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/Assembly/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
37 //===----------------------------------------------------------------------===//
38 // Helper functions.
39 //===----------------------------------------------------------------------===//
40
41 // atoull - Convert an ascii string of decimal digits into the unsigned long
42 // long representation... this does not have to do input error checking,
43 // because we know that the input will be matched by a suitable regex...
44 //
atoull(const char * Buffer,const char * End)45 uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
46 uint64_t Result = 0;
47 for (; Buffer != End; Buffer++) {
48 uint64_t OldRes = Result;
49 Result *= 10;
50 Result += *Buffer-'0';
51 if (Result < OldRes) { // Uh, oh, overflow detected!!!
52 Error("constant bigger than 64 bits detected!");
53 return 0;
54 }
55 }
56 return Result;
57 }
58
HexIntToVal(const char * Buffer,const char * End)59 uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
60 uint64_t Result = 0;
61 for (; Buffer != End; ++Buffer) {
62 uint64_t OldRes = Result;
63 Result *= 16;
64 Result += hexDigitValue(*Buffer);
65
66 if (Result < OldRes) { // Uh, oh, overflow detected!!!
67 Error("constant bigger than 64 bits detected!");
68 return 0;
69 }
70 }
71 return Result;
72 }
73
HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])74 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
75 uint64_t Pair[2]) {
76 Pair[0] = 0;
77 for (int i=0; i<16; i++, Buffer++) {
78 assert(Buffer != End);
79 Pair[0] *= 16;
80 Pair[0] += hexDigitValue(*Buffer);
81 }
82 Pair[1] = 0;
83 for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
84 Pair[1] *= 16;
85 Pair[1] += hexDigitValue(*Buffer);
86 }
87 if (Buffer != End)
88 Error("constant bigger than 128 bits detected!");
89 }
90
91 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
92 /// { low64, high16 } as usual for an APInt.
FP80HexToIntPair(const char * Buffer,const char * End,uint64_t Pair[2])93 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
94 uint64_t Pair[2]) {
95 Pair[1] = 0;
96 for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
97 assert(Buffer != End);
98 Pair[1] *= 16;
99 Pair[1] += hexDigitValue(*Buffer);
100 }
101 Pair[0] = 0;
102 for (int i=0; i<16; i++, Buffer++) {
103 Pair[0] *= 16;
104 Pair[0] += hexDigitValue(*Buffer);
105 }
106 if (Buffer != End)
107 Error("constant bigger than 128 bits detected!");
108 }
109
110 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
111 // appropriate character.
UnEscapeLexed(std::string & Str)112 static void UnEscapeLexed(std::string &Str) {
113 if (Str.empty()) return;
114
115 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
116 char *BOut = Buffer;
117 for (char *BIn = Buffer; BIn != EndBuffer; ) {
118 if (BIn[0] == '\\') {
119 if (BIn < EndBuffer-1 && BIn[1] == '\\') {
120 *BOut++ = '\\'; // Two \ becomes one
121 BIn += 2;
122 } else if (BIn < EndBuffer-2 &&
123 isxdigit(static_cast<unsigned char>(BIn[1])) &&
124 isxdigit(static_cast<unsigned char>(BIn[2]))) {
125 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
126 BIn += 3; // Skip over handled chars
127 ++BOut;
128 } else {
129 *BOut++ = *BIn++;
130 }
131 } else {
132 *BOut++ = *BIn++;
133 }
134 }
135 Str.resize(BOut-Buffer);
136 }
137
138 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
isLabelChar(char C)139 static bool isLabelChar(char C) {
140 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
141 C == '.' || C == '_';
142 }
143
144
145 /// isLabelTail - Return true if this pointer points to a valid end of a label.
isLabelTail(const char * CurPtr)146 static const char *isLabelTail(const char *CurPtr) {
147 while (1) {
148 if (CurPtr[0] == ':') return CurPtr+1;
149 if (!isLabelChar(CurPtr[0])) return 0;
150 ++CurPtr;
151 }
152 }
153
154
155
156 //===----------------------------------------------------------------------===//
157 // Lexer definition.
