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>
30using namespace llvm;
31
32bool 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//
45uint64_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
59uint64_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
74void 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.
93void 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.
112static 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$._0-9].
139static bool isLabelChar(char C) {
140 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
141 C == '.' || C == '_';
142}
143
144
145/// isLabelTail - Return true if this pointer points to a valid end of a label.
146static 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
160LLLexer::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
166std::string LLLexer::getFilename() const {
167 return CurBuf->getBufferIdentifier();
168}
169
170int 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
187lltok::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 '$':
222 if (const char *Ptr = isLabelTail(CurPtr)) {
223 CurPtr = Ptr;
224 StrVal.assign(TokStart, CurPtr-1);
225 return lltok::LabelStr;
226 }
227 return lltok::Error;
228 case ';':
229 SkipLineComment();
230 return LexToken();
231 case '!': return LexExclaim();
232 case '#': return LexHash();
233 case '0': case '1': case '2': case '3': case '4':
234 case '5': case '6': case '7': case '8': case '9':
235 case '-':
236 return LexDigitOrNegative();
237 case '=': return lltok::equal;
238 case '[': return lltok::lsquare;
239 case ']': return lltok::rsquare;
240 case '{': return lltok::lbrace;
241 case '}': return lltok::rbrace;
242 case '<': return lltok::less;
243 case '>': return lltok::greater;
244 case '(': return lltok::lparen;
245 case ')': return lltok::rparen;
246 case ',': return lltok::comma;
247 case '*': return lltok::star;
248 case '\\': return lltok::backslash;
249 }
250}
251
252void LLLexer::SkipLineComment() {
253 while (1) {
254 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
255 return;
256 }
257}
258
259/// LexAt - Lex all tokens that start with an @ character:
260/// GlobalVar @\"[^\"]*\"
261/// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
262/// GlobalVarID @[0-9]+
263lltok::Kind LLLexer::LexAt() {
264 // Handle AtStringConstant: @\"[^\"]*\"
265 if (CurPtr[0] == '"') {
266 ++CurPtr;
267
268 while (1) {
269 int CurChar = getNextChar();
270
271 if (CurChar == EOF) {
272 Error("end of file in global variable name");
273 return lltok::Error;
274 }
275 if (CurChar == '"') {
276 StrVal.assign(TokStart+2, CurPtr-1);
277 UnEscapeLexed(StrVal);
278 return lltok::GlobalVar;
279 }
280 }
281 }
282
283 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
284 if (ReadVarName())
285 return lltok::GlobalVar;
286
287 // Handle GlobalVarID: @[0-9]+
288 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
289 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
290 /*empty*/;
291
292 uint64_t Val = atoull(TokStart+1, CurPtr);
293 if ((unsigned)Val != Val)
294 Error("invalid value number (too large)!");
295 UIntVal = unsigned(Val);
296 return lltok::GlobalID;
297 }
298
299 return lltok::Error;
300}
301
302/// ReadString - Read a string until the closing quote.
303lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
304 const char *Start = CurPtr;
305 while (1) {
306 int CurChar = getNextChar();
307
308 if (CurChar == EOF) {
309 Error("end of file in string constant");
310 return lltok::Error;
311 }
312 if (CurChar == '"') {
313 StrVal.assign(Start, CurPtr-1);
314 UnEscapeLexed(StrVal);
315 return kind;
316 }
317 }
318}
319
320/// ReadVarName - Read the rest of a token containing a variable name.
321bool LLLexer::ReadVarName() {
322 const char *NameStart = CurPtr;
323 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
324 CurPtr[0] == '-' || CurPtr[0] == '$' ||
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$._0-9]*
341/// LocalVarID ::= %[0-9]+
342lltok::Kind LLLexer::LexPercent() {
343 // Handle LocalVarName: %\"[^\"]*\"
344 if (CurPtr[0] == '"') {
345 ++CurPtr;
346 return ReadString(lltok::LocalVar);
347 }
348
349 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
350 if (ReadVarName())
351 return lltok::LocalVar;
352
353 // Handle LocalVarID: %[0-9]+
354 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
355 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
356 /*empty*/;
357
358 uint64_t Val = atoull(TokStart+1, CurPtr);
359 if ((unsigned)Val != Val)
360 Error("invalid value number (too large)!");
361 UIntVal = unsigned(Val);
362 return lltok::LocalVarID;
363 }
364
365 return lltok::Error;
366}
367
368/// LexQuote - Lex all tokens that start with a " character:
369/// QuoteLabel "[^"]+":
370/// StringConstant "[^"]*"
371lltok::Kind LLLexer::LexQuote() {
372 lltok::Kind kind = ReadString(lltok::StringConstant);
373 if (kind == lltok::Error || kind == lltok::Eof)
374 return kind;
375
376 if (CurPtr[0] == ':') {
377 ++CurPtr;
378 kind = lltok::LabelStr;
379 }
380
381 return kind;
382}
383
384/// LexExclaim:
385/// !foo
386/// !
