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, 2005, 2007 8 Free Software Foundation, Inc. 9 10 This file is part of libopcodes. 11 12 This library is free software; you can redistribute it and/or modify 13 it under the terms of the GNU General Public License as published by 14 the Free Software Foundation; either version 3, or (at your option) 15 any later version. 16 17 It is distributed in the hope that it will be useful, but WITHOUT 18 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 19 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public 20 License for more details. 21 22 You should have received a copy of the GNU General Public License 23 along with this program; if not, write to the Free Software Foundation, Inc., 24 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */ 25 26 27 /* ??? Eventually more and more of this stuff can go to cpu-independent files. 28 Keep that in mind. */ 29 30 #include "sysdep.h" 31 #include <stdio.h> 32 #include "ansidecl.h" 33 #include "bfd.h" 34 #include "symcat.h" 35 #include "ip2k-desc.h" 36 #include "ip2k-opc.h" 37 #include "opintl.h" 38 #include "xregex.h" 39 #include "libiberty.h" 40 #include "safe-ctype.h" 41 42 #undef min 43 #define min(a,b) ((a) < (b) ? (a) : (b)) 44 #undef max 45 #define max(a,b) ((a) > (b) ? (a) : (b)) 46 47 static const char * parse_insn_normal 48 (CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *); 49 50 /* -- assembler routines inserted here. */ 51 52 /* -- asm.c */ 53 54 static const char * 55 parse_fr (CGEN_CPU_DESC cd, 56 const char **strp, 57 int opindex, 58 unsigned long *valuep) 59 { 60 const char *errmsg; 61 const char *old_strp; 62 char *afteroffset; 63 enum cgen_parse_operand_result result_type; 64 bfd_vma value; 65 extern CGEN_KEYWORD ip2k_cgen_opval_register_names; 66 bfd_vma tempvalue; 67 68 old_strp = *strp; 69 afteroffset = NULL; 70 71 /* Check here to see if you're about to try parsing a w as the first arg 72 and return an error if you are. */ 73 if ((strncmp (*strp, "w", 1) == 0) || (strncmp (*strp, "W", 1) == 0)) 74 { 75 (*strp)++; 76 77 if ((strncmp (*strp, ",", 1) == 0) || ISSPACE (**strp)) 78 { 79 /* We've been passed a w. Return with an error message so that 80 cgen will try the next parsing option. */ 81 errmsg = _("W keyword invalid in FR operand slot."); 82 return errmsg; 83 } 84 *strp = old_strp; 85 } 86 87 /* Attempt parse as register keyword. */ 88 errmsg = cgen_parse_keyword (cd, strp, & ip2k_cgen_opval_register_names, 89 (long *) valuep); 90 if (*strp != NULL 91 && errmsg == NULL) 92 return errmsg; 93 94 /* Attempt to parse for "(IP)". */ 95 afteroffset = strstr (*strp, "(IP)"); 96 97 if (afteroffset == NULL) 98 /* Make sure it's not in lower case. */ 99 afteroffset = strstr (*strp, "(ip)"); 100 101 if (afteroffset != NULL) 102 { 103 if (afteroffset != *strp) 104 { 105 /* Invalid offset present. */ 106 errmsg = _("offset(IP) is not a valid form"); 107 return errmsg; 108 } 109 else 110 { 111 *strp += 4; 112 *valuep = 0; 113 errmsg = NULL; 114 return errmsg; 115 } 116 } 117 118 /* Attempt to parse for DP. ex: mov w, offset(DP) 119 mov offset(DP),w */ 120 121 /* Try parsing it as an address and see what comes back. */ 122 afteroffset = strstr (*strp, "(DP)"); 123 124 if (afteroffset == NULL) 125 /* Maybe it's in lower case. */ 126 afteroffset = strstr (*strp, "(dp)"); 127 128 if (afteroffset != NULL) 129 { 130 if (afteroffset == *strp) 131 { 132 /* No offset present. Use 0 by default. */ 133 tempvalue = 0; 134 errmsg = NULL; 135 } 136 else 137 errmsg = cgen_parse_address (cd, strp, opindex, 138 BFD_RELOC_IP2K_FR_OFFSET, 139 & result_type, & tempvalue); 140 141 if (errmsg == NULL) 142 { 143 if (tempvalue <= 127) 144 { 145 /* Value is ok. Fix up the first 2 bits and return. */ 146 *valuep = 0x0100 | tempvalue; 147 *strp += 4; /* Skip over the (DP) in *strp. */ 148 return errmsg; 149 } 150 else 151 { 152 /* Found something there in front of (DP) but it's out 153 of range. */ 154 errmsg = _("(DP) offset out of range."); 155 return errmsg; 156 } 157 } 158 } 159 160 161 /* Attempt to parse for SP. ex: mov w, offset(SP) 162 mov offset(SP), w. */ 163 afteroffset = strstr (*strp, "(SP)"); 164 165 if (afteroffset == NULL) 166 /* Maybe it's in lower case. */ 167 afteroffset = strstr (*strp, "(sp)"); 168 169 if (afteroffset != NULL) 170 { 171 if (afteroffset == *strp) 172 { 173 /* No offset present. Use 0 by default. */ 174 tempvalue = 0; 175 errmsg = NULL; 176 } 177 else 178 errmsg = cgen_parse_address (cd, strp, opindex, 179 BFD_RELOC_IP2K_FR_OFFSET, 180 & result_type, & tempvalue); 181 182 if (errmsg == NULL) 183 { 184 if (tempvalue <= 127) 185 { 186 /* Value is ok. Fix up the first 2 bits and return. */ 187 *valuep = 0x0180 | tempvalue; 188 *strp += 4; /* Skip over the (SP) in *strp. */ 189 return errmsg; 190 } 191 else 192 { 193 /* Found something there in front of (SP) but it's out 194 of range. */ 195 errmsg = _("(SP) offset out of range."); 196 return errmsg; 197 } 198 } 199 } 200 201 /* Attempt to parse as an address. */ 202 *strp = old_strp; 203 errmsg = cgen_parse_address (cd, strp, opindex, BFD_RELOC_IP2K_FR9, 204 & result_type, & value); 205 if (errmsg == NULL) 206 { 207 *valuep = value; 208 209 /* If a parenthesis is found, warn about invalid form. */ 210 if (**strp == '(') 211 errmsg = _("illegal use of parentheses"); 212 213 /* If a numeric value is specified, ensure that it is between 214 1 and 255. */ 215 else if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) 216 { 217 if (value < 0x1 || value > 0xff) 218 errmsg = _("operand out of range (not between 1 and 255)"); 219 } 220 } 221 return errmsg; 222 } 223 224 static const char * 225 parse_addr16 (CGEN_CPU_DESC cd, 226 const char **strp, 227 int opindex, 228 unsigned long *valuep) 229 { 230 const char *errmsg; 231 enum cgen_parse_operand_result result_type; 232 bfd_reloc_code_real_type code = BFD_RELOC_NONE; 233 bfd_vma value; 234 235 if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16H) 236 code = BFD_RELOC_IP2K_HI8DATA; 237 else if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16L) 238 code = BFD_RELOC_IP2K_LO8DATA; 239 else 240 { 241 /* Something is very wrong. opindex has to be one of the above. */ 242 errmsg = _("parse_addr16: invalid opindex."); 243 return errmsg; 244 } 245 246 errmsg = cgen_parse_address (cd, strp, opindex, code, 247 & result_type, & value); 248 if (errmsg == NULL) 249 { 250 /* We either have a relocation or a number now. */ 251 if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) 252 { 253 /* We got a number back. */ 254 if (code == BFD_RELOC_IP2K_HI8DATA) 255 value >>= 8; 256 else 257 /* code = BFD_RELOC_IP2K_LOW8DATA. */ 258 value &= 0x00FF; 259 } 260 *valuep = value; 261 } 262 263 return errmsg; 264 } 265 266 static const char * 267 parse_addr16_cjp (CGEN_CPU_DESC cd, 268 const char **strp, 269 int opindex, 270 unsigned long *valuep) 271 { 272 const char *errmsg; 273 enum cgen_parse_operand_result result_type; 274 bfd_reloc_code_real_type code = BFD_RELOC_NONE; 275 bfd_vma value; 276 277 if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16CJP) 278 code = BFD_RELOC_IP2K_ADDR16CJP; 279 else if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16P) 280 code = BFD_RELOC_IP2K_PAGE3; 281 282 errmsg = cgen_parse_address (cd, strp, opindex, code, 283 & result_type, & value); 284 if (errmsg == NULL) 285 { 286 if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) 287 { 288 if ((value & 0x1) == 0) /* If the address is even .... */ 289 { 290 if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16CJP) 291 *valuep = (value >> 1) & 0x1FFF; /* Should mask be 1FFF? */ 292 else if (opindex == (CGEN_OPERAND_TYPE) IP2K_OPERAND_ADDR16P) 293 *valuep = (value >> 14) & 0x7; 294 } 295 else 296 errmsg = _("Byte address required. - must be even."); 297 } 298 else if (result_type == CGEN_PARSE_OPERAND_RESULT_QUEUED) 299 { 300 /* This will happen for things like (s2-s1) where s2 and s1 301 are labels. */ 302 *valuep = value; 303 } 304 else 305 errmsg = _("cgen_parse_address returned a symbol. Literal required."); 306 } 307 return errmsg; 308 } 309 310 static const char * 311 parse_lit8 (CGEN_CPU_DESC cd, 312 const char **strp, 313 int opindex, 314 long *valuep) 315 { 316 const char *errmsg; 317 enum cgen_parse_operand_result result_type; 318 bfd_reloc_code_real_type code = BFD_RELOC_NONE; 319 bfd_vma value; 320 321 /* Parse %OP relocating operators. */ 322 if (strncmp (*strp, "%bank", 5) == 0) 323 { 324 *strp += 5; 325 code = BFD_RELOC_IP2K_BANK; 326 } 327 else if (strncmp (*strp, "%lo8data", 8) == 0) 328 { 329 *strp += 8; 330 code = BFD_RELOC_IP2K_LO8DATA; 331 } 332 else if (strncmp (*strp, "%hi8data", 8) == 0) 333 { 334 *strp += 8; 335 code = BFD_RELOC_IP2K_HI8DATA; 336 } 337 else if (strncmp (*strp, "%ex8data", 8) == 0) 338 { 339 *strp += 8; 340 code = BFD_RELOC_IP2K_EX8DATA; 341 } 342 else if (strncmp (*strp, "%lo8insn", 8) == 0) 343 { 344 *strp += 8; 345 code = BFD_RELOC_IP2K_LO8INSN; 346 } 347 else if (strncmp (*strp, "%hi8insn", 8) == 0) 348 { 349 *strp += 8; 350 code = BFD_RELOC_IP2K_HI8INSN; 351 } 352 353 /* Parse %op operand. */ 354 if (code != BFD_RELOC_NONE) 355 { 356 errmsg = cgen_parse_address (cd, strp, opindex, code, 357 & result_type, & value); 358 if ((errmsg == NULL) && 359 (result_type != CGEN_PARSE_OPERAND_RESULT_QUEUED)) 360 errmsg = _("percent-operator operand is not a symbol"); 361 362 *valuep = value; 363 } 364 /* Parse as a number. */ 365 else 366 { 367 errmsg = cgen_parse_signed_integer (cd, strp, opindex, valuep); 368 369 /* Truncate to eight bits to accept both signed and unsigned input. */ 370 if (errmsg == NULL) 371 *valuep &= 0xFF; 372 } 373 374 return errmsg; 375 } 376 377 static const char * 378 parse_bit3 (CGEN_CPU_DESC cd, 379 const char **strp, 380 int opindex, 381 unsigned long *valuep) 382 { 383 const char *errmsg; 384 char mode = 0; 385 long count = 0; 386 unsigned long value; 387 388 if (strncmp (*strp, "%bit", 4) == 0) 389 { 390 *strp += 4; 391 mode = 1; 392 } 393 else if (strncmp (*strp, "%msbbit", 7) == 0) 394 { 395 *strp += 7; 396 mode = 1; 397 } 398 else if (strncmp (*strp, "%lsbbit", 7) == 0) 399 { 400 *strp += 7; 401 mode = 2; 402 } 403 404 errmsg = cgen_parse_unsigned_integer (cd, strp, opindex, valuep); 405 if (errmsg) 406 return errmsg; 407 408 if (mode) 409 { 410 value = * valuep; 411 if (value == 0) 412 { 413 errmsg = _("Attempt to find bit index of 0"); 414 return errmsg; 415 } 416 417 if (mode == 1) 418 { 419 count = 31; 420 while ((value & 0x80000000) == 0) 421 { 422 count--; 423 value <<= 1; 424 } 425 } 426 else if (mode == 2) 427 { 428 count = 0; 429 while ((value & 0x00000001) == 0) 430 { 431 count++; 432 value >>= 1; 433 } 434 } 435 436 *valuep = count; 437 } 438 439 return errmsg; 440 } 441 442 /* -- dis.c */ 443 444 const char * ip2k_cgen_parse_operand 445 (CGEN_CPU_DESC, int, const char **, CGEN_FIELDS *); 446 447 /* Main entry point for operand parsing. 