[RISCV] Fix mgather -> riscv.masked.strided.load combine not extending indices (...
[llvm-project.git] / llvm / lib / AsmParser / LLLexer.cpp
blobc8da3efbb68aff71d7727dac5985255e5de7768d
1 //===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // Implement the Lexer for .ll files.
11 //===----------------------------------------------------------------------===//
13 #include "llvm/AsmParser/LLLexer.h"
14 #include "llvm/ADT/APInt.h"
15 #include "llvm/ADT/STLExtras.h"
16 #include "llvm/ADT/StringExtras.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/IR/DerivedTypes.h"
19 #include "llvm/IR/Instruction.h"
20 #include "llvm/Support/ErrorHandling.h"
21 #include "llvm/Support/SourceMgr.h"
22 #include <cassert>
23 #include <cctype>
24 #include <cstdio>
26 using namespace llvm;
28 bool LLLexer::Error(LocTy ErrorLoc, const Twine &Msg) const {
29 ErrorInfo = SM.GetMessage(ErrorLoc, SourceMgr::DK_Error, Msg);
30 return true;
33 void LLLexer::Warning(LocTy WarningLoc, const Twine &Msg) const {
34 SM.PrintMessage(WarningLoc, SourceMgr::DK_Warning, Msg);
37 //===----------------------------------------------------------------------===//
38 // Helper functions.
39 //===----------------------------------------------------------------------===//
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...
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;
56 return Result;
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);
66 if (Result < OldRes) { // Uh, oh, overflow detected!!!
67 Error("constant bigger than 64 bits detected!");
68 return 0;
71 return Result;
74 void LLLexer::HexToIntPair(const char *Buffer, const char *End,
75 uint64_t Pair[2]) {
76 Pair[0] = 0;
77 if (End - Buffer >= 16) {
78 for (int i = 0; i < 16; i++, Buffer++) {
79 assert(Buffer != End);
80 Pair[0] *= 16;
81 Pair[0] += hexDigitValue(*Buffer);
84 Pair[1] = 0;
85 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
86 Pair[1] *= 16;
87 Pair[1] += hexDigitValue(*Buffer);
89 if (Buffer != End)
90 Error("constant bigger than 128 bits detected!");
93 /// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
94 /// { low64, high16 } as usual for an APInt.
95 void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
96 uint64_t Pair[2]) {
97 Pair[1] = 0;
98 for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
99 assert(Buffer != End);
100 Pair[1] *= 16;
101 Pair[1] += hexDigitValue(*Buffer);
103 Pair[0] = 0;
104 for (int i = 0; i < 16 && Buffer != End; i++, Buffer++) {
105 Pair[0] *= 16;
106 Pair[0] += hexDigitValue(*Buffer);
108 if (Buffer != End)
109 Error("constant bigger than 128 bits detected!");
112 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
113 // appropriate character.
114 static void UnEscapeLexed(std::string &Str) {
115 if (Str.empty()) return;
117 char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
118 char *BOut = Buffer;
119 for (char *BIn = Buffer; BIn != EndBuffer; ) {
120 if (BIn[0] == '\\') {
121 if (BIn < EndBuffer-1 && BIn[1] == '\\') {
122 *BOut++ = '\\'; // Two \ becomes one
123 BIn += 2;
124 } else if (BIn < EndBuffer-2 &&
125 isxdigit(static_cast<unsigned char>(BIn[1])) &&
126 isxdigit(static_cast<unsigned char>(BIn[2]))) {
127 *BOut = hexDigitValue(BIn[1]) * 16 + hexDigitValue(BIn[2]);
128 BIn += 3; // Skip over handled chars
129 ++BOut;
130 } else {
131 *BOut++ = *BIn++;
133 } else {
134 *BOut++ = *BIn++;
137 Str.resize(BOut-Buffer);
140 /// isLabelChar - Return true for [-a-zA-Z$._0-9].
141 static bool isLabelChar(char C) {
142 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
143 C == '.' || C == '_';
146 /// isLabelTail - Return true if this pointer points to a valid end of a label.
147 static const char *isLabelTail(const char *CurPtr) {
148 while (true) {
149 if (CurPtr[0] == ':') return CurPtr+1;
150 if (!isLabelChar(CurPtr[0])) return nullptr;
151 ++CurPtr;
155 //===----------------------------------------------------------------------===//
156 // Lexer definition.
157 //===----------------------------------------------------------------------===//
159 LLLexer::LLLexer(StringRef StartBuf, SourceMgr &SM, SMDiagnostic &Err,
160 LLVMContext &C)
161 : CurBuf(StartBuf), ErrorInfo(Err), SM(SM), Context(C) {
162 CurPtr = CurBuf.begin();
165 int LLLexer::getNextChar() {
166 char CurChar = *CurPtr++;
167 switch (CurChar) {
168 default: return (unsigned char)CurChar;
169 case 0:
170 // A nul character in the stream is either the end of the current buffer or
171 // a random nul in the file. Disambiguate that here.
