?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
AIX's assembler does not support local labels, use relative addressing instead
[qemu/mini2440.git] / target-arm / translate.c
blob9d52991b405fd046ee886246eb94f4bb8933fa2e
1 /*
2 * ARM translation
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 * Copyright (c) 2005-2007 CodeSourcery
6 * Copyright (c) 2007 OpenedHand, Ltd.
7 *
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2 of the License, or (at your option) any later version.
12 *
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, write to the Free Software
20 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21 */
22 #include <stdarg.h>
23 #include <stdlib.h>
24 #include <stdio.h>
25 #include <string.h>
26 #include <inttypes.h>
27
28 #include "cpu.h"
29 #include "exec-all.h"
30 #include "disas.h"
31 #include "tcg-op.h"
32 #include "qemu-log.h"
33
34 #include "helpers.h"
35 #define GEN_HELPER 1
36 #include "helpers.h"
37
38 #define ENABLE_ARCH_5J 0
39 #define ENABLE_ARCH_6 arm_feature(env, ARM_FEATURE_V6)
40 #define ENABLE_ARCH_6K arm_feature(env, ARM_FEATURE_V6K)
41 #define ENABLE_ARCH_6T2 arm_feature(env, ARM_FEATURE_THUMB2)
42 #define ENABLE_ARCH_7 arm_feature(env, ARM_FEATURE_V7)
43
44 #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0)
45
46 /* internal defines */
47 typedef struct DisasContext {
48 target_ulong pc;
49 int is_jmp;
50 /* Nonzero if this instruction has been conditionally skipped. */
51 int condjmp;
52 /* The label that will be jumped to when the instruction is skipped. */
53 int condlabel;
54 /* Thumb-2 condtional execution bits. */
55 int condexec_mask;
56 int condexec_cond;
57 struct TranslationBlock *tb;
58 int singlestep_enabled;
59 int thumb;
60 int is_mem;
61 #if !defined(CONFIG_USER_ONLY)
62 int user;
63 #endif
64 } DisasContext;
65
66 #if defined(CONFIG_USER_ONLY)
67 #define IS_USER(s) 1
68 #else
69 #define IS_USER(s) (s->user)
70 #endif
71
72 /* These instructions trap after executing, so defer them until after the
73 conditional executions state has been updated. */
74 #define DISAS_WFI 4
75 #define DISAS_SWI 5
76
77 static TCGv_ptr cpu_env;
78 /* We reuse the same 64-bit temporaries for efficiency. */
79 static TCGv_i64 cpu_V0, cpu_V1, cpu_M0;
80
81 /* FIXME: These should be removed. */
82 static TCGv cpu_T[2];
83 static TCGv cpu_F0s, cpu_F1s;
84 static TCGv_i64 cpu_F0d, cpu_F1d;
85
86 #define ICOUNT_TEMP cpu_T[0]
87 #include "gen-icount.h"
88
89 /* initialize TCG globals. */
90 void arm_translate_init(void)
91 {
92 cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
93
94 cpu_T[0] = tcg_global_reg_new_i32(TCG_AREG1, "T0");
95 cpu_T[1] = tcg_global_reg_new_i32(TCG_AREG2, "T1");
96
97 #define GEN_HELPER 2
98 #include "helpers.h"
99 }
100
101 /* The code generator doesn't like lots of temporaries, so maintain our own
102 cache for reuse within a function. */
103 #define MAX_TEMPS 8
104 static int num_temps;
105 static TCGv temps[MAX_TEMPS];
106
107 /* Allocate a temporary variable. */
108 static TCGv_i32 new_tmp(void)
109 {
110 TCGv tmp;
111 if (num_temps == MAX_TEMPS)
112 abort();
113
114 if (GET_TCGV_I32(temps[num_temps]))
115 return temps[num_temps++];
116
117 tmp = tcg_temp_new_i32();
118 temps[num_temps++] = tmp;
119 return tmp;
120 }
121
122 /* Release a temporary variable. */
123 static void dead_tmp(TCGv tmp)
124 {
125 int i;
126 num_temps--;
127 i = num_temps;
128 if (TCGV_EQUAL(temps[i], tmp))
129 return;
130
131 /* Shuffle this temp to the last slot. */
132 while (!TCGV_EQUAL(temps[i], tmp))
133 i--;
134 while (i < num_temps) {
135 temps[i] = temps[i + 1];
136 i++;
137 }
138 temps[i] = tmp;
139 }
140
141 static inline TCGv load_cpu_offset(int offset)
142 {
143 TCGv tmp = new_tmp();
144 tcg_gen_ld_i32(tmp, cpu_env, offset);
145 return tmp;
146 }
147
148 #define load_cpu_field(name) load_cpu_offset(offsetof(CPUState, name))
149
150 static inline void store_cpu_offset(TCGv var, int offset)
151 {
152 tcg_gen_st_i32(var, cpu_env, offset);
153 dead_tmp(var);
154 }
155
156 #define store_cpu_field(var, name) \
157 store_cpu_offset(var, offsetof(CPUState, name))
158
159 /* Set a variable to the value of a CPU register. */
160 static void load_reg_var(DisasContext *s, TCGv var, int reg)
161 {
162 if (reg == 15) {
163 uint32_t addr;
164 /* normaly, since we updated PC, we need only to add one insn */
165 if (s->thumb)
166 addr = (long)s->pc + 2;
167 else
168 addr = (long)s->pc + 4;
169 tcg_gen_movi_i32(var, addr);
170 } else {
171 tcg_gen_ld_i32(var, cpu_env, offsetof(CPUState, regs[reg]));
172 }
173 }
174
175 /* Create a new temporary and set it to the value of a CPU register. */
176 static inline TCGv load_reg(DisasContext *s, int reg)
177 {
178 TCGv tmp = new_tmp();
179 load_reg_var(s, tmp, reg);
180 return tmp;
181 }
182
183 /* Set a CPU register. The source must be a temporary and will be
184 marked as dead. */
185 static void store_reg(DisasContext *s, int reg, TCGv var)
186 {
187 if (reg == 15) {
188 tcg_gen_andi_i32(var, var, ~1);
189 s->is_jmp = DISAS_JUMP;
190 }
191 tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, regs[reg]));
192 dead_tmp(var);
193 }
194
195
196 /* Basic operations. */
197 #define gen_op_movl_T0_T1() tcg_gen_mov_i32(cpu_T[0], cpu_T[1])
198 #define gen_op_movl_T1_T0() tcg_gen_mov_i32(cpu_T[1], cpu_T[0])
199 #define gen_op_movl_T0_im(im) tcg_gen_movi_i32(cpu_T[0], im)
200 #define gen_op_movl_T1_im(im) tcg_gen_movi_i32(cpu_T[1], im)
201
202 #define gen_op_addl_T1_im(im) tcg_gen_addi_i32(cpu_T[1], cpu_T[1], im)
203 #define gen_op_addl_T0_T1() tcg_gen_add_i32(cpu_T[0], cpu_T[0], cpu_T[1])
204 #define gen_op_subl_T0_T1() tcg_gen_sub_i32(cpu_T[0], cpu_T[0], cpu_T[1])
205 #define gen_op_rsbl_T0_T1() tcg_gen_sub_i32(cpu_T[0], cpu_T[1], cpu_T[0])
206
207 #define gen_op_addl_T0_T1_cc() gen_helper_add_cc(cpu_T[0], cpu_T[0], cpu_T[1])
208 #define gen_op_adcl_T0_T1_cc() gen_helper_adc_cc(cpu_T[0], cpu_T[0], cpu_T[1])
209 #define gen_op_subl_T0_T1_cc() gen_helper_sub_cc(cpu_T[0], cpu_T[0], cpu_T[1])
210 #define gen_op_sbcl_T0_T1_cc() gen_helper_sbc_cc(cpu_T[0], cpu_T[0], cpu_T[1])
211 #define gen_op_rsbl_T0_T1_cc() gen_helper_sub_cc(cpu_T[0], cpu_T[1], cpu_T[0])
212 #define gen_op_rscl_T0_T1_cc() gen_helper_sbc_cc(cpu_T[0], cpu_T[1], cpu_T[0])
213
214 #define gen_op_andl_T0_T1() tcg_gen_and_i32(cpu_T[0], cpu_T[0], cpu_T[1])
215 #define gen_op_xorl_T0_T1() tcg_gen_xor_i32(cpu_T[0], cpu_T[0], cpu_T[1])
216 #define gen_op_orl_T0_T1() tcg_gen_or_i32(cpu_T[0], cpu_T[0], cpu_T[1])
217 #define gen_op_notl_T0() tcg_gen_not_i32(cpu_T[0], cpu_T[0])
218 #define gen_op_notl_T1() tcg_gen_not_i32(cpu_T[1], cpu_T[1])
219 #define gen_op_logic_T0_cc() gen_logic_CC(cpu_T[0]);
220 #define gen_op_logic_T1_cc() gen_logic_CC(cpu_T[1]);
221
222 #define gen_op_shll_T0_im(im) tcg_gen_shli_i32(cpu_T[0], cpu_T[0], im)
223 #define gen_op_shll_T1_im(im) tcg_gen_shli_i32(cpu_T[1], cpu_T[1], im)
224 #define gen_op_shrl_T1_im(im) tcg_gen_shri_i32(cpu_T[1], cpu_T[1], im)
225 #define gen_op_sarl_T1_im(im) tcg_gen_sari_i32(cpu_T[1], cpu_T[1], im)
226 #define gen_op_rorl_T1_im(im) tcg_gen_rori_i32(cpu_T[1], cpu_T[1], im)
227
228 /* Value extensions. */
229 #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var)
230 #define gen_uxth(var) tcg_gen_ext16u_i32(var, var)
231 #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
232 #define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
233
234 #define gen_sxtb16(var) gen_helper_sxtb16(var, var)
235 #define gen_uxtb16(var) gen_helper_uxtb16(var, var)
236
237 #define gen_op_mul_T0_T1() tcg_gen_mul_i32(cpu_T[0], cpu_T[0], cpu_T[1])
238
239 #define gen_set_cpsr(var, mask) gen_helper_cpsr_write(var, tcg_const_i32(mask))
240 /* Set NZCV flags from the high 4 bits of var. */
241 #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV)
242
243 static void gen_exception(int excp)
244 {
245 TCGv tmp = new_tmp();
246 tcg_gen_movi_i32(tmp, excp);
247 gen_helper_exception(tmp);
248 dead_tmp(tmp);
249 }
250
251 static void gen_smul_dual(TCGv a, TCGv b)
252 {
253 TCGv tmp1 = new_tmp();
254 TCGv tmp2 = new_tmp();
255 tcg_gen_ext16s_i32(tmp1, a);
256 tcg_gen_ext16s_i32(tmp2, b);
257 tcg_gen_mul_i32(tmp1, tmp1, tmp2);
258 dead_tmp(tmp2);
259 tcg_gen_sari_i32(a, a, 16);
260 tcg_gen_sari_i32(b, b, 16);
261 tcg_gen_mul_i32(b, b, a);
262 tcg_gen_mov_i32(a, tmp1);
263 dead_tmp(tmp1);
264 }
265
266 /* Byteswap each halfword. */
267 static void gen_rev16(TCGv var)
268 {
269 TCGv tmp = new_tmp();
270 tcg_gen_shri_i32(tmp, var, 8);
271 tcg_gen_andi_i32(tmp, tmp, 0x00ff00ff);
272 tcg_gen_shli_i32(var, var, 8);
273 tcg_gen_andi_i32(var, var, 0xff00ff00);
274 tcg_gen_or_i32(var, var, tmp);
275 dead_tmp(tmp);
276 }
277
278 /* Byteswap low halfword and sign extend. */
279 static void gen_revsh(TCGv var)
280 {
281 TCGv tmp = new_tmp();
282 tcg_gen_shri_i32(tmp, var, 8);
283 tcg_gen_andi_i32(tmp, tmp, 0x00ff);
284 tcg_gen_shli_i32(var, var, 8);
285 tcg_gen_ext8s_i32(var, var);
286 tcg_gen_or_i32(var, var, tmp);
287 dead_tmp(tmp);
288 }
289
290 /* Unsigned bitfield extract. */
291 static void gen_ubfx(TCGv var, int shift, uint32_t mask)
292 {
293 if (shift)
294 tcg_gen_shri_i32(var, var, shift);
295 tcg_gen_andi_i32(var, var, mask);
296 }
297
298 /* Signed bitfield extract. */
299 static void gen_sbfx(TCGv var, int shift, int width)
300 {
301 uint32_t signbit;
302
303 if (shift)
304 tcg_gen_sari_i32(var, var, shift);
305 if (shift + width < 32) {
306 signbit = 1u << (width - 1);
307 tcg_gen_andi_i32(var, var, (1u << width) - 1);
308 tcg_gen_xori_i32(var, var, signbit);
309 tcg_gen_subi_i32(var, var, signbit);
310 }
311 }
312
313 /* Bitfield insertion. Insert val into base. Clobbers base and val. */
314 static void gen_bfi(TCGv dest, TCGv base, TCGv val, int shift, uint32_t mask)
315 {
316 tcg_gen_andi_i32(val, val, mask);
317 tcg_gen_shli_i32(val, val, shift);
318 tcg_gen_andi_i32(base, base, ~(mask << shift));
319 tcg_gen_or_i32(dest, base, val);
320 }
321
322 /* Round the top 32 bits of a 64-bit value. */
323 static void gen_roundqd(TCGv a, TCGv b)
324 {
325 tcg_gen_shri_i32(a, a, 31);
326 tcg_gen_add_i32(a, a, b);
327 }
328
329 /* FIXME: Most targets have native widening multiplication.
330 It would be good to use that instead of a full wide multiply. */
331 /* 32x32->64 multiply. Marks inputs as dead. */
332 static TCGv_i64 gen_mulu_i64_i32(TCGv a, TCGv b)
333 {
334 TCGv_i64 tmp1 = tcg_temp_new_i64();
335 TCGv_i64 tmp2 = tcg_temp_new_i64();
336
337 tcg_gen_extu_i32_i64(tmp1, a);
338 dead_tmp(a);
339 tcg_gen_extu_i32_i64(tmp2, b);
340 dead_tmp(b);
341 tcg_gen_mul_i64(tmp1, tmp1, tmp2);
342 return tmp1;
343 }
344
345 static TCGv_i64 gen_muls_i64_i32(TCGv a, TCGv b)
346 {
347 TCGv_i64 tmp1 = tcg_temp_new_i64();
348 TCGv_i64 tmp2 = tcg_temp_new_i64();
349
350 tcg_gen_ext_i32_i64(tmp1, a);
351 dead_tmp(a);
352 tcg_gen_ext_i32_i64(tmp2, b);
353 dead_tmp(b);
354 tcg_gen_mul_i64(tmp1, tmp1, tmp2);
355 return tmp1;
356 }
357
358 /* Unsigned 32x32->64 multiply. */
359 static void gen_op_mull_T0_T1(void)
360 {
361 TCGv_i64 tmp1 = tcg_temp_new_i64();
362 TCGv_i64 tmp2 = tcg_temp_new_i64();
363
364 tcg_gen_extu_i32_i64(tmp1, cpu_T[0]);
365 tcg_gen_extu_i32_i64(tmp2, cpu_T[1]);
366 tcg_gen_mul_i64(tmp1, tmp1, tmp2);
367 tcg_gen_trunc_i64_i32(cpu_T[0], tmp1);
368 tcg_gen_shri_i64(tmp1, tmp1, 32);
369 tcg_gen_trunc_i64_i32(cpu_T[1], tmp1);
370 }
371
372 /* Signed 32x32->64 multiply. */
373 static void gen_imull(TCGv a, TCGv b)
374 {
375 TCGv_i64 tmp1 = tcg_temp_new_i64();
376 TCGv_i64 tmp2 = tcg_temp_new_i64();
377
378 tcg_gen_ext_i32_i64(tmp1, a);
379 tcg_gen_ext_i32_i64(tmp2, b);
380 tcg_gen_mul_i64(tmp1, tmp1, tmp2);
381 tcg_gen_trunc_i64_i32(a, tmp1);
382 tcg_gen_shri_i64(tmp1, tmp1, 32);
383 tcg_gen_trunc_i64_i32(b, tmp1);
384 }
385 #define gen_op_imull_T0_T1() gen_imull(cpu_T[0], cpu_T[1])
386
387 /* Swap low and high halfwords. */
388 static void gen_swap_half(TCGv var)
389 {
390 TCGv tmp = new_tmp();
391 tcg_gen_shri_i32(tmp, var, 16);
392 tcg_gen_shli_i32(var, var, 16);
393 tcg_gen_or_i32(var, var, tmp);
394 dead_tmp(tmp);
395 }
396
397 /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
398 tmp = (t0 ^ t1) & 0x8000;
399 t0 &= ~0x8000;
400 t1 &= ~0x8000;
401 t0 = (t0 + t1) ^ tmp;
402 */
403
404 static void gen_add16(TCGv t0, TCGv t1)
405 {
406 TCGv tmp = new_tmp();
407 tcg_gen_xor_i32(tmp, t0, t1);
408 tcg_gen_andi_i32(tmp, tmp, 0x8000);
409 tcg_gen_andi_i32(t0, t0, ~0x8000);
410 tcg_gen_andi_i32(t1, t1, ~0x8000);
411 tcg_gen_add_i32(t0, t0, t1);
412 tcg_gen_xor_i32(t0, t0, tmp);
413 dead_tmp(tmp);
414 dead_tmp(t1);
415 }
416
417 #define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, CF))
418
419 /* Set CF to the top bit of var. */
420 static void gen_set_CF_bit31(TCGv var)
421 {
422 TCGv tmp = new_tmp();
423 tcg_gen_shri_i32(tmp, var, 31);
424 gen_set_CF(var);
425 dead_tmp(tmp);
426 }
427
428 /* Set N and Z flags from var. */
429 static inline void gen_logic_CC(TCGv var)
430 {
431 tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, NF));
432 tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, ZF));
433 }
434
435 /* T0 += T1 + CF. */
436 static void gen_adc_T0_T1(void)
437 {
438 TCGv tmp;
439 gen_op_addl_T0_T1();
440 tmp = load_cpu_field(CF);
441 tcg_gen_add_i32(cpu_T[0], cpu_T[0], tmp);
442 dead_tmp(tmp);
443 }
444
445 /* dest = T0 - T1 + CF - 1. */
446 static void gen_sub_carry(TCGv dest, TCGv t0, TCGv t1)
447 {
448 TCGv tmp;
449 tcg_gen_sub_i32(dest, t0, t1);
450 tmp = load_cpu_field(CF);
451 tcg_gen_add_i32(dest, dest, tmp);
452 tcg_gen_subi_i32(dest, dest, 1);
453 dead_tmp(tmp);
454 }
455
456 #define gen_sbc_T0_T1() gen_sub_carry(cpu_T[0], cpu_T[0], cpu_T[1])
457 #define gen_rsc_T0_T1() gen_sub_carry(cpu_T[0], cpu_T[1], cpu_T[0])
458
459 /* T0 &= ~T1. Clobbers T1. */
460 /* FIXME: Implement bic natively. */
461 static inline void tcg_gen_bic_i32(TCGv dest, TCGv t0, TCGv t1)
462 {
463 TCGv tmp = new_tmp();
464 tcg_gen_not_i32(tmp, t1);
465 tcg_gen_and_i32(dest, t0, tmp);
466 dead_tmp(tmp);
467 }
468 static inline void gen_op_bicl_T0_T1(void)
469 {
470 gen_op_notl_T1();
471 gen_op_andl_T0_T1();
472 }
473
474 /* FIXME: Implement this natively. */
475 #define tcg_gen_abs_i32(t0, t1) gen_helper_abs(t0, t1)
476
477 /* FIXME: Implement this natively. */
478 static void tcg_gen_rori_i32(TCGv t0, TCGv t1, int i)
479 {
480 TCGv tmp;
481
482 if (i == 0)
483 return;
484
485 tmp = new_tmp();
486 tcg_gen_shri_i32(tmp, t1, i);
487 tcg_gen_shli_i32(t1, t1, 32 - i);
488 tcg_gen_or_i32(t0, t1, tmp);
489 dead_tmp(tmp);
490 }
491
492 static void shifter_out_im(TCGv var, int shift)
493 {
494 TCGv tmp = new_tmp();
495 if (shift == 0) {
496 tcg_gen_andi_i32(tmp, var, 1);
497 } else {
498 tcg_gen_shri_i32(tmp, var, shift);
499 if (shift != 31);
500 tcg_gen_andi_i32(tmp, tmp, 1);
501 }
502 gen_set_CF(tmp);
503 dead_tmp(tmp);
504 }
505
506 /* Shift by immediate. Includes special handling for shift == 0. */
507 static inline void gen_arm_shift_im(TCGv var, int shiftop, int shift, int flags)
508 {
509 switch (shiftop) {
510 case 0: /* LSL */
511 if (shift != 0) {
512 if (flags)
513 shifter_out_im(var, 32 - shift);
514 tcg_gen_shli_i32(var, var, shift);
515 }
516 break;
517 case 1: /* LSR */
518 if (shift == 0) {
519 if (flags) {
520 tcg_gen_shri_i32(var, var, 31);
521 gen_set_CF(var);
522 }
523 tcg_gen_movi_i32(var, 0);
524 } else {
525 if (flags)
526 shifter_out_im(var, shift - 1);
527 tcg_gen_shri_i32(var, var, shift);
528 }
529 break;
530 case 2: /* ASR */
531 if (shift == 0)
532 shift = 32;
533 if (flags)
534 shifter_out_im(var, shift - 1);
535 if (shift == 32)
536 shift = 31;
537 tcg_gen_sari_i32(var, var, shift);
538 break;
539 case 3: /* ROR/RRX */
540 if (shift != 0) {
541 if (flags)
542 shifter_out_im(var, shift - 1);
543 tcg_gen_rori_i32(var, var, shift); break;
544 } else {
545 TCGv tmp = load_cpu_field(CF);
546 if (flags)
547 shifter_out_im(var, 0);
548 tcg_gen_shri_i32(var, var, 1);
549 tcg_gen_shli_i32(tmp, tmp, 31);
550 tcg_gen_or_i32(var, var, tmp);
551 dead_tmp(tmp);
552 }
553 }
554 };
555
556 static inline void gen_arm_shift_reg(TCGv var, int shiftop,
557 TCGv shift, int flags)
558 {
559 if (flags) {
560 switch (shiftop) {
561 case 0: gen_helper_shl_cc(var, var, shift); break;
562 case 1: gen_helper_shr_cc(var, var, shift); break;
563 case 2: gen_helper_sar_cc(var, var, shift); break;
564 case 3: gen_helper_ror_cc(var, var, shift); break;
565 }
566 } else {
567 switch (shiftop) {
568 case 0: gen_helper_shl(var, var, shift); break;
569 case 1: gen_helper_shr(var, var, shift); break;
570 case 2: gen_helper_sar(var, var, shift); break;
571 case 3: gen_helper_ror(var, var, shift); break;
572 }
573 }
574 dead_tmp(shift);
575 }
576
577 #define PAS_OP(pfx) \
578 switch (op2) { \
579 case 0: gen_pas_helper(glue(pfx,add16)); break; \
580 case 1: gen_pas_helper(glue(pfx,addsubx)); break; \
581 case 2: gen_pas_helper(glue(pfx,subaddx)); break; \
582 case 3: gen_pas_helper(glue(pfx,sub16)); break; \
583 case 4: gen_pas_helper(glue(pfx,add8)); break; \
584 case 7: gen_pas_helper(glue(pfx,sub8)); break; \
585 }
586 static void gen_arm_parallel_addsub(int op1, int op2, TCGv a, TCGv b)
587 {
588 TCGv_ptr tmp;
589
590 switch (op1) {
591 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
592 case 1:
593 tmp = tcg_temp_new_ptr();
594 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
595 PAS_OP(s)
596 break;
597 case 5:
598 tmp = tcg_temp_new_ptr();
599 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
600 PAS_OP(u)
601 break;
602 #undef gen_pas_helper
603 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
604 case 2:
605 PAS_OP(q);
606 break;
607 case 3:
608 PAS_OP(sh);
609 break;
610 case 6:
611 PAS_OP(uq);
612 break;
613 case 7:
614 PAS_OP(uh);
615 break;
616 #undef gen_pas_helper
617 }
618 }
619 #undef PAS_OP
620
621 /* For unknown reasons Arm and Thumb-2 use arbitrarily different encodings. */
622 #define PAS_OP(pfx) \
623 switch (op2) { \
624 case 0: gen_pas_helper(glue(pfx,add8)); break; \
625 case 1: gen_pas_helper(glue(pfx,add16)); break; \
626 case 2: gen_pas_helper(glue(pfx,addsubx)); break; \
627 case 4: gen_pas_helper(glue(pfx,sub8)); break; \
628 case 5: gen_pas_helper(glue(pfx,sub16)); break; \
629 case 6: gen_pas_helper(glue(pfx,subaddx)); break; \
630 }
631 static void gen_thumb2_parallel_addsub(int op1, int op2, TCGv a, TCGv b)
632 {
633 TCGv_ptr tmp;
634
635 switch (op1) {
636 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
637 case 0:
638 tmp = tcg_temp_new_ptr();
639 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
640 PAS_OP(s)
641 break;
642 case 4:
643 tmp = tcg_temp_new_ptr();
644 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
645 PAS_OP(u)
646 break;
647 #undef gen_pas_helper
648 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
649 case 1:
650 PAS_OP(q);
651 break;
652 case 2:
653 PAS_OP(sh);
654 break;
655 case 5:
656 PAS_OP(uq);
657 break;
658 case 6:
659 PAS_OP(uh);
660 break;
661 #undef gen_pas_helper
662 }
663 }
664 #undef PAS_OP
665
666 static void gen_test_cc(int cc, int label)
667 {
668 TCGv tmp;
669 TCGv tmp2;
670 int inv;
671
672 switch (cc) {
673 case 0: /* eq: Z */
674 tmp = load_cpu_field(ZF);
675 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
676 break;
677 case 1: /* ne: !Z */
678 tmp = load_cpu_field(ZF);
679 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, label);
680 break;
681 case 2: /* cs: C */
682 tmp = load_cpu_field(CF);
683 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, label);
684 break;
685 case 3: /* cc: !C */
686 tmp = load_cpu_field(CF);
687 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
688 break;
689 case 4: /* mi: N */
690 tmp = load_cpu_field(NF);
691 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
692 break;
693 case 5: /* pl: !N */
694 tmp = load_cpu_field(NF);
695 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
696 break;
697 case 6: /* vs: V */
698 tmp = load_cpu_field(VF);
699 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
700 break;
701 case 7: /* vc: !V */
702 tmp = load_cpu_field(VF);
703 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
704 break;
705 case 8: /* hi: C && !Z */
706 inv = gen_new_label();
707 tmp = load_cpu_field(CF);
708 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, inv);
709 dead_tmp(tmp);
710 tmp = load_cpu_field(ZF);
711 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, label);
712 gen_set_label(inv);
713 break;
714 case 9: /* ls: !C || Z */
715 tmp = load_cpu_field(CF);
716 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
717 dead_tmp(tmp);
718 tmp = load_cpu_field(ZF);
719 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
720 break;
721 case 10: /* ge: N == V -> N ^ V == 0 */
722 tmp = load_cpu_field(VF);
723 tmp2 = load_cpu_field(NF);
724 tcg_gen_xor_i32(tmp, tmp, tmp2);
725 dead_tmp(tmp2);
726 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
727 break;
728 case 11: /* lt: N != V -> N ^ V != 0 */
729 tmp = load_cpu_field(VF);
730 tmp2 = load_cpu_field(NF);
731 tcg_gen_xor_i32(tmp, tmp, tmp2);
732 dead_tmp(tmp2);
733 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
734 break;
735 case 12: /* gt: !Z && N == V */
736 inv = gen_new_label();
737 tmp = load_cpu_field(ZF);
738 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, inv);
739 dead_tmp(tmp);
740 tmp = load_cpu_field(VF);
741 tmp2 = load_cpu_field(NF);
742 tcg_gen_xor_i32(tmp, tmp, tmp2);
743 dead_tmp(tmp2);
744 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
745 gen_set_label(inv);
746 break;
747 case 13: /* le: Z || N != V */
748 tmp = load_cpu_field(ZF);
749 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
750 dead_tmp(tmp);
751 tmp = load_cpu_field(VF);
752 tmp2 = load_cpu_field(NF);
753 tcg_gen_xor_i32(tmp, tmp, tmp2);
754 dead_tmp(tmp2);
755 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
756 break;
757 default:
758 fprintf(stderr, "Bad condition code 0x%x\n", cc);
759 abort();
760 }
761 dead_tmp(tmp);
762 }
763
764 static const uint8_t table_logic_cc[16] = {
765 1, /* and */
766 1, /* xor */
767 0, /* sub */
768 0, /* rsb */
769 0, /* add */
770 0, /* adc */
771 0, /* sbc */
772 0, /* rsc */
773 1, /* andl */
774 1, /* xorl */
775 0, /* cmp */
776 0, /* cmn */
777 1, /* orr */
778 1, /* mov */
779 1, /* bic */
780 1, /* mvn */
781 };
782
783 /* Set PC and Thumb state from an immediate address. */
784 static inline void gen_bx_im(DisasContext *s, uint32_t addr)
785 {
786 TCGv tmp;
787
788 s->is_jmp = DISAS_UPDATE;
789 tmp = new_tmp();
790 if (s->thumb != (addr & 1)) {
791 tcg_gen_movi_i32(tmp, addr & 1);
792 tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, thumb));
793 }
794 tcg_gen_movi_i32(tmp, addr & ~1);
795 tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, regs[15]));
796 dead_tmp(tmp);
797 }
798
799 /* Set PC and Thumb state from var. var is marked as dead. */
800 static inline void gen_bx(DisasContext *s, TCGv var)
801 {
802 TCGv tmp;
803
804 s->is_jmp = DISAS_UPDATE;
805 tmp = new_tmp();
806 tcg_gen_andi_i32(tmp, var, 1);
807 store_cpu_field(tmp, thumb);
808 tcg_gen_andi_i32(var, var, ~1);
809 store_cpu_field(var, regs[15]);
810 }
811
812 /* TODO: This should be removed. Use gen_bx instead. */
813 static inline void gen_bx_T0(DisasContext *s)
814 {
815 TCGv tmp = new_tmp();
816 tcg_gen_mov_i32(tmp, cpu_T[0]);
817 gen_bx(s, tmp);
818 }
819
820 #if defined(CONFIG_USER_ONLY)
821 #define gen_ldst(name, s) gen_op_##name##_raw()
822 #else
823 #define gen_ldst(name, s) do { \
824 s->is_mem = 1; \
825 if (IS_USER(s)) \
826 gen_op_##name##_user(); \
827 else \
828 gen_op_##name##_kernel(); \
829 } while (0)
830 #endif
831 static inline TCGv gen_ld8s(TCGv addr, int index)
832 {
833 TCGv tmp = new_tmp();
834 tcg_gen_qemu_ld8s(tmp, addr, index);
835 return tmp;
836 }
837 static inline TCGv gen_ld8u(TCGv addr, int index)
838 {
839 TCGv tmp = new_tmp();
840 tcg_gen_qemu_ld8u(tmp, addr, index);
841 return tmp;
842 }
843 static inline TCGv gen_ld16s(TCGv addr, int index)
844 {
845 TCGv tmp = new_tmp();
846 tcg_gen_qemu_ld16s(tmp, addr, index);
847 return tmp;
848 }
849 static inline TCGv gen_ld16u(TCGv addr, int index)
850 {
851 TCGv tmp = new_tmp();
852 tcg_gen_qemu_ld16u(tmp, addr, index);
853 return tmp;
854 }
855 static inline TCGv gen_ld32(TCGv addr, int index)
856 {
857 TCGv tmp = new_tmp();
858 tcg_gen_qemu_ld32u(tmp, addr, index);
859 return tmp;
860 }
861 static inline void gen_st8(TCGv val, TCGv addr, int index)
862 {
863 tcg_gen_qemu_st8(val, addr, index);
864 dead_tmp(val);
865 }
866 static inline void gen_st16(TCGv val, TCGv addr, int index)
867 {
868 tcg_gen_qemu_st16(val, addr, index);
869 dead_tmp(val);
870 }
871 static inline void gen_st32(TCGv val, TCGv addr, int index)
872 {
873 tcg_gen_qemu_st32(val, addr, index);
874 dead_tmp(val);
875 }
876
877 static inline void gen_movl_T0_reg(DisasContext *s, int reg)
878 {
879 load_reg_var(s, cpu_T[0], reg);
880 }
881
882 static inline void gen_movl_T1_reg(DisasContext *s, int reg)
883 {
884 load_reg_var(s, cpu_T[1], reg);
885 }
886
887 static inline void gen_movl_T2_reg(DisasContext *s, int reg)
888 {
889 load_reg_var(s, cpu_T[2], reg);
890 }
891
892 static inline void gen_set_pc_im(uint32_t val)
893 {
894 TCGv tmp = new_tmp();
895 tcg_gen_movi_i32(tmp, val);
896 store_cpu_field(tmp, regs[15]);
897 }
898
899 static inline void gen_movl_reg_TN(DisasContext *s, int reg, int t)
900 {
901 TCGv tmp;
902 if (reg == 15) {
903 tmp = new_tmp();
904 tcg_gen_andi_i32(tmp, cpu_T[t], ~1);
905 } else {
906 tmp = cpu_T[t];
907 }
908 tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, regs[reg]));
909 if (reg == 15) {
910 dead_tmp(tmp);
911 s->is_jmp = DISAS_JUMP;
912 }
913 }
914
915 static inline void gen_movl_reg_T0(DisasContext *s, int reg)
916 {
917 gen_movl_reg_TN(s, reg, 0);
918 }
919
920 static inline void gen_movl_reg_T1(DisasContext *s, int reg)
921 {
922 gen_movl_reg_TN(s, reg, 1);
923 }
924
925 /* Force a TB lookup after an instruction that changes the CPU state. */
926 static inline void gen_lookup_tb(DisasContext *s)
927 {
928 gen_op_movl_T0_im(s->pc);
929 gen_movl_reg_T0(s, 15);
930 s->is_jmp = DISAS_UPDATE;
931 }
932
933 static inline void gen_add_data_offset(DisasContext *s, unsigned int insn,
934 TCGv var)
935 {
936 int val, rm, shift, shiftop;
937 TCGv offset;
938
939 if (!(insn & (1 << 25))) {
940 /* immediate */
941 val = insn & 0xfff;
942 if (!(insn & (1 << 23)))
943 val = -val;
944 if (val != 0)
945 tcg_gen_addi_i32(var, var, val);
946 } else {
947 /* shift/register */
948 rm = (insn) & 0xf;
949 shift = (insn >> 7) & 0x1f;
950 shiftop = (insn >> 5) & 3;
951 offset = load_reg(s, rm);
952 gen_arm_shift_im(offset, shiftop, shift, 0);
953 if (!(insn & (1 << 23)))
954 tcg_gen_sub_i32(var, var, offset);
955 else
956 tcg_gen_add_i32(var, var, offset);
957 dead_tmp(offset);
958 }
959 }
960
961 static inline void gen_add_datah_offset(DisasContext *s, unsigned int insn,
962 int extra, TCGv var)
963 {
964 int val, rm;
965 TCGv offset;
966
967 if (insn & (1 << 22)) {
968 /* immediate */
969 val = (insn & 0xf) | ((insn >> 4) & 0xf0);
970 if (!(insn & (1 << 23)))
971 val = -val;
972 val += extra;
973 if (val != 0)
974 tcg_gen_addi_i32(var, var, val);
975 } else {
976 /* register */
977 if (extra)
978 tcg_gen_addi_i32(var, var, extra);
979 rm = (insn) & 0xf;
980 offset = load_reg(s, rm);
981 if (!(insn & (1 << 23)))
982 tcg_gen_sub_i32(var, var, offset);
983 else
984 tcg_gen_add_i32(var, var, offset);
985 dead_tmp(offset);
986 }
987 }
988
989 #define VFP_OP2(name) \
990 static inline void gen_vfp_##name(int dp) \
991 { \
992 if (dp) \
993 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, cpu_env); \
994 else \
995 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, cpu_env); \
996 }
997
998 #define VFP_OP1(name) \
999 static inline void gen_vfp_##name(int dp, int arg) \
1000 { \
1001 if (dp) \
1002 gen_op_vfp_##name##d(arg); \
1003 else \
1004 gen_op_vfp_##name##s(arg); \
1005 }
1006
1007 VFP_OP2(add)
1008 VFP_OP2(sub)
1009 VFP_OP2(mul)
1010 VFP_OP2(div)
1011
1012 #undef VFP_OP2
1013
1014 static inline void gen_vfp_abs(int dp)
1015 {
1016 if (dp)
1017 gen_helper_vfp_absd(cpu_F0d, cpu_F0d);
1018 else
1019 gen_helper_vfp_abss(cpu_F0s, cpu_F0s);
1020 }
1021
1022 static inline void gen_vfp_neg(int dp)
1023 {
1024 if (dp)
1025 gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
1026 else
1027 gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
1028 }
1029
1030 static inline void gen_vfp_sqrt(int dp)
1031 {
1032 if (dp)
1033 gen_helper_vfp_sqrtd(cpu_F0d, cpu_F0d, cpu_env);
1034 else
1035 gen_helper_vfp_sqrts(cpu_F0s, cpu_F0s, cpu_env);
1036 }
1037
1038 static inline void gen_vfp_cmp(int dp)
1039 {
1040 if (dp)
1041 gen_helper_vfp_cmpd(cpu_F0d, cpu_F1d, cpu_env);
1042 else
1043 gen_helper_vfp_cmps(cpu_F0s, cpu_F1s, cpu_env);
1044 }
1045
1046 static inline void gen_vfp_cmpe(int dp)
1047 {
1048 if (dp)
1049 gen_helper_vfp_cmped(cpu_F0d, cpu_F1d, cpu_env);
1050 else
1051 gen_helper_vfp_cmpes(cpu_F0s, cpu_F1s, cpu_env);
1052 }
1053
1054 static inline void gen_vfp_F1_ld0(int dp)
1055 {
1056 if (dp)
1057 tcg_gen_movi_i64(cpu_F1d, 0);
1058 else
1059 tcg_gen_movi_i32(cpu_F1s, 0);
1060 }
1061
1062 static inline void gen_vfp_uito(int dp)
1063 {
1064 if (dp)
1065 gen_helper_vfp_uitod(cpu_F0d, cpu_F0s, cpu_env);
1066 else
1067 gen_helper_vfp_uitos(cpu_F0s, cpu_F0s, cpu_env);
1068 }
1069
1070 static inline void gen_vfp_sito(int dp)
1071 {
1072 if (dp)
1073 gen_helper_vfp_sitod(cpu_F0d, cpu_F0s, cpu_env);
1074 else
1075 gen_helper_vfp_sitos(cpu_F0s, cpu_F0s, cpu_env);
1076 }
1077
1078 static inline void gen_vfp_toui(int dp)
1079 {
1080 if (dp)
1081 gen_helper_vfp_touid(cpu_F0s, cpu_F0d, cpu_env);
1082 else
1083 gen_helper_vfp_touis(cpu_F0s, cpu_F0s, cpu_env);
1084 }
1085
1086 static inline void gen_vfp_touiz(int dp)
1087 {
1088 if (dp)
1089 gen_helper_vfp_touizd(cpu_F0s, cpu_F0d, cpu_env);
1090 else
1091 gen_helper_vfp_touizs(cpu_F0s, cpu_F0s, cpu_env);
1092 }
1093
1094 static inline void gen_vfp_tosi(int dp)
1095 {
1096 if (dp)
1097 gen_helper_vfp_tosid(cpu_F0s, cpu_F0d, cpu_env);
1098 else
1099 gen_helper_vfp_tosis(cpu_F0s, cpu_F0s, cpu_env);
1100 }
1101
1102 static inline void gen_vfp_tosiz(int dp)
1103 {
1104 if (dp)
1105 gen_helper_vfp_tosizd(cpu_F0s, cpu_F0d, cpu_env);
1106 else
1107 gen_helper_vfp_tosizs(cpu_F0s, cpu_F0s, cpu_env);
1108 }
1109
1110 #define VFP_GEN_FIX(name) \
1111 static inline void gen_vfp_##name(int dp, int shift) \
1112 { \
1113 if (dp) \
1114 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tcg_const_i32(shift), cpu_env);\
1115 else \
1116 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, tcg_const_i32(shift), cpu_env);\
1117 }
1118 VFP_GEN_FIX(tosh)
1119 VFP_GEN_FIX(tosl)
1120 VFP_GEN_FIX(touh)
1121 VFP_GEN_FIX(toul)
1122 VFP_GEN_FIX(shto)
1123 VFP_GEN_FIX(slto)
1124 VFP_GEN_FIX(uhto)
1125 VFP_GEN_FIX(ulto)
1126 #undef VFP_GEN_FIX
1127
1128 static inline void gen_vfp_ld(DisasContext *s, int dp)
1129 {
1130 if (dp)
1131 tcg_gen_qemu_ld64(cpu_F0d, cpu_T[1], IS_USER(s));
1132 else
1133 tcg_gen_qemu_ld32u(cpu_F0s, cpu_T[1], IS_USER(s));
1134 }
1135
1136 static inline void gen_vfp_st(DisasContext *s, int dp)
1137 {
1138 if (dp)
1139 tcg_gen_qemu_st64(cpu_F0d, cpu_T[1], IS_USER(s));
1140 else
1141 tcg_gen_qemu_st32(cpu_F0s, cpu_T[1], IS_USER(s));
1142 }
1143
1144 static inline long
1145 vfp_reg_offset (int dp, int reg)
1146 {
1147 if (dp)
1148 return offsetof(CPUARMState, vfp.regs[reg]);
1149 else if (reg & 1) {
1150 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1151 + offsetof(CPU_DoubleU, l.upper);
1152 } else {
1153 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1154 + offsetof(CPU_DoubleU, l.lower);
1155 }
1156 }
1157
1158 /* Return the offset of a 32-bit piece of a NEON register.