387lltok::Kind LLLexer::LexExclaim() {
388 // Lex a metadata name as a MetadataVar.
389 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
390 CurPtr[0] == '-' || CurPtr[0] == '$' ||
391 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
392 ++CurPtr;
393 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
394 CurPtr[0] == '-' || CurPtr[0] == '$' ||
395 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
396 ++CurPtr;
397
398 StrVal.assign(TokStart+1, CurPtr); // Skip !
399 UnEscapeLexed(StrVal);
400 return lltok::MetadataVar;
401 }
402 return lltok::exclaim;
403}
404
405/// LexHash - Lex all tokens that start with a # character:
406/// AttrGrpID ::= #[0-9]+
407lltok::Kind LLLexer::LexHash() {
408 // Handle AttrGrpID: #[0-9]+
409 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
410 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
411 /*empty*/;
412
413 uint64_t Val = atoull(TokStart+1, CurPtr);
414 if ((unsigned)Val != Val)
415 Error("invalid value number (too large)!");
416 UIntVal = unsigned(Val);
417 return lltok::AttrGrpID;
418 }
419
420 return lltok::Error;
421}
422
423/// LexIdentifier: Handle several related productions:
424/// Label [-a-zA-Z$._0-9]+:
425/// IntegerType i[0-9]+
426/// Keyword sdiv, float, ...
427/// HexIntConstant [us]0x[0-9A-Fa-f]+
428lltok::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);
| 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>
30using namespace llvm;
31
32bool 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//
45uint64_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
59uint64_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
74void 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.
93void 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.
112static 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$._0-9].
139static bool isLabelChar(char C) {
140 return isalnum(static_cast<unsigned char>(C)) || C == '-' || C == '$' ||
141 C == '.' || C == '_';
142}
143
144
145/// isLabelTail - Return true if this pointer points to a valid end of a label.
146static 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
160LLLexer::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
166std::string LLLexer::getFilename() const {
167 return CurBuf->getBufferIdentifier();
168}
169
170int 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
187lltok::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 '$':
222 if (const char *Ptr = isLabelTail(CurPtr)) {
223 CurPtr = Ptr;
224 StrVal.assign(TokStart, CurPtr-1);
225 return lltok::LabelStr;
226 }
227 return lltok::Error;
228 case ';':
229 SkipLineComment();
230 return LexToken();
231 case '!': return LexExclaim();
232 case '#': return LexHash();
233 case '0': case '1': case '2': case '3': case '4':
234 case '5': case '6': case '7': case '8': case '9':
235 case '-':
236 return LexDigitOrNegative();
237 case '=': return lltok::equal;
238 case '[': return lltok::lsquare;
239 case ']': return lltok::rsquare;
240 case '{': return lltok::lbrace;
241 case '}': return lltok::rbrace;
242 case '<': return lltok::less;
243 case '>': return lltok::greater;
244 case '(': return lltok::lparen;
245 case ')': return lltok::rparen;
246 case ',': return lltok::comma;
247 case '*': return lltok::star;
248 case '\\': return lltok::backslash;
249 }
250}
251
252void LLLexer::SkipLineComment() {
253 while (1) {
254 if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
255 return;
256 }
257}
258
259/// LexAt - Lex all tokens that start with an @ character:
260/// GlobalVar @\"[^\"]*\"
261/// GlobalVar @[-a-zA-Z$._][-a-zA-Z$._0-9]*
262/// GlobalVarID @[0-9]+
263lltok::Kind LLLexer::LexAt() {
264 // Handle AtStringConstant: @\"[^\"]*\"
265 if (CurPtr[0] == '"') {
266 ++CurPtr;
267
268 while (1) {
269 int CurChar = getNextChar();
270
271 if (CurChar == EOF) {
272 Error("end of file in global variable name");
273 return lltok::Error;
274 }
275 if (CurChar == '"') {
276 StrVal.assign(TokStart+2, CurPtr-1);
277 UnEscapeLexed(StrVal);
278 return lltok::GlobalVar;
279 }
280 }
281 }
282
283 // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
284 if (ReadVarName())
285 return lltok::GlobalVar;
286
287 // Handle GlobalVarID: @[0-9]+
288 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
289 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
290 /*empty*/;
291
292 uint64_t Val = atoull(TokStart+1, CurPtr);
293 if ((unsigned)Val != Val)
294 Error("invalid value number (too large)!");
295 UIntVal = unsigned(Val);
296 return lltok::GlobalID;
297 }
298
299 return lltok::Error;
300}
301
302/// ReadString - Read a string until the closing quote.
