1 /*===-- Lexer.l - Scanner for llvm assembly files --------------*- C++ -*--===//
3 // The LLVM Compiler Infrastructure
5 // This file was developed by the LLVM research group and is distributed under
6 // the University of Illinois Open Source License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the flex scanner for LLVM assembly languages files.
12 //===----------------------------------------------------------------------===*/
14 %option prefix="llvmAsm"
17 %option never-interactive
22 %option outfile="Lexer.cpp"
28 #include "ParserInternals.h"
29 #include "llvm/Module.h"
31 #include "llvmAsmParser.h"
35 void set_scan_file(FILE * F){
36 yy_switch_to_buffer(yy_create_buffer( F, YY_BUF_SIZE ) );
38 void set_scan_string (const char * str) {
42 // Construct a token value for a non-obsolete token
43 #define RET_TOK(type, Enum, sym) \
44 llvmAsmlval.type = Instruction::Enum; \
47 // Construct a token value for an obsolete token
48 #define RET_TY(CTYPE, SYM) \
49 llvmAsmlval.PrimType = CTYPE;\
54 // TODO: All of the static identifiers are figured out by the lexer,
55 // these should be hashed to reduce the lexer size
58 // atoull - Convert an ascii string of decimal digits into the unsigned long
59 // long representation... this does not have to do input error checking,
60 // because we know that the input will be matched by a suitable regex...
62 static uint64_t atoull(const char *Buffer) {
64 for (; *Buffer; Buffer++) {
65 uint64_t OldRes = Result;
67 Result += *Buffer-'0';
68 if (Result < OldRes) // Uh, oh, overflow detected!!!
69 GenerateError("constant bigger than 64 bits detected!");
74 static uint64_t HexIntToVal(const char *Buffer) {
76 for (; *Buffer; ++Buffer) {
77 uint64_t OldRes = Result;
80 if (C >= '0' && C <= '9')
82 else if (C >= 'A' && C <= 'F')
84 else if (C >= 'a' && C <= 'f')
87 if (Result < OldRes) // Uh, oh, overflow detected!!!
88 GenerateError("constant bigger than 64 bits detected!");
94 // HexToFP - Convert the ascii string in hexidecimal format to the floating
95 // point representation of it.
97 static double HexToFP(const char *Buffer) {
98 // Behave nicely in the face of C TBAA rules... see:
99 // http://www.nullstone.com/htmls/category/aliastyp.htm
104 UIntToFP.UI = HexIntToVal(Buffer);
106 assert(sizeof(double) == sizeof(uint64_t) &&
107 "Data sizes incompatible on this target!");
108 return UIntToFP.FP; // Cast Hex constant to double
112 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
113 // appropriate character. If AllowNull is set to false, a \00 value will cause
114 // an exception to be thrown.
116 // If AllowNull is set to true, the return value of the function points to the
117 // last character of the string in memory.
119 char *UnEscapeLexed(char *Buffer, bool AllowNull) {
121 for (char *BIn = Buffer; *BIn; ) {
122 if (BIn[0] == '\\' && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
123 char Tmp = BIn[3]; BIn[3] = 0; // Terminate string
124 *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number
125 if (!AllowNull && !*BOut)
126 GenerateError("String literal cannot accept \\00 escape!");
128 BIn[3] = Tmp; // Restore character
129 BIn += 3; // Skip over handled chars
139 } // End llvm namespace
141 using namespace llvm;
143 #define YY_NEVER_INTERACTIVE 1
148 /* Comments start with a ; and go till end of line */
151 /* Local Values and Type identifiers start with a % sign */
152 LocalVarName %[-a-zA-Z$._][-a-zA-Z$._0-9]*
154 /* Global Value identifiers start with an @ sign */
155 GlobalVarName @[-a-zA-Z$._][-a-zA-Z$._0-9]*
157 /* Label identifiers end with a colon */
158 Label [-a-zA-Z$._0-9]+:
159 QuoteLabel \"[^\"]+\":
161 /* Quoted names can contain any character except " and \ */
162 StringConstant \"[^\"]*\"
163 AtStringConstant @\"[^\"]*\"
165 /* LocalVarID/GlobalVarID: match an unnamed local variable slot ID. */
169 /* Integer types are specified with i and a bitwidth */
172 /* E[PN]Integer: match positive and negative literal integer values. */
176 /* FPConstant - A Floating point constant.
178 FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
180 /* HexFPConstant - Floating point constant represented in IEEE format as a
181 * hexadecimal number for when exponential notation is not precise enough.
183 HexFPConstant 0x[0-9A-Fa-f]+
185 /* HexIntConstant - Hexadecimal constant generated by the CFE to avoid forcing
186 * it to deal with 64 bit numbers.
