Commit | Line | Data |
---|---|---|
a40cbfa3 NC |
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 | PARAMS ((CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *)); | |
47 | \f | |
48 | /* -- assembler routines inserted here. */ | |
49 | ||
50 | /* -- asm.c */ | |
51 | ||
52 | static const char * | |
53 | parse_fr (cd, strp, opindex, valuep) | |
54 | CGEN_CPU_DESC cd; | |
55 | const char **strp; | |
56 | int opindex; | |
57 | long *valuep; | |
58 | { | |
59 | const char *errmsg; | |
60 | char *old_strp; | |
61 | char *afteroffset; | |
62 | enum cgen_parse_operand_result result_type; | |
63 | bfd_vma value; | |
64 | extern CGEN_KEYWORD ip2k_cgen_opval_register_names; | |
65 | long tempvalue; | |
66 | ||
67 | old_strp = *strp; | |
68 | afteroffset = NULL; | |
69 | ||
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 | ||
88 | /* Attempt parse as register keyword. */ | |
89 | /* old_strp = *strp; */ | |
90 | ||
91 | errmsg = cgen_parse_keyword (cd, strp, & ip2k_cgen_opval_register_names, valuep); | |
92 | if ( *strp != NULL ) | |
93 | if (errmsg == NULL) | |
94 | return errmsg; | |
95 | ||
96 | /* Attempt to parse for "(IP)" */ | |
97 | afteroffset = strstr(*strp,"(IP)"); | |
98 | ||
99 | if ( afteroffset == NULL) | |
100 | { | |
101 | /* Make sure it's not in lower case */ | |
102 | afteroffset = strstr(*strp,"(ip)"); | |
103 | } | |
104 | ||
105 | if ( afteroffset != NULL ) | |
106 | { | |
107 | if ( afteroffset != *strp ) | |
108 | { | |
109 | /* Invalid offset present.*/ | |
110 | errmsg = _("offset(IP) is not a valid form"); | |
111 | return errmsg; | |
112 | } | |
113 | else | |
114 | { | |
115 | *strp += 4; | |
116 | *valuep = 0; | |
117 | errmsg = NULL; | |
118 | return errmsg; | |
119 | } | |
120 | } | |
121 | ||
122 | /* Attempt to parse for DP. ex: mov w, offset(DP) */ | |
123 | /* mov offset(DP),w */ | |
124 | ||
125 | /* Try parsing it as an address and see what comes back */ | |
126 | ||
127 | afteroffset = strstr(*strp,"(DP)"); | |
128 | ||
129 | if ( afteroffset == NULL) | |
130 | { | |
131 | /* Maybe it's in lower case */ | |
132 | afteroffset = strstr(*strp,"(dp)"); | |
133 | } | |
134 | ||
135 | if ( afteroffset != NULL ) | |
136 | { | |
137 | if ( afteroffset == *strp ) | |
138 | { | |
139 | /* No offset present. Use 0 by default. */ | |
140 | tempvalue = 0; | |
141 | errmsg = NULL; | |
142 | } | |
143 | else | |
144 | { | |
145 | errmsg = cgen_parse_address (cd, strp, opindex, BFD_RELOC_IP2K_FR_OFFSET, | |
146 | & result_type, & tempvalue); | |
147 | } | |
148 | ||
149 | if (errmsg == NULL) | |
150 | { | |
151 | if ( (tempvalue >= 0) && (tempvalue <= 127) ) | |
152 | { | |
153 | /* Value is ok. Fix up the first 2 bits and return */ | |
154 | *valuep = 0x0100 | tempvalue; | |
155 | *strp += 4; /* skip over the (DP) in *strp */ | |
156 | return errmsg; | |
157 | } else | |
158 | { | |
159 | /* Found something there in front of (DP) but it's out of range. */ | |
160 | errmsg = _("(DP) offset out of range."); | |
161 | return errmsg; | |
162 | } | |
163 | ||
164 | } | |
165 | } | |
166 | ||
167 | ||
168 | /* Attempt to parse for SP. ex: mov w, offset(SP) */ | |
169 | /* mov offset(SP), w */ | |
170 | ||
171 | ||
172 | afteroffset = strstr(*strp,"(SP)"); | |
173 | ||
174 | if (afteroffset == NULL) | |
175 | { | |
176 | /* Maybe it's in lower case. */ | |
177 | afteroffset = strstr(*strp, "(sp)"); | |
178 | } | |
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 | { | |
190 | errmsg = cgen_parse_address (cd, strp, opindex, BFD_RELOC_IP2K_FR_OFFSET, | |
191 | & result_type, & tempvalue); | |
192 | } | |
193 | if (errmsg == NULL) | |
194 | { | |
195 | if ( (tempvalue >= 0) && (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 | } else | |