172 if (CurPtr-1 != CurBuf.end())
173 return 0; // Just whitespace.
175 // Otherwise, return end of file.
176 --CurPtr; // Another call to lex will return EOF again.
177 return EOF;
181 lltok::Kind LLLexer::LexToken() {
182 while (true) {
183 TokStart = CurPtr;
185 int CurChar = getNextChar();
186 switch (CurChar) {
187 default:
188 // Handle letters: [a-zA-Z_]
189 if (isalpha(static_cast<unsigned char>(CurChar)) || CurChar == '_')
190 return LexIdentifier();
192 return lltok::Error;
193 case EOF: return lltok::Eof;
194 case 0:
195 case ' ':
196 case '\t':
197 case '\n':
198 case '\r':
199 // Ignore whitespace.
200 continue;
201 case '+': return LexPositive();
202 case '@': return LexAt();
203 case '$': return LexDollar();
204 case '%': return LexPercent();
205 case '"': return LexQuote();
206 case '.':
207 if (const char *Ptr = isLabelTail(CurPtr)) {
208 CurPtr = Ptr;
209 StrVal.assign(TokStart, CurPtr-1);
210 return lltok::LabelStr;
212 if (CurPtr[0] == '.' && CurPtr[1] == '.') {
213 CurPtr += 2;
214 return lltok::dotdotdot;
216 return lltok::Error;
217 case ';':
218 SkipLineComment();
219 continue;
220 case '!': return LexExclaim();
221 case '^':
222 return LexCaret();
223 case ':':
224 return lltok::colon;
225 case '#': return LexHash();
226 case '0': case '1': case '2': case '3': case '4':
227 case '5': case '6': case '7': case '8': case '9':
228 case '-':
229 return LexDigitOrNegative();
230 case '=': return lltok::equal;
231 case '[': return lltok::lsquare;
232 case ']': return lltok::rsquare;
233 case '{': return lltok::lbrace;
234 case '}': return lltok::rbrace;
235 case '<': return lltok::less;
236 case '>': return lltok::greater;
237 case '(': return lltok::lparen;
238 case ')': return lltok::rparen;
239 case ',': return lltok::comma;
240 case '*': return lltok::star;
241 case '|': return lltok::bar;
246 void LLLexer::SkipLineComment() {
247 while (true) {
248 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
249 return;
253 /// Lex all tokens that start with an @ character.
254 /// GlobalVar @\"[^\"]*\"
255 /// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
256 /// GlobalVarID @[0-9]+
257 lltok::Kind LLLexer::LexAt() {
258 return LexVar(lltok::GlobalVar, lltok::GlobalID);
261 lltok::Kind LLLexer::LexDollar() {
262 if (const char *Ptr = isLabelTail(TokStart)) {
263 CurPtr = Ptr;
264 StrVal.assign(TokStart, CurPtr - 1);
265 return lltok::LabelStr;
268 // Handle DollarStringConstant: $\"[^\"]*\"
269 if (CurPtr[0] == '"') {
270 ++CurPtr;
272 while (true) {
273 int CurChar = getNextChar();
275 if (CurChar == EOF) {
276 Error("end of file in COMDAT variable name");
277 return lltok::Error;
279 if (CurChar == '"') {
280 StrVal.assign(TokStart + 2, CurPtr - 1);
281 UnEscapeLexed(StrVal);
282 if (StringRef(StrVal).contains(0)) {
283 Error("Null bytes are not allowed in names");
284 return lltok::Error;
286 return lltok::ComdatVar;
291 // Handle ComdatVarName: $[-a-zA-Z$._][-a-zA-Z$._0-9]*
292 if (ReadVarName())
293 return lltok::ComdatVar;
295 return lltok::Error;
298 /// ReadString - Read a string until the closing quote.
299 lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
300 const char *Start = CurPtr;
301 while (true) {
302 int CurChar = getNextChar();
304 if (CurChar == EOF) {
305 Error("end of file in string constant");
306 return lltok::Error;
308 if (CurChar == '"') {
309 StrVal.assign(Start, CurPtr-1);
310 UnEscapeLexed(StrVal);
311 return kind;
316 /// ReadVarName - Read the rest of a token containing a variable name.
317 bool LLLexer::ReadVarName() {
318 const char *NameStart = CurPtr;
319 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
320 CurPtr[0] == '-' || CurPtr[0] == '$' ||
321 CurPtr[0] == '.' || CurPtr[0] == '_') {
322 ++CurPtr;
323 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
324 CurPtr[0] == '-' || CurPtr[0] == '$' ||
325 CurPtr[0] == '.' || CurPtr[0] == '_')
326 ++CurPtr;
328 StrVal.assign(NameStart, CurPtr);
329 return true;
331 return false;
334 // Lex an ID: [0-9]+. On success, the ID is stored in UIntVal and Token is
335 // returned, otherwise the Error token is returned.