158 //===----------------------------------------------------------------------===//
159
LLLexer(MemoryBuffer * StartBuf,SourceMgr & sm,SMDiagnostic & Err,LLVMContext & C)160 LLLexer::LLLexer(MemoryBuffer *StartBuf, SourceMgr &sm, SMDiagnostic &Err,
161 LLVMContext &C)
162 : CurBuf(StartBuf), ErrorInfo(Err), SM(sm), Context(C), APFloatVal(0.0) {
163 CurPtr = CurBuf->getBufferStart();
164 }
165
getFilename() const166 std::string LLLexer::getFilename() const {
167 return CurBuf->getBufferIdentifier();
168 }
169
getNextChar()170 int LLLexer::getNextChar() {
171 char CurChar = *CurPtr++;
172 switch (CurChar) {
173 default: return (unsigned char)CurChar;
174 case 0:
175 // A nul character in the stream is either the end of the current buffer or
176 // a random nul in the file. Disambiguate that here.
177 if (CurPtr-1 != CurBuf->getBufferEnd())
178 return 0; // Just whitespace.
179
180 // Otherwise, return end of file.
181 --CurPtr; // Another call to lex will return EOF again.
182 return EOF;
183 }
184 }
185
186
LexToken()187 lltok::Kind LLLexer::LexToken() {
188 TokStart = CurPtr;
189
190 int CurChar = getNextChar();
191 switch (CurChar) {
192 default:
193 // Handle letters: [a-zA-Z_]
194 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
195 return LexIdentifier();
196
197 return lltok::Error;
198 case EOF: return lltok::Eof;
199 case 0:
200 case ' ':
201 case '\t':
202 case '\n':
203 case '\r':
204 // Ignore whitespace.
205 return LexToken();
206 case '+': return LexPositive();
207 case '@': return LexAt();
208 case '%': return LexPercent();
209 case '"': return LexQuote();
210 case '.':
211 if (const char *Ptr = isLabelTail(CurPtr)) {
212 CurPtr = Ptr;
213 StrVal.assign(TokStart, CurPtr-1);
214 return lltok::LabelStr;
215 }
216 if (CurPtr[0] == '.' && CurPtr[1] == '.') {
217 CurPtr += 2;
218 return lltok::dotdotdot;
219 }
220 return lltok::Error;
221 case '$':
222 if (const char *Ptr = isLabelTail(CurPtr)) {
223 CurPtr = Ptr;
224 StrVal.assign(TokStart, CurPtr-1);
225 return lltok::LabelStr;
226 }
227 return lltok::Error;
228 case ';':
229 SkipLineComment();
230 return LexToken();
231 case '!': return LexExclaim();
232 case '#': return LexHash();
233 case '0': case '1': case '2': case '3': case '4':
234 case '5': case '6': case '7': case '8': case '9':
235 case '-':
236 return LexDigitOrNegative();
237 case '=': return lltok::equal;
238 case '[': return lltok::lsquare;
239 case ']': return lltok::rsquare;
240 case '{': return lltok::lbrace;
241 case '}': return lltok::rbrace;
242 case '<': return lltok::less;
243 case '>': return lltok::greater;
244 case '(': return lltok::lparen;
245 case ')': return lltok::rparen;
246 case ',': return lltok::comma;
247 case '*': return lltok::star;
248 case '\\': return lltok::backslash;
249 }
250 }
251
SkipLineComment()252 void LLLexer::SkipLineComment() {
253 while (1) {
254 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
255 return;
256 }
257 }
258
259 /// LexAt - Lex all tokens that start with an @ character:
260 /// GlobalVar @\"[^\"]*\"
261 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
262 /// GlobalVarID @[0-9]+
LexAt()263 lltok::Kind LLLexer::LexAt() {
264 // Handle AtStringConstant: @\"[^\"]*\"
265 if (CurPtr[0] == '"') {
266 ++CurPtr;
267
268 while (1) {
269 int CurChar = getNextChar();
270
271 if (CurChar == EOF) {
272 Error("end of file in global variable name");
273 return lltok::Error;
274 }
275 if (CurChar == '"') {
276 StrVal.assign(TokStart+2, CurPtr-1);
277 UnEscapeLexed(StrVal);
278 return lltok::GlobalVar;
279 }
280 }
281 }
282
283 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
284 if (ReadVarName())
285 return lltok::GlobalVar;
286
287 // Handle GlobalVarID: @[0-9]+
288 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
289 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
290 /*empty*/;
291
292 uint64_t Val = atoull(TokStart+1, CurPtr);
293 if ((unsigned)Val != Val)
294 Error("invalid value number (too large)!");
295 UIntVal = unsigned(Val);
296 return lltok::GlobalID;
297 }
298
299 return lltok::Error;
300 }
301
302 /// ReadString - Read a string until the closing quote.
