| 1 | /* Assembler interface for targets using CGEN. -*- C -*- |
| 2 | CGEN: Cpu tools GENerator |
| 3 | |
| 4 | THIS FILE IS MACHINE GENERATED WITH CGEN. |
| 5 | - the resultant file is machine generated, cgen-asm.in isn't |
| 6 | |
| 7 | Copyright 1996, 1997, 1998, 1999, 2000, 2001 Free Software Foundation, Inc. |
| 8 | |
| 9 | This file is part of the GNU Binutils and GDB, the GNU debugger. |
| 10 | |
| 11 | This program is free software; you can redistribute it and/or modify |
| 12 | it under the terms of the GNU General Public License as published by |
| 13 | the Free Software Foundation; either version 2, or (at your option) |
| 14 | any later version. |
| 15 | |
| 16 | This program is distributed in the hope that it will be useful, |
| 17 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 18 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 19 | GNU General Public License for more details. |
| 20 | |
| 21 | You should have received a copy of the GNU General Public License |
| 22 | along with this program; if not, write to the Free Software Foundation, Inc., |
| 23 | 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ |
| 24 | |
| 25 | /* ??? Eventually more and more of this stuff can go to cpu-independent files. |
| 26 | Keep that in mind. */ |
| 27 | |
| 28 | #include "sysdep.h" |
| 29 | #include <stdio.h> |
| 30 | #include "ansidecl.h" |
| 31 | #include "bfd.h" |
| 32 | #include "symcat.h" |
| 33 | #include "ip2k-desc.h" |
| 34 | #include "ip2k-opc.h" |
| 35 | #include "opintl.h" |
| 36 | #include "xregex.h" |
| 37 | #include "libiberty.h" |
| 38 | #include "safe-ctype.h" |
| 39 | |
| 40 | #undef min |
| 41 | #define min(a,b) ((a) < (b) ? (a) : (b)) |
| 42 | #undef max |
| 43 | #define max(a,b) ((a) > (b) ? (a) : (b)) |
| 44 | |
| 45 | static const char * parse_insn_normal |
| 46 | (CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *); |
| 47 | \f |
| 48 | /* -- assembler routines inserted here. */ |
| 49 | |
| 50 | /* -- asm.c */ |
| 51 | |
| 52 | #define PARSE_FUNC_DECL(name) \ |
| 53 | static const char *name (CGEN_CPU_DESC, const char **, int, long *) |
| 54 | #define PARSE_UFUNC_DECL(name) \ |
| 55 | static const char *name (CGEN_CPU_DESC, const char **, int, unsigned long *) |
| 56 | |
| 57 | PARSE_UFUNC_DECL (parse_fr); |
| 58 | PARSE_UFUNC_DECL (parse_addr16); |
| 59 | PARSE_UFUNC_DECL (parse_addr16_cjp); |
| 60 | PARSE_FUNC_DECL (parse_lit8); |
| 61 | PARSE_UFUNC_DECL (parse_bit3); |
| 62 | |
| 63 | |
| 64 | static const char * |
| 65 | parse_fr (cd, strp, opindex, valuep) |
| 66 | CGEN_CPU_DESC cd; |
| 67 | const char **strp; |
| 68 | int opindex; |
| 69 | unsigned long *valuep; |
| 70 | { |
| 71 | const char *errmsg; |
| 72 | const char *old_strp; |
| 73 | char *afteroffset; |
| 74 | enum cgen_parse_operand_result result_type; |
| 75 | bfd_vma value; |
| 76 | extern CGEN_KEYWORD ip2k_cgen_opval_register_names; |
| 77 | bfd_vma tempvalue; |
| 78 | |
| 79 | old_strp = *strp; |
| 80 | afteroffset = NULL; |
| 81 | |
| 82 | /* Check here to see if you're about to try parsing a w as the first arg |
| 83 | and return an error if you are. */ |
| 84 | if ((strncmp (*strp, "w", 1) == 0) || (strncmp (*strp, "W", 1) == 0)) |
| 85 | { |
| 86 | (*strp)++; |
| 87 | |
| 88 | if ((strncmp (*strp, ",", 1) == 0) || ISSPACE (**strp)) |
| 89 | { |
| 90 | /* We've been passed a w. Return with an error message so that |
| 91 | cgen will try the next parsing option. */ |
| 92 | errmsg = _("W keyword invalid in FR operand slot."); |
| 93 | return errmsg; |
| 94 | } |
| 95 | *strp = old_strp; |
| 96 | } |
| 97 | |
| 98 | /* Attempt parse as register keyword. */ |
| 99 | errmsg = cgen_parse_keyword (cd, strp, & ip2k_cgen_opval_register_names, |
| 100 | (long *) valuep); |
| 101 | if (*strp != NULL |
| 102 | && errmsg == NULL) |
| 103 | return errmsg; |
| 104 | |
| 105 | /* Attempt to parse for "(IP)". */ |
| 106 | afteroffset = strstr (*strp, "(IP)"); |
| 107 | |
| 108 | if (afteroffset == NULL) |
| 109 | /* Make sure it's not in lower case. */ |