1159 zero is the least significant end of the register. */
1160 static inline long
1161 neon_reg_offset (int reg, int n)
1162 {
1163 int sreg;
1164 sreg = reg * 2 + n;
1165 return vfp_reg_offset(0, sreg);
1166 }
1167
1168 /* FIXME: Remove these. */
1169 #define neon_T0 cpu_T[0]
1170 #define neon_T1 cpu_T[1]
1171 #define NEON_GET_REG(T, reg, n) \
1172 tcg_gen_ld_i32(neon_##T, cpu_env, neon_reg_offset(reg, n))
1173 #define NEON_SET_REG(T, reg, n) \
1174 tcg_gen_st_i32(neon_##T, cpu_env, neon_reg_offset(reg, n))
1175
1176 static TCGv neon_load_reg(int reg, int pass)
1177 {
1178 TCGv tmp = new_tmp();
1179 tcg_gen_ld_i32(tmp, cpu_env, neon_reg_offset(reg, pass));
1180 return tmp;
1181 }
1182
1183 static void neon_store_reg(int reg, int pass, TCGv var)
1184 {
1185 tcg_gen_st_i32(var, cpu_env, neon_reg_offset(reg, pass));
1186 dead_tmp(var);
1187 }
1188
1189 static inline void neon_load_reg64(TCGv_i64 var, int reg)
1190 {
1191 tcg_gen_ld_i64(var, cpu_env, vfp_reg_offset(1, reg));
1192 }
1193
1194 static inline void neon_store_reg64(TCGv_i64 var, int reg)
1195 {
1196 tcg_gen_st_i64(var, cpu_env, vfp_reg_offset(1, reg));
1197 }
1198
1199 #define tcg_gen_ld_f32 tcg_gen_ld_i32
1200 #define tcg_gen_ld_f64 tcg_gen_ld_i64
1201 #define tcg_gen_st_f32 tcg_gen_st_i32
1202 #define tcg_gen_st_f64 tcg_gen_st_i64
1203
1204 static inline void gen_mov_F0_vreg(int dp, int reg)
1205 {
1206 if (dp)
1207 tcg_gen_ld_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1208 else
1209 tcg_gen_ld_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1210 }
1211
1212 static inline void gen_mov_F1_vreg(int dp, int reg)
1213 {
1214 if (dp)
1215 tcg_gen_ld_f64(cpu_F1d, cpu_env, vfp_reg_offset(dp, reg));
1216 else
1217 tcg_gen_ld_f32(cpu_F1s, cpu_env, vfp_reg_offset(dp, reg));
1218 }
1219
1220 static inline void gen_mov_vreg_F0(int dp, int reg)
1221 {
1222 if (dp)
1223 tcg_gen_st_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1224 else
1225 tcg_gen_st_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1226 }
1227
1228 #define ARM_CP_RW_BIT (1 << 20)
1229
1230 static inline void iwmmxt_load_reg(TCGv_i64 var, int reg)
1231 {
1232 tcg_gen_ld_i64(var, cpu_env, offsetof(CPUState, iwmmxt.regs[reg]));
1233 }
1234
1235 static inline void iwmmxt_store_reg(TCGv_i64 var, int reg)
1236 {
1237 tcg_gen_st_i64(var, cpu_env, offsetof(CPUState, iwmmxt.regs[reg]));
1238 }
1239
1240 static inline void gen_op_iwmmxt_movl_wCx_T0(int reg)
1241 {
1242 tcg_gen_st_i32(cpu_T[0], cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
1243 }
1244
1245 static inline void gen_op_iwmmxt_movl_T0_wCx(int reg)
1246 {
1247 tcg_gen_ld_i32(cpu_T[0], cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
1248 }
1249
1250 static inline void gen_op_iwmmxt_movl_T1_wCx(int reg)
1251 {
1252 tcg_gen_ld_i32(cpu_T[1], cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
1253 }
1254
1255 static inline void gen_op_iwmmxt_movq_wRn_M0(int rn)
1256 {
1257 iwmmxt_store_reg(cpu_M0, rn);
1258 }
1259
1260 static inline void gen_op_iwmmxt_movq_M0_wRn(int rn)
1261 {
1262 iwmmxt_load_reg(cpu_M0, rn);
1263 }
1264
1265 static inline void gen_op_iwmmxt_orq_M0_wRn(int rn)
1266 {
1267 iwmmxt_load_reg(cpu_V1, rn);
1268 tcg_gen_or_i64(cpu_M0, cpu_M0, cpu_V1);
1269 }
1270
1271 static inline void gen_op_iwmmxt_andq_M0_wRn(int rn)
1272 {
1273 iwmmxt_load_reg(cpu_V1, rn);
1274 tcg_gen_and_i64(cpu_M0, cpu_M0, cpu_V1);
1275 }
1276
1277 static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn)
1278 {
1279 iwmmxt_load_reg(cpu_V1, rn);
1280 tcg_gen_xor_i64(cpu_M0, cpu_M0, cpu_V1);
1281 }
1282
1283 #define IWMMXT_OP(name) \
1284 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1285 { \
1286 iwmmxt_load_reg(cpu_V1, rn); \
1287 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
1288 }
1289
1290 #define IWMMXT_OP_ENV(name) \
1291 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1292 { \
1293 iwmmxt_load_reg(cpu_V1, rn); \
1294 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \
1295 }
1296
1297 #define IWMMXT_OP_ENV_SIZE(name) \
1298 IWMMXT_OP_ENV(name##b) \
1299 IWMMXT_OP_ENV(name##w) \
1300 IWMMXT_OP_ENV(name##l)
1301
1302 #define IWMMXT_OP_ENV1(name) \
1303 static inline void gen_op_iwmmxt_##name##_M0(void) \
1304 { \
1305 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \
1306 }
1307
1308 IWMMXT_OP(maddsq)
1309 IWMMXT_OP(madduq)
1310 IWMMXT_OP(sadb)
1311 IWMMXT_OP(sadw)
1312 IWMMXT_OP(mulslw)
1313 IWMMXT_OP(mulshw)
1314 IWMMXT_OP(mululw)
1315 IWMMXT_OP(muluhw)
1316 IWMMXT_OP(macsw)
1317 IWMMXT_OP(macuw)
1318
1319 IWMMXT_OP_ENV_SIZE(unpackl)
1320 IWMMXT_OP_ENV_SIZE(unpackh)
1321
1322 IWMMXT_OP_ENV1(unpacklub)
1323 IWMMXT_OP_ENV1(unpackluw)
1324 IWMMXT_OP_ENV1(unpacklul)
1325 IWMMXT_OP_ENV1(unpackhub)
1326 IWMMXT_OP_ENV1(unpackhuw)
1327 IWMMXT_OP_ENV1(unpackhul)
1328 IWMMXT_OP_ENV1(unpacklsb)
1329 IWMMXT_OP_ENV1(unpacklsw)
1330 IWMMXT_OP_ENV1(unpacklsl)
1331 IWMMXT_OP_ENV1(unpackhsb)
1332 IWMMXT_OP_ENV1(unpackhsw)
1333 IWMMXT_OP_ENV1(unpackhsl)
1334
1335 IWMMXT_OP_ENV_SIZE(cmpeq)
1336 IWMMXT_OP_ENV_SIZE(cmpgtu)
1337 IWMMXT_OP_ENV_SIZE(cmpgts)
1338
1339 IWMMXT_OP_ENV_SIZE(mins)
1340 IWMMXT_OP_ENV_SIZE(minu)
1341 IWMMXT_OP_ENV_SIZE(maxs)
1342 IWMMXT_OP_ENV_SIZE(maxu)
1343
1344 IWMMXT_OP_ENV_SIZE(subn)
1345 IWMMXT_OP_ENV_SIZE(addn)
1346 IWMMXT_OP_ENV_SIZE(subu)
1347 IWMMXT_OP_ENV_SIZE(addu)
1348 IWMMXT_OP_ENV_SIZE(subs)
1349 IWMMXT_OP_ENV_SIZE(adds)
1350
1351 IWMMXT_OP_ENV(avgb0)
1352 IWMMXT_OP_ENV(avgb1)
1353 IWMMXT_OP_ENV(avgw0)
1354 IWMMXT_OP_ENV(avgw1)
1355
1356 IWMMXT_OP(msadb)
1357
1358 IWMMXT_OP_ENV(packuw)
1359 IWMMXT_OP_ENV(packul)
1360 IWMMXT_OP_ENV(packuq)
1361 IWMMXT_OP_ENV(packsw)
1362 IWMMXT_OP_ENV(packsl)
1363 IWMMXT_OP_ENV(packsq)
1364
1365 static inline void gen_op_iwmmxt_muladdsl_M0_T0_T1(void)
1366 {
1367 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, cpu_T[0], cpu_T[1]);
1368 }
1369
1370 static inline void gen_op_iwmmxt_muladdsw_M0_T0_T1(void)
1371 {
1372 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, cpu_T[0], cpu_T[1]);
1373 }
1374
1375 static inline void gen_op_iwmmxt_muladdswl_M0_T0_T1(void)
1376 {
1377 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, cpu_T[0], cpu_T[1]);
1378 }
1379
1380 static inline void gen_op_iwmmxt_align_M0_T0_wRn(int rn)
1381 {
1382 iwmmxt_load_reg(cpu_V1, rn);
1383 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, cpu_T[0]);
1384 }
1385
1386 static inline void gen_op_iwmmxt_insr_M0_T0_T1(int shift)
1387 {
1388 TCGv tmp = tcg_const_i32(shift);
1389 gen_helper_iwmmxt_insr(cpu_M0, cpu_M0, cpu_T[0], cpu_T[1], tmp);
1390 }
1391
1392 static inline void gen_op_iwmmxt_extrsb_T0_M0(int shift)
1393 {
1394 tcg_gen_shri_i64(cpu_M0, cpu_M0, shift);
1395 tcg_gen_trunc_i64_i32(cpu_T[0], cpu_M0);
1396 tcg_gen_ext8s_i32(cpu_T[0], cpu_T[0]);
1397 }
1398
1399 static inline void gen_op_iwmmxt_extrsw_T0_M0(int shift)
1400 {
1401 tcg_gen_shri_i64(cpu_M0, cpu_M0, shift);
1402 tcg_gen_trunc_i64_i32(cpu_T[0], cpu_M0);
1403 tcg_gen_ext16s_i32(cpu_T[0], cpu_T[0]);
1404 }
1405
1406 static inline void gen_op_iwmmxt_extru_T0_M0(int shift, uint32_t mask)
1407 {
1408 tcg_gen_shri_i64(cpu_M0, cpu_M0, shift);
1409 tcg_gen_trunc_i64_i32(cpu_T[0], cpu_M0);
1410 if (mask != ~0u)
1411 tcg_gen_andi_i32(cpu_T[0], cpu_T[0], mask);
1412 }
1413
1414 static void gen_op_iwmmxt_set_mup(void)
1415 {
1416 TCGv tmp;
1417 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1418 tcg_gen_ori_i32(tmp, tmp, 2);
1419 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1420 }
1421
1422 static void gen_op_iwmmxt_set_cup(void)
1423 {
1424 TCGv tmp;
1425 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1426 tcg_gen_ori_i32(tmp, tmp, 1);
1427 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1428 }
1429
1430 static void gen_op_iwmmxt_setpsr_nz(void)
1431 {
1432 TCGv tmp = new_tmp();
1433 gen_helper_iwmmxt_setpsr_nz(tmp, cpu_M0);
1434 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCASF]);
1435 }
1436
1437 static inline void gen_op_iwmmxt_addl_M0_wRn(int rn)
1438 {
1439 iwmmxt_load_reg(cpu_V1, rn);
1440 tcg_gen_ext32u_i64(cpu_V1, cpu_V1);
1441 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1442 }
1443
1444
1445 static void gen_iwmmxt_movl_T0_T1_wRn(int rn)
1446 {
1447 iwmmxt_load_reg(cpu_V0, rn);
1448 tcg_gen_trunc_i64_i32(cpu_T[0], cpu_V0);
1449 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
1450 tcg_gen_trunc_i64_i32(cpu_T[1], cpu_V0);
1451 }
1452
1453 static void gen_iwmmxt_movl_wRn_T0_T1(int rn)
1454 {
1455 tcg_gen_concat_i32_i64(cpu_V0, cpu_T[0], cpu_T[1]);
1456 iwmmxt_store_reg(cpu_V0, rn);
1457 }
1458
1459 static inline int gen_iwmmxt_address(DisasContext *s, uint32_t insn)
1460 {
1461 int rd;
1462 uint32_t offset;
1463
1464 rd = (insn >> 16) & 0xf;
1465 gen_movl_T1_reg(s, rd);
1466
1467 offset = (insn & 0xff) << ((insn >> 7) & 2);
1468 if (insn & (1 << 24)) {
1469 /* Pre indexed */
1470 if (insn & (1 << 23))
1471 gen_op_addl_T1_im(offset);
1472 else
1473 gen_op_addl_T1_im(-offset);
1474
1475 if (insn & (1 << 21))
1476 gen_movl_reg_T1(s, rd);
1477 } else if (insn & (1 << 21)) {
1478 /* Post indexed */
1479 if (insn & (1 << 23))
1480 gen_op_movl_T0_im(offset);
1481 else
1482 gen_op_movl_T0_im(- offset);
1483 gen_op_addl_T0_T1();
1484 gen_movl_reg_T0(s, rd);
1485 } else if (!(insn & (1 << 23)))
1486 return 1;
1487 return 0;
1488 }
1489
1490 static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask)
1491 {
1492 int rd = (insn >> 0) & 0xf;
1493
1494 if (insn & (1 << 8))
1495 if (rd < ARM_IWMMXT_wCGR0 || rd > ARM_IWMMXT_wCGR3)
1496 return 1;
1497 else
1498 gen_op_iwmmxt_movl_T0_wCx(rd);
1499 else
1500 gen_iwmmxt_movl_T0_T1_wRn(rd);
1501
1502 gen_op_movl_T1_im(mask);
1503 gen_op_andl_T0_T1();
1504 return 0;
1505 }
1506
1507 /* Disassemble an iwMMXt instruction. Returns nonzero if an error occured
1508 (ie. an undefined instruction). */
1509 static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
1510 {
1511 int rd, wrd;
1512 int rdhi, rdlo, rd0, rd1, i;
1513 TCGv tmp;
1514
1515 if ((insn & 0x0e000e00) == 0x0c000000) {
1516 if ((insn & 0x0fe00ff0) == 0x0c400000) {
1517 wrd = insn & 0xf;
1518 rdlo = (insn >> 12) & 0xf;
1519 rdhi = (insn >> 16) & 0xf;
1520 if (insn & ARM_CP_RW_BIT) { /* TMRRC */
1521 gen_iwmmxt_movl_T0_T1_wRn(wrd);
1522 gen_movl_reg_T0(s, rdlo);
1523 gen_movl_reg_T1(s, rdhi);
1524 } else { /* TMCRR */
1525 gen_movl_T0_reg(s, rdlo);
1526 gen_movl_T1_reg(s, rdhi);
1527 gen_iwmmxt_movl_wRn_T0_T1(wrd);
1528 gen_op_iwmmxt_set_mup();
1529 }
1530 return 0;
1531 }
1532
1533 wrd = (insn >> 12) & 0xf;
1534 if (gen_iwmmxt_address(s, insn))
1535 return 1;
1536 if (insn & ARM_CP_RW_BIT) {
1537 if ((insn >> 28) == 0xf) { /* WLDRW wCx */
1538 tmp = gen_ld32(cpu_T[1], IS_USER(s));
1539 tcg_gen_mov_i32(cpu_T[0], tmp);
1540 dead_tmp(tmp);
1541 gen_op_iwmmxt_movl_wCx_T0(wrd);
1542 } else {
1543 i = 1;
1544 if (insn & (1 << 8)) {
1545 if (insn & (1 << 22)) { /* WLDRD */
1546 tcg_gen_qemu_ld64(cpu_M0, cpu_T[1], IS_USER(s));
1547 i = 0;
1548 } else { /* WLDRW wRd */
1549 tmp = gen_ld32(cpu_T[1], IS_USER(s));
1550 }
1551 } else {
1552 if (insn & (1 << 22)) { /* WLDRH */
1553 tmp = gen_ld16u(cpu_T[1], IS_USER(s));
1554 } else { /* WLDRB */
1555 tmp = gen_ld8u(cpu_T[1], IS_USER(s));
1556 }
1557 }
1558 if (i) {
1559 tcg_gen_extu_i32_i64(cpu_M0, tmp);
1560 dead_tmp(tmp);
1561 }
1562 gen_op_iwmmxt_movq_wRn_M0(wrd);
1563 }
1564 } else {
1565 if ((insn >> 28) == 0xf) { /* WSTRW wCx */
1566 gen_op_iwmmxt_movl_T0_wCx(wrd);
1567 tmp = new_tmp();
1568 tcg_gen_mov_i32(tmp, cpu_T[0]);
1569 gen_st32(tmp, cpu_T[1], IS_USER(s));
1570 } else {
1571 gen_op_iwmmxt_movq_M0_wRn(wrd);
1572 tmp = new_tmp();
1573 if (insn & (1 << 8)) {
1574 if (insn & (1 << 22)) { /* WSTRD */
1575 dead_tmp(tmp);
1576 tcg_gen_qemu_st64(cpu_M0, cpu_T[1], IS_USER(s));
1577 } else { /* WSTRW wRd */
1578 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1579 gen_st32(tmp, cpu_T[1], IS_USER(s));
1580 }
1581 } else {
1582 if (insn & (1 << 22)) { /* WSTRH */
1583 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1584 gen_st16(tmp, cpu_T[1], IS_USER(s));
1585 } else { /* WSTRB */
1586 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1587 gen_st8(tmp, cpu_T[1], IS_USER(s));
1588 }
1589 }
1590 }
1591 }
1592 return 0;
1593 }
1594
1595 if ((insn & 0x0f000000) != 0x0e000000)
1596 return 1;
1597
1598 switch (((insn >> 12) & 0xf00) | ((insn >> 4) & 0xff)) {
1599 case 0x000: /* WOR */
1600 wrd = (insn >> 12) & 0xf;
1601 rd0 = (insn >> 0) & 0xf;
1602 rd1 = (insn >> 16) & 0xf;
1603 gen_op_iwmmxt_movq_M0_wRn(rd0);
1604 gen_op_iwmmxt_orq_M0_wRn(rd1);
1605 gen_op_iwmmxt_setpsr_nz();
1606 gen_op_iwmmxt_movq_wRn_M0(wrd);
1607 gen_op_iwmmxt_set_mup();
1608 gen_op_iwmmxt_set_cup();
1609 break;
1610 case 0x011: /* TMCR */
1611 if (insn & 0xf)
1612 return 1;
1613 rd = (insn >> 12) & 0xf;
1614 wrd = (insn >> 16) & 0xf;
1615 switch (wrd) {
1616 case ARM_IWMMXT_wCID:
1617 case ARM_IWMMXT_wCASF:
1618 break;
1619 case ARM_IWMMXT_wCon:
1620 gen_op_iwmmxt_set_cup();
1621 /* Fall through. */
1622 case ARM_IWMMXT_wCSSF:
1623 gen_op_iwmmxt_movl_T0_wCx(wrd);
1624 gen_movl_T1_reg(s, rd);
1625 gen_op_bicl_T0_T1();
1626 gen_op_iwmmxt_movl_wCx_T0(wrd);
1627 break;
1628 case ARM_IWMMXT_wCGR0:
1629 case ARM_IWMMXT_wCGR1:
1630 case ARM_IWMMXT_wCGR2:
1631 case ARM_IWMMXT_wCGR3:
1632 gen_op_iwmmxt_set_cup();
1633 gen_movl_reg_T0(s, rd);
1634 gen_op_iwmmxt_movl_wCx_T0(wrd);
1635 break;
1636 default:
1637 return 1;
1638 }
1639 break;
1640 case 0x100: /* WXOR */
1641 wrd = (insn >> 12) & 0xf;
1642 rd0 = (insn >> 0) & 0xf;
1643 rd1 = (insn >> 16) & 0xf;
1644 gen_op_iwmmxt_movq_M0_wRn(rd0);
1645 gen_op_iwmmxt_xorq_M0_wRn(rd1);
1646 gen_op_iwmmxt_setpsr_nz();
1647 gen_op_iwmmxt_movq_wRn_M0(wrd);
1648 gen_op_iwmmxt_set_mup();
1649 gen_op_iwmmxt_set_cup();
1650 break;
1651 case 0x111: /* TMRC */
1652 if (insn & 0xf)
1653 return 1;
1654 rd = (insn >> 12) & 0xf;
1655 wrd = (insn >> 16) & 0xf;
1656 gen_op_iwmmxt_movl_T0_wCx(wrd);
1657 gen_movl_reg_T0(s, rd);
1658 break;
1659 case 0x300: /* WANDN */
1660 wrd = (insn >> 12) & 0xf;
1661 rd0 = (insn >> 0) & 0xf;
1662 rd1 = (insn >> 16) & 0xf;
1663 gen_op_iwmmxt_movq_M0_wRn(rd0);
1664 tcg_gen_neg_i64(cpu_M0, cpu_M0);
1665 gen_op_iwmmxt_andq_M0_wRn(rd1);
1666 gen_op_iwmmxt_setpsr_nz();
1667 gen_op_iwmmxt_movq_wRn_M0(wrd);
1668 gen_op_iwmmxt_set_mup();
1669 gen_op_iwmmxt_set_cup();
1670 break;
1671 case 0x200: /* WAND */
1672 wrd = (insn >> 12) & 0xf;
1673 rd0 = (insn >> 0) & 0xf;
1674 rd1 = (insn >> 16) & 0xf;
1675 gen_op_iwmmxt_movq_M0_wRn(rd0);
1676 gen_op_iwmmxt_andq_M0_wRn(rd1);
1677 gen_op_iwmmxt_setpsr_nz();
1678 gen_op_iwmmxt_movq_wRn_M0(wrd);
1679 gen_op_iwmmxt_set_mup();
1680 gen_op_iwmmxt_set_cup();
1681 break;
1682 case 0x810: case 0xa10: /* WMADD */
1683 wrd = (insn >> 12) & 0xf;
1684 rd0 = (insn >> 0) & 0xf;
1685 rd1 = (insn >> 16) & 0xf;
1686 gen_op_iwmmxt_movq_M0_wRn(rd0);
1687 if (insn & (1 << 21))
1688 gen_op_iwmmxt_maddsq_M0_wRn(rd1);
1689 else
1690 gen_op_iwmmxt_madduq_M0_wRn(rd1);
1691 gen_op_iwmmxt_movq_wRn_M0(wrd);
1692 gen_op_iwmmxt_set_mup();
1693 break;
1694 case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */
1695 wrd = (insn >> 12) & 0xf;
1696 rd0 = (insn >> 16) & 0xf;
1697 rd1 = (insn >> 0) & 0xf;
1698 gen_op_iwmmxt_movq_M0_wRn(rd0);
1699 switch ((insn >> 22) & 3) {
1700 case 0:
1701 gen_op_iwmmxt_unpacklb_M0_wRn(rd1);
1702 break;
1703 case 1:
1704 gen_op_iwmmxt_unpacklw_M0_wRn(rd1);
1705 break;
1706 case 2:
1707 gen_op_iwmmxt_unpackll_M0_wRn(rd1);
1708 break;
1709 case 3:
1710 return 1;
1711 }
1712 gen_op_iwmmxt_movq_wRn_M0(wrd);
1713 gen_op_iwmmxt_set_mup();
1714 gen_op_iwmmxt_set_cup();
1715 break;
1716 case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */
1717 wrd = (insn >> 12) & 0xf;
1718 rd0 = (insn >> 16) & 0xf;
1719 rd1 = (insn >> 0) & 0xf;
1720 gen_op_iwmmxt_movq_M0_wRn(rd0);
1721 switch ((insn >> 22) & 3) {
1722 case 0:
1723 gen_op_iwmmxt_unpackhb_M0_wRn(rd1);
1724 break;
1725 case 1:
1726 gen_op_iwmmxt_unpackhw_M0_wRn(rd1);
1727 break;
1728 case 2:
1729 gen_op_iwmmxt_unpackhl_M0_wRn(rd1);
1730 break;
1731 case 3:
1732 return 1;
1733 }
1734 gen_op_iwmmxt_movq_wRn_M0(wrd);
1735 gen_op_iwmmxt_set_mup();
1736 gen_op_iwmmxt_set_cup();
1737 break;
1738 case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */
1739 wrd = (insn >> 12) & 0xf;
1740 rd0 = (insn >> 16) & 0xf;
1741 rd1 = (insn >> 0) & 0xf;
1742 gen_op_iwmmxt_movq_M0_wRn(rd0);
1743 if (insn & (1 << 22))
1744 gen_op_iwmmxt_sadw_M0_wRn(rd1);
1745 else
1746 gen_op_iwmmxt_sadb_M0_wRn(rd1);
1747 if (!(insn & (1 << 20)))
1748 gen_op_iwmmxt_addl_M0_wRn(wrd);
1749 gen_op_iwmmxt_movq_wRn_M0(wrd);
1750 gen_op_iwmmxt_set_mup();
1751 break;
1752 case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */
1753 wrd = (insn >> 12) & 0xf;
1754 rd0 = (insn >> 16) & 0xf;
1755 rd1 = (insn >> 0) & 0xf;
1756 gen_op_iwmmxt_movq_M0_wRn(rd0);
1757 if (insn & (1 << 21)) {
1758 if (insn & (1 << 20))
1759 gen_op_iwmmxt_mulshw_M0_wRn(rd1);
1760 else
1761 gen_op_iwmmxt_mulslw_M0_wRn(rd1);
1762 } else {
1763 if (insn & (1 << 20))
1764 gen_op_iwmmxt_muluhw_M0_wRn(rd1);
1765 else
1766 gen_op_iwmmxt_mululw_M0_wRn(rd1);
1767 }
1768 gen_op_iwmmxt_movq_wRn_M0(wrd);
1769 gen_op_iwmmxt_set_mup();
1770 break;
1771 case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */
1772 wrd = (insn >> 12) & 0xf;
1773 rd0 = (insn >> 16) & 0xf;
1774 rd1 = (insn >> 0) & 0xf;
1775 gen_op_iwmmxt_movq_M0_wRn(rd0);
1776 if (insn & (1 << 21))
1777 gen_op_iwmmxt_macsw_M0_wRn(rd1);
1778 else
1779 gen_op_iwmmxt_macuw_M0_wRn(rd1);
1780 if (!(insn & (1 << 20))) {
1781 iwmmxt_load_reg(cpu_V1, wrd);
1782 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1783 }
1784 gen_op_iwmmxt_movq_wRn_M0(wrd);
1785 gen_op_iwmmxt_set_mup();
1786 break;
1787 case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */
1788 wrd = (insn >> 12) & 0xf;
1789 rd0 = (insn >> 16) & 0xf;
1790 rd1 = (insn >> 0) & 0xf;
1791 gen_op_iwmmxt_movq_M0_wRn(rd0);
1792 switch ((insn >> 22) & 3) {
1793 case 0:
1794 gen_op_iwmmxt_cmpeqb_M0_wRn(rd1);
1795 break;
1796 case 1:
1797 gen_op_iwmmxt_cmpeqw_M0_wRn(rd1);
1798 break;
1799 case 2:
1800 gen_op_iwmmxt_cmpeql_M0_wRn(rd1);
1801 break;
1802 case 3:
1803 return 1;
1804 }
1805 gen_op_iwmmxt_movq_wRn_M0(wrd);
1806 gen_op_iwmmxt_set_mup();
1807 gen_op_iwmmxt_set_cup();
1808 break;
1809 case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */
1810 wrd = (insn >> 12) & 0xf;
1811 rd0 = (insn >> 16) & 0xf;
1812 rd1 = (insn >> 0) & 0xf;
1813 gen_op_iwmmxt_movq_M0_wRn(rd0);
1814 if (insn & (1 << 22)) {
1815 if (insn & (1 << 20))
1816 gen_op_iwmmxt_avgw1_M0_wRn(rd1);
1817 else
1818 gen_op_iwmmxt_avgw0_M0_wRn(rd1);
1819 } else {
1820 if (insn & (1 << 20))
1821 gen_op_iwmmxt_avgb1_M0_wRn(rd1);
1822 else
1823 gen_op_iwmmxt_avgb0_M0_wRn(rd1);
1824 }
1825 gen_op_iwmmxt_movq_wRn_M0(wrd);
1826 gen_op_iwmmxt_set_mup();
1827 gen_op_iwmmxt_set_cup();
1828 break;
1829 case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */
1830 wrd = (insn >> 12) & 0xf;
1831 rd0 = (insn >> 16) & 0xf;
1832 rd1 = (insn >> 0) & 0xf;
1833 gen_op_iwmmxt_movq_M0_wRn(rd0);
1834 gen_op_iwmmxt_movl_T0_wCx(ARM_IWMMXT_wCGR0 + ((insn >> 20) & 3));
1835 gen_op_movl_T1_im(7);
1836 gen_op_andl_T0_T1();
1837 gen_op_iwmmxt_align_M0_T0_wRn(rd1);
1838 gen_op_iwmmxt_movq_wRn_M0(wrd);
1839 gen_op_iwmmxt_set_mup();
1840 break;
1841 case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */
1842 rd = (insn >> 12) & 0xf;
1843 wrd = (insn >> 16) & 0xf;
1844 gen_movl_T0_reg(s, rd);
1845 gen_op_iwmmxt_movq_M0_wRn(wrd);
1846 switch ((insn >> 6) & 3) {
1847 case 0:
1848 gen_op_movl_T1_im(0xff);
1849 gen_op_iwmmxt_insr_M0_T0_T1((insn & 7) << 3);
1850 break;
1851 case 1:
1852 gen_op_movl_T1_im(0xffff);
1853 gen_op_iwmmxt_insr_M0_T0_T1((insn & 3) << 4);
1854 break;
1855 case 2:
1856 gen_op_movl_T1_im(0xffffffff);
1857 gen_op_iwmmxt_insr_M0_T0_T1((insn & 1) << 5);
1858 break;
1859 case 3:
1860 return 1;
1861 }
1862 gen_op_iwmmxt_movq_wRn_M0(wrd);
1863 gen_op_iwmmxt_set_mup();
1864 break;
1865 case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */
1866 rd = (insn >> 12) & 0xf;
1867 wrd = (insn >> 16) & 0xf;
1868 if (rd == 15)
1869 return 1;
1870 gen_op_iwmmxt_movq_M0_wRn(wrd);
1871 switch ((insn >> 22) & 3) {
1872 case 0:
1873 if (insn & 8)
1874 gen_op_iwmmxt_extrsb_T0_M0((insn & 7) << 3);
1875 else {
1876 gen_op_iwmmxt_extru_T0_M0((insn & 7) << 3, 0xff);
1877 }
1878 break;
1879 case 1:
1880 if (insn & 8)
1881 gen_op_iwmmxt_extrsw_T0_M0((insn & 3) << 4);
1882 else {
1883 gen_op_iwmmxt_extru_T0_M0((insn & 3) << 4, 0xffff);
1884 }
1885 break;
1886 case 2:
1887 gen_op_iwmmxt_extru_T0_M0((insn & 1) << 5, ~0u);
1888 break;
1889 case 3:
1890 return 1;
1891 }
1892 gen_movl_reg_T0(s, rd);
1893 break;
1894 case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */
1895 if ((insn & 0x000ff008) != 0x0003f000)
1896 return 1;
1897 gen_op_iwmmxt_movl_T1_wCx(ARM_IWMMXT_wCASF);
1898 switch ((insn >> 22) & 3) {
1899 case 0:
1900 gen_op_shrl_T1_im(((insn & 7) << 2) + 0);
1901 break;
1902 case 1:
1903 gen_op_shrl_T1_im(((insn & 3) << 3) + 4);
1904 break;
1905 case 2:
1906 gen_op_shrl_T1_im(((insn & 1) << 4) + 12);
1907 break;
1908 case 3:
1909 return 1;
1910 }
1911 gen_op_shll_T1_im(28);
1912 gen_set_nzcv(cpu_T[1]);
1913 break;
1914 case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */
1915 rd = (insn >> 12) & 0xf;
1916 wrd = (insn >> 16) & 0xf;
1917 gen_movl_T0_reg(s, rd);
1918 switch ((insn >> 6) & 3) {
1919 case 0:
1920 gen_helper_iwmmxt_bcstb(cpu_M0, cpu_T[0]);
1921 break;
1922 case 1:
1923 gen_helper_iwmmxt_bcstw(cpu_M0, cpu_T[0]);
1924 break;
1925 case 2:
1926 gen_helper_iwmmxt_bcstl(cpu_M0, cpu_T[0]);
1927 break;
1928 case 3:
1929 return 1;
1930 }
1931 gen_op_iwmmxt_movq_wRn_M0(wrd);
1932 gen_op_iwmmxt_set_mup();
1933 break;
1934 case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */
1935 if ((insn & 0x000ff00f) != 0x0003f000)
1936 return 1;
1937 gen_op_iwmmxt_movl_T1_wCx(ARM_IWMMXT_wCASF);
1938 switch ((insn >> 22) & 3) {
1939 case 0:
1940 for (i = 0; i < 7; i ++) {
1941 gen_op_shll_T1_im(4);
1942 gen_op_andl_T0_T1();
1943 }
1944 break;
1945 case 1:
1946 for (i = 0; i < 3; i ++) {
1947 gen_op_shll_T1_im(8);
1948 gen_op_andl_T0_T1();
1949 }
1950 break;
1951 case 2:
1952 gen_op_shll_T1_im(16);
1953 gen_op_andl_T0_T1();
1954 break;
1955 case 3:
1956 return 1;
1957 }
1958 gen_set_nzcv(cpu_T[0]);
1959 break;
1960 case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */
1961 wrd = (insn >> 12) & 0xf;
1962 rd0 = (insn >> 16) & 0xf;
1963 gen_op_iwmmxt_movq_M0_wRn(rd0);
1964 switch ((insn >> 22) & 3) {
1965 case 0:
1966 gen_helper_iwmmxt_addcb(cpu_M0, cpu_M0);
1967 break;
1968 case 1:
1969 gen_helper_iwmmxt_addcw(cpu_M0, cpu_M0);
1970 break;
1971 case 2:
1972 gen_helper_iwmmxt_addcl(cpu_M0, cpu_M0);
1973 break;
1974 case 3:
1975 return 1;
1976 }
1977 gen_op_iwmmxt_movq_wRn_M0(wrd);
1978 gen_op_iwmmxt_set_mup();
1979 break;
1980 case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */
1981 if ((insn & 0x000ff00f) != 0x0003f000)
1982 return 1;
1983 gen_op_iwmmxt_movl_T1_wCx(ARM_IWMMXT_wCASF);
1984 switch ((insn >> 22) & 3) {
1985 case 0:
1986 for (i = 0; i < 7; i ++) {
1987 gen_op_shll_T1_im(4);
1988 gen_op_orl_T0_T1();
1989 }
1990 break;
1991 case 1:
1992 for (i = 0; i < 3; i ++) {
1993 gen_op_shll_T1_im(8);
1994 gen_op_orl_T0_T1();
1995 }
1996 break;
1997 case 2:
1998 gen_op_shll_T1_im(16);
1999 gen_op_orl_T0_T1();
2000 break;
2001 case 3:
2002 return 1;
2003 }
2004 gen_set_nzcv(cpu_T[0]);
2005 break;
2006 case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */
2007 rd = (insn >> 12) & 0xf;
2008 rd0 = (insn >> 16) & 0xf;
2009 if ((insn & 0xf) != 0)
2010 return 1;
2011 gen_op_iwmmxt_movq_M0_wRn(rd0);
2012 switch ((insn >> 22) & 3) {
2013 case 0:
2014 gen_helper_iwmmxt_msbb(cpu_T[0], cpu_M0);
2015 break;
2016 case 1:
2017 gen_helper_iwmmxt_msbw(cpu_T[0], cpu_M0);
2018 break;
2019 case 2:
2020 gen_helper_iwmmxt_msbl(cpu_T[0], cpu_M0);
2021 break;
2022 case 3:
2023 return 1;
2024 }
2025 gen_movl_reg_T0(s, rd);
2026 break;
2027 case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */
2028 case 0x906: case 0xb06: case 0xd06: case 0xf06:
2029 wrd = (insn >> 12) & 0xf;
2030 rd0 = (insn >> 16) & 0xf;
2031 rd1 = (insn >> 0) & 0xf;
2032 gen_op_iwmmxt_movq_M0_wRn(rd0);
2033 switch ((insn >> 22) & 3) {
2034 case 0:
2035 if (insn & (1 << 21))
2036 gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1);
2037 else
2038 gen_op_iwmmxt_cmpgtub_M0_wRn(rd1);
2039 break;
2040 case 1:
2041 if (insn & (1 << 21))
2042 gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1);
2043 else
2044 gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1);
2045 break;
2046 case 2:
2047 if (insn & (1 << 21))
2048 gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1);
2049 else
2050 gen_op_iwmmxt_cmpgtul_M0_wRn(rd1);
2051 break;
2052 case 3:
2053 return 1;
2054 }
2055 gen_op_iwmmxt_movq_wRn_M0(wrd);
2056 gen_op_iwmmxt_set_mup();
2057 gen_op_iwmmxt_set_cup();
2058 break;
2059 case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */
2060 case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e:
2061 wrd = (insn >> 12) & 0xf;
2062 rd0 = (insn >> 16) & 0xf;
2063 gen_op_iwmmxt_movq_M0_wRn(rd0);
2064 switch ((insn >> 22) & 3) {
2065 case 0:
2066 if (insn & (1 << 21))
2067 gen_op_iwmmxt_unpacklsb_M0();
2068 else
2069 gen_op_iwmmxt_unpacklub_M0();
2070 break;
2071 case 1:
2072 if (insn & (1 << 21))
2073 gen_op_iwmmxt_unpacklsw_M0();
2074 else
2075 gen_op_iwmmxt_unpackluw_M0();
2076 break;
2077 case 2:
2078 if (insn & (1 << 21))
2079 gen_op_iwmmxt_unpacklsl_M0();
2080 else
2081 gen_op_iwmmxt_unpacklul_M0();
2082 break;
2083 case 3:
2084 return 1;
2085 }
2086 gen_op_iwmmxt_movq_wRn_M0(wrd);
2087 gen_op_iwmmxt_set_mup();
2088 gen_op_iwmmxt_set_cup();
2089 break;
2090 case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */
2091 case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c:
2092 wrd = (insn >> 12) & 0xf;
2093 rd0 = (insn >> 16) & 0xf;
2094 gen_op_iwmmxt_movq_M0_wRn(rd0);
2095 switch ((insn >> 22) & 3) {
