?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
a9mpcore: Convert to QOM realize
[qemu/ar7.git] / target-arm / translate.c
blob5f003e785e00a8bebc27baa75f5b809454e26ebc
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, see <http://www.gnu.org/licenses/>.
20 */
21 #include <stdarg.h>
22 #include <stdlib.h>
23 #include <stdio.h>
24 #include <string.h>
25 #include <inttypes.h>
26
27 #include "cpu.h"
28 #include "disas/disas.h"
29 #include "tcg-op.h"
30 #include "qemu/log.h"
31 #include "qemu/bitops.h"
32
33 #include "helper.h"
34 #define GEN_HELPER 1
35 #include "helper.h"
36
37 #define ENABLE_ARCH_4T arm_feature(env, ARM_FEATURE_V4T)
38 #define ENABLE_ARCH_5 arm_feature(env, ARM_FEATURE_V5)
39 /* currently all emulated v5 cores are also v5TE, so don't bother */
40 #define ENABLE_ARCH_5TE arm_feature(env, ARM_FEATURE_V5)
41 #define ENABLE_ARCH_5J 0
42 #define ENABLE_ARCH_6 arm_feature(env, ARM_FEATURE_V6)
43 #define ENABLE_ARCH_6K arm_feature(env, ARM_FEATURE_V6K)
44 #define ENABLE_ARCH_6T2 arm_feature(env, ARM_FEATURE_THUMB2)
45 #define ENABLE_ARCH_7 arm_feature(env, ARM_FEATURE_V7)
46 #define ENABLE_ARCH_8 arm_feature(env, ARM_FEATURE_V8)
47
48 #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0)
49
50 #include "translate.h"
51 static uint32_t gen_opc_condexec_bits[OPC_BUF_SIZE];
52
53 #if defined(CONFIG_USER_ONLY)
54 #define IS_USER(s) 1
55 #else
56 #define IS_USER(s) (s->user)
57 #endif
58
59 /* These instructions trap after executing, so defer them until after the
60 conditional execution state has been updated. */
61 #define DISAS_WFI 4
62 #define DISAS_SWI 5
63
64 TCGv_ptr cpu_env;
65 /* We reuse the same 64-bit temporaries for efficiency. */
66 static TCGv_i64 cpu_V0, cpu_V1, cpu_M0;
67 static TCGv_i32 cpu_R[16];
68 static TCGv_i32 cpu_CF, cpu_NF, cpu_VF, cpu_ZF;
69 static TCGv_i32 cpu_exclusive_addr;
70 static TCGv_i32 cpu_exclusive_val;
71 static TCGv_i32 cpu_exclusive_high;
72 #ifdef CONFIG_USER_ONLY
73 static TCGv_i32 cpu_exclusive_test;
74 static TCGv_i32 cpu_exclusive_info;
75 #endif
76
77 /* FIXME: These should be removed. */
78 static TCGv_i32 cpu_F0s, cpu_F1s;
79 static TCGv_i64 cpu_F0d, cpu_F1d;
80
81 #include "exec/gen-icount.h"
82
83 static const char *regnames[] =
84 { "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
85 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "pc" };
86
87 /* initialize TCG globals. */
88 void arm_translate_init(void)
89 {
90 int i;
91
92 cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
93
94 for (i = 0; i < 16; i++) {
95 cpu_R[i] = tcg_global_mem_new_i32(TCG_AREG0,
96 offsetof(CPUARMState, regs[i]),
97 regnames[i]);
98 }
99 cpu_CF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, CF), "CF");
100 cpu_NF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, NF), "NF");
101 cpu_VF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, VF), "VF");
102 cpu_ZF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, ZF), "ZF");
103
104 cpu_exclusive_addr = tcg_global_mem_new_i32(TCG_AREG0,
105 offsetof(CPUARMState, exclusive_addr), "exclusive_addr");
106 cpu_exclusive_val = tcg_global_mem_new_i32(TCG_AREG0,
107 offsetof(CPUARMState, exclusive_val), "exclusive_val");
108 cpu_exclusive_high = tcg_global_mem_new_i32(TCG_AREG0,
109 offsetof(CPUARMState, exclusive_high), "exclusive_high");
110 #ifdef CONFIG_USER_ONLY
111 cpu_exclusive_test = tcg_global_mem_new_i32(TCG_AREG0,
112 offsetof(CPUARMState, exclusive_test), "exclusive_test");
113 cpu_exclusive_info = tcg_global_mem_new_i32(TCG_AREG0,
114 offsetof(CPUARMState, exclusive_info), "exclusive_info");
115 #endif
116
117 a64_translate_init();
118 }
119
120 static inline TCGv_i32 load_cpu_offset(int offset)
121 {
122 TCGv_i32 tmp = tcg_temp_new_i32();
123 tcg_gen_ld_i32(tmp, cpu_env, offset);
124 return tmp;
125 }
126
127 #define load_cpu_field(name) load_cpu_offset(offsetof(CPUARMState, name))
128
129 static inline void store_cpu_offset(TCGv_i32 var, int offset)
130 {
131 tcg_gen_st_i32(var, cpu_env, offset);
132 tcg_temp_free_i32(var);
133 }
134
135 #define store_cpu_field(var, name) \
136 store_cpu_offset(var, offsetof(CPUARMState, name))
137
138 /* Set a variable to the value of a CPU register. */
139 static void load_reg_var(DisasContext *s, TCGv_i32 var, int reg)
140 {
141 if (reg == 15) {
142 uint32_t addr;
143 /* normally, since we updated PC, we need only to add one insn */
144 if (s->thumb)
145 addr = (long)s->pc + 2;
146 else
147 addr = (long)s->pc + 4;
148 tcg_gen_movi_i32(var, addr);
149 } else {
150 tcg_gen_mov_i32(var, cpu_R[reg]);
151 }
152 }
153
154 /* Create a new temporary and set it to the value of a CPU register. */
155 static inline TCGv_i32 load_reg(DisasContext *s, int reg)
156 {
157 TCGv_i32 tmp = tcg_temp_new_i32();
158 load_reg_var(s, tmp, reg);
159 return tmp;
160 }
161
162 /* Set a CPU register. The source must be a temporary and will be
163 marked as dead. */
164 static void store_reg(DisasContext *s, int reg, TCGv_i32 var)
165 {
166 if (reg == 15) {
167 tcg_gen_andi_i32(var, var, ~1);
168 s->is_jmp = DISAS_JUMP;
169 }
170 tcg_gen_mov_i32(cpu_R[reg], var);
171 tcg_temp_free_i32(var);
172 }
173
174 /* Value extensions. */
175 #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var)
176 #define gen_uxth(var) tcg_gen_ext16u_i32(var, var)
177 #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
178 #define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
179
180 #define gen_sxtb16(var) gen_helper_sxtb16(var, var)
181 #define gen_uxtb16(var) gen_helper_uxtb16(var, var)
182
183
184 static inline void gen_set_cpsr(TCGv_i32 var, uint32_t mask)
185 {
186 TCGv_i32 tmp_mask = tcg_const_i32(mask);
187 gen_helper_cpsr_write(cpu_env, var, tmp_mask);
188 tcg_temp_free_i32(tmp_mask);
189 }
190 /* Set NZCV flags from the high 4 bits of var. */
191 #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV)
192
193 static void gen_exception(int excp)
194 {
195 TCGv_i32 tmp = tcg_temp_new_i32();
196 tcg_gen_movi_i32(tmp, excp);
197 gen_helper_exception(cpu_env, tmp);
198 tcg_temp_free_i32(tmp);
199 }
200
201 static void gen_smul_dual(TCGv_i32 a, TCGv_i32 b)
202 {
203 TCGv_i32 tmp1 = tcg_temp_new_i32();
204 TCGv_i32 tmp2 = tcg_temp_new_i32();
205 tcg_gen_ext16s_i32(tmp1, a);
206 tcg_gen_ext16s_i32(tmp2, b);
207 tcg_gen_mul_i32(tmp1, tmp1, tmp2);
208 tcg_temp_free_i32(tmp2);
209 tcg_gen_sari_i32(a, a, 16);
210 tcg_gen_sari_i32(b, b, 16);
211 tcg_gen_mul_i32(b, b, a);
212 tcg_gen_mov_i32(a, tmp1);
213 tcg_temp_free_i32(tmp1);
214 }
215
216 /* Byteswap each halfword. */
217 static void gen_rev16(TCGv_i32 var)
218 {
219 TCGv_i32 tmp = tcg_temp_new_i32();
220 tcg_gen_shri_i32(tmp, var, 8);
221 tcg_gen_andi_i32(tmp, tmp, 0x00ff00ff);
222 tcg_gen_shli_i32(var, var, 8);
223 tcg_gen_andi_i32(var, var, 0xff00ff00);
224 tcg_gen_or_i32(var, var, tmp);
225 tcg_temp_free_i32(tmp);
226 }
227
228 /* Byteswap low halfword and sign extend. */
229 static void gen_revsh(TCGv_i32 var)
230 {
231 tcg_gen_ext16u_i32(var, var);
232 tcg_gen_bswap16_i32(var, var);
233 tcg_gen_ext16s_i32(var, var);
234 }
235
236 /* Unsigned bitfield extract. */
237 static void gen_ubfx(TCGv_i32 var, int shift, uint32_t mask)
238 {
239 if (shift)
240 tcg_gen_shri_i32(var, var, shift);
241 tcg_gen_andi_i32(var, var, mask);
242 }
243
244 /* Signed bitfield extract. */
245 static void gen_sbfx(TCGv_i32 var, int shift, int width)
246 {
247 uint32_t signbit;
248
249 if (shift)
250 tcg_gen_sari_i32(var, var, shift);
251 if (shift + width < 32) {
252 signbit = 1u << (width - 1);
253 tcg_gen_andi_i32(var, var, (1u << width) - 1);
254 tcg_gen_xori_i32(var, var, signbit);
255 tcg_gen_subi_i32(var, var, signbit);
256 }
257 }
258
259 /* Return (b << 32) + a. Mark inputs as dead */
260 static TCGv_i64 gen_addq_msw(TCGv_i64 a, TCGv_i32 b)
261 {
262 TCGv_i64 tmp64 = tcg_temp_new_i64();
263
264 tcg_gen_extu_i32_i64(tmp64, b);
265 tcg_temp_free_i32(b);
266 tcg_gen_shli_i64(tmp64, tmp64, 32);
267 tcg_gen_add_i64(a, tmp64, a);
268
269 tcg_temp_free_i64(tmp64);
270 return a;
271 }
272
273 /* Return (b << 32) - a. Mark inputs as dead. */
274 static TCGv_i64 gen_subq_msw(TCGv_i64 a, TCGv_i32 b)
275 {
276 TCGv_i64 tmp64 = tcg_temp_new_i64();
277
278 tcg_gen_extu_i32_i64(tmp64, b);
279 tcg_temp_free_i32(b);
280 tcg_gen_shli_i64(tmp64, tmp64, 32);
281 tcg_gen_sub_i64(a, tmp64, a);
282
283 tcg_temp_free_i64(tmp64);
284 return a;
285 }
286
287 /* 32x32->64 multiply. Marks inputs as dead. */
288 static TCGv_i64 gen_mulu_i64_i32(TCGv_i32 a, TCGv_i32 b)
289 {
290 TCGv_i32 lo = tcg_temp_new_i32();
291 TCGv_i32 hi = tcg_temp_new_i32();
292 TCGv_i64 ret;
293
294 tcg_gen_mulu2_i32(lo, hi, a, b);
295 tcg_temp_free_i32(a);
296 tcg_temp_free_i32(b);
297
298 ret = tcg_temp_new_i64();
299 tcg_gen_concat_i32_i64(ret, lo, hi);
300 tcg_temp_free_i32(lo);
301 tcg_temp_free_i32(hi);
302
303 return ret;
304 }
305
306 static TCGv_i64 gen_muls_i64_i32(TCGv_i32 a, TCGv_i32 b)
307 {
308 TCGv_i32 lo = tcg_temp_new_i32();
309 TCGv_i32 hi = tcg_temp_new_i32();
310 TCGv_i64 ret;
311
312 tcg_gen_muls2_i32(lo, hi, a, b);
313 tcg_temp_free_i32(a);
314 tcg_temp_free_i32(b);
315
316 ret = tcg_temp_new_i64();
317 tcg_gen_concat_i32_i64(ret, lo, hi);
318 tcg_temp_free_i32(lo);
319 tcg_temp_free_i32(hi);
320
321 return ret;
322 }
323
324 /* Swap low and high halfwords. */
325 static void gen_swap_half(TCGv_i32 var)
326 {
327 TCGv_i32 tmp = tcg_temp_new_i32();
328 tcg_gen_shri_i32(tmp, var, 16);
329 tcg_gen_shli_i32(var, var, 16);
330 tcg_gen_or_i32(var, var, tmp);
331 tcg_temp_free_i32(tmp);
332 }
333
334 /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
335 tmp = (t0 ^ t1) & 0x8000;
336 t0 &= ~0x8000;
337 t1 &= ~0x8000;
338 t0 = (t0 + t1) ^ tmp;
339 */
340
341 static void gen_add16(TCGv_i32 t0, TCGv_i32 t1)
342 {
343 TCGv_i32 tmp = tcg_temp_new_i32();
344 tcg_gen_xor_i32(tmp, t0, t1);
345 tcg_gen_andi_i32(tmp, tmp, 0x8000);
346 tcg_gen_andi_i32(t0, t0, ~0x8000);
347 tcg_gen_andi_i32(t1, t1, ~0x8000);
348 tcg_gen_add_i32(t0, t0, t1);
349 tcg_gen_xor_i32(t0, t0, tmp);
350 tcg_temp_free_i32(tmp);
351 tcg_temp_free_i32(t1);
352 }
353
354 /* Set CF to the top bit of var. */
355 static void gen_set_CF_bit31(TCGv_i32 var)
356 {
357 tcg_gen_shri_i32(cpu_CF, var, 31);
358 }
359
360 /* Set N and Z flags from var. */
361 static inline void gen_logic_CC(TCGv_i32 var)
362 {
363 tcg_gen_mov_i32(cpu_NF, var);
364 tcg_gen_mov_i32(cpu_ZF, var);
365 }
366
367 /* T0 += T1 + CF. */
368 static void gen_adc(TCGv_i32 t0, TCGv_i32 t1)
369 {
370 tcg_gen_add_i32(t0, t0, t1);
371 tcg_gen_add_i32(t0, t0, cpu_CF);
372 }
373
374 /* dest = T0 + T1 + CF. */
375 static void gen_add_carry(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
376 {
377 tcg_gen_add_i32(dest, t0, t1);
378 tcg_gen_add_i32(dest, dest, cpu_CF);
379 }
380
381 /* dest = T0 - T1 + CF - 1. */
382 static void gen_sub_carry(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
383 {
384 tcg_gen_sub_i32(dest, t0, t1);
385 tcg_gen_add_i32(dest, dest, cpu_CF);
386 tcg_gen_subi_i32(dest, dest, 1);
387 }
388
389 /* dest = T0 + T1. Compute C, N, V and Z flags */
390 static void gen_add_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
391 {
392 TCGv_i32 tmp = tcg_temp_new_i32();
393 tcg_gen_movi_i32(tmp, 0);
394 tcg_gen_add2_i32(cpu_NF, cpu_CF, t0, tmp, t1, tmp);
395 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
396 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0);
397 tcg_gen_xor_i32(tmp, t0, t1);
398 tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
399 tcg_temp_free_i32(tmp);
400 tcg_gen_mov_i32(dest, cpu_NF);
401 }
402
403 /* dest = T0 + T1 + CF. Compute C, N, V and Z flags */
404 static void gen_adc_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
405 {
406 TCGv_i32 tmp = tcg_temp_new_i32();
407 if (TCG_TARGET_HAS_add2_i32) {
408 tcg_gen_movi_i32(tmp, 0);
409 tcg_gen_add2_i32(cpu_NF, cpu_CF, t0, tmp, cpu_CF, tmp);
410 tcg_gen_add2_i32(cpu_NF, cpu_CF, cpu_NF, cpu_CF, t1, tmp);
411 } else {
412 TCGv_i64 q0 = tcg_temp_new_i64();
413 TCGv_i64 q1 = tcg_temp_new_i64();
414 tcg_gen_extu_i32_i64(q0, t0);
415 tcg_gen_extu_i32_i64(q1, t1);
416 tcg_gen_add_i64(q0, q0, q1);
417 tcg_gen_extu_i32_i64(q1, cpu_CF);
418 tcg_gen_add_i64(q0, q0, q1);
419 tcg_gen_extr_i64_i32(cpu_NF, cpu_CF, q0);
420 tcg_temp_free_i64(q0);
421 tcg_temp_free_i64(q1);
422 }
423 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
424 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0);
425 tcg_gen_xor_i32(tmp, t0, t1);
426 tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
427 tcg_temp_free_i32(tmp);
428 tcg_gen_mov_i32(dest, cpu_NF);
429 }
430
431 /* dest = T0 - T1. Compute C, N, V and Z flags */
432 static void gen_sub_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
433 {
434 TCGv_i32 tmp;
435 tcg_gen_sub_i32(cpu_NF, t0, t1);
436 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
437 tcg_gen_setcond_i32(TCG_COND_GEU, cpu_CF, t0, t1);
438 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0);
439 tmp = tcg_temp_new_i32();
440 tcg_gen_xor_i32(tmp, t0, t1);
441 tcg_gen_and_i32(cpu_VF, cpu_VF, tmp);
442 tcg_temp_free_i32(tmp);
443 tcg_gen_mov_i32(dest, cpu_NF);
444 }
445
446 /* dest = T0 + ~T1 + CF. Compute C, N, V and Z flags */
447 static void gen_sbc_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
448 {
449 TCGv_i32 tmp = tcg_temp_new_i32();
450 tcg_gen_not_i32(tmp, t1);
451 gen_adc_CC(dest, t0, tmp);
452 tcg_temp_free_i32(tmp);
453 }
454
455 #define GEN_SHIFT(name) \
456 static void gen_##name(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) \
457 { \
458 TCGv_i32 tmp1, tmp2, tmp3; \
459 tmp1 = tcg_temp_new_i32(); \
460 tcg_gen_andi_i32(tmp1, t1, 0xff); \
461 tmp2 = tcg_const_i32(0); \
462 tmp3 = tcg_const_i32(0x1f); \
463 tcg_gen_movcond_i32(TCG_COND_GTU, tmp2, tmp1, tmp3, tmp2, t0); \
464 tcg_temp_free_i32(tmp3); \
465 tcg_gen_andi_i32(tmp1, tmp1, 0x1f); \
466 tcg_gen_##name##_i32(dest, tmp2, tmp1); \
467 tcg_temp_free_i32(tmp2); \
468 tcg_temp_free_i32(tmp1); \
469 }
470 GEN_SHIFT(shl)
471 GEN_SHIFT(shr)
472 #undef GEN_SHIFT
473
474 static void gen_sar(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
475 {
476 TCGv_i32 tmp1, tmp2;
477 tmp1 = tcg_temp_new_i32();
478 tcg_gen_andi_i32(tmp1, t1, 0xff);
479 tmp2 = tcg_const_i32(0x1f);
480 tcg_gen_movcond_i32(TCG_COND_GTU, tmp1, tmp1, tmp2, tmp2, tmp1);
481 tcg_temp_free_i32(tmp2);
482 tcg_gen_sar_i32(dest, t0, tmp1);
483 tcg_temp_free_i32(tmp1);
484 }
485
486 static void tcg_gen_abs_i32(TCGv_i32 dest, TCGv_i32 src)
487 {
488 TCGv_i32 c0 = tcg_const_i32(0);
489 TCGv_i32 tmp = tcg_temp_new_i32();
490 tcg_gen_neg_i32(tmp, src);
491 tcg_gen_movcond_i32(TCG_COND_GT, dest, src, c0, src, tmp);
492 tcg_temp_free_i32(c0);
493 tcg_temp_free_i32(tmp);
494 }
495
496 static void shifter_out_im(TCGv_i32 var, int shift)
497 {
498 if (shift == 0) {
499 tcg_gen_andi_i32(cpu_CF, var, 1);
500 } else {
501 tcg_gen_shri_i32(cpu_CF, var, shift);
502 if (shift != 31) {
503 tcg_gen_andi_i32(cpu_CF, cpu_CF, 1);
504 }
505 }
506 }
507
508 /* Shift by immediate. Includes special handling for shift == 0. */
509 static inline void gen_arm_shift_im(TCGv_i32 var, int shiftop,
510 int shift, int flags)
511 {
512 switch (shiftop) {
513 case 0: /* LSL */
514 if (shift != 0) {
515 if (flags)
516 shifter_out_im(var, 32 - shift);
517 tcg_gen_shli_i32(var, var, shift);
518 }
519 break;
520 case 1: /* LSR */
521 if (shift == 0) {
522 if (flags) {
523 tcg_gen_shri_i32(cpu_CF, var, 31);
524 }
525 tcg_gen_movi_i32(var, 0);
526 } else {
527 if (flags)
528 shifter_out_im(var, shift - 1);
529 tcg_gen_shri_i32(var, var, shift);
530 }
531 break;
532 case 2: /* ASR */
533 if (shift == 0)
534 shift = 32;
535 if (flags)
536 shifter_out_im(var, shift - 1);
537 if (shift == 32)
538 shift = 31;
539 tcg_gen_sari_i32(var, var, shift);
540 break;
541 case 3: /* ROR/RRX */
542 if (shift != 0) {
543 if (flags)
544 shifter_out_im(var, shift - 1);
545 tcg_gen_rotri_i32(var, var, shift); break;
546 } else {
547 TCGv_i32 tmp = tcg_temp_new_i32();
548 tcg_gen_shli_i32(tmp, cpu_CF, 31);
549 if (flags)
550 shifter_out_im(var, 0);
551 tcg_gen_shri_i32(var, var, 1);
552 tcg_gen_or_i32(var, var, tmp);
553 tcg_temp_free_i32(tmp);
554 }
555 }
556 };
557
558 static inline void gen_arm_shift_reg(TCGv_i32 var, int shiftop,
559 TCGv_i32 shift, int flags)
560 {
561 if (flags) {
562 switch (shiftop) {
563 case 0: gen_helper_shl_cc(var, cpu_env, var, shift); break;
564 case 1: gen_helper_shr_cc(var, cpu_env, var, shift); break;
565 case 2: gen_helper_sar_cc(var, cpu_env, var, shift); break;
566 case 3: gen_helper_ror_cc(var, cpu_env, var, shift); break;
567 }
568 } else {
569 switch (shiftop) {
570 case 0:
571 gen_shl(var, var, shift);
572 break;
573 case 1:
574 gen_shr(var, var, shift);
575 break;
576 case 2:
577 gen_sar(var, var, shift);
578 break;
579 case 3: tcg_gen_andi_i32(shift, shift, 0x1f);
580 tcg_gen_rotr_i32(var, var, shift); break;
581 }
582 }
583 tcg_temp_free_i32(shift);
584 }
585
586 #define PAS_OP(pfx) \
587 switch (op2) { \
588 case 0: gen_pas_helper(glue(pfx,add16)); break; \
589 case 1: gen_pas_helper(glue(pfx,addsubx)); break; \
590 case 2: gen_pas_helper(glue(pfx,subaddx)); break; \
591 case 3: gen_pas_helper(glue(pfx,sub16)); break; \
592 case 4: gen_pas_helper(glue(pfx,add8)); break; \
593 case 7: gen_pas_helper(glue(pfx,sub8)); break; \
594 }
595 static void gen_arm_parallel_addsub(int op1, int op2, TCGv_i32 a, TCGv_i32 b)
596 {
597 TCGv_ptr tmp;
598
599 switch (op1) {
600 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
601 case 1:
602 tmp = tcg_temp_new_ptr();
603 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
604 PAS_OP(s)
605 tcg_temp_free_ptr(tmp);
606 break;
607 case 5:
608 tmp = tcg_temp_new_ptr();
609 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
610 PAS_OP(u)
611 tcg_temp_free_ptr(tmp);
612 break;
613 #undef gen_pas_helper
614 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
615 case 2:
616 PAS_OP(q);
617 break;
618 case 3:
619 PAS_OP(sh);
620 break;
621 case 6:
622 PAS_OP(uq);
623 break;
624 case 7:
625 PAS_OP(uh);
626 break;
627 #undef gen_pas_helper
628 }
629 }
630 #undef PAS_OP
631
632 /* For unknown reasons Arm and Thumb-2 use arbitrarily different encodings. */
633 #define PAS_OP(pfx) \
634 switch (op1) { \
635 case 0: gen_pas_helper(glue(pfx,add8)); break; \
636 case 1: gen_pas_helper(glue(pfx,add16)); break; \
637 case 2: gen_pas_helper(glue(pfx,addsubx)); break; \
638 case 4: gen_pas_helper(glue(pfx,sub8)); break; \
639 case 5: gen_pas_helper(glue(pfx,sub16)); break; \
640 case 6: gen_pas_helper(glue(pfx,subaddx)); break; \
641 }
642 static void gen_thumb2_parallel_addsub(int op1, int op2, TCGv_i32 a, TCGv_i32 b)
643 {
644 TCGv_ptr tmp;
645
646 switch (op2) {
647 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
648 case 0:
649 tmp = tcg_temp_new_ptr();
650 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
651 PAS_OP(s)
652 tcg_temp_free_ptr(tmp);
653 break;
654 case 4:
655 tmp = tcg_temp_new_ptr();
656 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
657 PAS_OP(u)
658 tcg_temp_free_ptr(tmp);
659 break;
660 #undef gen_pas_helper
661 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
662 case 1:
663 PAS_OP(q);
664 break;
665 case 2:
666 PAS_OP(sh);
667 break;
668 case 5:
669 PAS_OP(uq);
670 break;
671 case 6:
672 PAS_OP(uh);
673 break;
674 #undef gen_pas_helper
675 }
676 }
677 #undef PAS_OP
678
679 static void gen_test_cc(int cc, int label)
680 {
681 TCGv_i32 tmp;
682 int inv;
683
684 switch (cc) {
685 case 0: /* eq: Z */
686 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, label);
687 break;
688 case 1: /* ne: !Z */
689 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_ZF, 0, label);
690 break;
691 case 2: /* cs: C */
692 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_CF, 0, label);
693 break;
694 case 3: /* cc: !C */
695 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_CF, 0, label);
696 break;
697 case 4: /* mi: N */
698 tcg_gen_brcondi_i32(TCG_COND_LT, cpu_NF, 0, label);
699 break;
700 case 5: /* pl: !N */
701 tcg_gen_brcondi_i32(TCG_COND_GE, cpu_NF, 0, label);
702 break;
703 case 6: /* vs: V */
704 tcg_gen_brcondi_i32(TCG_COND_LT, cpu_VF, 0, label);
705 break;
706 case 7: /* vc: !V */
707 tcg_gen_brcondi_i32(TCG_COND_GE, cpu_VF, 0, label);
708 break;
709 case 8: /* hi: C && !Z */
710 inv = gen_new_label();
711 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_CF, 0, inv);
712 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_ZF, 0, label);
713 gen_set_label(inv);
714 break;
715 case 9: /* ls: !C || Z */
716 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_CF, 0, label);
717 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, label);
718 break;
719 case 10: /* ge: N == V -> N ^ V == 0 */
720 tmp = tcg_temp_new_i32();
721 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
722 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
723 tcg_temp_free_i32(tmp);
724 break;
725 case 11: /* lt: N != V -> N ^ V != 0 */
726 tmp = tcg_temp_new_i32();
727 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
728 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
729 tcg_temp_free_i32(tmp);
730 break;
731 case 12: /* gt: !Z && N == V */
732 inv = gen_new_label();
733 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, inv);
734 tmp = tcg_temp_new_i32();
735 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
736 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
737 tcg_temp_free_i32(tmp);
738 gen_set_label(inv);
739 break;
740 case 13: /* le: Z || N != V */
741 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, label);
742 tmp = tcg_temp_new_i32();
743 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
744 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
745 tcg_temp_free_i32(tmp);
746 break;
747 default:
748 fprintf(stderr, "Bad condition code 0x%x\n", cc);
749 abort();
750 }
751 }
752
753 static const uint8_t table_logic_cc[16] = {
754 1, /* and */
755 1, /* xor */
756 0, /* sub */
757 0, /* rsb */
758 0, /* add */
759 0, /* adc */
760 0, /* sbc */
761 0, /* rsc */
762 1, /* andl */
763 1, /* xorl */
764 0, /* cmp */
765 0, /* cmn */
766 1, /* orr */
767 1, /* mov */
768 1, /* bic */
769 1, /* mvn */
770 };
771
772 /* Set PC and Thumb state from an immediate address. */
773 static inline void gen_bx_im(DisasContext *s, uint32_t addr)
774 {
775 TCGv_i32 tmp;
776
777 s->is_jmp = DISAS_UPDATE;
778 if (s->thumb != (addr & 1)) {
779 tmp = tcg_temp_new_i32();
780 tcg_gen_movi_i32(tmp, addr & 1);
781 tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUARMState, thumb));
782 tcg_temp_free_i32(tmp);
783 }
784 tcg_gen_movi_i32(cpu_R[15], addr & ~1);
785 }
786
787 /* Set PC and Thumb state from var. var is marked as dead. */
788 static inline void gen_bx(DisasContext *s, TCGv_i32 var)
789 {
790 s->is_jmp = DISAS_UPDATE;
791 tcg_gen_andi_i32(cpu_R[15], var, ~1);
792 tcg_gen_andi_i32(var, var, 1);
793 store_cpu_field(var, thumb);
794 }
795
796 /* Variant of store_reg which uses branch&exchange logic when storing
797 to r15 in ARM architecture v7 and above. The source must be a temporary
798 and will be marked as dead. */
799 static inline void store_reg_bx(CPUARMState *env, DisasContext *s,
800 int reg, TCGv_i32 var)
801 {
802 if (reg == 15 && ENABLE_ARCH_7) {
803 gen_bx(s, var);
804 } else {
805 store_reg(s, reg, var);
806 }
807 }
808
809 /* Variant of store_reg which uses branch&exchange logic when storing
810 * to r15 in ARM architecture v5T and above. This is used for storing
811 * the results of a LDR/LDM/POP into r15, and corresponds to the cases
812 * in the ARM ARM which use the LoadWritePC() pseudocode function. */
813 static inline void store_reg_from_load(CPUARMState *env, DisasContext *s,
814 int reg, TCGv_i32 var)
815 {
816 if (reg == 15 && ENABLE_ARCH_5) {
817 gen_bx(s, var);
818 } else {
819 store_reg(s, reg, var);
820 }
821 }
822
823 /* Abstractions of "generate code to do a guest load/store for
824 * AArch32", where a vaddr is always 32 bits (and is zero
825 * extended if we're a 64 bit core) and data is also
826 * 32 bits unless specifically doing a 64 bit access.