202 | { | |
203 | /* Found something there in front of (SP) but it's out of range. */ | |
204 | errmsg = _("(SP) offset out of range."); | |
205 | return errmsg; | |
206 | } | |
207 | ||
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 | ||
222 | if (**strp == '(') | |
223 | { | |
224 | errmsg = _("illegal use of parentheses"); | |
225 | } | |
226 | /* if a numeric value is specified, ensure that it is between 1 and 255 */ | |
227 | else if (result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) | |
228 | { | |
229 | if (value < 0x1 || value > 0xff) | |
230 | errmsg = _("operand out of range (not between 1 and 255)"); | |
231 | } | |
232 | } | |
233 | return errmsg; | |
234 | } | |
235 | ||
236 | static const char * | |
237 | parse_addr16 (cd, strp, opindex, valuep) | |
238 | CGEN_CPU_DESC cd; | |
239 | const char **strp; | |
240 | int opindex; | |
241 | long *valuep; | |
242 | { | |
243 | const char *errmsg; | |
244 | enum cgen_parse_operand_result result_type; | |
245 | bfd_reloc_code_real_type code = BFD_RELOC_NONE; | |
246 | long value; | |
247 | ||
248 | if ( opindex == (CGEN_OPERAND_TYPE)IP2K_OPERAND_ADDR16H ) | |
249 | code = BFD_RELOC_IP2K_HI8DATA; | |
250 | else if ( opindex == (CGEN_OPERAND_TYPE)IP2K_OPERAND_ADDR16L ) | |
251 | code = BFD_RELOC_IP2K_LO8DATA; | |
252 | else | |
253 | { | |
254 | /* Something is very wrong. opindex has to be one of the above. */ | |
255 | errmsg = _("parse_addr16: invalid opindex."); | |
256 | return errmsg; | |
257 | } | |
258 | ||
259 | errmsg = cgen_parse_address (cd, strp, opindex, code, | |
260 | & result_type, & value); | |
261 | if (errmsg == NULL) | |
262 | { | |
263 | /* We either have a relocation or a number now. */ | |
264 | if ( result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER ) | |
265 | { | |
266 | /* We got a number back. */ | |
267 | if ( code == BFD_RELOC_IP2K_HI8DATA ) | |
268 | value >>= 8; | |
269 | else /* code = BFD_RELOC_IP2K_LOW8DATA */ | |
270 | value &= 0x00FF; | |
271 | } | |
272 | *valuep = value; | |
273 | } | |
274 | ||
275 | return errmsg; | |
276 | } | |
277 | ||
278 | ||
279 | static const char * | |
280 | parse_addr16_p (cd, strp, opindex, valuep) | |
281 | CGEN_CPU_DESC cd; | |
282 | const char **strp; | |
283 | int opindex; | |
284 | long *valuep; | |
285 | { | |
286 | const char *errmsg; | |
287 | enum cgen_parse_operand_result result_type; | |
288 | bfd_reloc_code_real_type code = BFD_RELOC_IP2K_PAGE3; | |
289 | long value; | |
290 | ||
291 | errmsg = cgen_parse_address (cd, strp, opindex, code, | |
292 | & result_type, & value); | |
293 | if (errmsg == NULL) | |
294 | { | |
295 | if ( result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER ) | |
296 | *valuep = (value >> 13) & 0x7; | |
297 | else if ( result_type == CGEN_PARSE_OPERAND_RESULT_QUEUED ) | |
298 | *valuep = value; | |
299 | } | |
300 | return errmsg; | |
301 | } | |
302 | ||
303 | ||
304 | static const char * | |
305 | parse_addr16_cjp (cd, strp, opindex, valuep) | |
306 | CGEN_CPU_DESC cd; | |
307 | const char **strp; | |
308 | int opindex; | |
309 | long *valuep; | |
310 | { | |
311 | const char *errmsg; | |
312 | enum cgen_parse_operand_result result_type; | |
313 | bfd_reloc_code_real_type code = BFD_RELOC_NONE; | |
314 | long value; | |
315 | ||
316 | if ( opindex == (CGEN_OPERAND_TYPE)IP2K_OPERAND_ADDR16CJP ) | |
317 | code = BFD_RELOC_IP2K_ADDR16CJP; | |
318 | else if ( opindex == (CGEN_OPERAND_TYPE)IP2K_OPERAND_ADDR16P ) | |
319 | code = BFD_RELOC_IP2K_PAGE3; | |
320 | ||
321 | errmsg = cgen_parse_address (cd, strp, opindex, code, | |
322 | & result_type, & value); | |
323 | if (errmsg == NULL) | |
