LLLexer.cpp revision 1b25cb2416c46a6cebf2a6c52235e9fe46a10d11 
1//===- LLLexer.cpp - Lexer for .ll Files ----------------------------------===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// Implement the Lexer for .ll files.
11//
12//===----------------------------------------------------------------------===//
13
14#include "LLLexer.h"
15#include "llvm/DerivedTypes.h"
16#include "llvm/Instruction.h"
17#include "llvm/Support/MemoryBuffer.h"
18#include "llvm/Support/MathExtras.h"
19#include "llvm/Support/raw_ostream.h"
20#include "llvm/Assembly/Parser.h"
21#include <cstdlib>
22#include <cstring>
23using namespace llvm;
24
25bool LLLexer::Error(LocTy ErrorLoc, const std::string &Msg) const {
26  // Scan backward to find the start of the line.
27  const char *LineStart = ErrorLoc;
28  while (LineStart != CurBuf->getBufferStart() &&
29         LineStart[-1] != '\n' && LineStart[-1] != '\r')
30    --LineStart;
31  // Get the end of the line.
32  const char *LineEnd = ErrorLoc;
33  while (LineEnd != CurBuf->getBufferEnd() &&
34         LineEnd[0] != '\n' && LineEnd[0] != '\r')
35    ++LineEnd;
36
37  unsigned LineNo = 1;
38  for (const char *FP = CurBuf->getBufferStart(); FP != ErrorLoc; ++FP)
39    if (*FP == '\n') ++LineNo;
40
41  std::string LineContents(LineStart, LineEnd);
42  ErrorInfo.setError(Msg, LineNo, ErrorLoc-LineStart, LineContents);
43  return true;
44}
45
46//===----------------------------------------------------------------------===//
47// Helper functions.
48//===----------------------------------------------------------------------===//
49
50// atoull - Convert an ascii string of decimal digits into the unsigned long
51// long representation... this does not have to do input error checking,
52// because we know that the input will be matched by a suitable regex...
53//
54uint64_t LLLexer::atoull(const char *Buffer, const char *End) {
55  uint64_t Result = 0;
56  for (; Buffer != End; Buffer++) {
57    uint64_t OldRes = Result;
58    Result *= 10;
59    Result += *Buffer-'0';
60    if (Result < OldRes) {  // Uh, oh, overflow detected!!!
61      Error("constant bigger than 64 bits detected!");
62      return 0;
63    }
64  }
65  return Result;
66}
67
68uint64_t LLLexer::HexIntToVal(const char *Buffer, const char *End) {
69  uint64_t Result = 0;
70  for (; Buffer != End; ++Buffer) {
71    uint64_t OldRes = Result;
72    Result *= 16;
73    char C = *Buffer;
74    if (C >= '0' && C <= '9')
75      Result += C-'0';
76    else if (C >= 'A' && C <= 'F')
77      Result += C-'A'+10;
78    else if (C >= 'a' && C <= 'f')
79      Result += C-'a'+10;
80
81    if (Result < OldRes) {   // Uh, oh, overflow detected!!!
82      Error("constant bigger than 64 bits detected!");
83      return 0;
84    }
85  }
86  return Result;
87}
88
89void LLLexer::HexToIntPair(const char *Buffer, const char *End,
90                           uint64_t Pair[2]) {
91  Pair[0] = 0;
92  for (int i=0; i<16; i++, Buffer++) {
93    assert(Buffer != End);
94    Pair[0] *= 16;
95    char C = *Buffer;
96    if (C >= '0' && C <= '9')
97      Pair[0] += C-'0';
98    else if (C >= 'A' && C <= 'F')
99      Pair[0] += C-'A'+10;
100    else if (C >= 'a' && C <= 'f')
101      Pair[0] += C-'a'+10;
102  }
103  Pair[1] = 0;
104  for (int i=0; i<16 && Buffer != End; i++, Buffer++) {
105    Pair[1] *= 16;
106    char C = *Buffer;
107    if (C >= '0' && C <= '9')
108      Pair[1] += C-'0';
109    else if (C >= 'A' && C <= 'F')
110      Pair[1] += C-'A'+10;
111    else if (C >= 'a' && C <= 'f')
112      Pair[1] += C-'a'+10;
113  }
114  if (Buffer != End)
115    Error("constant bigger than 128 bits detected!");
116}
117
118/// FP80HexToIntPair - translate an 80 bit FP80 number (20 hexits) into
119/// { low64, high16 } as usual for an APInt.