827 * These functions work like tcg_gen_qemu_{ld,st}* except
828 * that their arguments are TCGv_i32 rather than TCGv.
829 */
830 #if TARGET_LONG_BITS == 32
831
832 #define DO_GEN_LD(OP) \
833 static inline void gen_aa32_##OP(TCGv_i32 val, TCGv_i32 addr, int index) \
834 { \
835 tcg_gen_qemu_##OP(val, addr, index); \
836 }
837
838 #define DO_GEN_ST(OP) \
839 static inline void gen_aa32_##OP(TCGv_i32 val, TCGv_i32 addr, int index) \
840 { \
841 tcg_gen_qemu_##OP(val, addr, index); \
842 }
843
844 static inline void gen_aa32_ld64(TCGv_i64 val, TCGv_i32 addr, int index)
845 {
846 tcg_gen_qemu_ld64(val, addr, index);
847 }
848
849 static inline void gen_aa32_st64(TCGv_i64 val, TCGv_i32 addr, int index)
850 {
851 tcg_gen_qemu_st64(val, addr, index);
852 }
853
854 #else
855
856 #define DO_GEN_LD(OP) \
857 static inline void gen_aa32_##OP(TCGv_i32 val, TCGv_i32 addr, int index) \
858 { \
859 TCGv addr64 = tcg_temp_new(); \
860 TCGv val64 = tcg_temp_new(); \
861 tcg_gen_extu_i32_i64(addr64, addr); \
862 tcg_gen_qemu_##OP(val64, addr64, index); \
863 tcg_temp_free(addr64); \
864 tcg_gen_trunc_i64_i32(val, val64); \
865 tcg_temp_free(val64); \
866 }
867
868 #define DO_GEN_ST(OP) \
869 static inline void gen_aa32_##OP(TCGv_i32 val, TCGv_i32 addr, int index) \
870 { \
871 TCGv addr64 = tcg_temp_new(); \
872 TCGv val64 = tcg_temp_new(); \
873 tcg_gen_extu_i32_i64(addr64, addr); \
874 tcg_gen_extu_i32_i64(val64, val); \
875 tcg_gen_qemu_##OP(val64, addr64, index); \
876 tcg_temp_free(addr64); \
877 tcg_temp_free(val64); \
878 }
879
880 static inline void gen_aa32_ld64(TCGv_i64 val, TCGv_i32 addr, int index)
881 {
882 TCGv addr64 = tcg_temp_new();
883 tcg_gen_extu_i32_i64(addr64, addr);
884 tcg_gen_qemu_ld64(val, addr64, index);
885 tcg_temp_free(addr64);
886 }
887
888 static inline void gen_aa32_st64(TCGv_i64 val, TCGv_i32 addr, int index)
889 {
890 TCGv addr64 = tcg_temp_new();
891 tcg_gen_extu_i32_i64(addr64, addr);
892 tcg_gen_qemu_st64(val, addr64, index);
893 tcg_temp_free(addr64);
894 }
895
896 #endif
897
898 DO_GEN_LD(ld8s)
899 DO_GEN_LD(ld8u)
900 DO_GEN_LD(ld16s)
901 DO_GEN_LD(ld16u)
902 DO_GEN_LD(ld32u)
903 DO_GEN_ST(st8)
904 DO_GEN_ST(st16)
905 DO_GEN_ST(st32)
906
907 static inline void gen_set_pc_im(DisasContext *s, target_ulong val)
908 {
909 if (s->aarch64) {
910 gen_a64_set_pc_im(val);
911 } else {
912 tcg_gen_movi_i32(cpu_R[15], val);
913 }
914 }
915
916 /* Force a TB lookup after an instruction that changes the CPU state. */
917 static inline void gen_lookup_tb(DisasContext *s)
918 {
919 tcg_gen_movi_i32(cpu_R[15], s->pc & ~1);
920 s->is_jmp = DISAS_UPDATE;
921 }
922
923 static inline void gen_add_data_offset(DisasContext *s, unsigned int insn,
924 TCGv_i32 var)
925 {
926 int val, rm, shift, shiftop;
927 TCGv_i32 offset;
928
929 if (!(insn & (1 << 25))) {
930 /* immediate */
931 val = insn & 0xfff;
932 if (!(insn & (1 << 23)))
933 val = -val;
934 if (val != 0)
935 tcg_gen_addi_i32(var, var, val);
936 } else {
937 /* shift/register */
938 rm = (insn) & 0xf;
939 shift = (insn >> 7) & 0x1f;
940 shiftop = (insn >> 5) & 3;
941 offset = load_reg(s, rm);
942 gen_arm_shift_im(offset, shiftop, shift, 0);
943 if (!(insn & (1 << 23)))
944 tcg_gen_sub_i32(var, var, offset);
945 else
946 tcg_gen_add_i32(var, var, offset);
947 tcg_temp_free_i32(offset);
948 }
949 }
950
951 static inline void gen_add_datah_offset(DisasContext *s, unsigned int insn,
952 int extra, TCGv_i32 var)
953 {
954 int val, rm;
955 TCGv_i32 offset;
956
957 if (insn & (1 << 22)) {
958 /* immediate */
959 val = (insn & 0xf) | ((insn >> 4) & 0xf0);
960 if (!(insn & (1 << 23)))
961 val = -val;
962 val += extra;
963 if (val != 0)
964 tcg_gen_addi_i32(var, var, val);
965 } else {
966 /* register */
967 if (extra)
968 tcg_gen_addi_i32(var, var, extra);
969 rm = (insn) & 0xf;
970 offset = load_reg(s, rm);
971 if (!(insn & (1 << 23)))
972 tcg_gen_sub_i32(var, var, offset);
973 else
974 tcg_gen_add_i32(var, var, offset);
975 tcg_temp_free_i32(offset);
976 }
977 }
978
979 static TCGv_ptr get_fpstatus_ptr(int neon)
980 {
981 TCGv_ptr statusptr = tcg_temp_new_ptr();
982 int offset;
983 if (neon) {
984 offset = offsetof(CPUARMState, vfp.standard_fp_status);
985 } else {
986 offset = offsetof(CPUARMState, vfp.fp_status);
987 }
988 tcg_gen_addi_ptr(statusptr, cpu_env, offset);
989 return statusptr;
990 }
991
992 #define VFP_OP2(name) \
993 static inline void gen_vfp_##name(int dp) \
994 { \
995 TCGv_ptr fpst = get_fpstatus_ptr(0); \
996 if (dp) { \
997 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, fpst); \
998 } else { \
999 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, fpst); \
1000 } \
1001 tcg_temp_free_ptr(fpst); \
1002 }
1003
1004 VFP_OP2(add)
1005 VFP_OP2(sub)
1006 VFP_OP2(mul)
1007 VFP_OP2(div)
1008
1009 #undef VFP_OP2
1010
1011 static inline void gen_vfp_F1_mul(int dp)
1012 {
1013 /* Like gen_vfp_mul() but put result in F1 */
1014 TCGv_ptr fpst = get_fpstatus_ptr(0);
1015 if (dp) {
1016 gen_helper_vfp_muld(cpu_F1d, cpu_F0d, cpu_F1d, fpst);
1017 } else {
1018 gen_helper_vfp_muls(cpu_F1s, cpu_F0s, cpu_F1s, fpst);
1019 }
1020 tcg_temp_free_ptr(fpst);
1021 }
1022
1023 static inline void gen_vfp_F1_neg(int dp)
1024 {
1025 /* Like gen_vfp_neg() but put result in F1 */
1026 if (dp) {
1027 gen_helper_vfp_negd(cpu_F1d, cpu_F0d);
1028 } else {
1029 gen_helper_vfp_negs(cpu_F1s, cpu_F0s);
1030 }
1031 }
1032
1033 static inline void gen_vfp_abs(int dp)
1034 {
1035 if (dp)
1036 gen_helper_vfp_absd(cpu_F0d, cpu_F0d);
1037 else
1038 gen_helper_vfp_abss(cpu_F0s, cpu_F0s);
1039 }
1040
1041 static inline void gen_vfp_neg(int dp)
1042 {
1043 if (dp)
1044 gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
1045 else
1046 gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
1047 }
1048
1049 static inline void gen_vfp_sqrt(int dp)
1050 {
1051 if (dp)
1052 gen_helper_vfp_sqrtd(cpu_F0d, cpu_F0d, cpu_env);
1053 else
1054 gen_helper_vfp_sqrts(cpu_F0s, cpu_F0s, cpu_env);
1055 }
1056
1057 static inline void gen_vfp_cmp(int dp)
1058 {
1059 if (dp)
1060 gen_helper_vfp_cmpd(cpu_F0d, cpu_F1d, cpu_env);
1061 else
1062 gen_helper_vfp_cmps(cpu_F0s, cpu_F1s, cpu_env);
1063 }
1064
1065 static inline void gen_vfp_cmpe(int dp)
1066 {
1067 if (dp)
1068 gen_helper_vfp_cmped(cpu_F0d, cpu_F1d, cpu_env);
1069 else
1070 gen_helper_vfp_cmpes(cpu_F0s, cpu_F1s, cpu_env);
1071 }
1072
1073 static inline void gen_vfp_F1_ld0(int dp)
1074 {
1075 if (dp)
1076 tcg_gen_movi_i64(cpu_F1d, 0);
1077 else
1078 tcg_gen_movi_i32(cpu_F1s, 0);
1079 }
1080
1081 #define VFP_GEN_ITOF(name) \
1082 static inline void gen_vfp_##name(int dp, int neon) \
1083 { \
1084 TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
1085 if (dp) { \
1086 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0s, statusptr); \
1087 } else { \
1088 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, statusptr); \
1089 } \
1090 tcg_temp_free_ptr(statusptr); \
1091 }
1092
1093 VFP_GEN_ITOF(uito)
1094 VFP_GEN_ITOF(sito)
1095 #undef VFP_GEN_ITOF
1096
1097 #define VFP_GEN_FTOI(name) \
1098 static inline void gen_vfp_##name(int dp, int neon) \
1099 { \
1100 TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
1101 if (dp) { \
1102 gen_helper_vfp_##name##d(cpu_F0s, cpu_F0d, statusptr); \
1103 } else { \
1104 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, statusptr); \
1105 } \
1106 tcg_temp_free_ptr(statusptr); \
1107 }
1108
1109 VFP_GEN_FTOI(toui)
1110 VFP_GEN_FTOI(touiz)
1111 VFP_GEN_FTOI(tosi)
1112 VFP_GEN_FTOI(tosiz)
1113 #undef VFP_GEN_FTOI
1114
1115 #define VFP_GEN_FIX(name) \
1116 static inline void gen_vfp_##name(int dp, int shift, int neon) \
1117 { \
1118 TCGv_i32 tmp_shift = tcg_const_i32(shift); \
1119 TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
1120 if (dp) { \
1121 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tmp_shift, statusptr); \
1122 } else { \
1123 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, tmp_shift, statusptr); \
1124 } \
1125 tcg_temp_free_i32(tmp_shift); \
1126 tcg_temp_free_ptr(statusptr); \
1127 }
1128 VFP_GEN_FIX(tosh)
1129 VFP_GEN_FIX(tosl)
1130 VFP_GEN_FIX(touh)
1131 VFP_GEN_FIX(toul)
1132 VFP_GEN_FIX(shto)
1133 VFP_GEN_FIX(slto)
1134 VFP_GEN_FIX(uhto)
1135 VFP_GEN_FIX(ulto)
1136 #undef VFP_GEN_FIX
1137
1138 static inline void gen_vfp_ld(DisasContext *s, int dp, TCGv_i32 addr)
1139 {
1140 if (dp) {
1141 gen_aa32_ld64(cpu_F0d, addr, IS_USER(s));
1142 } else {
1143 gen_aa32_ld32u(cpu_F0s, addr, IS_USER(s));
1144 }
1145 }
1146
1147 static inline void gen_vfp_st(DisasContext *s, int dp, TCGv_i32 addr)
1148 {
1149 if (dp) {
1150 gen_aa32_st64(cpu_F0d, addr, IS_USER(s));
1151 } else {
1152 gen_aa32_st32(cpu_F0s, addr, IS_USER(s));
1153 }
1154 }
1155
1156 static inline long
1157 vfp_reg_offset (int dp, int reg)
1158 {
1159 if (dp)
1160 return offsetof(CPUARMState, vfp.regs[reg]);
1161 else if (reg & 1) {
1162 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1163 + offsetof(CPU_DoubleU, l.upper);
1164 } else {
1165 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1166 + offsetof(CPU_DoubleU, l.lower);
1167 }
1168 }
1169
1170 /* Return the offset of a 32-bit piece of a NEON register.
1171 zero is the least significant end of the register. */
1172 static inline long
1173 neon_reg_offset (int reg, int n)
1174 {
1175 int sreg;
1176 sreg = reg * 2 + n;
1177 return vfp_reg_offset(0, sreg);
1178 }
1179
1180 static TCGv_i32 neon_load_reg(int reg, int pass)
1181 {
1182 TCGv_i32 tmp = tcg_temp_new_i32();
1183 tcg_gen_ld_i32(tmp, cpu_env, neon_reg_offset(reg, pass));
1184 return tmp;
1185 }
1186
1187 static void neon_store_reg(int reg, int pass, TCGv_i32 var)
1188 {
1189 tcg_gen_st_i32(var, cpu_env, neon_reg_offset(reg, pass));
1190 tcg_temp_free_i32(var);
1191 }
1192
1193 static inline void neon_load_reg64(TCGv_i64 var, int reg)
1194 {
1195 tcg_gen_ld_i64(var, cpu_env, vfp_reg_offset(1, reg));
1196 }
1197
1198 static inline void neon_store_reg64(TCGv_i64 var, int reg)
1199 {
1200 tcg_gen_st_i64(var, cpu_env, vfp_reg_offset(1, reg));
1201 }
1202
1203 #define tcg_gen_ld_f32 tcg_gen_ld_i32
1204 #define tcg_gen_ld_f64 tcg_gen_ld_i64
1205 #define tcg_gen_st_f32 tcg_gen_st_i32
1206 #define tcg_gen_st_f64 tcg_gen_st_i64
1207
1208 static inline void gen_mov_F0_vreg(int dp, int reg)
1209 {
1210 if (dp)
1211 tcg_gen_ld_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1212 else
1213 tcg_gen_ld_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1214 }
1215
1216 static inline void gen_mov_F1_vreg(int dp, int reg)
1217 {
1218 if (dp)
1219 tcg_gen_ld_f64(cpu_F1d, cpu_env, vfp_reg_offset(dp, reg));
1220 else
1221 tcg_gen_ld_f32(cpu_F1s, cpu_env, vfp_reg_offset(dp, reg));
1222 }
1223
1224 static inline void gen_mov_vreg_F0(int dp, int reg)
1225 {
1226 if (dp)
1227 tcg_gen_st_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1228 else
1229 tcg_gen_st_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1230 }
1231
1232 #define ARM_CP_RW_BIT (1 << 20)
1233
1234 static inline void iwmmxt_load_reg(TCGv_i64 var, int reg)
1235 {
1236 tcg_gen_ld_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg]));
1237 }
1238
1239 static inline void iwmmxt_store_reg(TCGv_i64 var, int reg)
1240 {
1241 tcg_gen_st_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg]));
1242 }
1243
1244 static inline TCGv_i32 iwmmxt_load_creg(int reg)
1245 {
1246 TCGv_i32 var = tcg_temp_new_i32();
1247 tcg_gen_ld_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg]));
1248 return var;
1249 }
1250
1251 static inline void iwmmxt_store_creg(int reg, TCGv_i32 var)
1252 {
1253 tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg]));
1254 tcg_temp_free_i32(var);
1255 }
1256
1257 static inline void gen_op_iwmmxt_movq_wRn_M0(int rn)
1258 {
1259 iwmmxt_store_reg(cpu_M0, rn);
1260 }
1261
1262 static inline void gen_op_iwmmxt_movq_M0_wRn(int rn)
1263 {
1264 iwmmxt_load_reg(cpu_M0, rn);
1265 }
1266
1267 static inline void gen_op_iwmmxt_orq_M0_wRn(int rn)
1268 {
1269 iwmmxt_load_reg(cpu_V1, rn);
1270 tcg_gen_or_i64(cpu_M0, cpu_M0, cpu_V1);
1271 }
1272
1273 static inline void gen_op_iwmmxt_andq_M0_wRn(int rn)
1274 {
1275 iwmmxt_load_reg(cpu_V1, rn);
1276 tcg_gen_and_i64(cpu_M0, cpu_M0, cpu_V1);
1277 }
1278
1279 static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn)
1280 {
1281 iwmmxt_load_reg(cpu_V1, rn);
1282 tcg_gen_xor_i64(cpu_M0, cpu_M0, cpu_V1);
1283 }
1284
1285 #define IWMMXT_OP(name) \
1286 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1287 { \
1288 iwmmxt_load_reg(cpu_V1, rn); \
1289 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
1290 }
1291
1292 #define IWMMXT_OP_ENV(name) \
1293 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1294 { \
1295 iwmmxt_load_reg(cpu_V1, rn); \
1296 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \
1297 }
1298
1299 #define IWMMXT_OP_ENV_SIZE(name) \
1300 IWMMXT_OP_ENV(name##b) \
1301 IWMMXT_OP_ENV(name##w) \
1302 IWMMXT_OP_ENV(name##l)
1303
1304 #define IWMMXT_OP_ENV1(name) \
1305 static inline void gen_op_iwmmxt_##name##_M0(void) \
1306 { \
1307 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \
1308 }
1309
1310 IWMMXT_OP(maddsq)
1311 IWMMXT_OP(madduq)
1312 IWMMXT_OP(sadb)
1313 IWMMXT_OP(sadw)
1314 IWMMXT_OP(mulslw)
1315 IWMMXT_OP(mulshw)
1316 IWMMXT_OP(mululw)
1317 IWMMXT_OP(muluhw)
1318 IWMMXT_OP(macsw)
1319 IWMMXT_OP(macuw)
1320
1321 IWMMXT_OP_ENV_SIZE(unpackl)
1322 IWMMXT_OP_ENV_SIZE(unpackh)
1323
1324 IWMMXT_OP_ENV1(unpacklub)
1325 IWMMXT_OP_ENV1(unpackluw)
1326 IWMMXT_OP_ENV1(unpacklul)
1327 IWMMXT_OP_ENV1(unpackhub)
1328 IWMMXT_OP_ENV1(unpackhuw)
1329 IWMMXT_OP_ENV1(unpackhul)
1330 IWMMXT_OP_ENV1(unpacklsb)
1331 IWMMXT_OP_ENV1(unpacklsw)
1332 IWMMXT_OP_ENV1(unpacklsl)
1333 IWMMXT_OP_ENV1(unpackhsb)
1334 IWMMXT_OP_ENV1(unpackhsw)
1335 IWMMXT_OP_ENV1(unpackhsl)
1336
1337 IWMMXT_OP_ENV_SIZE(cmpeq)
1338 IWMMXT_OP_ENV_SIZE(cmpgtu)
1339 IWMMXT_OP_ENV_SIZE(cmpgts)
1340
1341 IWMMXT_OP_ENV_SIZE(mins)
1342 IWMMXT_OP_ENV_SIZE(minu)
1343 IWMMXT_OP_ENV_SIZE(maxs)
1344 IWMMXT_OP_ENV_SIZE(maxu)
1345
1346 IWMMXT_OP_ENV_SIZE(subn)
1347 IWMMXT_OP_ENV_SIZE(addn)
1348 IWMMXT_OP_ENV_SIZE(subu)
1349 IWMMXT_OP_ENV_SIZE(addu)
1350 IWMMXT_OP_ENV_SIZE(subs)
1351 IWMMXT_OP_ENV_SIZE(adds)
1352
1353 IWMMXT_OP_ENV(avgb0)
1354 IWMMXT_OP_ENV(avgb1)
1355 IWMMXT_OP_ENV(avgw0)
1356 IWMMXT_OP_ENV(avgw1)
1357
1358 IWMMXT_OP(msadb)
1359
1360 IWMMXT_OP_ENV(packuw)
1361 IWMMXT_OP_ENV(packul)
1362 IWMMXT_OP_ENV(packuq)
1363 IWMMXT_OP_ENV(packsw)
1364 IWMMXT_OP_ENV(packsl)
1365 IWMMXT_OP_ENV(packsq)
1366
1367 static void gen_op_iwmmxt_set_mup(void)
1368 {
1369 TCGv_i32 tmp;
1370 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1371 tcg_gen_ori_i32(tmp, tmp, 2);
1372 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1373 }
1374
1375 static void gen_op_iwmmxt_set_cup(void)
1376 {
1377 TCGv_i32 tmp;
1378 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1379 tcg_gen_ori_i32(tmp, tmp, 1);
1380 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1381 }
1382
1383 static void gen_op_iwmmxt_setpsr_nz(void)
1384 {
1385 TCGv_i32 tmp = tcg_temp_new_i32();
1386 gen_helper_iwmmxt_setpsr_nz(tmp, cpu_M0);
1387 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCASF]);
1388 }
1389
1390 static inline void gen_op_iwmmxt_addl_M0_wRn(int rn)
1391 {
1392 iwmmxt_load_reg(cpu_V1, rn);
1393 tcg_gen_ext32u_i64(cpu_V1, cpu_V1);
1394 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1395 }
1396
1397 static inline int gen_iwmmxt_address(DisasContext *s, uint32_t insn,
1398 TCGv_i32 dest)
1399 {
1400 int rd;
1401 uint32_t offset;
1402 TCGv_i32 tmp;
1403
1404 rd = (insn >> 16) & 0xf;
1405 tmp = load_reg(s, rd);
1406
1407 offset = (insn & 0xff) << ((insn >> 7) & 2);
1408 if (insn & (1 << 24)) {
1409 /* Pre indexed */
1410 if (insn & (1 << 23))
1411 tcg_gen_addi_i32(tmp, tmp, offset);
1412 else
1413 tcg_gen_addi_i32(tmp, tmp, -offset);
1414 tcg_gen_mov_i32(dest, tmp);
1415 if (insn & (1 << 21))
1416 store_reg(s, rd, tmp);
1417 else
1418 tcg_temp_free_i32(tmp);
1419 } else if (insn & (1 << 21)) {
1420 /* Post indexed */
1421 tcg_gen_mov_i32(dest, tmp);
1422 if (insn & (1 << 23))
1423 tcg_gen_addi_i32(tmp, tmp, offset);
1424 else
1425 tcg_gen_addi_i32(tmp, tmp, -offset);
1426 store_reg(s, rd, tmp);
1427 } else if (!(insn & (1 << 23)))
1428 return 1;
1429 return 0;
1430 }
1431
1432 static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask, TCGv_i32 dest)
1433 {
1434 int rd = (insn >> 0) & 0xf;
1435 TCGv_i32 tmp;
1436
1437 if (insn & (1 << 8)) {
1438 if (rd < ARM_IWMMXT_wCGR0 || rd > ARM_IWMMXT_wCGR3) {
1439 return 1;
1440 } else {
1441 tmp = iwmmxt_load_creg(rd);
1442 }
1443 } else {
1444 tmp = tcg_temp_new_i32();
1445 iwmmxt_load_reg(cpu_V0, rd);
1446 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
1447 }
1448 tcg_gen_andi_i32(tmp, tmp, mask);
1449 tcg_gen_mov_i32(dest, tmp);
1450 tcg_temp_free_i32(tmp);
1451 return 0;
1452 }
1453
1454 /* Disassemble an iwMMXt instruction. Returns nonzero if an error occurred
1455 (ie. an undefined instruction). */
1456 static int disas_iwmmxt_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
1457 {
1458 int rd, wrd;
1459 int rdhi, rdlo, rd0, rd1, i;
1460 TCGv_i32 addr;
1461 TCGv_i32 tmp, tmp2, tmp3;
1462
1463 if ((insn & 0x0e000e00) == 0x0c000000) {
1464 if ((insn & 0x0fe00ff0) == 0x0c400000) {
1465 wrd = insn & 0xf;
1466 rdlo = (insn >> 12) & 0xf;
1467 rdhi = (insn >> 16) & 0xf;
1468 if (insn & ARM_CP_RW_BIT) { /* TMRRC */
1469 iwmmxt_load_reg(cpu_V0, wrd);
1470 tcg_gen_trunc_i64_i32(cpu_R[rdlo], cpu_V0);
1471 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
1472 tcg_gen_trunc_i64_i32(cpu_R[rdhi], cpu_V0);
1473 } else { /* TMCRR */
1474 tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]);
1475 iwmmxt_store_reg(cpu_V0, wrd);
1476 gen_op_iwmmxt_set_mup();
1477 }
1478 return 0;
1479 }
1480
1481 wrd = (insn >> 12) & 0xf;
1482 addr = tcg_temp_new_i32();
1483 if (gen_iwmmxt_address(s, insn, addr)) {