324 | { | |
325 | if ( result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER ) | |
326 | { | |
327 | if ( (value & 0x1) == 0) /* If the address is even .... */ | |
328 | { | |
329 | if ( opindex == (CGEN_OPERAND_TYPE)IP2K_OPERAND_ADDR16CJP ) | |
330 | *valuep = (value >> 1) & 0x1FFF; /* Should mask be 1FFF? */ | |
331 | else if ( opindex == (CGEN_OPERAND_TYPE)IP2K_OPERAND_ADDR16P ) | |
332 | *valuep = (value >> 14) & 0x7; | |
333 | } | |
334 | else | |
335 | errmsg = _("Byte address required. - must be even."); | |
336 | }else if ( result_type == CGEN_PARSE_OPERAND_RESULT_QUEUED ) | |
337 | { | |
338 | /* This will happen for things like (s2-s1) where s2 and s1 */ | |
339 | /* are labels. */ | |
340 | *valuep = value; | |
341 | } | |
342 | else | |
343 | errmsg = _("cgen_parse_address returned a symbol. Literal required."); | |
344 | } | |
345 | return errmsg; | |
346 | } | |
347 | ||
348 | ||
349 | static const char * | |
350 | parse_lit8 (cd, strp, opindex, valuep) | |
351 | CGEN_CPU_DESC cd; | |
352 | const char **strp; | |
353 | int opindex; | |
354 | long *valuep; | |
355 | { | |
356 | const char *errmsg; | |
357 | enum cgen_parse_operand_result result_type; | |
358 | bfd_reloc_code_real_type code = BFD_RELOC_NONE; | |
359 | long value; | |
360 | ||
361 | /* Parse %OP relocating operators. */ | |
362 | if (strncmp (*strp, "%bank", 5) == 0) | |
363 | { | |
364 | *strp += 5; | |
365 | code = BFD_RELOC_IP2K_BANK; | |
366 | } | |
367 | else if (strncmp (*strp, "%lo8data", 8) == 0) | |
368 | { | |
369 | *strp += 8; | |
370 | code = BFD_RELOC_IP2K_LO8DATA; | |
371 | } | |
372 | else if (strncmp (*strp, "%hi8data", 8) == 0) | |
373 | { | |
374 | *strp += 8; | |
375 | code = BFD_RELOC_IP2K_HI8DATA; | |
376 | } | |
377 | else if (strncmp (*strp, "%ex8data", 8) == 0) | |
378 | { | |
379 | *strp += 8; | |
380 | code = BFD_RELOC_IP2K_EX8DATA; | |
381 | } | |
382 | else if (strncmp (*strp, "%lo8insn", 8) == 0) | |
383 | { | |
384 | *strp += 8; | |
385 | code = BFD_RELOC_IP2K_LO8INSN; | |
386 | } | |
387 | else if (strncmp (*strp, "%hi8insn", 8) == 0) | |
388 | { | |
389 | *strp += 8; | |
390 | code = BFD_RELOC_IP2K_HI8INSN; | |
391 | } | |
392 | ||
393 | ||
394 | /* Parse %op operand. */ | |
395 | if (code != BFD_RELOC_NONE) | |
396 | { | |
397 | errmsg = cgen_parse_address (cd, strp, opindex, code, | |
398 | & result_type, & value); | |
399 | if ((errmsg == NULL) && | |
400 | (result_type != CGEN_PARSE_OPERAND_RESULT_QUEUED)) | |
401 | errmsg = _("%operator operand is not a symbol"); | |
402 | ||
403 | *valuep = value; | |
404 | } | |
405 | /* Parse as a number. */ | |
406 | else | |
407 | { | |
408 | errmsg = cgen_parse_signed_integer (cd, strp, opindex, valuep); | |
409 | ||
410 | /* Truncate to eight bits to accept both signed and unsigned input. */ | |
411 | if (errmsg == NULL) | |
412 | *valuep &= 0xFF; | |
413 | } | |
414 | ||
415 | return errmsg; | |
416 | } | |
417 | ||
418 | static const char * | |
419 | parse_bit3 (cd, strp, opindex, valuep) | |
420 | CGEN_CPU_DESC cd; | |
421 | const char **strp; | |
422 | int opindex; | |
423 | long *valuep; | |
424 | { | |
425 | const char *errmsg; | |
426 | char mode = 0; | |
427 | long count = 0; | |
428 | unsigned long value; | |
429 | ||
430 | if (strncmp (*strp, "%bit", 4) == 0) | |
431 | { | |
432 | *strp += 4; | |
433 | mode = 1; | |
434 | } | |
435 | else if (strncmp (*strp, "%msbbit", 7) == 0) | |
436 | { | |
437 | *strp += 7; | |
438 | mode = 1; | |
439 | } | |
440 | else if (strncmp (*strp, "%lsbbit", 7) == 0) | |
441 | { | |
442 | *strp += 7; | |
443 | mode = 2; | |
444 | } | |
445 | ||
446 | errmsg = cgen_parse_signed_integer (cd, strp, opindex, valuep); | |
