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 // Behave nicely in the face of C TBAA rules... see:
74 // http://www.nullstone.com/htmls/category/aliastyp.htm
76 char *ProxyPointer = (char*)&Result;
77 return *(double*)ProxyPointer; // Cast Hex constant to double
81 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
82 // appropriate character. If AllowNull is set to false, a \00 value will cause
83 // an exception to be thrown.
85 // If AllowNull is set to true, the return value of the function points to the
86 // last character of the string in memory.
88 char *UnEscapeLexed(char *Buffer, bool AllowNull) {
90 for (char *BIn = Buffer; *BIn; ) {
91 if (BIn[0] == '\\' && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
92 char Tmp = BIn[3]; BIn[3] = 0; // Terminate string
93 *BOut = strtol(BIn+1, 0, 16); // Convert to number
94 if (!AllowNull && !*BOut)
95 ThrowException("String literal cannot accept \\00 escape!");
97 BIn[3] = Tmp; // Restore character
98 BIn += 3; // Skip over handled chars
108 #define YY_NEVER_INTERACTIVE 1
113 /* Comments start with a ; and go till end of line */
116 /* Variable(Value) identifiers start with a % sign */
117 VarID %[-a-zA-Z$._][-a-zA-Z$._0-9]*
119 /* Label identifiers end with a colon */
120 Label [-a-zA-Z$._0-9]+:
122 /* Quoted names can contain any character except " and \ */
123 StringConstant \"[^\"]+\"
126 /* [PN]Integer: match positive and negative literal integer values that
127 * are preceeded by a '%' character. These represent unnamed variable slots.
133 /* E[PN]Integer: match positive and negative literal integer values */
137 /* FPConstant - A Floating point constant.
139 FPConstant [-+]?[0-9]+[.][0-9]*([eE][-+]?[0-9]+)?
141 /* HexFPConstant - Floating point constant represented in IEEE format as a
142 * hexadecimal number for when exponential notation is not precise enough.
144 HexFPConstant 0x[0-9A-Fa-f]+
147 {Comment} { /* Ignore comments for now */ }
149 begin { return BEGINTOK; }
150 end { return ENDTOK; }
151 true { return TRUE; }
152 false { return FALSE; }
153 declare { return DECLARE; }
154 global { return GLOBAL; }
155 constant { return CONSTANT; }
156 const { return CONST; }
157 internal { return INTERNAL; }
158 uninitialized { return UNINIT; }
159 implementation { return IMPLEMENTATION; }
160 \.\.\. { return DOTDOTDOT; }
161 string { return STRING; }
162 null { return NULL_TOK; }
164 except { return EXCEPT; }
166 void { llvmAsmlval.PrimType = Type::VoidTy ; return VOID; }
167 bool { llvmAsmlval.PrimType = Type::BoolTy ; return BOOL; }
168 sbyte { llvmAsmlval.PrimType = Type::SByteTy ; return SBYTE; }
169 ubyte { llvmAsmlval.PrimType = Type::UByteTy ; return UBYTE; }
170 short { llvmAsmlval.PrimType = Type::ShortTy ; return SHORT; }
171 ushort { llvmAsmlval.PrimType = Type::UShortTy; return USHORT; }
172 int { llvmAsmlval.PrimType = Type::IntTy ; return INT; }
173 uint { llvmAsmlval.PrimType = Type::UIntTy ; return UINT; }
174 long { llvmAsmlval.PrimType = Type::LongTy ; return LONG; }
175 ulong { llvmAsmlval.PrimType = Type::ULongTy ; return ULONG; }