336 lltok::Kind LLLexer::LexUIntID(lltok::Kind Token) {
337 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
338 return lltok::Error;
340 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
341 /*empty*/;
343 uint64_t Val = atoull(TokStart + 1, CurPtr);
344 if ((unsigned)Val != Val)
345 Error("invalid value number (too large)!");
346 UIntVal = unsigned(Val);
347 return Token;
350 lltok::Kind LLLexer::LexVar(lltok::Kind Var, lltok::Kind VarID) {
351 // Handle StringConstant: \"[^\"]*\"
352 if (CurPtr[0] == '"') {
353 ++CurPtr;
355 while (true) {
356 int CurChar = getNextChar();
358 if (CurChar == EOF) {
359 Error("end of file in global variable name");
360 return lltok::Error;
362 if (CurChar == '"') {
363 StrVal.assign(TokStart+2, CurPtr-1);
364 UnEscapeLexed(StrVal);
365 if (StringRef(StrVal).contains(0)) {
366 Error("Null bytes are not allowed in names");
367 return lltok::Error;
369 return Var;
374 // Handle VarName: [-a-zA-Z$._][-a-zA-Z$._0-9]*
375 if (ReadVarName())
376 return Var;
378 // Handle VarID: [0-9]+
379 return LexUIntID(VarID);
382 /// Lex all tokens that start with a % character.
383 /// LocalVar ::= %\"[^\"]*\"
384 /// LocalVar ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
385 /// LocalVarID ::= %[0-9]+
386 lltok::Kind LLLexer::LexPercent() {
387 return LexVar(lltok::LocalVar, lltok::LocalVarID);
390 /// Lex all tokens that start with a " character.
391 /// QuoteLabel "[^"]+":
392 /// StringConstant "[^"]*"
393 lltok::Kind LLLexer::LexQuote() {
394 lltok::Kind kind = ReadString(lltok::StringConstant);
395 if (kind == lltok::Error || kind == lltok::Eof)
396 return kind;
398 if (CurPtr[0] == ':') {
399 ++CurPtr;
400 if (StringRef(StrVal).contains(0)) {
401 Error("Null bytes are not allowed in names");
402 kind = lltok::Error;
403 } else {
404 kind = lltok::LabelStr;
408 return kind;
411 /// Lex all tokens that start with a ! character.
412 /// !foo
413 /// !
414 lltok::Kind LLLexer::LexExclaim() {
415 // Lex a metadata name as a MetadataVar.
416 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
417 CurPtr[0] == '-' || CurPtr[0] == '$' ||
418 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
419 ++CurPtr;
420 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
421 CurPtr[0] == '-' || CurPtr[0] == '$' ||
422 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
423 ++CurPtr;
425 StrVal.assign(TokStart+1, CurPtr); // Skip !
426 UnEscapeLexed(StrVal);
427 return lltok::MetadataVar;
429 return lltok::exclaim;
432 /// Lex all tokens that start with a ^ character.
433 /// SummaryID ::= ^[0-9]+
434 lltok::Kind LLLexer::LexCaret() {
435 // Handle SummaryID: ^[0-9]+
436 return LexUIntID(lltok::SummaryID);
439 /// Lex all tokens that start with a # character.
440 /// AttrGrpID ::= #[0-9]+
441 lltok::Kind LLLexer::LexHash() {
442 // Handle AttrGrpID: #[0-9]+
443 return LexUIntID(lltok::AttrGrpID);
446 /// Lex a label, integer type, keyword, or hexadecimal integer constant.
447 /// Label [-a-zA-Z$._0-9]+:
448 /// IntegerType i[0-9]+
449 /// Keyword sdiv, float, ...
450 /// HexIntConstant [us]0x[0-9A-Fa-f]+
451 lltok::Kind LLLexer::LexIdentifier() {
452 const char *StartChar = CurPtr;
453 const char *IntEnd = CurPtr[-1] == 'i' ? nullptr : StartChar;
454 const char *KeywordEnd = nullptr;
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;
465 // If we stopped due to a colon, unless we were directed to ignore it,
466 // this really is a label.
467 if (!IgnoreColonInIdentifiers && *CurPtr == ':') {
468 StrVal.assign(StartChar-1, CurPtr++);
469 return lltok::LabelStr;
472 // Otherwise, this wasn't a label. If this was valid as an integer type,
473 // return it.
474 if (!IntEnd) IntEnd = CurPtr;
475 if (IntEnd != StartChar) {
476 CurPtr = IntEnd;
477 uint64_t NumBits = atoull(StartChar, CurPtr);
478 if (NumBits < IntegerType::MIN_INT_BITS ||
479 NumBits > IntegerType::MAX_INT_BITS) {
480 Error("bitwidth for integer type out of range!");
481 return lltok::Error;
483 TyVal = IntegerType::get(Context, NumBits);
484 return lltok::Type;
487 // Otherwise, this was a letter sequence. See which keyword this is.