ReadString(lltok::Kind kind)303 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
304 const char *Start = CurPtr;
305 while (1) {
306 int CurChar = getNextChar();
307
308 if (CurChar == EOF) {
309 Error("end of file in string constant");
310 return lltok::Error;
311 }
312 if (CurChar == '"') {
313 StrVal.assign(Start, CurPtr-1);
314 UnEscapeLexed(StrVal);
315 return kind;
316 }
317 }
318 }
319
320 /// ReadVarName - Read the rest of a token containing a variable name.
ReadVarName()321 bool LLLexer::ReadVarName() {
322 const char *NameStart = CurPtr;
323 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
324 CurPtr[0] == '-' || CurPtr[0] == '$' ||
325 CurPtr[0] == '.' || CurPtr[0] == '_') {
326 ++CurPtr;
327 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
328 CurPtr[0] == '-' || CurPtr[0] == '$' ||
329 CurPtr[0] == '.' || CurPtr[0] == '_')
330 ++CurPtr;
331
332 StrVal.assign(NameStart, CurPtr);
333 return true;
334 }
335 return false;
336 }
337
338 /// LexPercent - Lex all tokens that start with a % character:
339 /// LocalVar ::= %\"[^\"]*\"
340 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
341 /// LocalVarID ::= %[0-9]+
LexPercent()342 lltok::Kind LLLexer::LexPercent() {
343 // Handle LocalVarName: %\"[^\"]*\"
344 if (CurPtr[0] == '"') {
345 ++CurPtr;
346 return ReadString(lltok::LocalVar);
347 }
348
349 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
350 if (ReadVarName())
351 return lltok::LocalVar;
352
353 // Handle LocalVarID: %[0-9]+
354 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
355 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
356 /*empty*/;
357
358 uint64_t Val = atoull(TokStart+1, CurPtr);
359 if ((unsigned)Val != Val)
360 Error("invalid value number (too large)!");
361 UIntVal = unsigned(Val);
362 return lltok::LocalVarID;
363 }
364
365 return lltok::Error;
366 }
367
368 /// LexQuote - Lex all tokens that start with a " character:
369 /// QuoteLabel "[^"]+":
370 /// StringConstant "[^"]*"
LexQuote()371 lltok::Kind LLLexer::LexQuote() {
372 lltok::Kind kind = ReadString(lltok::StringConstant);
373 if (kind == lltok::Error || kind == lltok::Eof)
374 return kind;
375
376 if (CurPtr[0] == ':') {
377 ++CurPtr;
378 kind = lltok::LabelStr;
379 }
380
381 return kind;
382 }
383
384 /// LexExclaim:
385 /// !foo
386 /// !
LexExclaim()387 lltok::Kind LLLexer::LexExclaim() {
388 // Lex a metadata name as a MetadataVar.