188 HexIntConstant [us]0x[0-9A-Fa-f]+
192 {Comment} { /* Ignore comments for now */ }
194 begin { return BEGINTOK; }
195 end { return ENDTOK; }
196 true { return TRUETOK; }
197 false { return FALSETOK; }
198 declare { return DECLARE; }
199 define { return DEFINE; }
200 global { return GLOBAL; }
201 constant { return CONSTANT; }
202 internal { return INTERNAL; }
203 linkonce { return LINKONCE; }
204 weak { return WEAK; }
205 appending { return APPENDING; }
206 dllimport { return DLLIMPORT; }
207 dllexport { return DLLEXPORT; }
208 hidden { return HIDDEN; }
209 protected { return PROTECTED; }
210 extern_weak { return EXTERN_WEAK; }
211 external { return EXTERNAL; }
212 thread_local { return THREAD_LOCAL; }
213 zeroinitializer { return ZEROINITIALIZER; }
214 \.\.\. { return DOTDOTDOT; }
215 undef { return UNDEF; }
216 null { return NULL_TOK; }
218 tail { return TAIL; }
219 target { return TARGET; }
220 triple { return TRIPLE; }
221 deplibs { return DEPLIBS; }
222 datalayout { return DATALAYOUT; }
223 volatile { return VOLATILE; }
224 align { return ALIGN; }
225 section { return SECTION; }
226 alias { return ALIAS; }
227 module { return MODULE; }
228 asm { return ASM_TOK; }
229 sideeffect { return SIDEEFFECT; }
231 cc { return CC_TOK; }
232 ccc { return CCC_TOK; }
233 fastcc { return FASTCC_TOK; }
234 coldcc { return COLDCC_TOK; }
235 x86_stdcallcc { return X86_STDCALLCC_TOK; }
236 x86_fastcallcc { return X86_FASTCALLCC_TOK; }
238 inreg { return INREG; }
239 sret { return SRET; }
240 nounwind { return NOUNWIND; }
241 noreturn { return NORETURN; }
243 void { RET_TY(Type::VoidTy, VOID); }
244 float { RET_TY(Type::FloatTy, FLOAT); }
245 double { RET_TY(Type::DoubleTy,DOUBLE);}
246 label { RET_TY(Type::LabelTy, LABEL); }
247 type { return TYPE; }
248 opaque { return OPAQUE; }
249 {IntegerType} { uint64_t NumBits = atoull(yytext+1);
250 if (NumBits < IntegerType::MIN_INT_BITS ||
251 NumBits > IntegerType::MAX_INT_BITS)
252 GenerateError("Bitwidth for integer type out of range!");
253 const Type* Ty = IntegerType::get(NumBits);
257 add { RET_TOK(BinaryOpVal, Add, ADD); }
258 sub { RET_TOK(BinaryOpVal, Sub, SUB); }
259 mul { RET_TOK(BinaryOpVal, Mul, MUL); }
260 udiv { RET_TOK(BinaryOpVal, UDiv, UDIV); }
261 sdiv { RET_TOK(BinaryOpVal, SDiv, SDIV); }
262 fdiv { RET_TOK(BinaryOpVal, FDiv, FDIV); }
263 urem { RET_TOK(BinaryOpVal, URem, UREM); }
264 srem { RET_TOK(BinaryOpVal, SRem, SREM); }
265 frem { RET_TOK(BinaryOpVal, FRem, FREM); }
266 shl { RET_TOK(BinaryOpVal, Shl, SHL); }
267 lshr { RET_TOK(BinaryOpVal, LShr, LSHR); }
268 ashr { RET_TOK(BinaryOpVal, AShr, ASHR); }
269 and { RET_TOK(BinaryOpVal, And, AND); }
270 or { RET_TOK(BinaryOpVal, Or , OR ); }
271 xor { RET_TOK(BinaryOpVal, Xor, XOR); }
272 icmp { RET_TOK(OtherOpVal, ICmp, ICMP); }
273 fcmp { RET_TOK(OtherOpVal, FCmp, FCMP); }
296 phi { RET_TOK(OtherOpVal, PHI, PHI_TOK); }
297 call { RET_TOK(OtherOpVal, Call, CALL); }
298 trunc { RET_TOK(CastOpVal, Trunc, TRUNC); }
299 zext { RET_TOK(CastOpVal, ZExt, ZEXT); }
300 sext { RET_TOK(CastOpVal, SExt, SEXT); }
301 fptrunc { RET_TOK(CastOpVal, FPTrunc, FPTRUNC); }
302 fpext { RET_TOK(CastOpVal, FPExt, FPEXT); }
303 uitofp { RET_TOK(CastOpVal, UIToFP, UITOFP); }
304 sitofp { RET_TOK(CastOpVal, SIToFP, SITOFP); }
305 fptoui { RET_TOK(CastOpVal, FPToUI, FPTOUI); }
306 fptosi { RET_TOK(CastOpVal, FPToSI, FPTOSI); }
307 inttoptr { RET_TOK(CastOpVal, IntToPtr, INTTOPTR); }
308 ptrtoint { RET_TOK(CastOpVal, PtrToInt, PTRTOINT); }
309 bitcast { RET_TOK(CastOpVal, BitCast, BITCAST); }
310 select { RET_TOK(OtherOpVal, Select, SELECT); }
311 va_arg { RET_TOK(OtherOpVal, VAArg , VAARG); }
312 ret { RET_TOK(TermOpVal, Ret, RET); }
