1 /*===-- Lexer.l - Scanner for llvm assembly files --------------*- C++ -*--===//
3 // This file implements the flex scanner for LLVM assembly languages files.
5 //===----------------------------------------------------------------------===*/
7 %option prefix="llvmAsm"
10 %option never-interactive
15 %option outfile="Lexer.cpp"
21 #include "ParserInternals.h"
22 #include "llvm/Module.h"
24 #include "llvmAsmParser.h"
28 #define RET_TOK(type, Enum, sym) \
29 llvmAsmlval.type = Instruction::Enum; return sym
32 // TODO: All of the static identifiers are figured out by the lexer,
33 // these should be hashed to reduce the lexer size
36 // atoull - Convert an ascii string of decimal digits into the unsigned long
37 // long representation... this does not have to do input error checking,
38 // because we know that the input will be matched by a suitable regex...
40 static uint64_t atoull(const char *Buffer) {
42 for (; *Buffer; Buffer++) {
43 uint64_t OldRes = Result;
45 Result += *Buffer-'0';
46 if (Result < OldRes) // Uh, oh, overflow detected!!!
47 ThrowException("constant bigger than 64 bits detected!");
52 static uint64_t HexIntToVal(const char *Buffer) {
54 for (; *Buffer; ++Buffer) {
55 uint64_t OldRes = Result;
58 if (C >= '0' && C <= '9')
60 else if (C >= 'A' && C <= 'F')
62 else if (C >= 'a' && C <= 'f')
65 if (Result < OldRes) // Uh, oh, overflow detected!!!
66 ThrowException("constant bigger than 64 bits detected!");
72 // HexToFP - Convert the ascii string in hexidecimal format to the floating
73 // point representation of it.
75 static double HexToFP(const char *Buffer) {
76 // Behave nicely in the face of C TBAA rules... see:
77 // http://www.nullstone.com/htmls/category/aliastyp.htm
82 UIntToFP.UI = HexIntToVal(Buffer);
84 assert(sizeof(double) == sizeof(uint64_t) &&
85 "Data sizes incompatible on this target!");
86 return UIntToFP.FP; // Cast Hex constant to double
90 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
91 // appropriate character. If AllowNull is set to false, a \00 value will cause
92 // an exception to be thrown.
94 // If AllowNull is set to true, the return value of the function points to the
95 // last character of the string in memory.
97 char *UnEscapeLexed(char *Buffer, bool AllowNull) {
99 for (char *BIn = Buffer; *BIn; ) {
100 if (BIn[0] == '\\' && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
101 char Tmp = BIn[3]; BIn[3] = 0; // Terminate string
102 *BOut = strtol(BIn+1, 0, 16); // Convert to number
103 if (!AllowNull && !*BOut)
104 ThrowException("String literal cannot accept \\00 escape!");
106 BIn[3] = Tmp; // Restore character
107 BIn += 3; // Skip over handled chars
117 #define YY_NEVER_INTERACTIVE 1
122 /* Comments start with a ; and go till end of line */
125 /* Variable(Value) identifiers start with a % sign */
126 VarID %[-a-zA-Z$._][-a-zA-Z$._0-9]*
128 /* Label identifiers end with a colon */
129 Label [-a-zA-Z$._0-9]+:
131 /* Quoted names can contain any character except " and \ */
132 StringConstant \"[^\"]+\"
135 /* [PN]Integer: match positive and negative literal integer values that
136 * are preceeded by a '%' character. These represent unnamed variable slots.
142 /* E[PN]Integer: match positive and negative literal integer values */
146 /* FPConstant - A Floating point constant.
148 FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
150 /* HexFPConstant - Floating point constant represented in IEEE format as a
151 * hexadecimal number for when exponential notation is not precise enough.
153 HexFPConstant 0x[0-9A-Fa-f]+
155 /* HexIntConstant - Hexadecimal constant generated by the CFE to avoid forcing
156 * it to deal with 64 bit numbers.