488 if (!KeywordEnd) KeywordEnd = CurPtr;
489 CurPtr = KeywordEnd;
490 --StartChar;
491 StringRef Keyword(StartChar, CurPtr - StartChar);
493 #define KEYWORD(STR) \
494 do { \
495 if (Keyword == #STR) \
496 return lltok::kw_##STR; \
497 } while (false)
499 KEYWORD(true); KEYWORD(false);
500 KEYWORD(declare); KEYWORD(define);
501 KEYWORD(global); KEYWORD(constant);
503 KEYWORD(dso_local);
504 KEYWORD(dso_preemptable);
506 KEYWORD(private);
507 KEYWORD(internal);
508 KEYWORD(available_externally);
509 KEYWORD(linkonce);
510 KEYWORD(linkonce_odr);
511 KEYWORD(weak); // Use as a linkage, and a modifier for "cmpxchg".
512 KEYWORD(weak_odr);
513 KEYWORD(appending);
514 KEYWORD(dllimport);
515 KEYWORD(dllexport);
516 KEYWORD(common);
517 KEYWORD(default);
518 KEYWORD(hidden);
519 KEYWORD(protected);
520 KEYWORD(unnamed_addr);
521 KEYWORD(local_unnamed_addr);
522 KEYWORD(externally_initialized);
523 KEYWORD(extern_weak);
524 KEYWORD(external);
525 KEYWORD(thread_local);
526 KEYWORD(localdynamic);
527 KEYWORD(initialexec);
528 KEYWORD(localexec);
529 KEYWORD(zeroinitializer);
530 KEYWORD(undef);
531 KEYWORD(null);
532 KEYWORD(none);
533 KEYWORD(poison);
534 KEYWORD(to);
535 KEYWORD(caller);
536 KEYWORD(within);
537 KEYWORD(from);
538 KEYWORD(tail);
539 KEYWORD(musttail);
540 KEYWORD(notail);
541 KEYWORD(target);
542 KEYWORD(triple);
543 KEYWORD(source_filename);
544 KEYWORD(unwind);
545 KEYWORD(datalayout);
546 KEYWORD(volatile);
547 KEYWORD(atomic);
548 KEYWORD(unordered);
549 KEYWORD(monotonic);
550 KEYWORD(acquire);
551 KEYWORD(release);
552 KEYWORD(acq_rel);
553 KEYWORD(seq_cst);
554 KEYWORD(syncscope);
556 KEYWORD(nnan);
557 KEYWORD(ninf);
558 KEYWORD(nsz);
559 KEYWORD(arcp);
560 KEYWORD(contract);
561 KEYWORD(reassoc);
562 KEYWORD(afn);
563 KEYWORD(fast);
564 KEYWORD(nuw);
565 KEYWORD(nsw);
566 KEYWORD(exact);
567 KEYWORD(disjoint);
568 KEYWORD(inbounds);
569 KEYWORD(nneg);
570 KEYWORD(inrange);
571 KEYWORD(addrspace);
572 KEYWORD(section);
573 KEYWORD(partition);
574 KEYWORD(code_model);
575 KEYWORD(alias);
576 KEYWORD(ifunc);
577 KEYWORD(module);
578 KEYWORD(asm);
579 KEYWORD(sideeffect);
580 KEYWORD(inteldialect);
581 KEYWORD(gc);
582 KEYWORD(prefix);
583 KEYWORD(prologue);
585 KEYWORD(no_sanitize_address);
586 KEYWORD(no_sanitize_hwaddress);
587 KEYWORD(sanitize_address_dyninit);
589 KEYWORD(ccc);
590 KEYWORD(fastcc);
591 KEYWORD(coldcc);
592 KEYWORD(cfguard_checkcc);
593 KEYWORD(x86_stdcallcc);
594 KEYWORD(x86_fastcallcc);
595 KEYWORD(x86_thiscallcc);
596 KEYWORD(x86_vectorcallcc);
597 KEYWORD(arm_apcscc);
598 KEYWORD(arm_aapcscc);
599 KEYWORD(arm_aapcs_vfpcc);
600 KEYWORD(aarch64_vector_pcs);
601 KEYWORD(aarch64_sve_vector_pcs);
602 KEYWORD(aarch64_sme_preservemost_from_x0);