120void LLLexer::FP80HexToIntPair(const char *Buffer, const char *End,
121                           uint64_t Pair[2]) {
122  Pair[1] = 0;
123  for (int i=0; i<4 && Buffer != End; i++, Buffer++) {
124    assert(Buffer != End);
125    Pair[1] *= 16;
126    char C = *Buffer;
127    if (C >= '0' && C <= '9')
128      Pair[1] += C-'0';
129    else if (C >= 'A' && C <= 'F')
130      Pair[1] += C-'A'+10;
131    else if (C >= 'a' && C <= 'f')
132      Pair[1] += C-'a'+10;
133  }
134  Pair[0] = 0;
135  for (int i=0; i<16; i++, Buffer++) {
136    Pair[0] *= 16;
137    char C = *Buffer;
138    if (C >= '0' && C <= '9')
139      Pair[0] += C-'0';
140    else if (C >= 'A' && C <= 'F')
141      Pair[0] += C-'A'+10;
142    else if (C >= 'a' && C <= 'f')
143      Pair[0] += C-'a'+10;
144  }
145  if (Buffer != End)
146    Error("constant bigger than 128 bits detected!");
147}
148
149// UnEscapeLexed - Run through the specified buffer and change \xx codes to the
150// appropriate character.
151static void UnEscapeLexed(std::string &Str) {
152  if (Str.empty()) return;
153
154  char *Buffer = &Str[0], *EndBuffer = Buffer+Str.size();
155  char *BOut = Buffer;
156  for (char *BIn = Buffer; BIn != EndBuffer; ) {
157    if (BIn[0] == '\\') {
158      if (BIn < EndBuffer-1 && BIn[1] == '\\') {
159        *BOut++ = '\\'; // Two \ becomes one
160        BIn += 2;
161      } else if (BIn < EndBuffer-2 && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
162        char Tmp = BIn[3]; BIn[3] = 0;      // Terminate string
163        *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number
164        BIn[3] = Tmp;                       // Restore character
165        BIn += 3;                           // Skip over handled chars
166        ++BOut;
167      } else {
168        *BOut++ = *BIn++;
169      }
170    } else {
171      *BOut++ = *BIn++;
172    }
173  }
174  Str.resize(BOut-Buffer);
175}
176
177/// isLabelChar - Return true for [-a-zA-Z$._0-9].
178static bool isLabelChar(char C) {
179  return isalnum(C) || C == '-' || C == '$' || C == '.' || C == '_';
180}
181
182
183/// isLabelTail - Return true if this pointer points to a valid end of a label.
184static const char *isLabelTail(const char *CurPtr) {
185  while (1) {
186    if (CurPtr[0] == ':') return CurPtr+1;
187    if (!isLabelChar(CurPtr[0])) return 0;
188    ++CurPtr;
189  }
190}
191
192
193
194//===----------------------------------------------------------------------===//
195// Lexer definition.
196//===----------------------------------------------------------------------===//
197
198LLLexer::LLLexer(MemoryBuffer *StartBuf, ParseError &Err)
199  : CurBuf(StartBuf), ErrorInfo(Err), APFloatVal(0.0) {
200  CurPtr = CurBuf->getBufferStart();
201}
202
203std::string LLLexer::getFilename() const {
204  return CurBuf->getBufferIdentifier();
205}
206
207int LLLexer::getNextChar() {
208  char CurChar = *CurPtr++;
209  switch (CurChar) {
210  default: return (unsigned char)CurChar;
211  case 0:
212    // A nul character in the stream is either the end of the current buffer or
213    // a random nul in the file.  Disambiguate that here.
214    if (CurPtr-1 != CurBuf->getBufferEnd())
215      return 0;  // Just whitespace.
216
217    // Otherwise, return end of file.
218    --CurPtr;  // Another call to lex will return EOF again.