448 449 This function is basically just a big switch statement. Earlier versions 450 used tables to look up the function to use, but 451 - if the table contains both assembler and disassembler functions then 452 the disassembler contains much of the assembler and vice-versa, 453 - there's a lot of inlining possibilities as things grow, 454 - using a switch statement avoids the function call overhead. 455 456 This function could be moved into `parse_insn_normal', but keeping it 457 separate makes clear the interface between `parse_insn_normal' and each of 458 the handlers. */ 459 460 const char * 461 ip2k_cgen_parse_operand (CGEN_CPU_DESC cd, 462 int opindex, 463 const char ** strp, 464 CGEN_FIELDS * fields) 465 { 466 const char * errmsg = NULL; 467 /* Used by scalar operands that still need to be parsed. */ 468 long junk ATTRIBUTE_UNUSED; 469 470 switch (opindex) 471 { 472 case IP2K_OPERAND_ADDR16CJP : 473 errmsg = parse_addr16_cjp (cd, strp, IP2K_OPERAND_ADDR16CJP, (unsigned long *) (& fields->f_addr16cjp)); 474 break; 475 case IP2K_OPERAND_ADDR16H : 476 errmsg = parse_addr16 (cd, strp, IP2K_OPERAND_ADDR16H, (unsigned long *) (& fields->f_imm8)); 477 break; 478 case IP2K_OPERAND_ADDR16L : 479 errmsg = parse_addr16 (cd, strp, IP2K_OPERAND_ADDR16L, (unsigned long *) (& fields->f_imm8)); 480 break; 481 case IP2K_OPERAND_ADDR16P : 482 errmsg = parse_addr16_cjp (cd, strp, IP2K_OPERAND_ADDR16P, (unsigned long *) (& fields->f_page3)); 483 break; 484 case IP2K_OPERAND_BITNO : 485 errmsg = parse_bit3 (cd, strp, IP2K_OPERAND_BITNO, (unsigned long *) (& fields->f_bitno)); 486 break; 487 case IP2K_OPERAND_CBIT : 488 errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_CBIT, (unsigned long *) (& junk)); 489 break; 490 case IP2K_OPERAND_DCBIT : 491 errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_DCBIT, (unsigned long *) (& junk)); 492 break; 493 case IP2K_OPERAND_FR : 494 errmsg = parse_fr (cd, strp, IP2K_OPERAND_FR, (unsigned long *) (& fields->f_reg)); 495 break; 496 case IP2K_OPERAND_LIT8 : 497 errmsg = parse_lit8 (cd, strp, IP2K_OPERAND_LIT8, (long *) (& fields->f_imm8)); 498 break; 499 case IP2K_OPERAND_PABITS : 500 errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_PABITS, (unsigned long *) (& junk)); 501 break; 502 case IP2K_OPERAND_RETI3 : 503 errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_RETI3, (unsigned long *) (& fields->f_reti3)); 504 break; 505 case IP2K_OPERAND_ZBIT : 506 errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_ZBIT, (unsigned long *) (& junk)); 507 break; 508 509 default : 510 /* xgettext:c-format */ 511 fprintf (stderr, _("Unrecognized field %d while parsing.\n"), opindex); 512 abort (); 513 } 514 515 return errmsg; 516 } 517 518 cgen_parse_fn * const ip2k_cgen_parse_handlers[] = 519 { 520 parse_insn_normal, 521 }; 522 523 void 524 ip2k_cgen_init_asm (CGEN_CPU_DESC cd) 525 { 526 ip2k_cgen_init_opcode_table (cd); 527 ip2k_cgen_init_ibld_table (cd); 528 cd->parse_handlers = & ip2k_cgen_parse_handlers[0]; 529 cd->parse_operand = ip2k_cgen_parse_operand; 530 #ifdef CGEN_ASM_INIT_HOOK 531 CGEN_ASM_INIT_HOOK 532 #endif 533 } 534 535 536 537 /* Regex construction routine. 538 539 This translates an opcode syntax string into a regex string, 540 by replacing any non-character syntax element (such as an 541 opcode) with the pattern '.