| 110 | afteroffset = strstr (*strp, "(ip)"); |
| 111 | |
| 112 | if (afteroffset != NULL) |
| 113 | { |
| 114 | if (afteroffset != *strp) |
| 115 | { |
| 116 | /* Invalid offset present. */ |
| 117 | errmsg = _("offset(IP) is not a valid form"); |
| 118 | return errmsg; |
| 119 | } |
| 120 | else |
| 121 | { |
| 122 | *strp += 4; |
| 123 | *valuep = 0; |
| 124 | errmsg = NULL; |
| 125 | return errmsg; |
| 126 | } |
| 127 | } |
| 128 | |
| 129 | /* Attempt to parse for DP. ex: mov w, offset(DP) |
| 130 | mov offset(DP),w */ |
| 131 | |
| 132 | /* Try parsing it as an address and see what comes back. */ |
| 133 | afteroffset = strstr (*strp, "(DP)"); |
| 134 | |
| 135 | if (afteroffset == NULL) |
| 136 | /* Maybe it's in lower case. */ |
| 137 | afteroffset = strstr (*strp, "(dp)"); |
| 138 | |
| 139 | if (afteroffset != NULL) |
| 140 | { |
| 141 | if (afteroffset == *strp) |
| 142 | { |
| 143 | /* No offset present. Use 0 by default. */ |
| 144 | tempvalue = 0; |
| 145 | errmsg = NULL; |
| 146 | } |
| 147 | else |
| 148 | errmsg = cgen_parse_address (cd, strp, opindex, |
| 149 | BFD_RELOC_IP2K_FR_OFFSET, |
| 150 | & result_type, & tempvalue); |
| 151 | |
| 152 | if (errmsg == NULL) |
| 153 | { |
| 154 | if (tempvalue <= 127) |
| 155 | { |
| 156 | /* Value is ok. Fix up the first 2 bits and return. */ |
| 157 | *valuep = 0x0100 | tempvalue; |
| 158 | *strp += 4; /* skip over the (DP) in *strp. */ |
| 159 | return errmsg; |
| 160 | } |
| 161 | else |
| 162 | { |
| 163 | /* Found something there in front of (DP) but it's out |
| 164 | of range. */ |
| 165 | errmsg = _("(DP) offset out of range."); |
| 166 | return errmsg; |
| 167 | } |
| 168 | } |
| 169 | } |
| 170 | |
| 171 | |
| 172 | /* Attempt to parse for SP. ex: mov w, offset(SP) |
| 173 | mov offset(SP), w. */ |
| 174 | afteroffset = strstr (*strp, "(SP)"); |
| 175 | |
| 176 | if (afteroffset == NULL) |
| 177 | /* Maybe it's in lower case. */ |
| 178 | afteroffset = strstr (*strp, "(sp)"); |
| 179 | |
| 180 | if (afteroffset != NULL) |
| 181 | { |
| 182 | if (afteroffset == *strp) |
| 183 | { |
| 184 | /* No offset present. Use 0 by default. */ |
| 185 | tempvalue = 0; |
| 186 | errmsg = NULL; |
| 187 | } |
| 188 | else |
| 189 | errmsg = cgen_parse_address (cd, strp, opindex, |
| 190 | BFD_RELOC_IP2K_FR_OFFSET, |
| 191 | & result_type, & tempvalue); |
| 192 | |
| 193 | if (errmsg == NULL) |
| 194 | { |
| 195 | if (tempvalue <= 127) |
| 196 | { |
| 197 | /* Value is ok. Fix up the first 2 bits and return. */ |
| 198 | *valuep = 0x0180 | tempvalue; |
| 199 | *strp += 4; /* skip over the (SP) in *strp. */ |
| 200 | return errmsg; |
| 201 | } |
| 202 | else |
| 203 | { |
| 204 | /* Found something there in front of (SP) but it's out |
| 205 | of range. */ |
| 206 | errmsg = _("(SP) offset out of range."); |
| 207 | return errmsg; |
| 208 | } |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | /* Attempt to parse as an address. */ |
| 213 | *strp = old_strp; |
| 214 | errmsg = cgen_parse_address (cd, strp, opindex, BFD_RELOC_IP2K_FR9, |
| 215 | & result_type, & value); |
| 216 | if (errmsg == NULL) |
| 217 | { |
| 218 | *valuep = value; |
| 219 | |
| 220 | /* if a parenthesis is found, warn about invalid form. */ |
| 221 | if (**strp == '(') |
| 222 | errmsg = _("illegal use of parentheses"); |
| 223 | |
| 224 | /* if a numeric value is specified, ensure that it is between |
| 225 | 1 and 255. */ |
| 226 | else if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) |
| 227 | { |
| 228 | if (value < 0x1 || value > 0xff) |
| 229 | errmsg = _("operand out of range (not between 1 and 255)"); |
| 230 | } |
| 231 | } |
| 232 | return errmsg; |
| 233 | } |
| 234 | |
| 235 | static const char * |
| 236 | parse_addr16 (cd, strp, opindex, valuep) |
| 237 | CGEN_CPU_DESC cd; |
| 238 | const char **strp; |
| 239 | int opindex; |
| 240 | unsigned long *valuep; |
| 241 | { |
| 242 | const char *errmsg; |
| 243 | enum cgen_parse_operand_result result_type; |
| 244 | bfd_reloc_code_real_type code = BFD_RELOC_NONE; |
| 245 | bfd_vma value; |