219    return EOF;
220  }
221}
222
223
224lltok::Kind LLLexer::LexToken() {
225  TokStart = CurPtr;
226
227  int CurChar = getNextChar();
228  switch (CurChar) {
229  default:
230    // Handle letters: [a-zA-Z_]
231    if (isalpha(CurChar) || CurChar == '_')
232      return LexIdentifier();
233
234    return lltok::Error;
235  case EOF: return lltok::Eof;
236  case 0:
237  case ' ':
238  case '\t':
239  case '\n':
240  case '\r':
241    // Ignore whitespace.
242    return LexToken();
243  case '+': return LexPositive();
244  case '@': return LexAt();
245  case '%': return LexPercent();
246  case '"': return LexQuote();
247  case '.':
248    if (const char *Ptr = isLabelTail(CurPtr)) {
249      CurPtr = Ptr;
250      StrVal.assign(TokStart, CurPtr-1);
251      return lltok::LabelStr;
252    }
253    if (CurPtr[0] == '.' && CurPtr[1] == '.') {
254      CurPtr += 2;
255      return lltok::dotdotdot;
256    }
257    return lltok::Error;
258  case '$':
259    if (const char *Ptr = isLabelTail(CurPtr)) {
260      CurPtr = Ptr;
261      StrVal.assign(TokStart, CurPtr-1);
262      return lltok::LabelStr;
263    }
264    return lltok::Error;
265  case ';':
266    SkipLineComment();
267    return LexToken();
268  case '0': case '1': case '2': case '3': case '4':
269  case '5': case '6': case '7': case '8': case '9':
270  case '-':
271    return LexDigitOrNegative();
272  case '=': return lltok::equal;
273  case '[': return lltok::lsquare;
274  case ']': return lltok::rsquare;
275  case '{': return lltok::lbrace;
276  case '}': return lltok::rbrace;
277  case '<': return lltok::less;
278  case '>': return lltok::greater;
279  case '(': return lltok::lparen;
280  case ')': return lltok::rparen;
281  case ',': return lltok::comma;
282  case '*': return lltok::star;
283  case '\\': return lltok::backslash;
284  }
285}
286
287void LLLexer::SkipLineComment() {
288  while (1) {
289    if (CurPtr[0] == '\n' || CurPtr[0] == '\r' || getNextChar() == EOF)
290      return;
291  }
292}
293
294/// LexAt - Lex all tokens that start with an @ character:
295///   GlobalVar   @\"[^\"]*\"
296///   GlobalVar   @[-a-zA-Z$._][-a-zA-Z$._0-9]*
297///   GlobalVarID @[0-9]+
298lltok::Kind LLLexer::LexAt() {
299  // Handle AtStringConstant: @\"[^\"]*\"
300  if (CurPtr[0] == '"') {
301    ++CurPtr;
302
303    while (1) {
304      int CurChar = getNextChar();
305
306      if (CurChar == EOF) {
307        Error("end of file in global variable name");
308        return lltok::Error;
309      }
310      if (CurChar == '"') {
311        StrVal.assign(TokStart+2, CurPtr-1);
312        UnEscapeLexed(StrVal);
313        return lltok::GlobalVar;
314      }
315    }
316  }
317
318  // Handle GlobalVarName: @[-a-zA-Z$._][-a-zA-Z$._0-9]*
319  if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
320      CurPtr[0] == '.' || CurPtr[0] == '_') {
321    ++CurPtr;
322    while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
323           CurPtr[0] == '.' || CurPtr[0] == '_')
324      ++CurPtr;
325
326    StrVal.assign(TokStart+1, CurPtr);   // Skip @
327    return lltok::GlobalVar;
328  }
329
330  // Handle GlobalVarID: @[0-9]+
331  if (isdigit(CurPtr[0])) {
332    for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
333      /*empty*/;
334
335    uint64_t Val = atoull(TokStart+1, CurPtr);
336    if ((unsigned)Val != Val)
337      Error("invalid value number (too large)!");
338    UIntVal = unsigned(Val);
339    return lltok::GlobalID;
340  }
341