2096 case 0:
2097 if (insn & (1 << 21))
2098 gen_op_iwmmxt_unpackhsb_M0();
2099 else
2100 gen_op_iwmmxt_unpackhub_M0();
2101 break;
2102 case 1:
2103 if (insn & (1 << 21))
2104 gen_op_iwmmxt_unpackhsw_M0();
2105 else
2106 gen_op_iwmmxt_unpackhuw_M0();
2107 break;
2108 case 2:
2109 if (insn & (1 << 21))
2110 gen_op_iwmmxt_unpackhsl_M0();
2111 else
2112 gen_op_iwmmxt_unpackhul_M0();
2113 break;
2114 case 3:
2115 return 1;
2116 }
2117 gen_op_iwmmxt_movq_wRn_M0(wrd);
2118 gen_op_iwmmxt_set_mup();
2119 gen_op_iwmmxt_set_cup();
2120 break;
2121 case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */
2122 case 0x214: case 0x614: case 0xa14: case 0xe14:
2123 wrd = (insn >> 12) & 0xf;
2124 rd0 = (insn >> 16) & 0xf;
2125 gen_op_iwmmxt_movq_M0_wRn(rd0);
2126 if (gen_iwmmxt_shift(insn, 0xff))
2127 return 1;
2128 switch ((insn >> 22) & 3) {
2129 case 0:
2130 return 1;
2131 case 1:
2132 gen_helper_iwmmxt_srlw(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2133 break;
2134 case 2:
2135 gen_helper_iwmmxt_srll(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2136 break;
2137 case 3:
2138 gen_helper_iwmmxt_srlq(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2139 break;
2140 }
2141 gen_op_iwmmxt_movq_wRn_M0(wrd);
2142 gen_op_iwmmxt_set_mup();
2143 gen_op_iwmmxt_set_cup();
2144 break;
2145 case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */
2146 case 0x014: case 0x414: case 0x814: case 0xc14:
2147 wrd = (insn >> 12) & 0xf;
2148 rd0 = (insn >> 16) & 0xf;
2149 gen_op_iwmmxt_movq_M0_wRn(rd0);
2150 if (gen_iwmmxt_shift(insn, 0xff))
2151 return 1;
2152 switch ((insn >> 22) & 3) {
2153 case 0:
2154 return 1;
2155 case 1:
2156 gen_helper_iwmmxt_sraw(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2157 break;
2158 case 2:
2159 gen_helper_iwmmxt_sral(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2160 break;
2161 case 3:
2162 gen_helper_iwmmxt_sraq(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2163 break;
2164 }
2165 gen_op_iwmmxt_movq_wRn_M0(wrd);
2166 gen_op_iwmmxt_set_mup();
2167 gen_op_iwmmxt_set_cup();
2168 break;
2169 case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */
2170 case 0x114: case 0x514: case 0x914: case 0xd14:
2171 wrd = (insn >> 12) & 0xf;
2172 rd0 = (insn >> 16) & 0xf;
2173 gen_op_iwmmxt_movq_M0_wRn(rd0);
2174 if (gen_iwmmxt_shift(insn, 0xff))
2175 return 1;
2176 switch ((insn >> 22) & 3) {
2177 case 0:
2178 return 1;
2179 case 1:
2180 gen_helper_iwmmxt_sllw(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2181 break;
2182 case 2:
2183 gen_helper_iwmmxt_slll(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2184 break;
2185 case 3:
2186 gen_helper_iwmmxt_sllq(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2187 break;
2188 }
2189 gen_op_iwmmxt_movq_wRn_M0(wrd);
2190 gen_op_iwmmxt_set_mup();
2191 gen_op_iwmmxt_set_cup();
2192 break;
2193 case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */
2194 case 0x314: case 0x714: case 0xb14: case 0xf14:
2195 wrd = (insn >> 12) & 0xf;
2196 rd0 = (insn >> 16) & 0xf;
2197 gen_op_iwmmxt_movq_M0_wRn(rd0);
2198 switch ((insn >> 22) & 3) {
2199 case 0:
2200 return 1;
2201 case 1:
2202 if (gen_iwmmxt_shift(insn, 0xf))
2203 return 1;
2204 gen_helper_iwmmxt_rorw(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2205 break;
2206 case 2:
2207 if (gen_iwmmxt_shift(insn, 0x1f))
2208 return 1;
2209 gen_helper_iwmmxt_rorl(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2210 break;
2211 case 3:
2212 if (gen_iwmmxt_shift(insn, 0x3f))
2213 return 1;
2214 gen_helper_iwmmxt_rorq(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2215 break;
2216 }
2217 gen_op_iwmmxt_movq_wRn_M0(wrd);
2218 gen_op_iwmmxt_set_mup();
2219 gen_op_iwmmxt_set_cup();
2220 break;
2221 case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */
2222 case 0x916: case 0xb16: case 0xd16: case 0xf16:
2223 wrd = (insn >> 12) & 0xf;
2224 rd0 = (insn >> 16) & 0xf;
2225 rd1 = (insn >> 0) & 0xf;
2226 gen_op_iwmmxt_movq_M0_wRn(rd0);
2227 switch ((insn >> 22) & 3) {
2228 case 0:
2229 if (insn & (1 << 21))
2230 gen_op_iwmmxt_minsb_M0_wRn(rd1);
2231 else
2232 gen_op_iwmmxt_minub_M0_wRn(rd1);
2233 break;
2234 case 1:
2235 if (insn & (1 << 21))
2236 gen_op_iwmmxt_minsw_M0_wRn(rd1);
2237 else
2238 gen_op_iwmmxt_minuw_M0_wRn(rd1);
2239 break;
2240 case 2:
2241 if (insn & (1 << 21))
2242 gen_op_iwmmxt_minsl_M0_wRn(rd1);
2243 else
2244 gen_op_iwmmxt_minul_M0_wRn(rd1);
2245 break;
2246 case 3:
2247 return 1;
2248 }
2249 gen_op_iwmmxt_movq_wRn_M0(wrd);
2250 gen_op_iwmmxt_set_mup();
2251 break;
2252 case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */
2253 case 0x816: case 0xa16: case 0xc16: case 0xe16:
2254 wrd = (insn >> 12) & 0xf;
2255 rd0 = (insn >> 16) & 0xf;
2256 rd1 = (insn >> 0) & 0xf;
2257 gen_op_iwmmxt_movq_M0_wRn(rd0);
2258 switch ((insn >> 22) & 3) {
2259 case 0:
2260 if (insn & (1 << 21))
2261 gen_op_iwmmxt_maxsb_M0_wRn(rd1);
2262 else
2263 gen_op_iwmmxt_maxub_M0_wRn(rd1);
2264 break;
2265 case 1:
2266 if (insn & (1 << 21))
2267 gen_op_iwmmxt_maxsw_M0_wRn(rd1);
2268 else
2269 gen_op_iwmmxt_maxuw_M0_wRn(rd1);
2270 break;
2271 case 2:
2272 if (insn & (1 << 21))
2273 gen_op_iwmmxt_maxsl_M0_wRn(rd1);
2274 else
2275 gen_op_iwmmxt_maxul_M0_wRn(rd1);
2276 break;
2277 case 3:
2278 return 1;
2279 }
2280 gen_op_iwmmxt_movq_wRn_M0(wrd);
2281 gen_op_iwmmxt_set_mup();
2282 break;
2283 case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */
2284 case 0x402: case 0x502: case 0x602: case 0x702:
2285 wrd = (insn >> 12) & 0xf;
2286 rd0 = (insn >> 16) & 0xf;
2287 rd1 = (insn >> 0) & 0xf;
2288 gen_op_iwmmxt_movq_M0_wRn(rd0);
2289 gen_op_movl_T0_im((insn >> 20) & 3);
2290 gen_op_iwmmxt_align_M0_T0_wRn(rd1);
2291 gen_op_iwmmxt_movq_wRn_M0(wrd);
2292 gen_op_iwmmxt_set_mup();
2293 break;
2294 case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */
2295 case 0x41a: case 0x51a: case 0x61a: case 0x71a:
2296 case 0x81a: case 0x91a: case 0xa1a: case 0xb1a:
2297 case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a:
2298 wrd = (insn >> 12) & 0xf;
2299 rd0 = (insn >> 16) & 0xf;
2300 rd1 = (insn >> 0) & 0xf;
2301 gen_op_iwmmxt_movq_M0_wRn(rd0);
2302 switch ((insn >> 20) & 0xf) {
2303 case 0x0:
2304 gen_op_iwmmxt_subnb_M0_wRn(rd1);
2305 break;
2306 case 0x1:
2307 gen_op_iwmmxt_subub_M0_wRn(rd1);
2308 break;
2309 case 0x3:
2310 gen_op_iwmmxt_subsb_M0_wRn(rd1);
2311 break;
2312 case 0x4:
2313 gen_op_iwmmxt_subnw_M0_wRn(rd1);
2314 break;
2315 case 0x5:
2316 gen_op_iwmmxt_subuw_M0_wRn(rd1);
2317 break;
2318 case 0x7:
2319 gen_op_iwmmxt_subsw_M0_wRn(rd1);
2320 break;
2321 case 0x8:
2322 gen_op_iwmmxt_subnl_M0_wRn(rd1);
2323 break;
2324 case 0x9:
2325 gen_op_iwmmxt_subul_M0_wRn(rd1);
2326 break;
2327 case 0xb:
2328 gen_op_iwmmxt_subsl_M0_wRn(rd1);
2329 break;
2330 default:
2331 return 1;
2332 }
2333 gen_op_iwmmxt_movq_wRn_M0(wrd);
2334 gen_op_iwmmxt_set_mup();
2335 gen_op_iwmmxt_set_cup();
2336 break;
2337 case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */
2338 case 0x41e: case 0x51e: case 0x61e: case 0x71e:
2339 case 0x81e: case 0x91e: case 0xa1e: case 0xb1e:
2340 case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e:
2341 wrd = (insn >> 12) & 0xf;
2342 rd0 = (insn >> 16) & 0xf;
2343 gen_op_iwmmxt_movq_M0_wRn(rd0);
2344 gen_op_movl_T0_im(((insn >> 16) & 0xf0) | (insn & 0x0f));
2345 gen_helper_iwmmxt_shufh(cpu_M0, cpu_env, cpu_M0, cpu_T[0]);
2346 gen_op_iwmmxt_movq_wRn_M0(wrd);
2347 gen_op_iwmmxt_set_mup();
2348 gen_op_iwmmxt_set_cup();
2349 break;
2350 case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */
2351 case 0x418: case 0x518: case 0x618: case 0x718:
2352 case 0x818: case 0x918: case 0xa18: case 0xb18:
2353 case 0xc18: case 0xd18: case 0xe18: case 0xf18:
2354 wrd = (insn >> 12) & 0xf;
2355 rd0 = (insn >> 16) & 0xf;
2356 rd1 = (insn >> 0) & 0xf;
2357 gen_op_iwmmxt_movq_M0_wRn(rd0);
2358 switch ((insn >> 20) & 0xf) {
2359 case 0x0:
2360 gen_op_iwmmxt_addnb_M0_wRn(rd1);
2361 break;
2362 case 0x1:
2363 gen_op_iwmmxt_addub_M0_wRn(rd1);
2364 break;
2365 case 0x3:
2366 gen_op_iwmmxt_addsb_M0_wRn(rd1);
2367 break;
2368 case 0x4:
2369 gen_op_iwmmxt_addnw_M0_wRn(rd1);
2370 break;
2371 case 0x5:
2372 gen_op_iwmmxt_adduw_M0_wRn(rd1);
2373 break;
2374 case 0x7:
2375 gen_op_iwmmxt_addsw_M0_wRn(rd1);
2376 break;
2377 case 0x8:
2378 gen_op_iwmmxt_addnl_M0_wRn(rd1);
2379 break;
2380 case 0x9:
2381 gen_op_iwmmxt_addul_M0_wRn(rd1);
2382 break;
2383 case 0xb:
2384 gen_op_iwmmxt_addsl_M0_wRn(rd1);
2385 break;
2386 default:
2387 return 1;
2388 }
2389 gen_op_iwmmxt_movq_wRn_M0(wrd);
2390 gen_op_iwmmxt_set_mup();
2391 gen_op_iwmmxt_set_cup();
2392 break;
2393 case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */
2394 case 0x408: case 0x508: case 0x608: case 0x708:
2395 case 0x808: case 0x908: case 0xa08: case 0xb08:
2396 case 0xc08: case 0xd08: case 0xe08: case 0xf08:
2397 wrd = (insn >> 12) & 0xf;
2398 rd0 = (insn >> 16) & 0xf;
2399 rd1 = (insn >> 0) & 0xf;
2400 gen_op_iwmmxt_movq_M0_wRn(rd0);
2401 if (!(insn & (1 << 20)))
2402 return 1;
2403 switch ((insn >> 22) & 3) {
2404 case 0:
2405 return 1;
2406 case 1:
2407 if (insn & (1 << 21))
2408 gen_op_iwmmxt_packsw_M0_wRn(rd1);
2409 else
2410 gen_op_iwmmxt_packuw_M0_wRn(rd1);
2411 break;
2412 case 2:
2413 if (insn & (1 << 21))
2414 gen_op_iwmmxt_packsl_M0_wRn(rd1);
2415 else
2416 gen_op_iwmmxt_packul_M0_wRn(rd1);
2417 break;
2418 case 3:
2419 if (insn & (1 << 21))
2420 gen_op_iwmmxt_packsq_M0_wRn(rd1);
2421 else
2422 gen_op_iwmmxt_packuq_M0_wRn(rd1);
2423 break;
2424 }
2425 gen_op_iwmmxt_movq_wRn_M0(wrd);
2426 gen_op_iwmmxt_set_mup();
2427 gen_op_iwmmxt_set_cup();
2428 break;
2429 case 0x201: case 0x203: case 0x205: case 0x207:
2430 case 0x209: case 0x20b: case 0x20d: case 0x20f:
2431 case 0x211: case 0x213: case 0x215: case 0x217:
2432 case 0x219: case 0x21b: case 0x21d: case 0x21f:
2433 wrd = (insn >> 5) & 0xf;
2434 rd0 = (insn >> 12) & 0xf;
2435 rd1 = (insn >> 0) & 0xf;
2436 if (rd0 == 0xf || rd1 == 0xf)
2437 return 1;
2438 gen_op_iwmmxt_movq_M0_wRn(wrd);
2439 switch ((insn >> 16) & 0xf) {
2440 case 0x0: /* TMIA */
2441 gen_movl_T0_reg(s, rd0);
2442 gen_movl_T1_reg(s, rd1);
2443 gen_op_iwmmxt_muladdsl_M0_T0_T1();
2444 break;
2445 case 0x8: /* TMIAPH */
2446 gen_movl_T0_reg(s, rd0);
2447 gen_movl_T1_reg(s, rd1);
2448 gen_op_iwmmxt_muladdsw_M0_T0_T1();
2449 break;
2450 case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */
2451 gen_movl_T1_reg(s, rd0);
2452 if (insn & (1 << 16))
2453 gen_op_shrl_T1_im(16);
2454 gen_op_movl_T0_T1();
2455 gen_movl_T1_reg(s, rd1);
2456 if (insn & (1 << 17))
2457 gen_op_shrl_T1_im(16);
2458 gen_op_iwmmxt_muladdswl_M0_T0_T1();
2459 break;
2460 default:
2461 return 1;
2462 }
2463 gen_op_iwmmxt_movq_wRn_M0(wrd);
2464 gen_op_iwmmxt_set_mup();
2465 break;
2466 default:
2467 return 1;
2468 }
2469
2470 return 0;
2471 }
2472
2473 /* Disassemble an XScale DSP instruction. Returns nonzero if an error occured
2474 (ie. an undefined instruction). */
2475 static int disas_dsp_insn(CPUState *env, DisasContext *s, uint32_t insn)
2476 {
2477 int acc, rd0, rd1, rdhi, rdlo;
2478
2479 if ((insn & 0x0ff00f10) == 0x0e200010) {
2480 /* Multiply with Internal Accumulate Format */
2481 rd0 = (insn >> 12) & 0xf;
2482 rd1 = insn & 0xf;
2483 acc = (insn >> 5) & 7;
2484
2485 if (acc != 0)
2486 return 1;
2487
2488 switch ((insn >> 16) & 0xf) {
2489 case 0x0: /* MIA */
2490 gen_movl_T0_reg(s, rd0);
2491 gen_movl_T1_reg(s, rd1);
2492 gen_op_iwmmxt_muladdsl_M0_T0_T1();
2493 break;
2494 case 0x8: /* MIAPH */
2495 gen_movl_T0_reg(s, rd0);
2496 gen_movl_T1_reg(s, rd1);
2497 gen_op_iwmmxt_muladdsw_M0_T0_T1();
2498 break;
2499 case 0xc: /* MIABB */
2500 case 0xd: /* MIABT */
2501 case 0xe: /* MIATB */
2502 case 0xf: /* MIATT */
2503 gen_movl_T1_reg(s, rd0);
2504 if (insn & (1 << 16))
2505 gen_op_shrl_T1_im(16);
2506 gen_op_movl_T0_T1();
2507 gen_movl_T1_reg(s, rd1);
2508 if (insn & (1 << 17))
2509 gen_op_shrl_T1_im(16);
2510 gen_op_iwmmxt_muladdswl_M0_T0_T1();
2511 break;
2512 default:
2513 return 1;
2514 }
2515
2516 gen_op_iwmmxt_movq_wRn_M0(acc);
2517 return 0;
2518 }
2519
2520 if ((insn & 0x0fe00ff8) == 0x0c400000) {
2521 /* Internal Accumulator Access Format */
2522 rdhi = (insn >> 16) & 0xf;
2523 rdlo = (insn >> 12) & 0xf;
2524 acc = insn & 7;
2525
2526 if (acc != 0)
2527 return 1;
2528
2529 if (insn & ARM_CP_RW_BIT) { /* MRA */
2530 gen_iwmmxt_movl_T0_T1_wRn(acc);
2531 gen_movl_reg_T0(s, rdlo);
2532 gen_op_movl_T0_im((1 << (40 - 32)) - 1);
2533 gen_op_andl_T0_T1();
2534 gen_movl_reg_T0(s, rdhi);
2535 } else { /* MAR */
2536 gen_movl_T0_reg(s, rdlo);
2537 gen_movl_T1_reg(s, rdhi);
2538 gen_iwmmxt_movl_wRn_T0_T1(acc);
2539 }
2540 return 0;
2541 }
2542
2543 return 1;
2544 }
2545
2546 /* Disassemble system coprocessor instruction. Return nonzero if
2547 instruction is not defined. */
2548 static int disas_cp_insn(CPUState *env, DisasContext *s, uint32_t insn)
2549 {
2550 TCGv tmp;
2551 uint32_t rd = (insn >> 12) & 0xf;
2552 uint32_t cp = (insn >> 8) & 0xf;
2553 if (IS_USER(s)) {
2554 return 1;
2555 }
2556
2557 if (insn & ARM_CP_RW_BIT) {
2558 if (!env->cp[cp].cp_read)
2559 return 1;
2560 gen_set_pc_im(s->pc);
2561 tmp = new_tmp();
2562 gen_helper_get_cp(tmp, cpu_env, tcg_const_i32(insn));
2563 store_reg(s, rd, tmp);
2564 } else {
2565 if (!env->cp[cp].cp_write)
2566 return 1;
2567 gen_set_pc_im(s->pc);
2568 tmp = load_reg(s, rd);
2569 gen_helper_set_cp(cpu_env, tcg_const_i32(insn), tmp);
2570 dead_tmp(tmp);
2571 }
2572 return 0;
2573 }
2574
2575 static int cp15_user_ok(uint32_t insn)
2576 {
2577 int cpn = (insn >> 16) & 0xf;
2578 int cpm = insn & 0xf;
2579 int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38);
2580
2581 if (cpn == 13 && cpm == 0) {
2582 /* TLS register. */
2583 if (op == 2 || (op == 3 && (insn & ARM_CP_RW_BIT)))
2584 return 1;
2585 }
2586 if (cpn == 7) {
2587 /* ISB, DSB, DMB. */
2588 if ((cpm == 5 && op == 4)
2589 || (cpm == 10 && (op == 4 || op == 5)))
2590 return 1;
2591 }
2592 return 0;
2593 }
2594
2595 /* Disassemble system coprocessor (cp15) instruction. Return nonzero if
2596 instruction is not defined. */
2597 static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn)
2598 {
2599 uint32_t rd;
2600 TCGv tmp;
2601
2602 /* M profile cores use memory mapped registers instead of cp15. */
2603 if (arm_feature(env, ARM_FEATURE_M))
2604 return 1;
2605
2606 if ((insn & (1 << 25)) == 0) {
2607 if (insn & (1 << 20)) {
2608 /* mrrc */
2609 return 1;
2610 }
2611 /* mcrr. Used for block cache operations, so implement as no-op. */
2612 return 0;
2613 }
2614 if ((insn & (1 << 4)) == 0) {
2615 /* cdp */
2616 return 1;
2617 }
2618 if (IS_USER(s) && !cp15_user_ok(insn)) {
2619 return 1;
2620 }
2621 if ((insn & 0x0fff0fff) == 0x0e070f90
2622 || (insn & 0x0fff0fff) == 0x0e070f58) {
2623 /* Wait for interrupt. */
2624 gen_set_pc_im(s->pc);
2625 s->is_jmp = DISAS_WFI;
2626 return 0;
2627 }
2628 rd = (insn >> 12) & 0xf;
2629 if (insn & ARM_CP_RW_BIT) {
2630 tmp = new_tmp();
2631 gen_helper_get_cp15(tmp, cpu_env, tcg_const_i32(insn));
2632 /* If the destination register is r15 then sets condition codes. */
2633 if (rd != 15)
2634 store_reg(s, rd, tmp);
2635 else
2636 dead_tmp(tmp);
2637 } else {
2638 tmp = load_reg(s, rd);
2639 gen_helper_set_cp15(cpu_env, tcg_const_i32(insn), tmp);
2640 dead_tmp(tmp);
2641 /* Normally we would always end the TB here, but Linux
2642 * arch/arm/mach-pxa/sleep.S expects two instructions following
2643 * an MMU enable to execute from cache. Imitate this behaviour. */
2644 if (!arm_feature(env, ARM_FEATURE_XSCALE) ||
2645 (insn & 0x0fff0fff) != 0x0e010f10)
2646 gen_lookup_tb(s);
2647 }
2648 return 0;
2649 }
2650
2651 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
2652 #define VFP_SREG(insn, bigbit, smallbit) \
2653 ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
2654 #define VFP_DREG(reg, insn, bigbit, smallbit) do { \
2655 if (arm_feature(env, ARM_FEATURE_VFP3)) { \
2656 reg = (((insn) >> (bigbit)) & 0x0f) \
2657 | (((insn) >> ((smallbit) - 4)) & 0x10); \
2658 } else { \
2659 if (insn & (1 << (smallbit))) \
2660 return 1; \
2661 reg = ((insn) >> (bigbit)) & 0x0f; \
2662 }} while (0)
2663
2664 #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
2665 #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
2666 #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7)
2667 #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7)
2668 #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5)
2669 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5)
2670
2671 /* Move between integer and VFP cores. */
2672 static TCGv gen_vfp_mrs(void)
2673 {
2674 TCGv tmp = new_tmp();
2675 tcg_gen_mov_i32(tmp, cpu_F0s);
2676 return tmp;
2677 }
2678
2679 static void gen_vfp_msr(TCGv tmp)
2680 {
2681 tcg_gen_mov_i32(cpu_F0s, tmp);
2682 dead_tmp(tmp);
2683 }
2684
2685 static inline int
2686 vfp_enabled(CPUState * env)
2687 {
2688 return ((env->vfp.xregs[ARM_VFP_FPEXC] & (1 << 30)) != 0);
2689 }
2690
2691 static void gen_neon_dup_u8(TCGv var, int shift)
2692 {
2693 TCGv tmp = new_tmp();
2694 if (shift)
2695 tcg_gen_shri_i32(var, var, shift);
2696 tcg_gen_ext8u_i32(var, var);
2697 tcg_gen_shli_i32(tmp, var, 8);
2698 tcg_gen_or_i32(var, var, tmp);
2699 tcg_gen_shli_i32(tmp, var, 16);
2700 tcg_gen_or_i32(var, var, tmp);
2701 dead_tmp(tmp);
2702 }
2703
2704 static void gen_neon_dup_low16(TCGv var)
2705 {
2706 TCGv tmp = new_tmp();
2707 tcg_gen_ext16u_i32(var, var);
2708 tcg_gen_shli_i32(tmp, var, 16);
2709 tcg_gen_or_i32(var, var, tmp);
2710 dead_tmp(tmp);
2711 }
2712
2713 static void gen_neon_dup_high16(TCGv var)
2714 {
2715 TCGv tmp = new_tmp();
2716 tcg_gen_andi_i32(var, var, 0xffff0000);
2717 tcg_gen_shri_i32(tmp, var, 16);
2718 tcg_gen_or_i32(var, var, tmp);
2719 dead_tmp(tmp);
2720 }
2721
2722 /* Disassemble a VFP instruction. Returns nonzero if an error occured
2723 (ie. an undefined instruction). */
2724 static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
2725 {
2726 uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask;
2727 int dp, veclen;
2728 TCGv tmp;
2729 TCGv tmp2;
2730
2731 if (!arm_feature(env, ARM_FEATURE_VFP))
2732 return 1;
2733
2734 if (!vfp_enabled(env)) {
2735 /* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
2736 if ((insn & 0x0fe00fff) != 0x0ee00a10)
2737 return 1;
2738 rn = (insn >> 16) & 0xf;
2739 if (rn != ARM_VFP_FPSID && rn != ARM_VFP_FPEXC
2740 && rn != ARM_VFP_MVFR1 && rn != ARM_VFP_MVFR0)
2741 return 1;
2742 }
2743 dp = ((insn & 0xf00) == 0xb00);
2744 switch ((insn >> 24) & 0xf) {
2745 case 0xe:
2746 if (insn & (1 << 4)) {
2747 /* single register transfer */
2748 rd = (insn >> 12) & 0xf;
2749 if (dp) {
2750 int size;
2751 int pass;
2752
2753 VFP_DREG_N(rn, insn);
2754 if (insn & 0xf)
2755 return 1;
2756 if (insn & 0x00c00060
2757 && !arm_feature(env, ARM_FEATURE_NEON))
2758 return 1;
2759
2760 pass = (insn >> 21) & 1;
2761 if (insn & (1 << 22)) {
2762 size = 0;
2763 offset = ((insn >> 5) & 3) * 8;
2764 } else if (insn & (1 << 5)) {
2765 size = 1;
2766 offset = (insn & (1 << 6)) ? 16 : 0;
2767 } else {
2768 size = 2;
2769 offset = 0;
2770 }
2771 if (insn & ARM_CP_RW_BIT) {
2772 /* vfp->arm */
2773 tmp = neon_load_reg(rn, pass);
2774 switch (size) {
2775 case 0:
2776 if (offset)
2777 tcg_gen_shri_i32(tmp, tmp, offset);
2778 if (insn & (1 << 23))
2779 gen_uxtb(tmp);
2780 else
2781 gen_sxtb(tmp);
2782 break;
2783 case 1:
2784 if (insn & (1 << 23)) {
2785 if (offset) {
2786 tcg_gen_shri_i32(tmp, tmp, 16);
2787 } else {
2788 gen_uxth(tmp);
2789 }
2790 } else {
2791 if (offset) {
2792 tcg_gen_sari_i32(tmp, tmp, 16);
2793 } else {
2794 gen_sxth(tmp);
2795 }
2796 }
2797 break;
2798 case 2:
2799 break;
2800 }
2801 store_reg(s, rd, tmp);
2802 } else {
2803 /* arm->vfp */
2804 tmp = load_reg(s, rd);
2805 if (insn & (1 << 23)) {
2806 /* VDUP */
2807 if (size == 0) {
2808 gen_neon_dup_u8(tmp, 0);
2809 } else if (size == 1) {
2810 gen_neon_dup_low16(tmp);
2811 }
2812 tmp2 = new_tmp();
2813 tcg_gen_mov_i32(tmp2, tmp);
2814 neon_store_reg(rn, 0, tmp2);
2815 neon_store_reg(rn, 1, tmp);
2816 } else {
2817 /* VMOV */
2818 switch (size) {
2819 case 0:
2820 tmp2 = neon_load_reg(rn, pass);
2821 gen_bfi(tmp, tmp2, tmp, offset, 0xff);
2822 dead_tmp(tmp2);
2823 break;
2824 case 1:
2825 tmp2 = neon_load_reg(rn, pass);
2826 gen_bfi(tmp, tmp2, tmp, offset, 0xffff);
2827 dead_tmp(tmp2);
2828 break;
2829 case 2:
2830 break;
2831 }
2832 neon_store_reg(rn, pass, tmp);
2833 }
2834 }
2835 } else { /* !dp */
2836 if ((insn & 0x6f) != 0x00)
2837 return 1;
2838 rn = VFP_SREG_N(insn);
2839 if (insn & ARM_CP_RW_BIT) {
2840 /* vfp->arm */
2841 if (insn & (1 << 21)) {
2842 /* system register */
2843 rn >>= 1;
2844
2845 switch (rn) {
2846 case ARM_VFP_FPSID:
2847 /* VFP2 allows access to FSID from userspace.
2848 VFP3 restricts all id registers to privileged
2849 accesses. */
2850 if (IS_USER(s)
2851 && arm_feature(env, ARM_FEATURE_VFP3))
2852 return 1;
2853 tmp = load_cpu_field(vfp.xregs[rn]);
2854 break;
2855 case ARM_VFP_FPEXC:
2856 if (IS_USER(s))
2857 return 1;
2858 tmp = load_cpu_field(vfp.xregs[rn]);
2859 break;
2860 case ARM_VFP_FPINST:
2861 case ARM_VFP_FPINST2:
2862 /* Not present in VFP3. */
2863 if (IS_USER(s)
2864 || arm_feature(env, ARM_FEATURE_VFP3))
2865 return 1;
2866 tmp = load_cpu_field(vfp.xregs[rn]);
2867 break;
2868 case ARM_VFP_FPSCR:
2869 if (rd == 15) {
2870 tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]);
2871 tcg_gen_andi_i32(tmp, tmp, 0xf0000000);
2872 } else {
2873 tmp = new_tmp();
2874 gen_helper_vfp_get_fpscr(tmp, cpu_env);
2875 }
2876 break;
2877 case ARM_VFP_MVFR0:
2878 case ARM_VFP_MVFR1:
2879 if (IS_USER(s)
2880 || !arm_feature(env, ARM_FEATURE_VFP3))
2881 return 1;
2882 tmp = load_cpu_field(vfp.xregs[rn]);
2883 break;
2884 default:
2885 return 1;
2886 }
2887 } else {
2888 gen_mov_F0_vreg(0, rn);
2889 tmp = gen_vfp_mrs();
2890 }
2891 if (rd == 15) {
2892 /* Set the 4 flag bits in the CPSR. */
2893 gen_set_nzcv(tmp);
2894 dead_tmp(tmp);
2895 } else {
2896 store_reg(s, rd, tmp);
2897 }
2898 } else {
2899 /* arm->vfp */
2900 tmp = load_reg(s, rd);
2901 if (insn & (1 << 21)) {
2902 rn >>= 1;
2903 /* system register */
2904 switch (rn) {
2905 case ARM_VFP_FPSID:
2906 case ARM_VFP_MVFR0:
2907 case ARM_VFP_MVFR1:
2908 /* Writes are ignored. */
2909 break;
2910 case ARM_VFP_FPSCR:
2911 gen_helper_vfp_set_fpscr(cpu_env, tmp);
2912 dead_tmp(tmp);
2913 gen_lookup_tb(s);
2914 break;
2915 case ARM_VFP_FPEXC:
2916 if (IS_USER(s))
2917 return 1;
2918 store_cpu_field(tmp, vfp.xregs[rn]);
2919 gen_lookup_tb(s);
2920 break;
2921 case ARM_VFP_FPINST:
2922 case ARM_VFP_FPINST2:
2923 store_cpu_field(tmp, vfp.xregs[rn]);
2924 break;
2925 default:
2926 return 1;
2927 }
2928 } else {
2929 gen_vfp_msr(tmp);
2930 gen_mov_vreg_F0(0, rn);
2931 }
2932 }
2933 }
2934 } else {
2935 /* data processing */
2936 /* The opcode is in bits 23, 21, 20 and 6. */
2937 op = ((insn >> 20) & 8) | ((insn >> 19) & 6) | ((insn >> 6) & 1);
2938 if (dp) {
2939 if (op == 15) {
2940 /* rn is opcode */
2941 rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
2942 } else {
2943 /* rn is register number */
2944 VFP_DREG_N(rn, insn);
2945 }
2946
2947 if (op == 15 && (rn == 15 || rn > 17)) {
2948 /* Integer or single precision destination. */
2949 rd = VFP_SREG_D(insn);
2950 } else {
2951 VFP_DREG_D(rd, insn);
2952 }
2953
2954 if (op == 15 && (rn == 16 || rn == 17)) {
2955 /* Integer source. */
2956 rm = ((insn << 1) & 0x1e) | ((insn >> 5) & 1);
2957 } else {
2958 VFP_DREG_M(rm, insn);
2959 }
2960 } else {
2961 rn = VFP_SREG_N(insn);
2962 if (op == 15 && rn == 15) {
2963 /* Double precision destination. */
2964 VFP_DREG_D(rd, insn);
2965 } else {
2966 rd = VFP_SREG_D(insn);
2967 }
2968 rm = VFP_SREG_M(insn);
2969 }
2970
2971 veclen = env->vfp.vec_len;
2972 if (op == 15 && rn > 3)
2973 veclen = 0;
2974
2975 /* Shut up compiler warnings. */
2976 delta_m = 0;
2977 delta_d = 0;
2978 bank_mask = 0;
2979
2980 if (veclen > 0) {
2981 if (dp)
2982 bank_mask = 0xc;
2983 else
2984 bank_mask = 0x18;
2985
2986 /* Figure out what type of vector operation this is. */
2987 if ((rd & bank_mask) == 0) {
2988 /* scalar */
2989 veclen = 0;
2990 } else {
2991 if (dp)
2992 delta_d = (env->vfp.vec_stride >> 1) + 1;
2993 else
2994 delta_d = env->vfp.vec_stride + 1;
2995
2996 if ((rm & bank_mask) == 0) {
2997 /* mixed scalar/vector */
2998 delta_m = 0;
2999 } else {
3000 /* vector */
3001 delta_m = delta_d;
3002 }
3003 }
3004 }
3005
3006 /* Load the initial operands. */
3007 if (op == 15) {
3008 switch (rn) {
3009 case 16:
3010 case 17:
3011 /* Integer source */
3012 gen_mov_F0_vreg(0, rm);
3013 break;
3014 case 8:
3015 case 9:
3016 /* Compare */
3017 gen_mov_F0_vreg(dp, rd);
3018 gen_mov_F1_vreg(dp, rm);
3019 break;
3020 case 10:
3021 case 11:
3022 /* Compare with zero */
3023 gen_mov_F0_vreg(dp, rd);
3024 gen_vfp_F1_ld0(dp);
3025 break;
3026 case 20:
3027 case 21:
3028 case 22:
3029 case 23:
3030 /* Source and destination the same. */
3031 gen_mov_F0_vreg(dp, rd);
3032 break;
3033 default:
3034 /* One source operand. */
3035 gen_mov_F0_vreg(dp, rm);
3036 break;
3037 }
3038 } else {
3039 /* Two source operands. */
3040 gen_mov_F0_vreg(dp, rn);
3041 gen_mov_F1_vreg(dp, rm);
3042 }
3043
3044 for (;;) {
3045 /* Perform the calculation. */
3046 switch (op) {
3047 case 0: /* mac: fd + (fn * fm) */
3048 gen_vfp_mul(dp);
3049 gen_mov_F1_vreg(dp, rd);
3050 gen_vfp_add(dp);
3051 break;
3052 case 1: /* nmac: fd - (fn * fm) */
3053 gen_vfp_mul(dp);
3054 gen_vfp_neg(dp);
3055 gen_mov_F1_vreg(dp, rd);
3056 gen_vfp_add(dp);
3057 break;
3058 case 2: /* msc: -fd + (fn * fm) */
3059 gen_vfp_mul(dp);
3060 gen_mov_F1_vreg(dp, rd);
3061 gen_vfp_sub(dp);
3062 break;
3063 case 3: /* nmsc: -fd - (fn * fm) */
3064 gen_vfp_mul(dp);
3065 gen_vfp_neg(dp);
3066 gen_mov_F1_vreg(dp, rd);
3067 gen_vfp_sub(dp);
3068 break;
3069 case 4: /* mul: fn * fm */
3070 gen_vfp_mul(dp);
3071 break;
3072 case 5: /* nmul: -(fn * fm) */
3073 gen_vfp_mul(dp);
3074 gen_vfp_neg(dp);
3075 break;
3076 case 6: /* add: fn + fm */
3077 gen_vfp_add(dp);
