1 /*===-- UpgradeLexer.l - Scanner for 1.9 assembly files --------*- C++ -*--===//
3 // The LLVM Compiler Infrastructure
5 // This file was developed by Reid Spencer and is distributed under the
6 // University of Illinois Open Source License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the flex scanner for LLVM 1.9 assembly languages files.
12 //===----------------------------------------------------------------------===*/
14 %option prefix="Upgrade"
17 %option never-interactive
22 %option outfile="UpgradeLexer.cpp"
28 #include "UpgradeInternals.h"
29 #include "llvm/Module.h"
31 #include "UpgradeParser.h"
35 #define YY_INPUT(buf,result,max_size) \
37 if (LexInput->good() && !LexInput->eof()) { \
38 LexInput->read(buf,max_size); \
39 result = LexInput->gcount(); \
45 #define YY_NEVER_INTERACTIVE 1
47 // Construct a token value for a non-obsolete token
48 #define RET_TOK(type, Enum, sym) \
49 Upgradelval.type = Enum; \
52 #define RET_TY(sym,NewTY,sign) \
53 Upgradelval.PrimType.T = NewTY; \
54 Upgradelval.PrimType.S = sign; \
59 // TODO: All of the static identifiers are figured out by the lexer,
60 // these should be hashed to reduce the lexer size
62 // UnEscapeLexed - Run through the specified buffer and change \xx codes to the
63 // appropriate character. If AllowNull is set to false, a \00 value will cause
64 // an exception to be thrown.
66 // If AllowNull is set to true, the return value of the function points to the
67 // last character of the string in memory.
69 char *UnEscapeLexed(char *Buffer, bool AllowNull) {
71 for (char *BIn = Buffer; *BIn; ) {
72 if (BIn[0] == '\\' && isxdigit(BIn[1]) && isxdigit(BIn[2])) {
73 char Tmp = BIn[3]; BIn[3] = 0; // Terminate string
74 *BOut = (char)strtol(BIn+1, 0, 16); // Convert to number
75 if (!AllowNull && !*BOut)
76 error("String literal cannot accept \\00 escape!");
78 BIn[3] = Tmp; // Restore character
79 BIn += 3; // Skip over handled chars
89 // atoull - Convert an ascii string of decimal digits into the unsigned long
90 // long representation... this does not have to do input error checking,
91 // because we know that the input will be matched by a suitable regex...
93 static uint64_t atoull(const char *Buffer) {
95 for (; *Buffer; Buffer++) {
96 uint64_t OldRes = Result;
98 Result += *Buffer-'0';
99 if (Result < OldRes) // Uh, oh, overflow detected!!!
100 error("constant bigger than 64 bits detected!");
105 static uint64_t HexIntToVal(const char *Buffer) {
107 for (; *Buffer; ++Buffer) {
108 uint64_t OldRes = Result;
111 if (C >= '0' && C <= '9')
113 else if (C >= 'A' && C <= 'F')
115 else if (C >= 'a' && C <= 'f')
118 if (Result < OldRes) // Uh, oh, overflow detected!!!
119 error("constant bigger than 64 bits detected!");
125 // HexToFP - Convert the ascii string in hexidecimal format to the floating
126 // point representation of it.
128 static double HexToFP(const char *Buffer) {
129 // Behave nicely in the face of C TBAA rules... see:
130 // http://www.nullstone.com/htmls/category/aliastyp.htm
135 UIntToFP.UI = HexIntToVal(Buffer);
137 assert(sizeof(double) == sizeof(uint64_t) &&
138 "Data sizes incompatible on this target!");
139 return UIntToFP.FP; // Cast Hex constant to double
143 } // End llvm namespace
145 using namespace llvm;
151 /* Comments start with a ; and go till end of line */
154 /* Variable(Value) identifiers start with a % sign */
155 VarID %[-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 \"[^\"]*\"
165 /* [PN]Integer: match positive and negative literal integer values that
166 * are preceeded by a '%' character. These represent unnamed variable slots.