303lltok::Kind LLLexer::ReadString(lltok::Kind kind) {
304 const char *Start = CurPtr;
305 while (1) {
306 int CurChar = getNextChar();
307
308 if (CurChar == EOF) {
309 Error("end of file in string constant");
310 return lltok::Error;
311 }
312 if (CurChar == '"') {
313 StrVal.assign(Start, CurPtr-1);
314 UnEscapeLexed(StrVal);
315 return kind;
316 }
317 }
318}
319
320/// ReadVarName - Read the rest of a token containing a variable name.
321bool LLLexer::ReadVarName() {
322 const char *NameStart = CurPtr;
323 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
324 CurPtr[0] == '-' || CurPtr[0] == '$' ||
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$._0-9]*
341/// LocalVarID ::= %[0-9]+
342lltok::Kind LLLexer::LexPercent() {
343 // Handle LocalVarName: %\"[^\"]*\"
344 if (CurPtr[0] == '"') {
345 ++CurPtr;
346 return ReadString(lltok::LocalVar);
347 }
348
349 // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
350 if (ReadVarName())
351 return lltok::LocalVar;
352
353 // Handle LocalVarID: %[0-9]+
354 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
355 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
356 /*empty*/;
357
358 uint64_t Val = atoull(TokStart+1, CurPtr);
359 if ((unsigned)Val != Val)
360 Error("invalid value number (too large)!");
361 UIntVal = unsigned(Val);
362 return lltok::LocalVarID;
363 }
364
365 return lltok::Error;
366}
367
368/// LexQuote - Lex all tokens that start with a " character:
369/// QuoteLabel "[^"]+":
370/// StringConstant "[^"]*"
371lltok::Kind LLLexer::LexQuote() {
372 lltok::Kind kind = ReadString(lltok::StringConstant);
373 if (kind == lltok::Error || kind == lltok::Eof)
374 return kind;
375
376 if (CurPtr[0] == ':') {
377 ++CurPtr;
378 kind = lltok::LabelStr;
379 }
380
381 return kind;
382}
383
384/// LexExclaim:
385/// !foo
386/// !
387lltok::Kind LLLexer::LexExclaim() {
388 // Lex a metadata name as a MetadataVar.
389 if (isalpha(static_cast<unsigned char>(CurPtr[0])) ||
390 CurPtr[0] == '-' || CurPtr[0] == '$' ||
391 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\') {
392 ++CurPtr;
393 while (isalnum(static_cast<unsigned char>(CurPtr[0])) ||
394 CurPtr[0] == '-' || CurPtr[0] == '$' ||
395 CurPtr[0] == '.' || CurPtr[0] == '_' || CurPtr[0] == '\\')
396 ++CurPtr;
397
398 StrVal.assign(TokStart+1, CurPtr); // Skip !
399 UnEscapeLexed(StrVal);
400 return lltok::MetadataVar;
401 }
402 return lltok::exclaim;
403}
404
405/// LexHash - Lex all tokens that start with a # character:
406/// AttrGrpID ::= #[0-9]+
407lltok::Kind LLLexer::LexHash() {
408 // Handle AttrGrpID: #[0-9]+
409 if (isdigit(static_cast<unsigned char>(CurPtr[0]))) {
410 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
411 /*empty*/;
412
413 uint64_t Val = atoull(TokStart+1, CurPtr);
414 if ((unsigned)Val != Val)
415 Error("invalid value number (too large)!");
416 UIntVal = unsigned(Val);
417 return lltok::AttrGrpID;
418 }
419
420 return lltok::Error;
421}
422
423/// LexIdentifier: Handle several related productions:
424/// Label [-a-zA-Z$._0-9]+:
425/// IntegerType i[0-9]+
426/// Keyword sdiv, float, ...
427/// HexIntConstant [us]0x[0-9A-Fa-f]+
428lltok::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(returned);
586 KEYWORD(returns_twice);
587 KEYWORD(signext);
588 KEYWORD(sret);
589 KEYWORD(ssp);
590 KEYWORD(sspreq);
591 KEYWORD(sspstrong);
592 KEYWORD(sanitize_address);
593 KEYWORD(sanitize_thread);
594 KEYWORD(sanitize_memory);
595 KEYWORD(uwtable);
596 KEYWORD(zeroext);
597
598 KEYWORD(type);
599 KEYWORD(opaque);
600
601 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
602 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
603 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
604 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
605
606 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
607 KEYWORD(umin);
608
609 KEYWORD(x);
610 KEYWORD(blockaddress);
611
612 KEYWORD(personality);
613 KEYWORD(cleanup);
614 KEYWORD(catch);
615 KEYWORD(filter);
616#undef KEYWORD
617
618 // Keywords for types.