3078 break;
3079 case 7: /* sub: fn - fm */
3080 gen_vfp_sub(dp);
3081 break;
3082 case 8: /* div: fn / fm */
3083 gen_vfp_div(dp);
3084 break;
3085 case 14: /* fconst */
3086 if (!arm_feature(env, ARM_FEATURE_VFP3))
3087 return 1;
3088
3089 n = (insn << 12) & 0x80000000;
3090 i = ((insn >> 12) & 0x70) | (insn & 0xf);
3091 if (dp) {
3092 if (i & 0x40)
3093 i |= 0x3f80;
3094 else
3095 i |= 0x4000;
3096 n |= i << 16;
3097 tcg_gen_movi_i64(cpu_F0d, ((uint64_t)n) << 32);
3098 } else {
3099 if (i & 0x40)
3100 i |= 0x780;
3101 else
3102 i |= 0x800;
3103 n |= i << 19;
3104 tcg_gen_movi_i32(cpu_F0s, n);
3105 }
3106 break;
3107 case 15: /* extension space */
3108 switch (rn) {
3109 case 0: /* cpy */
3110 /* no-op */
3111 break;
3112 case 1: /* abs */
3113 gen_vfp_abs(dp);
3114 break;
3115 case 2: /* neg */
3116 gen_vfp_neg(dp);
3117 break;
3118 case 3: /* sqrt */
3119 gen_vfp_sqrt(dp);
3120 break;
3121 case 8: /* cmp */
3122 gen_vfp_cmp(dp);
3123 break;
3124 case 9: /* cmpe */
3125 gen_vfp_cmpe(dp);
3126 break;
3127 case 10: /* cmpz */
3128 gen_vfp_cmp(dp);
3129 break;
3130 case 11: /* cmpez */
3131 gen_vfp_F1_ld0(dp);
3132 gen_vfp_cmpe(dp);
3133 break;
3134 case 15: /* single<->double conversion */
3135 if (dp)
3136 gen_helper_vfp_fcvtsd(cpu_F0s, cpu_F0d, cpu_env);
3137 else
3138 gen_helper_vfp_fcvtds(cpu_F0d, cpu_F0s, cpu_env);
3139 break;
3140 case 16: /* fuito */
3141 gen_vfp_uito(dp);
3142 break;
3143 case 17: /* fsito */
3144 gen_vfp_sito(dp);
3145 break;
3146 case 20: /* fshto */
3147 if (!arm_feature(env, ARM_FEATURE_VFP3))
3148 return 1;
3149 gen_vfp_shto(dp, rm);
3150 break;
3151 case 21: /* fslto */
3152 if (!arm_feature(env, ARM_FEATURE_VFP3))
3153 return 1;
3154 gen_vfp_slto(dp, rm);
3155 break;
3156 case 22: /* fuhto */
3157 if (!arm_feature(env, ARM_FEATURE_VFP3))
3158 return 1;
3159 gen_vfp_uhto(dp, rm);
3160 break;
3161 case 23: /* fulto */
3162 if (!arm_feature(env, ARM_FEATURE_VFP3))
3163 return 1;
3164 gen_vfp_ulto(dp, rm);
3165 break;
3166 case 24: /* ftoui */
3167 gen_vfp_toui(dp);
3168 break;
3169 case 25: /* ftouiz */
3170 gen_vfp_touiz(dp);
3171 break;
3172 case 26: /* ftosi */
3173 gen_vfp_tosi(dp);
3174 break;
3175 case 27: /* ftosiz */
3176 gen_vfp_tosiz(dp);
3177 break;
3178 case 28: /* ftosh */
3179 if (!arm_feature(env, ARM_FEATURE_VFP3))
3180 return 1;
3181 gen_vfp_tosh(dp, rm);
3182 break;
3183 case 29: /* ftosl */
3184 if (!arm_feature(env, ARM_FEATURE_VFP3))
3185 return 1;
3186 gen_vfp_tosl(dp, rm);
3187 break;
3188 case 30: /* ftouh */
3189 if (!arm_feature(env, ARM_FEATURE_VFP3))
3190 return 1;
3191 gen_vfp_touh(dp, rm);
3192 break;
3193 case 31: /* ftoul */
3194 if (!arm_feature(env, ARM_FEATURE_VFP3))
3195 return 1;
3196 gen_vfp_toul(dp, rm);
3197 break;
3198 default: /* undefined */
3199 printf ("rn:%d\n", rn);
3200 return 1;
3201 }
3202 break;
3203 default: /* undefined */
3204 printf ("op:%d\n", op);
3205 return 1;
3206 }
3207
3208 /* Write back the result. */
3209 if (op == 15 && (rn >= 8 && rn <= 11))
3210 ; /* Comparison, do nothing. */
3211 else if (op == 15 && rn > 17)
3212 /* Integer result. */
3213 gen_mov_vreg_F0(0, rd);
3214 else if (op == 15 && rn == 15)
3215 /* conversion */
3216 gen_mov_vreg_F0(!dp, rd);
3217 else
3218 gen_mov_vreg_F0(dp, rd);
3219
3220 /* break out of the loop if we have finished */
3221 if (veclen == 0)
3222 break;
3223
3224 if (op == 15 && delta_m == 0) {
3225 /* single source one-many */
3226 while (veclen--) {
3227 rd = ((rd + delta_d) & (bank_mask - 1))
3228 | (rd & bank_mask);
3229 gen_mov_vreg_F0(dp, rd);
3230 }
3231 break;
3232 }
3233 /* Setup the next operands. */
3234 veclen--;
3235 rd = ((rd + delta_d) & (bank_mask - 1))
3236 | (rd & bank_mask);
3237
3238 if (op == 15) {
3239 /* One source operand. */
3240 rm = ((rm + delta_m) & (bank_mask - 1))
3241 | (rm & bank_mask);
3242 gen_mov_F0_vreg(dp, rm);
3243 } else {
3244 /* Two source operands. */
3245 rn = ((rn + delta_d) & (bank_mask - 1))
3246 | (rn & bank_mask);
3247 gen_mov_F0_vreg(dp, rn);
3248 if (delta_m) {
3249 rm = ((rm + delta_m) & (bank_mask - 1))
3250 | (rm & bank_mask);
3251 gen_mov_F1_vreg(dp, rm);
3252 }
3253 }
3254 }
3255 }
3256 break;
3257 case 0xc:
3258 case 0xd:
3259 if (dp && (insn & 0x03e00000) == 0x00400000) {
3260 /* two-register transfer */
3261 rn = (insn >> 16) & 0xf;
3262 rd = (insn >> 12) & 0xf;
3263 if (dp) {
3264 VFP_DREG_M(rm, insn);
3265 } else {
3266 rm = VFP_SREG_M(insn);
3267 }
3268
3269 if (insn & ARM_CP_RW_BIT) {
3270 /* vfp->arm */
3271 if (dp) {
3272 gen_mov_F0_vreg(0, rm * 2);
3273 tmp = gen_vfp_mrs();
3274 store_reg(s, rd, tmp);
3275 gen_mov_F0_vreg(0, rm * 2 + 1);
3276 tmp = gen_vfp_mrs();
3277 store_reg(s, rn, tmp);
3278 } else {
3279 gen_mov_F0_vreg(0, rm);
3280 tmp = gen_vfp_mrs();
3281 store_reg(s, rn, tmp);
3282 gen_mov_F0_vreg(0, rm + 1);
3283 tmp = gen_vfp_mrs();
3284 store_reg(s, rd, tmp);
3285 }
3286 } else {
3287 /* arm->vfp */
3288 if (dp) {
3289 tmp = load_reg(s, rd);
3290 gen_vfp_msr(tmp);
3291 gen_mov_vreg_F0(0, rm * 2);
3292 tmp = load_reg(s, rn);
3293 gen_vfp_msr(tmp);
3294 gen_mov_vreg_F0(0, rm * 2 + 1);
3295 } else {
3296 tmp = load_reg(s, rn);
3297 gen_vfp_msr(tmp);
3298 gen_mov_vreg_F0(0, rm);
3299 tmp = load_reg(s, rd);
3300 gen_vfp_msr(tmp);
3301 gen_mov_vreg_F0(0, rm + 1);
3302 }
3303 }
3304 } else {
3305 /* Load/store */
3306 rn = (insn >> 16) & 0xf;
3307 if (dp)
3308 VFP_DREG_D(rd, insn);
3309 else
3310 rd = VFP_SREG_D(insn);
3311 if (s->thumb && rn == 15) {
3312 gen_op_movl_T1_im(s->pc & ~2);
3313 } else {
3314 gen_movl_T1_reg(s, rn);
3315 }
3316 if ((insn & 0x01200000) == 0x01000000) {
3317 /* Single load/store */
3318 offset = (insn & 0xff) << 2;
3319 if ((insn & (1 << 23)) == 0)
3320 offset = -offset;
3321 gen_op_addl_T1_im(offset);
3322 if (insn & (1 << 20)) {
3323 gen_vfp_ld(s, dp);
3324 gen_mov_vreg_F0(dp, rd);
3325 } else {
3326 gen_mov_F0_vreg(dp, rd);
3327 gen_vfp_st(s, dp);
3328 }
3329 } else {
3330 /* load/store multiple */
3331 if (dp)
3332 n = (insn >> 1) & 0x7f;
3333 else
3334 n = insn & 0xff;
3335
3336 if (insn & (1 << 24)) /* pre-decrement */
3337 gen_op_addl_T1_im(-((insn & 0xff) << 2));
3338
3339 if (dp)
3340 offset = 8;
3341 else
3342 offset = 4;
3343 for (i = 0; i < n; i++) {
3344 if (insn & ARM_CP_RW_BIT) {
3345 /* load */
3346 gen_vfp_ld(s, dp);
3347 gen_mov_vreg_F0(dp, rd + i);
3348 } else {
3349 /* store */
3350 gen_mov_F0_vreg(dp, rd + i);
3351 gen_vfp_st(s, dp);
3352 }
3353 gen_op_addl_T1_im(offset);
3354 }
3355 if (insn & (1 << 21)) {
3356 /* writeback */
3357 if (insn & (1 << 24))
3358 offset = -offset * n;
3359 else if (dp && (insn & 1))
3360 offset = 4;
3361 else
3362 offset = 0;
3363
3364 if (offset != 0)
3365 gen_op_addl_T1_im(offset);
3366 gen_movl_reg_T1(s, rn);
3367 }
3368 }
3369 }
3370 break;
3371 default:
3372 /* Should never happen. */
3373 return 1;
3374 }
3375 return 0;
3376 }
3377
3378 static inline void gen_goto_tb(DisasContext *s, int n, uint32_t dest)
3379 {
3380 TranslationBlock *tb;
3381
3382 tb = s->tb;
3383 if ((tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK)) {
3384 tcg_gen_goto_tb(n);
3385 gen_set_pc_im(dest);
3386 tcg_gen_exit_tb((long)tb + n);
3387 } else {
3388 gen_set_pc_im(dest);
3389 tcg_gen_exit_tb(0);
3390 }
3391 }
3392
3393 static inline void gen_jmp (DisasContext *s, uint32_t dest)
3394 {
3395 if (unlikely(s->singlestep_enabled)) {
3396 /* An indirect jump so that we still trigger the debug exception. */
3397 if (s->thumb)
3398 dest |= 1;
3399 gen_bx_im(s, dest);
3400 } else {
3401 gen_goto_tb(s, 0, dest);
3402 s->is_jmp = DISAS_TB_JUMP;
3403 }
3404 }
3405
3406 static inline void gen_mulxy(TCGv t0, TCGv t1, int x, int y)
3407 {
3408 if (x)
3409 tcg_gen_sari_i32(t0, t0, 16);
3410 else
3411 gen_sxth(t0);
3412 if (y)
3413 tcg_gen_sari_i32(t1, t1, 16);
3414 else
3415 gen_sxth(t1);
3416 tcg_gen_mul_i32(t0, t0, t1);
3417 }
3418
3419 /* Return the mask of PSR bits set by a MSR instruction. */
3420 static uint32_t msr_mask(CPUState *env, DisasContext *s, int flags, int spsr) {
3421 uint32_t mask;
3422
3423 mask = 0;
3424 if (flags & (1 << 0))
3425 mask |= 0xff;
3426 if (flags & (1 << 1))
3427 mask |= 0xff00;
3428 if (flags & (1 << 2))
3429 mask |= 0xff0000;
3430 if (flags & (1 << 3))
3431 mask |= 0xff000000;
3432
3433 /* Mask out undefined bits. */
3434 mask &= ~CPSR_RESERVED;
3435 if (!arm_feature(env, ARM_FEATURE_V6))
3436 mask &= ~(CPSR_E | CPSR_GE);
3437 if (!arm_feature(env, ARM_FEATURE_THUMB2))
3438 mask &= ~CPSR_IT;
3439 /* Mask out execution state bits. */
3440 if (!spsr)
3441 mask &= ~CPSR_EXEC;
3442 /* Mask out privileged bits. */
3443 if (IS_USER(s))
3444 mask &= CPSR_USER;
3445 return mask;
3446 }
3447
3448 /* Returns nonzero if access to the PSR is not permitted. */
3449 static int gen_set_psr_T0(DisasContext *s, uint32_t mask, int spsr)
3450 {
3451 TCGv tmp;
3452 if (spsr) {
3453 /* ??? This is also undefined in system mode. */
3454 if (IS_USER(s))
3455 return 1;
3456
3457 tmp = load_cpu_field(spsr);
3458 tcg_gen_andi_i32(tmp, tmp, ~mask);
3459 tcg_gen_andi_i32(cpu_T[0], cpu_T[0], mask);
3460 tcg_gen_or_i32(tmp, tmp, cpu_T[0]);
3461 store_cpu_field(tmp, spsr);
3462 } else {
3463 gen_set_cpsr(cpu_T[0], mask);
3464 }
3465 gen_lookup_tb(s);
3466 return 0;
3467 }
3468
3469 /* Generate an old-style exception return. */
3470 static void gen_exception_return(DisasContext *s)
3471 {
3472 TCGv tmp;
3473 gen_movl_reg_T0(s, 15);
3474 tmp = load_cpu_field(spsr);
3475 gen_set_cpsr(tmp, 0xffffffff);
3476 dead_tmp(tmp);
3477 s->is_jmp = DISAS_UPDATE;
3478 }
3479
3480 /* Generate a v6 exception return. Marks both values as dead. */
3481 static void gen_rfe(DisasContext *s, TCGv pc, TCGv cpsr)
3482 {
3483 gen_set_cpsr(cpsr, 0xffffffff);
3484 dead_tmp(cpsr);
3485 store_reg(s, 15, pc);
3486 s->is_jmp = DISAS_UPDATE;
3487 }
3488
3489 static inline void
3490 gen_set_condexec (DisasContext *s)
3491 {
3492 if (s->condexec_mask) {
3493 uint32_t val = (s->condexec_cond << 4) | (s->condexec_mask >> 1);
3494 TCGv tmp = new_tmp();
3495 tcg_gen_movi_i32(tmp, val);
3496 store_cpu_field(tmp, condexec_bits);
3497 }
3498 }
3499
3500 static void gen_nop_hint(DisasContext *s, int val)
3501 {
3502 switch (val) {
3503 case 3: /* wfi */
3504 gen_set_pc_im(s->pc);
3505 s->is_jmp = DISAS_WFI;
3506 break;
3507 case 2: /* wfe */
3508 case 4: /* sev */
3509 /* TODO: Implement SEV and WFE. May help SMP performance. */
3510 default: /* nop */
3511 break;
3512 }
3513 }
3514
3515 /* These macros help make the code more readable when migrating from the
3516 old dyngen helpers. They should probably be removed when
3517 T0/T1 are removed. */
3518 #define CPU_T001 cpu_T[0], cpu_T[0], cpu_T[1]
3519 #define CPU_T0E01 cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]
3520
3521 #define CPU_V001 cpu_V0, cpu_V0, cpu_V1
3522
3523 static inline int gen_neon_add(int size)
3524 {
3525 switch (size) {
3526 case 0: gen_helper_neon_add_u8(CPU_T001); break;
3527 case 1: gen_helper_neon_add_u16(CPU_T001); break;
3528 case 2: gen_op_addl_T0_T1(); break;
3529 default: return 1;
3530 }
3531 return 0;
3532 }
3533
3534 static inline void gen_neon_rsb(int size)
3535 {
3536 switch (size) {
3537 case 0: gen_helper_neon_sub_u8(cpu_T[0], cpu_T[1], cpu_T[0]); break;
3538 case 1: gen_helper_neon_sub_u16(cpu_T[0], cpu_T[1], cpu_T[0]); break;
3539 case 2: gen_op_rsbl_T0_T1(); break;
3540 default: return;
3541 }
3542 }
3543
3544 /* 32-bit pairwise ops end up the same as the elementwise versions. */
3545 #define gen_helper_neon_pmax_s32 gen_helper_neon_max_s32
3546 #define gen_helper_neon_pmax_u32 gen_helper_neon_max_u32
3547 #define gen_helper_neon_pmin_s32 gen_helper_neon_min_s32
3548 #define gen_helper_neon_pmin_u32 gen_helper_neon_min_u32
3549
3550 /* FIXME: This is wrong. They set the wrong overflow bit. */
3551 #define gen_helper_neon_qadd_s32(a, e, b, c) gen_helper_add_saturate(a, b, c)
3552 #define gen_helper_neon_qadd_u32(a, e, b, c) gen_helper_add_usaturate(a, b, c)
3553 #define gen_helper_neon_qsub_s32(a, e, b, c) gen_helper_sub_saturate(a, b, c)
3554 #define gen_helper_neon_qsub_u32(a, e, b, c) gen_helper_sub_usaturate(a, b, c)
3555
3556 #define GEN_NEON_INTEGER_OP_ENV(name) do { \
3557 switch ((size << 1) | u) { \
3558 case 0: \
3559 gen_helper_neon_##name##_s8(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]); \
3560 break; \
3561 case 1: \
3562 gen_helper_neon_##name##_u8(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]); \
3563 break; \
3564 case 2: \
3565 gen_helper_neon_##name##_s16(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]); \
3566 break; \
3567 case 3: \
3568 gen_helper_neon_##name##_u16(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]); \
3569 break; \
3570 case 4: \
3571 gen_helper_neon_##name##_s32(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]); \
3572 break; \
3573 case 5: \
3574 gen_helper_neon_##name##_u32(cpu_T[0], cpu_env, cpu_T[0], cpu_T[1]); \
3575 break; \
3576 default: return 1; \
3577 }} while (0)
3578
3579 #define GEN_NEON_INTEGER_OP(name) do { \
3580 switch ((size << 1) | u) { \
3581 case 0: \
3582 gen_helper_neon_##name##_s8(cpu_T[0], cpu_T[0], cpu_T[1]); \
3583 break; \
3584 case 1: \
3585 gen_helper_neon_##name##_u8(cpu_T[0], cpu_T[0], cpu_T[1]); \
3586 break; \
3587 case 2: \
3588 gen_helper_neon_##name##_s16(cpu_T[0], cpu_T[0], cpu_T[1]); \
3589 break; \
3590 case 3: \
3591 gen_helper_neon_##name##_u16(cpu_T[0], cpu_T[0], cpu_T[1]); \
3592 break; \
3593 case 4: \
3594 gen_helper_neon_##name##_s32(cpu_T[0], cpu_T[0], cpu_T[1]); \
3595 break; \
3596 case 5: \
3597 gen_helper_neon_##name##_u32(cpu_T[0], cpu_T[0], cpu_T[1]); \
3598 break; \
3599 default: return 1; \
3600 }} while (0)
3601
3602 static inline void
3603 gen_neon_movl_scratch_T0(int scratch)
3604 {
3605 uint32_t offset;
3606
3607 offset = offsetof(CPUARMState, vfp.scratch[scratch]);
3608 tcg_gen_st_i32(cpu_T[0], cpu_env, offset);
3609 }
3610
3611 static inline void
3612 gen_neon_movl_scratch_T1(int scratch)
3613 {
3614 uint32_t offset;
3615
3616 offset = offsetof(CPUARMState, vfp.scratch[scratch]);
3617 tcg_gen_st_i32(cpu_T[1], cpu_env, offset);
3618 }
3619
3620 static inline void
3621 gen_neon_movl_T0_scratch(int scratch)
3622 {
3623 uint32_t offset;
3624
3625 offset = offsetof(CPUARMState, vfp.scratch[scratch]);
3626 tcg_gen_ld_i32(cpu_T[0], cpu_env, offset);
3627 }
3628
3629 static inline void
3630 gen_neon_movl_T1_scratch(int scratch)
3631 {
3632 uint32_t offset;
3633
3634 offset = offsetof(CPUARMState, vfp.scratch[scratch]);
3635 tcg_gen_ld_i32(cpu_T[1], cpu_env, offset);
3636 }
3637
3638 static inline void gen_neon_get_scalar(int size, int reg)
3639 {
3640 if (size == 1) {
3641 NEON_GET_REG(T0, reg >> 1, reg & 1);
3642 } else {
3643 NEON_GET_REG(T0, reg >> 2, (reg >> 1) & 1);
3644 if (reg & 1)
3645 gen_neon_dup_low16(cpu_T[0]);
3646 else
3647 gen_neon_dup_high16(cpu_T[0]);
3648 }
3649 }
3650
3651 static void gen_neon_unzip(int reg, int q, int tmp, int size)
3652 {
3653 int n;
3654
3655 for (n = 0; n < q + 1; n += 2) {
3656 NEON_GET_REG(T0, reg, n);
3657 NEON_GET_REG(T0, reg, n + n);
3658 switch (size) {
3659 case 0: gen_helper_neon_unzip_u8(); break;
3660 case 1: gen_helper_neon_zip_u16(); break; /* zip and unzip are the same. */
3661 case 2: /* no-op */; break;
3662 default: abort();
3663 }
3664 gen_neon_movl_scratch_T0(tmp + n);
3665 gen_neon_movl_scratch_T1(tmp + n + 1);
3666 }
3667 }
3668
3669 static struct {
3670 int nregs;
3671 int interleave;
3672 int spacing;
3673 } neon_ls_element_type[11] = {
3674 {4, 4, 1},
3675 {4, 4, 2},
3676 {4, 1, 1},
3677 {4, 2, 1},
3678 {3, 3, 1},
3679 {3, 3, 2},
3680 {3, 1, 1},
3681 {1, 1, 1},
3682 {2, 2, 1},
3683 {2, 2, 2},
3684 {2, 1, 1}
3685 };
3686
3687 /* Translate a NEON load/store element instruction. Return nonzero if the
3688 instruction is invalid. */
3689 static int disas_neon_ls_insn(CPUState * env, DisasContext *s, uint32_t insn)
3690 {
3691 int rd, rn, rm;
3692 int op;
3693 int nregs;
3694 int interleave;
3695 int stride;
3696 int size;
3697 int reg;
3698 int pass;
3699 int load;
3700 int shift;
3701 int n;
3702 TCGv tmp;
3703 TCGv tmp2;
3704
3705 if (!vfp_enabled(env))
3706 return 1;
3707 VFP_DREG_D(rd, insn);
3708 rn = (insn >> 16) & 0xf;
3709 rm = insn & 0xf;
3710 load = (insn & (1 << 21)) != 0;
3711 if ((insn & (1 << 23)) == 0) {
3712 /* Load store all elements. */
3713 op = (insn >> 8) & 0xf;
3714 size = (insn >> 6) & 3;
3715 if (op > 10 || size == 3)
3716 return 1;
3717 nregs = neon_ls_element_type[op].nregs;
3718 interleave = neon_ls_element_type[op].interleave;
3719 gen_movl_T1_reg(s, rn);
3720 stride = (1 << size) * interleave;
3721 for (reg = 0; reg < nregs; reg++) {
3722 if (interleave > 2 || (interleave == 2 && nregs == 2)) {
3723 gen_movl_T1_reg(s, rn);
3724 gen_op_addl_T1_im((1 << size) * reg);
3725 } else if (interleave == 2 && nregs == 4 && reg == 2) {
3726 gen_movl_T1_reg(s, rn);
3727 gen_op_addl_T1_im(1 << size);
3728 }
3729 for (pass = 0; pass < 2; pass++) {
3730 if (size == 2) {
3731 if (load) {
3732 tmp = gen_ld32(cpu_T[1], IS_USER(s));
3733 neon_store_reg(rd, pass, tmp);
3734 } else {
3735 tmp = neon_load_reg(rd, pass);
3736 gen_st32(tmp, cpu_T[1], IS_USER(s));
3737 }
3738 gen_op_addl_T1_im(stride);
3739 } else if (size == 1) {
3740 if (load) {
3741 tmp = gen_ld16u(cpu_T[1], IS_USER(s));
3742 gen_op_addl_T1_im(stride);
3743 tmp2 = gen_ld16u(cpu_T[1], IS_USER(s));
3744 gen_op_addl_T1_im(stride);
3745 gen_bfi(tmp, tmp, tmp2, 16, 0xffff);
3746 dead_tmp(tmp2);
3747 neon_store_reg(rd, pass, tmp);
3748 } else {
3749 tmp = neon_load_reg(rd, pass);
3750 tmp2 = new_tmp();
3751 tcg_gen_shri_i32(tmp2, tmp, 16);
3752 gen_st16(tmp, cpu_T[1], IS_USER(s));
3753 gen_op_addl_T1_im(stride);
3754 gen_st16(tmp2, cpu_T[1], IS_USER(s));
3755 gen_op_addl_T1_im(stride);
3756 }
3757 } else /* size == 0 */ {
3758 if (load) {
3759 TCGV_UNUSED(tmp2);
3760 for (n = 0; n < 4; n++) {
3761 tmp = gen_ld8u(cpu_T[1], IS_USER(s));
3762 gen_op_addl_T1_im(stride);
3763 if (n == 0) {
3764 tmp2 = tmp;
3765 } else {
3766 gen_bfi(tmp2, tmp2, tmp, n * 8, 0xff);
3767 dead_tmp(tmp);
3768 }
3769 }
3770 neon_store_reg(rd, pass, tmp2);
3771 } else {
3772 tmp2 = neon_load_reg(rd, pass);
3773 for (n = 0; n < 4; n++) {
3774 tmp = new_tmp();
3775 if (n == 0) {
3776 tcg_gen_mov_i32(tmp, tmp2);
3777 } else {
3778 tcg_gen_shri_i32(tmp, tmp2, n * 8);
3779 }
3780 gen_st8(tmp, cpu_T[1], IS_USER(s));
3781 gen_op_addl_T1_im(stride);
3782 }
3783 dead_tmp(tmp2);
3784 }
3785 }
3786 }
3787 rd += neon_ls_element_type[op].spacing;
3788 }
3789 stride = nregs * 8;
3790 } else {
3791 size = (insn >> 10) & 3;
3792 if (size == 3) {
3793 /* Load single element to all lanes. */
3794 if (!load)
3795 return 1;
3796 size = (insn >> 6) & 3;
3797 nregs = ((insn >> 8) & 3) + 1;
3798 stride = (insn & (1 << 5)) ? 2 : 1;
3799 gen_movl_T1_reg(s, rn);
3800 for (reg = 0; reg < nregs; reg++) {
3801 switch (size) {
3802 case 0:
3803 tmp = gen_ld8u(cpu_T[1], IS_USER(s));
3804 gen_neon_dup_u8(tmp, 0);
3805 break;
3806 case 1:
3807 tmp = gen_ld16u(cpu_T[1], IS_USER(s));
3808 gen_neon_dup_low16(tmp);
3809 break;
3810 case 2:
3811 tmp = gen_ld32(cpu_T[0], IS_USER(s));
3812 break;
3813 case 3:
3814 return 1;
3815 default: /* Avoid compiler warnings. */
3816 abort();
3817 }
3818 gen_op_addl_T1_im(1 << size);
3819 tmp2 = new_tmp();
3820 tcg_gen_mov_i32(tmp2, tmp);
3821 neon_store_reg(rd, 0, tmp2);
3822 neon_store_reg(rd, 1, tmp);
3823 rd += stride;
3824 }
3825 stride = (1 << size) * nregs;
3826 } else {
3827 /* Single element. */
3828 pass = (insn >> 7) & 1;
3829 switch (size) {
3830 case 0:
3831 shift = ((insn >> 5) & 3) * 8;
3832 stride = 1;
3833 break;
3834 case 1:
3835 shift = ((insn >> 6) & 1) * 16;
3836 stride = (insn & (1 << 5)) ? 2 : 1;
3837 break;
3838 case 2:
3839 shift = 0;
3840 stride = (insn & (1 << 6)) ? 2 : 1;
3841 break;
3842 default:
3843 abort();
3844 }
3845 nregs = ((insn >> 8) & 3) + 1;
3846 gen_movl_T1_reg(s, rn);
3847 for (reg = 0; reg < nregs; reg++) {
3848 if (load) {
3849 switch (size) {
3850 case 0:
3851 tmp = gen_ld8u(cpu_T[1], IS_USER(s));
3852 break;
3853 case 1:
3854 tmp = gen_ld16u(cpu_T[1], IS_USER(s));
3855 break;
3856 case 2:
3857 tmp = gen_ld32(cpu_T[1], IS_USER(s));
3858 break;
3859 default: /* Avoid compiler warnings. */
3860 abort();
3861 }
3862 if (size != 2) {
3863 tmp2 = neon_load_reg(rd, pass);
3864 gen_bfi(tmp, tmp2, tmp, shift, size ? 0xffff : 0xff);
3865 dead_tmp(tmp2);
3866 }
3867 neon_store_reg(rd, pass, tmp);
3868 } else { /* Store */
3869 tmp = neon_load_reg(rd, pass);
3870 if (shift)
3871 tcg_gen_shri_i32(tmp, tmp, shift);
3872 switch (size) {
3873 case 0:
3874 gen_st8(tmp, cpu_T[1], IS_USER(s));
3875 break;
3876 case 1:
3877 gen_st16(tmp, cpu_T[1], IS_USER(s));
3878 break;
3879 case 2:
3880 gen_st32(tmp, cpu_T[1], IS_USER(s));
3881 break;
3882 }
3883 }
3884 rd += stride;
3885 gen_op_addl_T1_im(1 << size);
3886 }
3887 stride = nregs * (1 << size);
3888 }
3889 }
3890 if (rm != 15) {
3891 TCGv base;
3892
3893 base = load_reg(s, rn);
3894 if (rm == 13) {
3895 tcg_gen_addi_i32(base, base, stride);
3896 } else {
3897 TCGv index;
3898 index = load_reg(s, rm);
3899 tcg_gen_add_i32(base, base, index);
3900 dead_tmp(index);
3901 }
3902 store_reg(s, rn, base);
3903 }
3904 return 0;
3905 }
3906
3907 /* Bitwise select. dest = c ? t : f. Clobbers T and F. */
3908 static void gen_neon_bsl(TCGv dest, TCGv t, TCGv f, TCGv c)
3909 {
3910 tcg_gen_and_i32(t, t, c);
3911 tcg_gen_bic_i32(f, f, c);
3912 tcg_gen_or_i32(dest, t, f);
3913 }
3914
3915 static inline void gen_neon_narrow(int size, TCGv dest, TCGv_i64 src)
3916 {
3917 switch (size) {
3918 case 0: gen_helper_neon_narrow_u8(dest, src); break;
3919 case 1: gen_helper_neon_narrow_u16(dest, src); break;
3920 case 2: tcg_gen_trunc_i64_i32(dest, src); break;
3921 default: abort();
3922 }
3923 }
3924
3925 static inline void gen_neon_narrow_sats(int size, TCGv dest, TCGv_i64 src)
3926 {
3927 switch (size) {
3928 case 0: gen_helper_neon_narrow_sat_s8(dest, cpu_env, src); break;
3929 case 1: gen_helper_neon_narrow_sat_s16(dest, cpu_env, src); break;
3930 case 2: gen_helper_neon_narrow_sat_s32(dest, cpu_env, src); break;
3931 default: abort();
3932 }
3933 }
3934
3935 static inline void gen_neon_narrow_satu(int size, TCGv dest, TCGv_i64 src)
3936 {
3937 switch (size) {
3938 case 0: gen_helper_neon_narrow_sat_u8(dest, cpu_env, src); break;
3939 case 1: gen_helper_neon_narrow_sat_u16(dest, cpu_env, src); break;
3940 case 2: gen_helper_neon_narrow_sat_u32(dest, cpu_env, src); break;
3941 default: abort();
3942 }
3943 }
3944
3945 static inline void gen_neon_shift_narrow(int size, TCGv var, TCGv shift,
3946 int q, int u)
3947 {
3948 if (q) {
3949 if (u) {
3950 switch (size) {
3951 case 1: gen_helper_neon_rshl_u16(var, var, shift); break;
3952 case 2: gen_helper_neon_rshl_u32(var, var, shift); break;
3953 default: abort();
3954 }
3955 } else {
3956 switch (size) {
3957 case 1: gen_helper_neon_rshl_s16(var, var, shift); break;
3958 case 2: gen_helper_neon_rshl_s32(var, var, shift); break;
3959 default: abort();
3960 }
3961 }
3962 } else {
3963 if (u) {
3964 switch (size) {
3965 case 1: gen_helper_neon_rshl_u16(var, var, shift); break;
3966 case 2: gen_helper_neon_rshl_u32(var, var, shift); break;
3967 default: abort();
3968 }
3969 } else {
3970 switch (size) {
3971 case 1: gen_helper_neon_shl_s16(var, var, shift); break;
3972 case 2: gen_helper_neon_shl_s32(var, var, shift); break;
3973 default: abort();
3974 }
3975 }
3976 }
3977 }
3978
3979 static inline void gen_neon_widen(TCGv_i64 dest, TCGv src, int size, int u)
3980 {
3981 if (u) {
3982 switch (size) {
3983 case 0: gen_helper_neon_widen_u8(dest, src); break;
3984 case 1: gen_helper_neon_widen_u16(dest, src); break;
3985 case 2: tcg_gen_extu_i32_i64(dest, src); break;
3986 default: abort();
3987 }
3988 } else {
3989 switch (size) {
3990 case 0: gen_helper_neon_widen_s8(dest, src); break;
3991 case 1: gen_helper_neon_widen_s16(dest, src); break;
3992 case 2: tcg_gen_ext_i32_i64(dest, src); break;
3993 default: abort();
3994 }
3995 }
3996 dead_tmp(src);
3997 }
3998
3999 static inline void gen_neon_addl(int size)
4000 {
4001 switch (size) {
4002 case 0: gen_helper_neon_addl_u16(CPU_V001); break;
4003 case 1: gen_helper_neon_addl_u32(CPU_V001); break;
4004 case 2: tcg_gen_add_i64(CPU_V001); break;
4005 default: abort();
4006 }
4007 }
4008
4009 static inline void gen_neon_subl(int size)
4010 {
4011 switch (size) {
4012 case 0: gen_helper_neon_subl_u16(CPU_V001); break;
4013 case 1: gen_helper_neon_subl_u32(CPU_V001); break;
4014 case 2: tcg_gen_sub_i64(CPU_V001); break;
4015 default: abort();
4016 }
4017 }
4018
4019 static inline void gen_neon_negl(TCGv_i64 var, int size)
4020 {
4021 switch (size) {
4022 case 0: gen_helper_neon_negl_u16(var, var); break;
4023 case 1: gen_helper_neon_negl_u32(var, var); break;
4024 case 2: gen_helper_neon_negl_u64(var, var); break;
4025 default: abort();
4026 }
4027 }
4028
4029 static inline void gen_neon_addl_saturate(TCGv_i64 op0, TCGv_i64 op1, int size)
4030 {
4031 switch (size) {
4032 case 1: gen_helper_neon_addl_saturate_s32(op0, cpu_env, op0, op1); break;
4033 case 2: gen_helper_neon_addl_saturate_s64(op0, cpu_env, op0, op1); break;
4034 default: abort();
4035 }
4036 }
4037
4038 static inline void gen_neon_mull(TCGv_i64 dest, TCGv a, TCGv b, int size, int u)
4039 {
4040 TCGv_i64 tmp;
4041
4042 switch ((size << 1) | u) {
4043 case 0: gen_helper_neon_mull_s8(dest, a, b); break;
4044 case 1: gen_helper_neon_mull_u8(dest, a, b); break;
4045 case 2: gen_helper_neon_mull_s16(dest, a, b); break;
4046 case 3: gen_helper_neon_mull_u16(dest, a, b); break;
4047 case 4:
4048 tmp = gen_muls_i64_i32(a, b);
4049 tcg_gen_mov_i64(dest, tmp);
4050 break;
4051 case 5:
4052 tmp = gen_mulu_i64_i32(a, b);
4053 tcg_gen_mov_i64(dest, tmp);
4054 break;
4055 default: abort();
4056 }
4057 if (size < 2) {
4058 dead_tmp(b);
4059 dead_tmp(a);
4060 }
4061 }
4062
4063 /* Translate a NEON data processing instruction. Return nonzero if the
4064 instruction is invalid.