1484 tcg_temp_free_i32(addr);
1485 return 1;
1486 }
1487 if (insn & ARM_CP_RW_BIT) {
1488 if ((insn >> 28) == 0xf) { /* WLDRW wCx */
1489 tmp = tcg_temp_new_i32();
1490 gen_aa32_ld32u(tmp, addr, IS_USER(s));
1491 iwmmxt_store_creg(wrd, tmp);
1492 } else {
1493 i = 1;
1494 if (insn & (1 << 8)) {
1495 if (insn & (1 << 22)) { /* WLDRD */
1496 gen_aa32_ld64(cpu_M0, addr, IS_USER(s));
1497 i = 0;
1498 } else { /* WLDRW wRd */
1499 tmp = tcg_temp_new_i32();
1500 gen_aa32_ld32u(tmp, addr, IS_USER(s));
1501 }
1502 } else {
1503 tmp = tcg_temp_new_i32();
1504 if (insn & (1 << 22)) { /* WLDRH */
1505 gen_aa32_ld16u(tmp, addr, IS_USER(s));
1506 } else { /* WLDRB */
1507 gen_aa32_ld8u(tmp, addr, IS_USER(s));
1508 }
1509 }
1510 if (i) {
1511 tcg_gen_extu_i32_i64(cpu_M0, tmp);
1512 tcg_temp_free_i32(tmp);
1513 }
1514 gen_op_iwmmxt_movq_wRn_M0(wrd);
1515 }
1516 } else {
1517 if ((insn >> 28) == 0xf) { /* WSTRW wCx */
1518 tmp = iwmmxt_load_creg(wrd);
1519 gen_aa32_st32(tmp, addr, IS_USER(s));
1520 } else {
1521 gen_op_iwmmxt_movq_M0_wRn(wrd);
1522 tmp = tcg_temp_new_i32();
1523 if (insn & (1 << 8)) {
1524 if (insn & (1 << 22)) { /* WSTRD */
1525 gen_aa32_st64(cpu_M0, addr, IS_USER(s));
1526 } else { /* WSTRW wRd */
1527 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1528 gen_aa32_st32(tmp, addr, IS_USER(s));
1529 }
1530 } else {
1531 if (insn & (1 << 22)) { /* WSTRH */
1532 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1533 gen_aa32_st16(tmp, addr, IS_USER(s));
1534 } else { /* WSTRB */
1535 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1536 gen_aa32_st8(tmp, addr, IS_USER(s));
1537 }
1538 }
1539 }
1540 tcg_temp_free_i32(tmp);
1541 }
1542 tcg_temp_free_i32(addr);
1543 return 0;
1544 }
1545
1546 if ((insn & 0x0f000000) != 0x0e000000)
1547 return 1;
1548
1549 switch (((insn >> 12) & 0xf00) | ((insn >> 4) & 0xff)) {
1550 case 0x000: /* WOR */
1551 wrd = (insn >> 12) & 0xf;
1552 rd0 = (insn >> 0) & 0xf;
1553 rd1 = (insn >> 16) & 0xf;
1554 gen_op_iwmmxt_movq_M0_wRn(rd0);
1555 gen_op_iwmmxt_orq_M0_wRn(rd1);
1556 gen_op_iwmmxt_setpsr_nz();
1557 gen_op_iwmmxt_movq_wRn_M0(wrd);
1558 gen_op_iwmmxt_set_mup();
1559 gen_op_iwmmxt_set_cup();
1560 break;
1561 case 0x011: /* TMCR */
1562 if (insn & 0xf)
1563 return 1;
1564 rd = (insn >> 12) & 0xf;
1565 wrd = (insn >> 16) & 0xf;
1566 switch (wrd) {
1567 case ARM_IWMMXT_wCID:
1568 case ARM_IWMMXT_wCASF:
1569 break;
1570 case ARM_IWMMXT_wCon:
1571 gen_op_iwmmxt_set_cup();
1572 /* Fall through. */
1573 case ARM_IWMMXT_wCSSF:
1574 tmp = iwmmxt_load_creg(wrd);
1575 tmp2 = load_reg(s, rd);
1576 tcg_gen_andc_i32(tmp, tmp, tmp2);
1577 tcg_temp_free_i32(tmp2);
1578 iwmmxt_store_creg(wrd, tmp);
1579 break;
1580 case ARM_IWMMXT_wCGR0:
1581 case ARM_IWMMXT_wCGR1:
1582 case ARM_IWMMXT_wCGR2:
1583 case ARM_IWMMXT_wCGR3:
1584 gen_op_iwmmxt_set_cup();
1585 tmp = load_reg(s, rd);
1586 iwmmxt_store_creg(wrd, tmp);
1587 break;
1588 default:
1589 return 1;
1590 }
1591 break;
1592 case 0x100: /* WXOR */
1593 wrd = (insn >> 12) & 0xf;
1594 rd0 = (insn >> 0) & 0xf;
1595 rd1 = (insn >> 16) & 0xf;
1596 gen_op_iwmmxt_movq_M0_wRn(rd0);
1597 gen_op_iwmmxt_xorq_M0_wRn(rd1);
1598 gen_op_iwmmxt_setpsr_nz();
1599 gen_op_iwmmxt_movq_wRn_M0(wrd);
1600 gen_op_iwmmxt_set_mup();
1601 gen_op_iwmmxt_set_cup();
1602 break;
1603 case 0x111: /* TMRC */
1604 if (insn & 0xf)
1605 return 1;
1606 rd = (insn >> 12) & 0xf;
1607 wrd = (insn >> 16) & 0xf;
1608 tmp = iwmmxt_load_creg(wrd);
1609 store_reg(s, rd, tmp);
1610 break;
1611 case 0x300: /* WANDN */
1612 wrd = (insn >> 12) & 0xf;
1613 rd0 = (insn >> 0) & 0xf;
1614 rd1 = (insn >> 16) & 0xf;
1615 gen_op_iwmmxt_movq_M0_wRn(rd0);
1616 tcg_gen_neg_i64(cpu_M0, cpu_M0);
1617 gen_op_iwmmxt_andq_M0_wRn(rd1);
1618 gen_op_iwmmxt_setpsr_nz();
1619 gen_op_iwmmxt_movq_wRn_M0(wrd);
1620 gen_op_iwmmxt_set_mup();
1621 gen_op_iwmmxt_set_cup();
1622 break;
1623 case 0x200: /* WAND */
1624 wrd = (insn >> 12) & 0xf;
1625 rd0 = (insn >> 0) & 0xf;
1626 rd1 = (insn >> 16) & 0xf;
1627 gen_op_iwmmxt_movq_M0_wRn(rd0);
1628 gen_op_iwmmxt_andq_M0_wRn(rd1);
1629 gen_op_iwmmxt_setpsr_nz();
1630 gen_op_iwmmxt_movq_wRn_M0(wrd);
1631 gen_op_iwmmxt_set_mup();
1632 gen_op_iwmmxt_set_cup();
1633 break;
1634 case 0x810: case 0xa10: /* WMADD */
1635 wrd = (insn >> 12) & 0xf;
1636 rd0 = (insn >> 0) & 0xf;
1637 rd1 = (insn >> 16) & 0xf;
1638 gen_op_iwmmxt_movq_M0_wRn(rd0);
1639 if (insn & (1 << 21))
1640 gen_op_iwmmxt_maddsq_M0_wRn(rd1);
1641 else
1642 gen_op_iwmmxt_madduq_M0_wRn(rd1);
1643 gen_op_iwmmxt_movq_wRn_M0(wrd);
1644 gen_op_iwmmxt_set_mup();
1645 break;
1646 case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */
1647 wrd = (insn >> 12) & 0xf;
1648 rd0 = (insn >> 16) & 0xf;
1649 rd1 = (insn >> 0) & 0xf;
1650 gen_op_iwmmxt_movq_M0_wRn(rd0);
1651 switch ((insn >> 22) & 3) {
1652 case 0:
1653 gen_op_iwmmxt_unpacklb_M0_wRn(rd1);
1654 break;
1655 case 1:
1656 gen_op_iwmmxt_unpacklw_M0_wRn(rd1);
1657 break;
1658 case 2:
1659 gen_op_iwmmxt_unpackll_M0_wRn(rd1);
1660 break;
1661 case 3:
1662 return 1;
1663 }
1664 gen_op_iwmmxt_movq_wRn_M0(wrd);
1665 gen_op_iwmmxt_set_mup();
1666 gen_op_iwmmxt_set_cup();
1667 break;
1668 case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */
1669 wrd = (insn >> 12) & 0xf;
1670 rd0 = (insn >> 16) & 0xf;
1671 rd1 = (insn >> 0) & 0xf;
1672 gen_op_iwmmxt_movq_M0_wRn(rd0);
1673 switch ((insn >> 22) & 3) {
1674 case 0:
1675 gen_op_iwmmxt_unpackhb_M0_wRn(rd1);
1676 break;
1677 case 1:
1678 gen_op_iwmmxt_unpackhw_M0_wRn(rd1);
1679 break;
1680 case 2:
1681 gen_op_iwmmxt_unpackhl_M0_wRn(rd1);
1682 break;
1683 case 3:
1684 return 1;
1685 }
1686 gen_op_iwmmxt_movq_wRn_M0(wrd);
1687 gen_op_iwmmxt_set_mup();
1688 gen_op_iwmmxt_set_cup();
1689 break;
1690 case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */
1691 wrd = (insn >> 12) & 0xf;
1692 rd0 = (insn >> 16) & 0xf;
1693 rd1 = (insn >> 0) & 0xf;
1694 gen_op_iwmmxt_movq_M0_wRn(rd0);
1695 if (insn & (1 << 22))
1696 gen_op_iwmmxt_sadw_M0_wRn(rd1);
1697 else
1698 gen_op_iwmmxt_sadb_M0_wRn(rd1);
1699 if (!(insn & (1 << 20)))
1700 gen_op_iwmmxt_addl_M0_wRn(wrd);
1701 gen_op_iwmmxt_movq_wRn_M0(wrd);
1702 gen_op_iwmmxt_set_mup();
1703 break;
1704 case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */
1705 wrd = (insn >> 12) & 0xf;
1706 rd0 = (insn >> 16) & 0xf;
1707 rd1 = (insn >> 0) & 0xf;
1708 gen_op_iwmmxt_movq_M0_wRn(rd0);
1709 if (insn & (1 << 21)) {
1710 if (insn & (1 << 20))
1711 gen_op_iwmmxt_mulshw_M0_wRn(rd1);
1712 else
1713 gen_op_iwmmxt_mulslw_M0_wRn(rd1);
1714 } else {
1715 if (insn & (1 << 20))
1716 gen_op_iwmmxt_muluhw_M0_wRn(rd1);
1717 else
1718 gen_op_iwmmxt_mululw_M0_wRn(rd1);
1719 }
1720 gen_op_iwmmxt_movq_wRn_M0(wrd);
1721 gen_op_iwmmxt_set_mup();
1722 break;
1723 case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */
1724 wrd = (insn >> 12) & 0xf;
1725 rd0 = (insn >> 16) & 0xf;
1726 rd1 = (insn >> 0) & 0xf;
1727 gen_op_iwmmxt_movq_M0_wRn(rd0);
1728 if (insn & (1 << 21))
1729 gen_op_iwmmxt_macsw_M0_wRn(rd1);
1730 else
1731 gen_op_iwmmxt_macuw_M0_wRn(rd1);
1732 if (!(insn & (1 << 20))) {
1733 iwmmxt_load_reg(cpu_V1, wrd);
1734 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1735 }
1736 gen_op_iwmmxt_movq_wRn_M0(wrd);
1737 gen_op_iwmmxt_set_mup();
1738 break;
1739 case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */
1740 wrd = (insn >> 12) & 0xf;
1741 rd0 = (insn >> 16) & 0xf;
1742 rd1 = (insn >> 0) & 0xf;
1743 gen_op_iwmmxt_movq_M0_wRn(rd0);
1744 switch ((insn >> 22) & 3) {
1745 case 0:
1746 gen_op_iwmmxt_cmpeqb_M0_wRn(rd1);
1747 break;
1748 case 1:
1749 gen_op_iwmmxt_cmpeqw_M0_wRn(rd1);
1750 break;
1751 case 2:
1752 gen_op_iwmmxt_cmpeql_M0_wRn(rd1);
1753 break;
1754 case 3:
1755 return 1;
1756 }
1757 gen_op_iwmmxt_movq_wRn_M0(wrd);
1758 gen_op_iwmmxt_set_mup();
1759 gen_op_iwmmxt_set_cup();
1760 break;
1761 case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */
1762 wrd = (insn >> 12) & 0xf;
1763 rd0 = (insn >> 16) & 0xf;
1764 rd1 = (insn >> 0) & 0xf;
1765 gen_op_iwmmxt_movq_M0_wRn(rd0);
1766 if (insn & (1 << 22)) {
1767 if (insn & (1 << 20))
1768 gen_op_iwmmxt_avgw1_M0_wRn(rd1);
1769 else
1770 gen_op_iwmmxt_avgw0_M0_wRn(rd1);
1771 } else {
1772 if (insn & (1 << 20))
1773 gen_op_iwmmxt_avgb1_M0_wRn(rd1);
1774 else
1775 gen_op_iwmmxt_avgb0_M0_wRn(rd1);
1776 }
1777 gen_op_iwmmxt_movq_wRn_M0(wrd);
1778 gen_op_iwmmxt_set_mup();
1779 gen_op_iwmmxt_set_cup();
1780 break;
1781 case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */
1782 wrd = (insn >> 12) & 0xf;
1783 rd0 = (insn >> 16) & 0xf;
1784 rd1 = (insn >> 0) & 0xf;
1785 gen_op_iwmmxt_movq_M0_wRn(rd0);
1786 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCGR0 + ((insn >> 20) & 3));
1787 tcg_gen_andi_i32(tmp, tmp, 7);
1788 iwmmxt_load_reg(cpu_V1, rd1);
1789 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp);
1790 tcg_temp_free_i32(tmp);
1791 gen_op_iwmmxt_movq_wRn_M0(wrd);
1792 gen_op_iwmmxt_set_mup();
1793 break;
1794 case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */
1795 if (((insn >> 6) & 3) == 3)
1796 return 1;
1797 rd = (insn >> 12) & 0xf;
1798 wrd = (insn >> 16) & 0xf;
1799 tmp = load_reg(s, rd);
1800 gen_op_iwmmxt_movq_M0_wRn(wrd);
1801 switch ((insn >> 6) & 3) {
1802 case 0:
1803 tmp2 = tcg_const_i32(0xff);
1804 tmp3 = tcg_const_i32((insn & 7) << 3);
1805 break;
1806 case 1:
1807 tmp2 = tcg_const_i32(0xffff);
1808 tmp3 = tcg_const_i32((insn & 3) << 4);
1809 break;
1810 case 2:
1811 tmp2 = tcg_const_i32(0xffffffff);
1812 tmp3 = tcg_const_i32((insn & 1) << 5);
1813 break;
1814 default:
1815 TCGV_UNUSED_I32(tmp2);
1816 TCGV_UNUSED_I32(tmp3);
1817 }
1818 gen_helper_iwmmxt_insr(cpu_M0, cpu_M0, tmp, tmp2, tmp3);
1819 tcg_temp_free_i32(tmp3);
1820 tcg_temp_free_i32(tmp2);
1821 tcg_temp_free_i32(tmp);
1822 gen_op_iwmmxt_movq_wRn_M0(wrd);
1823 gen_op_iwmmxt_set_mup();
1824 break;
1825 case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */
1826 rd = (insn >> 12) & 0xf;
1827 wrd = (insn >> 16) & 0xf;
1828 if (rd == 15 || ((insn >> 22) & 3) == 3)
1829 return 1;
1830 gen_op_iwmmxt_movq_M0_wRn(wrd);
1831 tmp = tcg_temp_new_i32();
1832 switch ((insn >> 22) & 3) {
1833 case 0:
1834 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 7) << 3);
1835 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1836 if (insn & 8) {
1837 tcg_gen_ext8s_i32(tmp, tmp);
1838 } else {
1839 tcg_gen_andi_i32(tmp, tmp, 0xff);
1840 }
1841 break;
1842 case 1:
1843 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 3) << 4);
1844 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1845 if (insn & 8) {
1846 tcg_gen_ext16s_i32(tmp, tmp);
1847 } else {
1848 tcg_gen_andi_i32(tmp, tmp, 0xffff);
1849 }
1850 break;
1851 case 2:
1852 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 1) << 5);
1853 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1854 break;
1855 }
1856 store_reg(s, rd, tmp);
1857 break;
1858 case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */
1859 if ((insn & 0x000ff008) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1860 return 1;
1861 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1862 switch ((insn >> 22) & 3) {
1863 case 0:
1864 tcg_gen_shri_i32(tmp, tmp, ((insn & 7) << 2) + 0);
1865 break;
1866 case 1:
1867 tcg_gen_shri_i32(tmp, tmp, ((insn & 3) << 3) + 4);
1868 break;
1869 case 2:
1870 tcg_gen_shri_i32(tmp, tmp, ((insn & 1) << 4) + 12);
1871 break;
1872 }
1873 tcg_gen_shli_i32(tmp, tmp, 28);
1874 gen_set_nzcv(tmp);
1875 tcg_temp_free_i32(tmp);
1876 break;
1877 case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */
1878 if (((insn >> 6) & 3) == 3)
1879 return 1;
1880 rd = (insn >> 12) & 0xf;
1881 wrd = (insn >> 16) & 0xf;
1882 tmp = load_reg(s, rd);
1883 switch ((insn >> 6) & 3) {
1884 case 0:
1885 gen_helper_iwmmxt_bcstb(cpu_M0, tmp);
1886 break;
1887 case 1:
1888 gen_helper_iwmmxt_bcstw(cpu_M0, tmp);
1889 break;
1890 case 2:
1891 gen_helper_iwmmxt_bcstl(cpu_M0, tmp);
1892 break;
1893 }
1894 tcg_temp_free_i32(tmp);
1895 gen_op_iwmmxt_movq_wRn_M0(wrd);
1896 gen_op_iwmmxt_set_mup();
1897 break;
1898 case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */
1899 if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1900 return 1;
1901 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1902 tmp2 = tcg_temp_new_i32();
1903 tcg_gen_mov_i32(tmp2, tmp);
1904 switch ((insn >> 22) & 3) {
1905 case 0:
1906 for (i = 0; i < 7; i ++) {
1907 tcg_gen_shli_i32(tmp2, tmp2, 4);
1908 tcg_gen_and_i32(tmp, tmp, tmp2);
1909 }
1910 break;
1911 case 1:
1912 for (i = 0; i < 3; i ++) {
1913 tcg_gen_shli_i32(tmp2, tmp2, 8);
1914 tcg_gen_and_i32(tmp, tmp, tmp2);
1915 }
1916 break;
1917 case 2:
1918 tcg_gen_shli_i32(tmp2, tmp2, 16);
1919 tcg_gen_and_i32(tmp, tmp, tmp2);
1920 break;
1921 }
1922 gen_set_nzcv(tmp);
1923 tcg_temp_free_i32(tmp2);
1924 tcg_temp_free_i32(tmp);
1925 break;
1926 case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */
1927 wrd = (insn >> 12) & 0xf;
1928 rd0 = (insn >> 16) & 0xf;
1929 gen_op_iwmmxt_movq_M0_wRn(rd0);
1930 switch ((insn >> 22) & 3) {
1931 case 0:
1932 gen_helper_iwmmxt_addcb(cpu_M0, cpu_M0);
1933 break;
1934 case 1:
1935 gen_helper_iwmmxt_addcw(cpu_M0, cpu_M0);
1936 break;
1937 case 2:
1938 gen_helper_iwmmxt_addcl(cpu_M0, cpu_M0);
1939 break;
1940 case 3:
1941 return 1;
1942 }
1943 gen_op_iwmmxt_movq_wRn_M0(wrd);
1944 gen_op_iwmmxt_set_mup();
1945 break;
1946 case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */
1947 if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1948 return 1;
1949 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1950 tmp2 = tcg_temp_new_i32();
1951 tcg_gen_mov_i32(tmp2, tmp);
1952 switch ((insn >> 22) & 3) {
1953 case 0:
1954 for (i = 0; i < 7; i ++) {
1955 tcg_gen_shli_i32(tmp2, tmp2, 4);
1956 tcg_gen_or_i32(tmp, tmp, tmp2);
1957 }
1958 break;
1959 case 1:
1960 for (i = 0; i < 3; i ++) {
1961 tcg_gen_shli_i32(tmp2, tmp2, 8);
1962 tcg_gen_or_i32(tmp, tmp, tmp2);
1963 }
1964 break;
1965 case 2:
1966 tcg_gen_shli_i32(tmp2, tmp2, 16);
1967 tcg_gen_or_i32(tmp, tmp, tmp2);
1968 break;
1969 }
1970 gen_set_nzcv(tmp);
1971 tcg_temp_free_i32(tmp2);
1972 tcg_temp_free_i32(tmp);
1973 break;
1974 case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */
1975 rd = (insn >> 12) & 0xf;
1976 rd0 = (insn >> 16) & 0xf;
1977 if ((insn & 0xf) != 0 || ((insn >> 22) & 3) == 3)
1978 return 1;
1979 gen_op_iwmmxt_movq_M0_wRn(rd0);
1980 tmp = tcg_temp_new_i32();
1981 switch ((insn >> 22) & 3) {
1982 case 0:
1983 gen_helper_iwmmxt_msbb(tmp, cpu_M0);
1984 break;
1985 case 1:
1986 gen_helper_iwmmxt_msbw(tmp, cpu_M0);
1987 break;
1988 case 2:
1989 gen_helper_iwmmxt_msbl(tmp, cpu_M0);
1990 break;
1991 }
1992 store_reg(s, rd, tmp);
1993 break;
1994 case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */
1995 case 0x906: case 0xb06: case 0xd06: case 0xf06:
1996 wrd = (insn >> 12) & 0xf;
1997 rd0 = (insn >> 16) & 0xf;
1998 rd1 = (insn >> 0) & 0xf;
1999 gen_op_iwmmxt_movq_M0_wRn(rd0);
2000 switch ((insn >> 22) & 3) {
2001 case 0:
2002 if (insn & (1 << 21))
2003 gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1);
2004 else
2005 gen_op_iwmmxt_cmpgtub_M0_wRn(rd1);
2006 break;
2007 case 1:
2008 if (insn & (1 << 21))
2009 gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1);
2010 else
2011 gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1);
2012 break;
2013 case 2:
2014 if (insn & (1 << 21))
2015 gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1);
2016 else
2017 gen_op_iwmmxt_cmpgtul_M0_wRn(rd1);
2018 break;
2019 case 3:
2020 return 1;
2021 }
2022 gen_op_iwmmxt_movq_wRn_M0(wrd);
2023 gen_op_iwmmxt_set_mup();
2024 gen_op_iwmmxt_set_cup();
2025 break;
2026 case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */
2027 case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e:
2028 wrd = (insn >> 12) & 0xf;
2029 rd0 = (insn >> 16) & 0xf;
2030 gen_op_iwmmxt_movq_M0_wRn(rd0);
2031 switch ((insn >> 22) & 3) {
2032 case 0:
2033 if (insn & (1 << 21))
2034 gen_op_iwmmxt_unpacklsb_M0();
2035 else
2036 gen_op_iwmmxt_unpacklub_M0();
2037 break;
2038 case 1:
2039 if (insn & (1 << 21))
2040 gen_op_iwmmxt_unpacklsw_M0();
2041 else
2042 gen_op_iwmmxt_unpackluw_M0();
2043 break;
2044 case 2:
2045 if (insn & (1 << 21))
2046 gen_op_iwmmxt_unpacklsl_M0();
2047 else
2048 gen_op_iwmmxt_unpacklul_M0();
2049 break;
2050 case 3:
2051 return 1;
2052 }
2053 gen_op_iwmmxt_movq_wRn_M0(wrd);
2054 gen_op_iwmmxt_set_mup();
2055 gen_op_iwmmxt_set_cup();
2056 break;
2057 case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */
2058 case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c:
2059 wrd = (insn >> 12) & 0xf;
2060 rd0 = (insn >> 16) & 0xf;
2061 gen_op_iwmmxt_movq_M0_wRn(rd0);
2062 switch ((insn >> 22) & 3) {
2063 case 0:
2064 if (insn & (1 << 21))
2065 gen_op_iwmmxt_unpackhsb_M0();
2066 else
2067 gen_op_iwmmxt_unpackhub_M0();
2068 break;
2069 case 1:
2070 if (insn & (1 << 21))
2071 gen_op_iwmmxt_unpackhsw_M0();
2072 else
2073 gen_op_iwmmxt_unpackhuw_M0();
2074 break;
2075 case 2:
2076 if (insn & (1 << 21))
2077 gen_op_iwmmxt_unpackhsl_M0();
2078 else
2079 gen_op_iwmmxt_unpackhul_M0();
2080 break;
2081 case 3:
2082 return 1;
2083 }
2084 gen_op_iwmmxt_movq_wRn_M0(wrd);
2085 gen_op_iwmmxt_set_mup();
2086 gen_op_iwmmxt_set_cup();
2087 break;
2088 case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */
2089 case 0x214: case 0x614: case 0xa14: case 0xe14:
2090 if (((insn >> 22) & 3) == 0)
2091 return 1;
2092 wrd = (insn >> 12) & 0xf;
2093 rd0 = (insn >> 16) & 0xf;
2094 gen_op_iwmmxt_movq_M0_wRn(rd0);
2095 tmp = tcg_temp_new_i32();
2096 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2097 tcg_temp_free_i32(tmp);
2098 return 1;
2099 }
2100 switch ((insn >> 22) & 3) {
2101 case 1:
2102 gen_helper_iwmmxt_srlw(cpu_M0, cpu_env, cpu_M0, tmp);
2103 break;
2104 case 2:
2105 gen_helper_iwmmxt_srll(cpu_M0, cpu_env, cpu_M0, tmp);
2106 break;
2107 case 3:
2108 gen_helper_iwmmxt_srlq(cpu_M0, cpu_env, cpu_M0, tmp);
2109 break;