342  return lltok::Error;
343}
344
345
346/// LexPercent - Lex all tokens that start with a % character:
347///   LocalVar   ::= %\"[^\"]*\"
348///   LocalVar   ::= %[-a-zA-Z$._][-a-zA-Z$._0-9]*
349///   LocalVarID ::= %[0-9]+
350lltok::Kind LLLexer::LexPercent() {
351  // Handle LocalVarName: %\"[^\"]*\"
352  if (CurPtr[0] == '"') {
353    ++CurPtr;
354
355    while (1) {
356      int CurChar = getNextChar();
357
358      if (CurChar == EOF) {
359        Error("end of file in string constant");
360        return lltok::Error;
361      }
362      if (CurChar == '"') {
363        StrVal.assign(TokStart+2, CurPtr-1);
364        UnEscapeLexed(StrVal);
365        return lltok::LocalVar;
366      }
367    }
368  }
369
370  // Handle LocalVarName: %[-a-zA-Z$._][-a-zA-Z$._0-9]*
371  if (isalpha(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
372      CurPtr[0] == '.' || CurPtr[0] == '_') {
373    ++CurPtr;
374    while (isalnum(CurPtr[0]) || CurPtr[0] == '-' || CurPtr[0] == '$' ||
375           CurPtr[0] == '.' || CurPtr[0] == '_')
376      ++CurPtr;
377
378    StrVal.assign(TokStart+1, CurPtr);   // Skip %
379    return lltok::LocalVar;
380  }
381
382  // Handle LocalVarID: %[0-9]+
383  if (isdigit(CurPtr[0])) {
384    for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
385      /*empty*/;
386
387    uint64_t Val = atoull(TokStart+1, CurPtr);
388    if ((unsigned)Val != Val)
389      Error("invalid value number (too large)!");
390    UIntVal = unsigned(Val);
391    return lltok::LocalVarID;
392  }
393
394  return lltok::Error;
395}
396
397/// LexQuote - Lex all tokens that start with a " character:
398///   QuoteLabel        "[^"]+":
399///   StringConstant    "[^"]*"
400lltok::Kind LLLexer::LexQuote() {
401  while (1) {
402    int CurChar = getNextChar();
403
404    if (CurChar == EOF) {
405      Error("end of file in quoted string");
406      return lltok::Error;
407    }
408
409    if (CurChar != '"') continue;
410
411    if (CurPtr[0] != ':') {
412      StrVal.assign(TokStart+1, CurPtr-1);
413      UnEscapeLexed(StrVal);
414      return lltok::StringConstant;
415    }
416
417    ++CurPtr;
418    StrVal.assign(TokStart+1, CurPtr-2);
419    UnEscapeLexed(StrVal);
420    return lltok::LabelStr;
421  }
422}
423
424static bool JustWhitespaceNewLine(const char *&Ptr) {
425  const char *ThisPtr = Ptr;
426  while (*ThisPtr == ' ' || *ThisPtr == '\t')
427    ++ThisPtr;
428  if (*ThisPtr == '\n' || *ThisPtr == '\r') {
429    Ptr = ThisPtr;
430    return true;
431  }
432  return false;
433}
434
435
436/// LexIdentifier: Handle several related productions:
437///    Label           [-a-zA-Z$._0-9]+:
438///    IntegerType     i[0-9]+
439///    Keyword         sdiv, float, ...
440///    HexIntConstant  [us]0x[0-9A-Fa-f]+
441lltok::Kind LLLexer::LexIdentifier() {
442  const char *StartChar = CurPtr;
443  const char *IntEnd = CurPtr[-1] == 'i' ? 0 : StartChar;
444  const char *KeywordEnd = 0;
445
446  for (; isLabelChar(*CurPtr); ++CurPtr) {
447    // If we decide this is an integer, remember the end of the sequence.
448    if (!IntEnd && !isdigit(*CurPtr)) IntEnd = CurPtr;
449    if (!KeywordEnd && !isalnum(*CurPtr) && *CurPtr != '_') KeywordEnd = CurPtr;
450  }
451
452  // If we stopped due to a colon, this really is a label.
453  if (*CurPtr == ':') {
454    StrVal.assign(StartChar-1, CurPtr++);
455    return lltok::LabelStr;
456  }
457
458  // Otherwise, this wasn't a label.  If this was valid as an integer type,
459  // return it.