447 | if (errmsg) { | |
448 | return errmsg; | |
449 | } | |
450 | ||
451 | if (mode) { | |
452 | value = (unsigned long) *valuep; | |
453 | if (value == 0) { | |
454 | errmsg = _("Attempt to find bit index of 0"); | |
455 | return errmsg; | |
456 | } | |
457 | ||
458 | if (mode == 1) { | |
459 | count = 31; | |
460 | while ((value & 0x80000000) == 0) { | |
461 | count--; | |
462 | value <<= 1; | |
463 | } | |
464 | } else if (mode == 2) { | |
465 | count = 0; | |
466 | while ((value & 0x00000001) == 0) { | |
467 | count++; | |
468 | value >>= 1; | |
469 | } | |
470 | } | |
471 | ||
472 | *valuep = count; | |
473 | } | |
474 | ||
475 | return errmsg; | |
476 | } | |
477 | ||
478 | ||
479 | /* -- dis.c */ | |
480 | ||
481 | const char * ip2k_cgen_parse_operand | |
482 | PARAMS ((CGEN_CPU_DESC, int, const char **, CGEN_FIELDS *)); | |
483 | ||
484 | /* Main entry point for operand parsing. | |
485 | ||
486 | This function is basically just a big switch statement. Earlier versions | |
487 | used tables to look up the function to use, but | |
488 | - if the table contains both assembler and disassembler functions then | |
489 | the disassembler contains much of the assembler and vice-versa, | |
490 | - there's a lot of inlining possibilities as things grow, | |
491 | - using a switch statement avoids the function call overhead. | |
492 | ||
493 | This function could be moved into `parse_insn_normal', but keeping it | |
494 | separate makes clear the interface between `parse_insn_normal' and each of | |
495 | the handlers. */ | |
496 | ||
497 | const char * | |
498 | ip2k_cgen_parse_operand (cd, opindex, strp, fields) | |
499 | CGEN_CPU_DESC cd; | |
500 | int opindex; | |
501 | const char ** strp; | |
502 | CGEN_FIELDS * fields; | |
503 | { | |
504 | const char * errmsg = NULL; | |
505 | /* Used by scalar operands that still need to be parsed. */ | |
506 | long junk ATTRIBUTE_UNUSED; | |
507 | ||
508 | switch (opindex) | |
509 | { | |
510 | case IP2K_OPERAND_ADDR16CJP : | |
511 | errmsg = parse_addr16_cjp (cd, strp, IP2K_OPERAND_ADDR16CJP, &fields->f_addr16cjp); | |
512 | break; | |
513 | case IP2K_OPERAND_ADDR16H : | |
514 | errmsg = parse_addr16 (cd, strp, IP2K_OPERAND_ADDR16H, &fields->f_imm8); | |
515 | break; | |
516 | case IP2K_OPERAND_ADDR16L : | |
517 | errmsg = parse_addr16 (cd, strp, IP2K_OPERAND_ADDR16L, &fields->f_imm8); | |
518 | break; | |
519 | case IP2K_OPERAND_ADDR16P : | |
520 | errmsg = parse_addr16_cjp (cd, strp, IP2K_OPERAND_ADDR16P, &fields->f_page3); | |
521 | break; | |
522 | case IP2K_OPERAND_BITNO : | |
523 | errmsg = parse_bit3 (cd, strp, IP2K_OPERAND_BITNO, &fields->f_bitno); | |
524 | break; | |
525 | case IP2K_OPERAND_CBIT : | |
526 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_CBIT, &junk); | |
527 | break; | |
528 | case IP2K_OPERAND_DCBIT : | |
529 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_DCBIT, &junk); | |
530 | break; | |
531 | case IP2K_OPERAND_FR : | |
532 | errmsg = parse_fr (cd, strp, IP2K_OPERAND_FR, &fields->f_reg); | |
533 | break; | |
534 | case IP2K_OPERAND_LIT8 : | |
535 | errmsg = parse_lit8 (cd, strp, IP2K_OPERAND_LIT8, &fields->f_imm8); | |
536 | break; | |
537 | case IP2K_OPERAND_PABITS : | |
538 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_PABITS, &junk); | |
539 | break; | |
540 | case IP2K_OPERAND_RETI3 : | |
541 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_RETI3, &fields->f_reti3); | |
542 | break; | |
543 | case IP2K_OPERAND_ZBIT : | |
544 | errmsg = cgen_parse_unsigned_integer (cd, strp, IP2K_OPERAND_ZBIT, &junk); | |
545 | break; | |
546 | ||
547 | default : | |
548 | /* xgettext:c-format */ | |