| 246 | |
| 247 | if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16H) |
| 248 | code = BFD_RELOC_IP2K_HI8DATA; |
| 249 | else if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16L) |
| 250 | code = BFD_RELOC_IP2K_LO8DATA; |
| 251 | else |
| 252 | { |
| 253 | /* Something is very wrong. opindex has to be one of the above. */ |
| 254 | errmsg = _("parse_addr16: invalid opindex."); |
| 255 | return errmsg; |
| 256 | } |
| 257 | |
| 258 | errmsg = cgen_parse_address (cd, strp, opindex, code, |
| 259 | & result_type, & value); |
| 260 | if (errmsg == NULL) |
| 261 | { |
| 262 | /* We either have a relocation or a number now. */ |
| 263 | if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) |
| 264 | { |
| 265 | /* We got a number back. */ |
| 266 | if (code == BFD_RELOC_IP2K_HI8DATA) |
| 267 | value >>= 8; |
| 268 | else /* code = BFD_RELOC_IP2K_LOW8DATA */ |
| 269 | value &= 0x00FF; |
| 270 | } |
| 271 | *valuep = value; |
| 272 | } |
| 273 | |
| 274 | return errmsg; |
| 275 | } |
| 276 | |
| 277 | |
| 278 | static const char * |
| 279 | parse_addr16_cjp (cd, strp, opindex, valuep) |
| 280 | CGEN_CPU_DESC cd; |
| 281 | const char **strp; |
| 282 | int opindex; |
| 283 | unsigned long *valuep; |
| 284 | { |
| 285 | const char *errmsg; |
| 286 | enum cgen_parse_operand_result result_type; |
| 287 | bfd_reloc_code_real_type code = BFD_RELOC_NONE; |
| 288 | bfd_vma value; |
| 289 | |
| 290 | if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16CJP) |
| 291 | code = BFD_RELOC_IP2K_ADDR16CJP; |
| 292 | else if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16P) |
| 293 | code = BFD_RELOC_IP2K_PAGE3; |
| 294 | |
| 295 | errmsg = cgen_parse_address (cd, strp, opindex, code, |
| 296 | & result_type, & value); |
| 297 | if (errmsg == NULL) |
| 298 | { |
| 299 | if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) |
| 300 | { |
| 301 | if ((value & 0x1) == 0) /* If the address is even .... */ |
| 302 | { |
| 303 | if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16CJP) |
| 304 | *valuep = (value >> 1) & 0x1FFF; /* Should mask be 1FFF? */ |
| 305 | else if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16P) |
| 306 | *valuep = (value >> 14) & 0x7; |
| 307 | } |
| 308 | else |
| 309 | errmsg = _("Byte address required. - must be even."); |
| 310 | } |
| 311 | else if (result_type == CGEN_PARSE_OPERAND_RESULT_QUEUED) |
| 312 | { |
| 313 | /* This will happen for things like (s2-s1) where s2 and s1 |
| 314 | are labels. */ |
| 315 | *valuep = value; |
| 316 | } |
| 317 | else |
| 318 | errmsg = _("cgen_parse_address returned a symbol. Literal required."); |
| 319 | } |
| 320 | return errmsg; |
| 321 | } |
| 322 | |
| 323 | |
| 324 | static const char * |
| 325 | parse_lit8 (cd, strp, opindex, valuep) |
| 326 | CGEN_CPU_DESC cd; |
| 327 | const char **strp; |
| 328 | int opindex; |
| 329 | long *valuep; |
| 330 | { |
| 331 | const char *errmsg; |
| 332 | enum cgen_parse_operand_result result_type; |
| 333 | bfd_reloc_code_real_type code = BFD_RELOC_NONE; |
| 334 | bfd_vma value; |
| 335 | |
| 336 | /* Parse %OP relocating operators. */ |
| 337 | if (strncmp (*strp, "%bank", 5) == 0) |
| 338 | { |
| 339 | *strp += 5; |
| 340 | code = BFD_RELOC_IP2K_BANK; |
| 341 | } |
| 342 | else if (strncmp (*strp, "%lo8data", 8) == 0) |
| 343 | { |
| 344 | *strp += 8; |
| 345 | code = BFD_RELOC_IP2K_LO8DATA; |
| 346 | } |
| 347 | else if (strncmp (*strp, "%hi8data", 8) == 0) |
| 348 | { |
| 349 | *strp += 8; |
| 350 | code = BFD_RELOC_IP2K_HI8DATA; |
| 351 | } |
| 352 | else if (strncmp (*strp, "%ex8data", 8) == 0) |
| 353 | { |
| 354 | *strp += 8; |
| 355 | code = BFD_RELOC_IP2K_EX8DATA; |
| 356 | } |
| 357 | else if (strncmp (*strp, "%lo8insn", 8) == 0) |
| 358 | { |
| 359 | *strp += 8; |
| 360 | code = BFD_RELOC_IP2K_LO8INSN; |
| 361 | } |
| 362 | else if (strncmp (*strp, "%hi8insn", 8) == 0) |
| 363 | { |
| 364 | *strp += 8; |
| 365 | code = BFD_RELOC_IP2K_HI8INSN; |
| 366 | } |
| 367 | |
| 368 | |
| 369 | /* Parse %op operand. */ |