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]+
191 {Comment} { /* Ignore comments for now */ }
193 begin { return BEGINTOK; }
194 end { return ENDTOK; }
195 true { return TRUETOK; }
196 false { return FALSETOK; }
197 declare { return DECLARE; }
198 global { return GLOBAL; }
199 constant { return CONSTANT; }
200 internal { return INTERNAL; }
201 linkonce { return LINKONCE; }
202 weak { return WEAK; }
203 appending { return APPENDING; }
204 dllimport { return DLLIMPORT; }
205 dllexport { return DLLEXPORT; }
206 extern_weak { return EXTERN_WEAK; }
207 uninitialized { return EXTERNAL; } /* Deprecated, turn into external */
208 external { return EXTERNAL; }
209 implementation { return IMPLEMENTATION; }
210 zeroinitializer { return ZEROINITIALIZER; }
211 \.\.\. { return DOTDOTDOT; }
212 undef { return UNDEF; }
213 null { return NULL_TOK; }
215 except { return EXCEPT; }
216 not { return NOT; } /* Deprecated, turned into XOR */
217 tail { return TAIL; }
218 target { return TARGET; }
219 triple { return TRIPLE; }
220 deplibs { return DEPLIBS; }
221 endian { return ENDIAN; }
222 pointersize { return POINTERSIZE; }
223 datalayout { return DATALAYOUT; }
224 little { return LITTLE; }
226 volatile { return VOLATILE; }
227 align { return ALIGN; }
228 section { return SECTION; }
229 module { return MODULE; }
230 asm { return ASM_TOK; }
231 sideeffect { return SIDEEFFECT; }
233 cc { return CC_TOK; }
234 ccc { return CCC_TOK; }
235 csretcc { return CSRETCC_TOK; }
236 fastcc { return FASTCC_TOK; }
237 coldcc { return COLDCC_TOK; }
238 x86_stdcallcc { return X86_STDCALLCC_TOK; }
239 x86_fastcallcc { return X86_FASTCALLCC_TOK; }
241 sbyte { RET_TY(SBYTE, Type::Int8Ty, Signed); }
242 ubyte { RET_TY(UBYTE, Type::Int8Ty, Unsigned); }
243 short { RET_TY(SHORT, Type::Int16Ty, Signed); }
244 ushort { RET_TY(USHORT, Type::Int16Ty, Unsigned); }
245 int { RET_TY(INT, Type::Int32Ty, Signed); }
246 uint { RET_TY(UINT, Type::Int32Ty, Unsigned); }
247 long { RET_TY(LONG, Type::Int64Ty, Signed); }
248 ulong { RET_TY(ULONG, Type::Int64Ty, Unsigned); }
249 void { RET_TY(VOID, Type::VoidTy, Signless ); }
250 bool { RET_TY(BOOL, Type::Int1Ty, Unsigned ); }
251 float { RET_TY(FLOAT, Type::FloatTy, Signless ); }
252 double { RET_TY(DOUBLE, Type::DoubleTy,Signless); }
253 label { RET_TY(LABEL, Type::LabelTy, Signless ); }
254 type { return TYPE; }
255 opaque { return OPAQUE; }
257 add { RET_TOK(BinaryOpVal, AddOp, ADD); }
258 sub { RET_TOK(BinaryOpVal, SubOp, SUB); }
259 mul { RET_TOK(BinaryOpVal, MulOp, MUL); }
260 div { RET_TOK(BinaryOpVal, DivOp, DIV); }
261 udiv { RET_TOK(BinaryOpVal, UDivOp, UDIV); }
262 sdiv { RET_TOK(BinaryOpVal, SDivOp, SDIV); }
263 fdiv { RET_TOK(BinaryOpVal, FDivOp, FDIV); }
264 rem { RET_TOK(BinaryOpVal, RemOp, REM); }
265 urem { RET_TOK(BinaryOpVal, URemOp, UREM); }
266 srem { RET_TOK(BinaryOpVal, SRemOp, SREM); }
267 frem { RET_TOK(BinaryOpVal, FRemOp, FREM); }
268 and { RET_TOK(BinaryOpVal, AndOp, AND); }
269 or { RET_TOK(BinaryOpVal, OrOp , OR ); }
270 xor { RET_TOK(BinaryOpVal, XorOp, XOR); }
271 setne { RET_TOK(BinaryOpVal, SetNE, SETNE); }
272 seteq { RET_TOK(BinaryOpVal, SetEQ, SETEQ); }
273 setlt { RET_TOK(BinaryOpVal, SetLT, SETLT); }
274 setgt { RET_TOK(BinaryOpVal, SetGT, SETGT); }
275 setle { RET_TOK(BinaryOpVal, SetLE, SETLE); }
276 setge { RET_TOK(BinaryOpVal, SetGE, SETGE); }
277 icmp { RET_TOK(OtherOpVal, ICmpOp, ICMP); }
278 fcmp { RET_TOK(OtherOpVal, FCmpOp, FCMP); }
301 phi { RET_TOK(OtherOpVal, PHIOp, PHI_TOK); }
302 call { RET_TOK(OtherOpVal, CallOp, CALL); }
303 cast { RET_TOK(CastOpVal, CastOp, CAST); }
304 trunc { RET_TOK(CastOpVal, TruncOp, TRUNC); }