619#define TYPEKEYWORD(STR, LLVMTY) \
620 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
621 TyVal = LLVMTY; return lltok::Type; }
622 TYPEKEYWORD("void", Type::getVoidTy(Context));
623 TYPEKEYWORD("half", Type::getHalfTy(Context));
624 TYPEKEYWORD("float", Type::getFloatTy(Context));
625 TYPEKEYWORD("double", Type::getDoubleTy(Context));
626 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
627 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
628 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
629 TYPEKEYWORD("label", Type::getLabelTy(Context));
630 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
631 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
632#undef TYPEKEYWORD
633
634 // Keywords for instructions.
635#define INSTKEYWORD(STR, Enum) \
636 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
637 UIntVal = Instruction::Enum; return lltok::kw_##STR; }
638
639 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
640 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
641 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
642 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
643 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
644 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
645 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
646 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
647
648 INSTKEYWORD(phi, PHI);
649 INSTKEYWORD(call, Call);
650 INSTKEYWORD(trunc, Trunc);
651 INSTKEYWORD(zext, ZExt);
652 INSTKEYWORD(sext, SExt);
653 INSTKEYWORD(fptrunc, FPTrunc);
654 INSTKEYWORD(fpext, FPExt);
655 INSTKEYWORD(uitofp, UIToFP);
656 INSTKEYWORD(sitofp, SIToFP);
657 INSTKEYWORD(fptoui, FPToUI);
658 INSTKEYWORD(fptosi, FPToSI);
659 INSTKEYWORD(inttoptr, IntToPtr);
660 INSTKEYWORD(ptrtoint, PtrToInt);
661 INSTKEYWORD(bitcast, BitCast);
662 INSTKEYWORD(select, Select);
663 INSTKEYWORD(va_arg, VAArg);
664 INSTKEYWORD(ret, Ret);
665 INSTKEYWORD(br, Br);
666 INSTKEYWORD(switch, Switch);
667 INSTKEYWORD(indirectbr, IndirectBr);
668 INSTKEYWORD(invoke, Invoke);
669 INSTKEYWORD(resume, Resume);
670 INSTKEYWORD(unreachable, Unreachable);
671
672 INSTKEYWORD(alloca, Alloca);
673 INSTKEYWORD(load, Load);
674 INSTKEYWORD(store, Store);
675 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
676 INSTKEYWORD(atomicrmw, AtomicRMW);
677 INSTKEYWORD(fence, Fence);
678 INSTKEYWORD(getelementptr, GetElementPtr);
679
680 INSTKEYWORD(extractelement, ExtractElement);
681 INSTKEYWORD(insertelement, InsertElement);
682 INSTKEYWORD(shufflevector, ShuffleVector);
683 INSTKEYWORD(extractvalue, ExtractValue);
684 INSTKEYWORD(insertvalue, InsertValue);
685 INSTKEYWORD(landingpad, LandingPad);
686#undef INSTKEYWORD
687
688 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
689 // the CFE to avoid forcing it to deal with 64-bit numbers.
690 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
691 TokStart[1] == '0' && TokStart[2] == 'x' &&
692 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
693 int len = CurPtr-TokStart-3;
694 uint32_t bits = len * 4;
695 APInt Tmp(bits, StringRef(TokStart+3, len), 16);
696 uint32_t activeBits = Tmp.getActiveBits();
697 if (activeBits > 0 && activeBits < bits)
698 Tmp = Tmp.trunc(activeBits);
699 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
700 return lltok::APSInt;
701 }
702
703 // If this is "cc1234", return this as just "cc".
704 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
705 CurPtr = TokStart+2;
706 return lltok::kw_cc;
707 }
708
709 // Finally, if this isn't known, return an error.
710 CurPtr = TokStart+1;
711 return lltok::Error;
712}
713
714
715/// Lex0x: Handle productions that start with 0x, knowing that it matches and
716/// that this is not a label:
717/// HexFPConstant 0x[0-9A-Fa-f]+
718/// HexFP80Constant 0xK[0-9A-Fa-f]+
719/// HexFP128Constant 0xL[0-9A-Fa-f]+
720/// HexPPC128Constant 0xM[0-9A-Fa-f]+
721/// HexHalfConstant 0xH[0-9A-Fa-f]+
722lltok::Kind LLLexer::Lex0x() {
723 CurPtr = TokStart + 2;
724
725 char Kind;
726 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
727 Kind = *CurPtr++;
728 } else {
729 Kind = 'J';
730 }
731
732 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
733 // Bad token, return it as an error.