| 370 | if (code != BFD_RELOC_NONE) |
| 371 | { |
| 372 | errmsg = cgen_parse_address (cd, strp, opindex, code, |
| 373 | & result_type, & value); |
| 374 | if ((errmsg == NULL) && |
| 375 | (result_type != CGEN_PARSE_OPERAND_RESULT_QUEUED)) |
| 376 | errmsg = _("%operator operand is not a symbol"); |
| 377 | |
| 378 | *valuep = value; |
| 379 | } |
| 380 | /* Parse as a number. */ |
| 381 | else |
| 382 | { |
| 383 | errmsg = cgen_parse_signed_integer (cd, strp, opindex, valuep); |
| 384 | |
| 385 | /* Truncate to eight bits to accept both signed and unsigned input. */ |
| 386 | if (errmsg == NULL) |
| 387 | *valuep &= 0xFF; |
| 388 | } |
| 389 | |
| 390 | return errmsg; |
| 391 | } |
| 392 | |
| 393 | static const char * |
| 394 | parse_bit3 (cd, strp, opindex, valuep) |
| 395 | CGEN_CPU_DESC cd; |
| 396 | const char **strp; |
| 397 | int opindex; |
| 398 | unsigned long *valuep; |
| 399 | { |
| 400 | const char *errmsg; |
| 401 | char mode = 0; |
| 402 | long count = 0; |
| 403 | unsigned long value; |
| 404 | |
| 405 | if (strncmp (*strp, "%bit", 4) == 0) |
| 406 | { |
| 407 | *strp += 4; |
| 408 | mode = 1; |
| 409 | } |
| 410 | else if (strncmp (*strp, "%msbbit", 7) == 0) |
| 411 | { |
| 412 | *strp += 7; |
| 413 | mode = 1; |
| 414 | } |
| 415 | else if (strncmp (*strp, "%lsbbit", 7) == 0) |
| 416 | { |
| 417 | *strp += 7; |
| 418 | mode = 2; |
| 419 | } |
| 420 | |
| 421 | errmsg = cgen_parse_unsigned_integer (cd, strp, opindex, valuep); |
| 422 | if (errmsg) |
| 423 | return errmsg; |
| 424 | |
| 425 | if (mode) |
| 426 | { |
| 427 | value = * valuep; |
| 428 | if (value == 0) |
| 429 | { |
| 430 | errmsg = _("Attempt to find bit index of 0"); |
| 431 | return errmsg; |
| 432 | } |
| 433 | |
| 434 | if (mode == 1) |
| 435 | { |
| 436 | count = 31; |
| 437 | while ((value & 0x80000000) == 0) |
| 438 | { |
| 439 | count--; |
| 440 | value <<= 1; |
| 441 | } |
| 442 | } |
| 443 | else if (mode == 2) |
| 444 | { |
| 445 | count = 0; |
| 446 | while ((value & 0x00000001) == 0) |
| 447 | { |
| 448 | count++; |
| 449 | value >>= 1; |
| 450 | } |
| 451 | } |
| 452 | |
| 453 | *valuep = count; |
| 454 | } |
| 455 | |
| 456 | return errmsg; |
| 457 | } |
| 458 | |
| 459 | |
| 460 | /* -- dis.c */ |
| 461 | |
| 462 | const char * ip2k_cgen_parse_operand |
| 463 | PARAMS ((CGEN_CPU_DESC, int, const char **, CGEN_FIELDS *)); |
| 464 | |
| 465 | /* Main entry point for operand parsing. |
| 466 | |
| 467 | This function is basically just a big switch statement. Earlier versions |
| 468 | used tables to look up the function to use, but |
| 469 | - if the table contains both assembler and disassembler functions then |
| 470 | the disassembler contains much of the assembler and vice-versa, |
| 471 | - there's a lot of inlining possibilities as things grow, |
| 472 | - using a switch statement avoids the function call overhead. |
| 473 | |
| 474 | This function could be moved into `parse_insn_normal', but keeping it |
| 475 | separate makes clear the interface between `parse_insn_normal' and each of |
| 476 | the handlers. */ |
| 477 | |
| 478 | const char * |
| 479 | ip2k_cgen_parse_operand (cd, opindex, strp, fields) |
| 480 | CGEN_CPU_DESC cd; |
| 481 | int opindex; |
| 482 | const char ** strp; |
| 483 | CGEN_FIELDS * fields; |
| 484 | { |
| 485 | const char * errmsg = NULL; |
| 486 | /* Used by scalar operands that still need to be parsed. */ |
| 487 | long junk ATTRIBUTE_UNUSED; |
| 488 | |
| 489 | switch (opindex) |
| 490 | { |
| 491 | case IP2K_OPERAND_ADDR16CJP : |
| 492 | errmsg = parse_addr16_cjp (cd, strp, IP2K_OPERAND_ADDR16CJP, (unsigned long *) (& fields->f_addr16cjp)); |
| 493 | break; |
| 494 | case IP2K_OPERAND_ADDR16H : |
| 495 | errmsg = parse_addr16 (cd, strp, IP2K_OPERAND_ADDR16H, (unsigned long *) (& fields->f_imm8)); |
| 496 | break; |
| 497 | case IP2K_OPERAND_ADDR16L : |
| 498 | errmsg = parse_addr16 (cd, strp, IP2K_OPERAND_ADDR16L, (unsigned long *) (& fields->f_imm8)); |
| 499 | break; |
| 500 | case IP2K_OPERAND_ADDR16P : |