603 KEYWORD(aarch64_sme_preservemost_from_x2);
604 KEYWORD(msp430_intrcc);
605 KEYWORD(avr_intrcc);
606 KEYWORD(avr_signalcc);
607 KEYWORD(ptx_kernel);
608 KEYWORD(ptx_device);
609 KEYWORD(spir_kernel);
610 KEYWORD(spir_func);
611 KEYWORD(intel_ocl_bicc);
612 KEYWORD(x86_64_sysvcc);
613 KEYWORD(win64cc);
614 KEYWORD(x86_regcallcc);
615 KEYWORD(swiftcc);
616 KEYWORD(swifttailcc);
617 KEYWORD(anyregcc);
618 KEYWORD(preserve_mostcc);
619 KEYWORD(preserve_allcc);
620 KEYWORD(ghccc);
621 KEYWORD(x86_intrcc);
622 KEYWORD(hhvmcc);
623 KEYWORD(hhvm_ccc);
624 KEYWORD(cxx_fast_tlscc);
625 KEYWORD(amdgpu_vs);
626 KEYWORD(amdgpu_ls);
627 KEYWORD(amdgpu_hs);
628 KEYWORD(amdgpu_es);
629 KEYWORD(amdgpu_gs);
630 KEYWORD(amdgpu_ps);
631 KEYWORD(amdgpu_cs);
632 KEYWORD(amdgpu_cs_chain);
633 KEYWORD(amdgpu_cs_chain_preserve);
634 KEYWORD(amdgpu_kernel);
635 KEYWORD(amdgpu_gfx);
636 KEYWORD(tailcc);
637 KEYWORD(m68k_rtdcc);
638 KEYWORD(graalcc);
640 KEYWORD(cc);
641 KEYWORD(c);
643 KEYWORD(attributes);
644 KEYWORD(sync);
645 KEYWORD(async);
647 #define GET_ATTR_NAMES
648 #define ATTRIBUTE_ENUM(ENUM_NAME, DISPLAY_NAME) \
649 KEYWORD(DISPLAY_NAME);
650 #include "llvm/IR/Attributes.inc"
652 KEYWORD(read);
653 KEYWORD(write);
654 KEYWORD(readwrite);
655 KEYWORD(argmem);
656 KEYWORD(inaccessiblemem);
657 KEYWORD(argmemonly);
658 KEYWORD(inaccessiblememonly);
659 KEYWORD(inaccessiblemem_or_argmemonly);
661 // nofpclass attribute
662 KEYWORD(all);
663 KEYWORD(nan);
664 KEYWORD(snan);
665 KEYWORD(qnan);
666 KEYWORD(inf);
667 // ninf already a keyword
668 KEYWORD(pinf);
669 KEYWORD(norm);
670 KEYWORD(nnorm);
671 KEYWORD(pnorm);
672 // sub already a keyword
673 KEYWORD(nsub);
674 KEYWORD(psub);
675 KEYWORD(zero);
676 KEYWORD(nzero);
677 KEYWORD(pzero);
679 KEYWORD(type);
680 KEYWORD(opaque);
682 KEYWORD(comdat);
684 // Comdat types
685 KEYWORD(any);
686 KEYWORD(exactmatch);
687 KEYWORD(largest);
688 KEYWORD(nodeduplicate);
689 KEYWORD(samesize);
691 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
692 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
693 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
694 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
696 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
697 KEYWORD(umin); KEYWORD(fmax); KEYWORD(fmin);
698 KEYWORD(uinc_wrap);
699 KEYWORD(udec_wrap);
701 KEYWORD(splat);
702 KEYWORD(vscale);
703 KEYWORD(x);
704 KEYWORD(blockaddress);
705 KEYWORD(dso_local_equivalent);
706 KEYWORD(no_cfi);
708 // Metadata types.
709 KEYWORD(distinct);
711 // Use-list order directives.
712 KEYWORD(uselistorder);
713 KEYWORD(uselistorder_bb);
715 KEYWORD(personality);
716 KEYWORD(cleanup);
717 KEYWORD(catch);
718 KEYWORD(filter);
720 // Summary index keywords.