389 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
390 CurPtr[0] == '-' || CurPtr[0] == '$' ||
391 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
392 ++CurPtr;
393 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
394 CurPtr[0] == '-' || CurPtr[0] == '$' ||
395 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
396 ++CurPtr;
397
398 StrVal.assign(TokStart+1, CurPtr); // Skip !
399 UnEscapeLexed(StrVal);
400 return lltok::MetadataVar;
401 }
402 return lltok::exclaim;
403 }
404
405 /// LexHash - Lex all tokens that start with a # character:
406 /// AttrGrpID ::= #[0-9]+
LexHash()407 lltok::Kind LLLexer::LexHash() {
408 // Handle AttrGrpID: #[0-9]+
409 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
410 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
411 /*empty*/;
412
413 uint64_t Val = atoull(TokStart+1, CurPtr);
414 if ((unsigned)Val != Val)
415 Error("invalid value number (too large)!");
416 UIntVal = unsigned(Val);
417 return lltok::AttrGrpID;
418 }
419
420 return lltok::Error;
421 }
422
423 /// LexIdentifier: Handle several related productions:
424 /// Label [-a-zA-Z$._0-9]+:
425 /// IntegerType i[0-9]+
426 /// Keyword sdiv, float, ...
427 /// HexIntConstant [us]0x[0-9A-Fa-f]+
LexIdentifier()428 lltok::Kind LLLexer::LexIdentifier() {
429 const char *StartChar = CurPtr;
430 const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
431 const char *KeywordEnd = 0;
432
433 for (; isLabelChar(*CurPtr); ++CurPtr) {
434 // If we decide this is an integer, remember the end of the sequence.
435 if (!IntEnd && !isdigit(static_cast<unsigned char>(*CurPtr)))
436 IntEnd = CurPtr;
437 if (!KeywordEnd && !isalnum(static_cast<unsigned char>(*CurPtr)) &&
438 *CurPtr != '_')
439 KeywordEnd = CurPtr;
440 }
441
442 // If we stopped due to a colon, this really is a label.
443 if (*CurPtr == ':') {
444 StrVal.assign(StartChar-1, CurPtr++);
445 return lltok::LabelStr;
446 }
447
448 // Otherwise, this wasn't a label. If this was valid as an integer type,
449 // return it.
450 if (IntEnd == 0) IntEnd = CurPtr;
451 if (IntEnd != StartChar) {
452 CurPtr = IntEnd;
453 uint64_t NumBits = atoull(StartChar, CurPtr);
454 if (NumBits < IntegerType::MIN_INT_BITS ||
455 NumBits > IntegerType::MAX_INT_BITS) {
456 Error("bitwidth for integer type out of range!");
457 return lltok::Error;
458 }
459 TyVal = IntegerType::get(Context, NumBits);
460 return lltok::Type;
461 }
462
463 // Otherwise, this was a letter sequence. See which keyword this is.
464 if (KeywordEnd == 0) KeywordEnd = CurPtr;
465 CurPtr = KeywordEnd;
466 --StartChar;
467 unsigned Len = CurPtr-StartChar;
468 #define KEYWORD(STR) \
469 do { \
470 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
471 return lltok::kw_##STR; \
472 } while (0)
