Commit | Line | Data |
---|---|---|
4162bb66 | 1 | /* DO NOT EDIT! -*- buffer-read-only: t -*- vi:set ro: */ |
87e6d782 NC |
2 | /* Assembler interface for targets using CGEN. -*- C -*- |
3 | CGEN: Cpu tools GENerator | |
4 | ||
47b0e7ad NC |
5 | THIS FILE IS MACHINE GENERATED WITH CGEN. |
6 | - the resultant file is machine generated, cgen-asm.in isn't | |
87e6d782 | 7 | |
219d1afa | 8 | Copyright (C) 1996-2018 Free Software Foundation, Inc. |
87e6d782 | 9 | |
9b201bb5 | 10 | This file is part of libopcodes. |
87e6d782 | 11 | |
9b201bb5 | 12 | This library is free software; you can redistribute it and/or modify |
47b0e7ad | 13 | it under the terms of the GNU General Public License as published by |
9b201bb5 | 14 | the Free Software Foundation; either version 3, or (at your option) |
47b0e7ad | 15 | any later version. |
87e6d782 | 16 | |
9b201bb5 NC |
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. | |
87e6d782 | 21 | |
47b0e7ad NC |
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. */ | |
87e6d782 | 25 | |
9b201bb5 | 26 | |
87e6d782 NC |
27 | /* ??? Eventually more and more of this stuff can go to cpu-independent files. |
28 | Keep that in mind. */ | |
29 | ||
30 | #include "sysdep.h" | |
87e6d782 NC |
31 | #include <stdio.h> |
32 | #include "ansidecl.h" | |
33 | #include "bfd.h" | |
34 | #include "symcat.h" | |
73589c9d CS |
35 | #include "or1k-desc.h" |
36 | #include "or1k-opc.h" | |
87e6d782 | 37 | #include "opintl.h" |
fc7bc883 | 38 | #include "xregex.h" |
fc05c67f | 39 | #include "libiberty.h" |
37111cc7 | 40 | #include "safe-ctype.h" |
87e6d782 | 41 | |
37111cc7 | 42 | #undef min |
87e6d782 | 43 | #define min(a,b) ((a) < (b) ? (a) : (b)) |
37111cc7 | 44 | #undef max |
87e6d782 NC |
45 | #define max(a,b) ((a) > (b) ? (a) : (b)) |
46 | ||
47 | static const char * parse_insn_normal | |
ffead7ae | 48 | (CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *); |
87e6d782 | 49 | \f |
37111cc7 | 50 | /* -- assembler routines inserted here. */ |
87e6d782 NC |
51 | |
52 | /* -- asm.c */ | |
53 | ||
47b0e7ad | 54 | static const char * MISSING_CLOSING_PARENTHESIS = N_("missing `)'"); |
1c4f3780 RH |
55 | static const char * INVALID_STORE_RELOC = N_("relocation invalid for store"); |
56 | static const char * INVALID_RELOC_TYPE = N_("internal relocation type invalid"); | |
87e6d782 | 57 | |
47b0e7ad | 58 | #define CGEN_VERBOSE_ASSEMBLER_ERRORS |
0e2ee3ca | 59 | |
73589c9d CS |
60 | static const char * |
61 | parse_disp26 (CGEN_CPU_DESC cd, | |
c151b1c6 AM |
62 | const char ** strp, |
63 | int opindex, | |
64 | int opinfo, | |
65 | enum cgen_parse_operand_result * resultp, | |
66 | bfd_vma * valuep) | |
87e6d782 | 67 | { |
73589c9d CS |
68 | const char *errmsg = NULL; |
69 | enum cgen_parse_operand_result result_type; | |
87e6d782 | 70 | |
73589c9d CS |
71 | if (strncasecmp (*strp, "plt(", 4) == 0) |
72 | { | |
73 | bfd_vma value; | |
74 | ||
75 | *strp += 4; | |
76 | errmsg = cgen_parse_address (cd, strp, opindex, BFD_RELOC_OR1K_PLT26, | |
c151b1c6 | 77 | & result_type, & value); |
73589c9d | 78 | if (**strp != ')') |
c151b1c6 | 79 | return MISSING_CLOSING_PARENTHESIS; |
73589c9d CS |
80 | ++*strp; |
81 | if (errmsg == NULL | |
c151b1c6 AM |
82 | && result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) |
83 | value = (value >> 2) & 0xffff; | |
73589c9d CS |
84 | *valuep = value; |
85 | return errmsg; | |
86 | } | |
87 | return cgen_parse_address (cd, strp, opindex, opinfo, resultp, valuep); | |
88 | } | |
87e6d782 | 89 | |
1c4f3780 RH |
90 | enum { |
91 | RTYPE_LO = 0, | |
92 | RTYPE_HI = 1, | |
93 | RTYPE_AHI = 2, | |
94 | RTYPE_SLO = 3, | |
95 | ||
96 | RTYPE_GOT = (1 << 2), | |
97 | RTYPE_GOTPC = (2 << 2), | |
98 | RTYPE_GOTOFF = (3 << 2), | |
99 | RTYPE_TLSGD = (4 << 2), | |
