
802 // If the letter after the negative is a number, this is probably not a
803 // label.
804 if (!isdigit(CurPtr[0]))
805 return lltok::Error;
806
807 // Skip digits.
808 for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
809 /*empty*/;
810
811 // At this point, we need a '.'.
812 if (CurPtr[0] != '.') {
813 CurPtr = TokStart+1;
814 return lltok::Error;
815 }
816
817 ++CurPtr;
818
819 // Skip over [0-9]*([eE][-+]?[0-9]+)?
820 while (isdigit(CurPtr[0])) ++CurPtr;
821
822 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
823 if (isdigit(CurPtr[1]) ||
824 ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
825 CurPtr += 2;
826 while (isdigit(CurPtr[0])) ++CurPtr;
827 }
828 }
829
830 APFloatVal = APFloat(std::atof(TokStart));
831 return lltok::APFloat;
832 }
833