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"
23 #include "llvmAsmParser.h"
27 #define RET_TOK(type, Enum, sym) \
28 llvmAsmlval.type = Instruction::Enum; return sym
31 // TODO: All of the static identifiers are figured out by the lexer,
32 // these should be hashed to reduce the lexer size
35 // atoull - Convert an ascii string of decimal digits into the unsigned long
36 // long representation... this does not have to do input error checking,
37 // because we know that the input will be matched by a suitable regex...
39 static uint64_t atoull(const char *Buffer) {
41 for (; *Buffer; Buffer++) {
42 uint64_t OldRes = Result;
44 Result += *Buffer-'0';
45 if (Result < OldRes) // Uh, oh, overflow detected!!!
46 ThrowException("constant bigger than 64 bits detected!");
51 // HexToFP - Convert the ascii string in hexidecimal format to the floating
52 // point representation of it.
54 static double HexToFP(const char *Buffer) {
56 for (; *Buffer; ++Buffer) {
57 uint64_t OldRes = Result;
60 if (C >= '0' && C <= '9')
62 else if (C >= 'A' && C <= 'F')
64 else if (C >= 'a' && C <= 'f')
67 if (Result < OldRes) // Uh, oh, overflow detected!!!
68 ThrowException("constant bigger than 64 bits detected!");
71 assert(sizeof(double) == sizeof(Result) &&
72 "Data sizes incompatible on this target!");
73 void *ProxyPointer = &Result; // Break TBAA correctly
74 cerr << "VALUE: " << *(double*)ProxyPointer << "\n";
75 return *(double*)ProxyPointer; // Cast Hex constant to double
79 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
80 // appropriate character. If AllowNull is set to false, a \00 value will cause
81 // an exception to be thrown.
83 // If AllowNull is set to true, the return value of the function points to the
84 // last character of the string in memory.
86 char *UnEscapeLexed(char *Buffer, bool AllowNull = false) {
88 for (char *BIn = Buffer; *BIn; ) {
89 if (BIn[0] == '\\' && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
90 char Tmp = BIn[3]; BIn[3] = 0; // Terminate string
91 *BOut = strtol(BIn+1, 0, 16); // Convert to number
92 if (!AllowNull && !*BOut)
93 ThrowException("String literal cannot accept \\00 escape!");
95 BIn[3] = Tmp; // Restore character
96 BIn += 3; // Skip over handled chars
106 #define YY_NEVER_INTERACTIVE 1
111 /* Comments start with a ; and go till end of line */
114 /* Variable(Value) identifiers start with a % sign */
115 VarID %[-a-zA-Z$._][-a-zA-Z$._0-9]*
117 /* Label identifiers end with a colon */
118 Label [-a-zA-Z$._0-9]+:
120 /* Quoted names can contain any character except " and \ */
121 StringConstant \"[^\"]+\"
124 /* [PN]Integer: match positive and negative literal integer values that
125 * are preceeded by a '%' character. These represent unnamed variable slots.
131 /* E[PN]Integer: match positive and negative literal integer values */
135 /* FPConstant - A Floating point constant.
137 FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
139 /* HexFPConstant - Floating point constant represented in IEEE format as a
140 * hexadecimal number for when exponential notation is not precise enough.
142 HexFPConstant 0x[0-9A-Fa-f]+
145 {Comment} { /* Ignore comments for now */ }
147 begin { return BEGINTOK; }
149 true { return TRUE; }
150 false { return FALSE; }
151 declare { return DECLARE; }
152 global { return GLOBAL; }
153 constant { return CONSTANT; }
154 const { return CONST; }
155 internal { return INTERNAL; }
156 uninitialized { return UNINIT; }
157 implementation { return IMPLEMENTATION; }
158 \.\.\. { return DOTDOTDOT; }
159 string { return STRING; }
160 null { return NULL_TOK; }
162 except { return EXCEPT; }
164 void { llvmAsmlval.PrimType = Type::VoidTy ; return VOID; }