460  if (IntEnd == 0) IntEnd = CurPtr;
461  if (IntEnd != StartChar) {
462    CurPtr = IntEnd;
463    uint64_t NumBits = atoull(StartChar, CurPtr);
464    if (NumBits < IntegerType::MIN_INT_BITS ||
465        NumBits > IntegerType::MAX_INT_BITS) {
466      Error("bitwidth for integer type out of range!");
467      return lltok::Error;
468    }
469    TyVal = IntegerType::get(NumBits);
470    return lltok::Type;
471  }
472
473  // Otherwise, this was a letter sequence.  See which keyword this is.
474  if (KeywordEnd == 0) KeywordEnd = CurPtr;
475  CurPtr = KeywordEnd;
476  --StartChar;
477  unsigned Len = CurPtr-StartChar;
478#define KEYWORD(STR) \
479  if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) \
480    return lltok::kw_##STR;
481
482  KEYWORD(begin);   KEYWORD(end);
483  KEYWORD(true);    KEYWORD(false);
484  KEYWORD(declare); KEYWORD(define);
485  KEYWORD(global);  KEYWORD(constant);
486
487  KEYWORD(private);
488  KEYWORD(internal);
489  KEYWORD(linkonce);
490  KEYWORD(linkonce_odr);
491  KEYWORD(weak);
492  KEYWORD(weak_odr);
493  KEYWORD(appending);
494  KEYWORD(dllimport);
495  KEYWORD(dllexport);
496  KEYWORD(common);
497  KEYWORD(default);
498  KEYWORD(hidden);
499  KEYWORD(protected);
500  KEYWORD(extern_weak);
501  KEYWORD(external);
502  KEYWORD(thread_local);
503  KEYWORD(zeroinitializer);
504  KEYWORD(undef);
505  KEYWORD(null);
506  KEYWORD(to);
507  KEYWORD(tail);
508  KEYWORD(target);
509  KEYWORD(triple);
510  KEYWORD(deplibs);
511  KEYWORD(datalayout);
512  KEYWORD(volatile);
513  KEYWORD(align);
514  KEYWORD(addrspace);
515  KEYWORD(section);
516  KEYWORD(alias);
517  KEYWORD(module);
518  KEYWORD(asm);
519  KEYWORD(sideeffect);
520  KEYWORD(gc);
521
522  KEYWORD(ccc);
523  KEYWORD(fastcc);
524  KEYWORD(coldcc);
525  KEYWORD(x86_stdcallcc);
526  KEYWORD(x86_fastcallcc);
527  KEYWORD(cc);
528  KEYWORD(c);
529
530  KEYWORD(signext);
531  KEYWORD(zeroext);
532  KEYWORD(inreg);
533  KEYWORD(sret);
534  KEYWORD(nounwind);
535  KEYWORD(noreturn);
536  KEYWORD(noalias);
537  KEYWORD(nocapture);
538  KEYWORD(byval);
539  KEYWORD(nest);
540  KEYWORD(readnone);
541  KEYWORD(readonly);
542
543  KEYWORD(noinline);
544  KEYWORD(alwaysinline);
545  KEYWORD(optsize);
546  KEYWORD(ssp);
547  KEYWORD(sspreq);
548
549  KEYWORD(type);
550  KEYWORD(opaque);
551
552  KEYWORD(eq); KEYWORD(ne); KEYWORD(slt); KEYWORD(sgt); KEYWORD(sle);
553  KEYWORD(sge); KEYWORD(ult); KEYWORD(ugt); KEYWORD(ule); KEYWORD(uge);
554  KEYWORD(oeq); KEYWORD(one); KEYWORD(olt); KEYWORD(ogt); KEYWORD(ole);
555  KEYWORD(oge); KEYWORD(ord); KEYWORD(uno); KEYWORD(ueq); KEYWORD(une);
556
557  KEYWORD(x);
558#undef KEYWORD
559
560  // Keywords for types.