| 501 | errmsg = parse_addr16_cjp (cd, strp, IP2K_OPERAND_ADDR16P, (unsigned long *) (& fields->f_page3)); |
| 502 | break; |
| 503 | case IP2K_OPERAND_BITNO : |
| 504 | errmsg = parse_bit3 (cd, strp, IP2K_OPERAND_BITNO, (unsigned long *) (& fields->f_bitno)); |
| 505 | break; |
| 506 | case IP2K_OPERAND_CBIT : |
| 507 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_CBIT, (unsigned long *) (& junk)); |
| 508 | break; |
| 509 | case IP2K_OPERAND_DCBIT : |
| 510 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_DCBIT, (unsigned long *) (& junk)); |
| 511 | break; |
| 512 | case IP2K_OPERAND_FR : |
| 513 | errmsg = parse_fr (cd, strp, IP2K_OPERAND_FR, (unsigned long *) (& fields->f_reg)); |
| 514 | break; |
| 515 | case IP2K_OPERAND_LIT8 : |
| 516 | errmsg = parse_lit8 (cd, strp, IP2K_OPERAND_LIT8, (long *) (& fields->f_imm8)); |
| 517 | break; |
| 518 | case IP2K_OPERAND_PABITS : |
| 519 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_PABITS, (unsigned long *) (& junk)); |
| 520 | break; |
| 521 | case IP2K_OPERAND_RETI3 : |
| 522 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_RETI3, (unsigned long *) (& fields->f_reti3)); |
| 523 | break; |
| 524 | case IP2K_OPERAND_ZBIT : |
| 525 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_ZBIT, (unsigned long *) (& junk)); |
| 526 | break; |
| 527 | |
| 528 | default : |
| 529 | /* xgettext:c-format */ |
| 530 | fprintf (stderr, _("Unrecognized field %d while parsing.\n"), opindex); |
| 531 | abort (); |
| 532 | } |
| 533 | |
| 534 | return errmsg; |
| 535 | } |
| 536 | |
| 537 | cgen_parse_fn * const ip2k_cgen_parse_handlers[] = |
| 538 | { |
| 539 | parse_insn_normal, |
| 540 | }; |
| 541 | |
| 542 | void |
| 543 | ip2k_cgen_init_asm (cd) |
| 544 | CGEN_CPU_DESC cd; |
| 545 | { |
| 546 | ip2k_cgen_init_opcode_table (cd); |
| 547 | ip2k_cgen_init_ibld_table (cd); |
| 548 | cd->parse_handlers = & ip2k_cgen_parse_handlers[0]; |
| 549 | cd->parse_operand = ip2k_cgen_parse_operand; |
| 550 | } |
| 551 | |
| 552 | \f |
| 553 | |
| 554 | /* Regex construction routine. |
| 555 | |
| 556 | This translates an opcode syntax string into a regex string, |
| 557 | by replacing any non-character syntax element (such as an |
| 558 | opcode) with the pattern '.*' |
| 559 | |
| 560 | It then compiles the regex and stores it in the opcode, for |
| 561 | later use by ip2k_cgen_assemble_insn |
| 562 | |
| 563 | Returns NULL for success, an error message for failure. */ |
| 564 | |
| 565 | char * |
| 566 | ip2k_cgen_build_insn_regex (CGEN_INSN *insn) |
| 567 | { |
| 568 | CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn); |
| 569 | const char *mnem = CGEN_INSN_MNEMONIC (insn); |
| 570 | char rxbuf[CGEN_MAX_RX_ELEMENTS]; |
| 571 | char *rx = rxbuf; |
| 572 | const CGEN_SYNTAX_CHAR_TYPE *syn; |
| 573 | int reg_err; |
| 574 | |
| 575 | syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); |
| 576 | |
| 577 | /* Mnemonics come first in the syntax string. */ |
| 578 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) |
| 579 | return _("missing mnemonic in syntax string"); |
| 580 | ++syn; |
| 581 | |
| 582 | /* Generate a case sensitive regular expression that emulates case |
| 583 | insensitive matching in the "C" locale. We cannot generate a case |
| 584 | insensitive regular expression because in Turkish locales, 'i' and 'I' |
| 585 | are not equal modulo case conversion. */ |
| 586 | |
| 587 | /* Copy the literal mnemonic out of the insn. */ |
| 588 | for (; *mnem; mnem++) |
| 589 | { |
| 590 | char c = *mnem; |
| 591 | |
| 592 | if (ISALPHA (c)) |
| 593 | { |
| 594 | *rx++ = '['; |
| 595 | *rx++ = TOLOWER (c); |
| 596 | *rx++ = TOUPPER (c); |
| 597 | *rx++ = ']'; |
| 598 | } |
| 599 | else |
| 600 | *rx++ = c; |
| 601 | } |
| 602 | |
| 603 | /* Copy any remaining literals from the syntax string into the rx. */ |
| 604 | for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn) |
| 605 | { |
| 606 | if (CGEN_SYNTAX_CHAR_P (* syn)) |
| 607 | { |
| 608 | char c = CGEN_SYNTAX_CHAR (* syn); |
| 609 | |
| 610 | switch (c) |