734 CurPtr = TokStart+1;
735 return lltok::Error;
736 }
737
738 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
739 ++CurPtr;
740
741 if (Kind == 'J') {
742 // HexFPConstant - Floating point constant represented in IEEE format as a
743 // hexadecimal number for when exponential notation is not precise enough.
744 // Half, Float, and double only.
745 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
746 return lltok::APFloat;
747 }
748
749 uint64_t Pair[2];
750 switch (Kind) {
751 default: llvm_unreachable("Unknown kind!");
752 case 'K':
753 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
754 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
755 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair));
756 return lltok::APFloat;
757 case 'L':
758 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
759 HexToIntPair(TokStart+3, CurPtr, Pair);
760 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair));
761 return lltok::APFloat;
762 case 'M':
763 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
764 HexToIntPair(TokStart+3, CurPtr, Pair);
765 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair));
766 return lltok::APFloat;
767 case 'H':
768 APFloatVal = APFloat(APFloat::IEEEhalf,
769 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
770 return lltok::APFloat;
771 }
772}
773
774/// LexIdentifier: Handle several related productions:
775/// Label [-a-zA-Z$._0-9]+:
776/// NInteger -[0-9]+
777/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
778/// PInteger [0-9]+
779/// HexFPConstant 0x[0-9A-Fa-f]+
780/// HexFP80Constant 0xK[0-9A-Fa-f]+
781/// HexFP128Constant 0xL[0-9A-Fa-f]+
782/// HexPPC128Constant 0xM[0-9A-Fa-f]+
783lltok::Kind LLLexer::LexDigitOrNegative() {
784 // If the letter after the negative is not a number, this is probably a label.
785 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
786 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
787 // Okay, this is not a number after the -, it's probably a label.
788 if (const char *End = isLabelTail(CurPtr)) {
789 StrVal.assign(TokStart, End-1);
790 CurPtr = End;
791 return lltok::LabelStr;
792 }
793
794 return lltok::Error;
795 }
796
797 // At this point, it is either a label, int or fp constant.
798
799 // Skip digits, we have at least one.
800 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
801 /*empty*/;
802
803 // Check to see if this really is a label afterall, e.g. "-1:".
804 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
805 if (const char *End = isLabelTail(CurPtr)) {
806 StrVal.assign(TokStart, End-1);
807 CurPtr = End;
808 return lltok::LabelStr;
809 }
810 }
811
812 // If the next character is a '.', then it is a fp value, otherwise its
813 // integer.
814 if (CurPtr[0] != '.') {
815 if (TokStart[0] == '0' && TokStart[1] == 'x')
816 return Lex0x();
817 unsigned Len = CurPtr-TokStart;
818 uint32_t numBits = ((Len * 64) / 19) + 2;
819 APInt Tmp(numBits, StringRef(TokStart, Len), 10);
820 if (TokStart[0] == '-') {
821 uint32_t minBits = Tmp.getMinSignedBits();
822 if (minBits > 0 && minBits < numBits)
823 Tmp = Tmp.trunc(minBits);
824 APSIntVal = APSInt(Tmp, false);
825 } else {
826 uint32_t activeBits = Tmp.getActiveBits();
827 if (activeBits > 0 && activeBits < numBits)
828 Tmp = Tmp.trunc(activeBits);
829 APSIntVal = APSInt(Tmp, true);
830 }
831 return lltok::APSInt;
832 }
833
834 ++CurPtr;
835
836 // Skip over [0-9]*([eE][-+]?[0-9]+)?
837 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
838
839 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
840 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
841 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
842 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
843 CurPtr += 2;
844 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
845 }
846 }
847
848 APFloatVal = APFloat(std::atof(TokStart));
849 return lltok::APFloat;
850}
851
852/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
853lltok::Kind LLLexer::LexPositive() {
854 // If the letter after the negative is a number, this is probably not a
855 // label.
856 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
857 return lltok::Error;
858
859 // Skip digits.
860 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
861 /*empty*/;
862
863 // At this point, we need a '.'.
864 if (CurPtr[0] != '.') {
865 CurPtr = TokStart+1;
866 return lltok::Error;
867 }
868
869 ++CurPtr;
870
871 // Skip over [0-9]*([eE][-+]?[0-9]+)?