100 | RTYPE_TLSLDM = (5 << 2), | |
101 | RTYPE_DTPOFF = (6 << 2), | |
102 | RTYPE_GOTTPOFF = (7 << 2), | |
103 | RTYPE_TPOFF = (8 << 2), | |
104 | }; | |
105 | ||
106 | static const bfd_reloc_code_real_type or1k_imm16_relocs[][4] = { | |
107 | { BFD_RELOC_LO16, | |
108 | BFD_RELOC_HI16, | |
109 | BFD_RELOC_HI16_S, | |
110 | BFD_RELOC_OR1K_SLO16 }, | |
111 | { BFD_RELOC_OR1K_GOT16, | |
112 | BFD_RELOC_UNUSED, | |
113 | BFD_RELOC_UNUSED, | |
114 | BFD_RELOC_UNUSED }, | |
115 | { BFD_RELOC_OR1K_GOTPC_LO16, | |
116 | BFD_RELOC_OR1K_GOTPC_HI16, | |
117 | BFD_RELOC_UNUSED, | |
118 | BFD_RELOC_UNUSED }, | |
119 | { BFD_RELOC_LO16_GOTOFF, | |
120 | BFD_RELOC_HI16_GOTOFF, | |
121 | BFD_RELOC_HI16_S_GOTOFF, | |
122 | BFD_RELOC_OR1K_GOTOFF_SLO16 }, | |
123 | { BFD_RELOC_OR1K_TLS_GD_LO16, | |
124 | BFD_RELOC_OR1K_TLS_GD_HI16, | |
125 | BFD_RELOC_UNUSED, | |
126 | BFD_RELOC_UNUSED }, | |
127 | { BFD_RELOC_OR1K_TLS_LDM_LO16, | |
128 | BFD_RELOC_OR1K_TLS_LDM_HI16, | |
129 | BFD_RELOC_UNUSED, | |
130 | BFD_RELOC_UNUSED }, | |
131 | { BFD_RELOC_OR1K_TLS_LDO_LO16, | |
132 | BFD_RELOC_OR1K_TLS_LDO_HI16, | |
133 | BFD_RELOC_UNUSED, | |
134 | BFD_RELOC_UNUSED }, | |
135 | { BFD_RELOC_OR1K_TLS_IE_LO16, | |
136 | BFD_RELOC_OR1K_TLS_IE_HI16, | |
137 | BFD_RELOC_OR1K_TLS_IE_AHI16, | |
138 | BFD_RELOC_UNUSED }, | |
139 | { BFD_RELOC_OR1K_TLS_LE_LO16, | |
140 | BFD_RELOC_OR1K_TLS_LE_HI16, | |
141 | BFD_RELOC_OR1K_TLS_LE_AHI16, | |
142 | BFD_RELOC_OR1K_TLS_LE_SLO16 } | |
143 | }; | |
144 | ||
145 | static int | |
146 | parse_reloc(const char **strp) | |
87e6d782 | 147 | { |
1c4f3780 RH |
148 | const char *str = *strp; |
149 | int ret = 0; | |
73589c9d | 150 | |
1c4f3780 | 151 | if (strncasecmp (str, "got(", 4) == 0) |
73589c9d | 152 | { |
1c4f3780 RH |
153 | *strp = str + 4; |
154 | return RTYPE_GOT | RTYPE_LO; | |
73589c9d | 155 | } |
73589c9d | 156 | |
1c4f3780 | 157 | if (strncasecmp (str, "gotpc", 5) == 0) |
73589c9d | 158 | { |
1c4f3780 RH |
159 | str += 5; |
160 | ret = RTYPE_GOTPC; | |
73589c9d | 161 | } |
1c4f3780 | 162 | else if (strncasecmp (str, "gotoff", 6) == 0) |
73589c9d | 163 | { |
1c4f3780 RH |
164 | str += 6; |
165 | ret = RTYPE_GOTOFF; | |
87e6d782 | 166 | } |
1c4f3780 | 167 | else if (strncasecmp (str, "tlsgd", 5) == 0) |
87e6d782 | 168 | { |
1c4f3780 RH |
169 | str += 5; |
170 | ret = RTYPE_TLSGD; | |
73589c9d | 171 | } |
1c4f3780 | 172 | else if (strncasecmp (str, "tlsldm", 6) == 0) |
73589c9d | 173 | { |
1c4f3780 RH |
174 | str += 6; |
175 | ret = RTYPE_TLSLDM; | |
87e6d782 | 176 | } |
1c4f3780 | 177 | else if (strncasecmp (str, "dtpoff", 6) == 0) |
73589c9d | 178 | { |
1c4f3780 RH |
179 | str += 6; |
180 | ret = RTYPE_DTPOFF; | |
73589c9d | 181 | } |
1c4f3780 | 182 | else if (strncasecmp (str, "gottpoff", 8) == 0) |
73589c9d | 183 | { |
1c4f3780 RH |
184 | str += 8; |
185 | ret = RTYPE_GOTTPOFF; | |
73589c9d | 186 | } |
1c4f3780 | 187 | else if (strncasecmp (str, "tpoff", 5) == 0) |
73589c9d | 188 | { |
1c4f3780 RH |
189 | str += 5; |
190 | ret = RTYPE_TPOFF; | |
73589c9d | 191 | } |
87e6d782 | 192 | |
1c4f3780 | 193 | if (strncasecmp (str, "hi(", 3) == 0) |
73589c9d | 194 | { |
1c4f3780 RH |
195 | str += 3; |
196 | ret |= RTYPE_HI; | |
73589c9d | 197 | } |
1c4f3780 | 198 | else if (strncasecmp (str, "lo(", 3) == 0) |
73589c9d | 199 | { |
1c4f3780 RH |
200 | str += 3; |
201 | ret |= RTYPE_LO; | |
87e6d782 | 202 | } |
1c4f3780 | 203 | else if (strncasecmp (str, "ha(", 3) == 0) |
8f807746 | 204 | { |
1c4f3780 RH |
205 | str += 3; |
206 | ret |= RTYPE_AHI; | |
73589c9d | 207 | } |
1c4f3780 RH |
208 | else |
209 | return -1; | |
73589c9d | 210 | |
1c4f3780 RH |
211 | *strp = str; |
212 | return ret; | |
213 | } | |
8f807746 | 214 | |
1c4f3780 RH |
215 | static const char * |