165 bool { llvmAsmlval.PrimType = Type::BoolTy ; return BOOL; }
166 sbyte { llvmAsmlval.PrimType = Type::SByteTy ; return SBYTE; }
167 ubyte { llvmAsmlval.PrimType = Type::UByteTy ; return UBYTE; }
168 short { llvmAsmlval.PrimType = Type::ShortTy ; return SHORT; }
169 ushort { llvmAsmlval.PrimType = Type::UShortTy; return USHORT; }
170 int { llvmAsmlval.PrimType = Type::IntTy ; return INT; }
171 uint { llvmAsmlval.PrimType = Type::UIntTy ; return UINT; }
172 long { llvmAsmlval.PrimType = Type::LongTy ; return LONG; }
173 ulong { llvmAsmlval.PrimType = Type::ULongTy ; return ULONG; }
174 float { llvmAsmlval.PrimType = Type::FloatTy ; return FLOAT; }
175 double { llvmAsmlval.PrimType = Type::DoubleTy; return DOUBLE; }
176 type { llvmAsmlval.PrimType = Type::TypeTy ; return TYPE; }
177 label { llvmAsmlval.PrimType = Type::LabelTy ; return LABEL; }
178 opaque { return OPAQUE; }
181 not { RET_TOK(UnaryOpVal, Not, NOT); }
183 add { RET_TOK(BinaryOpVal, Add, ADD); }
184 sub { RET_TOK(BinaryOpVal, Sub, SUB); }
185 mul { RET_TOK(BinaryOpVal, Mul, MUL); }
186 div { RET_TOK(BinaryOpVal, Div, DIV); }
187 rem { RET_TOK(BinaryOpVal, Rem, REM); }
188 and { RET_TOK(BinaryOpVal, And, AND); }
189 or { RET_TOK(BinaryOpVal, Or , OR ); }
190 xor { RET_TOK(BinaryOpVal, Xor, XOR); }
191 setne { RET_TOK(BinaryOpVal, SetNE, SETNE); }
192 seteq { RET_TOK(BinaryOpVal, SetEQ, SETEQ); }
193 setlt { RET_TOK(BinaryOpVal, SetLT, SETLT); }
194 setgt { RET_TOK(BinaryOpVal, SetGT, SETGT); }
195 setle { RET_TOK(BinaryOpVal, SetLE, SETLE); }
196 setge { RET_TOK(BinaryOpVal, SetGE, SETGE); }
198 phi { RET_TOK(OtherOpVal, PHINode, PHI); }
199 call { RET_TOK(OtherOpVal, Call, CALL); }
200 cast { RET_TOK(OtherOpVal, Cast, CAST); }
201 shl { RET_TOK(OtherOpVal, Shl, SHL); }
202 shr { RET_TOK(OtherOpVal, Shr, SHR); }
204 ret { RET_TOK(TermOpVal, Ret, RET); }
205 br { RET_TOK(TermOpVal, Br, BR); }
206 switch { RET_TOK(TermOpVal, Switch, SWITCH); }
207 invoke { RET_TOK(TermOpVal, Invoke, INVOKE); }
210 malloc { RET_TOK(MemOpVal, Malloc, MALLOC); }
211 alloca { RET_TOK(MemOpVal, Alloca, ALLOCA); }
212 free { RET_TOK(MemOpVal, Free, FREE); }
213 load { RET_TOK(MemOpVal, Load, LOAD); }
214 store { RET_TOK(MemOpVal, Store, STORE); }
215 getelementptr { RET_TOK(MemOpVal, GetElementPtr, GETELEMENTPTR); }
219 UnEscapeLexed(yytext+1);
220 llvmAsmlval.StrVal = strdup(yytext+1); // Skip %
224 yytext[strlen(yytext)-1] = 0; // nuke colon
225 UnEscapeLexed(yytext);
226 llvmAsmlval.StrVal = strdup(yytext);
230 {StringConstant} { // Note that we cannot unescape a string constant here! The
231 // string constant might contain a \00 which would not be
232 // understood by the string stuff. It is valid to make a
233 // [sbyte] c"Hello World\00" constant, for example.
235 yytext[strlen(yytext)-1] = 0; // nuke end quote
236 llvmAsmlval.StrVal = strdup(yytext+1); // Nuke start quote
237 return STRINGCONSTANT;
241 {PInteger} { llvmAsmlval.UInt64Val = atoull(yytext); return EUINT64VAL; }
243 uint64_t Val = atoull(yytext+1);
244 // +1: we have bigger negative range
245 if (Val > (uint64_t)INT64_MAX+1)
246 ThrowException("Constant too large for signed 64 bits!");
247 llvmAsmlval.SInt64Val = -Val;
252 {EPInteger} { llvmAsmlval.UIntVal = atoull(yytext+1); return UINTVAL; }
254 uint64_t Val = atoull(yytext+2);
255 // +1: we have bigger negative range
256 if (Val > (uint64_t)INT32_MAX+1)
257 ThrowException("Constant too large for signed 32 bits!");
258 llvmAsmlval.SIntVal = -Val;
262 {FPConstant} { llvmAsmlval.FPVal = atof(yytext); return FPVAL; }
263 {HexFPConstant} { llvmAsmlval.FPVal = HexToFP(yytext); return FPVAL; }
265 [ \t\n] { /* Ignore whitespace */ }
266 . { return yytext[0]; }