872 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
873
874 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
875 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
876 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
877 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
878 CurPtr += 2;
879 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
880 }
881 }
882
883 APFloatVal = APFloat(std::atof(TokStart));
884 return lltok::APFloat;
885}
| 561
562 KEYWORD(cc);
563 KEYWORD(c);
564
565 KEYWORD(attributes);
566
567 KEYWORD(alwaysinline);
568 KEYWORD(byval);
569 KEYWORD(inlinehint);
570 KEYWORD(inreg);
571 KEYWORD(minsize);
572 KEYWORD(naked);
573 KEYWORD(nest);
574 KEYWORD(noalias);
575 KEYWORD(nobuiltin);
576 KEYWORD(nocapture);
577 KEYWORD(noduplicate);
578 KEYWORD(noimplicitfloat);
579 KEYWORD(noinline);
580 KEYWORD(nonlazybind);
581 KEYWORD(noredzone);
582 KEYWORD(noreturn);
583 KEYWORD(nounwind);
584 KEYWORD(optsize);
585 KEYWORD(readnone);
586 KEYWORD(readonly);
587 KEYWORD(returned);
588 KEYWORD(returns_twice);
589 KEYWORD(signext);
590 KEYWORD(sret);
591 KEYWORD(ssp);
592 KEYWORD(sspreq);
593 KEYWORD(sspstrong);
594 KEYWORD(sanitize_address);
595 KEYWORD(sanitize_thread);
596 KEYWORD(sanitize_memory);
597 KEYWORD(uwtable);
598 KEYWORD(zeroext);
599
600 KEYWORD(type);
601 KEYWORD(opaque);
602
603 KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
604 KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
605 KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
606 KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
607
608 KEYWORD(xchg); KEYWORD(nand); KEYWORD(max); KEYWORD(min); KEYWORD(umax);
609 KEYWORD(umin);
610
611 KEYWORD(x);
612 KEYWORD(blockaddress);
613
614 KEYWORD(personality);
615 KEYWORD(cleanup);
616 KEYWORD(catch);
617 KEYWORD(filter);
618#undef KEYWORD
619
620 // Keywords for types.
621#define TYPEKEYWORD(STR, LLVMTY) \
622 if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
623 TyVal = LLVMTY; return lltok::Type; }
624 TYPEKEYWORD("void", Type::getVoidTy(Context));
625 TYPEKEYWORD("half", Type::getHalfTy(Context));
626 TYPEKEYWORD("float", Type::getFloatTy(Context));
627 TYPEKEYWORD("double", Type::getDoubleTy(Context));
628 TYPEKEYWORD("x86_fp80", Type::getX86_FP80Ty(Context));
629 TYPEKEYWORD("fp128", Type::getFP128Ty(Context));
630 TYPEKEYWORD("ppc_fp128", Type::getPPC_FP128Ty(Context));
631 TYPEKEYWORD("label", Type::getLabelTy(Context));
632 TYPEKEYWORD("metadata", Type::getMetadataTy(Context));
633 TYPEKEYWORD("x86_mmx", Type::getX86_MMXTy(Context));
634#undef TYPEKEYWORD
635
636 // Keywords for instructions.
637#define INSTKEYWORD(STR, Enum) \
638 if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
639 UIntVal = Instruction::Enum; return lltok::kw_##STR; }
640
641 INSTKEYWORD(add, Add); INSTKEYWORD(fadd, FAdd);
642 INSTKEYWORD(sub, Sub); INSTKEYWORD(fsub, FSub);
643 INSTKEYWORD(mul, Mul); INSTKEYWORD(fmul, FMul);
644 INSTKEYWORD(udiv, UDiv); INSTKEYWORD(sdiv, SDiv); INSTKEYWORD(fdiv, FDiv);
645 INSTKEYWORD(urem, URem); INSTKEYWORD(srem, SRem); INSTKEYWORD(frem, FRem);
646 INSTKEYWORD(shl, Shl); INSTKEYWORD(lshr, LShr); INSTKEYWORD(ashr, AShr);
647 INSTKEYWORD(and, And); INSTKEYWORD(or, Or); INSTKEYWORD(xor, Xor);