721 KEYWORD(path);
722 KEYWORD(hash);
723 KEYWORD(gv);
724 KEYWORD(guid);
725 KEYWORD(name);
726 KEYWORD(summaries);
727 KEYWORD(flags);
728 KEYWORD(blockcount);
729 KEYWORD(linkage);
730 KEYWORD(visibility);
731 KEYWORD(notEligibleToImport);
732 KEYWORD(live);
733 KEYWORD(dsoLocal);
734 KEYWORD(canAutoHide);
735 KEYWORD(function);
736 KEYWORD(insts);
737 KEYWORD(funcFlags);
738 KEYWORD(readNone);
739 KEYWORD(readOnly);
740 KEYWORD(noRecurse);
741 KEYWORD(returnDoesNotAlias);
742 KEYWORD(noInline);
743 KEYWORD(alwaysInline);
744 KEYWORD(noUnwind);
745 KEYWORD(mayThrow);
746 KEYWORD(hasUnknownCall);
747 KEYWORD(mustBeUnreachable);
748 KEYWORD(calls);
749 KEYWORD(callee);
750 KEYWORD(params);
751 KEYWORD(param);
752 KEYWORD(hotness);
753 KEYWORD(unknown);
754 KEYWORD(critical);
755 KEYWORD(relbf);
756 KEYWORD(variable);
757 KEYWORD(vTableFuncs);
758 KEYWORD(virtFunc);
759 KEYWORD(aliasee);
760 KEYWORD(refs);
761 KEYWORD(typeIdInfo);
762 KEYWORD(typeTests);
763 KEYWORD(typeTestAssumeVCalls);
764 KEYWORD(typeCheckedLoadVCalls);
765 KEYWORD(typeTestAssumeConstVCalls);
766 KEYWORD(typeCheckedLoadConstVCalls);
767 KEYWORD(vFuncId);
768 KEYWORD(offset);
769 KEYWORD(args);
770 KEYWORD(typeid);
771 KEYWORD(typeidCompatibleVTable);
772 KEYWORD(summary);
773 KEYWORD(typeTestRes);
774 KEYWORD(kind);
775 KEYWORD(unsat);
776 KEYWORD(byteArray);
777 KEYWORD(inline);
778 KEYWORD(single);
779 KEYWORD(allOnes);
780 KEYWORD(sizeM1BitWidth);
781 KEYWORD(alignLog2);
782 KEYWORD(sizeM1);
783 KEYWORD(bitMask);
784 KEYWORD(inlineBits);
785 KEYWORD(vcall_visibility);
786 KEYWORD(wpdResolutions);
787 KEYWORD(wpdRes);
788 KEYWORD(indir);
789 KEYWORD(singleImpl);
790 KEYWORD(branchFunnel);
791 KEYWORD(singleImplName);
792 KEYWORD(resByArg);
793 KEYWORD(byArg);
794 KEYWORD(uniformRetVal);
795 KEYWORD(uniqueRetVal);
796 KEYWORD(virtualConstProp);
797 KEYWORD(info);
798 KEYWORD(byte);
799 KEYWORD(bit);
800 KEYWORD(varFlags);
801 KEYWORD(callsites);
802 KEYWORD(clones);
803 KEYWORD(stackIds);
804 KEYWORD(allocs);
805 KEYWORD(versions);
806 KEYWORD(memProf);
807 KEYWORD(notcold);
809 #undef KEYWORD
811 // Keywords for types.
812 #define TYPEKEYWORD(STR, LLVMTY) \
813 do { \
814 if (Keyword == STR) { \
815 TyVal = LLVMTY; \
816 return lltok::Type; \
818 } while (false)
820 TYPEKEYWORD("void", Type::getVoidTy(Context));
821 TYPEKEYWORD("half", Type::getHalfTy(Context));
822 TYPEKEYWORD("bfloat", Type::getBFloatTy(Context));
823 TYPEKEYWORD("float", Type::getFloatTy(Context));
824 TYPEKEYWORD("double", Type::getDoubleTy(Context));
825 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
826 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
827 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
828 TYPEKEYWORD("label", Type::getLabelTy(Context));
829 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
830 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
831 TYPEKEYWORD("x86_amx", Type::getX86_AMXTy(Context));
832 TYPEKEYWORD("token", Type::getTokenTy(Context));
833 TYPEKEYWORD("ptr", PointerType::getUnqual(Context));
835 #undef TYPEKEYWORD
837 // Keywords for instructions.
838 #define INSTKEYWORD(STR, Enum) \
839 do { \
840 if (Keyword == #STR) { \
841 UIntVal = Instruction::Enum; \
842 return lltok::kw_##STR; \
844 } while (false)
846 INSTKEYWORD(fneg, FNeg);
848 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
849 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
850 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
851 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
852 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
853 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
854 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
855 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
857 INSTKEYWORD(phi, PHI);
858 INSTKEYWORD(call, Call);
859 INSTKEYWORD(trunc, Trunc);
860 INSTKEYWORD(zext, ZExt);
861 INSTKEYWORD(sext, SExt);
862 INSTKEYWORD(fptrunc, FPTrunc);
863 INSTKEYWORD(fpext, FPExt);
864 INSTKEYWORD(uitofp, UIToFP);
865 INSTKEYWORD(sitofp, SIToFP);
866 INSTKEYWORD(fptoui, FPToUI);
867 INSTKEYWORD(fptosi, FPToSI);
868 INSTKEYWORD(inttoptr, IntToPtr);
869 INSTKEYWORD(ptrtoint, PtrToInt);
870 INSTKEYWORD(bitcast, BitCast);
871 INSTKEYWORD(addrspacecast, AddrSpaceCast);
872 INSTKEYWORD(select, Select);