561#define TYPEKEYWORD(STR, LLVMTY) \
562  if (Len == strlen(STR) && !memcmp(StartChar, STR, strlen(STR))) { \
563    TyVal = LLVMTY; return lltok::Type; }
564  TYPEKEYWORD("void",      Type::VoidTy);
565  TYPEKEYWORD("float",     Type::FloatTy);
566  TYPEKEYWORD("double",    Type::DoubleTy);
567  TYPEKEYWORD("x86_fp80",  Type::X86_FP80Ty);
568  TYPEKEYWORD("fp128",     Type::FP128Ty);
569  TYPEKEYWORD("ppc_fp128", Type::PPC_FP128Ty);
570  TYPEKEYWORD("label",     Type::LabelTy);
571#undef TYPEKEYWORD
572
573  // Handle special forms for autoupgrading.  Drop these in LLVM 3.0.  This is
574  // to avoid conflicting with the sext/zext instructions, below.
575  if (Len == 4 && !memcmp(StartChar, "sext", 4)) {
576    // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
577    if (JustWhitespaceNewLine(CurPtr))
578      return lltok::kw_signext;
579  } else if (Len == 4 && !memcmp(StartChar, "zext", 4)) {
580    // Scan CurPtr ahead, seeing if there is just whitespace before the newline.
581    if (JustWhitespaceNewLine(CurPtr))
582      return lltok::kw_zeroext;
583  }
584
585  // Keywords for instructions.
586#define INSTKEYWORD(STR, Enum) \
587  if (Len == strlen(#STR) && !memcmp(StartChar, #STR, strlen(#STR))) { \
588    UIntVal = Instruction::Enum; return lltok::kw_##STR; }
589
590  INSTKEYWORD(add,   Add);  INSTKEYWORD(sub,   Sub);  INSTKEYWORD(mul,   Mul);
591  INSTKEYWORD(udiv,  UDiv); INSTKEYWORD(sdiv,  SDiv); INSTKEYWORD(fdiv,  FDiv);
592  INSTKEYWORD(urem,  URem); INSTKEYWORD(srem,  SRem); INSTKEYWORD(frem,  FRem);
593  INSTKEYWORD(shl,   Shl);  INSTKEYWORD(lshr,  LShr); INSTKEYWORD(ashr,  AShr);
594  INSTKEYWORD(and,   And);  INSTKEYWORD(or,    Or);   INSTKEYWORD(xor,   Xor);
595  INSTKEYWORD(icmp,  ICmp); INSTKEYWORD(fcmp,  FCmp);
596  INSTKEYWORD(vicmp, VICmp); INSTKEYWORD(vfcmp, VFCmp);
597
598  INSTKEYWORD(phi,         PHI);
599  INSTKEYWORD(call,        Call);
600  INSTKEYWORD(trunc,       Trunc);
601  INSTKEYWORD(zext,        ZExt);
602  INSTKEYWORD(sext,        SExt);
603  INSTKEYWORD(fptrunc,     FPTrunc);
604  INSTKEYWORD(fpext,       FPExt);
605  INSTKEYWORD(uitofp,      UIToFP);
606  INSTKEYWORD(sitofp,      SIToFP);
607  INSTKEYWORD(fptoui,      FPToUI);
608  INSTKEYWORD(fptosi,      FPToSI);
609  INSTKEYWORD(inttoptr,    IntToPtr);
610  INSTKEYWORD(ptrtoint,    PtrToInt);
611  INSTKEYWORD(bitcast,     BitCast);
612  INSTKEYWORD(select,      Select);
613  INSTKEYWORD(va_arg,      VAArg);
614  INSTKEYWORD(ret,         Ret);
615  INSTKEYWORD(br,          Br);
616  INSTKEYWORD(switch,      Switch);
617  INSTKEYWORD(invoke,      Invoke);
618  INSTKEYWORD(unwind,      Unwind);
619  INSTKEYWORD(unreachable, Unreachable);
620
621  INSTKEYWORD(malloc,      Malloc);
622  INSTKEYWORD(alloca,      Alloca);
623  INSTKEYWORD(free,        Free);
624  INSTKEYWORD(load,        Load);
625  INSTKEYWORD(store,       Store);
626  INSTKEYWORD(getelementptr, GetElementPtr);
627
628  INSTKEYWORD(extractelement, ExtractElement);
629  INSTKEYWORD(insertelement,  InsertElement);
630  INSTKEYWORD(shufflevector,  ShuffleVector);
631  INSTKEYWORD(getresult,      ExtractValue);
632  INSTKEYWORD(extractvalue,   ExtractValue);
633  INSTKEYWORD(insertvalue,    InsertValue);
634#undef INSTKEYWORD
635
636  // Check for [us]0x[0-9A-Fa-f]+ which are Hexadecimal constant generated by
637  // the CFE to avoid forcing it to deal with 64-bit numbers.