4065 We process data in a mixture of 32-bit and 64-bit chunks.
4066 Mostly we use 32-bit chunks so we can use normal scalar instructions. */
4067
4068 static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
4069 {
4070 int op;
4071 int q;
4072 int rd, rn, rm;
4073 int size;
4074 int shift;
4075 int pass;
4076 int count;
4077 int pairwise;
4078 int u;
4079 int n;
4080 uint32_t imm;
4081 TCGv tmp;
4082 TCGv tmp2;
4083 TCGv tmp3;
4084 TCGv_i64 tmp64;
4085
4086 if (!vfp_enabled(env))
4087 return 1;
4088 q = (insn & (1 << 6)) != 0;
4089 u = (insn >> 24) & 1;
4090 VFP_DREG_D(rd, insn);
4091 VFP_DREG_N(rn, insn);
4092 VFP_DREG_M(rm, insn);
4093 size = (insn >> 20) & 3;
4094 if ((insn & (1 << 23)) == 0) {
4095 /* Three register same length. */
4096 op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1);
4097 if (size == 3 && (op == 1 || op == 5 || op == 8 || op == 9
4098 || op == 10 || op == 11 || op == 16)) {
4099 /* 64-bit element instructions. */
4100 for (pass = 0; pass < (q ? 2 : 1); pass++) {
4101 neon_load_reg64(cpu_V0, rn + pass);
4102 neon_load_reg64(cpu_V1, rm + pass);
4103 switch (op) {
4104 case 1: /* VQADD */
4105 if (u) {
4106 gen_helper_neon_add_saturate_u64(CPU_V001);
4107 } else {
4108 gen_helper_neon_add_saturate_s64(CPU_V001);
4109 }
4110 break;
4111 case 5: /* VQSUB */
4112 if (u) {
4113 gen_helper_neon_sub_saturate_u64(CPU_V001);
4114 } else {
4115 gen_helper_neon_sub_saturate_s64(CPU_V001);
4116 }
4117 break;
4118 case 8: /* VSHL */
4119 if (u) {
4120 gen_helper_neon_shl_u64(cpu_V0, cpu_V1, cpu_V0);
4121 } else {
4122 gen_helper_neon_shl_s64(cpu_V0, cpu_V1, cpu_V0);
4123 }
4124 break;
4125 case 9: /* VQSHL */
4126 if (u) {
4127 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
4128 cpu_V0, cpu_V0);
4129 } else {
4130 gen_helper_neon_qshl_s64(cpu_V1, cpu_env,
4131 cpu_V1, cpu_V0);
4132 }
4133 break;
4134 case 10: /* VRSHL */
4135 if (u) {
4136 gen_helper_neon_rshl_u64(cpu_V0, cpu_V1, cpu_V0);
4137 } else {
4138 gen_helper_neon_rshl_s64(cpu_V0, cpu_V1, cpu_V0);
4139 }
4140 break;
4141 case 11: /* VQRSHL */
4142 if (u) {
4143 gen_helper_neon_qrshl_u64(cpu_V0, cpu_env,
4144 cpu_V1, cpu_V0);
4145 } else {
4146 gen_helper_neon_qrshl_s64(cpu_V0, cpu_env,
4147 cpu_V1, cpu_V0);
4148 }
4149 break;
4150 case 16:
4151 if (u) {
4152 tcg_gen_sub_i64(CPU_V001);
4153 } else {
4154 tcg_gen_add_i64(CPU_V001);
4155 }
4156 break;
4157 default:
4158 abort();
4159 }
4160 neon_store_reg64(cpu_V0, rd + pass);
4161 }
4162 return 0;
4163 }
4164 switch (op) {
4165 case 8: /* VSHL */
4166 case 9: /* VQSHL */
4167 case 10: /* VRSHL */
4168 case 11: /* VQRSHL */
4169 {
4170 int rtmp;
4171 /* Shift instruction operands are reversed. */
4172 rtmp = rn;
4173 rn = rm;
4174 rm = rtmp;
4175 pairwise = 0;
4176 }
4177 break;
4178 case 20: /* VPMAX */
4179 case 21: /* VPMIN */
4180 case 23: /* VPADD */
4181 pairwise = 1;
4182 break;
4183 case 26: /* VPADD (float) */
4184 pairwise = (u && size < 2);
4185 break;
4186 case 30: /* VPMIN/VPMAX (float) */
4187 pairwise = u;
4188 break;
4189 default:
4190 pairwise = 0;
4191 break;
4192 }
4193 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4194
4195 if (pairwise) {
4196 /* Pairwise. */
4197 if (q)
4198 n = (pass & 1) * 2;
4199 else
4200 n = 0;
4201 if (pass < q + 1) {
4202 NEON_GET_REG(T0, rn, n);
4203 NEON_GET_REG(T1, rn, n + 1);
4204 } else {
4205 NEON_GET_REG(T0, rm, n);
4206 NEON_GET_REG(T1, rm, n + 1);
4207 }
4208 } else {
4209 /* Elementwise. */
4210 NEON_GET_REG(T0, rn, pass);
4211 NEON_GET_REG(T1, rm, pass);
4212 }
4213 switch (op) {
4214 case 0: /* VHADD */
4215 GEN_NEON_INTEGER_OP(hadd);
4216 break;
4217 case 1: /* VQADD */
4218 GEN_NEON_INTEGER_OP_ENV(qadd);
4219 break;
4220 case 2: /* VRHADD */
4221 GEN_NEON_INTEGER_OP(rhadd);
4222 break;
4223 case 3: /* Logic ops. */
4224 switch ((u << 2) | size) {
4225 case 0: /* VAND */
4226 gen_op_andl_T0_T1();
4227 break;
4228 case 1: /* BIC */
4229 gen_op_bicl_T0_T1();
4230 break;
4231 case 2: /* VORR */
4232 gen_op_orl_T0_T1();
4233 break;
4234 case 3: /* VORN */
4235 gen_op_notl_T1();
4236 gen_op_orl_T0_T1();
4237 break;
4238 case 4: /* VEOR */
4239 gen_op_xorl_T0_T1();
4240 break;
4241 case 5: /* VBSL */
4242 tmp = neon_load_reg(rd, pass);
4243 gen_neon_bsl(cpu_T[0], cpu_T[0], cpu_T[1], tmp);
4244 dead_tmp(tmp);
4245 break;
4246 case 6: /* VBIT */
4247 tmp = neon_load_reg(rd, pass);
4248 gen_neon_bsl(cpu_T[0], cpu_T[0], tmp, cpu_T[1]);
4249 dead_tmp(tmp);
4250 break;
4251 case 7: /* VBIF */
4252 tmp = neon_load_reg(rd, pass);
4253 gen_neon_bsl(cpu_T[0], tmp, cpu_T[0], cpu_T[1]);
4254 dead_tmp(tmp);
4255 break;
4256 }
4257 break;
4258 case 4: /* VHSUB */
4259 GEN_NEON_INTEGER_OP(hsub);
4260 break;
4261 case 5: /* VQSUB */
4262 GEN_NEON_INTEGER_OP_ENV(qsub);
4263 break;
4264 case 6: /* VCGT */
4265 GEN_NEON_INTEGER_OP(cgt);
4266 break;
4267 case 7: /* VCGE */
4268 GEN_NEON_INTEGER_OP(cge);
4269 break;
4270 case 8: /* VSHL */
4271 GEN_NEON_INTEGER_OP(shl);
4272 break;
4273 case 9: /* VQSHL */
4274 GEN_NEON_INTEGER_OP_ENV(qshl);
4275 break;
4276 case 10: /* VRSHL */
4277 GEN_NEON_INTEGER_OP(rshl);
4278 break;
4279 case 11: /* VQRSHL */
4280 GEN_NEON_INTEGER_OP_ENV(qrshl);
4281 break;
4282 case 12: /* VMAX */
4283 GEN_NEON_INTEGER_OP(max);
4284 break;
4285 case 13: /* VMIN */
4286 GEN_NEON_INTEGER_OP(min);
4287 break;
4288 case 14: /* VABD */
4289 GEN_NEON_INTEGER_OP(abd);
4290 break;
4291 case 15: /* VABA */
4292 GEN_NEON_INTEGER_OP(abd);
4293 NEON_GET_REG(T1, rd, pass);
4294 gen_neon_add(size);
4295 break;
4296 case 16:
4297 if (!u) { /* VADD */
4298 if (gen_neon_add(size))
4299 return 1;
4300 } else { /* VSUB */
4301 switch (size) {
4302 case 0: gen_helper_neon_sub_u8(CPU_T001); break;
4303 case 1: gen_helper_neon_sub_u16(CPU_T001); break;
4304 case 2: gen_op_subl_T0_T1(); break;
4305 default: return 1;
4306 }
4307 }
4308 break;
4309 case 17:
4310 if (!u) { /* VTST */
4311 switch (size) {
4312 case 0: gen_helper_neon_tst_u8(CPU_T001); break;
4313 case 1: gen_helper_neon_tst_u16(CPU_T001); break;
4314 case 2: gen_helper_neon_tst_u32(CPU_T001); break;
4315 default: return 1;
4316 }
4317 } else { /* VCEQ */
4318 switch (size) {
4319 case 0: gen_helper_neon_ceq_u8(CPU_T001); break;
4320 case 1: gen_helper_neon_ceq_u16(CPU_T001); break;
4321 case 2: gen_helper_neon_ceq_u32(CPU_T001); break;
4322 default: return 1;
4323 }
4324 }
4325 break;
4326 case 18: /* Multiply. */
4327 switch (size) {
4328 case 0: gen_helper_neon_mul_u8(CPU_T001); break;
4329 case 1: gen_helper_neon_mul_u16(CPU_T001); break;
4330 case 2: gen_op_mul_T0_T1(); break;
4331 default: return 1;
4332 }
4333 NEON_GET_REG(T1, rd, pass);
4334 if (u) { /* VMLS */
4335 gen_neon_rsb(size);
4336 } else { /* VMLA */
4337 gen_neon_add(size);
4338 }
4339 break;
4340 case 19: /* VMUL */
4341 if (u) { /* polynomial */
4342 gen_helper_neon_mul_p8(CPU_T001);
4343 } else { /* Integer */
4344 switch (size) {
4345 case 0: gen_helper_neon_mul_u8(CPU_T001); break;
4346 case 1: gen_helper_neon_mul_u16(CPU_T001); break;
4347 case 2: gen_op_mul_T0_T1(); break;
4348 default: return 1;
4349 }
4350 }
4351 break;
4352 case 20: /* VPMAX */
4353 GEN_NEON_INTEGER_OP(pmax);
4354 break;
4355 case 21: /* VPMIN */
4356 GEN_NEON_INTEGER_OP(pmin);
4357 break;
4358 case 22: /* Hultiply high. */
4359 if (!u) { /* VQDMULH */
4360 switch (size) {
4361 case 1: gen_helper_neon_qdmulh_s16(CPU_T0E01); break;
4362 case 2: gen_helper_neon_qdmulh_s32(CPU_T0E01); break;
4363 default: return 1;
4364 }
4365 } else { /* VQRDHMUL */
4366 switch (size) {
4367 case 1: gen_helper_neon_qrdmulh_s16(CPU_T0E01); break;
4368 case 2: gen_helper_neon_qrdmulh_s32(CPU_T0E01); break;
4369 default: return 1;
4370 }
4371 }
4372 break;
4373 case 23: /* VPADD */
4374 if (u)
4375 return 1;
4376 switch (size) {
4377 case 0: gen_helper_neon_padd_u8(CPU_T001); break;
4378 case 1: gen_helper_neon_padd_u16(CPU_T001); break;
4379 case 2: gen_op_addl_T0_T1(); break;
4380 default: return 1;
4381 }
4382 break;
4383 case 26: /* Floating point arithnetic. */
4384 switch ((u << 2) | size) {
4385 case 0: /* VADD */
4386 gen_helper_neon_add_f32(CPU_T001);
4387 break;
4388 case 2: /* VSUB */
4389 gen_helper_neon_sub_f32(CPU_T001);
4390 break;
4391 case 4: /* VPADD */
4392 gen_helper_neon_add_f32(CPU_T001);
4393 break;
4394 case 6: /* VABD */
4395 gen_helper_neon_abd_f32(CPU_T001);
4396 break;
4397 default:
4398 return 1;
4399 }
4400 break;
4401 case 27: /* Float multiply. */
4402 gen_helper_neon_mul_f32(CPU_T001);
4403 if (!u) {
4404 NEON_GET_REG(T1, rd, pass);
4405 if (size == 0) {
4406 gen_helper_neon_add_f32(CPU_T001);
4407 } else {
4408 gen_helper_neon_sub_f32(cpu_T[0], cpu_T[1], cpu_T[0]);
4409 }
4410 }
4411 break;
4412 case 28: /* Float compare. */
4413 if (!u) {
4414 gen_helper_neon_ceq_f32(CPU_T001);
4415 } else {
4416 if (size == 0)
4417 gen_helper_neon_cge_f32(CPU_T001);
4418 else
4419 gen_helper_neon_cgt_f32(CPU_T001);
4420 }
4421 break;
4422 case 29: /* Float compare absolute. */
4423 if (!u)
4424 return 1;
4425 if (size == 0)
4426 gen_helper_neon_acge_f32(CPU_T001);
4427 else
4428 gen_helper_neon_acgt_f32(CPU_T001);
4429 break;
4430 case 30: /* Float min/max. */
4431 if (size == 0)
4432 gen_helper_neon_max_f32(CPU_T001);
4433 else
4434 gen_helper_neon_min_f32(CPU_T001);
4435 break;
4436 case 31:
4437 if (size == 0)
4438 gen_helper_recps_f32(cpu_T[0], cpu_T[0], cpu_T[1], cpu_env);
4439 else
4440 gen_helper_rsqrts_f32(cpu_T[0], cpu_T[0], cpu_T[1], cpu_env);
4441 break;
4442 default:
4443 abort();
4444 }
4445 /* Save the result. For elementwise operations we can put it
4446 straight into the destination register. For pairwise operations
4447 we have to be careful to avoid clobbering the source operands. */
4448 if (pairwise && rd == rm) {
4449 gen_neon_movl_scratch_T0(pass);
4450 } else {
4451 NEON_SET_REG(T0, rd, pass);
4452 }
4453
4454 } /* for pass */
4455 if (pairwise && rd == rm) {
4456 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4457 gen_neon_movl_T0_scratch(pass);
4458 NEON_SET_REG(T0, rd, pass);
4459 }
4460 }
4461 /* End of 3 register same size operations. */
4462 } else if (insn & (1 << 4)) {
4463 if ((insn & 0x00380080) != 0) {
4464 /* Two registers and shift. */
4465 op = (insn >> 8) & 0xf;
4466 if (insn & (1 << 7)) {
4467 /* 64-bit shift. */
4468 size = 3;
4469 } else {
4470 size = 2;
4471 while ((insn & (1 << (size + 19))) == 0)
4472 size--;
4473 }
4474 shift = (insn >> 16) & ((1 << (3 + size)) - 1);
4475 /* To avoid excessive dumplication of ops we implement shift
4476 by immediate using the variable shift operations. */
4477 if (op < 8) {
4478 /* Shift by immediate:
4479 VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
4480 /* Right shifts are encoded as N - shift, where N is the
4481 element size in bits. */
4482 if (op <= 4)
4483 shift = shift - (1 << (size + 3));
4484 if (size == 3) {
4485 count = q + 1;
4486 } else {
4487 count = q ? 4: 2;
4488 }
4489 switch (size) {
4490 case 0:
4491 imm = (uint8_t) shift;
4492 imm |= imm << 8;
4493 imm |= imm << 16;
4494 break;
4495 case 1:
4496 imm = (uint16_t) shift;
4497 imm |= imm << 16;
4498 break;
4499 case 2:
4500 case 3:
4501 imm = shift;
4502 break;
4503 default:
4504 abort();
4505 }
4506
4507 for (pass = 0; pass < count; pass++) {
4508 if (size == 3) {
4509 neon_load_reg64(cpu_V0, rm + pass);
4510 tcg_gen_movi_i64(cpu_V1, imm);
4511 switch (op) {
4512 case 0: /* VSHR */
4513 case 1: /* VSRA */
4514 if (u)
4515 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
4516 else
4517 gen_helper_neon_shl_s64(cpu_V0, cpu_V0, cpu_V1);
4518 break;
4519 case 2: /* VRSHR */
4520 case 3: /* VRSRA */
4521 if (u)
4522 gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, cpu_V1);
4523 else
4524 gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, cpu_V1);
4525 break;
4526 case 4: /* VSRI */
4527 if (!u)
4528 return 1;
4529 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
4530 break;
4531 case 5: /* VSHL, VSLI */
4532 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
4533 break;
4534 case 6: /* VQSHL */
4535 if (u)
4536 gen_helper_neon_qshl_u64(cpu_V0, cpu_env, cpu_V0, cpu_V1);
4537 else
4538 gen_helper_neon_qshl_s64(cpu_V0, cpu_env, cpu_V0, cpu_V1);
4539 break;
4540 case 7: /* VQSHLU */
4541 gen_helper_neon_qshl_u64(cpu_V0, cpu_env, cpu_V0, cpu_V1);
4542 break;
4543 }
4544 if (op == 1 || op == 3) {
4545 /* Accumulate. */
4546 neon_load_reg64(cpu_V0, rd + pass);
4547 tcg_gen_add_i64(cpu_V0, cpu_V0, cpu_V1);
4548 } else if (op == 4 || (op == 5 && u)) {
4549 /* Insert */
4550 cpu_abort(env, "VS[LR]I.64 not implemented");
4551 }
4552 neon_store_reg64(cpu_V0, rd + pass);
4553 } else { /* size < 3 */
4554 /* Operands in T0 and T1. */
4555 gen_op_movl_T1_im(imm);
4556 NEON_GET_REG(T0, rm, pass);
4557 switch (op) {
4558 case 0: /* VSHR */
4559 case 1: /* VSRA */
4560 GEN_NEON_INTEGER_OP(shl);
4561 break;
4562 case 2: /* VRSHR */
4563 case 3: /* VRSRA */
4564 GEN_NEON_INTEGER_OP(rshl);
4565 break;
4566 case 4: /* VSRI */
4567 if (!u)
4568 return 1;
4569 GEN_NEON_INTEGER_OP(shl);
4570 break;
4571 case 5: /* VSHL, VSLI */
4572 switch (size) {
4573 case 0: gen_helper_neon_shl_u8(CPU_T001); break;
4574 case 1: gen_helper_neon_shl_u16(CPU_T001); break;
4575 case 2: gen_helper_neon_shl_u32(CPU_T001); break;
4576 default: return 1;
4577 }
4578 break;
4579 case 6: /* VQSHL */
4580 GEN_NEON_INTEGER_OP_ENV(qshl);
4581 break;
4582 case 7: /* VQSHLU */
4583 switch (size) {
4584 case 0: gen_helper_neon_qshl_u8(CPU_T0E01); break;
4585 case 1: gen_helper_neon_qshl_u16(CPU_T0E01); break;
4586 case 2: gen_helper_neon_qshl_u32(CPU_T0E01); break;
4587 default: return 1;
4588 }
4589 break;
4590 }
4591
4592 if (op == 1 || op == 3) {
4593 /* Accumulate. */
4594 NEON_GET_REG(T1, rd, pass);
4595 gen_neon_add(size);
4596 } else if (op == 4 || (op == 5 && u)) {
4597 /* Insert */
4598 switch (size) {
4599 case 0:
4600 if (op == 4)
4601 imm = 0xff >> -shift;
4602 else
4603 imm = (uint8_t)(0xff << shift);
4604 imm |= imm << 8;
4605 imm |= imm << 16;
4606 break;
4607 case 1:
4608 if (op == 4)
4609 imm = 0xffff >> -shift;
4610 else
4611 imm = (uint16_t)(0xffff << shift);
4612 imm |= imm << 16;
4613 break;
4614 case 2:
4615 if (op == 4)
4616 imm = 0xffffffffu >> -shift;
4617 else
4618 imm = 0xffffffffu << shift;
4619 break;
4620 default:
4621 abort();
4622 }
4623 tmp = neon_load_reg(rd, pass);
4624 tcg_gen_andi_i32(cpu_T[0], cpu_T[0], imm);
4625 tcg_gen_andi_i32(tmp, tmp, ~imm);
4626 tcg_gen_or_i32(cpu_T[0], cpu_T[0], tmp);
4627 }
4628 NEON_SET_REG(T0, rd, pass);
4629 }
4630 } /* for pass */
4631 } else if (op < 10) {
4632 /* Shift by immediate and narrow:
4633 VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
4634 shift = shift - (1 << (size + 3));
4635 size++;
4636 switch (size) {
4637 case 1:
4638 imm = (uint16_t)shift;
4639 imm |= imm << 16;
4640 tmp2 = tcg_const_i32(imm);
4641 TCGV_UNUSED_I64(tmp64);
4642 break;
4643 case 2:
4644 imm = (uint32_t)shift;
4645 tmp2 = tcg_const_i32(imm);
4646 TCGV_UNUSED_I64(tmp64);
4647 case 3:
4648 tmp64 = tcg_const_i64(shift);
4649 TCGV_UNUSED(tmp2);
4650 break;
4651 default:
4652 abort();
4653 }
4654
4655 for (pass = 0; pass < 2; pass++) {
4656 if (size == 3) {
4657 neon_load_reg64(cpu_V0, rm + pass);
4658 if (q) {
4659 if (u)
4660 gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, tmp64);
4661 else
4662 gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, tmp64);
4663 } else {
4664 if (u)
4665 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, tmp64);
4666 else
4667 gen_helper_neon_shl_s64(cpu_V0, cpu_V0, tmp64);
4668 }
4669 } else {
4670 tmp = neon_load_reg(rm + pass, 0);
4671 gen_neon_shift_narrow(size, tmp, tmp2, q, u);
4672 tmp3 = neon_load_reg(rm + pass, 1);
4673 gen_neon_shift_narrow(size, tmp3, tmp2, q, u);
4674 tcg_gen_concat_i32_i64(cpu_V0, tmp, tmp3);
4675 dead_tmp(tmp);
4676 dead_tmp(tmp3);
4677 }
4678 tmp = new_tmp();
4679 if (op == 8 && !u) {
4680 gen_neon_narrow(size - 1, tmp, cpu_V0);
4681 } else {
4682 if (op == 8)
4683 gen_neon_narrow_sats(size - 1, tmp, cpu_V0);
4684 else
4685 gen_neon_narrow_satu(size - 1, tmp, cpu_V0);
4686 }
4687 if (pass == 0) {
4688 tmp2 = tmp;
4689 } else {
4690 neon_store_reg(rd, 0, tmp2);
4691 neon_store_reg(rd, 1, tmp);
4692 }
4693 } /* for pass */
4694 } else if (op == 10) {
4695 /* VSHLL */
4696 if (q || size == 3)
4697 return 1;
4698 tmp = neon_load_reg(rm, 0);
4699 tmp2 = neon_load_reg(rm, 1);
4700 for (pass = 0; pass < 2; pass++) {
4701 if (pass == 1)
4702 tmp = tmp2;
4703
4704 gen_neon_widen(cpu_V0, tmp, size, u);
4705
4706 if (shift != 0) {
4707 /* The shift is less than the width of the source
4708 type, so we can just shift the whole register. */
4709 tcg_gen_shli_i64(cpu_V0, cpu_V0, shift);
4710 if (size < 2 || !u) {
4711 uint64_t imm64;
4712 if (size == 0) {
4713 imm = (0xffu >> (8 - shift));
4714 imm |= imm << 16;
4715 } else {
4716 imm = 0xffff >> (16 - shift);
4717 }
4718 imm64 = imm | (((uint64_t)imm) << 32);
4719 tcg_gen_andi_i64(cpu_V0, cpu_V0, imm64);
4720 }
4721 }
4722 neon_store_reg64(cpu_V0, rd + pass);
4723 }
4724 } else if (op == 15 || op == 16) {
4725 /* VCVT fixed-point. */
4726 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4727 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, pass));
4728 if (op & 1) {
4729 if (u)
4730 gen_vfp_ulto(0, shift);
4731 else
4732 gen_vfp_slto(0, shift);
4733 } else {
4734 if (u)
4735 gen_vfp_toul(0, shift);
4736 else
4737 gen_vfp_tosl(0, shift);
4738 }
4739 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, pass));
4740 }
4741 } else {
4742 return 1;
4743 }
4744 } else { /* (insn & 0x00380080) == 0 */
4745 int invert;
4746
4747 op = (insn >> 8) & 0xf;
4748 /* One register and immediate. */
4749 imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf);
4750 invert = (insn & (1 << 5)) != 0;
4751 switch (op) {
4752 case 0: case 1:
4753 /* no-op */
4754 break;
4755 case 2: case 3:
4756 imm <<= 8;
4757 break;
4758 case 4: case 5:
4759 imm <<= 16;
4760 break;
4761 case 6: case 7:
4762 imm <<= 24;
4763 break;
4764 case 8: case 9:
4765 imm |= imm << 16;
4766 break;
4767 case 10: case 11:
4768 imm = (imm << 8) | (imm << 24);
4769 break;
4770 case 12:
4771 imm = (imm < 8) | 0xff;
4772 break;
4773 case 13:
4774 imm = (imm << 16) | 0xffff;
4775 break;
4776 case 14:
4777 imm |= (imm << 8) | (imm << 16) | (imm << 24);
4778 if (invert)
4779 imm = ~imm;
4780 break;
4781 case 15:
4782 imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19)
4783 | ((imm & 0x40) ? (0x1f << 25) : (1 << 30));
4784 break;
4785 }
4786 if (invert)
4787 imm = ~imm;
4788
4789 if (op != 14 || !invert)
4790 gen_op_movl_T1_im(imm);
4791
4792 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4793 if (op & 1 && op < 12) {
4794 tmp = neon_load_reg(rd, pass);
4795 if (invert) {
4796 /* The immediate value has already been inverted, so
4797 BIC becomes AND. */
4798 tcg_gen_andi_i32(tmp, tmp, imm);
4799 } else {
4800 tcg_gen_ori_i32(tmp, tmp, imm);
4801 }
4802 } else {
4803 /* VMOV, VMVN. */
4804 tmp = new_tmp();
4805 if (op == 14 && invert) {
4806 uint32_t val;
4807 val = 0;
4808 for (n = 0; n < 4; n++) {
4809 if (imm & (1 << (n + (pass & 1) * 4)))
4810 val |= 0xff << (n * 8);
4811 }
4812 tcg_gen_movi_i32(tmp, val);
4813 } else {
4814 tcg_gen_movi_i32(tmp, imm);
4815 }
4816 }
4817 neon_store_reg(rd, pass, tmp);
4818 }
4819 }
4820 } else { /* (insn & 0x00800010 == 0x00800000) */
4821 if (size != 3) {
4822 op = (insn >> 8) & 0xf;
4823 if ((insn & (1 << 6)) == 0) {
4824 /* Three registers of different lengths. */
4825 int src1_wide;
4826 int src2_wide;
4827 int prewiden;
4828 /* prewiden, src1_wide, src2_wide */
4829 static const int neon_3reg_wide[16][3] = {
4830 {1, 0, 0}, /* VADDL */
4831 {1, 1, 0}, /* VADDW */
4832 {1, 0, 0}, /* VSUBL */
4833 {1, 1, 0}, /* VSUBW */
4834 {0, 1, 1}, /* VADDHN */
4835 {0, 0, 0}, /* VABAL */
4836 {0, 1, 1}, /* VSUBHN */
4837 {0, 0, 0}, /* VABDL */
4838 {0, 0, 0}, /* VMLAL */
4839 {0, 0, 0}, /* VQDMLAL */
4840 {0, 0, 0}, /* VMLSL */
4841 {0, 0, 0}, /* VQDMLSL */
4842 {0, 0, 0}, /* Integer VMULL */
4843 {0, 0, 0}, /* VQDMULL */
4844 {0, 0, 0} /* Polynomial VMULL */
4845 };
4846
4847 prewiden = neon_3reg_wide[op][0];
4848 src1_wide = neon_3reg_wide[op][1];
4849 src2_wide = neon_3reg_wide[op][2];
4850
4851 if (size == 0 && (op == 9 || op == 11 || op == 13))
4852 return 1;
4853
4854 /* Avoid overlapping operands. Wide source operands are
4855 always aligned so will never overlap with wide
4856 destinations in problematic ways. */
4857 if (rd == rm && !src2_wide) {
4858 NEON_GET_REG(T0, rm, 1);
4859 gen_neon_movl_scratch_T0(2);
4860 } else if (rd == rn && !src1_wide) {
4861 NEON_GET_REG(T0, rn, 1);
4862 gen_neon_movl_scratch_T0(2);
4863 }
4864 TCGV_UNUSED(tmp3);
4865 for (pass = 0; pass < 2; pass++) {
4866 if (src1_wide) {
4867 neon_load_reg64(cpu_V0, rn + pass);
4868 TCGV_UNUSED(tmp);
4869 } else {
4870 if (pass == 1 && rd == rn) {
4871 gen_neon_movl_T0_scratch(2);
4872 tmp = new_tmp();
4873 tcg_gen_mov_i32(tmp, cpu_T[0]);
4874 } else {
4875 tmp = neon_load_reg(rn, pass);
4876 }
4877 if (prewiden) {
4878 gen_neon_widen(cpu_V0, tmp, size, u);
4879 }
4880 }
4881 if (src2_wide) {
4882 neon_load_reg64(cpu_V1, rm + pass);
4883 TCGV_UNUSED(tmp2);
4884 } else {
4885 if (pass == 1 && rd == rm) {
4886 gen_neon_movl_T0_scratch(2);
4887 tmp2 = new_tmp();
4888 tcg_gen_mov_i32(tmp2, cpu_T[0]);
4889 } else {
4890 tmp2 = neon_load_reg(rm, pass);
4891 }
4892 if (prewiden) {
4893 gen_neon_widen(cpu_V1, tmp2, size, u);
4894 }
4895 }
4896 switch (op) {
4897 case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
4898 gen_neon_addl(size);
4899 break;
4900 case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHL, VRSUBHL */
4901 gen_neon_subl(size);
4902 break;
4903 case 5: case 7: /* VABAL, VABDL */
4904 switch ((size << 1) | u) {
4905 case 0:
4906 gen_helper_neon_abdl_s16(cpu_V0, tmp, tmp2);
4907 break;
4908 case 1:
4909 gen_helper_neon_abdl_u16(cpu_V0, tmp, tmp2);
4910 break;
4911 case 2:
4912 gen_helper_neon_abdl_s32(cpu_V0, tmp, tmp2);
4913 break;
4914 case 3:
4915 gen_helper_neon_abdl_u32(cpu_V0, tmp, tmp2);
4916 break;
4917 case 4:
4918 gen_helper_neon_abdl_s64(cpu_V0, tmp, tmp2);
4919 break;
4920 case 5:
4921 gen_helper_neon_abdl_u64(cpu_V0, tmp, tmp2);
4922 break;
4923 default: abort();
4924 }
4925 dead_tmp(tmp2);
4926 dead_tmp(tmp);
4927 break;
4928 case 8: case 9: case 10: case 11: case 12: case 13:
4929 /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
4930 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
4931 break;
4932 case 14: /* Polynomial VMULL */
4933 cpu_abort(env, "Polynomial VMULL not implemented");
4934
4935 default: /* 15 is RESERVED. */
4936 return 1;
4937 }
4938 if (op == 5 || op == 13 || (op >= 8 && op <= 11)) {
4939 /* Accumulate. */
4940 if (op == 10 || op == 11) {
4941 gen_neon_negl(cpu_V0, size);
4942 }
4943
4944 if (op != 13) {
4945 neon_load_reg64(cpu_V1, rd + pass);
4946 }
4947
4948 switch (op) {
4949 case 5: case 8: case 10: /* VABAL, VMLAL, VMLSL */
4950 gen_neon_addl(size);
4951 break;
4952 case 9: case 11: /* VQDMLAL, VQDMLSL */
4953 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
4954 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
4955 break;
4956 /* Fall through. */
4957 case 13: /* VQDMULL */
4958 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
4959 break;
4960 default:
4961 abort();
4962 }
4963 neon_store_reg64(cpu_V0, rd + pass);
4964 } else if (op == 4 || op == 6) {
4965 /* Narrowing operation. */
4966 tmp = new_tmp();
4967 if (u) {
4968 switch (size) {
4969 case 0:
4970 gen_helper_neon_narrow_high_u8(tmp, cpu_V0);
4971 break;
4972 case 1:
4973 gen_helper_neon_narrow_high_u16(tmp, cpu_V0);
4974 break;
4975 case 2:
4976 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
4977 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
4978 break;
4979 default: abort();
4980 }
4981 } else {
4982 switch (size) {
4983 case 0:
4984 gen_helper_neon_narrow_round_high_u8(tmp, cpu_V0);
4985 break;
4986 case 1:
4987 gen_helper_neon_narrow_round_high_u16(tmp, cpu_V0);
4988 break;
4989 case 2:
4990 tcg_gen_addi_i64(cpu_V0, cpu_V0, 1u << 31);
4991 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
4992 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
4993 break;
4994 default: abort();
4995 }
4996 }
4997 if (pass == 0) {
4998 tmp3 = tmp;
4999 } else {
5000 neon_store_reg(rd, 0, tmp3);
5001 neon_store_reg(rd, 1, tmp);
5002 }
5003 } else {
5004 /* Write back the result. */
5005 neon_store_reg64(cpu_V0, rd + pass);
5006 }
5007 }
5008 } else {
5009 /* Two registers and a scalar. */
5010 switch (op) {
5011 case 0: /* Integer VMLA scalar */
5012 case 1: /* Float VMLA scalar */
5013 case 4: /* Integer VMLS scalar */
5014 case 5: /* Floating point VMLS scalar */
5015 case 8: /* Integer VMUL scalar */
5016 case 9: /* Floating point VMUL scalar */
5017 case 12: /* VQDMULH scalar */
5018 case 13: /* VQRDMULH scalar */
5019 gen_neon_get_scalar(size, rm);
5020 gen_neon_movl_scratch_T0(0);
5021 for (pass = 0; pass < (u ? 4 : 2); pass++) {
5022 if (pass != 0)
5023 gen_neon_movl_T0_scratch(0);
5024 NEON_GET_REG(T1, rn, pass);
5025 if (op == 12) {
5026 if (size == 1) {
5027 gen_helper_neon_qdmulh_s16(CPU_T0E01);
5028 } else {
5029 gen_helper_neon_qdmulh_s32(CPU_T0E01);
5030 }
5031 } else if (op == 13) {
5032 if (size == 1) {
5033 gen_helper_neon_qrdmulh_s16(CPU_T0E01);
5034 } else {
5035 gen_helper_neon_qrdmulh_s32(CPU_T0E01);
5036 }
5037 } else if (op & 1) {
5038 gen_helper_neon_mul_f32(CPU_T001);
5039 } else {
5040 switch (size) {
5041 case 0: gen_helper_neon_mul_u8(CPU_T001); break;
5042 case 1: gen_helper_neon_mul_u16(CPU_T001); break;
5043 case 2: gen_op_mul_T0_T1(); break;
5044 default: return 1;
5045 }
5046 }
5047 if (op < 8) {
5048 /* Accumulate. */
5049 NEON_GET_REG(T1, rd, pass);
5050 switch (op) {
5051 case 0:
5052 gen_neon_add(size);
5053 break;
5054 case 1:
5055 gen_helper_neon_add_f32(CPU_T001);
5056 break;
5057 case 4:
5058 gen_neon_rsb(size);
5059 break;
5060 case 5:
5061 gen_helper_neon_sub_f32(cpu_T[0], cpu_T[1], cpu_T[0]);
5062 break;
5063 default:
5064 abort();
5065 }
5066 }
5067 NEON_SET_REG(T0, rd, pass);
5068 }
5069 break;
5070 case 2: /* VMLAL sclar */
5071 case 3: /* VQDMLAL scalar */
5072 case 6: /* VMLSL scalar */
5073 case 7: /* VQDMLSL scalar */
5074 case 10: /* VMULL scalar */
5075 case 11: /* VQDMULL scalar */
5076 if (size == 0 && (op == 3 || op == 7 || op == 11))
5077 return 1;
5078
5079 gen_neon_get_scalar(size, rm);
5080 NEON_GET_REG(T1, rn, 1);
5081
5082 for (pass = 0; pass < 2; pass++) {
5083 if (pass == 0) {
5084 tmp = neon_load_reg(rn, 0);
5085 } else {
5086 tmp = new_tmp();
5087 tcg_gen_mov_i32(tmp, cpu_T[1]);
5088 }
5089 tmp2 = new_tmp();
5090 tcg_gen_mov_i32(tmp2, cpu_T[0]);
5091 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
5092 if (op == 6 || op == 7) {
5093 gen_neon_negl(cpu_V0, size);
5094 }
5095 if (op != 11) {
5096 neon_load_reg64(cpu_V1, rd + pass);
5097 }
5098 switch (op) {
5099 case 2: case 6:
5100 gen_neon_addl(size);
5101 break;
5102 case 3: case 7:
5103 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5104 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
5105 break;
5106 case 10:
5107 /* no-op */
5108 break;
5109 case 11:
5110 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5111 break;
5112 default:
5113 abort();
5114 }
5115 neon_store_reg64(cpu_V0, rd + pass);
5116 }
5117 break;
5118 default: /* 14 and 15 are RESERVED */
5119 return 1;
5120 }
5121 }
5122 } else { /* size == 3 */
5123 if (!u) {
5124 /* Extract. */
5125 imm = (insn >> 8) & 0xf;
5126 count = q + 1;
5127
5128 if (imm > 7 && !q)
5129 return 1;
5130
5131 if (imm == 0) {
5132 neon_load_reg64(cpu_V0, rn);
5133 if (q) {
5134 neon_load_reg64(cpu_V1, rn + 1);
5135 }
5136 } else if (imm == 8) {
5137 neon_load_reg64(cpu_V0, rn + 1);
5138 if (q) {
5139 neon_load_reg64(cpu_V1, rm);
5140 }
5141 } else if (q) {
5142 tmp64 = tcg_temp_new_i64();
5143 if (imm < 8) {
5144 neon_load_reg64(cpu_V0, rn);
5145 neon_load_reg64(tmp64, rn + 1);
5146 } else {
5147 neon_load_reg64(cpu_V0, rn + 1);
5148 neon_load_reg64(tmp64, rm);
5149 }
5150 tcg_gen_shri_i64(cpu_V0, cpu_V0, (imm & 7) * 8);
5151 tcg_gen_shli_i64(cpu_V1, tmp64, 64 - ((imm & 7) * 8));
5152 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5153 if (imm < 8) {
5154 neon_load_reg64(cpu_V1, rm);
5155 } else {
5156 neon_load_reg64(cpu_V1, rm + 1);
5157 imm -= 8;
5158 }
5159 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
5160 tcg_gen_shri_i64(tmp64, tmp64, imm * 8);
5161 tcg_gen_or_i64(cpu_V1, cpu_V1, tmp64);
5162 } else {
5163 /* BUGFIX */
5164 neon_load_reg64(cpu_V0, rn);
5165 tcg_gen_shri_i64(cpu_V0, cpu_V0, imm * 8);
5166 neon_load_reg64(cpu_V1, rm);
5167 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
5168 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5169 }
5170 neon_store_reg64(cpu_V0, rd);
5171 if (q) {
5172 neon_store_reg64(cpu_V1, rd + 1);
5173 }
5174 } else if ((insn & (1 << 11)) == 0) {
5175 /* Two register misc. */
5176 op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf);
5177 size = (insn >> 18) & 3;
5178 switch (op) {
5179 case 0: /* VREV64 */
5180 if (size == 3)
5181 return 1;
5182 for (pass = 0; pass < (q ? 2 : 1); pass++) {
5183 NEON_GET_REG(T0, rm, pass * 2);
5184 NEON_GET_REG(T1, rm, pass * 2 + 1);
5185 switch (size) {
5186 case 0: tcg_gen_bswap_i32(cpu_T[0], cpu_T[0]); break;
5187 case 1: gen_swap_half(cpu_T[0]); break;
5188 case 2: /* no-op */ break;
5189 default: abort();
5190 }
5191 NEON_SET_REG(T0, rd, pass * 2 + 1);
5192 if (size == 2) {
5193 NEON_SET_REG(T1, rd, pass * 2);
5194 } else {
5195 gen_op_movl_T0_T1();
5196 switch (size) {
5197 case 0: tcg_gen_bswap_i32(cpu_T[0], cpu_T[0]); break;
5198 case 1: gen_swap_half(cpu_T[0]); break;
5199 default: abort();
5200 }
5201 NEON_SET_REG(T0, rd, pass * 2);
5202 }
5203 }
5204 break;
5205 case 4: case 5: /* VPADDL */
5206 case 12: case 13: /* VPADAL */
5207 if (size == 3)
5208 return 1;
5209 for (pass = 0; pass < q + 1; pass++) {
5210 tmp = neon_load_reg(rm, pass * 2);
5211 gen_neon_widen(cpu_V0, tmp, size, op & 1);
5212 tmp = neon_load_reg(rm, pass * 2 + 1);
5213 gen_neon_widen(cpu_V1, tmp, size, op & 1);
5214 switch (size) {
5215 case 0: gen_helper_neon_paddl_u16(CPU_V001); break;
5216 case 1: gen_helper_neon_paddl_u32(CPU_V001); break;
5217 case 2: tcg_gen_add_i64(CPU_V001); break;
5218 default: abort();
5219 }
5220 if (op >= 12) {
5221 /* Accumulate. */
5222 neon_load_reg64(cpu_V1, rd + pass);
5223 gen_neon_addl(size);
5224 }
5225 neon_store_reg64(cpu_V0, rd + pass);
5226 }
5227 break;
5228 case 33: /* VTRN */
5229 if (size == 2) {
5230 for (n = 0; n < (q ? 4 : 2); n += 2) {
5231 NEON_GET_REG(T0, rm, n);
5232 NEON_GET_REG(T1, rd, n + 1);
5233 NEON_SET_REG(T1, rm, n);
5234 NEON_SET_REG(T0, rd, n + 1);
5235 }
5236 } else {
5237 goto elementwise;
5238 }
5239 break;
5240 case 34: /* VUZP */
5241 /* Reg Before After
5242 Rd A3 A2 A1 A0 B2 B0 A2 A0
5243 Rm B3 B2 B1 B0 B3 B1 A3 A1
5244 */
5245 if (size == 3)
5246 return 1;
5247 gen_neon_unzip(rd, q, 0, size);
5248 gen_neon_unzip(rm, q, 4, size);
5249 if (q) {
5250 static int unzip_order_q[8] =
5251 {0, 2, 4, 6, 1, 3, 5, 7};
5252 for (n = 0; n < 8; n++) {
5253 int reg = (n < 4) ? rd : rm;
5254 gen_neon_movl_T0_scratch(unzip_order_q[n]);
5255 NEON_SET_REG(T0, reg, n % 4);
5256 }
5257 } else {
5258 static int unzip_order[4] =
5259 {0, 4, 1, 5};
5260 for (n = 0; n < 4; n++) {
5261 int reg = (n < 2) ? rd : rm;
5262 gen_neon_movl_T0_scratch(unzip_order[n]);
5263 NEON_SET_REG(T0, reg, n % 2);
5264 }
5265 }
5266 break;
5267 case 35: /* VZIP */
5268 /* Reg Before After
5269 Rd A3 A2 A1 A0 B1 A1 B0 A0
5270 Rm B3 B2 B1 B0 B3 A3 B2 A2
5271 */
5272 if (size == 3)
5273 return 1;
5274 count = (q ? 4 : 2);
5275 for (n = 0; n < count; n++) {
5276 NEON_GET_REG(T0, rd, n);
5277 NEON_GET_REG(T1, rd, n);
5278 switch (size) {
5279 case 0: gen_helper_neon_zip_u8(); break;
5280 case 1: gen_helper_neon_zip_u16(); break;
5281 case 2: /* no-op */; break;
5282 default: abort();
5283 }
5284 gen_neon_movl_scratch_T0(n * 2);
5285 gen_neon_movl_scratch_T1(n * 2 + 1);
5286 }
5287 for (n = 0; n < count * 2; n++) {
5288 int reg = (n < count) ? rd : rm;
5289 gen_neon_movl_T0_scratch(n);
5290 NEON_SET_REG(T0, reg, n % count);
5291 }
5292 break;
5293 case 36: case 37: /* VMOVN, VQMOVUN, VQMOVN */
5294 if (size == 3)
5295 return 1;
5296 TCGV_UNUSED(tmp2);
5297 for (pass = 0; pass < 2; pass++) {
5298 neon_load_reg64(cpu_V0, rm + pass);
5299 tmp = new_tmp();
5300 if (op == 36 && q == 0) {
5301 gen_neon_narrow(size, tmp, cpu_V0);
5302 } else if (q) {
5303 gen_neon_narrow_satu(size, tmp, cpu_V0);
5304 } else {
5305 gen_neon_narrow_sats(size, tmp, cpu_V0);
5306 }
5307 if (pass == 0) {
5308 tmp2 = tmp;
5309 } else {
5310 neon_store_reg(rd, 0, tmp2);
5311 neon_store_reg(rd, 1, tmp);
5312 }
5313 }
5314 break;
5315 case 38: /* VSHLL */
5316 if (q || size == 3)
5317 return 1;
5318 tmp = neon_load_reg(rm, 0);
5319 tmp2 = neon_load_reg(rm, 1);
5320 for (pass = 0; pass < 2; pass++) {
5321 if (pass == 1)
5322 tmp = tmp2;
5323 gen_neon_widen(cpu_V0, tmp, size, 1);
5324 neon_store_reg64(cpu_V0, rd + pass);
5325 }
5326 break;
5327 default:
5328 elementwise:
5329 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5330 if (op == 30 || op == 31 || op >= 58) {
5331 tcg_gen_ld_f32(cpu_F0s, cpu_env,
5332 neon_reg_offset(rm, pass));
5333 } else {
5334 NEON_GET_REG(T0, rm, pass);
5335 }
5336 switch (op) {
5337 case 1: /* VREV32 */
5338 switch (size) {
5339 case 0: tcg_gen_bswap_i32(cpu_T[0], cpu_T[0]); break;
5340 case 1: gen_swap_half(cpu_T[0]); break;
5341 default: return 1;
5342 }
5343 break;
5344 case 2: /* VREV16 */
5345 if (size != 0)
5346 return 1;
5347 gen_rev16(cpu_T[0]);
5348 break;
5349 case 8: /* CLS */
5350 switch (size) {
5351 case 0: gen_helper_neon_cls_s8(cpu_T[0], cpu_T[0]); break;
5352 case 1: gen_helper_neon_cls_s16(cpu_T[0], cpu_T[0]); break;
5353 case 2: gen_helper_neon_cls_s32(cpu_T[0], cpu_T[0]); break;
5354 default: return 1;
5355 }
5356 break;
5357 case 9: /* CLZ */
5358 switch (size) {
5359 case 0: gen_helper_neon_clz_u8(cpu_T[0], cpu_T[0]); break;
5360 case 1: gen_helper_neon_clz_u16(cpu_T[0], cpu_T[0]); break;
5361 case 2: gen_helper_clz(cpu_T[0], cpu_T[0]); break;
5362 default: return 1;
5363 }
5364 break;
5365 case 10: /* CNT */
5366 if (size != 0)
5367 return 1;
5368 gen_helper_neon_cnt_u8(cpu_T[0], cpu_T[0]);
5369 break;
5370 case 11: /* VNOT */
5371 if (size != 0)
5372 return 1;
5373 gen_op_notl_T0();
5374 break;
5375 case 14: /* VQABS */
5376 switch (size) {
5377 case 0: gen_helper_neon_qabs_s8(cpu_T[0], cpu_env, cpu_T[0]); break;
5378 case 1: gen_helper_neon_qabs_s16(cpu_T[0], cpu_env, cpu_T[0]); break;
5379 case 2: gen_helper_neon_qabs_s32(cpu_T[0], cpu_env, cpu_T[0]); break;
5380 default: return 1;
5381 }
5382 break;
5383 case 15: /* VQNEG */
5384 switch (size) {
5385 case 0: gen_helper_neon_qneg_s8(cpu_T[0], cpu_env, cpu_T[0]); break;
5386 case 1: gen_helper_neon_qneg_s16(cpu_T[0], cpu_env, cpu_T[0]); break;
5387 case 2: gen_helper_neon_qneg_s32(cpu_T[0], cpu_env, cpu_T[0]); break;
5388 default: return 1;
5389 }
5390 break;
5391 case 16: case 19: /* VCGT #0, VCLE #0 */
5392 gen_op_movl_T1_im(0);
5393 switch(size) {
5394 case 0: gen_helper_neon_cgt_s8(CPU_T001); break;
5395 case 1: gen_helper_neon_cgt_s16(CPU_T001); break;
5396 case 2: gen_helper_neon_cgt_s32(CPU_T001); break;
5397 default: return 1;
5398 }
5399 if (op == 19)
5400 gen_op_notl_T0();
5401 break;
5402 case 17: case 20: /* VCGE #0, VCLT #0 */
5403 gen_op_movl_T1_im(0);
5404 switch(size) {
5405 case 0: gen_helper_neon_cge_s8(CPU_T001); break;
5406 case 1: gen_helper_neon_cge_s16(CPU_T001); break;
5407 case 2: gen_helper_neon_cge_s32(CPU_T001); break;
5408 default: return 1;
5409 }
5410 if (op == 20)
5411 gen_op_notl_T0();
5412 break;
5413 case 18: /* VCEQ #0 */
5414 gen_op_movl_T1_im(0);
5415 switch(size) {
5416 case 0: gen_helper_neon_ceq_u8(CPU_T001); break;
5417 case 1: gen_helper_neon_ceq_u16(CPU_T001); break;
5418 case 2: gen_helper_neon_ceq_u32(CPU_T001); break;
5419 default: return 1;
5420 }
5421 break;
5422 case 22: /* VABS */
5423 switch(size) {
5424 case 0: gen_helper_neon_abs_s8(cpu_T[0], cpu_T[0]); break;
5425 case 1: gen_helper_neon_abs_s16(cpu_T[0], cpu_T[0]); break;
5426 case 2: tcg_gen_abs_i32(cpu_T[0], cpu_T[0]); break;
5427 default: return 1;
5428 }
5429 break;
5430 case 23: /* VNEG */
5431 gen_op_movl_T1_im(0);
5432 if (size == 3)