2110 }
2111 tcg_temp_free_i32(tmp);
2112 gen_op_iwmmxt_movq_wRn_M0(wrd);
2113 gen_op_iwmmxt_set_mup();
2114 gen_op_iwmmxt_set_cup();
2115 break;
2116 case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */
2117 case 0x014: case 0x414: case 0x814: case 0xc14:
2118 if (((insn >> 22) & 3) == 0)
2119 return 1;
2120 wrd = (insn >> 12) & 0xf;
2121 rd0 = (insn >> 16) & 0xf;
2122 gen_op_iwmmxt_movq_M0_wRn(rd0);
2123 tmp = tcg_temp_new_i32();
2124 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2125 tcg_temp_free_i32(tmp);
2126 return 1;
2127 }
2128 switch ((insn >> 22) & 3) {
2129 case 1:
2130 gen_helper_iwmmxt_sraw(cpu_M0, cpu_env, cpu_M0, tmp);
2131 break;
2132 case 2:
2133 gen_helper_iwmmxt_sral(cpu_M0, cpu_env, cpu_M0, tmp);
2134 break;
2135 case 3:
2136 gen_helper_iwmmxt_sraq(cpu_M0, cpu_env, cpu_M0, tmp);
2137 break;
2138 }
2139 tcg_temp_free_i32(tmp);
2140 gen_op_iwmmxt_movq_wRn_M0(wrd);
2141 gen_op_iwmmxt_set_mup();
2142 gen_op_iwmmxt_set_cup();
2143 break;
2144 case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */
2145 case 0x114: case 0x514: case 0x914: case 0xd14:
2146 if (((insn >> 22) & 3) == 0)
2147 return 1;
2148 wrd = (insn >> 12) & 0xf;
2149 rd0 = (insn >> 16) & 0xf;
2150 gen_op_iwmmxt_movq_M0_wRn(rd0);
2151 tmp = tcg_temp_new_i32();
2152 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2153 tcg_temp_free_i32(tmp);
2154 return 1;
2155 }
2156 switch ((insn >> 22) & 3) {
2157 case 1:
2158 gen_helper_iwmmxt_sllw(cpu_M0, cpu_env, cpu_M0, tmp);
2159 break;
2160 case 2:
2161 gen_helper_iwmmxt_slll(cpu_M0, cpu_env, cpu_M0, tmp);
2162 break;
2163 case 3:
2164 gen_helper_iwmmxt_sllq(cpu_M0, cpu_env, cpu_M0, tmp);
2165 break;
2166 }
2167 tcg_temp_free_i32(tmp);
2168 gen_op_iwmmxt_movq_wRn_M0(wrd);
2169 gen_op_iwmmxt_set_mup();
2170 gen_op_iwmmxt_set_cup();
2171 break;
2172 case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */
2173 case 0x314: case 0x714: case 0xb14: case 0xf14:
2174 if (((insn >> 22) & 3) == 0)
2175 return 1;
2176 wrd = (insn >> 12) & 0xf;
2177 rd0 = (insn >> 16) & 0xf;
2178 gen_op_iwmmxt_movq_M0_wRn(rd0);
2179 tmp = tcg_temp_new_i32();
2180 switch ((insn >> 22) & 3) {
2181 case 1:
2182 if (gen_iwmmxt_shift(insn, 0xf, tmp)) {
2183 tcg_temp_free_i32(tmp);
2184 return 1;
2185 }
2186 gen_helper_iwmmxt_rorw(cpu_M0, cpu_env, cpu_M0, tmp);
2187 break;
2188 case 2:
2189 if (gen_iwmmxt_shift(insn, 0x1f, tmp)) {
2190 tcg_temp_free_i32(tmp);
2191 return 1;
2192 }
2193 gen_helper_iwmmxt_rorl(cpu_M0, cpu_env, cpu_M0, tmp);
2194 break;
2195 case 3:
2196 if (gen_iwmmxt_shift(insn, 0x3f, tmp)) {
2197 tcg_temp_free_i32(tmp);
2198 return 1;
2199 }
2200 gen_helper_iwmmxt_rorq(cpu_M0, cpu_env, cpu_M0, tmp);
2201 break;
2202 }
2203 tcg_temp_free_i32(tmp);
2204 gen_op_iwmmxt_movq_wRn_M0(wrd);
2205 gen_op_iwmmxt_set_mup();
2206 gen_op_iwmmxt_set_cup();
2207 break;
2208 case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */
2209 case 0x916: case 0xb16: case 0xd16: case 0xf16:
2210 wrd = (insn >> 12) & 0xf;
2211 rd0 = (insn >> 16) & 0xf;
2212 rd1 = (insn >> 0) & 0xf;
2213 gen_op_iwmmxt_movq_M0_wRn(rd0);
2214 switch ((insn >> 22) & 3) {
2215 case 0:
2216 if (insn & (1 << 21))
2217 gen_op_iwmmxt_minsb_M0_wRn(rd1);
2218 else
2219 gen_op_iwmmxt_minub_M0_wRn(rd1);
2220 break;
2221 case 1:
2222 if (insn & (1 << 21))
2223 gen_op_iwmmxt_minsw_M0_wRn(rd1);
2224 else
2225 gen_op_iwmmxt_minuw_M0_wRn(rd1);
2226 break;
2227 case 2:
2228 if (insn & (1 << 21))
2229 gen_op_iwmmxt_minsl_M0_wRn(rd1);
2230 else
2231 gen_op_iwmmxt_minul_M0_wRn(rd1);
2232 break;
2233 case 3:
2234 return 1;
2235 }
2236 gen_op_iwmmxt_movq_wRn_M0(wrd);
2237 gen_op_iwmmxt_set_mup();
2238 break;
2239 case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */
2240 case 0x816: case 0xa16: case 0xc16: case 0xe16:
2241 wrd = (insn >> 12) & 0xf;
2242 rd0 = (insn >> 16) & 0xf;
2243 rd1 = (insn >> 0) & 0xf;
2244 gen_op_iwmmxt_movq_M0_wRn(rd0);
2245 switch ((insn >> 22) & 3) {
2246 case 0:
2247 if (insn & (1 << 21))
2248 gen_op_iwmmxt_maxsb_M0_wRn(rd1);
2249 else
2250 gen_op_iwmmxt_maxub_M0_wRn(rd1);
2251 break;
2252 case 1:
2253 if (insn & (1 << 21))
2254 gen_op_iwmmxt_maxsw_M0_wRn(rd1);
2255 else
2256 gen_op_iwmmxt_maxuw_M0_wRn(rd1);
2257 break;
2258 case 2:
2259 if (insn & (1 << 21))
2260 gen_op_iwmmxt_maxsl_M0_wRn(rd1);
2261 else
2262 gen_op_iwmmxt_maxul_M0_wRn(rd1);
2263 break;
2264 case 3:
2265 return 1;
2266 }
2267 gen_op_iwmmxt_movq_wRn_M0(wrd);
2268 gen_op_iwmmxt_set_mup();
2269 break;
2270 case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */
2271 case 0x402: case 0x502: case 0x602: case 0x702:
2272 wrd = (insn >> 12) & 0xf;
2273 rd0 = (insn >> 16) & 0xf;
2274 rd1 = (insn >> 0) & 0xf;
2275 gen_op_iwmmxt_movq_M0_wRn(rd0);
2276 tmp = tcg_const_i32((insn >> 20) & 3);
2277 iwmmxt_load_reg(cpu_V1, rd1);
2278 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp);
2279 tcg_temp_free_i32(tmp);
2280 gen_op_iwmmxt_movq_wRn_M0(wrd);
2281 gen_op_iwmmxt_set_mup();
2282 break;
2283 case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */
2284 case 0x41a: case 0x51a: case 0x61a: case 0x71a:
2285 case 0x81a: case 0x91a: case 0xa1a: case 0xb1a:
2286 case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a:
2287 wrd = (insn >> 12) & 0xf;
2288 rd0 = (insn >> 16) & 0xf;
2289 rd1 = (insn >> 0) & 0xf;
2290 gen_op_iwmmxt_movq_M0_wRn(rd0);
2291 switch ((insn >> 20) & 0xf) {
2292 case 0x0:
2293 gen_op_iwmmxt_subnb_M0_wRn(rd1);
2294 break;
2295 case 0x1:
2296 gen_op_iwmmxt_subub_M0_wRn(rd1);
2297 break;
2298 case 0x3:
2299 gen_op_iwmmxt_subsb_M0_wRn(rd1);
2300 break;
2301 case 0x4:
2302 gen_op_iwmmxt_subnw_M0_wRn(rd1);
2303 break;
2304 case 0x5:
2305 gen_op_iwmmxt_subuw_M0_wRn(rd1);
2306 break;
2307 case 0x7:
2308 gen_op_iwmmxt_subsw_M0_wRn(rd1);
2309 break;
2310 case 0x8:
2311 gen_op_iwmmxt_subnl_M0_wRn(rd1);
2312 break;
2313 case 0x9:
2314 gen_op_iwmmxt_subul_M0_wRn(rd1);
2315 break;
2316 case 0xb:
2317 gen_op_iwmmxt_subsl_M0_wRn(rd1);
2318 break;
2319 default:
2320 return 1;
2321 }
2322 gen_op_iwmmxt_movq_wRn_M0(wrd);
2323 gen_op_iwmmxt_set_mup();
2324 gen_op_iwmmxt_set_cup();
2325 break;
2326 case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */
2327 case 0x41e: case 0x51e: case 0x61e: case 0x71e:
2328 case 0x81e: case 0x91e: case 0xa1e: case 0xb1e:
2329 case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e:
2330 wrd = (insn >> 12) & 0xf;
2331 rd0 = (insn >> 16) & 0xf;
2332 gen_op_iwmmxt_movq_M0_wRn(rd0);
2333 tmp = tcg_const_i32(((insn >> 16) & 0xf0) | (insn & 0x0f));
2334 gen_helper_iwmmxt_shufh(cpu_M0, cpu_env, cpu_M0, tmp);
2335 tcg_temp_free_i32(tmp);
2336 gen_op_iwmmxt_movq_wRn_M0(wrd);
2337 gen_op_iwmmxt_set_mup();
2338 gen_op_iwmmxt_set_cup();
2339 break;
2340 case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */
2341 case 0x418: case 0x518: case 0x618: case 0x718:
2342 case 0x818: case 0x918: case 0xa18: case 0xb18:
2343 case 0xc18: case 0xd18: case 0xe18: case 0xf18:
2344 wrd = (insn >> 12) & 0xf;
2345 rd0 = (insn >> 16) & 0xf;
2346 rd1 = (insn >> 0) & 0xf;
2347 gen_op_iwmmxt_movq_M0_wRn(rd0);
2348 switch ((insn >> 20) & 0xf) {
2349 case 0x0:
2350 gen_op_iwmmxt_addnb_M0_wRn(rd1);
2351 break;
2352 case 0x1:
2353 gen_op_iwmmxt_addub_M0_wRn(rd1);
2354 break;
2355 case 0x3:
2356 gen_op_iwmmxt_addsb_M0_wRn(rd1);
2357 break;
2358 case 0x4:
2359 gen_op_iwmmxt_addnw_M0_wRn(rd1);
2360 break;
2361 case 0x5:
2362 gen_op_iwmmxt_adduw_M0_wRn(rd1);
2363 break;
2364 case 0x7:
2365 gen_op_iwmmxt_addsw_M0_wRn(rd1);
2366 break;
2367 case 0x8:
2368 gen_op_iwmmxt_addnl_M0_wRn(rd1);
2369 break;
2370 case 0x9:
2371 gen_op_iwmmxt_addul_M0_wRn(rd1);
2372 break;
2373 case 0xb:
2374 gen_op_iwmmxt_addsl_M0_wRn(rd1);
2375 break;
2376 default:
2377 return 1;
2378 }
2379 gen_op_iwmmxt_movq_wRn_M0(wrd);
2380 gen_op_iwmmxt_set_mup();
2381 gen_op_iwmmxt_set_cup();
2382 break;
2383 case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */
2384 case 0x408: case 0x508: case 0x608: case 0x708:
2385 case 0x808: case 0x908: case 0xa08: case 0xb08:
2386 case 0xc08: case 0xd08: case 0xe08: case 0xf08:
2387 if (!(insn & (1 << 20)) || ((insn >> 22) & 3) == 0)
2388 return 1;
2389 wrd = (insn >> 12) & 0xf;
2390 rd0 = (insn >> 16) & 0xf;
2391 rd1 = (insn >> 0) & 0xf;
2392 gen_op_iwmmxt_movq_M0_wRn(rd0);
2393 switch ((insn >> 22) & 3) {
2394 case 1:
2395 if (insn & (1 << 21))
2396 gen_op_iwmmxt_packsw_M0_wRn(rd1);
2397 else
2398 gen_op_iwmmxt_packuw_M0_wRn(rd1);
2399 break;
2400 case 2:
2401 if (insn & (1 << 21))
2402 gen_op_iwmmxt_packsl_M0_wRn(rd1);
2403 else
2404 gen_op_iwmmxt_packul_M0_wRn(rd1);
2405 break;
2406 case 3:
2407 if (insn & (1 << 21))
2408 gen_op_iwmmxt_packsq_M0_wRn(rd1);
2409 else
2410 gen_op_iwmmxt_packuq_M0_wRn(rd1);
2411 break;
2412 }
2413 gen_op_iwmmxt_movq_wRn_M0(wrd);
2414 gen_op_iwmmxt_set_mup();
2415 gen_op_iwmmxt_set_cup();
2416 break;
2417 case 0x201: case 0x203: case 0x205: case 0x207:
2418 case 0x209: case 0x20b: case 0x20d: case 0x20f:
2419 case 0x211: case 0x213: case 0x215: case 0x217:
2420 case 0x219: case 0x21b: case 0x21d: case 0x21f:
2421 wrd = (insn >> 5) & 0xf;
2422 rd0 = (insn >> 12) & 0xf;
2423 rd1 = (insn >> 0) & 0xf;
2424 if (rd0 == 0xf || rd1 == 0xf)
2425 return 1;
2426 gen_op_iwmmxt_movq_M0_wRn(wrd);
2427 tmp = load_reg(s, rd0);
2428 tmp2 = load_reg(s, rd1);
2429 switch ((insn >> 16) & 0xf) {
2430 case 0x0: /* TMIA */
2431 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2);
2432 break;
2433 case 0x8: /* TMIAPH */
2434 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2);
2435 break;
2436 case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */
2437 if (insn & (1 << 16))
2438 tcg_gen_shri_i32(tmp, tmp, 16);
2439 if (insn & (1 << 17))
2440 tcg_gen_shri_i32(tmp2, tmp2, 16);
2441 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2);
2442 break;
2443 default:
2444 tcg_temp_free_i32(tmp2);
2445 tcg_temp_free_i32(tmp);
2446 return 1;
2447 }
2448 tcg_temp_free_i32(tmp2);
2449 tcg_temp_free_i32(tmp);
2450 gen_op_iwmmxt_movq_wRn_M0(wrd);
2451 gen_op_iwmmxt_set_mup();
2452 break;
2453 default:
2454 return 1;
2455 }
2456
2457 return 0;
2458 }
2459
2460 /* Disassemble an XScale DSP instruction. Returns nonzero if an error occurred
2461 (ie. an undefined instruction). */
2462 static int disas_dsp_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
2463 {
2464 int acc, rd0, rd1, rdhi, rdlo;
2465 TCGv_i32 tmp, tmp2;
2466
2467 if ((insn & 0x0ff00f10) == 0x0e200010) {
2468 /* Multiply with Internal Accumulate Format */
2469 rd0 = (insn >> 12) & 0xf;
2470 rd1 = insn & 0xf;
2471 acc = (insn >> 5) & 7;
2472
2473 if (acc != 0)
2474 return 1;
2475
2476 tmp = load_reg(s, rd0);
2477 tmp2 = load_reg(s, rd1);
2478 switch ((insn >> 16) & 0xf) {
2479 case 0x0: /* MIA */
2480 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2);
2481 break;
2482 case 0x8: /* MIAPH */
2483 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2);
2484 break;
2485 case 0xc: /* MIABB */
2486 case 0xd: /* MIABT */
2487 case 0xe: /* MIATB */
2488 case 0xf: /* MIATT */
2489 if (insn & (1 << 16))
2490 tcg_gen_shri_i32(tmp, tmp, 16);
2491 if (insn & (1 << 17))
2492 tcg_gen_shri_i32(tmp2, tmp2, 16);
2493 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2);
2494 break;
2495 default:
2496 return 1;
2497 }
2498 tcg_temp_free_i32(tmp2);
2499 tcg_temp_free_i32(tmp);
2500
2501 gen_op_iwmmxt_movq_wRn_M0(acc);
2502 return 0;
2503 }
2504
2505 if ((insn & 0x0fe00ff8) == 0x0c400000) {
2506 /* Internal Accumulator Access Format */
2507 rdhi = (insn >> 16) & 0xf;
2508 rdlo = (insn >> 12) & 0xf;
2509 acc = insn & 7;
2510
2511 if (acc != 0)
2512 return 1;
2513
2514 if (insn & ARM_CP_RW_BIT) { /* MRA */
2515 iwmmxt_load_reg(cpu_V0, acc);
2516 tcg_gen_trunc_i64_i32(cpu_R[rdlo], cpu_V0);
2517 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
2518 tcg_gen_trunc_i64_i32(cpu_R[rdhi], cpu_V0);
2519 tcg_gen_andi_i32(cpu_R[rdhi], cpu_R[rdhi], (1 << (40 - 32)) - 1);
2520 } else { /* MAR */
2521 tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]);
2522 iwmmxt_store_reg(cpu_V0, acc);
2523 }
2524 return 0;
2525 }
2526
2527 return 1;
2528 }
2529
2530 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
2531 #define VFP_SREG(insn, bigbit, smallbit) \
2532 ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
2533 #define VFP_DREG(reg, insn, bigbit, smallbit) do { \
2534 if (arm_feature(env, ARM_FEATURE_VFP3)) { \
2535 reg = (((insn) >> (bigbit)) & 0x0f) \
2536 | (((insn) >> ((smallbit) - 4)) & 0x10); \
2537 } else { \
2538 if (insn & (1 << (smallbit))) \
2539 return 1; \
2540 reg = ((insn) >> (bigbit)) & 0x0f; \
2541 }} while (0)
2542
2543 #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
2544 #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
2545 #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7)
2546 #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7)
2547 #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5)
2548 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5)
2549
2550 /* Move between integer and VFP cores. */
2551 static TCGv_i32 gen_vfp_mrs(void)
2552 {
2553 TCGv_i32 tmp = tcg_temp_new_i32();
2554 tcg_gen_mov_i32(tmp, cpu_F0s);
2555 return tmp;
2556 }
2557
2558 static void gen_vfp_msr(TCGv_i32 tmp)
2559 {
2560 tcg_gen_mov_i32(cpu_F0s, tmp);
2561 tcg_temp_free_i32(tmp);
2562 }
2563
2564 static void gen_neon_dup_u8(TCGv_i32 var, int shift)
2565 {
2566 TCGv_i32 tmp = tcg_temp_new_i32();
2567 if (shift)
2568 tcg_gen_shri_i32(var, var, shift);
2569 tcg_gen_ext8u_i32(var, var);
2570 tcg_gen_shli_i32(tmp, var, 8);
2571 tcg_gen_or_i32(var, var, tmp);
2572 tcg_gen_shli_i32(tmp, var, 16);
2573 tcg_gen_or_i32(var, var, tmp);
2574 tcg_temp_free_i32(tmp);
2575 }
2576
2577 static void gen_neon_dup_low16(TCGv_i32 var)
2578 {
2579 TCGv_i32 tmp = tcg_temp_new_i32();
2580 tcg_gen_ext16u_i32(var, var);
2581 tcg_gen_shli_i32(tmp, var, 16);
2582 tcg_gen_or_i32(var, var, tmp);
2583 tcg_temp_free_i32(tmp);
2584 }
2585
2586 static void gen_neon_dup_high16(TCGv_i32 var)
2587 {
2588 TCGv_i32 tmp = tcg_temp_new_i32();
2589 tcg_gen_andi_i32(var, var, 0xffff0000);
2590 tcg_gen_shri_i32(tmp, var, 16);
2591 tcg_gen_or_i32(var, var, tmp);
2592 tcg_temp_free_i32(tmp);
2593 }
2594
2595 static TCGv_i32 gen_load_and_replicate(DisasContext *s, TCGv_i32 addr, int size)
2596 {
2597 /* Load a single Neon element and replicate into a 32 bit TCG reg */
2598 TCGv_i32 tmp = tcg_temp_new_i32();
2599 switch (size) {
2600 case 0:
2601 gen_aa32_ld8u(tmp, addr, IS_USER(s));
2602 gen_neon_dup_u8(tmp, 0);
2603 break;
2604 case 1:
2605 gen_aa32_ld16u(tmp, addr, IS_USER(s));
2606 gen_neon_dup_low16(tmp);
2607 break;
2608 case 2:
2609 gen_aa32_ld32u(tmp, addr, IS_USER(s));
2610 break;
2611 default: /* Avoid compiler warnings. */
2612 abort();
2613 }
2614 return tmp;
2615 }
2616
2617 /* Disassemble a VFP instruction. Returns nonzero if an error occurred
2618 (ie. an undefined instruction). */
2619 static int disas_vfp_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
2620 {
2621 uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask;
2622 int dp, veclen;
2623 TCGv_i32 addr;
2624 TCGv_i32 tmp;
2625 TCGv_i32 tmp2;
2626
2627 if (!arm_feature(env, ARM_FEATURE_VFP))
2628 return 1;
2629
2630 if (!s->vfp_enabled) {
2631 /* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
2632 if ((insn & 0x0fe00fff) != 0x0ee00a10)
2633 return 1;
2634 rn = (insn >> 16) & 0xf;
2635 if (rn != ARM_VFP_FPSID && rn != ARM_VFP_FPEXC
2636 && rn != ARM_VFP_MVFR1 && rn != ARM_VFP_MVFR0)
2637 return 1;
2638 }
2639 dp = ((insn & 0xf00) == 0xb00);
2640 switch ((insn >> 24) & 0xf) {
2641 case 0xe:
2642 if (insn & (1 << 4)) {
2643 /* single register transfer */
2644 rd = (insn >> 12) & 0xf;
2645 if (dp) {
2646 int size;
2647 int pass;
2648
2649 VFP_DREG_N(rn, insn);
2650 if (insn & 0xf)
2651 return 1;
2652 if (insn & 0x00c00060
2653 && !arm_feature(env, ARM_FEATURE_NEON))
2654 return 1;
2655
2656 pass = (insn >> 21) & 1;
2657 if (insn & (1 << 22)) {
2658 size = 0;
2659 offset = ((insn >> 5) & 3) * 8;
2660 } else if (insn & (1 << 5)) {
2661 size = 1;
2662 offset = (insn & (1 << 6)) ? 16 : 0;
2663 } else {
2664 size = 2;
2665 offset = 0;
2666 }
2667 if (insn & ARM_CP_RW_BIT) {
2668 /* vfp->arm */
2669 tmp = neon_load_reg(rn, pass);
2670 switch (size) {
2671 case 0:
2672 if (offset)
2673 tcg_gen_shri_i32(tmp, tmp, offset);
2674 if (insn & (1 << 23))
2675 gen_uxtb(tmp);
2676 else
2677 gen_sxtb(tmp);
2678 break;
2679 case 1:
2680 if (insn & (1 << 23)) {
2681 if (offset) {
2682 tcg_gen_shri_i32(tmp, tmp, 16);
2683 } else {
2684 gen_uxth(tmp);
2685 }
2686 } else {
2687 if (offset) {
2688 tcg_gen_sari_i32(tmp, tmp, 16);
2689 } else {
2690 gen_sxth(tmp);
2691 }
2692 }
2693 break;
2694 case 2:
2695 break;
2696 }
2697 store_reg(s, rd, tmp);
2698 } else {
2699 /* arm->vfp */
2700 tmp = load_reg(s, rd);
2701 if (insn & (1 << 23)) {
2702 /* VDUP */
2703 if (size == 0) {
2704 gen_neon_dup_u8(tmp, 0);
2705 } else if (size == 1) {
2706 gen_neon_dup_low16(tmp);
2707 }
2708 for (n = 0; n <= pass * 2; n++) {
2709 tmp2 = tcg_temp_new_i32();
2710 tcg_gen_mov_i32(tmp2, tmp);
2711 neon_store_reg(rn, n, tmp2);
2712 }
2713 neon_store_reg(rn, n, tmp);
2714 } else {
2715 /* VMOV */
2716 switch (size) {
2717 case 0:
2718 tmp2 = neon_load_reg(rn, pass);
2719 tcg_gen_deposit_i32(tmp, tmp2, tmp, offset, 8);
2720 tcg_temp_free_i32(tmp2);
2721 break;
2722 case 1:
2723 tmp2 = neon_load_reg(rn, pass);
2724 tcg_gen_deposit_i32(tmp, tmp2, tmp, offset, 16);
2725 tcg_temp_free_i32(tmp2);
2726 break;
2727 case 2:
2728 break;
2729 }
2730 neon_store_reg(rn, pass, tmp);
2731 }
2732 }
2733 } else { /* !dp */
2734 if ((insn & 0x6f) != 0x00)
2735 return 1;
2736 rn = VFP_SREG_N(insn);
2737 if (insn & ARM_CP_RW_BIT) {
2738 /* vfp->arm */
2739 if (insn & (1 << 21)) {
2740 /* system register */
2741 rn >>= 1;
2742
2743 switch (rn) {
2744 case ARM_VFP_FPSID:
2745 /* VFP2 allows access to FSID from userspace.