473
474 KEYWORD(true); KEYWORD(false);
475 KEYWORD(declare); KEYWORD(define);
476 KEYWORD(global); KEYWORD(constant);
477
478 KEYWORD(private);
479 KEYWORD(linker_private);
480 KEYWORD(linker_private_weak);
481 KEYWORD(linker_private_weak_def_auto); // FIXME: For backwards compatibility.
482 KEYWORD(internal);
483 KEYWORD(available_externally);
484 KEYWORD(linkonce);
485 KEYWORD(linkonce_odr);
486 KEYWORD(linkonce_odr_auto_hide);
487 KEYWORD(weak);
488 KEYWORD(weak_odr);
489 KEYWORD(appending);
490 KEYWORD(dllimport);
491 KEYWORD(dllexport);
492 KEYWORD(common);
493 KEYWORD(default);
494 KEYWORD(hidden);
495 KEYWORD(protected);
496 KEYWORD(unnamed_addr);
497 KEYWORD(externally_initialized);
498 KEYWORD(extern_weak);
499 KEYWORD(external);
500 KEYWORD(thread_local);
501 KEYWORD(localdynamic);
502 KEYWORD(initialexec);
503 KEYWORD(localexec);
504 KEYWORD(zeroinitializer);
505 KEYWORD(undef);
506 KEYWORD(null);
507 KEYWORD(to);
508 KEYWORD(tail);
509 KEYWORD(target);
510 KEYWORD(triple);
511 KEYWORD(unwind);
512 KEYWORD(deplibs); // FIXME: Remove in 4.0.
513 KEYWORD(datalayout);
514 KEYWORD(volatile);
515 KEYWORD(atomic);
516 KEYWORD(unordered);
517 KEYWORD(monotonic);
518 KEYWORD(acquire);
519 KEYWORD(release);
520 KEYWORD(acq_rel);
521 KEYWORD(seq_cst);
522 KEYWORD(singlethread);
523
524 KEYWORD(nnan);
525 KEYWORD(ninf);
526 KEYWORD(nsz);
527 KEYWORD(arcp);
528 KEYWORD(fast);
529 KEYWORD(nuw);
530 KEYWORD(nsw);
531 KEYWORD(exact);
532 KEYWORD(inbounds);
533 KEYWORD(align);
534 KEYWORD(addrspace);
535 KEYWORD(section);
536 KEYWORD(alias);
537 KEYWORD(module);
538 KEYWORD(asm);
539 KEYWORD(sideeffect);
540 KEYWORD(alignstack);
541 KEYWORD(inteldialect);
542 KEYWORD(gc);
543
544 KEYWORD(ccc);
545 KEYWORD(fastcc);
546 KEYWORD(coldcc);
547 KEYWORD(x86_stdcallcc);
548 KEYWORD(x86_fastcallcc);
549 KEYWORD(x86_thiscallcc);
550 KEYWORD(arm_apcscc);
551 KEYWORD(arm_aapcscc);
552 KEYWORD(arm_aapcs_vfpcc);
553 KEYWORD(msp430_intrcc);
554 KEYWORD(ptx_kernel);
555 KEYWORD(ptx_device);
556 KEYWORD(spir_kernel);
557 KEYWORD(spir_func);
558 KEYWORD(intel_ocl_bicc);
559
560 KEYWORD(cc);
561 KEYWORD(c);
562
563 KEYWORD(attributes);
564
565 KEYWORD(alwaysinline);
566 KEYWORD(byval);
567 KEYWORD(inlinehint);
568 KEYWORD(inreg);
569 KEYWORD(minsize);
570 KEYWORD(naked);
571 KEYWORD(nest);
572 KEYWORD(noalias);
573 KEYWORD(nobuiltin);
574 KEYWORD(nocapture);
575 KEYWORD(noduplicate);
576 KEYWORD(noimplicitfloat);
577 KEYWORD(noinline);
578 KEYWORD(nonlazybind);
579 KEYWORD(noredzone);
580 KEYWORD(noreturn);
581 KEYWORD(nounwind);
582 KEYWORD(optsize);
583 KEYWORD(readnone);
584 KEYWORD(readonly);
585 KEYWORD(returns_twice);
586 KEYWORD(signext);
587 KEYWORD(sret);
588 KEYWORD(ssp);
589 KEYWORD(sspreq);
590 KEYWORD(sspstrong);
591 KEYWORD(sanitize_address);
592 KEYWORD(sanitize_thread);
593 KEYWORD(sanitize_memory);
594 KEYWORD(uwtable);
595 KEYWORD(zeroext);
596
597 KEYWORD(type);
598 KEYWORD(opaque);
599
600 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
601 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
602 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
603 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
604
605 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
606 KEYWORD(umin);
607
608 KEYWORD(x);
609 KEYWORD(blockaddress);
610
611 KEYWORD(personality);
612 KEYWORD(cleanup);
613 KEYWORD(catch);
614 KEYWORD(filter);
615 #undef KEYWORD
616
617 // Keywords for types.
618 #define TYPEKEYWORD(STR, LLVMTY) \
619 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
620 TyVal = LLVMTY; return lltok::Type; }
621 TYPEKEYWORD("void", Type::getVoidTy(Context));
622 TYPEKEYWORD("half", Type::getHalfTy(Context));
623 TYPEKEYWORD("float", Type::getFloatTy(Context));
624 TYPEKEYWORD("double", Type::getDoubleTy(Context));
625 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
626 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
627 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
628 TYPEKEYWORD("label", Type::getLabelTy(Context));
629 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
630 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
631 #undef TYPEKEYWORD
632
633 // Keywords for instructions.