*' 542 543 It then compiles the regex and stores it in the opcode, for 544 later use by ip2k_cgen_assemble_insn 545 546 Returns NULL for success, an error message for failure. */ 547 548 char * 549 ip2k_cgen_build_insn_regex (CGEN_INSN *insn) 550 { 551 CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn); 552 const char *mnem = CGEN_INSN_MNEMONIC (insn); 553 char rxbuf[CGEN_MAX_RX_ELEMENTS]; 554 char *rx = rxbuf; 555 const CGEN_SYNTAX_CHAR_TYPE *syn; 556 int reg_err; 557 558 syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); 559 560 /* Mnemonics come first in the syntax string. */ 561 if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) 562 return _("missing mnemonic in syntax string"); 563 ++syn; 564 565 /* Generate a case sensitive regular expression that emulates case 566 insensitive matching in the "C" locale. We cannot generate a case 567 insensitive regular expression because in Turkish locales, 'i' and 'I' 568 are not equal modulo case conversion. */ 569 570 /* Copy the literal mnemonic out of the insn. */ 571 for (; *mnem; mnem++) 572 { 573 char c = *mnem; 574 575 if (ISALPHA (c)) 576 { 577 *rx++ = '['; 578 *rx++ = TOLOWER (c); 579 *rx++ = TOUPPER (c); 580 *rx++ = ']'; 581 } 582 else 583 *rx++ = c; 584 } 585 586 /* Copy any remaining literals from the syntax string into the rx. */ 587 for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn) 588 { 589 if (CGEN_SYNTAX_CHAR_P (* syn)) 590 { 591 char c = CGEN_SYNTAX_CHAR (* syn); 592 593 switch (c) 594 { 595 /* Escape any regex metacharacters in the syntax. */ 596 case '.': case '[': case '\\': 597 case '*': case '^': case '$': 598 599 #ifdef CGEN_ESCAPE_EXTENDED_REGEX 600 case '?': case '{': case '}': 601 case '(': case ')': case '*': 602 case '|': case '+': case ']': 603 #endif 604 *rx++ = '\\'; 605 *rx++ = c; 606 break; 607 608 default: 609 if (ISALPHA (c)) 610 { 611 *rx++ = '['; 612 *rx++ = TOLOWER (c); 613 *rx++ = TOUPPER (c); 614 *rx++ = ']'; 615 } 616 else 617 *rx++ = c; 618 break; 619 } 620 } 621 else 622 { 623 /* Replace non-syntax fields with globs. */ 624 *rx++ = '.'; 625 *rx++ = '*'; 626 } 627 } 628 629 /* Trailing whitespace ok. */ 630 * rx++ = '['; 631 * rx++ = ' '; 632 * rx++ = '\t'; 633 * rx++ = ']'; 634 * rx++ = '*'; 635 636 /* But anchor it after that. */ 637 * rx++ = '$'; 638 * rx = '\0'; 639 640 CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); 641 reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB); 642 643 if (reg_err == 0) 644 return NULL; 645 else 646 { 647 static char msg[80]; 648 649 regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); 650 regfree ((regex_t *) CGEN_INSN_RX (insn)); 651 free (CGEN_INSN_RX (insn)); 652 (CGEN_INSN_RX (insn)) = NULL; 653 return msg; 654 } 655 } 656 657 658 /* Default insn parser. 659 660 The syntax string is scanned and operands are parsed and stored in FIELDS. 661 Relocs are queued as we go via other callbacks. 662 663 ??? Note that this is currently an all-or-nothing parser. If we fail to 664 parse the instruction, we return 0 and the caller will start over from 665 the beginning. Backtracking will be necessary in parsing subexpressions, 666 but that can be handled there. Not handling backtracking here may get 667 expensive in the case of the m68k. Deal with later. 