2746 VFP3 restricts all id registers to privileged
2747 accesses. */
2748 if (IS_USER(s)
2749 && arm_feature(env, ARM_FEATURE_VFP3))
2750 return 1;
2751 tmp = load_cpu_field(vfp.xregs[rn]);
2752 break;
2753 case ARM_VFP_FPEXC:
2754 if (IS_USER(s))
2755 return 1;
2756 tmp = load_cpu_field(vfp.xregs[rn]);
2757 break;
2758 case ARM_VFP_FPINST:
2759 case ARM_VFP_FPINST2:
2760 /* Not present in VFP3. */
2761 if (IS_USER(s)
2762 || arm_feature(env, ARM_FEATURE_VFP3))
2763 return 1;
2764 tmp = load_cpu_field(vfp.xregs[rn]);
2765 break;
2766 case ARM_VFP_FPSCR:
2767 if (rd == 15) {
2768 tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]);
2769 tcg_gen_andi_i32(tmp, tmp, 0xf0000000);
2770 } else {
2771 tmp = tcg_temp_new_i32();
2772 gen_helper_vfp_get_fpscr(tmp, cpu_env);
2773 }
2774 break;
2775 case ARM_VFP_MVFR0:
2776 case ARM_VFP_MVFR1:
2777 if (IS_USER(s)
2778 || !arm_feature(env, ARM_FEATURE_MVFR))
2779 return 1;
2780 tmp = load_cpu_field(vfp.xregs[rn]);
2781 break;
2782 default:
2783 return 1;
2784 }
2785 } else {
2786 gen_mov_F0_vreg(0, rn);
2787 tmp = gen_vfp_mrs();
2788 }
2789 if (rd == 15) {
2790 /* Set the 4 flag bits in the CPSR. */
2791 gen_set_nzcv(tmp);
2792 tcg_temp_free_i32(tmp);
2793 } else {
2794 store_reg(s, rd, tmp);
2795 }
2796 } else {
2797 /* arm->vfp */
2798 if (insn & (1 << 21)) {
2799 rn >>= 1;
2800 /* system register */
2801 switch (rn) {
2802 case ARM_VFP_FPSID:
2803 case ARM_VFP_MVFR0:
2804 case ARM_VFP_MVFR1:
2805 /* Writes are ignored. */
2806 break;
2807 case ARM_VFP_FPSCR:
2808 tmp = load_reg(s, rd);
2809 gen_helper_vfp_set_fpscr(cpu_env, tmp);
2810 tcg_temp_free_i32(tmp);
2811 gen_lookup_tb(s);
2812 break;
2813 case ARM_VFP_FPEXC:
2814 if (IS_USER(s))
2815 return 1;
2816 /* TODO: VFP subarchitecture support.
2817 * For now, keep the EN bit only */
2818 tmp = load_reg(s, rd);
2819 tcg_gen_andi_i32(tmp, tmp, 1 << 30);
2820 store_cpu_field(tmp, vfp.xregs[rn]);
2821 gen_lookup_tb(s);
2822 break;
2823 case ARM_VFP_FPINST:
2824 case ARM_VFP_FPINST2:
2825 tmp = load_reg(s, rd);
2826 store_cpu_field(tmp, vfp.xregs[rn]);
2827 break;
2828 default:
2829 return 1;
2830 }
2831 } else {
2832 tmp = load_reg(s, rd);
2833 gen_vfp_msr(tmp);
2834 gen_mov_vreg_F0(0, rn);
2835 }
2836 }
2837 }
2838 } else {
2839 /* data processing */
2840 /* The opcode is in bits 23, 21, 20 and 6. */
2841 op = ((insn >> 20) & 8) | ((insn >> 19) & 6) | ((insn >> 6) & 1);
2842 if (dp) {
2843 if (op == 15) {
2844 /* rn is opcode */
2845 rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
2846 } else {
2847 /* rn is register number */
2848 VFP_DREG_N(rn, insn);
2849 }
2850
2851 if (op == 15 && (rn == 15 || ((rn & 0x1c) == 0x18))) {
2852 /* Integer or single precision destination. */
2853 rd = VFP_SREG_D(insn);
2854 } else {
2855 VFP_DREG_D(rd, insn);
2856 }
2857 if (op == 15 &&
2858 (((rn & 0x1c) == 0x10) || ((rn & 0x14) == 0x14))) {
2859 /* VCVT from int is always from S reg regardless of dp bit.
2860 * VCVT with immediate frac_bits has same format as SREG_M
2861 */
2862 rm = VFP_SREG_M(insn);
2863 } else {
2864 VFP_DREG_M(rm, insn);
2865 }
2866 } else {
2867 rn = VFP_SREG_N(insn);
2868 if (op == 15 && rn == 15) {
2869 /* Double precision destination. */
2870 VFP_DREG_D(rd, insn);
2871 } else {
2872 rd = VFP_SREG_D(insn);
2873 }
2874 /* NB that we implicitly rely on the encoding for the frac_bits
2875 * in VCVT of fixed to float being the same as that of an SREG_M
2876 */
2877 rm = VFP_SREG_M(insn);
2878 }
2879
2880 veclen = s->vec_len;
2881 if (op == 15 && rn > 3)
2882 veclen = 0;
2883
2884 /* Shut up compiler warnings. */
2885 delta_m = 0;
2886 delta_d = 0;
2887 bank_mask = 0;
2888
2889 if (veclen > 0) {
2890 if (dp)
2891 bank_mask = 0xc;
2892 else
2893 bank_mask = 0x18;
2894
2895 /* Figure out what type of vector operation this is. */
2896 if ((rd & bank_mask) == 0) {
2897 /* scalar */
2898 veclen = 0;
2899 } else {
2900 if (dp)
2901 delta_d = (s->vec_stride >> 1) + 1;
2902 else
2903 delta_d = s->vec_stride + 1;
2904
2905 if ((rm & bank_mask) == 0) {
2906 /* mixed scalar/vector */
2907 delta_m = 0;
2908 } else {
2909 /* vector */
2910 delta_m = delta_d;
2911 }
2912 }
2913 }
2914
2915 /* Load the initial operands. */
2916 if (op == 15) {
2917 switch (rn) {
2918 case 16:
2919 case 17:
2920 /* Integer source */
2921 gen_mov_F0_vreg(0, rm);
2922 break;
2923 case 8:
2924 case 9:
2925 /* Compare */
2926 gen_mov_F0_vreg(dp, rd);
2927 gen_mov_F1_vreg(dp, rm);
2928 break;
2929 case 10:
2930 case 11:
2931 /* Compare with zero */
2932 gen_mov_F0_vreg(dp, rd);
2933 gen_vfp_F1_ld0(dp);
2934 break;
2935 case 20:
2936 case 21:
2937 case 22:
2938 case 23:
2939 case 28:
2940 case 29:
2941 case 30:
2942 case 31:
2943 /* Source and destination the same. */
2944 gen_mov_F0_vreg(dp, rd);
2945 break;
2946 case 4:
2947 case 5:
2948 case 6:
2949 case 7:
2950 /* VCVTB, VCVTT: only present with the halfprec extension,
2951 * UNPREDICTABLE if bit 8 is set (we choose to UNDEF)
2952 */
2953 if (dp || !arm_feature(env, ARM_FEATURE_VFP_FP16)) {
2954 return 1;
2955 }
2956 /* Otherwise fall through */
2957 default:
2958 /* One source operand. */
2959 gen_mov_F0_vreg(dp, rm);
2960 break;
2961 }
2962 } else {
2963 /* Two source operands. */
2964 gen_mov_F0_vreg(dp, rn);
2965 gen_mov_F1_vreg(dp, rm);
2966 }
2967
2968 for (;;) {
2969 /* Perform the calculation. */
2970 switch (op) {
2971 case 0: /* VMLA: fd + (fn * fm) */
2972 /* Note that order of inputs to the add matters for NaNs */
2973 gen_vfp_F1_mul(dp);
2974 gen_mov_F0_vreg(dp, rd);
2975 gen_vfp_add(dp);
2976 break;
2977 case 1: /* VMLS: fd + -(fn * fm) */
2978 gen_vfp_mul(dp);
2979 gen_vfp_F1_neg(dp);
2980 gen_mov_F0_vreg(dp, rd);
2981 gen_vfp_add(dp);
2982 break;
2983 case 2: /* VNMLS: -fd + (fn * fm) */
2984 /* Note that it isn't valid to replace (-A + B) with (B - A)
2985 * or similar plausible looking simplifications
2986 * because this will give wrong results for NaNs.
2987 */
2988 gen_vfp_F1_mul(dp);
2989 gen_mov_F0_vreg(dp, rd);
2990 gen_vfp_neg(dp);
2991 gen_vfp_add(dp);
2992 break;
2993 case 3: /* VNMLA: -fd + -(fn * fm) */
2994 gen_vfp_mul(dp);
2995 gen_vfp_F1_neg(dp);
2996 gen_mov_F0_vreg(dp, rd);
2997 gen_vfp_neg(dp);
2998 gen_vfp_add(dp);
2999 break;
3000 case 4: /* mul: fn * fm */
3001 gen_vfp_mul(dp);
3002 break;
3003 case 5: /* nmul: -(fn * fm) */
3004 gen_vfp_mul(dp);
3005 gen_vfp_neg(dp);
3006 break;
3007 case 6: /* add: fn + fm */
3008 gen_vfp_add(dp);
3009 break;
3010 case 7: /* sub: fn - fm */
3011 gen_vfp_sub(dp);
3012 break;
3013 case 8: /* div: fn / fm */
3014 gen_vfp_div(dp);
3015 break;
3016 case 10: /* VFNMA : fd = muladd(-fd, fn, fm) */
3017 case 11: /* VFNMS : fd = muladd(-fd, -fn, fm) */
3018 case 12: /* VFMA : fd = muladd( fd, fn, fm) */
3019 case 13: /* VFMS : fd = muladd( fd, -fn, fm) */
3020 /* These are fused multiply-add, and must be done as one
3021 * floating point operation with no rounding between the
3022 * multiplication and addition steps.
3023 * NB that doing the negations here as separate steps is
3024 * correct : an input NaN should come out with its sign bit
3025 * flipped if it is a negated-input.
3026 */
3027 if (!arm_feature(env, ARM_FEATURE_VFP4)) {
3028 return 1;
3029 }
3030 if (dp) {
3031 TCGv_ptr fpst;
3032 TCGv_i64 frd;
3033 if (op & 1) {
3034 /* VFNMS, VFMS */
3035 gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
3036 }
3037 frd = tcg_temp_new_i64();
3038 tcg_gen_ld_f64(frd, cpu_env, vfp_reg_offset(dp, rd));
3039 if (op & 2) {
3040 /* VFNMA, VFNMS */
3041 gen_helper_vfp_negd(frd, frd);
3042 }
3043 fpst = get_fpstatus_ptr(0);
3044 gen_helper_vfp_muladdd(cpu_F0d, cpu_F0d,
3045 cpu_F1d, frd, fpst);
3046 tcg_temp_free_ptr(fpst);
3047 tcg_temp_free_i64(frd);
3048 } else {
3049 TCGv_ptr fpst;
3050 TCGv_i32 frd;
3051 if (op & 1) {
3052 /* VFNMS, VFMS */
3053 gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
3054 }
3055 frd = tcg_temp_new_i32();
3056 tcg_gen_ld_f32(frd, cpu_env, vfp_reg_offset(dp, rd));
3057 if (op & 2) {
3058 gen_helper_vfp_negs(frd, frd);
3059 }
3060 fpst = get_fpstatus_ptr(0);
3061 gen_helper_vfp_muladds(cpu_F0s, cpu_F0s,
3062 cpu_F1s, frd, fpst);
3063 tcg_temp_free_ptr(fpst);
3064 tcg_temp_free_i32(frd);
3065 }
3066 break;
3067 case 14: /* fconst */
3068 if (!arm_feature(env, ARM_FEATURE_VFP3))
3069 return 1;
3070
3071 n = (insn << 12) & 0x80000000;
3072 i = ((insn >> 12) & 0x70) | (insn & 0xf);
3073 if (dp) {
3074 if (i & 0x40)
3075 i |= 0x3f80;
3076 else
3077 i |= 0x4000;
3078 n |= i << 16;
3079 tcg_gen_movi_i64(cpu_F0d, ((uint64_t)n) << 32);
3080 } else {
3081 if (i & 0x40)
3082 i |= 0x780;
3083 else
3084 i |= 0x800;
3085 n |= i << 19;
3086 tcg_gen_movi_i32(cpu_F0s, n);
3087 }
3088 break;
3089 case 15: /* extension space */
3090 switch (rn) {
3091 case 0: /* cpy */
3092 /* no-op */
3093 break;
3094 case 1: /* abs */
3095 gen_vfp_abs(dp);
3096 break;
3097 case 2: /* neg */
3098 gen_vfp_neg(dp);
3099 break;
3100 case 3: /* sqrt */
3101 gen_vfp_sqrt(dp);
3102 break;
3103 case 4: /* vcvtb.f32.f16 */
3104 tmp = gen_vfp_mrs();
3105 tcg_gen_ext16u_i32(tmp, tmp);
3106 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp, cpu_env);
3107 tcg_temp_free_i32(tmp);
3108 break;
3109 case 5: /* vcvtt.f32.f16 */
3110 tmp = gen_vfp_mrs();
3111 tcg_gen_shri_i32(tmp, tmp, 16);
3112 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp, cpu_env);
3113 tcg_temp_free_i32(tmp);
3114 break;
3115 case 6: /* vcvtb.f16.f32 */
3116 tmp = tcg_temp_new_i32();
3117 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
3118 gen_mov_F0_vreg(0, rd);
3119 tmp2 = gen_vfp_mrs();
3120 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
3121 tcg_gen_or_i32(tmp, tmp, tmp2);
3122 tcg_temp_free_i32(tmp2);
3123 gen_vfp_msr(tmp);
3124 break;
3125 case 7: /* vcvtt.f16.f32 */
3126 tmp = tcg_temp_new_i32();
3127 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
3128 tcg_gen_shli_i32(tmp, tmp, 16);
3129 gen_mov_F0_vreg(0, rd);
3130 tmp2 = gen_vfp_mrs();
3131 tcg_gen_ext16u_i32(tmp2, tmp2);
3132 tcg_gen_or_i32(tmp, tmp, tmp2);
3133 tcg_temp_free_i32(tmp2);
3134 gen_vfp_msr(tmp);
3135 break;
3136 case 8: /* cmp */
3137 gen_vfp_cmp(dp);
3138 break;
3139 case 9: /* cmpe */
3140 gen_vfp_cmpe(dp);
3141 break;
3142 case 10: /* cmpz */
3143 gen_vfp_cmp(dp);
3144 break;
3145 case 11: /* cmpez */
3146 gen_vfp_F1_ld0(dp);
3147 gen_vfp_cmpe(dp);
3148 break;
3149 case 15: /* single<->double conversion */
3150 if (dp)
3151 gen_helper_vfp_fcvtsd(cpu_F0s, cpu_F0d, cpu_env);
3152 else
3153 gen_helper_vfp_fcvtds(cpu_F0d, cpu_F0s, cpu_env);
3154 break;
3155 case 16: /* fuito */
3156 gen_vfp_uito(dp, 0);
3157 break;
3158 case 17: /* fsito */
3159 gen_vfp_sito(dp, 0);
3160 break;
3161 case 20: /* fshto */
3162 if (!arm_feature(env, ARM_FEATURE_VFP3))
3163 return 1;
3164 gen_vfp_shto(dp, 16 - rm, 0);
3165 break;
3166 case 21: /* fslto */
3167 if (!arm_feature(env, ARM_FEATURE_VFP3))
3168 return 1;
3169 gen_vfp_slto(dp, 32 - rm, 0);
3170 break;
3171 case 22: /* fuhto */
3172 if (!arm_feature(env, ARM_FEATURE_VFP3))
3173 return 1;
3174 gen_vfp_uhto(dp, 16 - rm, 0);
3175 break;
3176 case 23: /* fulto */
3177 if (!arm_feature(env, ARM_FEATURE_VFP3))
3178 return 1;
3179 gen_vfp_ulto(dp, 32 - rm, 0);
3180 break;
3181 case 24: /* ftoui */
3182 gen_vfp_toui(dp, 0);
3183 break;
3184 case 25: /* ftouiz */
3185 gen_vfp_touiz(dp, 0);
3186 break;
3187 case 26: /* ftosi */
3188 gen_vfp_tosi(dp, 0);
3189 break;
3190 case 27: /* ftosiz */
3191 gen_vfp_tosiz(dp, 0);
3192 break;
3193 case 28: /* ftosh */
3194 if (!arm_feature(env, ARM_FEATURE_VFP3))
3195 return 1;
3196 gen_vfp_tosh(dp, 16 - rm, 0);
3197 break;
3198 case 29: /* ftosl */
3199 if (!arm_feature(env, ARM_FEATURE_VFP3))
3200 return 1;
3201 gen_vfp_tosl(dp, 32 - rm, 0);
3202 break;
3203 case 30: /* ftouh */
3204 if (!arm_feature(env, ARM_FEATURE_VFP3))
3205 return 1;
3206 gen_vfp_touh(dp, 16 - rm, 0);
3207 break;
3208 case 31: /* ftoul */
3209 if (!arm_feature(env, ARM_FEATURE_VFP3))
3210 return 1;
3211 gen_vfp_toul(dp, 32 - rm, 0);
3212 break;
3213 default: /* undefined */
3214 return 1;
3215 }
3216 break;
3217 default: /* undefined */
3218 return 1;
3219 }
3220
3221 /* Write back the result. */
3222 if (op == 15 && (rn >= 8 && rn <= 11))
3223 ; /* Comparison, do nothing. */
3224 else if (op == 15 && dp && ((rn & 0x1c) == 0x18))
3225 /* VCVT double to int: always integer result. */
3226 gen_mov_vreg_F0(0, rd);
3227 else if (op == 15 && rn == 15)
3228 /* conversion */
3229 gen_mov_vreg_F0(!dp, rd);
3230 else
3231 gen_mov_vreg_F0(dp, rd);
3232
3233 /* break out of the loop if we have finished */
3234 if (veclen == 0)
3235 break;
3236
3237 if (op == 15 && delta_m == 0) {
3238 /* single source one-many */
3239 while (veclen--) {
3240 rd = ((rd + delta_d) & (bank_mask - 1))
3241 | (rd & bank_mask);
3242 gen_mov_vreg_F0(dp, rd);
3243 }
3244 break;
3245 }
3246 /* Setup the next operands. */
3247 veclen--;
3248 rd = ((rd + delta_d) & (bank_mask - 1))
3249 | (rd & bank_mask);
3250
3251 if (op == 15) {
3252 /* One source operand. */
3253 rm = ((rm + delta_m) & (bank_mask - 1))
3254 | (rm & bank_mask);
3255 gen_mov_F0_vreg(dp, rm);
3256 } else {
3257 /* Two source operands. */
3258 rn = ((rn + delta_d) & (bank_mask - 1))
3259 | (rn & bank_mask);
3260 gen_mov_F0_vreg(dp, rn);
3261 if (delta_m) {
3262 rm = ((rm + delta_m) & (bank_mask - 1))
3263 | (rm & bank_mask);
3264 gen_mov_F1_vreg(dp, rm);
3265 }
3266 }
3267 }
3268 }
3269 break;
3270 case 0xc:
3271 case 0xd:
3272 if ((insn & 0x03e00000) == 0x00400000) {
3273 /* two-register transfer */
3274 rn = (insn >> 16) & 0xf;
3275 rd = (insn >> 12) & 0xf;
3276 if (dp) {
3277 VFP_DREG_M(rm, insn);
3278 } else {
3279 rm = VFP_SREG_M(insn);
3280 }
3281
3282 if (insn & ARM_CP_RW_BIT) {
3283 /* vfp->arm */
3284 if (dp) {
3285 gen_mov_F0_vreg(0, rm * 2);
3286 tmp = gen_vfp_mrs();
3287 store_reg(s, rd, tmp);
3288 gen_mov_F0_vreg(0, rm * 2 + 1);
3289 tmp = gen_vfp_mrs();
3290 store_reg(s, rn, tmp);
3291 } else {
3292 gen_mov_F0_vreg(0, rm);
3293 tmp = gen_vfp_mrs();
3294 store_reg(s, rd, tmp);
3295 gen_mov_F0_vreg(0, rm + 1);
3296 tmp = gen_vfp_mrs();
3297 store_reg(s, rn, tmp);
3298 }
3299 } else {
3300 /* arm->vfp */
3301 if (dp) {
3302 tmp = load_reg(s, rd);
3303 gen_vfp_msr(tmp);
3304 gen_mov_vreg_F0(0, rm * 2);
3305 tmp = load_reg(s, rn);
3306 gen_vfp_msr(tmp);
3307 gen_mov_vreg_F0(0, rm * 2 + 1);
3308 } else {
3309 tmp = load_reg(s, rd);
3310 gen_vfp_msr(tmp);
3311 gen_mov_vreg_F0(0, rm);
3312 tmp = load_reg(s, rn);
3313 gen_vfp_msr(tmp);
3314 gen_mov_vreg_F0(0, rm + 1);
3315 }
3316 }
3317 } else {
3318 /* Load/store */
3319 rn = (insn >> 16) & 0xf;
3320 if (dp)
3321 VFP_DREG_D(rd, insn);
3322 else
3323 rd = VFP_SREG_D(insn);
3324 if ((insn & 0x01200000) == 0x01000000) {
3325 /* Single load/store */
3326 offset = (insn & 0xff) << 2;
3327 if ((insn & (1 << 23)) == 0)
3328 offset = -offset;
3329 if (s->thumb && rn == 15) {
3330 /* This is actually UNPREDICTABLE */
3331 addr = tcg_temp_new_i32();
3332 tcg_gen_movi_i32(addr, s->pc & ~2);
3333 } else {
3334 addr = load_reg(s, rn);
3335 }
3336 tcg_gen_addi_i32(addr, addr, offset);
3337 if (insn & (1 << 20)) {
3338 gen_vfp_ld(s, dp, addr);
3339 gen_mov_vreg_F0(dp, rd);
3340 } else {
3341 gen_mov_F0_vreg(dp, rd);
3342 gen_vfp_st(s, dp, addr);
3343 }
3344 tcg_temp_free_i32(addr);
3345 } else {
3346 /* load/store multiple */
3347 int w = insn & (1 << 21);
3348 if (dp)
3349 n = (insn >> 1) & 0x7f;
3350 else
3351 n = insn & 0xff;
3352
3353 if (w && !(((insn >> 23) ^ (insn >> 24)) & 1)) {
3354 /* P == U , W == 1 => UNDEF */
3355 return 1;
3356 }
3357 if (n == 0 || (rd + n) > 32 || (dp && n > 16)) {
3358 /* UNPREDICTABLE cases for bad immediates: we choose to
3359 * UNDEF to avoid generating huge numbers of TCG ops
3360 */
3361 return 1;
3362 }
3363 if (rn == 15 && w) {
3364 /* writeback to PC is UNPREDICTABLE, we choose to UNDEF */
3365 return 1;
3366 }
3367
3368 if (s->thumb && rn == 15) {
3369 /* This is actually UNPREDICTABLE */
3370 addr = tcg_temp_new_i32();
3371 tcg_gen_movi_i32(addr, s->pc & ~2);
3372 } else {
3373 addr = load_reg(s, rn);
3374 }
3375 if (insn & (1 << 24)) /* pre-decrement */
3376 tcg_gen_addi_i32(addr, addr, -((insn & 0xff) << 2));
3377
3378 if (dp)
3379 offset = 8;
3380 else
3381 offset = 4;
3382 for (i = 0; i < n; i++) {
3383 if (insn & ARM_CP_RW_BIT) {
3384 /* load */
3385 gen_vfp_ld(s, dp, addr);
3386 gen_mov_vreg_F0(dp, rd + i);
3387 } else {
3388 /* store */
3389 gen_mov_F0_vreg(dp, rd + i);
3390 gen_vfp_st(s, dp, addr);
3391 }
3392 tcg_gen_addi_i32(addr, addr, offset);
3393 }
3394 if (w) {
3395 /* writeback */
3396 if (insn & (1 << 24))
3397 offset = -offset * n;
3398 else if (dp && (insn & 1))
3399 offset = 4;
3400 else
3401 offset = 0;
3402
3403 if (offset != 0)
3404 tcg_gen_addi_i32(addr, addr, offset);
3405 store_reg(s, rn, addr);
3406 } else {
3407 tcg_temp_free_i32(addr);
3408 }
3409 }
3410 }
3411 break;
3412 default:
3413 /* Should never happen. */
3414 return 1;
3415 }
3416 return 0;
3417 }
3418
3419 static inline void gen_goto_tb(DisasContext *s, int n, target_ulong dest)
3420 {
3421 TranslationBlock *tb;
3422
3423 tb = s->tb;
3424 if ((tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK)) {
3425 tcg_gen_goto_tb(n);
3426 gen_set_pc_im(s, dest);
3427 tcg_gen_exit_tb((uintptr_t)tb + n);
3428 } else {
3429 gen_set_pc_im(s, dest);
3430 tcg_gen_exit_tb(0);
3431 }
3432 }
3433
3434 static inline void gen_jmp (DisasContext *s, uint32_t dest)
3435 {
3436 if (unlikely(s->singlestep_enabled)) {
3437 /* An indirect jump so that we still trigger the debug exception. */
3438 if (s->thumb)
3439 dest |= 1;
3440 gen_bx_im(s, dest);
3441 } else {
3442 gen_goto_tb(s, 0, dest);
3443 s->is_jmp = DISAS_TB_JUMP;
3444 }
3445 }
3446
3447 static inline void gen_mulxy(TCGv_i32 t0, TCGv_i32 t1, int x, int y)
3448 {
3449 if (x)
3450 tcg_gen_sari_i32(t0, t0, 16);
3451 else
3452 gen_sxth(t0);
3453 if (y)
3454 tcg_gen_sari_i32(t1, t1, 16);
3455 else
3456 gen_sxth(t1);
3457 tcg_gen_mul_i32(t0, t0, t1);
3458 }
3459
3460 /* Return the mask of PSR bits set by a MSR instruction. */
3461 static uint32_t msr_mask(CPUARMState *env, DisasContext *s, int flags, int spsr) {
3462 uint32_t mask;
3463
3464 mask = 0;
3465 if (flags & (1 << 0))
3466 mask |= 0xff;
3467 if (flags & (1 << 1))
3468 mask |= 0xff00;
3469 if (flags & (1 << 2))
3470 mask |= 0xff0000;
3471 if (flags & (1 << 3))
3472 mask |= 0xff000000;
3473
3474 /* Mask out undefined bits. */
3475 mask &= ~CPSR_RESERVED;
3476 if (!arm_feature(env, ARM_FEATURE_V4T))
3477 mask &= ~CPSR_T;
3478 if (!arm_feature(env, ARM_FEATURE_V5))
3479 mask &= ~CPSR_Q; /* V5TE in reality*/
3480 if (!arm_feature(env, ARM_FEATURE_V6))
3481 mask &= ~(CPSR_E | CPSR_GE);
3482 if (!arm_feature(env, ARM_FEATURE_THUMB2))
3483 mask &= ~CPSR_IT;
3484 /* Mask out execution state bits. */
3485 if (!spsr)
3486 mask &= ~CPSR_EXEC;
3487 /* Mask out privileged bits. */
3488 if (IS_USER(s))
3489 mask &= CPSR_USER;
3490 return mask;
3491 }
3492
3493 /* Returns nonzero if access to the PSR is not permitted. Marks t0 as dead. */
3494 static int gen_set_psr(DisasContext *s, uint32_t mask, int spsr, TCGv_i32 t0)
3495 {
3496 TCGv_i32 tmp;
3497 if (spsr) {
3498 /* ??? This is also undefined in system mode. */
3499 if (IS_USER(s))
3500 return 1;
3501
3502 tmp = load_cpu_field(spsr);
3503 tcg_gen_andi_i32(tmp, tmp, ~mask);
3504 tcg_gen_andi_i32(t0, t0, mask);
3505 tcg_gen_or_i32(tmp, tmp, t0);
3506 store_cpu_field(tmp, spsr);
3507 } else {
3508 gen_set_cpsr(t0, mask);
3509 }
3510 tcg_temp_free_i32(t0);
3511 gen_lookup_tb(s);
3512 return 0;
3513 }
3514
3515 /* Returns nonzero if access to the PSR is not permitted. */
3516 static int gen_set_psr_im(DisasContext *s, uint32_t mask, int spsr, uint32_t val)
3517 {
3518 TCGv_i32 tmp;
3519 tmp = tcg_temp_new_i32();
3520 tcg_gen_movi_i32(tmp, val);
3521 return gen_set_psr(s, mask, spsr, tmp);
3522 }
3523
3524 /* Generate an old-style exception return. Marks pc as dead. */
3525 static void gen_exception_return(DisasContext *s, TCGv_i32 pc)
3526 {
3527 TCGv_i32 tmp;
3528 store_reg(s, 15, pc);
3529 tmp = load_cpu_field(spsr);
3530 gen_set_cpsr(tmp, 0xffffffff);
3531 tcg_temp_free_i32(tmp);
3532 s->is_jmp = DISAS_UPDATE;
3533 }
3534
3535 /* Generate a v6 exception return. Marks both values as dead. */
3536 static void gen_rfe(DisasContext *s, TCGv_i32 pc, TCGv_i32 cpsr)
3537 {
3538 gen_set_cpsr(cpsr, 0xffffffff);
3539 tcg_temp_free_i32(cpsr);
3540 store_reg(s, 15, pc);
3541 s->is_jmp = DISAS_UPDATE;
3542 }
3543
3544 static inline void
3545 gen_set_condexec (DisasContext *s)
3546 {