158 HexIntConstant [us]0x[0-9A-Fa-f]+
161 {Comment} { /* Ignore comments for now */ }
163 begin { return BEGINTOK; }
164 end { return ENDTOK; }
165 true { return TRUE; }
166 false { return FALSE; }
167 declare { return DECLARE; }
168 global { return GLOBAL; }
169 constant { return CONSTANT; }
170 const { return CONST; }
171 internal { return INTERNAL; }
172 linkonce { return LINKONCE; }
173 appending { return APPENDING; }
174 uninitialized { return EXTERNAL; } /* Deprecated, turn into external */
175 external { return EXTERNAL; }
176 implementation { return IMPLEMENTATION; }
177 \.\.\. { return DOTDOTDOT; }
178 null { return NULL_TOK; }
180 except { return EXCEPT; }
181 not { return NOT; } /* Deprecated, turned into XOR */
182 target { return TARGET; }
183 endian { return ENDIAN; }
184 pointersize { return POINTERSIZE; }
185 little { return LITTLE; }
188 void { llvmAsmlval.PrimType = Type::VoidTy ; return VOID; }
189 bool { llvmAsmlval.PrimType = Type::BoolTy ; return BOOL; }
190 sbyte { llvmAsmlval.PrimType = Type::SByteTy ; return SBYTE; }
191 ubyte { llvmAsmlval.PrimType = Type::UByteTy ; return UBYTE; }
192 short { llvmAsmlval.PrimType = Type::ShortTy ; return SHORT; }
193 ushort { llvmAsmlval.PrimType = Type::UShortTy; return USHORT; }
194 int { llvmAsmlval.PrimType = Type::IntTy ; return INT; }
195 uint { llvmAsmlval.PrimType = Type::UIntTy ; return UINT; }
196 long { llvmAsmlval.PrimType = Type::LongTy ; return LONG; }
197 ulong { llvmAsmlval.PrimType = Type::ULongTy ; return ULONG; }
198 float { llvmAsmlval.PrimType = Type::FloatTy ; return FLOAT; }
199 double { llvmAsmlval.PrimType = Type::DoubleTy; return DOUBLE; }
200 type { llvmAsmlval.PrimType = Type::TypeTy ; return TYPE; }
201 label { llvmAsmlval.PrimType = Type::LabelTy ; return LABEL; }
202 opaque { return OPAQUE; }
204 add { RET_TOK(BinaryOpVal, Add, ADD); }
205 sub { RET_TOK(BinaryOpVal, Sub, SUB); }
206 mul { RET_TOK(BinaryOpVal, Mul, MUL); }
207 div { RET_TOK(BinaryOpVal, Div, DIV); }
208 rem { RET_TOK(BinaryOpVal, Rem, REM); }
209 and { RET_TOK(BinaryOpVal, And, AND); }
210 or { RET_TOK(BinaryOpVal, Or , OR ); }
211 xor { RET_TOK(BinaryOpVal, Xor, XOR); }
212 setne { RET_TOK(BinaryOpVal, SetNE, SETNE); }
213 seteq { RET_TOK(BinaryOpVal, SetEQ, SETEQ); }
214 setlt { RET_TOK(BinaryOpVal, SetLT, SETLT); }
215 setgt { RET_TOK(BinaryOpVal, SetGT, SETGT); }
216 setle { RET_TOK(BinaryOpVal, SetLE, SETLE); }
217 setge { RET_TOK(BinaryOpVal, SetGE, SETGE); }
219 phi { RET_TOK(OtherOpVal, PHINode, PHI); }
220 call { RET_TOK(OtherOpVal, Call, CALL); }
221 cast { RET_TOK(OtherOpVal, Cast, CAST); }
222 shl { RET_TOK(OtherOpVal, Shl, SHL); }
223 shr { RET_TOK(OtherOpVal, Shr, SHR); }
224 va_arg { RET_TOK(OtherOpVal, VarArg, VA_ARG); }
226 ret { RET_TOK(TermOpVal, Ret, RET); }
227 br { RET_TOK(TermOpVal, Br, BR); }
228 switch { RET_TOK(TermOpVal, Switch, SWITCH); }
229 invoke { RET_TOK(TermOpVal, Invoke, INVOKE); }
232 malloc { RET_TOK(MemOpVal, Malloc, MALLOC); }
233 alloca { RET_TOK(MemOpVal, Alloca, ALLOCA); }
234 free { RET_TOK(MemOpVal, Free, FREE); }
235 load { RET_TOK(MemOpVal, Load, LOAD); }
236 store { RET_TOK(MemOpVal, Store, STORE); }
237 getelementptr { RET_TOK(MemOpVal, GetElementPtr, GETELEMENTPTR); }
241 UnEscapeLexed(yytext+1);
242 llvmAsmlval.StrVal = strdup(yytext+1); // Skip %
246 yytext[strlen(yytext)-1] = 0; // nuke colon
247 UnEscapeLexed(yytext);
248 llvmAsmlval.StrVal = strdup(yytext);
252 {StringConstant} { // Note that we cannot unescape a string constant here! The
253 // string constant might contain a \00 which would not be
254 // understood by the string stuff. It is valid to make a
255 // [sbyte] c"Hello World\00" constant, for example.
257 yytext[strlen(yytext)-1] = 0; // nuke end quote
258 llvmAsmlval.StrVal = strdup(yytext+1); // Nuke start quote
259 return STRINGCONSTANT;
263 {PInteger} { llvmAsmlval.UInt64Val = atoull(yytext); return EUINT64VAL; }
265 uint64_t Val = atoull(yytext+1);
266 // +1: we have bigger negative range
267 if (Val > (uint64_t)INT64_MAX+1)
268 ThrowException("Constant too large for signed 64 bits!");
269 llvmAsmlval.SInt64Val = -Val;
273 llvmAsmlval.UInt64Val = HexIntToVal(yytext+3);
274 return yytext[0] == 's' ? ESINT64VAL : EUINT64VAL;
277 {EPInteger} { llvmAsmlval.UIntVal = atoull(yytext+1); return UINTVAL; }
279 uint64_t Val = atoull(yytext+2);
280 // +1: we have bigger negative range
281 if (Val > (uint64_t)INT32_MAX+1)
282 ThrowException("Constant too large for signed 32 bits!");
283 llvmAsmlval.SIntVal = -Val;
287 {FPConstant} { llvmAsmlval.FPVal = atof(yytext); return FPVAL; }
288 {HexFPConstant} { llvmAsmlval.FPVal = HexToFP(yytext); return FPVAL; }
290 [ \t\n] { /* Ignore whitespace */ }
291 . { return yytext[0]; }