638  if ((TokStart[0] == 'u' || TokStart[0] == 's') &&
639      TokStart[1] == '0' && TokStart[2] == 'x' && isxdigit(TokStart[3])) {
640    int len = CurPtr-TokStart-3;
641    uint32_t bits = len * 4;
642    APInt Tmp(bits, TokStart+3, len, 16);
643    uint32_t activeBits = Tmp.getActiveBits();
644    if (activeBits > 0 && activeBits < bits)
645      Tmp.trunc(activeBits);
646    APSIntVal = APSInt(Tmp, TokStart[0] == 'u');
647    return lltok::APSInt;
648  }
649
650  // If this is "cc1234", return this as just "cc".
651  if (TokStart[0] == 'c' && TokStart[1] == 'c') {
652    CurPtr = TokStart+2;
653    return lltok::kw_cc;
654  }
655
656  // If this starts with "call", return it as CALL.  This is to support old
657  // broken .ll files.  FIXME: remove this with LLVM 3.0.
658  if (CurPtr-TokStart > 4 && !memcmp(TokStart, "call", 4)) {
659    CurPtr = TokStart+4;
660    UIntVal = Instruction::Call;
661    return lltok::kw_call;
662  }
663
664  // Finally, if this isn't known, return an error.
665  CurPtr = TokStart+1;
666  return lltok::Error;
667}
668
669
670/// Lex0x: Handle productions that start with 0x, knowing that it matches and
671/// that this is not a label:
672///    HexFPConstant     0x[0-9A-Fa-f]+
673///    HexFP80Constant   0xK[0-9A-Fa-f]+
674///    HexFP128Constant  0xL[0-9A-Fa-f]+
675///    HexPPC128Constant 0xM[0-9A-Fa-f]+
676lltok::Kind LLLexer::Lex0x() {
677  CurPtr = TokStart + 2;
678
679  char Kind;
680  if (CurPtr[0] >= 'K' && CurPtr[0] <= 'M') {
681    Kind = *CurPtr++;
682  } else {
683    Kind = 'J';
684  }
685
686  if (!isxdigit(CurPtr[0])) {
687    // Bad token, return it as an error.
688    CurPtr = TokStart+1;
689    return lltok::Error;
690  }
691
692  while (isxdigit(CurPtr[0]))
693    ++CurPtr;
694
695  if (Kind == 'J') {
696    // HexFPConstant - Floating point constant represented in IEEE format as a
697    // hexadecimal number for when exponential notation is not precise enough.
698    // Float and double only.
699    APFloatVal = APFloat(BitsToDouble(HexIntToVal(TokStart+2, CurPtr)));
700    return lltok::APFloat;
701  }
702
703  uint64_t Pair[2];
704  switch (Kind) {
705  default: assert(0 && "Unknown kind!");
706  case 'K':
707    // F80HexFPConstant - x87 long double in hexadecimal format (10 bytes)
708    FP80HexToIntPair(TokStart+3, CurPtr, Pair);
709    APFloatVal = APFloat(APInt(80, 2, Pair));
710    return lltok::APFloat;
711  case 'L':
712    // F128HexFPConstant - IEEE 128-bit in hexadecimal format (16 bytes)
713    HexToIntPair(TokStart+3, CurPtr, Pair);
714    APFloatVal = APFloat(APInt(128, 2, Pair), true);
715    return lltok::APFloat;
716  case 'M':
717    // PPC128HexFPConstant - PowerPC 128-bit in hexadecimal format (16 bytes)
718    HexToIntPair(TokStart+3, CurPtr, Pair);
719    APFloatVal = APFloat(APInt(128, 2, Pair));
720    return lltok::APFloat;
721  }
722}
723
724/// LexIdentifier: Handle several related productions:
725///    Label             [-a-zA-Z$._0-9]+:
726///    NInteger          -[0-9]+
727///    FPConstant        [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
728///    PInteger          [0-9]+
729///    HexFPConstant     0x[0-9A-Fa-f]+
730///    HexFP80Constant   0xK[0-9A-Fa-f]+
731///    HexFP128Constant  0xL[0-9A-Fa-f]+
732///    HexPPC128Constant 0xM[0-9A-Fa-f]+
733lltok::Kind LLLexer::LexDigitOrNegative() {
734  // If the letter after the negative is a number, this is probably a label.