3547 if (s->condexec_mask) {
3548 uint32_t val = (s->condexec_cond << 4) | (s->condexec_mask >> 1);
3549 TCGv_i32 tmp = tcg_temp_new_i32();
3550 tcg_gen_movi_i32(tmp, val);
3551 store_cpu_field(tmp, condexec_bits);
3552 }
3553 }
3554
3555 static void gen_exception_insn(DisasContext *s, int offset, int excp)
3556 {
3557 gen_set_condexec(s);
3558 gen_set_pc_im(s, s->pc - offset);
3559 gen_exception(excp);
3560 s->is_jmp = DISAS_JUMP;
3561 }
3562
3563 static void gen_nop_hint(DisasContext *s, int val)
3564 {
3565 switch (val) {
3566 case 3: /* wfi */
3567 gen_set_pc_im(s, s->pc);
3568 s->is_jmp = DISAS_WFI;
3569 break;
3570 case 2: /* wfe */
3571 case 4: /* sev */
3572 case 5: /* sevl */
3573 /* TODO: Implement SEV, SEVL and WFE. May help SMP performance. */
3574 default: /* nop */
3575 break;
3576 }
3577 }
3578
3579 #define CPU_V001 cpu_V0, cpu_V0, cpu_V1
3580
3581 static inline void gen_neon_add(int size, TCGv_i32 t0, TCGv_i32 t1)
3582 {
3583 switch (size) {
3584 case 0: gen_helper_neon_add_u8(t0, t0, t1); break;
3585 case 1: gen_helper_neon_add_u16(t0, t0, t1); break;
3586 case 2: tcg_gen_add_i32(t0, t0, t1); break;
3587 default: abort();
3588 }
3589 }
3590
3591 static inline void gen_neon_rsb(int size, TCGv_i32 t0, TCGv_i32 t1)
3592 {
3593 switch (size) {
3594 case 0: gen_helper_neon_sub_u8(t0, t1, t0); break;
3595 case 1: gen_helper_neon_sub_u16(t0, t1, t0); break;
3596 case 2: tcg_gen_sub_i32(t0, t1, t0); break;
3597 default: return;
3598 }
3599 }
3600
3601 /* 32-bit pairwise ops end up the same as the elementwise versions. */
3602 #define gen_helper_neon_pmax_s32 gen_helper_neon_max_s32
3603 #define gen_helper_neon_pmax_u32 gen_helper_neon_max_u32
3604 #define gen_helper_neon_pmin_s32 gen_helper_neon_min_s32
3605 #define gen_helper_neon_pmin_u32 gen_helper_neon_min_u32
3606
3607 #define GEN_NEON_INTEGER_OP_ENV(name) do { \
3608 switch ((size << 1) | u) { \
3609 case 0: \
3610 gen_helper_neon_##name##_s8(tmp, cpu_env, tmp, tmp2); \
3611 break; \
3612 case 1: \
3613 gen_helper_neon_##name##_u8(tmp, cpu_env, tmp, tmp2); \
3614 break; \
3615 case 2: \
3616 gen_helper_neon_##name##_s16(tmp, cpu_env, tmp, tmp2); \
3617 break; \
3618 case 3: \
3619 gen_helper_neon_##name##_u16(tmp, cpu_env, tmp, tmp2); \
3620 break; \
3621 case 4: \
3622 gen_helper_neon_##name##_s32(tmp, cpu_env, tmp, tmp2); \
3623 break; \
3624 case 5: \
3625 gen_helper_neon_##name##_u32(tmp, cpu_env, tmp, tmp2); \
3626 break; \
3627 default: return 1; \
3628 }} while (0)
3629
3630 #define GEN_NEON_INTEGER_OP(name) do { \
3631 switch ((size << 1) | u) { \
3632 case 0: \
3633 gen_helper_neon_##name##_s8(tmp, tmp, tmp2); \
3634 break; \
3635 case 1: \
3636 gen_helper_neon_##name##_u8(tmp, tmp, tmp2); \
3637 break; \
3638 case 2: \
3639 gen_helper_neon_##name##_s16(tmp, tmp, tmp2); \
3640 break; \
3641 case 3: \
3642 gen_helper_neon_##name##_u16(tmp, tmp, tmp2); \
3643 break; \
3644 case 4: \
3645 gen_helper_neon_##name##_s32(tmp, tmp, tmp2); \
3646 break; \
3647 case 5: \
3648 gen_helper_neon_##name##_u32(tmp, tmp, tmp2); \
3649 break; \
3650 default: return 1; \
3651 }} while (0)
3652
3653 static TCGv_i32 neon_load_scratch(int scratch)
3654 {
3655 TCGv_i32 tmp = tcg_temp_new_i32();
3656 tcg_gen_ld_i32(tmp, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch]));
3657 return tmp;
3658 }
3659
3660 static void neon_store_scratch(int scratch, TCGv_i32 var)
3661 {
3662 tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch]));
3663 tcg_temp_free_i32(var);
3664 }
3665
3666 static inline TCGv_i32 neon_get_scalar(int size, int reg)
3667 {
3668 TCGv_i32 tmp;
3669 if (size == 1) {
3670 tmp = neon_load_reg(reg & 7, reg >> 4);
3671 if (reg & 8) {
3672 gen_neon_dup_high16(tmp);
3673 } else {
3674 gen_neon_dup_low16(tmp);
3675 }
3676 } else {
3677 tmp = neon_load_reg(reg & 15, reg >> 4);
3678 }
3679 return tmp;
3680 }
3681
3682 static int gen_neon_unzip(int rd, int rm, int size, int q)
3683 {
3684 TCGv_i32 tmp, tmp2;
3685 if (!q && size == 2) {
3686 return 1;
3687 }
3688 tmp = tcg_const_i32(rd);
3689 tmp2 = tcg_const_i32(rm);
3690 if (q) {
3691 switch (size) {
3692 case 0:
3693 gen_helper_neon_qunzip8(cpu_env, tmp, tmp2);
3694 break;
3695 case 1:
3696 gen_helper_neon_qunzip16(cpu_env, tmp, tmp2);
3697 break;
3698 case 2:
3699 gen_helper_neon_qunzip32(cpu_env, tmp, tmp2);
3700 break;
3701 default:
3702 abort();
3703 }
3704 } else {
3705 switch (size) {
3706 case 0:
3707 gen_helper_neon_unzip8(cpu_env, tmp, tmp2);
3708 break;
3709 case 1:
3710 gen_helper_neon_unzip16(cpu_env, tmp, tmp2);
3711 break;
3712 default:
3713 abort();
3714 }
3715 }
3716 tcg_temp_free_i32(tmp);
3717 tcg_temp_free_i32(tmp2);
3718 return 0;
3719 }
3720
3721 static int gen_neon_zip(int rd, int rm, int size, int q)
3722 {
3723 TCGv_i32 tmp, tmp2;
3724 if (!q && size == 2) {
3725 return 1;
3726 }
3727 tmp = tcg_const_i32(rd);
3728 tmp2 = tcg_const_i32(rm);
3729 if (q) {
3730 switch (size) {
3731 case 0:
3732 gen_helper_neon_qzip8(cpu_env, tmp, tmp2);
3733 break;
3734 case 1:
3735 gen_helper_neon_qzip16(cpu_env, tmp, tmp2);
3736 break;
3737 case 2:
3738 gen_helper_neon_qzip32(cpu_env, tmp, tmp2);
3739 break;
3740 default:
3741 abort();
3742 }
3743 } else {
3744 switch (size) {
3745 case 0:
3746 gen_helper_neon_zip8(cpu_env, tmp, tmp2);
3747 break;
3748 case 1:
3749 gen_helper_neon_zip16(cpu_env, tmp, tmp2);
3750 break;
3751 default:
3752 abort();
3753 }
3754 }
3755 tcg_temp_free_i32(tmp);
3756 tcg_temp_free_i32(tmp2);
3757 return 0;
3758 }
3759
3760 static void gen_neon_trn_u8(TCGv_i32 t0, TCGv_i32 t1)
3761 {
3762 TCGv_i32 rd, tmp;
3763
3764 rd = tcg_temp_new_i32();
3765 tmp = tcg_temp_new_i32();
3766
3767 tcg_gen_shli_i32(rd, t0, 8);
3768 tcg_gen_andi_i32(rd, rd, 0xff00ff00);
3769 tcg_gen_andi_i32(tmp, t1, 0x00ff00ff);
3770 tcg_gen_or_i32(rd, rd, tmp);
3771
3772 tcg_gen_shri_i32(t1, t1, 8);
3773 tcg_gen_andi_i32(t1, t1, 0x00ff00ff);
3774 tcg_gen_andi_i32(tmp, t0, 0xff00ff00);
3775 tcg_gen_or_i32(t1, t1, tmp);
3776 tcg_gen_mov_i32(t0, rd);
3777
3778 tcg_temp_free_i32(tmp);
3779 tcg_temp_free_i32(rd);
3780 }
3781
3782 static void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1)
3783 {
3784 TCGv_i32 rd, tmp;
3785
3786 rd = tcg_temp_new_i32();
3787 tmp = tcg_temp_new_i32();
3788
3789 tcg_gen_shli_i32(rd, t0, 16);
3790 tcg_gen_andi_i32(tmp, t1, 0xffff);
3791 tcg_gen_or_i32(rd, rd, tmp);
3792 tcg_gen_shri_i32(t1, t1, 16);
3793 tcg_gen_andi_i32(tmp, t0, 0xffff0000);
3794 tcg_gen_or_i32(t1, t1, tmp);
3795 tcg_gen_mov_i32(t0, rd);
3796
3797 tcg_temp_free_i32(tmp);
3798 tcg_temp_free_i32(rd);
3799 }
3800
3801
3802 static struct {
3803 int nregs;
3804 int interleave;
3805 int spacing;
3806 } neon_ls_element_type[11] = {
3807 {4, 4, 1},
3808 {4, 4, 2},
3809 {4, 1, 1},
3810 {4, 2, 1},
3811 {3, 3, 1},
3812 {3, 3, 2},
3813 {3, 1, 1},
3814 {1, 1, 1},
3815 {2, 2, 1},
3816 {2, 2, 2},
3817 {2, 1, 1}
3818 };
3819
3820 /* Translate a NEON load/store element instruction. Return nonzero if the
3821 instruction is invalid. */
3822 static int disas_neon_ls_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
3823 {
3824 int rd, rn, rm;
3825 int op;
3826 int nregs;
3827 int interleave;
3828 int spacing;
3829 int stride;
3830 int size;
3831 int reg;
3832 int pass;
3833 int load;
3834 int shift;
3835 int n;
3836 TCGv_i32 addr;
3837 TCGv_i32 tmp;
3838 TCGv_i32 tmp2;
3839 TCGv_i64 tmp64;
3840
3841 if (!s->vfp_enabled)
3842 return 1;
3843 VFP_DREG_D(rd, insn);
3844 rn = (insn >> 16) & 0xf;
3845 rm = insn & 0xf;
3846 load = (insn & (1 << 21)) != 0;
3847 if ((insn & (1 << 23)) == 0) {
3848 /* Load store all elements. */
3849 op = (insn >> 8) & 0xf;
3850 size = (insn >> 6) & 3;
3851 if (op > 10)
3852 return 1;
3853 /* Catch UNDEF cases for bad values of align field */
3854 switch (op & 0xc) {
3855 case 4:
3856 if (((insn >> 5) & 1) == 1) {
3857 return 1;
3858 }
3859 break;
3860 case 8:
3861 if (((insn >> 4) & 3) == 3) {
3862 return 1;
3863 }
3864 break;
3865 default:
3866 break;
3867 }
3868 nregs = neon_ls_element_type[op].nregs;
3869 interleave = neon_ls_element_type[op].interleave;
3870 spacing = neon_ls_element_type[op].spacing;
3871 if (size == 3 && (interleave | spacing) != 1)
3872 return 1;
3873 addr = tcg_temp_new_i32();
3874 load_reg_var(s, addr, rn);
3875 stride = (1 << size) * interleave;
3876 for (reg = 0; reg < nregs; reg++) {
3877 if (interleave > 2 || (interleave == 2 && nregs == 2)) {
3878 load_reg_var(s, addr, rn);
3879 tcg_gen_addi_i32(addr, addr, (1 << size) * reg);
3880 } else if (interleave == 2 && nregs == 4 && reg == 2) {
3881 load_reg_var(s, addr, rn);
3882 tcg_gen_addi_i32(addr, addr, 1 << size);
3883 }
3884 if (size == 3) {
3885 tmp64 = tcg_temp_new_i64();
3886 if (load) {
3887 gen_aa32_ld64(tmp64, addr, IS_USER(s));
3888 neon_store_reg64(tmp64, rd);
3889 } else {
3890 neon_load_reg64(tmp64, rd);
3891 gen_aa32_st64(tmp64, addr, IS_USER(s));
3892 }
3893 tcg_temp_free_i64(tmp64);
3894 tcg_gen_addi_i32(addr, addr, stride);
3895 } else {
3896 for (pass = 0; pass < 2; pass++) {
3897 if (size == 2) {
3898 if (load) {
3899 tmp = tcg_temp_new_i32();
3900 gen_aa32_ld32u(tmp, addr, IS_USER(s));
3901 neon_store_reg(rd, pass, tmp);
3902 } else {
3903 tmp = neon_load_reg(rd, pass);
3904 gen_aa32_st32(tmp, addr, IS_USER(s));
3905 tcg_temp_free_i32(tmp);
3906 }
3907 tcg_gen_addi_i32(addr, addr, stride);
3908 } else if (size == 1) {
3909 if (load) {
3910 tmp = tcg_temp_new_i32();
3911 gen_aa32_ld16u(tmp, addr, IS_USER(s));
3912 tcg_gen_addi_i32(addr, addr, stride);
3913 tmp2 = tcg_temp_new_i32();
3914 gen_aa32_ld16u(tmp2, addr, IS_USER(s));
3915 tcg_gen_addi_i32(addr, addr, stride);
3916 tcg_gen_shli_i32(tmp2, tmp2, 16);
3917 tcg_gen_or_i32(tmp, tmp, tmp2);
3918 tcg_temp_free_i32(tmp2);
3919 neon_store_reg(rd, pass, tmp);
3920 } else {
3921 tmp = neon_load_reg(rd, pass);
3922 tmp2 = tcg_temp_new_i32();
3923 tcg_gen_shri_i32(tmp2, tmp, 16);
3924 gen_aa32_st16(tmp, addr, IS_USER(s));
3925 tcg_temp_free_i32(tmp);
3926 tcg_gen_addi_i32(addr, addr, stride);
3927 gen_aa32_st16(tmp2, addr, IS_USER(s));
3928 tcg_temp_free_i32(tmp2);
3929 tcg_gen_addi_i32(addr, addr, stride);
3930 }
3931 } else /* size == 0 */ {
3932 if (load) {
3933 TCGV_UNUSED_I32(tmp2);
3934 for (n = 0; n < 4; n++) {
3935 tmp = tcg_temp_new_i32();
3936 gen_aa32_ld8u(tmp, addr, IS_USER(s));
3937 tcg_gen_addi_i32(addr, addr, stride);
3938 if (n == 0) {
3939 tmp2 = tmp;
3940 } else {
3941 tcg_gen_shli_i32(tmp, tmp, n * 8);
3942 tcg_gen_or_i32(tmp2, tmp2, tmp);
3943 tcg_temp_free_i32(tmp);
3944 }
3945 }
3946 neon_store_reg(rd, pass, tmp2);
3947 } else {
3948 tmp2 = neon_load_reg(rd, pass);
3949 for (n = 0; n < 4; n++) {
3950 tmp = tcg_temp_new_i32();
3951 if (n == 0) {
3952 tcg_gen_mov_i32(tmp, tmp2);
3953 } else {
3954 tcg_gen_shri_i32(tmp, tmp2, n * 8);
3955 }
3956 gen_aa32_st8(tmp, addr, IS_USER(s));
3957 tcg_temp_free_i32(tmp);
3958 tcg_gen_addi_i32(addr, addr, stride);
3959 }
3960 tcg_temp_free_i32(tmp2);
3961 }
3962 }
3963 }
3964 }
3965 rd += spacing;
3966 }
3967 tcg_temp_free_i32(addr);
3968 stride = nregs * 8;
3969 } else {
3970 size = (insn >> 10) & 3;
3971 if (size == 3) {
3972 /* Load single element to all lanes. */
3973 int a = (insn >> 4) & 1;
3974 if (!load) {
3975 return 1;
3976 }
3977 size = (insn >> 6) & 3;
3978 nregs = ((insn >> 8) & 3) + 1;
3979
3980 if (size == 3) {
3981 if (nregs != 4 || a == 0) {
3982 return 1;
3983 }
3984 /* For VLD4 size==3 a == 1 means 32 bits at 16 byte alignment */
3985 size = 2;
3986 }
3987 if (nregs == 1 && a == 1 && size == 0) {
3988 return 1;
3989 }
3990 if (nregs == 3 && a == 1) {
3991 return 1;
3992 }
3993 addr = tcg_temp_new_i32();
3994 load_reg_var(s, addr, rn);
3995 if (nregs == 1) {
3996 /* VLD1 to all lanes: bit 5 indicates how many Dregs to write */
3997 tmp = gen_load_and_replicate(s, addr, size);
3998 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 0));
3999 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 1));
4000 if (insn & (1 << 5)) {
4001 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd + 1, 0));
4002 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd + 1, 1));
4003 }
4004 tcg_temp_free_i32(tmp);
4005 } else {
4006 /* VLD2/3/4 to all lanes: bit 5 indicates register stride */
4007 stride = (insn & (1 << 5)) ? 2 : 1;
4008 for (reg = 0; reg < nregs; reg++) {
4009 tmp = gen_load_and_replicate(s, addr, size);
4010 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 0));
4011 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 1));
4012 tcg_temp_free_i32(tmp);
4013 tcg_gen_addi_i32(addr, addr, 1 << size);
4014 rd += stride;
4015 }
4016 }
4017 tcg_temp_free_i32(addr);
4018 stride = (1 << size) * nregs;
4019 } else {
4020 /* Single element. */
4021 int idx = (insn >> 4) & 0xf;
4022 pass = (insn >> 7) & 1;
4023 switch (size) {
4024 case 0:
4025 shift = ((insn >> 5) & 3) * 8;
4026 stride = 1;
4027 break;
4028 case 1:
4029 shift = ((insn >> 6) & 1) * 16;
4030 stride = (insn & (1 << 5)) ? 2 : 1;
4031 break;
4032 case 2:
4033 shift = 0;
4034 stride = (insn & (1 << 6)) ? 2 : 1;
4035 break;
4036 default:
4037 abort();
4038 }
4039 nregs = ((insn >> 8) & 3) + 1;
4040 /* Catch the UNDEF cases. This is unavoidably a bit messy. */
4041 switch (nregs) {
4042 case 1:
4043 if (((idx & (1 << size)) != 0) ||
4044 (size == 2 && ((idx & 3) == 1 || (idx & 3) == 2))) {
4045 return 1;
4046 }
4047 break;
4048 case 3:
4049 if ((idx & 1) != 0) {
4050 return 1;
4051 }
4052 /* fall through */
4053 case 2:
4054 if (size == 2 && (idx & 2) != 0) {
4055 return 1;
4056 }
4057 break;
4058 case 4:
4059 if ((size == 2) && ((idx & 3) == 3)) {
4060 return 1;
4061 }
4062 break;
4063 default:
4064 abort();
4065 }
4066 if ((rd + stride * (nregs - 1)) > 31) {
4067 /* Attempts to write off the end of the register file
4068 * are UNPREDICTABLE; we choose to UNDEF because otherwise
4069 * the neon_load_reg() would write off the end of the array.
4070 */
4071 return 1;
4072 }
4073 addr = tcg_temp_new_i32();
4074 load_reg_var(s, addr, rn);
4075 for (reg = 0; reg < nregs; reg++) {
4076 if (load) {
4077 tmp = tcg_temp_new_i32();
4078 switch (size) {
4079 case 0:
4080 gen_aa32_ld8u(tmp, addr, IS_USER(s));
4081 break;
4082 case 1:
4083 gen_aa32_ld16u(tmp, addr, IS_USER(s));
4084 break;
4085 case 2:
4086 gen_aa32_ld32u(tmp, addr, IS_USER(s));
4087 break;
4088 default: /* Avoid compiler warnings. */
4089 abort();
4090 }
4091 if (size != 2) {
4092 tmp2 = neon_load_reg(rd, pass);
4093 tcg_gen_deposit_i32(tmp, tmp2, tmp,
4094 shift, size ? 16 : 8);
4095 tcg_temp_free_i32(tmp2);
4096 }
4097 neon_store_reg(rd, pass, tmp);
4098 } else { /* Store */
4099 tmp = neon_load_reg(rd, pass);
4100 if (shift)
4101 tcg_gen_shri_i32(tmp, tmp, shift);
4102 switch (size) {
4103 case 0:
4104 gen_aa32_st8(tmp, addr, IS_USER(s));
4105 break;
4106 case 1:
4107 gen_aa32_st16(tmp, addr, IS_USER(s));
4108 break;
4109 case 2:
4110 gen_aa32_st32(tmp, addr, IS_USER(s));
4111 break;
4112 }
4113 tcg_temp_free_i32(tmp);
4114 }
4115 rd += stride;
4116 tcg_gen_addi_i32(addr, addr, 1 << size);
4117 }
4118 tcg_temp_free_i32(addr);
4119 stride = nregs * (1 << size);
4120 }
4121 }
4122 if (rm != 15) {
4123 TCGv_i32 base;
4124
4125 base = load_reg(s, rn);
4126 if (rm == 13) {
4127 tcg_gen_addi_i32(base, base, stride);
4128 } else {
4129 TCGv_i32 index;
4130 index = load_reg(s, rm);
4131 tcg_gen_add_i32(base, base, index);
4132 tcg_temp_free_i32(index);
4133 }
4134 store_reg(s, rn, base);
4135 }
4136 return 0;
4137 }
4138
4139 /* Bitwise select. dest = c ? t : f. Clobbers T and F. */
4140 static void gen_neon_bsl(TCGv_i32 dest, TCGv_i32 t, TCGv_i32 f, TCGv_i32 c)
4141 {
4142 tcg_gen_and_i32(t, t, c);
4143 tcg_gen_andc_i32(f, f, c);
4144 tcg_gen_or_i32(dest, t, f);
4145 }
4146
4147 static inline void gen_neon_narrow(int size, TCGv_i32 dest, TCGv_i64 src)
4148 {
4149 switch (size) {
4150 case 0: gen_helper_neon_narrow_u8(dest, src); break;
4151 case 1: gen_helper_neon_narrow_u16(dest, src); break;
4152 case 2: tcg_gen_trunc_i64_i32(dest, src); break;
4153 default: abort();
4154 }
4155 }
4156
4157 static inline void gen_neon_narrow_sats(int size, TCGv_i32 dest, TCGv_i64 src)
4158 {
4159 switch (size) {
4160 case 0: gen_helper_neon_narrow_sat_s8(dest, cpu_env, src); break;
4161 case 1: gen_helper_neon_narrow_sat_s16(dest, cpu_env, src); break;
4162 case 2: gen_helper_neon_narrow_sat_s32(dest, cpu_env, src); break;
4163 default: abort();
4164 }
4165 }
4166
4167 static inline void gen_neon_narrow_satu(int size, TCGv_i32 dest, TCGv_i64 src)
4168 {
4169 switch (size) {
4170 case 0: gen_helper_neon_narrow_sat_u8(dest, cpu_env, src); break;
4171 case 1: gen_helper_neon_narrow_sat_u16(dest, cpu_env, src); break;
4172 case 2: gen_helper_neon_narrow_sat_u32(dest, cpu_env, src); break;
4173 default: abort();
4174 }
4175 }
4176
4177 static inline void gen_neon_unarrow_sats(int size, TCGv_i32 dest, TCGv_i64 src)
4178 {
4179 switch (size) {
4180 case 0: gen_helper_neon_unarrow_sat8(dest, cpu_env, src); break;
4181 case 1: gen_helper_neon_unarrow_sat16(dest, cpu_env, src); break;
4182 case 2: gen_helper_neon_unarrow_sat32(dest, cpu_env, src); break;
4183 default: abort();
4184 }
4185 }
4186
4187 static inline void gen_neon_shift_narrow(int size, TCGv_i32 var, TCGv_i32 shift,
4188 int q, int u)
4189 {
4190 if (q) {
4191 if (u) {
4192 switch (size) {
4193 case 1: gen_helper_neon_rshl_u16(var, var, shift); break;
4194 case 2: gen_helper_neon_rshl_u32(var, var, shift); break;
4195 default: abort();
4196 }
4197 } else {
4198 switch (size) {
4199 case 1: gen_helper_neon_rshl_s16(var, var, shift); break;
4200 case 2: gen_helper_neon_rshl_s32(var, var, shift); break;
4201 default: abort();
4202 }
4203 }
4204 } else {
4205 if (u) {
4206 switch (size) {
4207 case 1: gen_helper_neon_shl_u16(var, var, shift); break;
4208 case 2: gen_helper_neon_shl_u32(var, var, shift); break;
4209 default: abort();
4210 }
4211 } else {
4212 switch (size) {
4213 case 1: gen_helper_neon_shl_s16(var, var, shift); break;
4214 case 2: gen_helper_neon_shl_s32(var, var, shift); break;
4215 default: abort();
4216 }
4217 }
4218 }
4219 }
4220
4221 static inline void gen_neon_widen(TCGv_i64 dest, TCGv_i32 src, int size, int u)
4222 {
4223 if (u) {
4224 switch (size) {
4225 case 0: gen_helper_neon_widen_u8(dest, src); break;
4226 case 1: gen_helper_neon_widen_u16(dest, src); break;
4227 case 2: tcg_gen_extu_i32_i64(dest, src); break;
4228 default: abort();
4229 }
4230 } else {
4231 switch (size) {
4232 case 0: gen_helper_neon_widen_s8(dest, src); break;
4233 case 1: gen_helper_neon_widen_s16(dest, src); break;
4234 case 2: tcg_gen_ext_i32_i64(dest, src); break;
4235 default: abort();
4236 }
4237 }
4238 tcg_temp_free_i32(src);
4239 }
4240
4241 static inline void gen_neon_addl(int size)
4242 {
4243 switch (size) {
4244 case 0: gen_helper_neon_addl_u16(CPU_V001); break;
4245 case 1: gen_helper_neon_addl_u32(CPU_V001); break;
4246 case 2: tcg_gen_add_i64(CPU_V001); break;
4247 default: abort();
4248 }
4249 }
4250
4251 static inline void gen_neon_subl(int size)
4252 {
4253 switch (size) {
4254 case 0: gen_helper_neon_subl_u16(CPU_V001); break;
4255 case 1: gen_helper_neon_subl_u32(CPU_V001); break;
4256 case 2: tcg_gen_sub_i64(CPU_V001); break;
4257 default: abort();
4258 }
4259 }
4260
4261 static inline void gen_neon_negl(TCGv_i64 var, int size)
4262 {
4263 switch (size) {
4264 case 0: gen_helper_neon_negl_u16(var, var); break;
4265 case 1: gen_helper_neon_negl_u32(var, var); break;
4266 case 2:
4267 tcg_gen_neg_i64(var, var);
4268 break;
4269 default: abort();
4270 }
4271 }
4272
4273 static inline void gen_neon_addl_saturate(TCGv_i64 op0, TCGv_i64 op1, int size)
4274 {
4275 switch (size) {
4276 case 1: gen_helper_neon_addl_saturate_s32(op0, cpu_env, op0, op1); break;
4277 case 2: gen_helper_neon_addl_saturate_s64(op0, cpu_env, op0, op1); break;
4278 default: abort();
4279 }
4280 }
4281
4282 static inline void gen_neon_mull(TCGv_i64 dest, TCGv_i32 a, TCGv_i32 b,
4283 int size, int u)
4284 {
4285 TCGv_i64 tmp;
4286
4287 switch ((size << 1) | u) {
4288 case 0: gen_helper_neon_mull_s8(dest, a, b); break;
4289 case 1: gen_helper_neon_mull_u8(dest, a, b); break;
4290 case 2: gen_helper_neon_mull_s16(dest, a, b); break;
4291 case 3: gen_helper_neon_mull_u16(dest, a, b); break;
4292 case 4:
4293 tmp = gen_muls_i64_i32(a, b);
4294 tcg_gen_mov_i64(dest, tmp);
4295 tcg_temp_free_i64(tmp);
4296 break;
4297 case 5:
4298 tmp = gen_mulu_i64_i32(a, b);
4299 tcg_gen_mov_i64(dest, tmp);
4300 tcg_temp_free_i64(tmp);
4301 break;
4302 default: abort();
4303 }
4304
4305 /* gen_helper_neon_mull_[su]{8|16} do not free their parameters.