5433 return 1;
5434 gen_neon_rsb(size);
5435 break;
5436 case 24: case 27: /* Float VCGT #0, Float VCLE #0 */
5437 gen_op_movl_T1_im(0);
5438 gen_helper_neon_cgt_f32(CPU_T001);
5439 if (op == 27)
5440 gen_op_notl_T0();
5441 break;
5442 case 25: case 28: /* Float VCGE #0, Float VCLT #0 */
5443 gen_op_movl_T1_im(0);
5444 gen_helper_neon_cge_f32(CPU_T001);
5445 if (op == 28)
5446 gen_op_notl_T0();
5447 break;
5448 case 26: /* Float VCEQ #0 */
5449 gen_op_movl_T1_im(0);
5450 gen_helper_neon_ceq_f32(CPU_T001);
5451 break;
5452 case 30: /* Float VABS */
5453 gen_vfp_abs(0);
5454 break;
5455 case 31: /* Float VNEG */
5456 gen_vfp_neg(0);
5457 break;
5458 case 32: /* VSWP */
5459 NEON_GET_REG(T1, rd, pass);
5460 NEON_SET_REG(T1, rm, pass);
5461 break;
5462 case 33: /* VTRN */
5463 NEON_GET_REG(T1, rd, pass);
5464 switch (size) {
5465 case 0: gen_helper_neon_trn_u8(); break;
5466 case 1: gen_helper_neon_trn_u16(); break;
5467 case 2: abort();
5468 default: return 1;
5469 }
5470 NEON_SET_REG(T1, rm, pass);
5471 break;
5472 case 56: /* Integer VRECPE */
5473 gen_helper_recpe_u32(cpu_T[0], cpu_T[0], cpu_env);
5474 break;
5475 case 57: /* Integer VRSQRTE */
5476 gen_helper_rsqrte_u32(cpu_T[0], cpu_T[0], cpu_env);
5477 break;
5478 case 58: /* Float VRECPE */
5479 gen_helper_recpe_f32(cpu_F0s, cpu_F0s, cpu_env);
5480 break;
5481 case 59: /* Float VRSQRTE */
5482 gen_helper_rsqrte_f32(cpu_F0s, cpu_F0s, cpu_env);
5483 break;
5484 case 60: /* VCVT.F32.S32 */
5485 gen_vfp_tosiz(0);
5486 break;
5487 case 61: /* VCVT.F32.U32 */
5488 gen_vfp_touiz(0);
5489 break;
5490 case 62: /* VCVT.S32.F32 */
5491 gen_vfp_sito(0);
5492 break;
5493 case 63: /* VCVT.U32.F32 */
5494 gen_vfp_uito(0);
5495 break;
5496 default:
5497 /* Reserved: 21, 29, 39-56 */
5498 return 1;
5499 }
5500 if (op == 30 || op == 31 || op >= 58) {
5501 tcg_gen_st_f32(cpu_F0s, cpu_env,
5502 neon_reg_offset(rd, pass));
5503 } else {
5504 NEON_SET_REG(T0, rd, pass);
5505 }
5506 }
5507 break;
5508 }
5509 } else if ((insn & (1 << 10)) == 0) {
5510 /* VTBL, VTBX. */
5511 n = ((insn >> 5) & 0x18) + 8;
5512 if (insn & (1 << 6)) {
5513 tmp = neon_load_reg(rd, 0);
5514 } else {
5515 tmp = new_tmp();
5516 tcg_gen_movi_i32(tmp, 0);
5517 }
5518 tmp2 = neon_load_reg(rm, 0);
5519 gen_helper_neon_tbl(tmp2, tmp2, tmp, tcg_const_i32(rn),
5520 tcg_const_i32(n));
5521 dead_tmp(tmp);
5522 if (insn & (1 << 6)) {
5523 tmp = neon_load_reg(rd, 1);
5524 } else {
5525 tmp = new_tmp();
5526 tcg_gen_movi_i32(tmp, 0);
5527 }
5528 tmp3 = neon_load_reg(rm, 1);
5529 gen_helper_neon_tbl(tmp3, tmp3, tmp, tcg_const_i32(rn),
5530 tcg_const_i32(n));
5531 neon_store_reg(rd, 0, tmp2);
5532 neon_store_reg(rd, 1, tmp3);
5533 dead_tmp(tmp);
5534 } else if ((insn & 0x380) == 0) {
5535 /* VDUP */
5536 if (insn & (1 << 19)) {
5537 NEON_SET_REG(T0, rm, 1);
5538 } else {
5539 NEON_SET_REG(T0, rm, 0);
5540 }
5541 if (insn & (1 << 16)) {
5542 gen_neon_dup_u8(cpu_T[0], ((insn >> 17) & 3) * 8);
5543 } else if (insn & (1 << 17)) {
5544 if ((insn >> 18) & 1)
5545 gen_neon_dup_high16(cpu_T[0]);
5546 else
5547 gen_neon_dup_low16(cpu_T[0]);
5548 }
5549 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5550 NEON_SET_REG(T0, rd, pass);
5551 }
5552 } else {
5553 return 1;
5554 }
5555 }
5556 }
5557 return 0;
5558 }
5559
5560 static int disas_coproc_insn(CPUState * env, DisasContext *s, uint32_t insn)
5561 {
5562 int cpnum;
5563
5564 cpnum = (insn >> 8) & 0xf;
5565 if (arm_feature(env, ARM_FEATURE_XSCALE)
5566 && ((env->cp15.c15_cpar ^ 0x3fff) & (1 << cpnum)))
5567 return 1;
5568
5569 switch (cpnum) {
5570 case 0:
5571 case 1:
5572 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
5573 return disas_iwmmxt_insn(env, s, insn);
5574 } else if (arm_feature(env, ARM_FEATURE_XSCALE)) {
5575 return disas_dsp_insn(env, s, insn);
5576 }
5577 return 1;
5578 case 10:
5579 case 11:
5580 return disas_vfp_insn (env, s, insn);
5581 case 15:
5582 return disas_cp15_insn (env, s, insn);
5583 default:
5584 /* Unknown coprocessor. See if the board has hooked it. */
5585 return disas_cp_insn (env, s, insn);
5586 }
5587 }
5588
5589
5590 /* Store a 64-bit value to a register pair. Clobbers val. */
5591 static void gen_storeq_reg(DisasContext *s, int rlow, int rhigh, TCGv_i64 val)
5592 {
5593 TCGv tmp;
5594 tmp = new_tmp();
5595 tcg_gen_trunc_i64_i32(tmp, val);
5596 store_reg(s, rlow, tmp);
5597 tmp = new_tmp();
5598 tcg_gen_shri_i64(val, val, 32);
5599 tcg_gen_trunc_i64_i32(tmp, val);
5600 store_reg(s, rhigh, tmp);
5601 }
5602
5603 /* load a 32-bit value from a register and perform a 64-bit accumulate. */
5604 static void gen_addq_lo(DisasContext *s, TCGv_i64 val, int rlow)
5605 {
5606 TCGv_i64 tmp;
5607 TCGv tmp2;
5608
5609 /* Load value and extend to 64 bits. */
5610 tmp = tcg_temp_new_i64();
5611 tmp2 = load_reg(s, rlow);
5612 tcg_gen_extu_i32_i64(tmp, tmp2);
5613 dead_tmp(tmp2);
5614 tcg_gen_add_i64(val, val, tmp);
5615 }
5616
5617 /* load and add a 64-bit value from a register pair. */
5618 static void gen_addq(DisasContext *s, TCGv_i64 val, int rlow, int rhigh)
5619 {
5620 TCGv_i64 tmp;
5621 TCGv tmpl;
5622 TCGv tmph;
5623
5624 /* Load 64-bit value rd:rn. */
5625 tmpl = load_reg(s, rlow);
5626 tmph = load_reg(s, rhigh);
5627 tmp = tcg_temp_new_i64();
5628 tcg_gen_concat_i32_i64(tmp, tmpl, tmph);
5629 dead_tmp(tmpl);
5630 dead_tmp(tmph);
5631 tcg_gen_add_i64(val, val, tmp);
5632 }
5633
5634 /* Set N and Z flags from a 64-bit value. */
5635 static void gen_logicq_cc(TCGv_i64 val)
5636 {
5637 TCGv tmp = new_tmp();
5638 gen_helper_logicq_cc(tmp, val);
5639 gen_logic_CC(tmp);
5640 dead_tmp(tmp);
5641 }
5642
5643 static void disas_arm_insn(CPUState * env, DisasContext *s)
5644 {
5645 unsigned int cond, insn, val, op1, i, shift, rm, rs, rn, rd, sh;
5646 TCGv tmp;
5647 TCGv tmp2;
5648 TCGv tmp3;
5649 TCGv addr;
5650 TCGv_i64 tmp64;
5651
5652 insn = ldl_code(s->pc);
5653 s->pc += 4;
5654
5655 /* M variants do not implement ARM mode. */
5656 if (IS_M(env))
5657 goto illegal_op;
5658 cond = insn >> 28;
5659 if (cond == 0xf){
5660 /* Unconditional instructions. */
5661 if (((insn >> 25) & 7) == 1) {
5662 /* NEON Data processing. */
5663 if (!arm_feature(env, ARM_FEATURE_NEON))
5664 goto illegal_op;
5665
5666 if (disas_neon_data_insn(env, s, insn))
5667 goto illegal_op;
5668 return;
5669 }
5670 if ((insn & 0x0f100000) == 0x04000000) {
5671 /* NEON load/store. */
5672 if (!arm_feature(env, ARM_FEATURE_NEON))
5673 goto illegal_op;
5674
5675 if (disas_neon_ls_insn(env, s, insn))
5676 goto illegal_op;
5677 return;
5678 }
5679 if ((insn & 0x0d70f000) == 0x0550f000)
5680 return; /* PLD */
5681 else if ((insn & 0x0ffffdff) == 0x01010000) {
5682 ARCH(6);
5683 /* setend */
5684 if (insn & (1 << 9)) {
5685 /* BE8 mode not implemented. */
5686 goto illegal_op;
5687 }
5688 return;
5689 } else if ((insn & 0x0fffff00) == 0x057ff000) {
5690 switch ((insn >> 4) & 0xf) {
5691 case 1: /* clrex */
5692 ARCH(6K);
5693 gen_helper_clrex(cpu_env);
5694 return;
5695 case 4: /* dsb */
5696 case 5: /* dmb */
5697 case 6: /* isb */
5698 ARCH(7);
5699 /* We don't emulate caches so these are a no-op. */
5700 return;
5701 default:
5702 goto illegal_op;
5703 }
5704 } else if ((insn & 0x0e5fffe0) == 0x084d0500) {
5705 /* srs */
5706 uint32_t offset;
5707 if (IS_USER(s))
5708 goto illegal_op;
5709 ARCH(6);
5710 op1 = (insn & 0x1f);
5711 if (op1 == (env->uncached_cpsr & CPSR_M)) {
5712 addr = load_reg(s, 13);
5713 } else {
5714 addr = new_tmp();
5715 gen_helper_get_r13_banked(addr, cpu_env, tcg_const_i32(op1));
5716 }
5717 i = (insn >> 23) & 3;
5718 switch (i) {
5719 case 0: offset = -4; break; /* DA */
5720 case 1: offset = -8; break; /* DB */
5721 case 2: offset = 0; break; /* IA */
5722 case 3: offset = 4; break; /* IB */
5723 default: abort();
5724 }
5725 if (offset)
5726 tcg_gen_addi_i32(addr, addr, offset);
5727 tmp = load_reg(s, 14);
5728 gen_st32(tmp, addr, 0);
5729 tmp = new_tmp();
5730 gen_helper_cpsr_read(tmp);
5731 tcg_gen_addi_i32(addr, addr, 4);
5732 gen_st32(tmp, addr, 0);
5733 if (insn & (1 << 21)) {
5734 /* Base writeback. */
5735 switch (i) {
5736 case 0: offset = -8; break;
5737 case 1: offset = -4; break;
5738 case 2: offset = 4; break;
5739 case 3: offset = 0; break;
5740 default: abort();
5741 }
5742 if (offset)
5743 tcg_gen_addi_i32(addr, tmp, offset);
5744 if (op1 == (env->uncached_cpsr & CPSR_M)) {
5745 gen_movl_reg_T1(s, 13);
5746 } else {
5747 gen_helper_set_r13_banked(cpu_env, tcg_const_i32(op1), cpu_T[1]);
5748 }
5749 } else {
5750 dead_tmp(addr);
5751 }
5752 } else if ((insn & 0x0e5fffe0) == 0x081d0a00) {
5753 /* rfe */
5754 uint32_t offset;
5755 if (IS_USER(s))
5756 goto illegal_op;
5757 ARCH(6);
5758 rn = (insn >> 16) & 0xf;
5759 addr = load_reg(s, rn);
5760 i = (insn >> 23) & 3;
5761 switch (i) {
5762 case 0: offset = -4; break; /* DA */
5763 case 1: offset = -8; break; /* DB */
5764 case 2: offset = 0; break; /* IA */
5765 case 3: offset = 4; break; /* IB */
5766 default: abort();
5767 }
5768 if (offset)
5769 tcg_gen_addi_i32(addr, addr, offset);
5770 /* Load PC into tmp and CPSR into tmp2. */
5771 tmp = gen_ld32(addr, 0);
5772 tcg_gen_addi_i32(addr, addr, 4);
5773 tmp2 = gen_ld32(addr, 0);
5774 if (insn & (1 << 21)) {
5775 /* Base writeback. */
5776 switch (i) {
5777 case 0: offset = -8; break;
5778 case 1: offset = -4; break;
5779 case 2: offset = 4; break;
5780 case 3: offset = 0; break;
5781 default: abort();
5782 }
5783 if (offset)
5784 tcg_gen_addi_i32(addr, addr, offset);
5785 store_reg(s, rn, addr);
5786 } else {
5787 dead_tmp(addr);
5788 }
5789 gen_rfe(s, tmp, tmp2);
5790 } else if ((insn & 0x0e000000) == 0x0a000000) {
5791 /* branch link and change to thumb (blx <offset>) */
5792 int32_t offset;
5793
5794 val = (uint32_t)s->pc;
5795 tmp = new_tmp();
5796 tcg_gen_movi_i32(tmp, val);
5797 store_reg(s, 14, tmp);
5798 /* Sign-extend the 24-bit offset */
5799 offset = (((int32_t)insn) << 8) >> 8;
5800 /* offset * 4 + bit24 * 2 + (thumb bit) */
5801 val += (offset << 2) | ((insn >> 23) & 2) | 1;
5802 /* pipeline offset */
5803 val += 4;
5804 gen_bx_im(s, val);
5805 return;
5806 } else if ((insn & 0x0e000f00) == 0x0c000100) {
5807 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
5808 /* iWMMXt register transfer. */
5809 if (env->cp15.c15_cpar & (1 << 1))
5810 if (!disas_iwmmxt_insn(env, s, insn))
5811 return;
5812 }
5813 } else if ((insn & 0x0fe00000) == 0x0c400000) {
5814 /* Coprocessor double register transfer. */
5815 } else if ((insn & 0x0f000010) == 0x0e000010) {
5816 /* Additional coprocessor register transfer. */
5817 } else if ((insn & 0x0ff10020) == 0x01000000) {
5818 uint32_t mask;
5819 uint32_t val;
5820 /* cps (privileged) */
5821 if (IS_USER(s))
5822 return;
5823 mask = val = 0;
5824 if (insn & (1 << 19)) {
5825 if (insn & (1 << 8))
5826 mask |= CPSR_A;
5827 if (insn & (1 << 7))
5828 mask |= CPSR_I;
5829 if (insn & (1 << 6))
5830 mask |= CPSR_F;
5831 if (insn & (1 << 18))
5832 val |= mask;
5833 }
5834 if (insn & (1 << 17)) {
5835 mask |= CPSR_M;
5836 val |= (insn & 0x1f);
5837 }
5838 if (mask) {
5839 gen_op_movl_T0_im(val);
5840 gen_set_psr_T0(s, mask, 0);
5841 }
5842 return;
5843 }
5844 goto illegal_op;
5845 }
5846 if (cond != 0xe) {
5847 /* if not always execute, we generate a conditional jump to
5848 next instruction */
5849 s->condlabel = gen_new_label();
5850 gen_test_cc(cond ^ 1, s->condlabel);
5851 s->condjmp = 1;
5852 }
5853 if ((insn & 0x0f900000) == 0x03000000) {
5854 if ((insn & (1 << 21)) == 0) {
5855 ARCH(6T2);
5856 rd = (insn >> 12) & 0xf;
5857 val = ((insn >> 4) & 0xf000) | (insn & 0xfff);
5858 if ((insn & (1 << 22)) == 0) {
5859 /* MOVW */
5860 tmp = new_tmp();
5861 tcg_gen_movi_i32(tmp, val);
5862 } else {
5863 /* MOVT */
5864 tmp = load_reg(s, rd);
5865 tcg_gen_ext16u_i32(tmp, tmp);
5866 tcg_gen_ori_i32(tmp, tmp, val << 16);
5867 }
5868 store_reg(s, rd, tmp);
5869 } else {
5870 if (((insn >> 12) & 0xf) != 0xf)
5871 goto illegal_op;
5872 if (((insn >> 16) & 0xf) == 0) {
5873 gen_nop_hint(s, insn & 0xff);
5874 } else {
5875 /* CPSR = immediate */
5876 val = insn & 0xff;
5877 shift = ((insn >> 8) & 0xf) * 2;
5878 if (shift)
5879 val = (val >> shift) | (val << (32 - shift));
5880 gen_op_movl_T0_im(val);
5881 i = ((insn & (1 << 22)) != 0);
5882 if (gen_set_psr_T0(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i))
5883 goto illegal_op;
5884 }
5885 }
5886 } else if ((insn & 0x0f900000) == 0x01000000
5887 && (insn & 0x00000090) != 0x00000090) {
5888 /* miscellaneous instructions */
5889 op1 = (insn >> 21) & 3;
5890 sh = (insn >> 4) & 0xf;
5891 rm = insn & 0xf;
5892 switch (sh) {
5893 case 0x0: /* move program status register */
5894 if (op1 & 1) {
5895 /* PSR = reg */
5896 gen_movl_T0_reg(s, rm);
5897 i = ((op1 & 2) != 0);
5898 if (gen_set_psr_T0(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i))
5899 goto illegal_op;
5900 } else {
5901 /* reg = PSR */
5902 rd = (insn >> 12) & 0xf;
5903 if (op1 & 2) {
5904 if (IS_USER(s))
5905 goto illegal_op;
5906 tmp = load_cpu_field(spsr);
5907 } else {
5908 tmp = new_tmp();
5909 gen_helper_cpsr_read(tmp);
5910 }
5911 store_reg(s, rd, tmp);
5912 }
5913 break;
5914 case 0x1:
5915 if (op1 == 1) {
5916 /* branch/exchange thumb (bx). */
5917 tmp = load_reg(s, rm);
5918 gen_bx(s, tmp);
5919 } else if (op1 == 3) {
5920 /* clz */
5921 rd = (insn >> 12) & 0xf;
5922 tmp = load_reg(s, rm);
5923 gen_helper_clz(tmp, tmp);
5924 store_reg(s, rd, tmp);
5925 } else {
5926 goto illegal_op;
5927 }
5928 break;
5929 case 0x2:
5930 if (op1 == 1) {
5931 ARCH(5J); /* bxj */
5932 /* Trivial implementation equivalent to bx. */
5933 tmp = load_reg(s, rm);
5934 gen_bx(s, tmp);
5935 } else {
5936 goto illegal_op;
5937 }
5938 break;
5939 case 0x3:
5940 if (op1 != 1)
5941 goto illegal_op;
5942
5943 /* branch link/exchange thumb (blx) */
5944 tmp = load_reg(s, rm);
5945 tmp2 = new_tmp();
5946 tcg_gen_movi_i32(tmp2, s->pc);
5947 store_reg(s, 14, tmp2);
5948 gen_bx(s, tmp);
5949 break;
5950 case 0x5: /* saturating add/subtract */
5951 rd = (insn >> 12) & 0xf;
5952 rn = (insn >> 16) & 0xf;
5953 tmp = load_reg(s, rm);
5954 tmp2 = load_reg(s, rn);
5955 if (op1 & 2)
5956 gen_helper_double_saturate(tmp2, tmp2);
5957 if (op1 & 1)
5958 gen_helper_sub_saturate(tmp, tmp, tmp2);
5959 else
5960 gen_helper_add_saturate(tmp, tmp, tmp2);
5961 dead_tmp(tmp2);
5962 store_reg(s, rd, tmp);
5963 break;
5964 case 7: /* bkpt */
5965 gen_set_condexec(s);
5966 gen_set_pc_im(s->pc - 4);
5967 gen_exception(EXCP_BKPT);
5968 s->is_jmp = DISAS_JUMP;
5969 break;
5970 case 0x8: /* signed multiply */
5971 case 0xa:
5972 case 0xc:
5973 case 0xe:
5974 rs = (insn >> 8) & 0xf;
5975 rn = (insn >> 12) & 0xf;
5976 rd = (insn >> 16) & 0xf;
5977 if (op1 == 1) {
5978 /* (32 * 16) >> 16 */
5979 tmp = load_reg(s, rm);
5980 tmp2 = load_reg(s, rs);
5981 if (sh & 4)
5982 tcg_gen_sari_i32(tmp2, tmp2, 16);
5983 else
5984 gen_sxth(tmp2);
5985 tmp64 = gen_muls_i64_i32(tmp, tmp2);
5986 tcg_gen_shri_i64(tmp64, tmp64, 16);
5987 tmp = new_tmp();
5988 tcg_gen_trunc_i64_i32(tmp, tmp64);
5989 if ((sh & 2) == 0) {
5990 tmp2 = load_reg(s, rn);
5991 gen_helper_add_setq(tmp, tmp, tmp2);
5992 dead_tmp(tmp2);
5993 }
5994 store_reg(s, rd, tmp);
5995 } else {
5996 /* 16 * 16 */
5997 tmp = load_reg(s, rm);
5998 tmp2 = load_reg(s, rs);
5999 gen_mulxy(tmp, tmp2, sh & 2, sh & 4);
6000 dead_tmp(tmp2);
6001 if (op1 == 2) {
6002 tmp64 = tcg_temp_new_i64();
6003 tcg_gen_ext_i32_i64(tmp64, tmp);
6004 dead_tmp(tmp);
6005 gen_addq(s, tmp64, rn, rd);
6006 gen_storeq_reg(s, rn, rd, tmp64);
6007 } else {
6008 if (op1 == 0) {
6009 tmp2 = load_reg(s, rn);
6010 gen_helper_add_setq(tmp, tmp, tmp2);
6011 dead_tmp(tmp2);
6012 }
6013 store_reg(s, rd, tmp);
6014 }
6015 }
6016 break;
6017 default:
6018 goto illegal_op;
6019 }
6020 } else if (((insn & 0x0e000000) == 0 &&
6021 (insn & 0x00000090) != 0x90) ||
6022 ((insn & 0x0e000000) == (1 << 25))) {
6023 int set_cc, logic_cc, shiftop;
6024
6025 op1 = (insn >> 21) & 0xf;
6026 set_cc = (insn >> 20) & 1;
6027 logic_cc = table_logic_cc[op1] & set_cc;
6028
6029 /* data processing instruction */
6030 if (insn & (1 << 25)) {
6031 /* immediate operand */
6032 val = insn & 0xff;
6033 shift = ((insn >> 8) & 0xf) * 2;
6034 if (shift)
6035 val = (val >> shift) | (val << (32 - shift));
6036 gen_op_movl_T1_im(val);
6037 if (logic_cc && shift)
6038 gen_set_CF_bit31(cpu_T[1]);
6039 } else {
6040 /* register */
6041 rm = (insn) & 0xf;
6042 gen_movl_T1_reg(s, rm);
6043 shiftop = (insn >> 5) & 3;
6044 if (!(insn & (1 << 4))) {
6045 shift = (insn >> 7) & 0x1f;
6046 gen_arm_shift_im(cpu_T[1], shiftop, shift, logic_cc);
6047 } else {
6048 rs = (insn >> 8) & 0xf;
6049 tmp = load_reg(s, rs);
6050 gen_arm_shift_reg(cpu_T[1], shiftop, tmp, logic_cc);
6051 }
6052 }
6053 if (op1 != 0x0f && op1 != 0x0d) {
6054 rn = (insn >> 16) & 0xf;
6055 gen_movl_T0_reg(s, rn);
6056 }
6057 rd = (insn >> 12) & 0xf;
6058 switch(op1) {
6059 case 0x00:
6060 gen_op_andl_T0_T1();
6061 gen_movl_reg_T0(s, rd);
6062 if (logic_cc)
6063 gen_op_logic_T0_cc();
6064 break;
6065 case 0x01:
6066 gen_op_xorl_T0_T1();
6067 gen_movl_reg_T0(s, rd);
6068 if (logic_cc)
6069 gen_op_logic_T0_cc();
6070 break;
6071 case 0x02:
6072 if (set_cc && rd == 15) {
6073 /* SUBS r15, ... is used for exception return. */
6074 if (IS_USER(s))
6075 goto illegal_op;
6076 gen_op_subl_T0_T1_cc();
6077 gen_exception_return(s);
6078 } else {
6079 if (set_cc)
6080 gen_op_subl_T0_T1_cc();
6081 else
6082 gen_op_subl_T0_T1();
6083 gen_movl_reg_T0(s, rd);
6084 }
6085 break;
6086 case 0x03:
6087 if (set_cc)
6088 gen_op_rsbl_T0_T1_cc();
6089 else
6090 gen_op_rsbl_T0_T1();
6091 gen_movl_reg_T0(s, rd);
6092 break;
6093 case 0x04:
6094 if (set_cc)
6095 gen_op_addl_T0_T1_cc();
6096 else
6097 gen_op_addl_T0_T1();
6098 gen_movl_reg_T0(s, rd);
6099 break;
6100 case 0x05:
6101 if (set_cc)
6102 gen_op_adcl_T0_T1_cc();
6103 else
6104 gen_adc_T0_T1();
6105 gen_movl_reg_T0(s, rd);
6106 break;
6107 case 0x06:
6108 if (set_cc)
6109 gen_op_sbcl_T0_T1_cc();
6110 else
6111 gen_sbc_T0_T1();
6112 gen_movl_reg_T0(s, rd);
6113 break;
6114 case 0x07:
6115 if (set_cc)
6116 gen_op_rscl_T0_T1_cc();
6117 else
6118 gen_rsc_T0_T1();
6119 gen_movl_reg_T0(s, rd);
6120 break;
6121 case 0x08:
6122 if (set_cc) {
6123 gen_op_andl_T0_T1();
6124 gen_op_logic_T0_cc();
6125 }
6126 break;
6127 case 0x09:
6128 if (set_cc) {
6129 gen_op_xorl_T0_T1();
6130 gen_op_logic_T0_cc();
6131 }
6132 break;
6133 case 0x0a:
6134 if (set_cc) {
6135 gen_op_subl_T0_T1_cc();
6136 }
6137 break;
6138 case 0x0b:
6139 if (set_cc) {
6140 gen_op_addl_T0_T1_cc();
6141 }
6142 break;
6143 case 0x0c:
6144 gen_op_orl_T0_T1();
6145 gen_movl_reg_T0(s, rd);
6146 if (logic_cc)
6147 gen_op_logic_T0_cc();
6148 break;
6149 case 0x0d:
6150 if (logic_cc && rd == 15) {
6151 /* MOVS r15, ... is used for exception return. */
6152 if (IS_USER(s))
6153 goto illegal_op;
6154 gen_op_movl_T0_T1();
6155 gen_exception_return(s);
6156 } else {
6157 gen_movl_reg_T1(s, rd);
6158 if (logic_cc)
6159 gen_op_logic_T1_cc();
6160 }
6161 break;
6162 case 0x0e:
6163 gen_op_bicl_T0_T1();
6164 gen_movl_reg_T0(s, rd);
6165 if (logic_cc)
6166 gen_op_logic_T0_cc();
6167 break;
6168 default:
6169 case 0x0f:
6170 gen_op_notl_T1();
6171 gen_movl_reg_T1(s, rd);
6172 if (logic_cc)
6173 gen_op_logic_T1_cc();
6174 break;
6175 }
6176 } else {
6177 /* other instructions */
6178 op1 = (insn >> 24) & 0xf;
6179 switch(op1) {
6180 case 0x0:
6181 case 0x1:
6182 /* multiplies, extra load/stores */
6183 sh = (insn >> 5) & 3;
6184 if (sh == 0) {
6185 if (op1 == 0x0) {
6186 rd = (insn >> 16) & 0xf;
6187 rn = (insn >> 12) & 0xf;
6188 rs = (insn >> 8) & 0xf;
6189 rm = (insn) & 0xf;
6190 op1 = (insn >> 20) & 0xf;
6191 switch (op1) {
6192 case 0: case 1: case 2: case 3: case 6:
6193 /* 32 bit mul */
6194 tmp = load_reg(s, rs);
6195 tmp2 = load_reg(s, rm);
6196 tcg_gen_mul_i32(tmp, tmp, tmp2);
6197 dead_tmp(tmp2);
6198 if (insn & (1 << 22)) {
6199 /* Subtract (mls) */
6200 ARCH(6T2);
6201 tmp2 = load_reg(s, rn);
6202 tcg_gen_sub_i32(tmp, tmp2, tmp);
6203 dead_tmp(tmp2);
6204 } else if (insn & (1 << 21)) {
6205 /* Add */
6206 tmp2 = load_reg(s, rn);
6207 tcg_gen_add_i32(tmp, tmp, tmp2);
6208 dead_tmp(tmp2);
6209 }
6210 if (insn & (1 << 20))
6211 gen_logic_CC(tmp);
6212 store_reg(s, rd, tmp);
6213 break;
6214 default:
6215 /* 64 bit mul */
6216 tmp = load_reg(s, rs);
6217 tmp2 = load_reg(s, rm);
6218 if (insn & (1 << 22))
6219 tmp64 = gen_muls_i64_i32(tmp, tmp2);
6220 else
6221 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
6222 if (insn & (1 << 21)) /* mult accumulate */
6223 gen_addq(s, tmp64, rn, rd);
6224 if (!(insn & (1 << 23))) { /* double accumulate */
6225 ARCH(6);
6226 gen_addq_lo(s, tmp64, rn);
6227 gen_addq_lo(s, tmp64, rd);
6228 }
6229 if (insn & (1 << 20))
6230 gen_logicq_cc(tmp64);
6231 gen_storeq_reg(s, rn, rd, tmp64);
6232 break;
6233 }
6234 } else {
6235 rn = (insn >> 16) & 0xf;
6236 rd = (insn >> 12) & 0xf;
6237 if (insn & (1 << 23)) {
6238 /* load/store exclusive */
6239 op1 = (insn >> 21) & 0x3;
6240 if (op1)
6241 ARCH(6K);
6242 else
6243 ARCH(6);
6244 gen_movl_T1_reg(s, rn);
6245 addr = cpu_T[1];
6246 if (insn & (1 << 20)) {
6247 gen_helper_mark_exclusive(cpu_env, cpu_T[1]);
6248 switch (op1) {
6249 case 0: /* ldrex */
6250 tmp = gen_ld32(addr, IS_USER(s));
6251 break;
6252 case 1: /* ldrexd */
6253 tmp = gen_ld32(addr, IS_USER(s));
6254 store_reg(s, rd, tmp);
6255 tcg_gen_addi_i32(addr, addr, 4);
6256 tmp = gen_ld32(addr, IS_USER(s));
6257 rd++;
6258 break;
6259 case 2: /* ldrexb */
6260 tmp = gen_ld8u(addr, IS_USER(s));
6261 break;
6262 case 3: /* ldrexh */
6263 tmp = gen_ld16u(addr, IS_USER(s));
6264 break;
6265 default:
6266 abort();
6267 }
6268 store_reg(s, rd, tmp);
6269 } else {
6270 int label = gen_new_label();
6271 rm = insn & 0xf;
6272 gen_helper_test_exclusive(cpu_T[0], cpu_env, addr);
6273 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_T[0],
6274 0, label);
6275 tmp = load_reg(s,rm);
6276 switch (op1) {
6277 case 0: /* strex */
6278 gen_st32(tmp, addr, IS_USER(s));
6279 break;
6280 case 1: /* strexd */
6281 gen_st32(tmp, addr, IS_USER(s));
6282 tcg_gen_addi_i32(addr, addr, 4);
6283 tmp = load_reg(s, rm + 1);
6284 gen_st32(tmp, addr, IS_USER(s));
6285 break;
6286 case 2: /* strexb */
6287 gen_st8(tmp, addr, IS_USER(s));
6288 break;
6289 case 3: /* strexh */
6290 gen_st16(tmp, addr, IS_USER(s));
6291 break;
6292 default:
6293 abort();
6294 }
6295 gen_set_label(label);
6296 gen_movl_reg_T0(s, rd);
6297 }
6298 } else {
6299 /* SWP instruction */
6300 rm = (insn) & 0xf;
6301
6302 /* ??? This is not really atomic. However we know
6303 we never have multiple CPUs running in parallel,
6304 so it is good enough. */
6305 addr = load_reg(s, rn);
6306 tmp = load_reg(s, rm);
6307 if (insn & (1 << 22)) {
6308 tmp2 = gen_ld8u(addr, IS_USER(s));
6309 gen_st8(tmp, addr, IS_USER(s));
6310 } else {
6311 tmp2 = gen_ld32(addr, IS_USER(s));
6312 gen_st32(tmp, addr, IS_USER(s));
6313 }
6314 dead_tmp(addr);
6315 store_reg(s, rd, tmp2);
6316 }
6317 }
6318 } else {
6319 int address_offset;
6320 int load;
6321 /* Misc load/store */
6322 rn = (insn >> 16) & 0xf;
6323 rd = (insn >> 12) & 0xf;
6324 addr = load_reg(s, rn);
6325 if (insn & (1 << 24))
6326 gen_add_datah_offset(s, insn, 0, addr);
6327 address_offset = 0;
6328 if (insn & (1 << 20)) {
6329 /* load */
6330 switch(sh) {
6331 case 1:
6332 tmp = gen_ld16u(addr, IS_USER(s));
6333 break;
6334 case 2:
6335 tmp = gen_ld8s(addr, IS_USER(s));
6336 break;
6337 default:
6338 case 3:
6339 tmp = gen_ld16s(addr, IS_USER(s));
6340 break;
6341 }
6342 load = 1;
6343 } else if (sh & 2) {
6344 /* doubleword */
6345 if (sh & 1) {
6346 /* store */
6347 tmp = load_reg(s, rd);
6348 gen_st32(tmp, addr, IS_USER(s));
6349 tcg_gen_addi_i32(addr, addr, 4);
6350 tmp = load_reg(s, rd + 1);
6351 gen_st32(tmp, addr, IS_USER(s));
6352 load = 0;
6353 } else {
6354 /* load */
6355 tmp = gen_ld32(addr, IS_USER(s));
6356 store_reg(s, rd, tmp);
6357 tcg_gen_addi_i32(addr, addr, 4);
6358 tmp = gen_ld32(addr, IS_USER(s));
6359 rd++;
6360 load = 1;
6361 }
6362 address_offset = -4;
6363 } else {
6364 /* store */
6365 tmp = load_reg(s, rd);
6366 gen_st16(tmp, addr, IS_USER(s));
6367 load = 0;
6368 }
6369 /* Perform base writeback before the loaded value to
6370 ensure correct behavior with overlapping index registers.
6371 ldrd with base writeback is is undefined if the
6372 destination and index registers overlap. */
6373 if (!(insn & (1 << 24))) {
6374 gen_add_datah_offset(s, insn, address_offset, addr);
6375 store_reg(s, rn, addr);
6376 } else if (insn & (1 << 21)) {
6377 if (address_offset)
6378 tcg_gen_addi_i32(addr, addr, address_offset);
6379 store_reg(s, rn, addr);
6380 } else {
6381 dead_tmp(addr);
6382 }
6383 if (load) {
6384 /* Complete the load. */
6385 store_reg(s, rd, tmp);
6386 }
6387 }
6388 break;
6389 case 0x4:
6390 case 0x5:
6391 goto do_ldst;
6392 case 0x6:
6393 case 0x7:
6394 if (insn & (1 << 4)) {
6395 ARCH(6);
6396 /* Armv6 Media instructions. */
6397 rm = insn & 0xf;
6398 rn = (insn >> 16) & 0xf;
6399 rd = (insn >> 12) & 0xf;
6400 rs = (insn >> 8) & 0xf;
6401 switch ((insn >> 23) & 3) {
6402 case 0: /* Parallel add/subtract. */
6403 op1 = (insn >> 20) & 7;
6404 tmp = load_reg(s, rn);
6405 tmp2 = load_reg(s, rm);
6406 sh = (insn >> 5) & 7;
6407 if ((op1 & 3) == 0 || sh == 5 || sh == 6)
6408 goto illegal_op;
6409 gen_arm_parallel_addsub(op1, sh, tmp, tmp2);
6410 dead_tmp(tmp2);
6411 store_reg(s, rd, tmp);
6412 break;
6413 case 1:
6414 if ((insn & 0x00700020) == 0) {
6415 /* Halfword pack. */
6416 tmp = load_reg(s, rn);
6417 tmp2 = load_reg(s, rm);
6418 shift = (insn >> 7) & 0x1f;
6419 if (insn & (1 << 6)) {
6420 /* pkhtb */
6421 if (shift == 0)
6422 shift = 31;
6423 tcg_gen_sari_i32(tmp2, tmp2, shift);
6424 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
6425 tcg_gen_ext16u_i32(tmp2, tmp2);
6426 } else {
6427 /* pkhbt */
6428 if (shift)
6429 tcg_gen_shli_i32(tmp2, tmp2, shift);
6430 tcg_gen_ext16u_i32(tmp, tmp);
6431 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
6432 }
6433 tcg_gen_or_i32(tmp, tmp, tmp2);
6434 dead_tmp(tmp2);
6435 store_reg(s, rd, tmp);
6436 } else if ((insn & 0x00200020) == 0x00200000) {
6437 /* [us]sat */
6438 tmp = load_reg(s, rm);
6439 shift = (insn >> 7) & 0x1f;
6440 if (insn & (1 << 6)) {
6441 if (shift == 0)
6442 shift = 31;
6443 tcg_gen_sari_i32(tmp, tmp, shift);
6444 } else {
6445 tcg_gen_shli_i32(tmp, tmp, shift);
6446 }
6447 sh = (insn >> 16) & 0x1f;
6448 if (sh != 0) {
6449 if (insn & (1 << 22))
6450 gen_helper_usat(tmp, tmp, tcg_const_i32(sh));
6451 else
6452 gen_helper_ssat(tmp, tmp, tcg_const_i32(sh));
6453 }
6454 store_reg(s, rd, tmp);
6455 } else if ((insn & 0x00300fe0) == 0x00200f20) {
6456 /* [us]sat16 */
6457 tmp = load_reg(s, rm);
6458 sh = (insn >> 16) & 0x1f;
6459 if (sh != 0) {
6460 if (insn & (1 << 22))
6461 gen_helper_usat16(tmp, tmp, tcg_const_i32(sh));
6462 else
6463 gen_helper_ssat16(tmp, tmp, tcg_const_i32(sh));
6464 }
6465 store_reg(s, rd, tmp);
6466 } else if ((insn & 0x00700fe0) == 0x00000fa0) {
6467 /* Select bytes. */
6468 tmp = load_reg(s, rn);
6469 tmp2 = load_reg(s, rm);
6470 tmp3 = new_tmp();
6471 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUState, GE));
6472 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
6473 dead_tmp(tmp3);
6474 dead_tmp(tmp2);
6475 store_reg(s, rd, tmp);
6476 } else if ((insn & 0x000003e0) == 0x00000060) {
6477 tmp = load_reg(s, rm);
6478 shift = (insn >> 10) & 3;
6479 /* ??? In many cases it's not neccessary to do a
6480 rotate, a shift is sufficient. */
6481 if (shift != 0)
6482 tcg_gen_rori_i32(tmp, tmp, shift * 8);
6483 op1 = (insn >> 20) & 7;
6484 switch (op1) {
6485 case 0: gen_sxtb16(tmp); break;
6486 case 2: gen_sxtb(tmp); break;
6487 case 3: gen_sxth(tmp); break;
6488 case 4: gen_uxtb16(tmp); break;
6489 case 6: gen_uxtb(tmp); break;
6490 case 7: gen_uxth(tmp); break;
6491 default: goto illegal_op;
6492 }
6493 if (rn != 15) {
6494 tmp2 = load_reg(s, rn);
6495 if ((op1 & 3) == 0) {
6496 gen_add16(tmp, tmp2);
6497 } else {
6498 tcg_gen_add_i32(tmp, tmp, tmp2);
6499 dead_tmp(tmp2);
6500 }
6501 }
6502 store_reg(s, rd, tmp);
6503 } else if ((insn & 0x003f0f60) == 0x003f0f20) {
6504 /* rev */
6505 tmp = load_reg(s, rm);
6506 if (insn & (1 << 22)) {
6507 if (insn & (1 << 7)) {
6508 gen_revsh(tmp);
6509 } else {
6510 ARCH(6T2);
6511 gen_helper_rbit(tmp, tmp);
6512 }
6513 } else {
6514 if (insn & (1 << 7))
6515 gen_rev16(tmp);
6516 else
6517 tcg_gen_bswap_i32(tmp, tmp);
6518 }
6519 store_reg(s, rd, tmp);
6520 } else {
6521 goto illegal_op;
6522 }
6523 break;
6524 case 2: /* Multiplies (Type 3). */
6525 tmp = load_reg(s, rm);
6526 tmp2 = load_reg(s, rs);
6527 if (insn & (1 << 20)) {
6528 /* Signed multiply most significant [accumulate]. */
6529 tmp64 = gen_muls_i64_i32(tmp, tmp2);
6530 if (insn & (1 << 5))
6531 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
6532 tcg_gen_shri_i64(tmp64, tmp64, 32);
6533 tmp = new_tmp();
6534 tcg_gen_trunc_i64_i32(tmp, tmp64);
6535 if (rn != 15) {
6536 tmp2 = load_reg(s, rn);
6537 if (insn & (1 << 6)) {
6538 tcg_gen_sub_i32(tmp, tmp, tmp2);
6539 } else {
6540 tcg_gen_add_i32(tmp, tmp, tmp2);
6541 }
6542 dead_tmp(tmp2);
6543 }
6544 store_reg(s, rd, tmp);
6545 } else {
6546 if (insn & (1 << 5))
6547 gen_swap_half(tmp2);
6548 gen_smul_dual(tmp, tmp2);
6549 /* This addition cannot overflow. */
6550 if (insn & (1 << 6)) {
6551 tcg_gen_sub_i32(tmp, tmp, tmp2);
6552 } else {
6553 tcg_gen_add_i32(tmp, tmp, tmp2);
6554 }
6555 dead_tmp(tmp2);
6556 if (insn & (1 << 22)) {
6557 /* smlald, smlsld */
6558 tmp64 = tcg_temp_new_i64();
6559 tcg_gen_ext_i32_i64(tmp64, tmp);
6560 dead_tmp(tmp);
6561 gen_addq(s, tmp64, rd, rn);
6562 gen_storeq_reg(s, rd, rn, tmp64);
6563 } else {
6564 /* smuad, smusd, smlad, smlsd */
6565 if (rd != 15)
6566 {
6567 tmp2 = load_reg(s, rd);
6568 gen_helper_add_setq(tmp, tmp, tmp2);
6569 dead_tmp(tmp2);
6570 }
6571 store_reg(s, rn, tmp);
6572 }
6573 }
6574 break;
6575 case 3:
6576 op1 = ((insn >> 17) & 0x38) | ((insn >> 5) & 7);
6577 switch (op1) {
6578 case 0: /* Unsigned sum of absolute differences. */
6579 ARCH(6);
6580 tmp = load_reg(s, rm);
6581 tmp2 = load_reg(s, rs);
6582 gen_helper_usad8(tmp, tmp, tmp2);
6583 dead_tmp(tmp2);
6584 if (rn != 15) {
6585 tmp2 = load_reg(s, rn);
6586 tcg_gen_add_i32(tmp, tmp, tmp2);
6587 dead_tmp(tmp2);
6588 }
6589 store_reg(s, rd, tmp);
6590 break;
6591 case 0x20: case 0x24: case 0x28: case 0x2c:
6592 /* Bitfield insert/clear. */
6593 ARCH(6T2);
6594 shift = (insn >> 7) & 0x1f;
6595 i = (insn >> 16) & 0x1f;
6596 i = i + 1 - shift;
6597 if (rm == 15) {
6598 tmp = new_tmp();
6599 tcg_gen_movi_i32(tmp, 0);
6600 } else {
6601 tmp = load_reg(s, rm);
6602 }
6603 if (i != 32) {
6604 tmp2 = load_reg(s, rd);
6605 gen_bfi(tmp, tmp2, tmp, shift, (1u << i) - 1);
6606 dead_tmp(tmp2);
6607 }
6608 store_reg(s, rd, tmp);
6609 break;
6610 case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
6611 case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
6612 tmp = load_reg(s, rm);
6613 shift = (insn >> 7) & 0x1f;
6614 i = ((insn >> 16) & 0x1f) + 1;
6615 if (shift + i > 32)
6616 goto illegal_op;
6617 if (i < 32) {
6618 if (op1 & 0x20) {
6619 gen_ubfx(tmp, shift, (1u << i) - 1);
6620 } else {
6621 gen_sbfx(tmp, shift, i);
6622 }
6623 }
6624 store_reg(s, rd, tmp);
6625 break;
6626 default:
6627 goto illegal_op;
6628 }
6629 break;
6630 }
6631 break;
6632 }
6633 do_ldst:
6634 /* Check for undefined extension instructions
6635 * per the ARM Bible IE:
6636 * xxxx 0111 1111 xxxx xxxx xxxx 1111 xxxx
6637 */
6638 sh = (0xf << 20) | (0xf << 4);
6639 if (op1 == 0x7 && ((insn & sh) == sh))
6640 {
6641 goto illegal_op;
6642 }
6643 /* load/store byte/word */
6644 rn = (insn >> 16) & 0xf;
6645 rd = (insn >> 12) & 0xf;
6646 tmp2 = load_reg(s, rn);
6647 i = (IS_USER(s) || (insn & 0x01200000) == 0x00200000);
6648 if (insn & (1 << 24))
6649 gen_add_data_offset(s, insn, tmp2);
6650 if (insn & (1 << 20)) {
6651 /* load */
6652 s->is_mem = 1;
6653 if (insn & (1 << 22)) {
6654 tmp = gen_ld8u(tmp2, i);
6655 } else {
6656 tmp = gen_ld32(tmp2, i);
6657 }
6658 } else {
6659 /* store */
6660 tmp = load_reg(s, rd);
6661 if (insn & (1 << 22))
6662 gen_st8(tmp, tmp2, i);
6663 else
6664 gen_st32(tmp, tmp2, i);
6665 }
6666 if (!(insn & (1 << 24))) {
6667 gen_add_data_offset(s, insn, tmp2);
6668 store_reg(s, rn, tmp2);
6669 } else if (insn & (1 << 21)) {
6670 store_reg(s, rn, tmp2);
6671 } else {
6672 dead_tmp(tmp2);
6673 }
6674 if (insn & (1 << 20)) {
6675 /* Complete the load. */
6676 if (rd == 15)
6677 gen_bx(s, tmp);
6678 else
6679 store_reg(s, rd, tmp);
6680 }
6681 break;
6682 case 0x08:
6683 case 0x09:
6684 {
6685 int j, n, user, loaded_base;
6686 TCGv loaded_var;
6687 /* load/store multiple words */
6688 /* XXX: store correct base if write back */
6689 user = 0;
6690 if (insn & (1 << 22)) {
6691 if (IS_USER(s))
6692 goto illegal_op; /* only usable in supervisor mode */
6693
6694 if ((insn & (1 << 15)) == 0)
6695 user = 1;
6696 }
6697 rn = (insn >> 16) & 0xf;
6698 addr = load_reg(s, rn);
6699
6700 /* compute total size */
6701 loaded_base = 0;
6702 TCGV_UNUSED(loaded_var);
6703 n = 0;
6704 for(i=0;i<16;i++) {
6705 if (insn & (1 << i))
6706 n++;
6707 }
6708 /* XXX: test invalid n == 0 case ? */
6709 if (insn & (1 << 23)) {
6710 if (insn & (1 << 24)) {
6711 /* pre increment */
6712 tcg_gen_addi_i32(addr, addr, 4);
6713 } else {
6714 /* post increment */
6715 }
6716 } else {
6717 if (insn & (1 << 24)) {
6718 /* pre decrement */
6719 tcg_gen_addi_i32(addr, addr, -(n * 4));
6720 } else {
6721 /* post decrement */
6722 if (n != 1)
6723 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
6724 }
6725 }
6726 j = 0;
6727 for(i=0;i<16;i++) {
6728 if (insn & (1 << i)) {
6729 if (insn & (1 << 20)) {
6730 /* load */
6731 tmp = gen_ld32(addr, IS_USER(s));
6732 if (i == 15) {
6733 gen_bx(s, tmp);
6734 } else if (user) {
6735 gen_helper_set_user_reg(tcg_const_i32(i), tmp);
6736 dead_tmp(tmp);
6737 } else if (i == rn) {
6738 loaded_var = tmp;
6739 loaded_base = 1;
6740 } else {
6741 store_reg(s, i, tmp);
6742 }
6743 } else {
6744 /* store */
6745 if (i == 15) {
6746 /* special case: r15 = PC + 8 */
6747 val = (long)s->pc + 4;
6748 tmp = new_tmp();
6749 tcg_gen_movi_i32(tmp, val);
6750 } else if (user) {
6751 tmp = new_tmp();
6752 gen_helper_get_user_reg(tmp, tcg_const_i32(i));
6753 } else {
6754 tmp = load_reg(s, i);
6755 }
6756 gen_st32(tmp, addr, IS_USER(s));
6757 }
6758 j++;
6759 /* no need to add after the last transfer */
6760 if (j != n)
6761 tcg_gen_addi_i32(addr, addr, 4);
6762 }
6763 }
6764 if (insn & (1 << 21)) {
6765 /* write back */
6766 if (insn & (1 << 23)) {
6767 if (insn & (1 << 24)) {
6768 /* pre increment */
6769 } else {
6770 /* post increment */
6771 tcg_gen_addi_i32(addr, addr, 4);
6772 }
6773 } else {
6774 if (insn & (1 << 24)) {
6775 /* pre decrement */
6776 if (n != 1)
6777 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
6778 } else {
6779 /* post decrement */
6780 tcg_gen_addi_i32(addr, addr, -(n * 4));
6781 }
6782 }
6783 store_reg(s, rn, addr);
6784 } else {
6785 dead_tmp(addr);
6786 }
6787 if (loaded_base) {
6788 store_reg(s, rn, loaded_var);
6789 }
6790 if ((insn & (1 << 22)) && !user) {
6791 /* Restore CPSR from SPSR. */
6792 tmp = load_cpu_field(spsr);
6793 gen_set_cpsr(tmp, 0xffffffff);
6794 dead_tmp(tmp);
6795 s->is_jmp = DISAS_UPDATE;
6796 }
6797 }
6798 break;
6799 case 0xa:
6800 case 0xb:
6801 {
6802 int32_t offset;
6803
6804 /* branch (and link) */
6805 val = (int32_t)s->pc;
6806 if (insn & (1 << 24)) {
6807 tmp = new_tmp();
6808 tcg_gen_movi_i32(tmp, val);
6809 store_reg(s, 14, tmp);
6810 }
6811 offset = (((int32_t)insn << 8) >> 8);
6812 val += (offset << 2) + 4;
6813 gen_jmp(s, val);
6814 }
6815 break;
6816 case 0xc:
6817 case 0xd:
6818 case 0xe:
6819 /* Coprocessor. */
6820 if (disas_coproc_insn(env, s, insn))
6821 goto illegal_op;
6822 break;
6823 case 0xf:
6824 /* swi */
6825 gen_set_pc_im(s->pc);
6826 s->is_jmp = DISAS_SWI;
6827 break;
6828 default:
6829 illegal_op:
6830 gen_set_condexec(s);
6831 gen_set_pc_im(s->pc - 4);
6832 gen_exception(EXCP_UDEF);
6833 s->is_jmp = DISAS_JUMP;
6834 break;
6835 }
6836 }
6837 }
6838
6839 /* Return true if this is a Thumb-2 logical op. */
6840 static int
6841 thumb2_logic_op(int op)
6842 {
6843 return (op < 8);
6844 }
6845
6846 /* Generate code for a Thumb-2 data processing operation. If CONDS is nonzero
6847 then set condition code flags based on the result of the operation.