| 611 | { |
| 612 | /* Escape any regex metacharacters in the syntax. */ |
| 613 | case '.': case '[': case '\\': |
| 614 | case '*': case '^': case '$': |
| 615 | |
| 616 | #ifdef CGEN_ESCAPE_EXTENDED_REGEX |
| 617 | case '?': case '{': case '}': |
| 618 | case '(': case ')': case '*': |
| 619 | case '|': case '+': case ']': |
| 620 | #endif |
| 621 | *rx++ = '\\'; |
| 622 | *rx++ = c; |
| 623 | break; |
| 624 | |
| 625 | default: |
| 626 | if (ISALPHA (c)) |
| 627 | { |
| 628 | *rx++ = '['; |
| 629 | *rx++ = TOLOWER (c); |
| 630 | *rx++ = TOUPPER (c); |
| 631 | *rx++ = ']'; |
| 632 | } |
| 633 | else |
| 634 | *rx++ = c; |
| 635 | break; |
| 636 | } |
| 637 | } |
| 638 | else |
| 639 | { |
| 640 | /* Replace non-syntax fields with globs. */ |
| 641 | *rx++ = '.'; |
| 642 | *rx++ = '*'; |
| 643 | } |
| 644 | } |
| 645 | |
| 646 | /* Trailing whitespace ok. */ |
| 647 | * rx++ = '['; |
| 648 | * rx++ = ' '; |
| 649 | * rx++ = '\t'; |
| 650 | * rx++ = ']'; |
| 651 | * rx++ = '*'; |
| 652 | |
| 653 | /* But anchor it after that. */ |
| 654 | * rx++ = '$'; |
| 655 | * rx = '\0'; |
| 656 | |
| 657 | CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); |
| 658 | reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB); |
| 659 | |
| 660 | if (reg_err == 0) |
| 661 | return NULL; |
| 662 | else |
| 663 | { |
| 664 | static char msg[80]; |
| 665 | |
| 666 | regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); |
| 667 | regfree ((regex_t *) CGEN_INSN_RX (insn)); |
| 668 | free (CGEN_INSN_RX (insn)); |
| 669 | (CGEN_INSN_RX (insn)) = NULL; |
| 670 | return msg; |
| 671 | } |
| 672 | } |
| 673 | |
| 674 | \f |
| 675 | /* Default insn parser. |
| 676 | |
| 677 | The syntax string is scanned and operands are parsed and stored in FIELDS. |
| 678 | Relocs are queued as we go via other callbacks. |
| 679 | |
| 680 | ??? Note that this is currently an all-or-nothing parser. If we fail to |
| 681 | parse the instruction, we return 0 and the caller will start over from |
| 682 | the beginning. Backtracking will be necessary in parsing subexpressions, |
| 683 | but that can be handled there. Not handling backtracking here may get |
| 684 | expensive in the case of the m68k. Deal with later. |
| 685 | |
| 686 | Returns NULL for success, an error message for failure. */ |
| 687 | |
| 688 | static const char * |
| 689 | parse_insn_normal (CGEN_CPU_DESC cd, |
| 690 | const CGEN_INSN *insn, |
| 691 | const char **strp, |
| 692 | CGEN_FIELDS *fields) |
| 693 | { |
| 694 | /* ??? Runtime added insns not handled yet. */ |
| 695 | const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn); |
| 696 | const char *str = *strp; |
| 697 | const char *errmsg; |
| 698 | const char *p; |
| 699 | const CGEN_SYNTAX_CHAR_TYPE * syn; |
| 700 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 701 | /* FIXME: wip */ |
| 702 | int past_opcode_p; |
| 703 | #endif |
| 704 | |
| 705 | /* For now we assume the mnemonic is first (there are no leading operands). |
| 706 | We can parse it without needing to set up operand parsing. |
| 707 | GAS's input scrubber will ensure mnemonics are lowercase, but we may |
| 708 | not be called from GAS. */ |
| 709 | p = CGEN_INSN_MNEMONIC (insn); |
| 710 | while (*p && TOLOWER (*p) == TOLOWER (*str)) |
| 711 | ++p, ++str; |
| 712 | |
| 713 | if (* p) |
| 714 | return _("unrecognized instruction"); |
| 715 | |
| 716 | #ifndef CGEN_MNEMONIC_OPERANDS |
| 717 | if (* str && ! ISSPACE (* str)) |
| 718 | return _("unrecognized instruction"); |
| 719 | #endif |
| 720 | |
| 721 | CGEN_INIT_PARSE (cd); |
| 722 | cgen_init_parse_operand (cd); |
| 723 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 724 | past_opcode_p = 0; |
| 725 | #endif |
| 726 | |
| 727 | /* We don't check for (*str != '\0') here because we want to parse |
| 728 | any trailing fake arguments in the syntax string. */ |
| 729 | syn = CGEN_SYNTAX_STRING (syntax); |
| 730 | |
| 731 | /* Mnemonics come first for now, ensure valid string. */ |
| 732 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) |