634 #define INSTKEYWORD(STR, Enum) \
635 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
636 UIntVal = Instruction::Enum; return lltok::kw_##STR; }
637
638 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
639 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
640 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
641 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
642 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
643 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
644 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
645 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
646
647 INSTKEYWORD(phi, PHI);
648 INSTKEYWORD(call, Call);
649 INSTKEYWORD(trunc, Trunc);
650 INSTKEYWORD(zext, ZExt);
651 INSTKEYWORD(sext, SExt);
652 INSTKEYWORD(fptrunc, FPTrunc);
653 INSTKEYWORD(fpext, FPExt);
654 INSTKEYWORD(uitofp, UIToFP);
655 INSTKEYWORD(sitofp, SIToFP);
656 INSTKEYWORD(fptoui, FPToUI);
657 INSTKEYWORD(fptosi, FPToSI);
658 INSTKEYWORD(inttoptr, IntToPtr);
659 INSTKEYWORD(ptrtoint, PtrToInt);
660 INSTKEYWORD(bitcast, BitCast);
661 INSTKEYWORD(select, Select);
662 INSTKEYWORD(va_arg, VAArg);
663 INSTKEYWORD(ret, Ret);
664 INSTKEYWORD(br, Br);
665 INSTKEYWORD(switch, Switch);
666 INSTKEYWORD(indirectbr, IndirectBr);
667 INSTKEYWORD(invoke, Invoke);
668 INSTKEYWORD(resume, Resume);
669 INSTKEYWORD(unreachable, Unreachable);
670
671 INSTKEYWORD(alloca, Alloca);
672 INSTKEYWORD(load, Load);
673 INSTKEYWORD(store, Store);
674 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
675 INSTKEYWORD(atomicrmw, AtomicRMW);
676 INSTKEYWORD(fence, Fence);
677 INSTKEYWORD(getelementptr, GetElementPtr);
678
679 INSTKEYWORD(extractelement, ExtractElement);
680 INSTKEYWORD(insertelement, InsertElement);
681 INSTKEYWORD(shufflevector, ShuffleVector);
682 INSTKEYWORD(extractvalue, ExtractValue);
683 INSTKEYWORD(insertvalue, InsertValue);
684 INSTKEYWORD(landingpad, LandingPad);
685 #undef INSTKEYWORD
686
687 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
688 // the CFE to avoid forcing it to deal with 64-bit numbers.
689 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
690 TokStart[1] == '0' && TokStart[2] == 'x' &&
691 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
692 int len = CurPtr-TokStart-3;
693 uint32_t bits = len * 4;
694 APInt Tmp(bits, StringRef(TokStart+3, len), 16);
695 uint32_t activeBits = Tmp.getActiveBits();
696 if (activeBits > 0 && activeBits < bits)
697 Tmp = Tmp.trunc(activeBits);
698 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
699 return lltok::APSInt;
700 }
701
702 // If this is "cc1234", return this as just "cc".
703 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
704 CurPtr = TokStart+2;
705 return lltok::kw_cc;
706 }
707
708 // Finally, if this isn't known, return an error.
709 CurPtr = TokStart+1;
710 return lltok::Error;
711 }
712
713
714 /// Lex0x: Handle productions that start with 0x, knowing that it matches and
715 /// that this is not a label:
716 /// HexFPConstant 0x[0-9A-Fa-f]+
717 /// HexFP80Constant 0xK[0-9A-Fa-f]+
718 /// HexFP128Constant 0xL[0-9A-Fa-f]+
719 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
720 /// HexHalfConstant 0xH[0-9A-Fa-f]+
Lex0x()721 lltok::Kind LLLexer::Lex0x() {
722 CurPtr = TokStart + 2;
723
724 char Kind;
725 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
726 Kind = *CurPtr++;
727 } else {
728 Kind = 'J';
729 }
730
731 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
732 // Bad token, return it as an error.
733 CurPtr = TokStart+1;
734 return lltok::Error;
735 }
736
737 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
738 ++CurPtr;
739
740 if (Kind == 'J') {
741 // HexFPConstant - Floating point constant represented in IEEE format as a
742 // hexadecimal number for when exponential notation is not precise enough.