549 | fprintf (stderr, _("Unrecognized field %d while parsing.\n"), opindex); | |
550 | abort (); | |
551 | } | |
552 | ||
553 | return errmsg; | |
554 | } | |
555 | ||
556 | cgen_parse_fn * const ip2k_cgen_parse_handlers[] = | |
557 | { | |
558 | parse_insn_normal, | |
559 | }; | |
560 | ||
561 | void | |
562 | ip2k_cgen_init_asm (cd) | |
563 | CGEN_CPU_DESC cd; | |
564 | { | |
565 | ip2k_cgen_init_opcode_table (cd); | |
566 | ip2k_cgen_init_ibld_table (cd); | |
567 | cd->parse_handlers = & ip2k_cgen_parse_handlers[0]; | |
568 | cd->parse_operand = ip2k_cgen_parse_operand; | |
569 | } | |
570 | ||
571 | \f | |
572 | ||
573 | /* Regex construction routine. | |
574 | ||
575 | This translates an opcode syntax string into a regex string, | |
576 | by replacing any non-character syntax element (such as an | |
577 | opcode) with the pattern '.*' | |
578 | ||
579 | It then compiles the regex and stores it in the opcode, for | |
580 | later use by ip2k_cgen_assemble_insn | |
581 | ||
582 | Returns NULL for success, an error message for failure. */ | |
583 | ||
584 | char * | |
585 | ip2k_cgen_build_insn_regex (insn) | |
586 | CGEN_INSN *insn; | |
587 | { | |
588 | CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn); | |
589 | const char *mnem = CGEN_INSN_MNEMONIC (insn); | |
590 | char rxbuf[CGEN_MAX_RX_ELEMENTS]; | |
591 | char *rx = rxbuf; | |
592 | const CGEN_SYNTAX_CHAR_TYPE *syn; | |
593 | int reg_err; | |
594 | ||
595 | syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); | |
596 | ||
597 | /* Mnemonics come first in the syntax string. */ | |
598 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) | |
599 | return _("missing mnemonic in syntax string"); | |
600 | ++syn; | |
601 | ||
602 | /* Generate a case sensitive regular expression that emulates case | |
603 | insensitive matching in the "C" locale. We cannot generate a case | |
604 | insensitive regular expression because in Turkish locales, 'i' and 'I' | |
605 | are not equal modulo case conversion. */ | |
606 | ||
607 | /* Copy the literal mnemonic out of the insn. */ | |
608 | for (; *mnem; mnem++) | |
609 | { | |
610 | char c = *mnem; | |
611 | ||
612 | if (ISALPHA (c)) | |
613 | { | |
614 | *rx++ = '['; | |
615 | *rx++ = TOLOWER (c); | |
616 | *rx++ = TOUPPER (c); | |
617 | *rx++ = ']'; | |
618 | } | |
619 | else | |
620 | *rx++ = c; | |
621 | } | |
622 | ||
623 | /* Copy any remaining literals from the syntax string into the rx. */ | |
624 | for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn) | |
625 | { | |
626 | if (CGEN_SYNTAX_CHAR_P (* syn)) | |
627 | { | |
628 | char c = CGEN_SYNTAX_CHAR (* syn); | |
629 | ||
630 | switch (c) | |
631 | { | |
632 | /* Escape any regex metacharacters in the syntax. */ | |
633 | case '.': case '[': case '\\': | |
634 | case '*': case '^': case '$': | |
635 | ||
636 | #ifdef CGEN_ESCAPE_EXTENDED_REGEX | |
637 | case '?': case '{': case '}': | |
638 | case '(': case ')': case '*': | |
639 | case '|': case '+': case ']': | |
640 | #endif | |
641 | *rx++ = '\\'; | |
642 | *rx++ = c; | |
643 | break; | |
644 | ||
645 | default: | |
646 | if (ISALPHA (c)) | |
647 | { | |
648 | *rx++ = '['; | |
649 | *rx++ = TOLOWER (c); | |
650 | *rx++ = TOUPPER (c); | |
651 | *rx++ = ']'; | |
652 | } | |
653 | else | |
654 | *rx++ = c; | |
655 | break; | |
656 | } | |
657 | } | |
658 | else | |
659 | { | |
660 | /* Replace non-syntax fields with globs. */ | |
661 | *rx++ = '.'; | |
662 | *rx++ = '*'; | |
663 | } | |
664 | } | |
665 | ||
666 | /* Trailing whitespace ok. */ | |
667 | * rx++ = '['; | |
668 | * rx++ = ' '; | |
669 | * rx++ = '\t'; | |
670 | * rx++ = ']'; | |
671 | * rx++ = '*'; | |
672 | ||
673 | /* But anchor it after that. */ | |
674 | * rx++ = '$'; | |