873 INSTKEYWORD(va_arg, VAArg);
874 INSTKEYWORD(ret, Ret);
875 INSTKEYWORD(br, Br);
876 INSTKEYWORD(switch, Switch);
877 INSTKEYWORD(indirectbr, IndirectBr);
878 INSTKEYWORD(invoke, Invoke);
879 INSTKEYWORD(resume, Resume);
880 INSTKEYWORD(unreachable, Unreachable);
881 INSTKEYWORD(callbr, CallBr);
883 INSTKEYWORD(alloca, Alloca);
884 INSTKEYWORD(load, Load);
885 INSTKEYWORD(store, Store);
886 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
887 INSTKEYWORD(atomicrmw, AtomicRMW);
888 INSTKEYWORD(fence, Fence);
889 INSTKEYWORD(getelementptr, GetElementPtr);
891 INSTKEYWORD(extractelement, ExtractElement);
892 INSTKEYWORD(insertelement, InsertElement);
893 INSTKEYWORD(shufflevector, ShuffleVector);
894 INSTKEYWORD(extractvalue, ExtractValue);
895 INSTKEYWORD(insertvalue, InsertValue);
896 INSTKEYWORD(landingpad, LandingPad);
897 INSTKEYWORD(cleanupret, CleanupRet);
898 INSTKEYWORD(catchret, CatchRet);
899 INSTKEYWORD(catchswitch, CatchSwitch);
900 INSTKEYWORD(catchpad, CatchPad);
901 INSTKEYWORD(cleanuppad, CleanupPad);
903 INSTKEYWORD(freeze, Freeze);
905 #undef INSTKEYWORD
907 #define DWKEYWORD(TYPE, TOKEN) \
908 do { \
909 if (Keyword.starts_with("DW_" #TYPE "_")) { \
910 StrVal.assign(Keyword.begin(), Keyword.end()); \
911 return lltok::TOKEN; \
913 } while (false)
915 DWKEYWORD(TAG, DwarfTag);
916 DWKEYWORD(ATE, DwarfAttEncoding);
917 DWKEYWORD(VIRTUALITY, DwarfVirtuality);
918 DWKEYWORD(LANG, DwarfLang);
919 DWKEYWORD(CC, DwarfCC);
920 DWKEYWORD(OP, DwarfOp);
921 DWKEYWORD(MACINFO, DwarfMacinfo);
923 #undef DWKEYWORD
925 if (Keyword.starts_with("DIFlag")) {
926 StrVal.assign(Keyword.begin(), Keyword.end());
927 return lltok::DIFlag;
930 if (Keyword.starts_with("DISPFlag")) {
931 StrVal.assign(Keyword.begin(), Keyword.end());
932 return lltok::DISPFlag;
935 if (Keyword.starts_with("CSK_")) {
936 StrVal.assign(Keyword.begin(), Keyword.end());
937 return lltok::ChecksumKind;
940 if (Keyword == "NoDebug" || Keyword == "FullDebug" ||
941 Keyword == "LineTablesOnly" || Keyword == "DebugDirectivesOnly") {
942 StrVal.assign(Keyword.begin(), Keyword.end());
943 return lltok::EmissionKind;
946 if (Keyword == "GNU" || Keyword == "Apple" || Keyword == "None" ||
947 Keyword == "Default") {
948 StrVal.assign(Keyword.begin(), Keyword.end());
949 return lltok::NameTableKind;
952 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
953 // the CFE to avoid forcing it to deal with 64-bit numbers.
954 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
955 TokStart[1] == '0' && TokStart[2] == 'x' &&
956 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
957 int len = CurPtr-TokStart-3;
958 uint32_t bits = len * 4;
959 StringRef HexStr(TokStart + 3, len);
960 if (!all_of(HexStr, isxdigit)) {
961 // Bad token, return it as an error.
962 CurPtr = TokStart+3;
963 return lltok::Error;
965 APInt Tmp(bits, HexStr, 16);
966 uint32_t activeBits = Tmp.getActiveBits();
967 if (activeBits > 0 && activeBits < bits)
968 Tmp = Tmp.trunc(activeBits);
969 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
970 return lltok::APSInt;
973 // If this is "cc1234", return this as just "cc".
974 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
975 CurPtr = TokStart+2;
976 return lltok::kw_cc;
979 // Finally, if this isn't known, return an error.
980 CurPtr = TokStart+1;
981 return lltok::Error;
984 /// Lex all tokens that start with a 0x prefix, knowing they match and are not
985 /// labels.
986 /// HexFPConstant 0x[0-9A-Fa-f]+
987 /// HexFP80Constant 0xK[0-9A-Fa-f]+
988 /// HexFP128Constant 0xL[0-9A-Fa-f]+
989 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
990 /// HexHalfConstant 0xH[0-9A-Fa-f]+
991 /// HexBFloatConstant 0xR[0-9A-Fa-f]+
992 lltok::Kind LLLexer::Lex0x() {
993 CurPtr = TokStart + 2;
995 char Kind;
996 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H' ||
997 CurPtr[0] == 'R') {
998 Kind = *CurPtr++;
999 } else {
1000 Kind = 'J';
1003 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
1004 // Bad token, return it as an error.
1005 CurPtr = TokStart+1;
1006 return lltok::Error;
1009 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
1010 ++CurPtr;
1012 if (Kind == 'J') {
1013 // HexFPConstant - Floating point constant represented in IEEE format as a
1014 // hexadecimal number for when exponential notation is not precise enough.