216 | parse_imm16 (CGEN_CPU_DESC cd, const char **strp, int opindex, | |
217 | long *valuep, int splitp) | |
218 | { | |
219 | const char *errmsg; | |
220 | enum cgen_parse_operand_result result_type; | |
221 | bfd_reloc_code_real_type reloc = BFD_RELOC_UNUSED; | |
222 | int reloc_type; | |
223 | bfd_vma ret; | |
73589c9d | 224 | |
1c4f3780 | 225 | if (**strp == '#') |
73589c9d | 226 | ++*strp; |
1c4f3780 RH |
227 | |
228 | reloc_type = parse_reloc (strp); | |
229 | if (reloc_type >= 0) | |
230 | { | |
231 | if (splitp) | |
232 | { | |
233 | if ((reloc_type & 3) == RTYPE_LO && reloc_type != RTYPE_GOT) | |
234 | reloc_type |= RTYPE_SLO; | |
235 | else | |
236 | return INVALID_STORE_RELOC; | |
73589c9d | 237 | } |
1c4f3780 RH |
238 | reloc = or1k_imm16_relocs[reloc_type >> 2][reloc_type & 3]; |
239 | } | |
240 | ||
241 | if (reloc != BFD_RELOC_UNUSED) | |
73589c9d CS |
242 | { |
243 | bfd_vma value; | |
244 | ||
1c4f3780 | 245 | errmsg = cgen_parse_address (cd, strp, opindex, reloc, |
c151b1c6 | 246 | &result_type, &value); |
73589c9d | 247 | if (**strp != ')') |
1c4f3780 | 248 | errmsg = MISSING_CLOSING_PARENTHESIS; |
73589c9d | 249 | ++*strp; |
73589c9d | 250 | |
1c4f3780 | 251 | ret = value; |
73589c9d | 252 | |
1c4f3780 RH |
253 | if (errmsg == NULL && result_type == CGEN_PARSE_OPERAND_RESULT_NUMBER) |
254 | switch (reloc_type & 3) | |
73589c9d | 255 | { |
1c4f3780 RH |
256 | case RTYPE_AHI: |
257 | ret += 0x8000; | |
258 | /* FALLTHRU */ | |
259 | case RTYPE_HI: | |
260 | ret >>= 16; | |
261 | /* FALLTHRU */ | |
262 | case RTYPE_LO: | |
263 | case RTYPE_SLO: | |
264 | ret &= 0xffff; | |
265 | ret = (ret ^ 0x8000) - 0x8000; | |
266 | break; | |
267 | default: | |
268 | errmsg = INVALID_RELOC_TYPE; | |
269 | } | |
73589c9d CS |
270 | } |
271 | else | |
272 | { | |
273 | long value; | |
274 | errmsg = cgen_parse_signed_integer (cd, strp, opindex, &value); | |
275 | ret = value; | |
276 | } | |
277 | ||
278 | if (errmsg == NULL) | |
279 | *valuep = ret; | |
280 | ||
281 | return errmsg; | |
282 | } | |
283 | ||
284 | static const char * | |
1c4f3780 RH |
285 | parse_simm16 (CGEN_CPU_DESC cd, const char **strp, int opindex, long *valuep) |
286 | { | |
287 | return parse_imm16(cd, strp, opindex, (long *) valuep, 0); | |
288 | } | |
289 | ||
290 | static const char * | |
291 | parse_simm16_split (CGEN_CPU_DESC cd, const char **strp, int opindex, | |
292 | long *valuep) | |
73589c9d | 293 | { |
1c4f3780 RH |
294 | return parse_imm16(cd, strp, opindex, (long *) valuep, 1); |
295 | } | |
c151b1c6 | 296 | |
1c4f3780 RH |
297 | static const char * |
298 | parse_uimm16 (CGEN_CPU_DESC cd, const char **strp, int opindex, | |
299 | unsigned long *valuep) | |
300 | { | |
301 | const char *errmsg = parse_imm16(cd, strp, opindex, (long *) valuep, 0); | |
302 | if (errmsg == NULL) | |
303 | *valuep &= 0xffff; | |
304 | return errmsg; | |
305 | } | |
306 | ||
307 | static const char * | |
308 | parse_uimm16_split (CGEN_CPU_DESC cd, const char **strp, int opindex, | |
309 | unsigned long *valuep) | |
310 | { | |
311 | const char *errmsg = parse_imm16(cd, strp, opindex, (long *) valuep, 1); | |
73589c9d CS |
312 | if (errmsg == NULL) |
313 | *valuep &= 0xffff; | |
87e6d782 NC |
314 | return errmsg; |
315 | } | |
316 | ||
317 | /* -- */ | |
318 | ||
73589c9d | 319 | const char * or1k_cgen_parse_operand |
47b0e7ad | 320 | (CGEN_CPU_DESC, int, const char **, CGEN_FIELDS *); |
0e2ee3ca | 321 | |
87e6d782 NC |
322 | /* Main entry point for operand parsing. |
323 | ||
324 | This function is basically just a big switch statement. Earlier versions | |
325 | used tables to look up the function to use, but | |
326 | - if the table contains both assembler and disassembler functions then | |