675 | * rx = '\0'; | |
676 | ||
677 | CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); | |
678 | reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB); | |
679 | ||
680 | if (reg_err == 0) | |
681 | return NULL; | |
682 | else | |
683 | { | |
684 | static char msg[80]; | |
685 | ||
686 | regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); | |
687 | regfree ((regex_t *) CGEN_INSN_RX (insn)); | |
688 | free (CGEN_INSN_RX (insn)); | |
689 | (CGEN_INSN_RX (insn)) = NULL; | |
690 | return msg; | |
691 | } | |
692 | } | |
693 | ||
694 | \f | |
695 | /* Default insn parser. | |
696 | ||
697 | The syntax string is scanned and operands are parsed and stored in FIELDS. | |
698 | Relocs are queued as we go via other callbacks. | |
699 | ||
700 | ??? Note that this is currently an all-or-nothing parser. If we fail to | |
701 | parse the instruction, we return 0 and the caller will start over from | |
702 | the beginning. Backtracking will be necessary in parsing subexpressions, | |
703 | but that can be handled there. Not handling backtracking here may get | |
704 | expensive in the case of the m68k. Deal with later. | |
705 | ||
706 | Returns NULL for success, an error message for failure. */ | |
707 | ||
708 | static const char * | |
709 | parse_insn_normal (cd, insn, strp, fields) | |
710 | CGEN_CPU_DESC cd; | |
711 | const CGEN_INSN *insn; | |
712 | const char **strp; | |
713 | CGEN_FIELDS *fields; | |
714 | { | |
715 | /* ??? Runtime added insns not handled yet. */ | |
716 | const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn); | |
717 | const char *str = *strp; | |
718 | const char *errmsg; | |
719 | const char *p; | |
720 | const CGEN_SYNTAX_CHAR_TYPE * syn; | |
721 | #ifdef CGEN_MNEMONIC_OPERANDS | |
722 | /* FIXME: wip */ | |
723 | int past_opcode_p; | |
724 | #endif | |
725 | ||
726 | /* For now we assume the mnemonic is first (there are no leading operands). | |
727 | We can parse it without needing to set up operand parsing. | |
728 | GAS's input scrubber will ensure mnemonics are lowercase, but we may | |
729 | not be called from GAS. */ | |
730 | p = CGEN_INSN_MNEMONIC (insn); | |
731 | while (*p && TOLOWER (*p) == TOLOWER (*str)) | |
732 | ++p, ++str; | |
733 | ||
734 | if (* p) | |
735 | return _("unrecognized instruction"); | |
736 | ||
737 | #ifndef CGEN_MNEMONIC_OPERANDS | |
738 | if (* str && ! ISSPACE (* str)) | |
739 | return _("unrecognized instruction"); | |
740 | #endif | |
741 | ||
742 | CGEN_INIT_PARSE (cd); | |
743 | cgen_init_parse_operand (cd); | |
744 | #ifdef CGEN_MNEMONIC_OPERANDS | |
745 | past_opcode_p = 0; | |
746 | #endif | |
747 | ||
748 | /* We don't check for (*str != '\0') here because we want to parse | |
749 | any trailing fake arguments in the syntax string. */ | |
750 | syn = CGEN_SYNTAX_STRING (syntax); | |
751 | ||
752 | /* Mnemonics come first for now, ensure valid string. */ | |
753 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) | |
754 | abort (); | |
755 | ||
756 | ++syn; | |
757 | ||
758 | while (* syn != 0) | |
759 | { | |
760 | /* Non operand chars must match exactly. */ | |
761 | if (CGEN_SYNTAX_CHAR_P (* syn)) | |
762 | { | |
763 | /* FIXME: While we allow for non-GAS callers above, we assume the | |
764 | first char after the mnemonic part is a space. */ | |
765 | /* FIXME: We also take inappropriate advantage of the fact that | |
766 | GAS's input scrubber will remove extraneous blanks. */ | |
767 | if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn))) | |
768 | { | |
769 | #ifdef CGEN_MNEMONIC_OPERANDS | |
770 | if (CGEN_SYNTAX_CHAR(* syn) == ' ') | |
771 | past_opcode_p = 1; | |
772 | #endif | |
773 | ++ syn; | |
774 | ++ str; | |