6848 If SHIFTER_OUT is nonzero then set the carry flag for logical operations
6849 to the high bit of T1.
6850 Returns zero if the opcode is valid. */
6851
6852 static int
6853 gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out)
6854 {
6855 int logic_cc;
6856
6857 logic_cc = 0;
6858 switch (op) {
6859 case 0: /* and */
6860 gen_op_andl_T0_T1();
6861 logic_cc = conds;
6862 break;
6863 case 1: /* bic */
6864 gen_op_bicl_T0_T1();
6865 logic_cc = conds;
6866 break;
6867 case 2: /* orr */
6868 gen_op_orl_T0_T1();
6869 logic_cc = conds;
6870 break;
6871 case 3: /* orn */
6872 gen_op_notl_T1();
6873 gen_op_orl_T0_T1();
6874 logic_cc = conds;
6875 break;
6876 case 4: /* eor */
6877 gen_op_xorl_T0_T1();
6878 logic_cc = conds;
6879 break;
6880 case 8: /* add */
6881 if (conds)
6882 gen_op_addl_T0_T1_cc();
6883 else
6884 gen_op_addl_T0_T1();
6885 break;
6886 case 10: /* adc */
6887 if (conds)
6888 gen_op_adcl_T0_T1_cc();
6889 else
6890 gen_adc_T0_T1();
6891 break;
6892 case 11: /* sbc */
6893 if (conds)
6894 gen_op_sbcl_T0_T1_cc();
6895 else
6896 gen_sbc_T0_T1();
6897 break;
6898 case 13: /* sub */
6899 if (conds)
6900 gen_op_subl_T0_T1_cc();
6901 else
6902 gen_op_subl_T0_T1();
6903 break;
6904 case 14: /* rsb */
6905 if (conds)
6906 gen_op_rsbl_T0_T1_cc();
6907 else
6908 gen_op_rsbl_T0_T1();
6909 break;
6910 default: /* 5, 6, 7, 9, 12, 15. */
6911 return 1;
6912 }
6913 if (logic_cc) {
6914 gen_op_logic_T0_cc();
6915 if (shifter_out)
6916 gen_set_CF_bit31(cpu_T[1]);
6917 }
6918 return 0;
6919 }
6920
6921 /* Translate a 32-bit thumb instruction. Returns nonzero if the instruction
6922 is not legal. */
6923 static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
6924 {
6925 uint32_t insn, imm, shift, offset;
6926 uint32_t rd, rn, rm, rs;
6927 TCGv tmp;
6928 TCGv tmp2;
6929 TCGv tmp3;
6930 TCGv addr;
6931 TCGv_i64 tmp64;
6932 int op;
6933 int shiftop;
6934 int conds;
6935 int logic_cc;
6936
6937 if (!(arm_feature(env, ARM_FEATURE_THUMB2)
6938 || arm_feature (env, ARM_FEATURE_M))) {
6939 /* Thumb-1 cores may need to treat bl and blx as a pair of
6940 16-bit instructions to get correct prefetch abort behavior. */
6941 insn = insn_hw1;
6942 if ((insn & (1 << 12)) == 0) {
6943 /* Second half of blx. */
6944 offset = ((insn & 0x7ff) << 1);
6945 tmp = load_reg(s, 14);
6946 tcg_gen_addi_i32(tmp, tmp, offset);
6947 tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
6948
6949 tmp2 = new_tmp();
6950 tcg_gen_movi_i32(tmp2, s->pc | 1);
6951 store_reg(s, 14, tmp2);
6952 gen_bx(s, tmp);
6953 return 0;
6954 }
6955 if (insn & (1 << 11)) {
6956 /* Second half of bl. */
6957 offset = ((insn & 0x7ff) << 1) | 1;
6958 tmp = load_reg(s, 14);
6959 tcg_gen_addi_i32(tmp, tmp, offset);
6960
6961 tmp2 = new_tmp();
6962 tcg_gen_movi_i32(tmp2, s->pc | 1);
6963 store_reg(s, 14, tmp2);
6964 gen_bx(s, tmp);
6965 return 0;
6966 }
6967 if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
6968 /* Instruction spans a page boundary. Implement it as two
6969 16-bit instructions in case the second half causes an
6970 prefetch abort. */
6971 offset = ((int32_t)insn << 21) >> 9;
6972 gen_op_movl_T0_im(s->pc + 2 + offset);
6973 gen_movl_reg_T0(s, 14);
6974 return 0;
6975 }
6976 /* Fall through to 32-bit decode. */
6977 }
6978
6979 insn = lduw_code(s->pc);
6980 s->pc += 2;
6981 insn |= (uint32_t)insn_hw1 << 16;
6982
6983 if ((insn & 0xf800e800) != 0xf000e800) {
6984 ARCH(6T2);
6985 }
6986
6987 rn = (insn >> 16) & 0xf;
6988 rs = (insn >> 12) & 0xf;
6989 rd = (insn >> 8) & 0xf;
6990 rm = insn & 0xf;
6991 switch ((insn >> 25) & 0xf) {
6992 case 0: case 1: case 2: case 3:
6993 /* 16-bit instructions. Should never happen. */
6994 abort();
6995 case 4:
6996 if (insn & (1 << 22)) {
6997 /* Other load/store, table branch. */
6998 if (insn & 0x01200000) {
6999 /* Load/store doubleword. */
7000 if (rn == 15) {
7001 addr = new_tmp();
7002 tcg_gen_movi_i32(addr, s->pc & ~3);
7003 } else {
7004 addr = load_reg(s, rn);
7005 }
7006 offset = (insn & 0xff) * 4;
7007 if ((insn & (1 << 23)) == 0)
7008 offset = -offset;
7009 if (insn & (1 << 24)) {
7010 tcg_gen_addi_i32(addr, addr, offset);
7011 offset = 0;
7012 }
7013 if (insn & (1 << 20)) {
7014 /* ldrd */
7015 tmp = gen_ld32(addr, IS_USER(s));
7016 store_reg(s, rs, tmp);
7017 tcg_gen_addi_i32(addr, addr, 4);
7018 tmp = gen_ld32(addr, IS_USER(s));
7019 store_reg(s, rd, tmp);
7020 } else {
7021 /* strd */
7022 tmp = load_reg(s, rs);
7023 gen_st32(tmp, addr, IS_USER(s));
7024 tcg_gen_addi_i32(addr, addr, 4);
7025 tmp = load_reg(s, rd);
7026 gen_st32(tmp, addr, IS_USER(s));
7027 }
7028 if (insn & (1 << 21)) {
7029 /* Base writeback. */
7030 if (rn == 15)
7031 goto illegal_op;
7032 tcg_gen_addi_i32(addr, addr, offset - 4);
7033 store_reg(s, rn, addr);
7034 } else {
7035 dead_tmp(addr);
7036 }
7037 } else if ((insn & (1 << 23)) == 0) {
7038 /* Load/store exclusive word. */
7039 gen_movl_T1_reg(s, rn);
7040 addr = cpu_T[1];
7041 if (insn & (1 << 20)) {
7042 gen_helper_mark_exclusive(cpu_env, cpu_T[1]);
7043 tmp = gen_ld32(addr, IS_USER(s));
7044 store_reg(s, rd, tmp);
7045 } else {
7046 int label = gen_new_label();
7047 gen_helper_test_exclusive(cpu_T[0], cpu_env, addr);
7048 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_T[0],
7049 0, label);
7050 tmp = load_reg(s, rs);
7051 gen_st32(tmp, cpu_T[1], IS_USER(s));
7052 gen_set_label(label);
7053 gen_movl_reg_T0(s, rd);
7054 }
7055 } else if ((insn & (1 << 6)) == 0) {
7056 /* Table Branch. */
7057 if (rn == 15) {
7058 addr = new_tmp();
7059 tcg_gen_movi_i32(addr, s->pc);
7060 } else {
7061 addr = load_reg(s, rn);
7062 }
7063 tmp = load_reg(s, rm);
7064 tcg_gen_add_i32(addr, addr, tmp);
7065 if (insn & (1 << 4)) {
7066 /* tbh */
7067 tcg_gen_add_i32(addr, addr, tmp);
7068 dead_tmp(tmp);
7069 tmp = gen_ld16u(addr, IS_USER(s));
7070 } else { /* tbb */
7071 dead_tmp(tmp);
7072 tmp = gen_ld8u(addr, IS_USER(s));
7073 }
7074 dead_tmp(addr);
7075 tcg_gen_shli_i32(tmp, tmp, 1);
7076 tcg_gen_addi_i32(tmp, tmp, s->pc);
7077 store_reg(s, 15, tmp);
7078 } else {
7079 /* Load/store exclusive byte/halfword/doubleword. */
7080 /* ??? These are not really atomic. However we know
7081 we never have multiple CPUs running in parallel,
7082 so it is good enough. */
7083 op = (insn >> 4) & 0x3;
7084 /* Must use a global reg for the address because we have
7085 a conditional branch in the store instruction. */
7086 gen_movl_T1_reg(s, rn);
7087 addr = cpu_T[1];
7088 if (insn & (1 << 20)) {
7089 gen_helper_mark_exclusive(cpu_env, addr);
7090 switch (op) {
7091 case 0:
7092 tmp = gen_ld8u(addr, IS_USER(s));
7093 break;
7094 case 1:
7095 tmp = gen_ld16u(addr, IS_USER(s));
7096 break;
7097 case 3:
7098 tmp = gen_ld32(addr, IS_USER(s));
7099 tcg_gen_addi_i32(addr, addr, 4);
7100 tmp2 = gen_ld32(addr, IS_USER(s));
7101 store_reg(s, rd, tmp2);
7102 break;
7103 default:
7104 goto illegal_op;
7105 }
7106 store_reg(s, rs, tmp);
7107 } else {
7108 int label = gen_new_label();
7109 /* Must use a global that is not killed by the branch. */
7110 gen_helper_test_exclusive(cpu_T[0], cpu_env, addr);
7111 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_T[0], 0, label);
7112 tmp = load_reg(s, rs);
7113 switch (op) {
7114 case 0:
7115 gen_st8(tmp, addr, IS_USER(s));
7116 break;
7117 case 1:
7118 gen_st16(tmp, addr, IS_USER(s));
7119 break;
7120 case 3:
7121 gen_st32(tmp, addr, IS_USER(s));
7122 tcg_gen_addi_i32(addr, addr, 4);
7123 tmp = load_reg(s, rd);
7124 gen_st32(tmp, addr, IS_USER(s));
7125 break;
7126 default:
7127 goto illegal_op;
7128 }
7129 gen_set_label(label);
7130 gen_movl_reg_T0(s, rm);
7131 }
7132 }
7133 } else {
7134 /* Load/store multiple, RFE, SRS. */
7135 if (((insn >> 23) & 1) == ((insn >> 24) & 1)) {
7136 /* Not available in user mode. */
7137 if (IS_USER(s))
7138 goto illegal_op;
7139 if (insn & (1 << 20)) {
7140 /* rfe */
7141 addr = load_reg(s, rn);
7142 if ((insn & (1 << 24)) == 0)
7143 tcg_gen_addi_i32(addr, addr, -8);
7144 /* Load PC into tmp and CPSR into tmp2. */
7145 tmp = gen_ld32(addr, 0);
7146 tcg_gen_addi_i32(addr, addr, 4);
7147 tmp2 = gen_ld32(addr, 0);
7148 if (insn & (1 << 21)) {
7149 /* Base writeback. */
7150 if (insn & (1 << 24)) {
7151 tcg_gen_addi_i32(addr, addr, 4);
7152 } else {
7153 tcg_gen_addi_i32(addr, addr, -4);
7154 }
7155 store_reg(s, rn, addr);
7156 } else {
7157 dead_tmp(addr);
7158 }
7159 gen_rfe(s, tmp, tmp2);
7160 } else {
7161 /* srs */
7162 op = (insn & 0x1f);
7163 if (op == (env->uncached_cpsr & CPSR_M)) {
7164 addr = load_reg(s, 13);
7165 } else {
7166 addr = new_tmp();
7167 gen_helper_get_r13_banked(addr, cpu_env, tcg_const_i32(op));
7168 }
7169 if ((insn & (1 << 24)) == 0) {
7170 tcg_gen_addi_i32(addr, addr, -8);
7171 }
7172 tmp = load_reg(s, 14);
7173 gen_st32(tmp, addr, 0);
7174 tcg_gen_addi_i32(addr, addr, 4);
7175 tmp = new_tmp();
7176 gen_helper_cpsr_read(tmp);
7177 gen_st32(tmp, addr, 0);
7178 if (insn & (1 << 21)) {
7179 if ((insn & (1 << 24)) == 0) {
7180 tcg_gen_addi_i32(addr, addr, -4);
7181 } else {
7182 tcg_gen_addi_i32(addr, addr, 4);
7183 }
7184 if (op == (env->uncached_cpsr & CPSR_M)) {
7185 store_reg(s, 13, addr);
7186 } else {
7187 gen_helper_set_r13_banked(cpu_env,
7188 tcg_const_i32(op), addr);
7189 }
7190 } else {
7191 dead_tmp(addr);
7192 }
7193 }
7194 } else {
7195 int i;
7196 /* Load/store multiple. */
7197 addr = load_reg(s, rn);
7198 offset = 0;
7199 for (i = 0; i < 16; i++) {
7200 if (insn & (1 << i))
7201 offset += 4;
7202 }
7203 if (insn & (1 << 24)) {
7204 tcg_gen_addi_i32(addr, addr, -offset);
7205 }
7206
7207 for (i = 0; i < 16; i++) {
7208 if ((insn & (1 << i)) == 0)
7209 continue;
7210 if (insn & (1 << 20)) {
7211 /* Load. */
7212 tmp = gen_ld32(addr, IS_USER(s));
7213 if (i == 15) {
7214 gen_bx(s, tmp);
7215 } else {
7216 store_reg(s, i, tmp);
7217 }
7218 } else {
7219 /* Store. */
7220 tmp = load_reg(s, i);
7221 gen_st32(tmp, addr, IS_USER(s));
7222 }
7223 tcg_gen_addi_i32(addr, addr, 4);
7224 }
7225 if (insn & (1 << 21)) {
7226 /* Base register writeback. */
7227 if (insn & (1 << 24)) {
7228 tcg_gen_addi_i32(addr, addr, -offset);
7229 }
7230 /* Fault if writeback register is in register list. */
7231 if (insn & (1 << rn))
7232 goto illegal_op;
7233 store_reg(s, rn, addr);
7234 } else {
7235 dead_tmp(addr);
7236 }
7237 }
7238 }
7239 break;
7240 case 5: /* Data processing register constant shift. */
7241 if (rn == 15)
7242 gen_op_movl_T0_im(0);
7243 else
7244 gen_movl_T0_reg(s, rn);
7245 gen_movl_T1_reg(s, rm);
7246 op = (insn >> 21) & 0xf;
7247 shiftop = (insn >> 4) & 3;
7248 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
7249 conds = (insn & (1 << 20)) != 0;
7250 logic_cc = (conds && thumb2_logic_op(op));
7251 gen_arm_shift_im(cpu_T[1], shiftop, shift, logic_cc);
7252 if (gen_thumb2_data_op(s, op, conds, 0))
7253 goto illegal_op;
7254 if (rd != 15)
7255 gen_movl_reg_T0(s, rd);
7256 break;
7257 case 13: /* Misc data processing. */
7258 op = ((insn >> 22) & 6) | ((insn >> 7) & 1);
7259 if (op < 4 && (insn & 0xf000) != 0xf000)
7260 goto illegal_op;
7261 switch (op) {
7262 case 0: /* Register controlled shift. */
7263 tmp = load_reg(s, rn);
7264 tmp2 = load_reg(s, rm);
7265 if ((insn & 0x70) != 0)
7266 goto illegal_op;
7267 op = (insn >> 21) & 3;
7268 logic_cc = (insn & (1 << 20)) != 0;
7269 gen_arm_shift_reg(tmp, op, tmp2, logic_cc);
7270 if (logic_cc)
7271 gen_logic_CC(tmp);
7272 store_reg(s, rd, tmp);
7273 break;
7274 case 1: /* Sign/zero extend. */
7275 tmp = load_reg(s, rm);
7276 shift = (insn >> 4) & 3;
7277 /* ??? In many cases it's not neccessary to do a
7278 rotate, a shift is sufficient. */
7279 if (shift != 0)
7280 tcg_gen_rori_i32(tmp, tmp, shift * 8);
7281 op = (insn >> 20) & 7;
7282 switch (op) {
7283 case 0: gen_sxth(tmp); break;
7284 case 1: gen_uxth(tmp); break;
7285 case 2: gen_sxtb16(tmp); break;
7286 case 3: gen_uxtb16(tmp); break;
7287 case 4: gen_sxtb(tmp); break;
7288 case 5: gen_uxtb(tmp); break;
7289 default: goto illegal_op;
7290 }
7291 if (rn != 15) {
7292 tmp2 = load_reg(s, rn);
7293 if ((op >> 1) == 1) {
7294 gen_add16(tmp, tmp2);
7295 } else {
7296 tcg_gen_add_i32(tmp, tmp, tmp2);
7297 dead_tmp(tmp2);
7298 }
7299 }
7300 store_reg(s, rd, tmp);
7301 break;
7302 case 2: /* SIMD add/subtract. */
7303 op = (insn >> 20) & 7;
7304 shift = (insn >> 4) & 7;
7305 if ((op & 3) == 3 || (shift & 3) == 3)
7306 goto illegal_op;
7307 tmp = load_reg(s, rn);
7308 tmp2 = load_reg(s, rm);
7309 gen_thumb2_parallel_addsub(op, shift, tmp, tmp2);
7310 dead_tmp(tmp2);
7311 store_reg(s, rd, tmp);
7312 break;
7313 case 3: /* Other data processing. */
7314 op = ((insn >> 17) & 0x38) | ((insn >> 4) & 7);
7315 if (op < 4) {
7316 /* Saturating add/subtract. */
7317 tmp = load_reg(s, rn);
7318 tmp2 = load_reg(s, rm);
7319 if (op & 2)
7320 gen_helper_double_saturate(tmp, tmp);
7321 if (op & 1)
7322 gen_helper_sub_saturate(tmp, tmp2, tmp);
7323 else
7324 gen_helper_add_saturate(tmp, tmp, tmp2);
7325 dead_tmp(tmp2);
7326 } else {
7327 tmp = load_reg(s, rn);
7328 switch (op) {
7329 case 0x0a: /* rbit */
7330 gen_helper_rbit(tmp, tmp);
7331 break;
7332 case 0x08: /* rev */
7333 tcg_gen_bswap_i32(tmp, tmp);
7334 break;
7335 case 0x09: /* rev16 */
7336 gen_rev16(tmp);
7337 break;
7338 case 0x0b: /* revsh */
7339 gen_revsh(tmp);
7340 break;
7341 case 0x10: /* sel */
7342 tmp2 = load_reg(s, rm);
7343 tmp3 = new_tmp();
7344 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUState, GE));
7345 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
7346 dead_tmp(tmp3);
7347 dead_tmp(tmp2);
7348 break;
7349 case 0x18: /* clz */
7350 gen_helper_clz(tmp, tmp);
7351 break;
7352 default:
7353 goto illegal_op;
7354 }
7355 }
7356 store_reg(s, rd, tmp);
7357 break;
7358 case 4: case 5: /* 32-bit multiply. Sum of absolute differences. */
7359 op = (insn >> 4) & 0xf;
7360 tmp = load_reg(s, rn);
7361 tmp2 = load_reg(s, rm);
7362 switch ((insn >> 20) & 7) {
7363 case 0: /* 32 x 32 -> 32 */
7364 tcg_gen_mul_i32(tmp, tmp, tmp2);
7365 dead_tmp(tmp2);
7366 if (rs != 15) {
7367 tmp2 = load_reg(s, rs);
7368 if (op)
7369 tcg_gen_sub_i32(tmp, tmp2, tmp);
7370 else
7371 tcg_gen_add_i32(tmp, tmp, tmp2);
7372 dead_tmp(tmp2);
7373 }
7374 break;
7375 case 1: /* 16 x 16 -> 32 */
7376 gen_mulxy(tmp, tmp2, op & 2, op & 1);
7377 dead_tmp(tmp2);
7378 if (rs != 15) {
7379 tmp2 = load_reg(s, rs);
7380 gen_helper_add_setq(tmp, tmp, tmp2);
7381 dead_tmp(tmp2);
7382 }
7383 break;
7384 case 2: /* Dual multiply add. */
7385 case 4: /* Dual multiply subtract. */
7386 if (op)
7387 gen_swap_half(tmp2);
7388 gen_smul_dual(tmp, tmp2);
7389 /* This addition cannot overflow. */
7390 if (insn & (1 << 22)) {
7391 tcg_gen_sub_i32(tmp, tmp, tmp2);
7392 } else {
7393 tcg_gen_add_i32(tmp, tmp, tmp2);
7394 }
7395 dead_tmp(tmp2);
7396 if (rs != 15)
7397 {
7398 tmp2 = load_reg(s, rs);
7399 gen_helper_add_setq(tmp, tmp, tmp2);
7400 dead_tmp(tmp2);
7401 }
7402 break;
7403 case 3: /* 32 * 16 -> 32msb */
7404 if (op)
7405 tcg_gen_sari_i32(tmp2, tmp2, 16);
7406 else
7407 gen_sxth(tmp2);
7408 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7409 tcg_gen_shri_i64(tmp64, tmp64, 16);
7410 tmp = new_tmp();
7411 tcg_gen_trunc_i64_i32(tmp, tmp64);
7412 if (rs != 15)
7413 {
7414 tmp2 = load_reg(s, rs);
7415 gen_helper_add_setq(tmp, tmp, tmp2);
7416 dead_tmp(tmp2);
7417 }
7418 break;
7419 case 5: case 6: /* 32 * 32 -> 32msb */
7420 gen_imull(tmp, tmp2);
7421 if (insn & (1 << 5)) {
7422 gen_roundqd(tmp, tmp2);
7423 dead_tmp(tmp2);
7424 } else {
7425 dead_tmp(tmp);
7426 tmp = tmp2;
7427 }
7428 if (rs != 15) {
7429 tmp2 = load_reg(s, rs);
7430 if (insn & (1 << 21)) {
7431 tcg_gen_add_i32(tmp, tmp, tmp2);
7432 } else {
7433 tcg_gen_sub_i32(tmp, tmp2, tmp);
7434 }
7435 dead_tmp(tmp2);
7436 }
7437 break;
7438 case 7: /* Unsigned sum of absolute differences. */
7439 gen_helper_usad8(tmp, tmp, tmp2);
7440 dead_tmp(tmp2);
7441 if (rs != 15) {
7442 tmp2 = load_reg(s, rs);
7443 tcg_gen_add_i32(tmp, tmp, tmp2);
7444 dead_tmp(tmp2);
7445 }
7446 break;
7447 }
7448 store_reg(s, rd, tmp);
7449 break;
7450 case 6: case 7: /* 64-bit multiply, Divide. */
7451 op = ((insn >> 4) & 0xf) | ((insn >> 16) & 0x70);
7452 tmp = load_reg(s, rn);
7453 tmp2 = load_reg(s, rm);
7454 if ((op & 0x50) == 0x10) {
7455 /* sdiv, udiv */
7456 if (!arm_feature(env, ARM_FEATURE_DIV))
7457 goto illegal_op;
7458 if (op & 0x20)
7459 gen_helper_udiv(tmp, tmp, tmp2);
7460 else
7461 gen_helper_sdiv(tmp, tmp, tmp2);
7462 dead_tmp(tmp2);
7463 store_reg(s, rd, tmp);
7464 } else if ((op & 0xe) == 0xc) {
7465 /* Dual multiply accumulate long. */
7466 if (op & 1)
7467 gen_swap_half(tmp2);
7468 gen_smul_dual(tmp, tmp2);
7469 if (op & 0x10) {
7470 tcg_gen_sub_i32(tmp, tmp, tmp2);
7471 } else {
7472 tcg_gen_add_i32(tmp, tmp, tmp2);
7473 }
7474 dead_tmp(tmp2);
7475 /* BUGFIX */
7476 tmp64 = tcg_temp_new_i64();
7477 tcg_gen_ext_i32_i64(tmp64, tmp);
7478 dead_tmp(tmp);
7479 gen_addq(s, tmp64, rs, rd);
7480 gen_storeq_reg(s, rs, rd, tmp64);
7481 } else {
7482 if (op & 0x20) {
7483 /* Unsigned 64-bit multiply */
7484 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
7485 } else {
7486 if (op & 8) {
7487 /* smlalxy */
7488 gen_mulxy(tmp, tmp2, op & 2, op & 1);
7489 dead_tmp(tmp2);
7490 tmp64 = tcg_temp_new_i64();
7491 tcg_gen_ext_i32_i64(tmp64, tmp);
7492 dead_tmp(tmp);
7493 } else {
7494 /* Signed 64-bit multiply */
7495 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7496 }
7497 }
7498 if (op & 4) {
7499 /* umaal */
7500 gen_addq_lo(s, tmp64, rs);
7501 gen_addq_lo(s, tmp64, rd);
7502 } else if (op & 0x40) {
7503 /* 64-bit accumulate. */
7504 gen_addq(s, tmp64, rs, rd);
7505 }
7506 gen_storeq_reg(s, rs, rd, tmp64);
7507 }
7508 break;
7509 }
7510 break;
7511 case 6: case 7: case 14: case 15:
7512 /* Coprocessor. */
7513 if (((insn >> 24) & 3) == 3) {
7514 /* Translate into the equivalent ARM encoding. */
7515 insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4);
7516 if (disas_neon_data_insn(env, s, insn))
7517 goto illegal_op;
7518 } else {
7519 if (insn & (1 << 28))
7520 goto illegal_op;
7521 if (disas_coproc_insn (env, s, insn))
7522 goto illegal_op;
7523 }
7524 break;
7525 case 8: case 9: case 10: case 11:
7526 if (insn & (1 << 15)) {
7527 /* Branches, misc control. */
7528 if (insn & 0x5000) {
7529 /* Unconditional branch. */
7530 /* signextend(hw1[10:0]) -> offset[:12]. */
7531 offset = ((int32_t)insn << 5) >> 9 & ~(int32_t)0xfff;
7532 /* hw1[10:0] -> offset[11:1]. */
7533 offset |= (insn & 0x7ff) << 1;
7534 /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
7535 offset[24:22] already have the same value because of the
7536 sign extension above. */
7537 offset ^= ((~insn) & (1 << 13)) << 10;
7538 offset ^= ((~insn) & (1 << 11)) << 11;
7539
7540 if (insn & (1 << 14)) {
7541 /* Branch and link. */
7542 gen_op_movl_T1_im(s->pc | 1);
7543 gen_movl_reg_T1(s, 14);
7544 }
7545
7546 offset += s->pc;
7547 if (insn & (1 << 12)) {
7548 /* b/bl */
7549 gen_jmp(s, offset);
7550 } else {
7551 /* blx */
7552 offset &= ~(uint32_t)2;
7553 gen_bx_im(s, offset);
7554 }
7555 } else if (((insn >> 23) & 7) == 7) {
7556 /* Misc control */
7557 if (insn & (1 << 13))
7558 goto illegal_op;
7559
7560 if (insn & (1 << 26)) {
7561 /* Secure monitor call (v6Z) */
7562 goto illegal_op; /* not implemented. */
7563 } else {
7564 op = (insn >> 20) & 7;
7565 switch (op) {
7566 case 0: /* msr cpsr. */
7567 if (IS_M(env)) {
7568 tmp = load_reg(s, rn);
7569 addr = tcg_const_i32(insn & 0xff);
7570 gen_helper_v7m_msr(cpu_env, addr, tmp);
7571 gen_lookup_tb(s);
7572 break;
7573 }
7574 /* fall through */
7575 case 1: /* msr spsr. */
7576 if (IS_M(env))
7577 goto illegal_op;
7578 gen_movl_T0_reg(s, rn);
7579 if (gen_set_psr_T0(s,
7580 msr_mask(env, s, (insn >> 8) & 0xf, op == 1),
7581 op == 1))
7582 goto illegal_op;
7583 break;
7584 case 2: /* cps, nop-hint. */
7585 if (((insn >> 8) & 7) == 0) {
7586 gen_nop_hint(s, insn & 0xff);
7587 }
7588 /* Implemented as NOP in user mode. */
7589 if (IS_USER(s))
7590 break;
7591 offset = 0;
7592 imm = 0;
7593 if (insn & (1 << 10)) {
7594 if (insn & (1 << 7))
7595 offset |= CPSR_A;
7596 if (insn & (1 << 6))
7597 offset |= CPSR_I;
7598 if (insn & (1 << 5))
7599 offset |= CPSR_F;
7600 if (insn & (1 << 9))
7601 imm = CPSR_A | CPSR_I | CPSR_F;
7602 }
7603 if (insn & (1 << 8)) {
7604 offset |= 0x1f;
7605 imm |= (insn & 0x1f);
7606 }
7607 if (offset) {
7608 gen_op_movl_T0_im(imm);
7609 gen_set_psr_T0(s, offset, 0);
7610 }
7611 break;
7612 case 3: /* Special control operations. */
7613 op = (insn >> 4) & 0xf;
7614 switch (op) {
7615 case 2: /* clrex */
7616 gen_helper_clrex(cpu_env);
7617 break;
7618 case 4: /* dsb */
7619 case 5: /* dmb */
7620 case 6: /* isb */
7621 /* These execute as NOPs. */
7622 ARCH(7);
7623 break;
7624 default:
7625 goto illegal_op;
7626 }
7627 break;
7628 case 4: /* bxj */
7629 /* Trivial implementation equivalent to bx. */
7630 tmp = load_reg(s, rn);
7631 gen_bx(s, tmp);
7632 break;
7633 case 5: /* Exception return. */
7634 /* Unpredictable in user mode. */
7635 goto illegal_op;
7636 case 6: /* mrs cpsr. */
7637 tmp = new_tmp();
7638 if (IS_M(env)) {
7639 addr = tcg_const_i32(insn & 0xff);
7640 gen_helper_v7m_mrs(tmp, cpu_env, addr);
7641 } else {
7642 gen_helper_cpsr_read(tmp);
7643 }
7644 store_reg(s, rd, tmp);
7645 break;
7646 case 7: /* mrs spsr. */
7647 /* Not accessible in user mode. */
7648 if (IS_USER(s) || IS_M(env))
7649 goto illegal_op;
7650 tmp = load_cpu_field(spsr);
7651 store_reg(s, rd, tmp);
7652 break;
7653 }
7654 }
7655 } else {
7656 /* Conditional branch. */
7657 op = (insn >> 22) & 0xf;
7658 /* Generate a conditional jump to next instruction. */
7659 s->condlabel = gen_new_label();
7660 gen_test_cc(op ^ 1, s->condlabel);
7661 s->condjmp = 1;
7662
7663 /* offset[11:1] = insn[10:0] */
7664 offset = (insn & 0x7ff) << 1;
7665 /* offset[17:12] = insn[21:16]. */
7666 offset |= (insn & 0x003f0000) >> 4;
7667 /* offset[31:20] = insn[26]. */
7668 offset |= ((int32_t)((insn << 5) & 0x80000000)) >> 11;
7669 /* offset[18] = insn[13]. */
7670 offset |= (insn & (1 << 13)) << 5;
7671 /* offset[19] = insn[11]. */
7672 offset |= (insn & (1 << 11)) << 8;
7673
7674 /* jump to the offset */
7675 gen_jmp(s, s->pc + offset);
7676 }
7677 } else {
7678 /* Data processing immediate. */
7679 if (insn & (1 << 25)) {
7680 if (insn & (1 << 24)) {
7681 if (insn & (1 << 20))
7682 goto illegal_op;
7683 /* Bitfield/Saturate. */
7684 op = (insn >> 21) & 7;
7685 imm = insn & 0x1f;
7686 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
7687 if (rn == 15) {
7688 tmp = new_tmp();
7689 tcg_gen_movi_i32(tmp, 0);
7690 } else {
7691 tmp = load_reg(s, rn);
7692 }
7693 switch (op) {
7694 case 2: /* Signed bitfield extract. */
7695 imm++;
7696 if (shift + imm > 32)
7697 goto illegal_op;
7698 if (imm < 32)
7699 gen_sbfx(tmp, shift, imm);
7700 break;
7701 case 6: /* Unsigned bitfield extract. */
7702 imm++;
7703 if (shift + imm > 32)
7704 goto illegal_op;
7705 if (imm < 32)
7706 gen_ubfx(tmp, shift, (1u << imm) - 1);
7707 break;
7708 case 3: /* Bitfield insert/clear. */
7709 if (imm < shift)
7710 goto illegal_op;
7711 imm = imm + 1 - shift;
7712 if (imm != 32) {
7713 tmp2 = load_reg(s, rd);
7714 gen_bfi(tmp, tmp2, tmp, shift, (1u << imm) - 1);
7715 dead_tmp(tmp2);
7716 }
7717 break;
7718 case 7:
7719 goto illegal_op;
7720 default: /* Saturate. */
7721 if (shift) {
7722 if (op & 1)
7723 tcg_gen_sari_i32(tmp, tmp, shift);
7724 else
7725 tcg_gen_shli_i32(tmp, tmp, shift);
7726 }
7727 tmp2 = tcg_const_i32(imm);
7728 if (op & 4) {
7729 /* Unsigned. */
7730 if ((op & 1) && shift == 0)
7731 gen_helper_usat16(tmp, tmp, tmp2);
7732 else
7733 gen_helper_usat(tmp, tmp, tmp2);
7734 } else {
7735 /* Signed. */
7736 if ((op & 1) && shift == 0)
7737 gen_helper_ssat16(tmp, tmp, tmp2);
7738 else
7739 gen_helper_ssat(tmp, tmp, tmp2);
7740 }
7741 break;
7742 }
7743 store_reg(s, rd, tmp);
7744 } else {
7745 imm = ((insn & 0x04000000) >> 15)
7746 | ((insn & 0x7000) >> 4) | (insn & 0xff);
7747 if (insn & (1 << 22)) {
7748 /* 16-bit immediate. */
7749 imm |= (insn >> 4) & 0xf000;
7750 if (insn & (1 << 23)) {
7751 /* movt */
7752 tmp = load_reg(s, rd);
7753 tcg_gen_ext16u_i32(tmp, tmp);
7754 tcg_gen_ori_i32(tmp, tmp, imm << 16);
7755 } else {
7756 /* movw */
7757 tmp = new_tmp();
7758 tcg_gen_movi_i32(tmp, imm);
7759 }
7760 } else {
7761 /* Add/sub 12-bit immediate. */
7762 if (rn == 15) {
7763 offset = s->pc & ~(uint32_t)3;
7764 if (insn & (1 << 23))
7765 offset -= imm;
7766 else
7767 offset += imm;
7768 tmp = new_tmp();
7769 tcg_gen_movi_i32(tmp, offset);
7770 } else {
7771 tmp = load_reg(s, rn);
7772 if (insn & (1 << 23))
7773 tcg_gen_subi_i32(tmp, tmp, imm);
7774 else
7775 tcg_gen_addi_i32(tmp, tmp, imm);
7776 }
7777 }
7778 store_reg(s, rd, tmp);
7779 }
7780 } else {
7781 int shifter_out = 0;
7782 /* modified 12-bit immediate. */
7783 shift = ((insn & 0x04000000) >> 23) | ((insn & 0x7000) >> 12);
7784 imm = (insn & 0xff);
7785 switch (shift) {
7786 case 0: /* XY */
7787 /* Nothing to do. */
7788 break;
7789 case 1: /* 00XY00XY */
7790 imm |= imm << 16;
7791 break;
7792 case 2: /* XY00XY00 */
7793 imm |= imm << 16;
7794 imm <<= 8;
7795 break;
7796 case 3: /* XYXYXYXY */
7797 imm |= imm << 16;
7798 imm |= imm << 8;
7799 break;
7800 default: /* Rotated constant. */
7801 shift = (shift << 1) | (imm >> 7);
7802 imm |= 0x80;
7803 imm = imm << (32 - shift);
7804 shifter_out = 1;
7805 break;
7806 }
7807 gen_op_movl_T1_im(imm);
7808 rn = (insn >> 16) & 0xf;
7809 if (rn == 15)
7810 gen_op_movl_T0_im(0);
7811 else
7812 gen_movl_T0_reg(s, rn);
7813 op = (insn >> 21) & 0xf;
7814 if (gen_thumb2_data_op(s, op, (insn & (1 << 20)) != 0,
7815 shifter_out))
7816 goto illegal_op;
7817 rd = (insn >> 8) & 0xf;
7818 if (rd != 15) {
7819 gen_movl_reg_T0(s, rd);
7820 }
7821 }
7822 }
7823 break;
7824 case 12: /* Load/store single data item. */
7825 {
7826 int postinc = 0;
7827 int writeback = 0;
7828 int user;
7829 if ((insn & 0x01100000) == 0x01000000) {
7830 if (disas_neon_ls_insn(env, s, insn))
7831 goto illegal_op;
7832 break;
7833 }
7834 user = IS_USER(s);
7835 if (rn == 15) {
7836 addr = new_tmp();
7837 /* PC relative. */
7838 /* s->pc has already been incremented by 4. */
7839 imm = s->pc & 0xfffffffc;
7840 if (insn & (1 << 23))
7841 imm += insn & 0xfff;
7842 else
7843 imm -= insn & 0xfff;
7844 tcg_gen_movi_i32(addr, imm);
7845 } else {
7846 addr = load_reg(s, rn);
7847 if (insn & (1 << 23)) {
7848 /* Positive offset. */
7849 imm = insn & 0xfff;
7850 tcg_gen_addi_i32(addr, addr, imm);
7851 } else {
7852 op = (insn >> 8) & 7;
7853 imm = insn & 0xff;
7854 switch (op) {
7855 case 0: case 8: /* Shifted Register. */
7856 shift = (insn >> 4) & 0xf;
7857 if (shift > 3)
7858 goto illegal_op;
7859 tmp = load_reg(s, rm);
7860 if (shift)
7861 tcg_gen_shli_i32(tmp, tmp, shift);
7862 tcg_gen_add_i32(addr, addr, tmp);