1015 // Half, BFloat, Float, and double only.
1016 APFloatVal = APFloat(APFloat::IEEEdouble(),
1017 APInt(64, HexIntToVal(TokStart + 2, CurPtr)));
1018 return lltok::APFloat;
1021 uint64_t Pair[2];
1022 switch (Kind) {
1023 default: llvm_unreachable("Unknown kind!");
1024 case 'K':
1025 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
1026 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
1027 APFloatVal = APFloat(APFloat::x87DoubleExtended(), APInt(80, Pair));
1028 return lltok::APFloat;
1029 case 'L':
1030 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
1031 HexToIntPair(TokStart+3, CurPtr, Pair);
1032 APFloatVal = APFloat(APFloat::IEEEquad(), APInt(128, Pair));
1033 return lltok::APFloat;
1034 case 'M':
1035 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
1036 HexToIntPair(TokStart+3, CurPtr, Pair);
1037 APFloatVal = APFloat(APFloat::PPCDoubleDouble(), APInt(128, Pair));
1038 return lltok::APFloat;
1039 case 'H':
1040 APFloatVal = APFloat(APFloat::IEEEhalf(),
1041 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
1042 return lltok::APFloat;
1043 case 'R':
1044 // Brain floating point
1045 APFloatVal = APFloat(APFloat::BFloat(),
1046 APInt(16, HexIntToVal(TokStart + 3, CurPtr)));
1047 return lltok::APFloat;
1051 /// Lex tokens for a label or a numeric constant, possibly starting with -.
1052 /// Label [-a-zA-Z$._0-9]+:
1053 /// NInteger -[0-9]+
1054 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1055 /// PInteger [0-9]+
1056 /// HexFPConstant 0x[0-9A-Fa-f]+
1057 /// HexFP80Constant 0xK[0-9A-Fa-f]+
1058 /// HexFP128Constant 0xL[0-9A-Fa-f]+
1059 /// HexPPC128Constant 0xM[0-9A-Fa-f]+
1060 lltok::Kind LLLexer::LexDigitOrNegative() {
1061 // If the letter after the negative is not a number, this is probably a label.
1062 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
1063 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
1064 // Okay, this is not a number after the -, it's probably a label.
1065 if (const char *End = isLabelTail(CurPtr)) {
1066 StrVal.assign(TokStart, End-1);
1067 CurPtr = End;
1068 return lltok::LabelStr;
1071 return lltok::Error;
1074 // At this point, it is either a label, int or fp constant.
1076 // Skip digits, we have at least one.
1077 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1078 /*empty*/;
1080 // Check if this is a fully-numeric label:
1081 if (isdigit(TokStart[0]) && CurPtr[0] == ':') {
1082 uint64_t Val = atoull(TokStart, CurPtr);
1083 ++CurPtr; // Skip the colon.
1084 if ((unsigned)Val != Val)
1085 Error("invalid value number (too large)!");
1086 UIntVal = unsigned(Val);
1087 return lltok::LabelID;
1090 // Check to see if this really is a string label, e.g. "-1:".
1091 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
1092 if (const char *End = isLabelTail(CurPtr)) {
1093 StrVal.assign(TokStart, End-1);
1094 CurPtr = End;
1095 return lltok::LabelStr;
1099 // If the next character is a '.', then it is a fp value, otherwise its
1100 // integer.
1101 if (CurPtr[0] != '.') {
1102 if (TokStart[0] == '0' && TokStart[1] == 'x')
1103 return Lex0x();
1104 APSIntVal = APSInt(StringRef(TokStart, CurPtr - TokStart));
1105 return lltok::APSInt;
1108 ++CurPtr;
1110 // Skip over [0-9]*([eE][-+]?[0-9]+)?
1111 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1113 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1114 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1115 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1116 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1117 CurPtr += 2;
1118 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1122 APFloatVal = APFloat(APFloat::IEEEdouble(),
1123 StringRef(TokStart, CurPtr - TokStart));
1124 return lltok::APFloat;
1127 /// Lex a floating point constant starting with +.
1128 /// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
1129 lltok::Kind LLLexer::LexPositive() {
1130 // If the letter after the negative is a number, this is probably not a
1131 // label.
1132 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
1133 return lltok::Error;
1135 // Skip digits.
1136 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
1137 /*empty*/;
1139 // At this point, we need a '.'.
1140 if (CurPtr[0] != '.') {
1141 CurPtr = TokStart+1;
1142 return lltok::Error;
1145 ++CurPtr;
1147 // Skip over [0-9]*([eE][-+]?[0-9]+)?
1148 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1150 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
1151 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
1152 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
1153 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
1154 CurPtr += 2;
1155 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
1159 APFloatVal = APFloat(APFloat::IEEEdouble(),
1160 StringRef(TokStart, CurPtr - TokStart));
1161 return lltok::APFloat;