176 float { llvmAsmlval.PrimType = Type::FloatTy ; return FLOAT; }
177 double { llvmAsmlval.PrimType = Type::DoubleTy; return DOUBLE; }
178 type { llvmAsmlval.PrimType = Type::TypeTy ; return TYPE; }
179 label { llvmAsmlval.PrimType = Type::LabelTy ; return LABEL; }
180 opaque { return OPAQUE; }
183 not { RET_TOK(UnaryOpVal, Not, NOT); }
185 add { RET_TOK(BinaryOpVal, Add, ADD); }
186 sub { RET_TOK(BinaryOpVal, Sub, SUB); }
187 mul { RET_TOK(BinaryOpVal, Mul, MUL); }
188 div { RET_TOK(BinaryOpVal, Div, DIV); }
189 rem { RET_TOK(BinaryOpVal, Rem, REM); }
190 and { RET_TOK(BinaryOpVal, And, AND); }
191 or { RET_TOK(BinaryOpVal, Or , OR ); }
192 xor { RET_TOK(BinaryOpVal, Xor, XOR); }
193 setne { RET_TOK(BinaryOpVal, SetNE, SETNE); }
194 seteq { RET_TOK(BinaryOpVal, SetEQ, SETEQ); }
195 setlt { RET_TOK(BinaryOpVal, SetLT, SETLT); }
196 setgt { RET_TOK(BinaryOpVal, SetGT, SETGT); }
197 setle { RET_TOK(BinaryOpVal, SetLE, SETLE); }
198 setge { RET_TOK(BinaryOpVal, SetGE, SETGE); }
200 phi { RET_TOK(OtherOpVal, PHINode, PHI); }
201 call { RET_TOK(OtherOpVal, Call, CALL); }
202 cast { RET_TOK(OtherOpVal, Cast, CAST); }
203 shl { RET_TOK(OtherOpVal, Shl, SHL); }
204 shr { RET_TOK(OtherOpVal, Shr, SHR); }
206 ret { RET_TOK(TermOpVal, Ret, RET); }
207 br { RET_TOK(TermOpVal, Br, BR); }
208 switch { RET_TOK(TermOpVal, Switch, SWITCH); }
209 invoke { RET_TOK(TermOpVal, Invoke, INVOKE); }
212 malloc { RET_TOK(MemOpVal, Malloc, MALLOC); }
213 alloca { RET_TOK(MemOpVal, Alloca, ALLOCA); }
214 free { RET_TOK(MemOpVal, Free, FREE); }
215 load { RET_TOK(MemOpVal, Load, LOAD); }
216 store { RET_TOK(MemOpVal, Store, STORE); }
217 getelementptr { RET_TOK(MemOpVal, GetElementPtr, GETELEMENTPTR); }
221 UnEscapeLexed(yytext+1);
222 llvmAsmlval.StrVal = strdup(yytext+1); // Skip %
226 yytext[strlen(yytext)-1] = 0; // nuke colon
227 UnEscapeLexed(yytext);
228 llvmAsmlval.StrVal = strdup(yytext);
232 {StringConstant} { // Note that we cannot unescape a string constant here! The
233 // string constant might contain a \00 which would not be
234 // understood by the string stuff. It is valid to make a
235 // [sbyte] c"Hello World\00" constant, for example.
237 yytext[strlen(yytext)-1] = 0; // nuke end quote
238 llvmAsmlval.StrVal = strdup(yytext+1); // Nuke start quote
239 return STRINGCONSTANT;
243 {PInteger} { llvmAsmlval.UInt64Val = atoull(yytext); return EUINT64VAL; }
245 uint64_t Val = atoull(yytext+1);
246 // +1: we have bigger negative range
247 if (Val > (uint64_t)INT64_MAX+1)
248 ThrowException("Constant too large for signed 64 bits!");
249 llvmAsmlval.SInt64Val = -Val;
254 {EPInteger} { llvmAsmlval.UIntVal = atoull(yytext+1); return UINTVAL; }
256 uint64_t Val = atoull(yytext+2);
257 // +1: we have bigger negative range
258 if (Val > (uint64_t)INT32_MAX+1)
259 ThrowException("Constant too large for signed 32 bits!");
260 llvmAsmlval.SIntVal = -Val;
264 {FPConstant} { llvmAsmlval.FPVal = atof(yytext); return FPVAL; }
265 {HexFPConstant} { llvmAsmlval.FPVal = HexToFP(yytext); return FPVAL; }
267 [ \t\n] { /* Ignore whitespace */ }
268 . { return yytext[0]; }