| 733 | abort (); |
| 734 | |
| 735 | ++syn; |
| 736 | |
| 737 | while (* syn != 0) |
| 738 | { |
| 739 | /* Non operand chars must match exactly. */ |
| 740 | if (CGEN_SYNTAX_CHAR_P (* syn)) |
| 741 | { |
| 742 | /* FIXME: While we allow for non-GAS callers above, we assume the |
| 743 | first char after the mnemonic part is a space. */ |
| 744 | /* FIXME: We also take inappropriate advantage of the fact that |
| 745 | GAS's input scrubber will remove extraneous blanks. */ |
| 746 | if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn))) |
| 747 | { |
| 748 | #ifdef CGEN_MNEMONIC_OPERANDS |
| 749 | if (CGEN_SYNTAX_CHAR(* syn) == ' ') |
| 750 | past_opcode_p = 1; |
| 751 | #endif |
| 752 | ++ syn; |
| 753 | ++ str; |
| 754 | } |
| 755 | else if (*str) |
| 756 | { |
| 757 | /* Syntax char didn't match. Can't be this insn. */ |
| 758 | static char msg [80]; |
| 759 | |
| 760 | /* xgettext:c-format */ |
| 761 | sprintf (msg, _("syntax error (expected char `%c', found `%c')"), |
| 762 | CGEN_SYNTAX_CHAR(*syn), *str); |
| 763 | return msg; |
| 764 | } |
| 765 | else |
| 766 | { |
| 767 | /* Ran out of input. */ |
| 768 | static char msg [80]; |
| 769 | |
| 770 | /* xgettext:c-format */ |
| 771 | sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"), |
| 772 | CGEN_SYNTAX_CHAR(*syn)); |
| 773 | return msg; |
| 774 | } |
| 775 | continue; |
| 776 | } |
| 777 | |
| 778 | /* We have an operand of some sort. */ |
| 779 | errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), |
| 780 | &str, fields); |
| 781 | if (errmsg) |
| 782 | return errmsg; |
| 783 | |
| 784 | /* Done with this operand, continue with next one. */ |
| 785 | ++ syn; |
| 786 | } |
| 787 | |
| 788 | /* If we're at the end of the syntax string, we're done. */ |
| 789 | if (* syn == 0) |
| 790 | { |
| 791 | /* FIXME: For the moment we assume a valid `str' can only contain |
| 792 | blanks now. IE: We needn't try again with a longer version of |
| 793 | the insn and it is assumed that longer versions of insns appear |
| 794 | before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */ |
| 795 | while (ISSPACE (* str)) |
| 796 | ++ str; |
| 797 | |
| 798 | if (* str != '\0') |
| 799 | return _("junk at end of line"); /* FIXME: would like to include `str' */ |
| 800 | |
| 801 | return NULL; |
| 802 | } |
| 803 | |
| 804 | /* We couldn't parse it. */ |
| 805 | return _("unrecognized instruction"); |
| 806 | } |
| 807 | \f |
| 808 | /* Main entry point. |
| 809 | This routine is called for each instruction to be assembled. |
| 810 | STR points to the insn to be assembled. |
| 811 | We assume all necessary tables have been initialized. |
| 812 | The assembled instruction, less any fixups, is stored in BUF. |
| 813 | Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value |
| 814 | still needs to be converted to target byte order, otherwise BUF is an array |
| 815 | of bytes in target byte order. |
| 816 | The result is a pointer to the insn's entry in the opcode table, |
| 817 | or NULL if an error occured (an error message will have already been |
| 818 | printed). |
| 819 | |
| 820 | Note that when processing (non-alias) macro-insns, |
| 821 | this function recurses. |
| 822 | |
| 823 | ??? It's possible to make this cpu-independent. |
| 824 | One would have to deal with a few minor things. |
| 825 | At this point in time doing so would be more of a curiosity than useful |
| 826 | [for example this file isn't _that_ big], but keeping the possibility in |
| 827 | mind helps keep the design clean. */ |
| 828 | |
| 829 | const CGEN_INSN * |
| 830 | ip2k_cgen_assemble_insn (CGEN_CPU_DESC cd, |
| 831 | const char *str, |
| 832 | CGEN_FIELDS *fields, |
| 833 | CGEN_INSN_BYTES_PTR buf, |
| 834 | char **errmsg) |
| 835 | { |
| 836 | const char *start; |
| 837 | CGEN_INSN_LIST *ilist; |
| 838 | const char *parse_errmsg = NULL; |
| 839 | const char *insert_errmsg = NULL; |
| 840 | int recognized_mnemonic = 0; |
| 841 | |
| 842 | /* Skip leading white space. */ |
| 843 | while (ISSPACE (* str)) |