648 INSTKEYWORD(icmp, ICmp); INSTKEYWORD(fcmp, FCmp);
649
650 INSTKEYWORD(phi, PHI);
651 INSTKEYWORD(call, Call);
652 INSTKEYWORD(trunc, Trunc);
653 INSTKEYWORD(zext, ZExt);
654 INSTKEYWORD(sext, SExt);
655 INSTKEYWORD(fptrunc, FPTrunc);
656 INSTKEYWORD(fpext, FPExt);
657 INSTKEYWORD(uitofp, UIToFP);
658 INSTKEYWORD(sitofp, SIToFP);
659 INSTKEYWORD(fptoui, FPToUI);
660 INSTKEYWORD(fptosi, FPToSI);
661 INSTKEYWORD(inttoptr, IntToPtr);
662 INSTKEYWORD(ptrtoint, PtrToInt);
663 INSTKEYWORD(bitcast, BitCast);
664 INSTKEYWORD(select, Select);
665 INSTKEYWORD(va_arg, VAArg);
666 INSTKEYWORD(ret, Ret);
667 INSTKEYWORD(br, Br);
668 INSTKEYWORD(switch, Switch);
669 INSTKEYWORD(indirectbr, IndirectBr);
670 INSTKEYWORD(invoke, Invoke);
671 INSTKEYWORD(resume, Resume);
672 INSTKEYWORD(unreachable, Unreachable);
673
674 INSTKEYWORD(alloca, Alloca);
675 INSTKEYWORD(load, Load);
676 INSTKEYWORD(store, Store);
677 INSTKEYWORD(cmpxchg, AtomicCmpXchg);
678 INSTKEYWORD(atomicrmw, AtomicRMW);
679 INSTKEYWORD(fence, Fence);
680 INSTKEYWORD(getelementptr, GetElementPtr);
681
682 INSTKEYWORD(extractelement, ExtractElement);
683 INSTKEYWORD(insertelement, InsertElement);
684 INSTKEYWORD(shufflevector, ShuffleVector);
685 INSTKEYWORD(extractvalue, ExtractValue);
686 INSTKEYWORD(insertvalue, InsertValue);
687 INSTKEYWORD(landingpad, LandingPad);
688#undef INSTKEYWORD
689
690 // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
691 // the CFE to avoid forcing it to deal with 64-bit numbers.
692 if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
693 TokStart[1] == '0' && TokStart[2] == 'x' &&
694 isxdigit(static_cast<unsigned char>(TokStart[3]))) {
695 int len = CurPtr-TokStart-3;
696 uint32_t bits = len * 4;
697 APInt Tmp(bits, StringRef(TokStart+3, len), 16);
698 uint32_t activeBits = Tmp.getActiveBits();
699 if (activeBits > 0 && activeBits < bits)
700 Tmp = Tmp.trunc(activeBits);
701 APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
702 return lltok::APSInt;
703 }
704
705 // If this is "cc1234", return this as just "cc".
706 if (TokStart[0] == 'c' && TokStart[1] == 'c') {
707 CurPtr = TokStart+2;
708 return lltok::kw_cc;
709 }
710
711 // Finally, if this isn't known, return an error.
712 CurPtr = TokStart+1;
713 return lltok::Error;
714}
715
716
717/// Lex0x: Handle productions that start with 0x, knowing that it matches and
718/// that this is not a label:
719/// HexFPConstant 0x[0-9A-Fa-f]+
720/// HexFP80Constant 0xK[0-9A-Fa-f]+
721/// HexFP128Constant 0xL[0-9A-Fa-f]+
722/// HexPPC128Constant 0xM[0-9A-Fa-f]+
723/// HexHalfConstant 0xH[0-9A-Fa-f]+
724lltok::Kind LLLexer::Lex0x() {
725 CurPtr = TokStart + 2;
726
727 char Kind;
728 if ((CurPtr[0] >= 'K' && CurPtr[0] <= 'M') || CurPtr[0] == 'H') {
729 Kind = *CurPtr++;
730 } else {
731 Kind = 'J';
732 }
733
734 if (!isxdigit(static_cast<unsigned char>(CurPtr[0]))) {
735 // Bad token, return it as an error.
736 CurPtr = TokStart+1;
737 return lltok::Error;
738 }
739
740 while (isxdigit(static_cast<unsigned char>(CurPtr[0])))
741 ++CurPtr;
742
743 if (Kind == 'J') {
744 // HexFPConstant - Floating point constant represented in IEEE format as a
745 // hexadecimal number for when exponential notation is not precise enough.
746 // Half, Float, and double only.