327 | the disassembler contains much of the assembler and vice-versa, | |
328 | - there's a lot of inlining possibilities as things grow, | |
329 | - using a switch statement avoids the function call overhead. | |
330 | ||
331 | This function could be moved into `parse_insn_normal', but keeping it | |
332 | separate makes clear the interface between `parse_insn_normal' and each of | |
9a2e995d | 333 | the handlers. */ |
87e6d782 NC |
334 | |
335 | const char * | |
73589c9d | 336 | or1k_cgen_parse_operand (CGEN_CPU_DESC cd, |
47b0e7ad NC |
337 | int opindex, |
338 | const char ** strp, | |
339 | CGEN_FIELDS * fields) | |
87e6d782 NC |
340 | { |
341 | const char * errmsg = NULL; | |
342 | /* Used by scalar operands that still need to be parsed. */ | |
fc05c67f | 343 | long junk ATTRIBUTE_UNUSED; |
87e6d782 NC |
344 | |
345 | switch (opindex) | |
346 | { | |
73589c9d | 347 | case OR1K_OPERAND_DISP26 : |
87e6d782 | 348 | { |
9494d739 | 349 | bfd_vma value = 0; |
73589c9d | 350 | errmsg = parse_disp26 (cd, strp, OR1K_OPERAND_DISP26, 0, NULL, & value); |
87e6d782 NC |
351 | fields->f_disp26 = value; |
352 | } | |
353 | break; | |
73589c9d CS |
354 | case OR1K_OPERAND_RA : |
355 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_gpr, & fields->f_r2); | |
356 | break; | |
357 | case OR1K_OPERAND_RADF : | |
358 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fdr, & fields->f_r1); | |
87e6d782 | 359 | break; |
73589c9d CS |
360 | case OR1K_OPERAND_RASF : |
361 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fsr, & fields->f_r2); | |
87e6d782 | 362 | break; |
73589c9d CS |
363 | case OR1K_OPERAND_RB : |
364 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_gpr, & fields->f_r3); | |
87e6d782 | 365 | break; |
73589c9d CS |
366 | case OR1K_OPERAND_RBDF : |
367 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fdr, & fields->f_r1); | |
87e6d782 | 368 | break; |
73589c9d CS |
369 | case OR1K_OPERAND_RBSF : |
370 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fsr, & fields->f_r3); | |
87e6d782 | 371 | break; |
73589c9d CS |
372 | case OR1K_OPERAND_RD : |
373 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_gpr, & fields->f_r1); | |
87e6d782 | 374 | break; |
73589c9d CS |
375 | case OR1K_OPERAND_RDDF : |
376 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fdr, & fields->f_r1); | |
87e6d782 | 377 | break; |
73589c9d CS |
378 | case OR1K_OPERAND_RDSF : |
379 | errmsg = cgen_parse_keyword (cd, strp, & or1k_cgen_opval_h_fsr, & fields->f_r1); | |
87e6d782 | 380 | break; |
73589c9d CS |
381 | case OR1K_OPERAND_SIMM16 : |
382 | errmsg = parse_simm16 (cd, strp, OR1K_OPERAND_SIMM16, (long *) (& fields->f_simm16)); | |
87e6d782 | 383 | break; |
73589c9d | 384 | case OR1K_OPERAND_SIMM16_SPLIT : |
1c4f3780 | 385 | errmsg = parse_simm16_split (cd, strp, OR1K_OPERAND_SIMM16_SPLIT, (long *) (& fields->f_simm16_split)); |
87e6d782 | 386 | break; |
73589c9d CS |
387 | case OR1K_OPERAND_UIMM16 : |
388 | errmsg = parse_uimm16 (cd, strp, OR1K_OPERAND_UIMM16, (unsigned long *) (& fields->f_uimm16)); | |
389 | break; | |
390 | case OR1K_OPERAND_UIMM16_SPLIT : | |
1c4f3780 | 391 | errmsg = parse_uimm16_split (cd, strp, OR1K_OPERAND_UIMM16_SPLIT, (unsigned long *) (& fields->f_uimm16_split)); |
73589c9d CS |
392 | break; |
393 | case OR1K_OPERAND_UIMM6 : | |
394 | errmsg = cgen_parse_unsigned_integer (cd, strp, OR1K_OPERAND_UIMM6, (unsigned long *) (& fields->f_uimm6)); | |
87e6d782 NC |
395 | break; |
396 | ||
397 | default : | |
398 | /* xgettext:c-format */ | |
a6743a54 AM |
399 | opcodes_error_handler |
400 | (_("internal error: unrecognized field %d while parsing"), | |
401 | opindex); | |
87e6d782 NC |
402 | abort (); |
403 | } | |
404 | ||
405 | return errmsg; | |
406 | } | |
407 | ||