313 br { RET_TOK(TermOpVal, Br, BR); }
314 switch { RET_TOK(TermOpVal, Switch, SWITCH); }
315 invoke { RET_TOK(TermOpVal, Invoke, INVOKE); }
316 unwind { RET_TOK(TermOpVal, Unwind, UNWIND); }
317 unreachable { RET_TOK(TermOpVal, Unreachable, UNREACHABLE); }
319 malloc { RET_TOK(MemOpVal, Malloc, MALLOC); }
320 alloca { RET_TOK(MemOpVal, Alloca, ALLOCA); }
321 free { RET_TOK(MemOpVal, Free, FREE); }
322 load { RET_TOK(MemOpVal, Load, LOAD); }
323 store { RET_TOK(MemOpVal, Store, STORE); }
324 getelementptr { RET_TOK(MemOpVal, GetElementPtr, GETELEMENTPTR); }
326 extractelement { RET_TOK(OtherOpVal, ExtractElement, EXTRACTELEMENT); }
327 insertelement { RET_TOK(OtherOpVal, InsertElement, INSERTELEMENT); }
328 shufflevector { RET_TOK(OtherOpVal, ShuffleVector, SHUFFLEVECTOR); }
332 UnEscapeLexed(yytext+1);
333 llvmAsmlval.StrVal = strdup(yytext+1); // Skip %
337 UnEscapeLexed(yytext+1);
338 llvmAsmlval.StrVal = strdup(yytext+1); // Skip @
342 yytext[strlen(yytext)-1] = 0; // nuke colon
343 UnEscapeLexed(yytext);
344 llvmAsmlval.StrVal = strdup(yytext);
348 yytext[strlen(yytext)-2] = 0; // nuke colon, end quote
349 UnEscapeLexed(yytext+1);
350 llvmAsmlval.StrVal = strdup(yytext+1);
354 {StringConstant} { // Note that we cannot unescape a string constant here! The
355 // string constant might contain a \00 which would not be
356 // understood by the string stuff. It is valid to make a
357 // [sbyte] c"Hello World\00" constant, for example.
359 yytext[strlen(yytext)-1] = 0; // nuke end quote
360 llvmAsmlval.StrVal = strdup(yytext+1); // Nuke start quote
361 return STRINGCONSTANT;
364 yytext[strlen(yytext)-1] = 0; // nuke end quote
365 llvmAsmlval.StrVal = strdup(yytext+2); // Nuke @, quote
366 return ATSTRINGCONSTANT;
369 {PInteger} { int len = strlen(yytext);
370 uint32_t numBits = ((len * 64) / 19) + 1;
371 APInt Tmp(numBits, yytext, len, 10);
372 uint32_t activeBits = Tmp.getActiveBits();
373 if (activeBits > 0 && activeBits < numBits)
374 Tmp.trunc(activeBits);
375 if (Tmp.getBitWidth() > 64) {
376 llvmAsmlval.APIntVal = new APInt(Tmp);
379 llvmAsmlval.UInt64Val = Tmp.getZExtValue();
383 {NInteger} { int len = strlen(yytext);
384 uint32_t numBits = (((len-1) * 64) / 19) + 2;
385 APInt Tmp(numBits, yytext, len, 10);
386 uint32_t minBits = Tmp.getMinSignedBits();
387 if (minBits > 0 && minBits < numBits)
389 if (Tmp.getBitWidth() > 64) {
390 llvmAsmlval.APIntVal = new APInt(Tmp);
393 llvmAsmlval.SInt64Val = Tmp.getSExtValue();
398 {HexIntConstant} { int len = strlen(yytext+3) - 3;
399 uint32_t bits = len * 4;
400 APInt Tmp(bits, yytext+3, len, 16);
401 uint32_t activeBits = Tmp.getActiveBits();
402 if (activeBits > 0 && activeBits < bits)
403 Tmp.trunc(activeBits);
404 if (Tmp.getBitWidth() > 64) {
405 llvmAsmlval.APIntVal = new APInt(Tmp);
406 return yytext[0] == 's' ? ESAPINTVAL : EUAPINTVAL;
407 } else if (yytext[0] == 's') {
408 llvmAsmlval.SInt64Val = Tmp.getSExtValue();
411 llvmAsmlval.UInt64Val = Tmp.getZExtValue();
417 uint64_t Val = atoull(yytext+1);
418 if ((unsigned)Val != Val)
419 GenerateError("Invalid value number (too large)!");
420 llvmAsmlval.UIntVal = unsigned(Val);
424 uint64_t Val = atoull(yytext+1);
425 if ((unsigned)Val != Val)
426 GenerateError("Invalid value number (too large)!");
427 llvmAsmlval.UIntVal = unsigned(Val);
431 {FPConstant} { llvmAsmlval.FPVal = atof(yytext); return FPVAL; }
432 {HexFPConstant} { llvmAsmlval.FPVal = HexToFP(yytext); return FPVAL; }
435 /* Make sure to free the internal buffers for flex when we are
436 * done reading our input!
438 yy_delete_buffer(YY_CURRENT_BUFFER);
442 [ \r\t\n] { /* Ignore whitespace */ }
443 . { return yytext[0]; }