747 APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
748 return lltok::APFloat;
749 }
750
751 uint64_t Pair[2];
752 switch (Kind) {
753 default: llvm_unreachable("Unknown kind!");
754 case 'K':
755 // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
756 FP80HexToIntPair(TokStart+3, CurPtr, Pair);
757 APFloatVal = APFloat(APFloat::x87DoubleExtended, APInt(80, Pair));
758 return lltok::APFloat;
759 case 'L':
760 // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
761 HexToIntPair(TokStart+3, CurPtr, Pair);
762 APFloatVal = APFloat(APFloat::IEEEquad, APInt(128, Pair));
763 return lltok::APFloat;
764 case 'M':
765 // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
766 HexToIntPair(TokStart+3, CurPtr, Pair);
767 APFloatVal = APFloat(APFloat::PPCDoubleDouble, APInt(128, Pair));
768 return lltok::APFloat;
769 case 'H':
770 APFloatVal = APFloat(APFloat::IEEEhalf,
771 APInt(16,HexIntToVal(TokStart+3, CurPtr)));
772 return lltok::APFloat;
773 }
774}
775
776/// LexIdentifier: Handle several related productions:
777/// Label [-a-zA-Z$._0-9]+:
778/// NInteger -[0-9]+
779/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
780/// PInteger [0-9]+
781/// HexFPConstant 0x[0-9A-Fa-f]+
782/// HexFP80Constant 0xK[0-9A-Fa-f]+
783/// HexFP128Constant 0xL[0-9A-Fa-f]+
784/// HexPPC128Constant 0xM[0-9A-Fa-f]+
785lltok::Kind LLLexer::LexDigitOrNegative() {
786 // If the letter after the negative is not a number, this is probably a label.
787 if (!isdigit(static_cast<unsigned char>(TokStart[0])) &&
788 !isdigit(static_cast<unsigned char>(CurPtr[0]))) {
789 // Okay, this is not a number after the -, it's probably a label.
790 if (const char *End = isLabelTail(CurPtr)) {
791 StrVal.assign(TokStart, End-1);
792 CurPtr = End;
793 return lltok::LabelStr;
794 }
795
796 return lltok::Error;
797 }
798
799 // At this point, it is either a label, int or fp constant.
800
801 // Skip digits, we have at least one.
802 for (; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
803 /*empty*/;
804
805 // Check to see if this really is a label afterall, e.g. "-1:".
806 if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
807 if (const char *End = isLabelTail(CurPtr)) {
808 StrVal.assign(TokStart, End-1);
809 CurPtr = End;
810 return lltok::LabelStr;
811 }
812 }
813
814 // If the next character is a '.', then it is a fp value, otherwise its
815 // integer.
816 if (CurPtr[0] != '.') {
817 if (TokStart[0] == '0' && TokStart[1] == 'x')
818 return Lex0x();
819 unsigned Len = CurPtr-TokStart;
820 uint32_t numBits = ((Len * 64) / 19) + 2;
821 APInt Tmp(numBits, StringRef(TokStart, Len), 10);
822 if (TokStart[0] == '-') {
823 uint32_t minBits = Tmp.getMinSignedBits();
824 if (minBits > 0 && minBits < numBits)
825 Tmp = Tmp.trunc(minBits);
826 APSIntVal = APSInt(Tmp, false);
827 } else {
828 uint32_t activeBits = Tmp.getActiveBits();
829 if (activeBits > 0 && activeBits < numBits)
830 Tmp = Tmp.trunc(activeBits);
831 APSIntVal = APSInt(Tmp, true);
832 }
833 return lltok::APSInt;
834 }
835
836 ++CurPtr;
837
838 // Skip over [0-9]*([eE][-+]?[0-9]+)?
839 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
840
841 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
842 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
843 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
844 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
845 CurPtr += 2;
846 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
847 }
848 }
849
850 APFloatVal = APFloat(std::atof(TokStart));
851 return lltok::APFloat;
852}
853
854/// FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
855lltok::Kind LLLexer::LexPositive() {
856 // If the letter after the negative is a number, this is probably not a
857 // label.
858 if (!isdigit(static_cast<unsigned char>(CurPtr[0])))
859 return lltok::Error;
860
861 // Skip digits.
862 for (++CurPtr; isdigit(static_cast<unsigned char>(CurPtr[0])); ++CurPtr)
863 /*empty*/;
864
865 // At this point, we need a '.'.
866 if (CurPtr[0] != '.') {
867 CurPtr = TokStart+1;
868 return lltok::Error;
869 }
870
871 ++CurPtr;
872
873 // Skip over [0-9]*([eE][-+]?[0-9]+)?
874 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
875
876 if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
877 if (isdigit(static_cast<unsigned char>(CurPtr[1])) ||
878 ((CurPtr[1] == '-' || CurPtr[1] == '+') &&
879 isdigit(static_cast<unsigned char>(CurPtr[2])))) {
880 CurPtr += 2;
881 while (isdigit(static_cast<unsigned char>(CurPtr[0]))) ++CurPtr;
882 }
883 }
884
885 APFloatVal = APFloat(std::atof(TokStart));
886 return lltok::APFloat;
887}
|