43e65147 | 408 | cgen_parse_fn * const or1k_cgen_parse_handlers[] = |
87e6d782 NC |
409 | { |
410 | parse_insn_normal, | |
411 | }; | |
412 | ||
413 | void | |
73589c9d | 414 | or1k_cgen_init_asm (CGEN_CPU_DESC cd) |
87e6d782 | 415 | { |
73589c9d CS |
416 | or1k_cgen_init_opcode_table (cd); |
417 | or1k_cgen_init_ibld_table (cd); | |
418 | cd->parse_handlers = & or1k_cgen_parse_handlers[0]; | |
419 | cd->parse_operand = or1k_cgen_parse_operand; | |
1620f33d AM |
420 | #ifdef CGEN_ASM_INIT_HOOK |
421 | CGEN_ASM_INIT_HOOK | |
422 | #endif | |
87e6d782 NC |
423 | } |
424 | ||
fc7bc883 RH |
425 | \f |
426 | ||
37111cc7 | 427 | /* Regex construction routine. |
fc7bc883 | 428 | |
37111cc7 NC |
429 | This translates an opcode syntax string into a regex string, |
430 | by replacing any non-character syntax element (such as an | |
431 | opcode) with the pattern '.*' | |
fc7bc883 | 432 | |
37111cc7 | 433 | It then compiles the regex and stores it in the opcode, for |
73589c9d | 434 | later use by or1k_cgen_assemble_insn |
fc7bc883 | 435 | |
37111cc7 | 436 | Returns NULL for success, an error message for failure. */ |
fc7bc883 | 437 | |
43e65147 | 438 | char * |
73589c9d | 439 | or1k_cgen_build_insn_regex (CGEN_INSN *insn) |
43e65147 | 440 | { |
fc05c67f | 441 | CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn); |
fc7bc883 | 442 | const char *mnem = CGEN_INSN_MNEMONIC (insn); |
fc7bc883 RH |
443 | char rxbuf[CGEN_MAX_RX_ELEMENTS]; |
444 | char *rx = rxbuf; | |
445 | const CGEN_SYNTAX_CHAR_TYPE *syn; | |
446 | int reg_err; | |
447 | ||
448 | syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc)); | |
449 | ||
f3a55c17 NC |
450 | /* Mnemonics come first in the syntax string. */ |
451 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) | |
452 | return _("missing mnemonic in syntax string"); | |
fc7bc883 RH |
453 | ++syn; |
454 | ||
f3a55c17 NC |
455 | /* Generate a case sensitive regular expression that emulates case |
456 | insensitive matching in the "C" locale. We cannot generate a case | |
457 | insensitive regular expression because in Turkish locales, 'i' and 'I' | |
458 | are not equal modulo case conversion. */ | |
fc7bc883 | 459 | |
f3a55c17 NC |
460 | /* Copy the literal mnemonic out of the insn. */ |
461 | for (; *mnem; mnem++) | |
462 | { | |
463 | char c = *mnem; | |
464 | ||
465 | if (ISALPHA (c)) | |
466 | { | |
467 | *rx++ = '['; | |
468 | *rx++ = TOLOWER (c); | |
469 | *rx++ = TOUPPER (c); | |
470 | *rx++ = ']'; | |
471 | } | |
472 | else | |
473 | *rx++ = c; | |
474 | } | |
475 | ||
476 | /* Copy any remaining literals from the syntax string into the rx. */ | |
477 | for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn) | |
fc7bc883 | 478 | { |
43e65147 | 479 | if (CGEN_SYNTAX_CHAR_P (* syn)) |
fc7bc883 | 480 | { |
f3a55c17 NC |
481 | char c = CGEN_SYNTAX_CHAR (* syn); |
482 | ||
43e65147 | 483 | switch (c) |
f3a55c17 NC |
484 | { |
485 | /* Escape any regex metacharacters in the syntax. */ | |
43e65147 L |
486 | case '.': case '[': case '\\': |
487 | case '*': case '^': case '$': | |
fc7bc883 RH |
488 | |
489 | #ifdef CGEN_ESCAPE_EXTENDED_REGEX | |
43e65147 | 490 | case '?': case '{': case '}': |
f3a55c17 NC |
491 | case '(': case ')': case '*': |
492 | case '|': case '+': case ']': | |
fc7bc883 | 493 | #endif |
f3a55c17 NC |
494 | *rx++ = '\\'; |
495 | *rx++ = c; | |
496 | break; | |
497 | ||
498 | default: | |
499 | if (ISALPHA (c)) | |
500 | { | |
501 | *rx++ = '['; | |
502 | *rx++ = TOLOWER (c); | |
503 | *rx++ = TOUPPER (c); | |
504 | *rx++ = ']'; | |
505 | } | |
506 | else | |
507 | *rx++ = c; | |
508 | break; | |
509 | } | |
fc7bc883 RH |
510 | } |
511 | else | |
512 | { | |
f3a55c17 NC |
513 | /* Replace non-syntax fields with globs. */ |
514 | *rx++ = '.'; | |