775 | } | |
776 | else if (*str) | |
777 | { | |
778 | /* Syntax char didn't match. Can't be this insn. */ | |
779 | static char msg [80]; | |
780 | ||
781 | /* xgettext:c-format */ | |
782 | sprintf (msg, _("syntax error (expected char `%c', found `%c')"), | |
783 | CGEN_SYNTAX_CHAR(*syn), *str); | |
784 | return msg; | |
785 | } | |
786 | else | |
787 | { | |
788 | /* Ran out of input. */ | |
789 | static char msg [80]; | |
790 | ||
791 | /* xgettext:c-format */ | |
792 | sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"), | |
793 | CGEN_SYNTAX_CHAR(*syn)); | |
794 | return msg; | |
795 | } | |
796 | continue; | |
797 | } | |
798 | ||
799 | /* We have an operand of some sort. */ | |
800 | errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), | |
801 | &str, fields); | |
802 | if (errmsg) | |
803 | return errmsg; | |
804 | ||
805 | /* Done with this operand, continue with next one. */ | |
806 | ++ syn; | |
807 | } | |
808 | ||
809 | /* If we're at the end of the syntax string, we're done. */ | |
810 | if (* syn == 0) | |
811 | { | |
812 | /* FIXME: For the moment we assume a valid `str' can only contain | |
813 | blanks now. IE: We needn't try again with a longer version of | |
814 | the insn and it is assumed that longer versions of insns appear | |
815 | before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */ | |
816 | while (ISSPACE (* str)) | |
817 | ++ str; | |
818 | ||
819 | if (* str != '\0') | |
820 | return _("junk at end of line"); /* FIXME: would like to include `str' */ | |
821 | ||
822 | return NULL; | |
823 | } | |
824 | ||
825 | /* We couldn't parse it. */ | |
826 | return _("unrecognized instruction"); | |
827 | } | |
828 | \f | |
829 | /* Main entry point. | |
830 | This routine is called for each instruction to be assembled. | |
831 | STR points to the insn to be assembled. | |
832 | We assume all necessary tables have been initialized. | |
833 | The assembled instruction, less any fixups, is stored in BUF. | |
834 | Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value | |
835 | still needs to be converted to target byte order, otherwise BUF is an array | |
836 | of bytes in target byte order. | |
837 | The result is a pointer to the insn's entry in the opcode table, | |
838 | or NULL if an error occured (an error message will have already been | |
839 | printed). | |
840 | ||
841 | Note that when processing (non-alias) macro-insns, | |
842 | this function recurses. | |
843 | ||
844 | ??? It's possible to make this cpu-independent. | |
845 | One would have to deal with a few minor things. | |
846 | At this point in time doing so would be more of a curiosity than useful | |
847 | [for example this file isn't _that_ big], but keeping the possibility in | |
848 | mind helps keep the design clean. */ | |
849 | ||
850 | const CGEN_INSN * | |
851 | ip2k_cgen_assemble_insn (cd, str, fields, buf, errmsg) | |
852 | CGEN_CPU_DESC cd; | |
853 | const char *str; | |
854 | CGEN_FIELDS *fields; | |
855 | CGEN_INSN_BYTES_PTR buf; | |
856 | char **errmsg; | |
857 | { | |
858 | const char *start; | |
859 | CGEN_INSN_LIST *ilist; | |
860 | const char *parse_errmsg = NULL; | |
861 | const char *insert_errmsg = NULL; | |
862 | int recognized_mnemonic = 0; | |
863 | ||
864 | /* Skip leading white space. */ | |
865 | while (ISSPACE (* str)) | |
866 | ++ str; | |
867 | ||
868 | /* The instructions are stored in hashed lists. | |
869 | Get the first in the list. */ | |
870 | ilist = CGEN_ASM_LOOKUP_INSN (cd, str); | |
871 | ||
872 | /* Keep looking until we find a match. */ | |
873 | start = str; | |
874 | for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) | |
875 | { | |
876 | const CGEN_INSN *insn = ilist->insn; | |