735  if (!isdigit(TokStart[0]) && !isdigit(CurPtr[0])) {
736    // Okay, this is not a number after the -, it's probably a label.
737    if (const char *End = isLabelTail(CurPtr)) {
738      StrVal.assign(TokStart, End-1);
739      CurPtr = End;
740      return lltok::LabelStr;
741    }
742
743    return lltok::Error;
744  }
745
746  // At this point, it is either a label, int or fp constant.
747
748  // Skip digits, we have at least one.
749  for (; isdigit(CurPtr[0]); ++CurPtr)
750    /*empty*/;
751
752  // Check to see if this really is a label afterall, e.g. "-1:".
753  if (isLabelChar(CurPtr[0]) || CurPtr[0] == ':') {
754    if (const char *End = isLabelTail(CurPtr)) {
755      StrVal.assign(TokStart, End-1);
756      CurPtr = End;
757      return lltok::LabelStr;
758    }
759  }
760
761  // If the next character is a '.', then it is a fp value, otherwise its
762  // integer.
763  if (CurPtr[0] != '.') {
764    if (TokStart[0] == '0' && TokStart[1] == 'x')
765      return Lex0x();
766    unsigned Len = CurPtr-TokStart;
767    uint32_t numBits = ((Len * 64) / 19) + 2;
768    APInt Tmp(numBits, TokStart, Len, 10);
769    if (TokStart[0] == '-') {
770      uint32_t minBits = Tmp.getMinSignedBits();
771      if (minBits > 0 && minBits < numBits)
772        Tmp.trunc(minBits);
773      APSIntVal = APSInt(Tmp, false);
774    } else {
775      uint32_t activeBits = Tmp.getActiveBits();
776      if (activeBits > 0 && activeBits < numBits)
777        Tmp.trunc(activeBits);
778      APSIntVal = APSInt(Tmp, true);
779    }
780    return lltok::APSInt;
781  }
782
783  ++CurPtr;
784
785  // Skip over [0-9]*([eE][-+]?[0-9]+)?
786  while (isdigit(CurPtr[0])) ++CurPtr;
787
788  if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
789    if (isdigit(CurPtr[1]) ||
790        ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
791      CurPtr += 2;
792      while (isdigit(CurPtr[0])) ++CurPtr;
793    }
794  }
795
796  APFloatVal = APFloat(atof(TokStart));
797  return lltok::APFloat;
798}
799
800///    FPConstant  [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
801lltok::Kind LLLexer::LexPositive() {
802  // If the letter after the negative is a number, this is probably not a
803  // label.
804  if (!isdigit(CurPtr[0]))
805    return lltok::Error;
806
807  // Skip digits.
808  for (++CurPtr; isdigit(CurPtr[0]); ++CurPtr)
809    /*empty*/;
810
811  // At this point, we need a '.'.
812  if (CurPtr[0] != '.') {
813    CurPtr = TokStart+1;
814    return lltok::Error;
815  }
816
817  ++CurPtr;
818
819  // Skip over [0-9]*([eE][-+]?[0-9]+)?
820  while (isdigit(CurPtr[0])) ++CurPtr;
821
822  if (CurPtr[0] == 'e' || CurPtr[0] == 'E') {
823    if (isdigit(CurPtr[1]) ||
824        ((CurPtr[1] == '-' || CurPtr[1] == '+') && isdigit(CurPtr[2]))) {
825      CurPtr += 2;
826      while (isdigit(CurPtr[0])) ++CurPtr;
827    }
828  }
829
830  APFloatVal = APFloat(atof(TokStart));
831  return lltok::APFloat;
832}
833