515 | *rx++ = '*'; | |
fc7bc883 RH |
516 | } |
517 | } | |
518 | ||
f3a55c17 | 519 | /* Trailing whitespace ok. */ |
43e65147 L |
520 | * rx++ = '['; |
521 | * rx++ = ' '; | |
522 | * rx++ = '\t'; | |
523 | * rx++ = ']'; | |
524 | * rx++ = '*'; | |
fc7bc883 | 525 | |
f3a55c17 | 526 | /* But anchor it after that. */ |
43e65147 | 527 | * rx++ = '$'; |
fc7bc883 RH |
528 | * rx = '\0'; |
529 | ||
530 | CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t)); | |
f3a55c17 | 531 | reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB); |
fc7bc883 | 532 | |
43e65147 | 533 | if (reg_err == 0) |
fc7bc883 RH |
534 | return NULL; |
535 | else | |
536 | { | |
537 | static char msg[80]; | |
f3a55c17 | 538 | |
fc7bc883 RH |
539 | regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80); |
540 | regfree ((regex_t *) CGEN_INSN_RX (insn)); | |
541 | free (CGEN_INSN_RX (insn)); | |
542 | (CGEN_INSN_RX (insn)) = NULL; | |
37111cc7 | 543 | return msg; |
fc7bc883 RH |
544 | } |
545 | } | |
546 | ||
87e6d782 NC |
547 | \f |
548 | /* Default insn parser. | |
549 | ||
550 | The syntax string is scanned and operands are parsed and stored in FIELDS. | |
551 | Relocs are queued as we go via other callbacks. | |
552 | ||
553 | ??? Note that this is currently an all-or-nothing parser. If we fail to | |
554 | parse the instruction, we return 0 and the caller will start over from | |
555 | the beginning. Backtracking will be necessary in parsing subexpressions, | |
556 | but that can be handled there. Not handling backtracking here may get | |
557 | expensive in the case of the m68k. Deal with later. | |
558 | ||
f3a55c17 | 559 | Returns NULL for success, an error message for failure. */ |
87e6d782 NC |
560 | |
561 | static const char * | |
ffead7ae MM |
562 | parse_insn_normal (CGEN_CPU_DESC cd, |
563 | const CGEN_INSN *insn, | |
564 | const char **strp, | |
565 | CGEN_FIELDS *fields) | |
87e6d782 NC |
566 | { |
567 | /* ??? Runtime added insns not handled yet. */ | |
568 | const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn); | |
569 | const char *str = *strp; | |
570 | const char *errmsg; | |
571 | const char *p; | |
572 | const CGEN_SYNTAX_CHAR_TYPE * syn; | |
573 | #ifdef CGEN_MNEMONIC_OPERANDS | |
574 | /* FIXME: wip */ | |
575 | int past_opcode_p; | |
576 | #endif | |
577 | ||
578 | /* For now we assume the mnemonic is first (there are no leading operands). | |
579 | We can parse it without needing to set up operand parsing. | |
580 | GAS's input scrubber will ensure mnemonics are lowercase, but we may | |
581 | not be called from GAS. */ | |
582 | p = CGEN_INSN_MNEMONIC (insn); | |
37111cc7 | 583 | while (*p && TOLOWER (*p) == TOLOWER (*str)) |
87e6d782 NC |
584 | ++p, ++str; |
585 | ||
586 | if (* p) | |
587 | return _("unrecognized instruction"); | |
588 | ||
589 | #ifndef CGEN_MNEMONIC_OPERANDS | |
37111cc7 | 590 | if (* str && ! ISSPACE (* str)) |
87e6d782 NC |
591 | return _("unrecognized instruction"); |
592 | #endif | |
593 | ||
594 | CGEN_INIT_PARSE (cd); | |
595 | cgen_init_parse_operand (cd); | |
596 | #ifdef CGEN_MNEMONIC_OPERANDS | |
597 | past_opcode_p = 0; | |
598 | #endif | |
599 | ||
600 | /* We don't check for (*str != '\0') here because we want to parse | |
601 | any trailing fake arguments in the syntax string. */ | |
602 | syn = CGEN_SYNTAX_STRING (syntax); | |
603 | ||
604 | /* Mnemonics come first for now, ensure valid string. */ | |
605 | if (! CGEN_SYNTAX_MNEMONIC_P (* syn)) | |
606 | abort (); | |
607 | ||
608 | ++syn; | |
609 | ||
610 | while (* syn != 0) | |
611 | { | |
612 | /* Non operand chars must match exactly. */ | |
613 | if (CGEN_SYNTAX_CHAR_P (* syn)) | |
614 | { | |
615 | /* FIXME: While we allow for non-GAS callers above, we assume the | |