4306 Don't forget to clean them now. */
4307 if (size < 2) {
4308 tcg_temp_free_i32(a);
4309 tcg_temp_free_i32(b);
4310 }
4311 }
4312
4313 static void gen_neon_narrow_op(int op, int u, int size,
4314 TCGv_i32 dest, TCGv_i64 src)
4315 {
4316 if (op) {
4317 if (u) {
4318 gen_neon_unarrow_sats(size, dest, src);
4319 } else {
4320 gen_neon_narrow(size, dest, src);
4321 }
4322 } else {
4323 if (u) {
4324 gen_neon_narrow_satu(size, dest, src);
4325 } else {
4326 gen_neon_narrow_sats(size, dest, src);
4327 }
4328 }
4329 }
4330
4331 /* Symbolic constants for op fields for Neon 3-register same-length.
4332 * The values correspond to bits [11:8,4]; see the ARM ARM DDI0406B
4333 * table A7-9.
4334 */
4335 #define NEON_3R_VHADD 0
4336 #define NEON_3R_VQADD 1
4337 #define NEON_3R_VRHADD 2
4338 #define NEON_3R_LOGIC 3 /* VAND,VBIC,VORR,VMOV,VORN,VEOR,VBIF,VBIT,VBSL */
4339 #define NEON_3R_VHSUB 4
4340 #define NEON_3R_VQSUB 5
4341 #define NEON_3R_VCGT 6
4342 #define NEON_3R_VCGE 7
4343 #define NEON_3R_VSHL 8
4344 #define NEON_3R_VQSHL 9
4345 #define NEON_3R_VRSHL 10
4346 #define NEON_3R_VQRSHL 11
4347 #define NEON_3R_VMAX 12
4348 #define NEON_3R_VMIN 13
4349 #define NEON_3R_VABD 14
4350 #define NEON_3R_VABA 15
4351 #define NEON_3R_VADD_VSUB 16
4352 #define NEON_3R_VTST_VCEQ 17
4353 #define NEON_3R_VML 18 /* VMLA, VMLAL, VMLS, VMLSL */
4354 #define NEON_3R_VMUL 19
4355 #define NEON_3R_VPMAX 20
4356 #define NEON_3R_VPMIN 21
4357 #define NEON_3R_VQDMULH_VQRDMULH 22
4358 #define NEON_3R_VPADD 23
4359 #define NEON_3R_VFM 25 /* VFMA, VFMS : float fused multiply-add */
4360 #define NEON_3R_FLOAT_ARITH 26 /* float VADD, VSUB, VPADD, VABD */
4361 #define NEON_3R_FLOAT_MULTIPLY 27 /* float VMLA, VMLS, VMUL */
4362 #define NEON_3R_FLOAT_CMP 28 /* float VCEQ, VCGE, VCGT */
4363 #define NEON_3R_FLOAT_ACMP 29 /* float VACGE, VACGT, VACLE, VACLT */
4364 #define NEON_3R_FLOAT_MINMAX 30 /* float VMIN, VMAX */
4365 #define NEON_3R_VRECPS_VRSQRTS 31 /* float VRECPS, VRSQRTS */
4366
4367 static const uint8_t neon_3r_sizes[] = {
4368 [NEON_3R_VHADD] = 0x7,
4369 [NEON_3R_VQADD] = 0xf,
4370 [NEON_3R_VRHADD] = 0x7,
4371 [NEON_3R_LOGIC] = 0xf, /* size field encodes op type */
4372 [NEON_3R_VHSUB] = 0x7,
4373 [NEON_3R_VQSUB] = 0xf,
4374 [NEON_3R_VCGT] = 0x7,
4375 [NEON_3R_VCGE] = 0x7,
4376 [NEON_3R_VSHL] = 0xf,
4377 [NEON_3R_VQSHL] = 0xf,
4378 [NEON_3R_VRSHL] = 0xf,
4379 [NEON_3R_VQRSHL] = 0xf,
4380 [NEON_3R_VMAX] = 0x7,
4381 [NEON_3R_VMIN] = 0x7,
4382 [NEON_3R_VABD] = 0x7,
4383 [NEON_3R_VABA] = 0x7,
4384 [NEON_3R_VADD_VSUB] = 0xf,
4385 [NEON_3R_VTST_VCEQ] = 0x7,
4386 [NEON_3R_VML] = 0x7,
4387 [NEON_3R_VMUL] = 0x7,
4388 [NEON_3R_VPMAX] = 0x7,
4389 [NEON_3R_VPMIN] = 0x7,
4390 [NEON_3R_VQDMULH_VQRDMULH] = 0x6,
4391 [NEON_3R_VPADD] = 0x7,
4392 [NEON_3R_VFM] = 0x5, /* size bit 1 encodes op */
4393 [NEON_3R_FLOAT_ARITH] = 0x5, /* size bit 1 encodes op */
4394 [NEON_3R_FLOAT_MULTIPLY] = 0x5, /* size bit 1 encodes op */
4395 [NEON_3R_FLOAT_CMP] = 0x5, /* size bit 1 encodes op */
4396 [NEON_3R_FLOAT_ACMP] = 0x5, /* size bit 1 encodes op */
4397 [NEON_3R_FLOAT_MINMAX] = 0x5, /* size bit 1 encodes op */
4398 [NEON_3R_VRECPS_VRSQRTS] = 0x5, /* size bit 1 encodes op */
4399 };
4400
4401 /* Symbolic constants for op fields for Neon 2-register miscellaneous.
4402 * The values correspond to bits [17:16,10:7]; see the ARM ARM DDI0406B
4403 * table A7-13.
4404 */
4405 #define NEON_2RM_VREV64 0
4406 #define NEON_2RM_VREV32 1
4407 #define NEON_2RM_VREV16 2
4408 #define NEON_2RM_VPADDL 4
4409 #define NEON_2RM_VPADDL_U 5
4410 #define NEON_2RM_VCLS 8
4411 #define NEON_2RM_VCLZ 9
4412 #define NEON_2RM_VCNT 10
4413 #define NEON_2RM_VMVN 11
4414 #define NEON_2RM_VPADAL 12
4415 #define NEON_2RM_VPADAL_U 13
4416 #define NEON_2RM_VQABS 14
4417 #define NEON_2RM_VQNEG 15
4418 #define NEON_2RM_VCGT0 16
4419 #define NEON_2RM_VCGE0 17
4420 #define NEON_2RM_VCEQ0 18
4421 #define NEON_2RM_VCLE0 19
4422 #define NEON_2RM_VCLT0 20
4423 #define NEON_2RM_VABS 22
4424 #define NEON_2RM_VNEG 23
4425 #define NEON_2RM_VCGT0_F 24
4426 #define NEON_2RM_VCGE0_F 25
4427 #define NEON_2RM_VCEQ0_F 26
4428 #define NEON_2RM_VCLE0_F 27
4429 #define NEON_2RM_VCLT0_F 28
4430 #define NEON_2RM_VABS_F 30
4431 #define NEON_2RM_VNEG_F 31
4432 #define NEON_2RM_VSWP 32
4433 #define NEON_2RM_VTRN 33
4434 #define NEON_2RM_VUZP 34
4435 #define NEON_2RM_VZIP 35
4436 #define NEON_2RM_VMOVN 36 /* Includes VQMOVN, VQMOVUN */
4437 #define NEON_2RM_VQMOVN 37 /* Includes VQMOVUN */
4438 #define NEON_2RM_VSHLL 38
4439 #define NEON_2RM_VCVT_F16_F32 44
4440 #define NEON_2RM_VCVT_F32_F16 46
4441 #define NEON_2RM_VRECPE 56
4442 #define NEON_2RM_VRSQRTE 57
4443 #define NEON_2RM_VRECPE_F 58
4444 #define NEON_2RM_VRSQRTE_F 59
4445 #define NEON_2RM_VCVT_FS 60
4446 #define NEON_2RM_VCVT_FU 61
4447 #define NEON_2RM_VCVT_SF 62
4448 #define NEON_2RM_VCVT_UF 63
4449
4450 static int neon_2rm_is_float_op(int op)
4451 {
4452 /* Return true if this neon 2reg-misc op is float-to-float */
4453 return (op == NEON_2RM_VABS_F || op == NEON_2RM_VNEG_F ||
4454 op >= NEON_2RM_VRECPE_F);
4455 }
4456
4457 /* Each entry in this array has bit n set if the insn allows
4458 * size value n (otherwise it will UNDEF). Since unallocated
4459 * op values will have no bits set they always UNDEF.
4460 */
4461 static const uint8_t neon_2rm_sizes[] = {
4462 [NEON_2RM_VREV64] = 0x7,
4463 [NEON_2RM_VREV32] = 0x3,
4464 [NEON_2RM_VREV16] = 0x1,
4465 [NEON_2RM_VPADDL] = 0x7,
4466 [NEON_2RM_VPADDL_U] = 0x7,
4467 [NEON_2RM_VCLS] = 0x7,
4468 [NEON_2RM_VCLZ] = 0x7,
4469 [NEON_2RM_VCNT] = 0x1,
4470 [NEON_2RM_VMVN] = 0x1,
4471 [NEON_2RM_VPADAL] = 0x7,
4472 [NEON_2RM_VPADAL_U] = 0x7,
4473 [NEON_2RM_VQABS] = 0x7,
4474 [NEON_2RM_VQNEG] = 0x7,
4475 [NEON_2RM_VCGT0] = 0x7,
4476 [NEON_2RM_VCGE0] = 0x7,
4477 [NEON_2RM_VCEQ0] = 0x7,
4478 [NEON_2RM_VCLE0] = 0x7,
4479 [NEON_2RM_VCLT0] = 0x7,
4480 [NEON_2RM_VABS] = 0x7,
4481 [NEON_2RM_VNEG] = 0x7,
4482 [NEON_2RM_VCGT0_F] = 0x4,
4483 [NEON_2RM_VCGE0_F] = 0x4,
4484 [NEON_2RM_VCEQ0_F] = 0x4,
4485 [NEON_2RM_VCLE0_F] = 0x4,
4486 [NEON_2RM_VCLT0_F] = 0x4,
4487 [NEON_2RM_VABS_F] = 0x4,
4488 [NEON_2RM_VNEG_F] = 0x4,
4489 [NEON_2RM_VSWP] = 0x1,
4490 [NEON_2RM_VTRN] = 0x7,
4491 [NEON_2RM_VUZP] = 0x7,
4492 [NEON_2RM_VZIP] = 0x7,
4493 [NEON_2RM_VMOVN] = 0x7,
4494 [NEON_2RM_VQMOVN] = 0x7,
4495 [NEON_2RM_VSHLL] = 0x7,
4496 [NEON_2RM_VCVT_F16_F32] = 0x2,
4497 [NEON_2RM_VCVT_F32_F16] = 0x2,
4498 [NEON_2RM_VRECPE] = 0x4,
4499 [NEON_2RM_VRSQRTE] = 0x4,
4500 [NEON_2RM_VRECPE_F] = 0x4,
4501 [NEON_2RM_VRSQRTE_F] = 0x4,
4502 [NEON_2RM_VCVT_FS] = 0x4,
4503 [NEON_2RM_VCVT_FU] = 0x4,
4504 [NEON_2RM_VCVT_SF] = 0x4,
4505 [NEON_2RM_VCVT_UF] = 0x4,
4506 };
4507
4508 /* Translate a NEON data processing instruction. Return nonzero if the
4509 instruction is invalid.
4510 We process data in a mixture of 32-bit and 64-bit chunks.
4511 Mostly we use 32-bit chunks so we can use normal scalar instructions. */
4512
4513 static int disas_neon_data_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
4514 {
4515 int op;
4516 int q;
4517 int rd, rn, rm;
4518 int size;
4519 int shift;
4520 int pass;
4521 int count;
4522 int pairwise;
4523 int u;
4524 uint32_t imm, mask;
4525 TCGv_i32 tmp, tmp2, tmp3, tmp4, tmp5;
4526 TCGv_i64 tmp64;
4527
4528 if (!s->vfp_enabled)
4529 return 1;
4530 q = (insn & (1 << 6)) != 0;
4531 u = (insn >> 24) & 1;
4532 VFP_DREG_D(rd, insn);
4533 VFP_DREG_N(rn, insn);
4534 VFP_DREG_M(rm, insn);
4535 size = (insn >> 20) & 3;
4536 if ((insn & (1 << 23)) == 0) {
4537 /* Three register same length. */
4538 op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1);
4539 /* Catch invalid op and bad size combinations: UNDEF */
4540 if ((neon_3r_sizes[op] & (1 << size)) == 0) {
4541 return 1;
4542 }
4543 /* All insns of this form UNDEF for either this condition or the
4544 * superset of cases "Q==1"; we catch the latter later.
4545 */
4546 if (q && ((rd | rn | rm) & 1)) {
4547 return 1;
4548 }
4549 if (size == 3 && op != NEON_3R_LOGIC) {
4550 /* 64-bit element instructions. */
4551 for (pass = 0; pass < (q ? 2 : 1); pass++) {
4552 neon_load_reg64(cpu_V0, rn + pass);
4553 neon_load_reg64(cpu_V1, rm + pass);
4554 switch (op) {
4555 case NEON_3R_VQADD:
4556 if (u) {
4557 gen_helper_neon_qadd_u64(cpu_V0, cpu_env,
4558 cpu_V0, cpu_V1);
4559 } else {
4560 gen_helper_neon_qadd_s64(cpu_V0, cpu_env,
4561 cpu_V0, cpu_V1);
4562 }
4563 break;
4564 case NEON_3R_VQSUB:
4565 if (u) {
4566 gen_helper_neon_qsub_u64(cpu_V0, cpu_env,
4567 cpu_V0, cpu_V1);
4568 } else {
4569 gen_helper_neon_qsub_s64(cpu_V0, cpu_env,
4570 cpu_V0, cpu_V1);
4571 }
4572 break;
4573 case NEON_3R_VSHL:
4574 if (u) {
4575 gen_helper_neon_shl_u64(cpu_V0, cpu_V1, cpu_V0);
4576 } else {
4577 gen_helper_neon_shl_s64(cpu_V0, cpu_V1, cpu_V0);
4578 }
4579 break;
4580 case NEON_3R_VQSHL:
4581 if (u) {
4582 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
4583 cpu_V1, cpu_V0);
4584 } else {
4585 gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
4586 cpu_V1, cpu_V0);
4587 }
4588 break;
4589 case NEON_3R_VRSHL:
4590 if (u) {
4591 gen_helper_neon_rshl_u64(cpu_V0, cpu_V1, cpu_V0);
4592 } else {
4593 gen_helper_neon_rshl_s64(cpu_V0, cpu_V1, cpu_V0);
4594 }
4595 break;
4596 case NEON_3R_VQRSHL:
4597 if (u) {
4598 gen_helper_neon_qrshl_u64(cpu_V0, cpu_env,
4599 cpu_V1, cpu_V0);
4600 } else {
4601 gen_helper_neon_qrshl_s64(cpu_V0, cpu_env,
4602 cpu_V1, cpu_V0);
4603 }
4604 break;
4605 case NEON_3R_VADD_VSUB:
4606 if (u) {
4607 tcg_gen_sub_i64(CPU_V001);
4608 } else {
4609 tcg_gen_add_i64(CPU_V001);
4610 }
4611 break;
4612 default:
4613 abort();
4614 }
4615 neon_store_reg64(cpu_V0, rd + pass);
4616 }
4617 return 0;
4618 }
4619 pairwise = 0;
4620 switch (op) {
4621 case NEON_3R_VSHL:
4622 case NEON_3R_VQSHL:
4623 case NEON_3R_VRSHL:
4624 case NEON_3R_VQRSHL:
4625 {
4626 int rtmp;
4627 /* Shift instruction operands are reversed. */
4628 rtmp = rn;
4629 rn = rm;
4630 rm = rtmp;
4631 }
4632 break;
4633 case NEON_3R_VPADD:
4634 if (u) {
4635 return 1;
4636 }
4637 /* Fall through */
4638 case NEON_3R_VPMAX:
4639 case NEON_3R_VPMIN:
4640 pairwise = 1;
4641 break;
4642 case NEON_3R_FLOAT_ARITH:
4643 pairwise = (u && size < 2); /* if VPADD (float) */
4644 break;
4645 case NEON_3R_FLOAT_MINMAX:
4646 pairwise = u; /* if VPMIN/VPMAX (float) */
4647 break;
4648 case NEON_3R_FLOAT_CMP:
4649 if (!u && size) {
4650 /* no encoding for U=0 C=1x */
4651 return 1;
4652 }
4653 break;
4654 case NEON_3R_FLOAT_ACMP:
4655 if (!u) {
4656 return 1;
4657 }
4658 break;
4659 case NEON_3R_VRECPS_VRSQRTS:
4660 if (u) {
4661 return 1;
4662 }
4663 break;
4664 case NEON_3R_VMUL:
4665 if (u && (size != 0)) {
4666 /* UNDEF on invalid size for polynomial subcase */
4667 return 1;
4668 }
4669 break;
4670 case NEON_3R_VFM:
4671 if (!arm_feature(env, ARM_FEATURE_VFP4) || u) {
4672 return 1;
4673 }
4674 break;
4675 default:
4676 break;
4677 }
4678
4679 if (pairwise && q) {
4680 /* All the pairwise insns UNDEF if Q is set */
4681 return 1;
4682 }
4683
4684 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4685
4686 if (pairwise) {
4687 /* Pairwise. */
4688 if (pass < 1) {
4689 tmp = neon_load_reg(rn, 0);
4690 tmp2 = neon_load_reg(rn, 1);
4691 } else {
4692 tmp = neon_load_reg(rm, 0);
4693 tmp2 = neon_load_reg(rm, 1);
4694 }
4695 } else {
4696 /* Elementwise. */
4697 tmp = neon_load_reg(rn, pass);
4698 tmp2 = neon_load_reg(rm, pass);
4699 }
4700 switch (op) {
4701 case NEON_3R_VHADD:
4702 GEN_NEON_INTEGER_OP(hadd);
4703 break;
4704 case NEON_3R_VQADD:
4705 GEN_NEON_INTEGER_OP_ENV(qadd);
4706 break;
4707 case NEON_3R_VRHADD:
4708 GEN_NEON_INTEGER_OP(rhadd);
4709 break;
4710 case NEON_3R_LOGIC: /* Logic ops. */
4711 switch ((u << 2) | size) {
4712 case 0: /* VAND */
4713 tcg_gen_and_i32(tmp, tmp, tmp2);
4714 break;
4715 case 1: /* BIC */
4716 tcg_gen_andc_i32(tmp, tmp, tmp2);
4717 break;
4718 case 2: /* VORR */
4719 tcg_gen_or_i32(tmp, tmp, tmp2);
4720 break;
4721 case 3: /* VORN */
4722 tcg_gen_orc_i32(tmp, tmp, tmp2);
4723 break;
4724 case 4: /* VEOR */
4725 tcg_gen_xor_i32(tmp, tmp, tmp2);
4726 break;
4727 case 5: /* VBSL */
4728 tmp3 = neon_load_reg(rd, pass);
4729 gen_neon_bsl(tmp, tmp, tmp2, tmp3);
4730 tcg_temp_free_i32(tmp3);
4731 break;
4732 case 6: /* VBIT */
4733 tmp3 = neon_load_reg(rd, pass);
4734 gen_neon_bsl(tmp, tmp, tmp3, tmp2);
4735 tcg_temp_free_i32(tmp3);
4736 break;
4737 case 7: /* VBIF */
4738 tmp3 = neon_load_reg(rd, pass);
4739 gen_neon_bsl(tmp, tmp3, tmp, tmp2);
4740 tcg_temp_free_i32(tmp3);
4741 break;
4742 }
4743 break;
4744 case NEON_3R_VHSUB:
4745 GEN_NEON_INTEGER_OP(hsub);
4746 break;
4747 case NEON_3R_VQSUB:
4748 GEN_NEON_INTEGER_OP_ENV(qsub);
4749 break;
4750 case NEON_3R_VCGT:
4751 GEN_NEON_INTEGER_OP(cgt);
4752 break;
4753 case NEON_3R_VCGE:
4754 GEN_NEON_INTEGER_OP(cge);
4755 break;
4756 case NEON_3R_VSHL:
4757 GEN_NEON_INTEGER_OP(shl);
4758 break;
4759 case NEON_3R_VQSHL:
4760 GEN_NEON_INTEGER_OP_ENV(qshl);
4761 break;
4762 case NEON_3R_VRSHL:
4763 GEN_NEON_INTEGER_OP(rshl);
4764 break;
4765 case NEON_3R_VQRSHL:
4766 GEN_NEON_INTEGER_OP_ENV(qrshl);
4767 break;
4768 case NEON_3R_VMAX:
4769 GEN_NEON_INTEGER_OP(max);
4770 break;
4771 case NEON_3R_VMIN:
4772 GEN_NEON_INTEGER_OP(min);
4773 break;
4774 case NEON_3R_VABD:
4775 GEN_NEON_INTEGER_OP(abd);
4776 break;
4777 case NEON_3R_VABA:
4778 GEN_NEON_INTEGER_OP(abd);
4779 tcg_temp_free_i32(tmp2);
4780 tmp2 = neon_load_reg(rd, pass);
4781 gen_neon_add(size, tmp, tmp2);
4782 break;
4783 case NEON_3R_VADD_VSUB:
4784 if (!u) { /* VADD */
4785 gen_neon_add(size, tmp, tmp2);
4786 } else { /* VSUB */
4787 switch (size) {
4788 case 0: gen_helper_neon_sub_u8(tmp, tmp, tmp2); break;
4789 case 1: gen_helper_neon_sub_u16(tmp, tmp, tmp2); break;
4790 case 2: tcg_gen_sub_i32(tmp, tmp, tmp2); break;
4791 default: abort();
4792 }
4793 }
4794 break;
4795 case NEON_3R_VTST_VCEQ:
4796 if (!u) { /* VTST */
4797 switch (size) {
4798 case 0: gen_helper_neon_tst_u8(tmp, tmp, tmp2); break;
4799 case 1: gen_helper_neon_tst_u16(tmp, tmp, tmp2); break;
4800 case 2: gen_helper_neon_tst_u32(tmp, tmp, tmp2); break;
4801 default: abort();
4802 }
4803 } else { /* VCEQ */
4804 switch (size) {
4805 case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break;
4806 case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break;
4807 case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break;
4808 default: abort();
4809 }
4810 }
4811 break;
4812 case NEON_3R_VML: /* VMLA, VMLAL, VMLS,VMLSL */
4813 switch (size) {
4814 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
4815 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
4816 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
4817 default: abort();
4818 }
4819 tcg_temp_free_i32(tmp2);
4820 tmp2 = neon_load_reg(rd, pass);
4821 if (u) { /* VMLS */
4822 gen_neon_rsb(size, tmp, tmp2);
4823 } else { /* VMLA */
4824 gen_neon_add(size, tmp, tmp2);
4825 }
4826 break;
4827 case NEON_3R_VMUL:
4828 if (u) { /* polynomial */
4829 gen_helper_neon_mul_p8(tmp, tmp, tmp2);
4830 } else { /* Integer */
4831 switch (size) {
4832 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
4833 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
4834 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
4835 default: abort();
4836 }
4837 }
4838 break;
4839 case NEON_3R_VPMAX:
4840 GEN_NEON_INTEGER_OP(pmax);
4841 break;
4842 case NEON_3R_VPMIN:
4843 GEN_NEON_INTEGER_OP(pmin);
4844 break;
4845 case NEON_3R_VQDMULH_VQRDMULH: /* Multiply high. */
4846 if (!u) { /* VQDMULH */
4847 switch (size) {
4848 case 1:
4849 gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
4850 break;
4851 case 2:
4852 gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
4853 break;
4854 default: abort();
4855 }
4856 } else { /* VQRDMULH */
4857 switch (size) {
4858 case 1:
4859 gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
4860 break;
4861 case 2:
4862 gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
4863 break;
4864 default: abort();
4865 }
4866 }
4867 break;
4868 case NEON_3R_VPADD:
4869 switch (size) {
4870 case 0: gen_helper_neon_padd_u8(tmp, tmp, tmp2); break;
4871 case 1: gen_helper_neon_padd_u16(tmp, tmp, tmp2); break;
4872 case 2: tcg_gen_add_i32(tmp, tmp, tmp2); break;
4873 default: abort();
4874 }
4875 break;
4876 case NEON_3R_FLOAT_ARITH: /* Floating point arithmetic. */
4877 {
4878 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
4879 switch ((u << 2) | size) {
4880 case 0: /* VADD */
4881 case 4: /* VPADD */
4882 gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
4883 break;
4884 case 2: /* VSUB */
4885 gen_helper_vfp_subs(tmp, tmp, tmp2, fpstatus);
4886 break;
4887 case 6: /* VABD */
4888 gen_helper_neon_abd_f32(tmp, tmp, tmp2, fpstatus);
4889 break;
4890 default:
4891 abort();
4892 }
4893 tcg_temp_free_ptr(fpstatus);
4894 break;
4895 }
4896 case NEON_3R_FLOAT_MULTIPLY:
4897 {
4898 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
4899 gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus);
4900 if (!u) {
4901 tcg_temp_free_i32(tmp2);
4902 tmp2 = neon_load_reg(rd, pass);
4903 if (size == 0) {
4904 gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
4905 } else {
4906 gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus);
4907 }
4908 }
4909 tcg_temp_free_ptr(fpstatus);
4910 break;
4911 }
4912 case NEON_3R_FLOAT_CMP:
4913 {
4914 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
4915 if (!u) {
4916 gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus);
4917 } else {
4918 if (size == 0) {
4919 gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus);
4920 } else {
4921 gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus);
4922 }
4923 }
4924 tcg_temp_free_ptr(fpstatus);
4925 break;
4926 }
4927 case NEON_3R_FLOAT_ACMP:
4928 {
4929 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
4930 if (size == 0) {
4931 gen_helper_neon_acge_f32(tmp, tmp, tmp2, fpstatus);
4932 } else {
4933 gen_helper_neon_acgt_f32(tmp, tmp, tmp2, fpstatus);
4934 }
4935 tcg_temp_free_ptr(fpstatus);
4936 break;
4937 }
4938 case NEON_3R_FLOAT_MINMAX:
4939 {
4940 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
4941 if (size == 0) {
4942 gen_helper_neon_max_f32(tmp, tmp, tmp2, fpstatus);
4943 } else {
4944 gen_helper_neon_min_f32(tmp, tmp, tmp2, fpstatus);
4945 }
4946 tcg_temp_free_ptr(fpstatus);
4947 break;
4948 }
4949 case NEON_3R_VRECPS_VRSQRTS:
4950 if (size == 0)
4951 gen_helper_recps_f32(tmp, tmp, tmp2, cpu_env);
4952 else