668 669 Returns NULL for success, an error message for failure. */ 670 671 static const char * 672 parse_insn_normal (CGEN_CPU_DESC cd, 673 const CGEN_INSN *insn, 674 const char **strp, 675 CGEN_FIELDS *fields) 676 { 677 /* ??? Runtime added insns not handled yet. */ 678 const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn); 679 const char *str = *strp; 680 const char *errmsg; 681 const char *p; 682 const CGEN_SYNTAX_CHAR_TYPE * syn; 683 #ifdef CGEN_MNEMONIC_OPERANDS 684 /* FIXME: wip */ 685 int past_opcode_p; 686 #endif 687 688 /* For now we assume the mnemonic is first (there are no leading operands). 689 We can parse it without needing to set up operand parsing. 690 GAS's input scrubber will ensure mnemonics are lowercase, but we may 691 not be called from GAS. */ 692 p = CGEN_INSN_MNEMONIC (insn); 693 while (*p && TOLOWER (*p) == TOLOWER (*str)) 694 ++p, ++str; 695 696 if (* p) 697 return _("unrecognized instruction"); 698 699 #ifndef CGEN_MNEMONIC_OPERANDS 700 if (* str && ! ISSPACE (* str)) 701 return _("unrecognized instruction"); 702 #endif 703 704 CGEN_INIT_PARSE (cd); 705 cgen_init_parse_operand (cd); 706 #ifdef CGEN_MNEMONIC_OPERANDS 707 past_opcode_p = 0; 708 #endif 709 710 /* We don't check for (*str != '\0') here because we want to parse 711 any trailing fake arguments in the syntax string. */ 712 syn = CGEN_SYNTAX_STRING (syntax); 713 714 /* Mnemonics come first for now, ensure valid string. */ 715 if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) 716 abort (); 717 718 ++syn; 719 720 while (* syn != 0) 721 { 722 /* Non operand chars must match exactly. */ 723 if (CGEN_SYNTAX_CHAR_P (* syn)) 724 { 725 /* FIXME: While we allow for non-GAS callers above, we assume the 726 first char after the mnemonic part is a space. */ 727 /* FIXME: We also take inappropriate advantage of the fact that 728 GAS's input scrubber will remove extraneous blanks. */ 729 if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn))) 730 { 731 #ifdef CGEN_MNEMONIC_OPERANDS 732 if (CGEN_SYNTAX_CHAR(* syn) == ' ') 733 past_opcode_p = 1; 734 #endif 735 ++ syn; 736 ++ str; 737 } 738 else if (*str) 739 { 740 /* Syntax char didn't match. Can't be this insn. */ 741 static char msg [80]; 742 743 /* xgettext:c-format */ 744 sprintf (msg, _("syntax error (expected char `%c', found `%c')"), 745 CGEN_SYNTAX_CHAR(*syn), *str); 746 return msg; 747 } 748 else 749 { 750 /* Ran out of input. */ 751 static char msg [80]; 752 753 /* xgettext:c-format */ 754 sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"), 755 CGEN_SYNTAX_CHAR(*syn)); 756 return msg; 757 } 758 continue; 759 } 760 761 /* We have an operand of some sort. */ 762 errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), 763 &str, fields); 764 if (errmsg) 765 return errmsg; 766 767 /* Done with this operand, continue with next one. */ 768 ++ syn; 769 } 770 771 /* If we're at the end of the syntax string, we're done. */ 772 if (* syn == 0) 773 { 774 /* FIXME: For the moment we assume a valid `str' can only contain 775 blanks now. IE: We needn't try again with a longer version of 776 the insn and it is assumed that longer versions of insns appear 777 before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */ 778 while (ISSPACE (* str)) 779 ++ str; 780 781 if (* str != '\0') 782 return _("junk at end of line"); /* FIXME: would like to include `str' */ 783 784 return NULL; 785 } 786 787 /* We couldn't parse it. */ 788 return _("unrecognized instruction"); 789 } 790 791 /* Main entry point. 792 This routine is called for each instruction to be assembled. 793 STR points to the insn to be assembled. 794 We assume all necessary tables have been initialized. 795 The assembled instruction, less any fixups, is stored in BUF. 796 Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value 797 still needs to be converted to target byte order, otherwise BUF is an array 798 of bytes in target byte order. 