616 | first char after the mnemonic part is a space. */ | |
617 | /* FIXME: We also take inappropriate advantage of the fact that | |
618 | GAS's input scrubber will remove extraneous blanks. */ | |
37111cc7 | 619 | if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn))) |
87e6d782 NC |
620 | { |
621 | #ifdef CGEN_MNEMONIC_OPERANDS | |
622 | if (CGEN_SYNTAX_CHAR(* syn) == ' ') | |
623 | past_opcode_p = 1; | |
624 | #endif | |
625 | ++ syn; | |
626 | ++ str; | |
627 | } | |
628 | else if (*str) | |
629 | { | |
630 | /* Syntax char didn't match. Can't be this insn. */ | |
631 | static char msg [80]; | |
f3a55c17 | 632 | |
87e6d782 NC |
633 | /* xgettext:c-format */ |
634 | sprintf (msg, _("syntax error (expected char `%c', found `%c')"), | |
635 | CGEN_SYNTAX_CHAR(*syn), *str); | |
636 | return msg; | |
637 | } | |
638 | else | |
639 | { | |
640 | /* Ran out of input. */ | |
641 | static char msg [80]; | |
f3a55c17 | 642 | |
87e6d782 NC |
643 | /* xgettext:c-format */ |
644 | sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"), | |
645 | CGEN_SYNTAX_CHAR(*syn)); | |
646 | return msg; | |
647 | } | |
648 | continue; | |
649 | } | |
650 | ||
c7e2358a AM |
651 | #ifdef CGEN_MNEMONIC_OPERANDS |
652 | (void) past_opcode_p; | |
653 | #endif | |
87e6d782 | 654 | /* We have an operand of some sort. */ |
c7e2358a | 655 | errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn), &str, fields); |
87e6d782 NC |
656 | if (errmsg) |
657 | return errmsg; | |
658 | ||
659 | /* Done with this operand, continue with next one. */ | |
660 | ++ syn; | |
661 | } | |
662 | ||
663 | /* If we're at the end of the syntax string, we're done. */ | |
664 | if (* syn == 0) | |
665 | { | |
666 | /* FIXME: For the moment we assume a valid `str' can only contain | |
667 | blanks now. IE: We needn't try again with a longer version of | |
668 | the insn and it is assumed that longer versions of insns appear | |
669 | before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */ | |
37111cc7 | 670 | while (ISSPACE (* str)) |
87e6d782 NC |
671 | ++ str; |
672 | ||
673 | if (* str != '\0') | |
674 | return _("junk at end of line"); /* FIXME: would like to include `str' */ | |
675 | ||
676 | return NULL; | |
677 | } | |
678 | ||
679 | /* We couldn't parse it. */ | |
680 | return _("unrecognized instruction"); | |
681 | } | |
682 | \f | |
683 | /* Main entry point. | |
684 | This routine is called for each instruction to be assembled. | |
685 | STR points to the insn to be assembled. | |
686 | We assume all necessary tables have been initialized. | |
687 | The assembled instruction, less any fixups, is stored in BUF. | |
688 | Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value | |
689 | still needs to be converted to target byte order, otherwise BUF is an array | |
690 | of bytes in target byte order. | |
691 | The result is a pointer to the insn's entry in the opcode table, | |
692 | or NULL if an error occured (an error message will have already been | |
693 | printed). | |
694 | ||
695 | Note that when processing (non-alias) macro-insns, | |
696 | this function recurses. | |
697 | ||
698 | ??? It's possible to make this cpu-independent. | |
699 | One would have to deal with a few minor things. | |
700 | At this point in time doing so would be more of a curiosity than useful | |
701 | [for example this file isn't _that_ big], but keeping the possibility in | |
702 | mind helps keep the design clean. */ | |
703 | ||
704 | const CGEN_INSN * | |
73589c9d | 705 | or1k_cgen_assemble_insn (CGEN_CPU_DESC cd, |
ffead7ae MM |
706 | const char *str, |
707 | CGEN_FIELDS *fields, | |
708 | CGEN_INSN_BYTES_PTR buf, | |
709 | char **errmsg) | |
87e6d782 NC |
710 | { |
711 | const char *start; | |