877 | recognized_mnemonic = 1; | |
878 | ||
879 | #ifdef CGEN_VALIDATE_INSN_SUPPORTED | |
880 | /* Not usually needed as unsupported opcodes | |
881 | shouldn't be in the hash lists. */ | |
882 | /* Is this insn supported by the selected cpu? */ | |
883 | if (! ip2k_cgen_insn_supported (cd, insn)) | |
884 | continue; | |
885 | #endif | |
886 | /* If the RELAX attribute is set, this is an insn that shouldn't be | |
887 | chosen immediately. Instead, it is used during assembler/linker | |
888 | relaxation if possible. */ | |
889 | if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAX) != 0) | |
890 | continue; | |
891 | ||
892 | str = start; | |
893 | ||
894 | /* Skip this insn if str doesn't look right lexically. */ | |
895 | if (CGEN_INSN_RX (insn) != NULL && | |
896 | regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) | |
897 | continue; | |
898 | ||
899 | /* Allow parse/insert handlers to obtain length of insn. */ | |
900 | CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); | |
901 | ||
902 | parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields); | |
903 | if (parse_errmsg != NULL) | |
904 | continue; | |
905 | ||
906 | /* ??? 0 is passed for `pc'. */ | |
907 | insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf, | |
908 | (bfd_vma) 0); | |
909 | if (insert_errmsg != NULL) | |
910 | continue; | |
911 | ||
912 | /* It is up to the caller to actually output the insn and any | |
913 | queued relocs. */ | |
914 | return insn; | |
915 | } | |
916 | ||
917 | { | |
918 | static char errbuf[150]; | |
919 | #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS | |
920 | const char *tmp_errmsg; | |
921 | ||
922 | /* If requesting verbose error messages, use insert_errmsg. | |
923 | Failing that, use parse_errmsg. */ | |
924 | tmp_errmsg = (insert_errmsg ? insert_errmsg : | |
925 | parse_errmsg ? parse_errmsg : | |
926 | recognized_mnemonic ? | |
927 | _("unrecognized form of instruction") : | |
928 | _("unrecognized instruction")); | |
929 | ||
930 | if (strlen (start) > 50) | |
931 | /* xgettext:c-format */ | |
932 | sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start); | |
933 | else | |
934 | /* xgettext:c-format */ | |
935 | sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start); | |
936 | #else | |
937 | if (strlen (start) > 50) | |
938 | /* xgettext:c-format */ | |
939 | sprintf (errbuf, _("bad instruction `%.50s...'"), start); | |
940 | else | |
941 | /* xgettext:c-format */ | |
942 | sprintf (errbuf, _("bad instruction `%.50s'"), start); | |
943 | #endif | |
944 | ||
945 | *errmsg = errbuf; | |
946 | return NULL; | |
947 | } | |
948 | } | |
949 | \f | |
950 | #if 0 /* This calls back to GAS which we can't do without care. */ | |
951 | ||
952 | /* Record each member of OPVALS in the assembler's symbol table. | |
953 | This lets GAS parse registers for us. | |
954 | ??? Interesting idea but not currently used. */ | |
955 | ||
956 | /* Record each member of OPVALS in the assembler's symbol table. | |
957 | FIXME: Not currently used. */ | |
958 | ||
959 | void | |
960 | ip2k_cgen_asm_hash_keywords (cd, opvals) | |
961 | CGEN_CPU_DESC cd; | |
962 | CGEN_KEYWORD *opvals; | |
963 | { | |
964 | CGEN_KEYWORD_SEARCH search = cgen_keyword_search_init (opvals, NULL); | |
965 | const CGEN_KEYWORD_ENTRY * ke; | |
966 | ||
967 | while ((ke = cgen_keyword_search_next (& search)) != NULL) | |
968 | { | |
969 | #if 0 /* Unnecessary, should be done in the search routine. */ | |
970 | if (! ip2k_cgen_opval_supported (ke)) | |
971 | continue; | |
972 | #endif | |
973 | cgen_asm_record_register (cd, ke->name, ke->value); | |
974 | } | |
975 | } | |
976 | ||
977 | #endif /* 0 */ |