| 844 | ++ str; |
| 845 | |
| 846 | /* The instructions are stored in hashed lists. |
| 847 | Get the first in the list. */ |
| 848 | ilist = CGEN_ASM_LOOKUP_INSN (cd, str); |
| 849 | |
| 850 | /* Keep looking until we find a match. */ |
| 851 | start = str; |
| 852 | for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) |
| 853 | { |
| 854 | const CGEN_INSN *insn = ilist->insn; |
| 855 | recognized_mnemonic = 1; |
| 856 | |
| 857 | #ifdef CGEN_VALIDATE_INSN_SUPPORTED |
| 858 | /* Not usually needed as unsupported opcodes |
| 859 | shouldn't be in the hash lists. */ |
| 860 | /* Is this insn supported by the selected cpu? */ |
| 861 | if (! ip2k_cgen_insn_supported (cd, insn)) |
| 862 | continue; |
| 863 | #endif |
| 864 | /* If the RELAXED attribute is set, this is an insn that shouldn't be |
| 865 | chosen immediately. Instead, it is used during assembler/linker |
| 866 | relaxation if possible. */ |
| 867 | if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0) |
| 868 | continue; |
| 869 | |
| 870 | str = start; |
| 871 | |
| 872 | /* Skip this insn if str doesn't look right lexically. */ |
| 873 | if (CGEN_INSN_RX (insn) != NULL && |
| 874 | regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) |
| 875 | continue; |
| 876 | |
| 877 | /* Allow parse/insert handlers to obtain length of insn. */ |
| 878 | CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); |
| 879 | |
| 880 | parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields); |
| 881 | if (parse_errmsg != NULL) |
| 882 | continue; |
| 883 | |
| 884 | /* ??? 0 is passed for `pc'. */ |
| 885 | insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf, |
| 886 | (bfd_vma) 0); |
| 887 | if (insert_errmsg != NULL) |
| 888 | continue; |
| 889 | |
| 890 | /* It is up to the caller to actually output the insn and any |
| 891 | queued relocs. */ |
| 892 | return insn; |
| 893 | } |
| 894 | |
| 895 | { |
| 896 | static char errbuf[150]; |
| 897 | #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS |
| 898 | const char *tmp_errmsg; |
| 899 | |
| 900 | /* If requesting verbose error messages, use insert_errmsg. |
| 901 | Failing that, use parse_errmsg. */ |
| 902 | tmp_errmsg = (insert_errmsg ? insert_errmsg : |
| 903 | parse_errmsg ? parse_errmsg : |
| 904 | recognized_mnemonic ? |
| 905 | _("unrecognized form of instruction") : |
| 906 | _("unrecognized instruction")); |
| 907 | |
| 908 | if (strlen (start) > 50) |
| 909 | /* xgettext:c-format */ |
| 910 | sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start); |
| 911 | else |
| 912 | /* xgettext:c-format */ |
| 913 | sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start); |
| 914 | #else |
| 915 | if (strlen (start) > 50) |
| 916 | /* xgettext:c-format */ |
| 917 | sprintf (errbuf, _("bad instruction `%.50s...'"), start); |
| 918 | else |
| 919 | /* xgettext:c-format */ |
| 920 | sprintf (errbuf, _("bad instruction `%.50s'"), start); |
| 921 | #endif |
| 922 | |
| 923 | *errmsg = errbuf; |
| 924 | return NULL; |
| 925 | } |
| 926 | } |
| 927 | \f |
| 928 | #if 0 /* This calls back to GAS which we can't do without care. */ |
| 929 | |
| 930 | /* Record each member of OPVALS in the assembler's symbol table. |
| 931 | This lets GAS parse registers for us. |
| 932 | ??? Interesting idea but not currently used. */ |
| 933 | |
| 934 | /* Record each member of OPVALS in the assembler's symbol table. |
| 935 | FIXME: Not currently used. */ |
| 936 | |
| 937 | void |
| 938 | ip2k_cgen_asm_hash_keywords (CGEN_CPU_DESC cd, CGEN_KEYWORD *opvals) |
| 939 | { |
| 940 | CGEN_KEYWORD_SEARCH search = cgen_keyword_search_init (opvals, NULL); |
| 941 | const CGEN_KEYWORD_ENTRY * ke; |
| 942 | |
| 943 | while ((ke = cgen_keyword_search_next (& search)) != NULL) |
| 944 | { |
| 945 | #if 0 /* Unnecessary, should be done in the search routine. */ |
| 946 | if (! ip2k_cgen_opval_supported (ke)) |
| 947 | continue; |
| 948 | #endif |
| 949 | cgen_asm_record_register (cd, ke->name, ke->value); |
| 950 | } |
| 951 | } |
| 952 | |
| 953 | #endif /* 0 */ |