7863 dead_tmp(tmp);
7864 break;
7865 case 4: /* Negative offset. */
7866 tcg_gen_addi_i32(addr, addr, -imm);
7867 break;
7868 case 6: /* User privilege. */
7869 tcg_gen_addi_i32(addr, addr, imm);
7870 user = 1;
7871 break;
7872 case 1: /* Post-decrement. */
7873 imm = -imm;
7874 /* Fall through. */
7875 case 3: /* Post-increment. */
7876 postinc = 1;
7877 writeback = 1;
7878 break;
7879 case 5: /* Pre-decrement. */
7880 imm = -imm;
7881 /* Fall through. */
7882 case 7: /* Pre-increment. */
7883 tcg_gen_addi_i32(addr, addr, imm);
7884 writeback = 1;
7885 break;
7886 default:
7887 goto illegal_op;
7888 }
7889 }
7890 }
7891 op = ((insn >> 21) & 3) | ((insn >> 22) & 4);
7892 if (insn & (1 << 20)) {
7893 /* Load. */
7894 if (rs == 15 && op != 2) {
7895 if (op & 2)
7896 goto illegal_op;
7897 /* Memory hint. Implemented as NOP. */
7898 } else {
7899 switch (op) {
7900 case 0: tmp = gen_ld8u(addr, user); break;
7901 case 4: tmp = gen_ld8s(addr, user); break;
7902 case 1: tmp = gen_ld16u(addr, user); break;
7903 case 5: tmp = gen_ld16s(addr, user); break;
7904 case 2: tmp = gen_ld32(addr, user); break;
7905 default: goto illegal_op;
7906 }
7907 if (rs == 15) {
7908 gen_bx(s, tmp);
7909 } else {
7910 store_reg(s, rs, tmp);
7911 }
7912 }
7913 } else {
7914 /* Store. */
7915 if (rs == 15)
7916 goto illegal_op;
7917 tmp = load_reg(s, rs);
7918 switch (op) {
7919 case 0: gen_st8(tmp, addr, user); break;
7920 case 1: gen_st16(tmp, addr, user); break;
7921 case 2: gen_st32(tmp, addr, user); break;
7922 default: goto illegal_op;
7923 }
7924 }
7925 if (postinc)
7926 tcg_gen_addi_i32(addr, addr, imm);
7927 if (writeback) {
7928 store_reg(s, rn, addr);
7929 } else {
7930 dead_tmp(addr);
7931 }
7932 }
7933 break;
7934 default:
7935 goto illegal_op;
7936 }
7937 return 0;
7938 illegal_op:
7939 return 1;
7940 }
7941
7942 static void disas_thumb_insn(CPUState *env, DisasContext *s)
7943 {
7944 uint32_t val, insn, op, rm, rn, rd, shift, cond;
7945 int32_t offset;
7946 int i;
7947 TCGv tmp;
7948 TCGv tmp2;
7949 TCGv addr;
7950
7951 if (s->condexec_mask) {
7952 cond = s->condexec_cond;
7953 s->condlabel = gen_new_label();
7954 gen_test_cc(cond ^ 1, s->condlabel);
7955 s->condjmp = 1;
7956 }
7957
7958 insn = lduw_code(s->pc);
7959 s->pc += 2;
7960
7961 switch (insn >> 12) {
7962 case 0: case 1:
7963 rd = insn & 7;
7964 op = (insn >> 11) & 3;
7965 if (op == 3) {
7966 /* add/subtract */
7967 rn = (insn >> 3) & 7;
7968 gen_movl_T0_reg(s, rn);
7969 if (insn & (1 << 10)) {
7970 /* immediate */
7971 gen_op_movl_T1_im((insn >> 6) & 7);
7972 } else {
7973 /* reg */
7974 rm = (insn >> 6) & 7;
7975 gen_movl_T1_reg(s, rm);
7976 }
7977 if (insn & (1 << 9)) {
7978 if (s->condexec_mask)
7979 gen_op_subl_T0_T1();
7980 else
7981 gen_op_subl_T0_T1_cc();
7982 } else {
7983 if (s->condexec_mask)
7984 gen_op_addl_T0_T1();
7985 else
7986 gen_op_addl_T0_T1_cc();
7987 }
7988 gen_movl_reg_T0(s, rd);
7989 } else {
7990 /* shift immediate */
7991 rm = (insn >> 3) & 7;
7992 shift = (insn >> 6) & 0x1f;
7993 tmp = load_reg(s, rm);
7994 gen_arm_shift_im(tmp, op, shift, s->condexec_mask == 0);
7995 if (!s->condexec_mask)
7996 gen_logic_CC(tmp);
7997 store_reg(s, rd, tmp);
7998 }
7999 break;
8000 case 2: case 3:
8001 /* arithmetic large immediate */
8002 op = (insn >> 11) & 3;
8003 rd = (insn >> 8) & 0x7;
8004 if (op == 0) {
8005 gen_op_movl_T0_im(insn & 0xff);
8006 } else {
8007 gen_movl_T0_reg(s, rd);
8008 gen_op_movl_T1_im(insn & 0xff);
8009 }
8010 switch (op) {
8011 case 0: /* mov */
8012 if (!s->condexec_mask)
8013 gen_op_logic_T0_cc();
8014 break;
8015 case 1: /* cmp */
8016 gen_op_subl_T0_T1_cc();
8017 break;
8018 case 2: /* add */
8019 if (s->condexec_mask)
8020 gen_op_addl_T0_T1();
8021 else
8022 gen_op_addl_T0_T1_cc();
8023 break;
8024 case 3: /* sub */
8025 if (s->condexec_mask)
8026 gen_op_subl_T0_T1();
8027 else
8028 gen_op_subl_T0_T1_cc();
8029 break;
8030 }
8031 if (op != 1)
8032 gen_movl_reg_T0(s, rd);
8033 break;
8034 case 4:
8035 if (insn & (1 << 11)) {
8036 rd = (insn >> 8) & 7;
8037 /* load pc-relative. Bit 1 of PC is ignored. */
8038 val = s->pc + 2 + ((insn & 0xff) * 4);
8039 val &= ~(uint32_t)2;
8040 addr = new_tmp();
8041 tcg_gen_movi_i32(addr, val);
8042 tmp = gen_ld32(addr, IS_USER(s));
8043 dead_tmp(addr);
8044 store_reg(s, rd, tmp);
8045 break;
8046 }
8047 if (insn & (1 << 10)) {
8048 /* data processing extended or blx */
8049 rd = (insn & 7) | ((insn >> 4) & 8);
8050 rm = (insn >> 3) & 0xf;
8051 op = (insn >> 8) & 3;
8052 switch (op) {
8053 case 0: /* add */
8054 gen_movl_T0_reg(s, rd);
8055 gen_movl_T1_reg(s, rm);
8056 gen_op_addl_T0_T1();
8057 gen_movl_reg_T0(s, rd);
8058 break;
8059 case 1: /* cmp */
8060 gen_movl_T0_reg(s, rd);
8061 gen_movl_T1_reg(s, rm);
8062 gen_op_subl_T0_T1_cc();
8063 break;
8064 case 2: /* mov/cpy */
8065 gen_movl_T0_reg(s, rm);
8066 gen_movl_reg_T0(s, rd);
8067 break;
8068 case 3:/* branch [and link] exchange thumb register */
8069 tmp = load_reg(s, rm);
8070 if (insn & (1 << 7)) {
8071 val = (uint32_t)s->pc | 1;
8072 tmp2 = new_tmp();
8073 tcg_gen_movi_i32(tmp2, val);
8074 store_reg(s, 14, tmp2);
8075 }
8076 gen_bx(s, tmp);
8077 break;
8078 }
8079 break;
8080 }
8081
8082 /* data processing register */
8083 rd = insn & 7;
8084 rm = (insn >> 3) & 7;
8085 op = (insn >> 6) & 0xf;
8086 if (op == 2 || op == 3 || op == 4 || op == 7) {
8087 /* the shift/rotate ops want the operands backwards */
8088 val = rm;
8089 rm = rd;
8090 rd = val;
8091 val = 1;
8092 } else {
8093 val = 0;
8094 }
8095
8096 if (op == 9) /* neg */
8097 gen_op_movl_T0_im(0);
8098 else if (op != 0xf) /* mvn doesn't read its first operand */
8099 gen_movl_T0_reg(s, rd);
8100
8101 gen_movl_T1_reg(s, rm);
8102 switch (op) {
8103 case 0x0: /* and */
8104 gen_op_andl_T0_T1();
8105 if (!s->condexec_mask)
8106 gen_op_logic_T0_cc();
8107 break;
8108 case 0x1: /* eor */
8109 gen_op_xorl_T0_T1();
8110 if (!s->condexec_mask)
8111 gen_op_logic_T0_cc();
8112 break;
8113 case 0x2: /* lsl */
8114 if (s->condexec_mask) {
8115 gen_helper_shl(cpu_T[1], cpu_T[1], cpu_T[0]);
8116 } else {
8117 gen_helper_shl_cc(cpu_T[1], cpu_T[1], cpu_T[0]);
8118 gen_op_logic_T1_cc();
8119 }
8120 break;
8121 case 0x3: /* lsr */
8122 if (s->condexec_mask) {
8123 gen_helper_shr(cpu_T[1], cpu_T[1], cpu_T[0]);
8124 } else {
8125 gen_helper_shr_cc(cpu_T[1], cpu_T[1], cpu_T[0]);
8126 gen_op_logic_T1_cc();
8127 }
8128 break;
8129 case 0x4: /* asr */
8130 if (s->condexec_mask) {
8131 gen_helper_sar(cpu_T[1], cpu_T[1], cpu_T[0]);
8132 } else {
8133 gen_helper_sar_cc(cpu_T[1], cpu_T[1], cpu_T[0]);
8134 gen_op_logic_T1_cc();
8135 }
8136 break;
8137 case 0x5: /* adc */
8138 if (s->condexec_mask)
8139 gen_adc_T0_T1();
8140 else
8141 gen_op_adcl_T0_T1_cc();
8142 break;
8143 case 0x6: /* sbc */
8144 if (s->condexec_mask)
8145 gen_sbc_T0_T1();
8146 else
8147 gen_op_sbcl_T0_T1_cc();
8148 break;
8149 case 0x7: /* ror */
8150 if (s->condexec_mask) {
8151 gen_helper_ror(cpu_T[1], cpu_T[1], cpu_T[0]);
8152 } else {
8153 gen_helper_ror_cc(cpu_T[1], cpu_T[1], cpu_T[0]);
8154 gen_op_logic_T1_cc();
8155 }
8156 break;
8157 case 0x8: /* tst */
8158 gen_op_andl_T0_T1();
8159 gen_op_logic_T0_cc();
8160 rd = 16;
8161 break;
8162 case 0x9: /* neg */
8163 if (s->condexec_mask)
8164 tcg_gen_neg_i32(cpu_T[0], cpu_T[1]);
8165 else
8166 gen_op_subl_T0_T1_cc();
8167 break;
8168 case 0xa: /* cmp */
8169 gen_op_subl_T0_T1_cc();
8170 rd = 16;
8171 break;
8172 case 0xb: /* cmn */
8173 gen_op_addl_T0_T1_cc();
8174 rd = 16;
8175 break;
8176 case 0xc: /* orr */
8177 gen_op_orl_T0_T1();
8178 if (!s->condexec_mask)
8179 gen_op_logic_T0_cc();
8180 break;
8181 case 0xd: /* mul */
8182 gen_op_mull_T0_T1();
8183 if (!s->condexec_mask)
8184 gen_op_logic_T0_cc();
8185 break;
8186 case 0xe: /* bic */
8187 gen_op_bicl_T0_T1();
8188 if (!s->condexec_mask)
8189 gen_op_logic_T0_cc();
8190 break;
8191 case 0xf: /* mvn */
8192 gen_op_notl_T1();
8193 if (!s->condexec_mask)
8194 gen_op_logic_T1_cc();
8195 val = 1;
8196 rm = rd;
8197 break;
8198 }
8199 if (rd != 16) {
8200 if (val)
8201 gen_movl_reg_T1(s, rm);
8202 else
8203 gen_movl_reg_T0(s, rd);
8204 }
8205 break;
8206
8207 case 5:
8208 /* load/store register offset. */
8209 rd = insn & 7;
8210 rn = (insn >> 3) & 7;
8211 rm = (insn >> 6) & 7;
8212 op = (insn >> 9) & 7;
8213 addr = load_reg(s, rn);
8214 tmp = load_reg(s, rm);
8215 tcg_gen_add_i32(addr, addr, tmp);
8216 dead_tmp(tmp);
8217
8218 if (op < 3) /* store */
8219 tmp = load_reg(s, rd);
8220
8221 switch (op) {
8222 case 0: /* str */
8223 gen_st32(tmp, addr, IS_USER(s));
8224 break;
8225 case 1: /* strh */
8226 gen_st16(tmp, addr, IS_USER(s));
8227 break;
8228 case 2: /* strb */
8229 gen_st8(tmp, addr, IS_USER(s));
8230 break;
8231 case 3: /* ldrsb */
8232 tmp = gen_ld8s(addr, IS_USER(s));
8233 break;
8234 case 4: /* ldr */
8235 tmp = gen_ld32(addr, IS_USER(s));
8236 break;
8237 case 5: /* ldrh */
8238 tmp = gen_ld16u(addr, IS_USER(s));
8239 break;
8240 case 6: /* ldrb */
8241 tmp = gen_ld8u(addr, IS_USER(s));
8242 break;
8243 case 7: /* ldrsh */
8244 tmp = gen_ld16s(addr, IS_USER(s));
8245 break;
8246 }
8247 if (op >= 3) /* load */
8248 store_reg(s, rd, tmp);
8249 dead_tmp(addr);
8250 break;
8251
8252 case 6:
8253 /* load/store word immediate offset */
8254 rd = insn & 7;
8255 rn = (insn >> 3) & 7;
8256 addr = load_reg(s, rn);
8257 val = (insn >> 4) & 0x7c;
8258 tcg_gen_addi_i32(addr, addr, val);
8259
8260 if (insn & (1 << 11)) {
8261 /* load */
8262 tmp = gen_ld32(addr, IS_USER(s));
8263 store_reg(s, rd, tmp);
8264 } else {
8265 /* store */
8266 tmp = load_reg(s, rd);
8267 gen_st32(tmp, addr, IS_USER(s));
8268 }
8269 dead_tmp(addr);
8270 break;
8271
8272 case 7:
8273 /* load/store byte immediate offset */
8274 rd = insn & 7;
8275 rn = (insn >> 3) & 7;
8276 addr = load_reg(s, rn);
8277 val = (insn >> 6) & 0x1f;
8278 tcg_gen_addi_i32(addr, addr, val);
8279
8280 if (insn & (1 << 11)) {
8281 /* load */
8282 tmp = gen_ld8u(addr, IS_USER(s));
8283 store_reg(s, rd, tmp);
8284 } else {
8285 /* store */
8286 tmp = load_reg(s, rd);
8287 gen_st8(tmp, addr, IS_USER(s));
8288 }
8289 dead_tmp(addr);
8290 break;
8291
8292 case 8:
8293 /* load/store halfword immediate offset */
8294 rd = insn & 7;
8295 rn = (insn >> 3) & 7;
8296 addr = load_reg(s, rn);
8297 val = (insn >> 5) & 0x3e;
8298 tcg_gen_addi_i32(addr, addr, val);
8299
8300 if (insn & (1 << 11)) {
8301 /* load */
8302 tmp = gen_ld16u(addr, IS_USER(s));
8303 store_reg(s, rd, tmp);
8304 } else {
8305 /* store */
8306 tmp = load_reg(s, rd);
8307 gen_st16(tmp, addr, IS_USER(s));
8308 }
8309 dead_tmp(addr);
8310 break;
8311
8312 case 9:
8313 /* load/store from stack */
8314 rd = (insn >> 8) & 7;
8315 addr = load_reg(s, 13);
8316 val = (insn & 0xff) * 4;
8317 tcg_gen_addi_i32(addr, addr, val);
8318
8319 if (insn & (1 << 11)) {
8320 /* load */
8321 tmp = gen_ld32(addr, IS_USER(s));
8322 store_reg(s, rd, tmp);
8323 } else {
8324 /* store */
8325 tmp = load_reg(s, rd);
8326 gen_st32(tmp, addr, IS_USER(s));
8327 }
8328 dead_tmp(addr);
8329 break;
8330
8331 case 10:
8332 /* add to high reg */
8333 rd = (insn >> 8) & 7;
8334 if (insn & (1 << 11)) {
8335 /* SP */
8336 tmp = load_reg(s, 13);
8337 } else {
8338 /* PC. bit 1 is ignored. */
8339 tmp = new_tmp();
8340 tcg_gen_movi_i32(tmp, (s->pc + 2) & ~(uint32_t)2);
8341 }
8342 val = (insn & 0xff) * 4;
8343 tcg_gen_addi_i32(tmp, tmp, val);
8344 store_reg(s, rd, tmp);
8345 break;
8346
8347 case 11:
8348 /* misc */
8349 op = (insn >> 8) & 0xf;
8350 switch (op) {
8351 case 0:
8352 /* adjust stack pointer */
8353 tmp = load_reg(s, 13);
8354 val = (insn & 0x7f) * 4;
8355 if (insn & (1 << 7))
8356 val = -(int32_t)val;
8357 tcg_gen_addi_i32(tmp, tmp, val);
8358 store_reg(s, 13, tmp);
8359 break;
8360
8361 case 2: /* sign/zero extend. */
8362 ARCH(6);
8363 rd = insn & 7;
8364 rm = (insn >> 3) & 7;
8365 tmp = load_reg(s, rm);
8366 switch ((insn >> 6) & 3) {
8367 case 0: gen_sxth(tmp); break;
8368 case 1: gen_sxtb(tmp); break;
8369 case 2: gen_uxth(tmp); break;
8370 case 3: gen_uxtb(tmp); break;
8371 }
8372 store_reg(s, rd, tmp);
8373 break;
8374 case 4: case 5: case 0xc: case 0xd:
8375 /* push/pop */
8376 addr = load_reg(s, 13);
8377 if (insn & (1 << 8))
8378 offset = 4;
8379 else
8380 offset = 0;
8381 for (i = 0; i < 8; i++) {
8382 if (insn & (1 << i))
8383 offset += 4;
8384 }
8385 if ((insn & (1 << 11)) == 0) {
8386 tcg_gen_addi_i32(addr, addr, -offset);
8387 }
8388 for (i = 0; i < 8; i++) {
8389 if (insn & (1 << i)) {
8390 if (insn & (1 << 11)) {
8391 /* pop */
8392 tmp = gen_ld32(addr, IS_USER(s));
8393 store_reg(s, i, tmp);
8394 } else {
8395 /* push */
8396 tmp = load_reg(s, i);
8397 gen_st32(tmp, addr, IS_USER(s));
8398 }
8399 /* advance to the next address. */
8400 tcg_gen_addi_i32(addr, addr, 4);
8401 }
8402 }
8403 TCGV_UNUSED(tmp);
8404 if (insn & (1 << 8)) {
8405 if (insn & (1 << 11)) {
8406 /* pop pc */
8407 tmp = gen_ld32(addr, IS_USER(s));
8408 /* don't set the pc until the rest of the instruction
8409 has completed */
8410 } else {
8411 /* push lr */
8412 tmp = load_reg(s, 14);
8413 gen_st32(tmp, addr, IS_USER(s));
8414 }
8415 tcg_gen_addi_i32(addr, addr, 4);
8416 }
8417 if ((insn & (1 << 11)) == 0) {
8418 tcg_gen_addi_i32(addr, addr, -offset);
8419 }
8420 /* write back the new stack pointer */
8421 store_reg(s, 13, addr);
8422 /* set the new PC value */
8423 if ((insn & 0x0900) == 0x0900)
8424 gen_bx(s, tmp);
8425 break;
8426
8427 case 1: case 3: case 9: case 11: /* czb */
8428 rm = insn & 7;
8429 tmp = load_reg(s, rm);
8430 s->condlabel = gen_new_label();
8431 s->condjmp = 1;
8432 if (insn & (1 << 11))
8433 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, s->condlabel);
8434 else
8435 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, s->condlabel);
8436 dead_tmp(tmp);
8437 offset = ((insn & 0xf8) >> 2) | (insn & 0x200) >> 3;
8438 val = (uint32_t)s->pc + 2;
8439 val += offset;
8440 gen_jmp(s, val);
8441 break;
8442
8443 case 15: /* IT, nop-hint. */
8444 if ((insn & 0xf) == 0) {
8445 gen_nop_hint(s, (insn >> 4) & 0xf);
8446 break;
8447 }
8448 /* If Then. */
8449 s->condexec_cond = (insn >> 4) & 0xe;
8450 s->condexec_mask = insn & 0x1f;
8451 /* No actual code generated for this insn, just setup state. */
8452 break;
8453
8454 case 0xe: /* bkpt */
8455 gen_set_condexec(s);
8456 gen_set_pc_im(s->pc - 2);
8457 gen_exception(EXCP_BKPT);
8458 s->is_jmp = DISAS_JUMP;
8459 break;
8460
8461 case 0xa: /* rev */
8462 ARCH(6);
8463 rn = (insn >> 3) & 0x7;
8464 rd = insn & 0x7;
8465 tmp = load_reg(s, rn);
8466 switch ((insn >> 6) & 3) {
8467 case 0: tcg_gen_bswap_i32(tmp, tmp); break;
8468 case 1: gen_rev16(tmp); break;
8469 case 3: gen_revsh(tmp); break;
8470 default: goto illegal_op;
8471 }
8472 store_reg(s, rd, tmp);
8473 break;
8474
8475 case 6: /* cps */
8476 ARCH(6);
8477 if (IS_USER(s))
8478 break;
8479 if (IS_M(env)) {
8480 tmp = tcg_const_i32((insn & (1 << 4)) != 0);
8481 /* PRIMASK */
8482 if (insn & 1) {
8483 addr = tcg_const_i32(16);
8484 gen_helper_v7m_msr(cpu_env, addr, tmp);
8485 }
8486 /* FAULTMASK */
8487 if (insn & 2) {
8488 addr = tcg_const_i32(17);
8489 gen_helper_v7m_msr(cpu_env, addr, tmp);
8490 }
8491 gen_lookup_tb(s);
8492 } else {
8493 if (insn & (1 << 4))
8494 shift = CPSR_A | CPSR_I | CPSR_F;
8495 else
8496 shift = 0;
8497
8498 val = ((insn & 7) << 6) & shift;
8499 gen_op_movl_T0_im(val);
8500 gen_set_psr_T0(s, shift, 0);
8501 }
8502 break;
8503
8504 default:
8505 goto undef;
8506 }
8507 break;
8508
8509 case 12:
8510 /* load/store multiple */
8511 rn = (insn >> 8) & 0x7;
8512 addr = load_reg(s, rn);
8513 for (i = 0; i < 8; i++) {
8514 if (insn & (1 << i)) {
8515 if (insn & (1 << 11)) {
8516 /* load */
8517 tmp = gen_ld32(addr, IS_USER(s));
8518 store_reg(s, i, tmp);
8519 } else {
8520 /* store */
8521 tmp = load_reg(s, i);
8522 gen_st32(tmp, addr, IS_USER(s));
8523 }
8524 /* advance to the next address */
8525 tcg_gen_addi_i32(addr, addr, 4);
8526 }
8527 }
8528 /* Base register writeback. */
8529 if ((insn & (1 << rn)) == 0) {
8530 store_reg(s, rn, addr);
8531 } else {
8532 dead_tmp(addr);
8533 }
8534 break;
8535
8536 case 13:
8537 /* conditional branch or swi */
8538 cond = (insn >> 8) & 0xf;
8539 if (cond == 0xe)
8540 goto undef;
8541
8542 if (cond == 0xf) {
8543 /* swi */
8544 gen_set_condexec(s);
8545 gen_set_pc_im(s->pc);
8546 s->is_jmp = DISAS_SWI;
8547 break;
8548 }
8549 /* generate a conditional jump to next instruction */
8550 s->condlabel = gen_new_label();
8551 gen_test_cc(cond ^ 1, s->condlabel);
8552 s->condjmp = 1;
8553 gen_movl_T1_reg(s, 15);
8554
8555 /* jump to the offset */
8556 val = (uint32_t)s->pc + 2;
8557 offset = ((int32_t)insn << 24) >> 24;
8558 val += offset << 1;
8559 gen_jmp(s, val);
8560 break;
8561
8562 case 14:
8563 if (insn & (1 << 11)) {
8564 if (disas_thumb2_insn(env, s, insn))
8565 goto undef32;
8566 break;
8567 }
8568 /* unconditional branch */
8569 val = (uint32_t)s->pc;
8570 offset = ((int32_t)insn << 21) >> 21;
8571 val += (offset << 1) + 2;
8572 gen_jmp(s, val);
8573 break;
8574
8575 case 15:
8576 if (disas_thumb2_insn(env, s, insn))
8577 goto undef32;
8578 break;
8579 }
8580 return;
8581 undef32:
8582 gen_set_condexec(s);
8583 gen_set_pc_im(s->pc - 4);
8584 gen_exception(EXCP_UDEF);
8585 s->is_jmp = DISAS_JUMP;
8586 return;
8587 illegal_op:
8588 undef:
8589 gen_set_condexec(s);
8590 gen_set_pc_im(s->pc - 2);
8591 gen_exception(EXCP_UDEF);
8592 s->is_jmp = DISAS_JUMP;
8593 }
8594
8595 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
8596 basic block 'tb'. If search_pc is TRUE, also generate PC
8597 information for each intermediate instruction. */
8598 static inline void gen_intermediate_code_internal(CPUState *env,
8599 TranslationBlock *tb,
8600 int search_pc)
8601 {
8602 DisasContext dc1, *dc = &dc1;
8603 uint16_t *gen_opc_end;
8604 int j, lj;
8605 target_ulong pc_start;
8606 uint32_t next_page_start;
8607 int num_insns;
8608 int max_insns;
8609
8610 /* generate intermediate code */
8611 num_temps = 0;
8612 memset(temps, 0, sizeof(temps));
8613
8614 pc_start = tb->pc;
8615
8616 dc->tb = tb;
8617
8618 gen_opc_end = gen_opc_buf + OPC_MAX_SIZE;
8619
8620 dc->is_jmp = DISAS_NEXT;
8621 dc->pc = pc_start;
8622 dc->singlestep_enabled = env->singlestep_enabled;
8623 dc->condjmp = 0;
8624 dc->thumb = env->thumb;
8625 dc->condexec_mask = (env->condexec_bits & 0xf) << 1;
8626 dc->condexec_cond = env->condexec_bits >> 4;
8627 dc->is_mem = 0;
8628 #if !defined(CONFIG_USER_ONLY)
8629 if (IS_M(env)) {
8630 dc->user = ((env->v7m.exception == 0) && (env->v7m.control & 1));
8631 } else {
8632 dc->user = (env->uncached_cpsr & 0x1f) == ARM_CPU_MODE_USR;
8633 }
8634 #endif
8635 cpu_F0s = tcg_temp_new_i32();
8636 cpu_F1s = tcg_temp_new_i32();
8637 cpu_F0d = tcg_temp_new_i64();
8638 cpu_F1d = tcg_temp_new_i64();
8639 cpu_V0 = cpu_F0d;
8640 cpu_V1 = cpu_F1d;
8641 /* FIXME: cpu_M0 can probably be the same as cpu_V0. */
8642 cpu_M0 = tcg_temp_new_i64();
8643 next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
8644 lj = -1;
8645 num_insns = 0;
8646 max_insns = tb->cflags & CF_COUNT_MASK;
8647 if (max_insns == 0)
8648 max_insns = CF_COUNT_MASK;
8649
8650 gen_icount_start();
8651 /* Reset the conditional execution bits immediately. This avoids
8652 complications trying to do it at the end of the block. */
8653 if (env->condexec_bits)
8654 {
8655 TCGv tmp = new_tmp();
8656 tcg_gen_movi_i32(tmp, 0);
8657 store_cpu_field(tmp, condexec_bits);
8658 }
8659 do {
8660 #ifdef CONFIG_USER_ONLY
8661 /* Intercept jump to the magic kernel page. */
8662 if (dc->pc >= 0xffff0000) {
8663 /* We always get here via a jump, so know we are not in a
8664 conditional execution block. */
8665 gen_exception(EXCP_KERNEL_TRAP);
8666 dc->is_jmp = DISAS_UPDATE;
8667 break;
8668 }
8669 #else
8670 if (dc->pc >= 0xfffffff0 && IS_M(env)) {
8671 /* We always get here via a jump, so know we are not in a
8672 conditional execution block. */
8673 gen_exception(EXCP_EXCEPTION_EXIT);
8674 dc->is_jmp = DISAS_UPDATE;
8675 break;
8676 }
8677 #endif
8678
8679 if (env->nb_breakpoints > 0) {
8680 for(j = 0; j < env->nb_breakpoints; j++) {
8681 if (env->breakpoints[j] == dc->pc) {
8682 gen_set_condexec(dc);
8683 gen_set_pc_im(dc->pc);
8684 gen_exception(EXCP_DEBUG);
8685 dc->is_jmp = DISAS_JUMP;
8686 /* Advance PC so that clearing the breakpoint will
8687 invalidate this TB. */
8688 dc->pc += 2;
8689 goto done_generating;
8690 break;
8691 }
8692 }
8693 }
8694 if (search_pc) {
8695 j = gen_opc_ptr - gen_opc_buf;
8696 if (lj < j) {
8697 lj++;
8698 while (lj < j)
8699 gen_opc_instr_start[lj++] = 0;
8700 }
8701 gen_opc_pc[lj] = dc->pc;
8702 gen_opc_instr_start[lj] = 1;
8703 gen_opc_icount[lj] = num_insns;
8704 }
8705
8706 if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO))
8707 gen_io_start();
8708
8709 if (env->thumb) {
8710 disas_thumb_insn(env, dc);
8711 if (dc->condexec_mask) {
8712 dc->condexec_cond = (dc->condexec_cond & 0xe)
8713 | ((dc->condexec_mask >> 4) & 1);
8714 dc->condexec_mask = (dc->condexec_mask << 1) & 0x1f;
8715 if (dc->condexec_mask == 0) {
8716 dc->condexec_cond = 0;
8717 }
8718 }
8719 } else {
8720 disas_arm_insn(env, dc);
8721 }
8722 if (num_temps) {
8723 fprintf(stderr, "Internal resource leak before %08x\n", dc->pc);
8724 num_temps = 0;
8725 }
8726
8727 if (dc->condjmp && !dc->is_jmp) {
8728 gen_set_label(dc->condlabel);
8729 dc->condjmp = 0;
8730 }
8731 /* Terminate the TB on memory ops if watchpoints are present. */
8732 /* FIXME: This should be replacd by the deterministic execution
8733 * IRQ raising bits. */
8734 if (dc->is_mem && env->nb_watchpoints)
8735 break;
8736
8737 /* Translation stops when a conditional branch is enoutered.
8738 * Otherwise the subsequent code could get translated several times.
8739 * Also stop translation when a page boundary is reached. This
8740 * ensures prefetch aborts occur at the right place. */
8741 num_insns ++;
8742 } while (!dc->is_jmp && gen_opc_ptr < gen_opc_end &&
8743 !env->singlestep_enabled &&
8744 dc->pc < next_page_start &&
8745 num_insns < max_insns);
8746
8747 if (tb->cflags & CF_LAST_IO) {
8748 if (dc->condjmp) {
8749 /* FIXME: This can theoretically happen with self-modifying
8750 code. */
8751 cpu_abort(env, "IO on conditional branch instruction");
8752 }
8753 gen_io_end();
8754 }
8755
8756 /* At this stage dc->condjmp will only be set when the skipped
8757 instruction was a conditional branch or trap, and the PC has
8758 already been written. */
8759 if (unlikely(env->singlestep_enabled)) {
8760 /* Make sure the pc is updated, and raise a debug exception. */
8761 if (dc->condjmp) {
8762 gen_set_condexec(dc);
8763 if (dc->is_jmp == DISAS_SWI) {
8764 gen_exception(EXCP_SWI);
8765 } else {
8766 gen_exception(EXCP_DEBUG);
8767 }
8768 gen_set_label(dc->condlabel);
8769 }
8770 if (dc->condjmp || !dc->is_jmp) {
8771 gen_set_pc_im(dc->pc);
8772 dc->condjmp = 0;
8773 }
8774 gen_set_condexec(dc);
8775 if (dc->is_jmp == DISAS_SWI && !dc->condjmp) {
8776 gen_exception(EXCP_SWI);
8777 } else {
8778 /* FIXME: Single stepping a WFI insn will not halt
8779 the CPU. */
8780 gen_exception(EXCP_DEBUG);
8781 }
8782 } else {
8783 /* While branches must always occur at the end of an IT block,
8784 there are a few other things that can cause us to terminate
8785 the TB in the middel of an IT block:
8786 - Exception generating instructions (bkpt, swi, undefined).
8787 - Page boundaries.
8788 - Hardware watchpoints.
8789 Hardware breakpoints have already been handled and skip this code.
8790 */
8791 gen_set_condexec(dc);
8792 switch(dc->is_jmp) {
8793 case DISAS_NEXT:
8794 gen_goto_tb(dc, 1, dc->pc);
8795 break;
8796 default:
8797 case DISAS_JUMP:
8798 case DISAS_UPDATE:
8799 /* indicate that the hash table must be used to find the next TB */
8800 tcg_gen_exit_tb(0);
8801 break;
8802 case DISAS_TB_JUMP:
8803 /* nothing more to generate */
8804 break;
8805 case DISAS_WFI:
8806 gen_helper_wfi();
8807 break;
8808 case DISAS_SWI:
8809 gen_exception(EXCP_SWI);
8810 break;
8811 }
8812 if (dc->condjmp) {
8813 gen_set_label(dc->condlabel);
8814 gen_set_condexec(dc);
8815 gen_goto_tb(dc, 1, dc->pc);
8816 dc->condjmp = 0;
8817 }
8818 }
8819
8820 done_generating:
8821 gen_icount_end(tb, num_insns);
8822 *gen_opc_ptr = INDEX_op_end;
8823
8824 #ifdef DEBUG_DISAS
8825 if (loglevel & CPU_LOG_TB_IN_ASM) {
8826 fprintf(logfile, "----------------\n");
8827 fprintf(logfile, "IN: %s\n", lookup_symbol(pc_start));
8828 target_disas(logfile, pc_start, dc->pc - pc_start, env->thumb);
8829 fprintf(logfile, "\n");
8830 }
8831 #endif
8832 if (search_pc) {
8833 j = gen_opc_ptr - gen_opc_buf;
8834 lj++;
8835 while (lj <= j)
8836 gen_opc_instr_start[lj++] = 0;
8837 } else {
8838 tb->size = dc->pc - pc_start;
8839 tb->icount = num_insns;
8840 }
8841 }
8842
8843 void gen_intermediate_code(CPUState *env, TranslationBlock *tb)
8844 {
8845 gen_intermediate_code_internal(env, tb, 0);
8846 }
8847
8848 void gen_intermediate_code_pc(CPUState *env, TranslationBlock *tb)
8849 {
8850 gen_intermediate_code_internal(env, tb, 1);
8851 }
8852
8853 static const char *cpu_mode_names[16] = {
8854 "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
8855 "???", "???", "???", "und", "???", "???", "???", "sys"
8856 };
8857
8858 void cpu_dump_state(CPUState *env, FILE *f,
8859 int (*cpu_fprintf)(FILE *f, const char *fmt, ...),
8860 int flags)
8861 {
8862 int i;
8863 #if 0
8864 union {
8865 uint32_t i;
8866 float s;
8867 } s0, s1;
8868 CPU_DoubleU d;
8869 /* ??? This assumes float64 and double have the same layout.
8870 Oh well, it's only debug dumps. */
8871 union {
8872 float64 f64;
8873 double d;
8874 } d0;
8875 #endif
8876 uint32_t psr;
8877
8878 for(i=0;i<16;i++) {
8879 cpu_fprintf(f, "R%02d=%08x", i, env->regs[i]);
8880 if ((i % 4) == 3)
8881 cpu_fprintf(f, "\n");
8882 else
8883 cpu_fprintf(f, " ");
8884 }
8885 psr = cpsr_read(env);
8886 cpu_fprintf(f, "PSR=%08x %c%c%c%c %c %s%d\n",
8887 psr,
8888 psr & (1 << 31) ? 'N' : '-',
8889 psr & (1 << 30) ? 'Z' : '-',
8890 psr & (1 << 29) ? 'C' : '-',
8891 psr & (1 << 28) ? 'V' : '-',
8892 psr & CPSR_T ? 'T' : 'A',
8893 cpu_mode_names[psr & 0xf], (psr & 0x10) ? 32 : 26);
8894
8895 #if 0
8896 for (i = 0; i < 16; i++) {
8897 d.d = env->vfp.regs[i];
8898 s0.i = d.l.lower;
8899 s1.i = d.l.upper;
8900 d0.f64 = d.d;
8901 cpu_fprintf(f, "s%02d=%08x(%8g) s%02d=%08x(%8g) d%02d=%08x%08x(%8g)\n",
8902 i * 2, (int)s0.i, s0.s,
8903 i * 2 + 1, (int)s1.i, s1.s,
8904 i, (int)(uint32_t)d.l.upper, (int)(uint32_t)d.l.lower,
8905 d0.d);
8906 }
8907 cpu_fprintf(f, "FPSCR: %08x\n", (int)env->vfp.xregs[ARM_VFP_FPSCR]);
8908 #endif
8909 }
8910
8911 void gen_pc_load(CPUState *env, TranslationBlock *tb,
8912 unsigned long searched_pc, int pc_pos, void *puc)
8913 {
8914 env->regs[15] = gen_opc_pc[pc_pos];
8915 }
Samsung s3c2440 and arm920t support for the mini2440 dev board