743 // Half, Float, and double only.
744 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
745 return lltok::APFloat;
746 }
747
748 uint64_t Pair[2];
749 switch (Kind) {
750 default: llvm_unreachable("Unknown kind!");
751 case 'K':
752 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
753 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
754 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair));
755 return lltok::APFloat;
756 case 'L':
757 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
758 HexToIntPair(TokStart+3, CurPtr, Pair);
759 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair));
760 return lltok::APFloat;
761 case 'M':
762 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
763 HexToIntPair(TokStart+3, CurPtr, Pair);
764 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair));
765 return lltok::APFloat;
766 case 'H':
767 APFloatVal = APFloat(APFloat::IEEEhalf,
768 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
769 return lltok::APFloat;
770 }
771 }
772
773 /// LexIdentifier: Handle several related productions:
774 /// Label [-a-zA-Z$._0-9]+:
775 /// NInteger -[0-9]+
776 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
777 /// PInteger [0-9]+
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]+
LexDigitOrNegative()782 lltok::Kind LLLexer::LexDigitOrNegative() {
783 // If the letter after the negative is not a number, this is probably a label.
784 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
785 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
786 // Okay, this is not a number after the -, it's probably a label.
787 if (const char *End = isLabelTail(CurPtr)) {
788 StrVal.assign(TokStart, End-1);
789 CurPtr = End;
790 return lltok::LabelStr;
791 }
792
793 return lltok::Error;
794 }
795
796 // At this point, it is either a label, int or fp constant.
797
798 // Skip digits, we have at least one.
799 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
800 /*empty*/;
801
802 // Check to see if this really is a label afterall, e.g. "-1:".
803 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
804 if (const char *End = isLabelTail(CurPtr)) {
805 StrVal.assign(TokStart, End-1);
806 CurPtr = End;
807 return lltok::LabelStr;
808 }
809 }
810
811 // If the next character is a '.', then it is a fp value, otherwise its
812 // integer.
813 if (CurPtr[0] != '.') {
814 if (TokStart[0] == '0' && TokStart[1] == 'x')
815 return Lex0x();
816 unsigned Len = CurPtr-TokStart;
817 uint32_t numBits = ((Len * 64) / 19) + 2;
818 APInt Tmp(numBits, StringRef(TokStart, Len), 10);
819 if (TokStart[0] == '-') {
820 uint32_t minBits = Tmp.getMinSignedBits();
821 if (minBits > 0 && minBits < numBits)
822 Tmp = Tmp.trunc(minBits);
823 APSIntVal = APSInt(Tmp, false);
824 } else {
825 uint32_t activeBits = Tmp.getActiveBits();
826 if (activeBits > 0 && activeBits < numBits)
827 Tmp = Tmp.trunc(activeBits);
828 APSIntVal = APSInt(Tmp, true);
829 }
830 return lltok::APSInt;
831 }
832
833 ++CurPtr;
834
835 // Skip over [0-9]*([eE][-+]?[0-9]+)?
836 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
837
838 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
839 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
840 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
841 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
842 CurPtr += 2;
843 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
844 }
845 }
846
847 APFloatVal = APFloat(std::atof(TokStart));
848 return lltok::APFloat;
849 }
850
851 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
LexPositive()852 lltok::Kind LLLexer::LexPositive() {
853 // If the letter after the negative is a number, this is probably not a
854 // label.
855 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
856 return lltok::Error;
857
858 // Skip digits.
859 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
860 /*empty*/;
861
862 // At this point, we need a '.'.
863 if (CurPtr[0] != '.') {
864 CurPtr = TokStart+1;
865 return lltok::Error;
866 }
867
868 ++CurPtr;
869
870 // Skip over [0-9]*([eE][-+]?[0-9]+)?
871 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
872
873 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
874 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
875 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
876 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
877 CurPtr += 2;
878 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
879 }
880 }
881
882 APFloatVal = APFloat(std::atof(TokStart));
883 return lltok::APFloat;
884 }
885