712 | CGEN_INSN_LIST *ilist; | |
713 | const char *parse_errmsg = NULL; | |
714 | const char *insert_errmsg = NULL; | |
fc7bc883 | 715 | int recognized_mnemonic = 0; |
87e6d782 NC |
716 | |
717 | /* Skip leading white space. */ | |
37111cc7 | 718 | while (ISSPACE (* str)) |
87e6d782 NC |
719 | ++ str; |
720 | ||
721 | /* The instructions are stored in hashed lists. | |
722 | Get the first in the list. */ | |
723 | ilist = CGEN_ASM_LOOKUP_INSN (cd, str); | |
724 | ||
725 | /* Keep looking until we find a match. */ | |
87e6d782 NC |
726 | start = str; |
727 | for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist)) | |
728 | { | |
729 | const CGEN_INSN *insn = ilist->insn; | |
fc7bc883 | 730 | recognized_mnemonic = 1; |
87e6d782 | 731 | |
43e65147 | 732 | #ifdef CGEN_VALIDATE_INSN_SUPPORTED |
f3a55c17 NC |
733 | /* Not usually needed as unsupported opcodes |
734 | shouldn't be in the hash lists. */ | |
87e6d782 | 735 | /* Is this insn supported by the selected cpu? */ |
73589c9d | 736 | if (! or1k_cgen_insn_supported (cd, insn)) |
87e6d782 NC |
737 | continue; |
738 | #endif | |
b11dcf4e | 739 | /* If the RELAXED attribute is set, this is an insn that shouldn't be |
87e6d782 NC |
740 | chosen immediately. Instead, it is used during assembler/linker |
741 | relaxation if possible. */ | |
b11dcf4e | 742 | if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0) |
87e6d782 NC |
743 | continue; |
744 | ||
745 | str = start; | |
746 | ||
f3a55c17 | 747 | /* Skip this insn if str doesn't look right lexically. */ |
fc7bc883 RH |
748 | if (CGEN_INSN_RX (insn) != NULL && |
749 | regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH) | |
750 | continue; | |
751 | ||
87e6d782 NC |
752 | /* Allow parse/insert handlers to obtain length of insn. */ |
753 | CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn); | |
754 | ||
755 | parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields); | |
756 | if (parse_errmsg != NULL) | |
757 | continue; | |
758 | ||
f3a55c17 | 759 | /* ??? 0 is passed for `pc'. */ |
87e6d782 NC |
760 | insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf, |
761 | (bfd_vma) 0); | |
762 | if (insert_errmsg != NULL) | |
763 | continue; | |
764 | ||
765 | /* It is up to the caller to actually output the insn and any | |
766 | queued relocs. */ | |
767 | return insn; | |
768 | } | |
769 | ||
770 | { | |
771 | static char errbuf[150]; | |
87e6d782 | 772 | const char *tmp_errmsg; |
c7e2358a AM |
773 | #ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS |
774 | #define be_verbose 1 | |
87e6d782 | 775 | #else |
c7e2358a | 776 | #define be_verbose 0 |
87e6d782 | 777 | #endif |
c7e2358a AM |
778 | |
779 | if (be_verbose) | |
780 | { | |
781 | /* If requesting verbose error messages, use insert_errmsg. | |
782 | Failing that, use parse_errmsg. */ | |
783 | tmp_errmsg = (insert_errmsg ? insert_errmsg : | |
784 | parse_errmsg ? parse_errmsg : | |
785 | recognized_mnemonic ? | |
786 | _("unrecognized form of instruction") : | |
787 | _("unrecognized instruction")); | |
788 | ||
789 | if (strlen (start) > 50) | |
790 | /* xgettext:c-format */ | |
791 | sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start); | |
43e65147 | 792 | else |
c7e2358a AM |
793 | /* xgettext:c-format */ |
794 | sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start); | |
795 | } | |
796 | else | |
797 | { | |
798 | if (strlen (start) > 50) | |
799 | /* xgettext:c-format */ | |
800 | sprintf (errbuf, _("bad instruction `%.50s...'"), start); | |
43e65147 | 801 | else |
c7e2358a AM |
802 | /* xgettext:c-format */ |
803 | sprintf (errbuf, _("bad instruction `%.50s'"), start); | |
804 | } | |
43e65147 | 805 | |
87e6d782 NC |
806 | *errmsg = errbuf; |
807 | return NULL; | |
808 | } | |
809 | } |