305 zext { RET_TOK(CastOpVal, ZExtOp , ZEXT); }
306 sext { RET_TOK(CastOpVal, SExtOp, SEXT); }
307 fptrunc { RET_TOK(CastOpVal, FPTruncOp, FPTRUNC); }
308 fpext { RET_TOK(CastOpVal, FPExtOp, FPEXT); }
309 fptoui { RET_TOK(CastOpVal, FPToUIOp, FPTOUI); }
310 fptosi { RET_TOK(CastOpVal, FPToSIOp, FPTOSI); }
311 uitofp { RET_TOK(CastOpVal, UIToFPOp, UITOFP); }
312 sitofp { RET_TOK(CastOpVal, SIToFPOp, SITOFP); }
313 ptrtoint { RET_TOK(CastOpVal, PtrToIntOp, PTRTOINT); }
314 inttoptr { RET_TOK(CastOpVal, IntToPtrOp, INTTOPTR); }
315 bitcast { RET_TOK(CastOpVal, BitCastOp, BITCAST); }
316 select { RET_TOK(OtherOpVal, SelectOp, SELECT); }
317 shl { RET_TOK(OtherOpVal, ShlOp, SHL); }
318 shr { RET_TOK(OtherOpVal, ShrOp, SHR); }
319 lshr { RET_TOK(OtherOpVal, LShrOp, LSHR); }
320 ashr { RET_TOK(OtherOpVal, AShrOp, ASHR); }
321 vanext { return VANEXT_old; }
322 vaarg { return VAARG_old; }
323 va_arg { RET_TOK(OtherOpVal, VAArg , VAARG); }
324 ret { RET_TOK(TermOpVal, RetOp, RET); }
325 br { RET_TOK(TermOpVal, BrOp, BR); }
326 switch { RET_TOK(TermOpVal, SwitchOp, SWITCH); }
327 invoke { RET_TOK(TermOpVal, InvokeOp, INVOKE); }
328 unwind { return UNWIND; }
329 unreachable { RET_TOK(TermOpVal, UnreachableOp, UNREACHABLE); }
331 malloc { RET_TOK(MemOpVal, MallocOp, MALLOC); }
332 alloca { RET_TOK(MemOpVal, AllocaOp, ALLOCA); }
333 free { RET_TOK(MemOpVal, FreeOp, FREE); }
334 load { RET_TOK(MemOpVal, LoadOp, LOAD); }
335 store { RET_TOK(MemOpVal, StoreOp, STORE); }
336 getelementptr { RET_TOK(MemOpVal, GetElementPtrOp, GETELEMENTPTR); }
338 extractelement { RET_TOK(OtherOpVal, ExtractElementOp, EXTRACTELEMENT); }
339 insertelement { RET_TOK(OtherOpVal, InsertElementOp, INSERTELEMENT); }
340 shufflevector { RET_TOK(OtherOpVal, ShuffleVectorOp, SHUFFLEVECTOR); }
344 UnEscapeLexed(yytext+1);
345 Upgradelval.StrVal = strdup(yytext+1); // Skip %
349 yytext[strlen(yytext)-1] = 0; // nuke colon
350 UnEscapeLexed(yytext);
351 Upgradelval.StrVal = strdup(yytext);
355 yytext[strlen(yytext)-2] = 0; // nuke colon, end quote
356 UnEscapeLexed(yytext+1);
357 Upgradelval.StrVal = strdup(yytext+1);
361 {StringConstant} { // Note that we cannot unescape a string constant here! The
362 // string constant might contain a \00 which would not be
363 // understood by the string stuff. It is valid to make a
364 // [sbyte] c"Hello World\00" constant, for example.
366 yytext[strlen(yytext)-1] = 0; // nuke end quote
367 Upgradelval.StrVal = strdup(yytext+1); // Nuke start quote
368 return STRINGCONSTANT;
372 {PInteger} { Upgradelval.UInt64Val = atoull(yytext); return EUINT64VAL; }
374 uint64_t Val = atoull(yytext+1);
375 // +1: we have bigger negative range
376 if (Val > (uint64_t)INT64_MAX+1)
377 error("Constant too large for signed 64 bits!");
378 Upgradelval.SInt64Val = -Val;
382 Upgradelval.UInt64Val = HexIntToVal(yytext+3);
383 return yytext[0] == 's' ? ESINT64VAL : EUINT64VAL;
387 uint64_t Val = atoull(yytext+1);
388 if ((unsigned)Val != Val)
389 error("Invalid value number (too large)!");
390 Upgradelval.UIntVal = unsigned(Val);
394 uint64_t Val = atoull(yytext+2);
395 // +1: we have bigger negative range
396 if (Val > (uint64_t)INT32_MAX+1)
397 error("Constant too large for signed 32 bits!");
398 Upgradelval.SIntVal = (int)-Val;
402 {FPConstant} { Upgradelval.FPVal = atof(yytext); return FPVAL; }
403 {HexFPConstant} { Upgradelval.FPVal = HexToFP(yytext); return FPVAL; }
406 /* Make sure to free the internal buffers for flex when we are
407 * done reading our input!
409 yy_delete_buffer(YY_CURRENT_BUFFER);
413 [ \r\t\n] { /* Ignore whitespace */ }
414 . { return yytext[0]; }