799 The result is a pointer to the insn's entry in the opcode table, 800 or NULL if an error occured (an error message will have already been 801 printed). 802 803 Note that when processing (non-alias) macro-insns, 804 this function recurses. 805 806 ??? It's possible to make this cpu-independent. 807 One would have to deal with a few minor things. 808 At this point in time doing so would be more of a curiosity than useful 809 [for example this file isn't _that_ big], but keeping the possibility in 810 mind helps keep the design clean. */ 811 812 const CGEN_INSN * 813 ip2k_cgen_assemble_insn (CGEN_CPU_DESC cd, 814 const char *str, 815 CGEN_FIELDS *fields, 816 CGEN_INSN_BYTES_PTR buf, 817 char **errmsg) 818 { 819 const char *start; 820 CGEN_INSN_LIST *ilist; 821 const char *parse_errmsg = NULL; 822 const char *insert_errmsg = NULL; 823 int recognized_mnemonic = 0; 824 825 /* Skip leading white space. */ 826 while (ISSPACE (* str)) 827 ++ str; 828 829 /* The instructions are stored in hashed lists. 830 Get the first in the list. */ 831 ilist = CGEN_ASM_LOOKUP_INSN (cd, str); 832 833 /* Keep looking until we find a match. */ 834 start = str; 835 for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) 836 { 837 const CGEN_INSN *insn = ilist->insn; 838 recognized_mnemonic = 1; 839 840 #ifdef CGEN_VALIDATE_INSN_SUPPORTED 841 /* Not usually needed as unsupported opcodes 842 shouldn't be in the hash lists. */ 843 /* Is this insn supported by the selected cpu? */ 844 if (! ip2k_cgen_insn_supported (cd, insn)) 845 continue; 846 #endif 847 /* If the RELAXED attribute is set, this is an insn that shouldn't be 848 chosen immediately. Instead, it is used during assembler/linker 849 relaxation if possible. */ 850 if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0) 851 continue; 852 853 str = start; 854 855 /* Skip this insn if str doesn't look right lexically. */ 856 if (CGEN_INSN_RX (insn) != NULL && 857 regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) 858 continue; 859 860 /* Allow parse/insert handlers to obtain length of insn. */ 861 CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); 862 863 parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields); 864 if (parse_errmsg != NULL) 865 continue; 866 867 /* ??? 0 is passed for `pc'. */ 868 insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf, 869 (bfd_vma) 0); 870 if (insert_errmsg != NULL) 871 continue; 872 873 /* It is up to the caller to actually output the insn and any 874 queued relocs. */ 875 return insn; 876 } 877 878 { 879 static char errbuf[150]; 880 #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS 881 const char *tmp_errmsg; 882 883 /* If requesting verbose error messages, use insert_errmsg. 884 Failing that, use parse_errmsg. */ 885 tmp_errmsg = (insert_errmsg ? insert_errmsg : 886 parse_errmsg ? parse_errmsg : 887 recognized_mnemonic ? 888 _("unrecognized form of instruction") : 889 _("unrecognized instruction")); 890 891 if (strlen (start) > 50) 892 /* xgettext:c-format */ 893 sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start); 894 else 895 /* xgettext:c-format */ 896 sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start); 897 #else 898 if (strlen (start) > 50) 899 /* xgettext:c-format */ 900 sprintf (errbuf, _("bad instruction `%.50s...'"), start); 901 else 902 /* xgettext:c-format */ 903 sprintf (errbuf, _("bad instruction `%.50s'"), start); 904 #endif 905 906 *errmsg = errbuf; 907 return NULL; 908 } 909 } 910