4953 gen_helper_rsqrts_f32(tmp, tmp, tmp2, cpu_env);
4954 break;
4955 case NEON_3R_VFM:
4956 {
4957 /* VFMA, VFMS: fused multiply-add */
4958 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
4959 TCGv_i32 tmp3 = neon_load_reg(rd, pass);
4960 if (size) {
4961 /* VFMS */
4962 gen_helper_vfp_negs(tmp, tmp);
4963 }
4964 gen_helper_vfp_muladds(tmp, tmp, tmp2, tmp3, fpstatus);
4965 tcg_temp_free_i32(tmp3);
4966 tcg_temp_free_ptr(fpstatus);
4967 break;
4968 }
4969 default:
4970 abort();
4971 }
4972 tcg_temp_free_i32(tmp2);
4973
4974 /* Save the result. For elementwise operations we can put it
4975 straight into the destination register. For pairwise operations
4976 we have to be careful to avoid clobbering the source operands. */
4977 if (pairwise && rd == rm) {
4978 neon_store_scratch(pass, tmp);
4979 } else {
4980 neon_store_reg(rd, pass, tmp);
4981 }
4982
4983 } /* for pass */
4984 if (pairwise && rd == rm) {
4985 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4986 tmp = neon_load_scratch(pass);
4987 neon_store_reg(rd, pass, tmp);
4988 }
4989 }
4990 /* End of 3 register same size operations. */
4991 } else if (insn & (1 << 4)) {
4992 if ((insn & 0x00380080) != 0) {
4993 /* Two registers and shift. */
4994 op = (insn >> 8) & 0xf;
4995 if (insn & (1 << 7)) {
4996 /* 64-bit shift. */
4997 if (op > 7) {
4998 return 1;
4999 }
5000 size = 3;
5001 } else {
5002 size = 2;
5003 while ((insn & (1 << (size + 19))) == 0)
5004 size--;
5005 }
5006 shift = (insn >> 16) & ((1 << (3 + size)) - 1);
5007 /* To avoid excessive duplication of ops we implement shift
5008 by immediate using the variable shift operations. */
5009 if (op < 8) {
5010 /* Shift by immediate:
5011 VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
5012 if (q && ((rd | rm) & 1)) {
5013 return 1;
5014 }
5015 if (!u && (op == 4 || op == 6)) {
5016 return 1;
5017 }
5018 /* Right shifts are encoded as N - shift, where N is the
5019 element size in bits. */
5020 if (op <= 4)
5021 shift = shift - (1 << (size + 3));
5022 if (size == 3) {
5023 count = q + 1;
5024 } else {
5025 count = q ? 4: 2;
5026 }
5027 switch (size) {
5028 case 0:
5029 imm = (uint8_t) shift;
5030 imm |= imm << 8;
5031 imm |= imm << 16;
5032 break;
5033 case 1:
5034 imm = (uint16_t) shift;
5035 imm |= imm << 16;
5036 break;
5037 case 2:
5038 case 3:
5039 imm = shift;
5040 break;
5041 default:
5042 abort();
5043 }
5044
5045 for (pass = 0; pass < count; pass++) {
5046 if (size == 3) {
5047 neon_load_reg64(cpu_V0, rm + pass);
5048 tcg_gen_movi_i64(cpu_V1, imm);
5049 switch (op) {
5050 case 0: /* VSHR */
5051 case 1: /* VSRA */
5052 if (u)
5053 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
5054 else
5055 gen_helper_neon_shl_s64(cpu_V0, cpu_V0, cpu_V1);
5056 break;
5057 case 2: /* VRSHR */
5058 case 3: /* VRSRA */
5059 if (u)
5060 gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, cpu_V1);
5061 else
5062 gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, cpu_V1);
5063 break;
5064 case 4: /* VSRI */
5065 case 5: /* VSHL, VSLI */
5066 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
5067 break;
5068 case 6: /* VQSHLU */
5069 gen_helper_neon_qshlu_s64(cpu_V0, cpu_env,
5070 cpu_V0, cpu_V1);
5071 break;
5072 case 7: /* VQSHL */
5073 if (u) {
5074 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
5075 cpu_V0, cpu_V1);
5076 } else {
5077 gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
5078 cpu_V0, cpu_V1);
5079 }
5080 break;
5081 }
5082 if (op == 1 || op == 3) {
5083 /* Accumulate. */
5084 neon_load_reg64(cpu_V1, rd + pass);
5085 tcg_gen_add_i64(cpu_V0, cpu_V0, cpu_V1);
5086 } else if (op == 4 || (op == 5 && u)) {
5087 /* Insert */
5088 neon_load_reg64(cpu_V1, rd + pass);
5089 uint64_t mask;
5090 if (shift < -63 || shift > 63) {
5091 mask = 0;
5092 } else {
5093 if (op == 4) {
5094 mask = 0xffffffffffffffffull >> -shift;
5095 } else {
5096 mask = 0xffffffffffffffffull << shift;
5097 }
5098 }
5099 tcg_gen_andi_i64(cpu_V1, cpu_V1, ~mask);
5100 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5101 }
5102 neon_store_reg64(cpu_V0, rd + pass);
5103 } else { /* size < 3 */
5104 /* Operands in T0 and T1. */
5105 tmp = neon_load_reg(rm, pass);
5106 tmp2 = tcg_temp_new_i32();
5107 tcg_gen_movi_i32(tmp2, imm);
5108 switch (op) {
5109 case 0: /* VSHR */
5110 case 1: /* VSRA */
5111 GEN_NEON_INTEGER_OP(shl);
5112 break;
5113 case 2: /* VRSHR */
5114 case 3: /* VRSRA */
5115 GEN_NEON_INTEGER_OP(rshl);
5116 break;
5117 case 4: /* VSRI */
5118 case 5: /* VSHL, VSLI */
5119 switch (size) {
5120 case 0: gen_helper_neon_shl_u8(tmp, tmp, tmp2); break;
5121 case 1: gen_helper_neon_shl_u16(tmp, tmp, tmp2); break;
5122 case 2: gen_helper_neon_shl_u32(tmp, tmp, tmp2); break;
5123 default: abort();
5124 }
5125 break;
5126 case 6: /* VQSHLU */
5127 switch (size) {
5128 case 0:
5129 gen_helper_neon_qshlu_s8(tmp, cpu_env,
5130 tmp, tmp2);
5131 break;
5132 case 1:
5133 gen_helper_neon_qshlu_s16(tmp, cpu_env,
5134 tmp, tmp2);
5135 break;
5136 case 2:
5137 gen_helper_neon_qshlu_s32(tmp, cpu_env,
5138 tmp, tmp2);
5139 break;
5140 default:
5141 abort();
5142 }
5143 break;
5144 case 7: /* VQSHL */
5145 GEN_NEON_INTEGER_OP_ENV(qshl);
5146 break;
5147 }
5148 tcg_temp_free_i32(tmp2);
5149
5150 if (op == 1 || op == 3) {
5151 /* Accumulate. */
5152 tmp2 = neon_load_reg(rd, pass);
5153 gen_neon_add(size, tmp, tmp2);
5154 tcg_temp_free_i32(tmp2);
5155 } else if (op == 4 || (op == 5 && u)) {
5156 /* Insert */
5157 switch (size) {
5158 case 0:
5159 if (op == 4)
5160 mask = 0xff >> -shift;
5161 else
5162 mask = (uint8_t)(0xff << shift);
5163 mask |= mask << 8;
5164 mask |= mask << 16;
5165 break;
5166 case 1:
5167 if (op == 4)
5168 mask = 0xffff >> -shift;
5169 else
5170 mask = (uint16_t)(0xffff << shift);
5171 mask |= mask << 16;
5172 break;
5173 case 2:
5174 if (shift < -31 || shift > 31) {
5175 mask = 0;
5176 } else {
5177 if (op == 4)
5178 mask = 0xffffffffu >> -shift;
5179 else
5180 mask = 0xffffffffu << shift;
5181 }
5182 break;
5183 default:
5184 abort();
5185 }
5186 tmp2 = neon_load_reg(rd, pass);
5187 tcg_gen_andi_i32(tmp, tmp, mask);
5188 tcg_gen_andi_i32(tmp2, tmp2, ~mask);
5189 tcg_gen_or_i32(tmp, tmp, tmp2);
5190 tcg_temp_free_i32(tmp2);
5191 }
5192 neon_store_reg(rd, pass, tmp);
5193 }
5194 } /* for pass */
5195 } else if (op < 10) {
5196 /* Shift by immediate and narrow:
5197 VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
5198 int input_unsigned = (op == 8) ? !u : u;
5199 if (rm & 1) {
5200 return 1;
5201 }
5202 shift = shift - (1 << (size + 3));
5203 size++;
5204 if (size == 3) {
5205 tmp64 = tcg_const_i64(shift);
5206 neon_load_reg64(cpu_V0, rm);
5207 neon_load_reg64(cpu_V1, rm + 1);
5208 for (pass = 0; pass < 2; pass++) {
5209 TCGv_i64 in;
5210 if (pass == 0) {
5211 in = cpu_V0;
5212 } else {
5213 in = cpu_V1;
5214 }
5215 if (q) {
5216 if (input_unsigned) {
5217 gen_helper_neon_rshl_u64(cpu_V0, in, tmp64);
5218 } else {
5219 gen_helper_neon_rshl_s64(cpu_V0, in, tmp64);
5220 }
5221 } else {
5222 if (input_unsigned) {
5223 gen_helper_neon_shl_u64(cpu_V0, in, tmp64);
5224 } else {
5225 gen_helper_neon_shl_s64(cpu_V0, in, tmp64);
5226 }
5227 }
5228 tmp = tcg_temp_new_i32();
5229 gen_neon_narrow_op(op == 8, u, size - 1, tmp, cpu_V0);
5230 neon_store_reg(rd, pass, tmp);
5231 } /* for pass */
5232 tcg_temp_free_i64(tmp64);
5233 } else {
5234 if (size == 1) {
5235 imm = (uint16_t)shift;
5236 imm |= imm << 16;
5237 } else {
5238 /* size == 2 */
5239 imm = (uint32_t)shift;
5240 }
5241 tmp2 = tcg_const_i32(imm);
5242 tmp4 = neon_load_reg(rm + 1, 0);
5243 tmp5 = neon_load_reg(rm + 1, 1);
5244 for (pass = 0; pass < 2; pass++) {
5245 if (pass == 0) {
5246 tmp = neon_load_reg(rm, 0);
5247 } else {
5248 tmp = tmp4;
5249 }
5250 gen_neon_shift_narrow(size, tmp, tmp2, q,
5251 input_unsigned);
5252 if (pass == 0) {
5253 tmp3 = neon_load_reg(rm, 1);
5254 } else {
5255 tmp3 = tmp5;
5256 }
5257 gen_neon_shift_narrow(size, tmp3, tmp2, q,
5258 input_unsigned);
5259 tcg_gen_concat_i32_i64(cpu_V0, tmp, tmp3);
5260 tcg_temp_free_i32(tmp);
5261 tcg_temp_free_i32(tmp3);
5262 tmp = tcg_temp_new_i32();
5263 gen_neon_narrow_op(op == 8, u, size - 1, tmp, cpu_V0);
5264 neon_store_reg(rd, pass, tmp);
5265 } /* for pass */
5266 tcg_temp_free_i32(tmp2);
5267 }
5268 } else if (op == 10) {
5269 /* VSHLL, VMOVL */
5270 if (q || (rd & 1)) {
5271 return 1;
5272 }
5273 tmp = neon_load_reg(rm, 0);
5274 tmp2 = neon_load_reg(rm, 1);
5275 for (pass = 0; pass < 2; pass++) {
5276 if (pass == 1)
5277 tmp = tmp2;
5278
5279 gen_neon_widen(cpu_V0, tmp, size, u);
5280
5281 if (shift != 0) {
5282 /* The shift is less than the width of the source
5283 type, so we can just shift the whole register. */
5284 tcg_gen_shli_i64(cpu_V0, cpu_V0, shift);
5285 /* Widen the result of shift: we need to clear
5286 * the potential overflow bits resulting from
5287 * left bits of the narrow input appearing as
5288 * right bits of left the neighbour narrow
5289 * input. */
5290 if (size < 2 || !u) {
5291 uint64_t imm64;
5292 if (size == 0) {
5293 imm = (0xffu >> (8 - shift));
5294 imm |= imm << 16;
5295 } else if (size == 1) {
5296 imm = 0xffff >> (16 - shift);
5297 } else {
5298 /* size == 2 */
5299 imm = 0xffffffff >> (32 - shift);
5300 }
5301 if (size < 2) {
5302 imm64 = imm | (((uint64_t)imm) << 32);
5303 } else {
5304 imm64 = imm;
5305 }
5306 tcg_gen_andi_i64(cpu_V0, cpu_V0, ~imm64);
5307 }
5308 }
5309 neon_store_reg64(cpu_V0, rd + pass);
5310 }
5311 } else if (op >= 14) {
5312 /* VCVT fixed-point. */
5313 if (!(insn & (1 << 21)) || (q && ((rd | rm) & 1))) {
5314 return 1;
5315 }
5316 /* We have already masked out the must-be-1 top bit of imm6,
5317 * hence this 32-shift where the ARM ARM has 64-imm6.
5318 */
5319 shift = 32 - shift;
5320 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5321 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, pass));
5322 if (!(op & 1)) {
5323 if (u)
5324 gen_vfp_ulto(0, shift, 1);
5325 else
5326 gen_vfp_slto(0, shift, 1);
5327 } else {
5328 if (u)
5329 gen_vfp_toul(0, shift, 1);
5330 else
5331 gen_vfp_tosl(0, shift, 1);
5332 }
5333 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, pass));
5334 }
5335 } else {
5336 return 1;
5337 }
5338 } else { /* (insn & 0x00380080) == 0 */
5339 int invert;
5340 if (q && (rd & 1)) {
5341 return 1;
5342 }
5343
5344 op = (insn >> 8) & 0xf;
5345 /* One register and immediate. */
5346 imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf);
5347 invert = (insn & (1 << 5)) != 0;
5348 /* Note that op = 2,3,4,5,6,7,10,11,12,13 imm=0 is UNPREDICTABLE.
5349 * We choose to not special-case this and will behave as if a
5350 * valid constant encoding of 0 had been given.
5351 */
5352 switch (op) {
5353 case 0: case 1:
5354 /* no-op */
5355 break;
5356 case 2: case 3:
5357 imm <<= 8;
5358 break;
5359 case 4: case 5:
5360 imm <<= 16;
5361 break;
5362 case 6: case 7:
5363 imm <<= 24;
5364 break;
5365 case 8: case 9:
5366 imm |= imm << 16;
5367 break;
5368 case 10: case 11:
5369 imm = (imm << 8) | (imm << 24);
5370 break;
5371 case 12:
5372 imm = (imm << 8) | 0xff;
5373 break;
5374 case 13:
5375 imm = (imm << 16) | 0xffff;
5376 break;
5377 case 14:
5378 imm |= (imm << 8) | (imm << 16) | (imm << 24);
5379 if (invert)
5380 imm = ~imm;
5381 break;
5382 case 15:
5383 if (invert) {
5384 return 1;
5385 }
5386 imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19)
5387 | ((imm & 0x40) ? (0x1f << 25) : (1 << 30));
5388 break;
5389 }
5390 if (invert)
5391 imm = ~imm;
5392
5393 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5394 if (op & 1 && op < 12) {
5395 tmp = neon_load_reg(rd, pass);
5396 if (invert) {
5397 /* The immediate value has already been inverted, so
5398 BIC becomes AND. */
5399 tcg_gen_andi_i32(tmp, tmp, imm);
5400 } else {
5401 tcg_gen_ori_i32(tmp, tmp, imm);
5402 }
5403 } else {
5404 /* VMOV, VMVN. */
5405 tmp = tcg_temp_new_i32();
5406 if (op == 14 && invert) {
5407 int n;
5408 uint32_t val;
5409 val = 0;
5410 for (n = 0; n < 4; n++) {
5411 if (imm & (1 << (n + (pass & 1) * 4)))
5412 val |= 0xff << (n * 8);
5413 }
5414 tcg_gen_movi_i32(tmp, val);
5415 } else {
5416 tcg_gen_movi_i32(tmp, imm);
5417 }
5418 }
5419 neon_store_reg(rd, pass, tmp);
5420 }
5421 }
5422 } else { /* (insn & 0x00800010 == 0x00800000) */
5423 if (size != 3) {
5424 op = (insn >> 8) & 0xf;
5425 if ((insn & (1 << 6)) == 0) {
5426 /* Three registers of different lengths. */
5427 int src1_wide;
5428 int src2_wide;
5429 int prewiden;
5430 /* undefreq: bit 0 : UNDEF if size != 0
5431 * bit 1 : UNDEF if size == 0
5432 * bit 2 : UNDEF if U == 1
5433 * Note that [1:0] set implies 'always UNDEF'
5434 */
5435 int undefreq;
5436 /* prewiden, src1_wide, src2_wide, undefreq */
5437 static const int neon_3reg_wide[16][4] = {
5438 {1, 0, 0, 0}, /* VADDL */
5439 {1, 1, 0, 0}, /* VADDW */
5440 {1, 0, 0, 0}, /* VSUBL */
5441 {1, 1, 0, 0}, /* VSUBW */
5442 {0, 1, 1, 0}, /* VADDHN */
5443 {0, 0, 0, 0}, /* VABAL */
5444 {0, 1, 1, 0}, /* VSUBHN */
5445 {0, 0, 0, 0}, /* VABDL */
5446 {0, 0, 0, 0}, /* VMLAL */
5447 {0, 0, 0, 6}, /* VQDMLAL */
5448 {0, 0, 0, 0}, /* VMLSL */
5449 {0, 0, 0, 6}, /* VQDMLSL */
5450 {0, 0, 0, 0}, /* Integer VMULL */
5451 {0, 0, 0, 2}, /* VQDMULL */
5452 {0, 0, 0, 5}, /* Polynomial VMULL */
5453 {0, 0, 0, 3}, /* Reserved: always UNDEF */
5454 };
5455
5456 prewiden = neon_3reg_wide[op][0];
5457 src1_wide = neon_3reg_wide[op][1];
5458 src2_wide = neon_3reg_wide[op][2];
5459 undefreq = neon_3reg_wide[op][3];
5460
5461 if (((undefreq & 1) && (size != 0)) ||
5462 ((undefreq & 2) && (size == 0)) ||
5463 ((undefreq & 4) && u)) {
5464 return 1;
5465 }
5466 if ((src1_wide && (rn & 1)) ||
5467 (src2_wide && (rm & 1)) ||
5468 (!src2_wide && (rd & 1))) {
5469 return 1;
5470 }
5471
5472 /* Avoid overlapping operands. Wide source operands are
5473 always aligned so will never overlap with wide
5474 destinations in problematic ways. */
5475 if (rd == rm && !src2_wide) {
5476 tmp = neon_load_reg(rm, 1);
5477 neon_store_scratch(2, tmp);
5478 } else if (rd == rn && !src1_wide) {
5479 tmp = neon_load_reg(rn, 1);
5480 neon_store_scratch(2, tmp);
5481 }
5482 TCGV_UNUSED_I32(tmp3);
5483 for (pass = 0; pass < 2; pass++) {
5484 if (src1_wide) {
5485 neon_load_reg64(cpu_V0, rn + pass);
5486 TCGV_UNUSED_I32(tmp);
5487 } else {
5488 if (pass == 1 && rd == rn) {
5489 tmp = neon_load_scratch(2);
5490 } else {
5491 tmp = neon_load_reg(rn, pass);
5492 }
5493 if (prewiden) {
5494 gen_neon_widen(cpu_V0, tmp, size, u);
5495 }
5496 }
5497 if (src2_wide) {
5498 neon_load_reg64(cpu_V1, rm + pass);
5499 TCGV_UNUSED_I32(tmp2);
5500 } else {
5501 if (pass == 1 && rd == rm) {
5502 tmp2 = neon_load_scratch(2);
5503 } else {
5504 tmp2 = neon_load_reg(rm, pass);
5505 }
5506 if (prewiden) {
5507 gen_neon_widen(cpu_V1, tmp2, size, u);
5508 }
5509 }
5510 switch (op) {
5511 case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
5512 gen_neon_addl(size);
5513 break;
5514 case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */
5515 gen_neon_subl(size);
5516 break;
5517 case 5: case 7: /* VABAL, VABDL */
5518 switch ((size << 1) | u) {
5519 case 0:
5520 gen_helper_neon_abdl_s16(cpu_V0, tmp, tmp2);
5521 break;
5522 case 1:
5523 gen_helper_neon_abdl_u16(cpu_V0, tmp, tmp2);
5524 break;
5525 case 2:
5526 gen_helper_neon_abdl_s32(cpu_V0, tmp, tmp2);
5527 break;
5528 case 3:
5529 gen_helper_neon_abdl_u32(cpu_V0, tmp, tmp2);
5530 break;
5531 case 4:
5532 gen_helper_neon_abdl_s64(cpu_V0, tmp, tmp2);
5533 break;
5534 case 5:
5535 gen_helper_neon_abdl_u64(cpu_V0, tmp, tmp2);
5536 break;
5537 default: abort();
5538 }
5539 tcg_temp_free_i32(tmp2);
5540 tcg_temp_free_i32(tmp);
5541 break;
5542 case 8: case 9: case 10: case 11: case 12: case 13:
5543 /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
5544 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
5545 break;
5546 case 14: /* Polynomial VMULL */
5547 gen_helper_neon_mull_p8(cpu_V0, tmp, tmp2);
5548 tcg_temp_free_i32(tmp2);
5549 tcg_temp_free_i32(tmp);
5550 break;
5551 default: /* 15 is RESERVED: caught earlier */
5552 abort();
5553 }
5554 if (op == 13) {
5555 /* VQDMULL */
5556 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5557 neon_store_reg64(cpu_V0, rd + pass);
5558 } else if (op == 5 || (op >= 8 && op <= 11)) {
5559 /* Accumulate. */
5560 neon_load_reg64(cpu_V1, rd + pass);
5561 switch (op) {
5562 case 10: /* VMLSL */
5563 gen_neon_negl(cpu_V0, size);
5564 /* Fall through */
5565 case 5: case 8: /* VABAL, VMLAL */
5566 gen_neon_addl(size);
5567 break;
5568 case 9: case 11: /* VQDMLAL, VQDMLSL */
5569 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5570 if (op == 11) {
5571 gen_neon_negl(cpu_V0, size);
5572 }
5573 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
5574 break;
5575 default:
5576 abort();
5577 }
5578 neon_store_reg64(cpu_V0, rd + pass);
5579 } else if (op == 4 || op == 6) {
5580 /* Narrowing operation. */
5581 tmp = tcg_temp_new_i32();
5582 if (!u) {
5583 switch (size) {
5584 case 0:
5585 gen_helper_neon_narrow_high_u8(tmp, cpu_V0);
5586 break;
5587 case 1:
5588 gen_helper_neon_narrow_high_u16(tmp, cpu_V0);
5589 break;
5590 case 2:
5591 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
5592 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
5593 break;
5594 default: abort();
5595 }
5596 } else {
5597 switch (size) {
5598 case 0:
5599 gen_helper_neon_narrow_round_high_u8(tmp, cpu_V0);
5600 break;
5601 case 1:
5602 gen_helper_neon_narrow_round_high_u16(tmp, cpu_V0);
5603 break;
5604 case 2:
5605 tcg_gen_addi_i64(cpu_V0, cpu_V0, 1u << 31);
5606 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
5607 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
5608 break;
5609 default: abort();
5610 }
5611 }
5612 if (pass == 0) {
5613 tmp3 = tmp;
5614 } else {
5615 neon_store_reg(rd, 0, tmp3);
5616 neon_store_reg(rd, 1, tmp);
5617 }
5618 } else {
5619 /* Write back the result. */
5620 neon_store_reg64(cpu_V0, rd + pass);
5621 }
5622 }
5623 } else {
5624 /* Two registers and a scalar. NB that for ops of this form
5625 * the ARM ARM labels bit 24 as Q, but it is in our variable
5626 * 'u', not 'q'.
5627 */
5628 if (size == 0) {
5629 return 1;
5630 }
5631 switch (op) {
5632 case 1: /* Float VMLA scalar */
5633 case 5: /* Floating point VMLS scalar */
5634 case 9: /* Floating point VMUL scalar */
5635 if (size == 1) {
5636 return 1;
5637 }
5638 /* fall through */
5639 case 0: /* Integer VMLA scalar */
5640 case 4: /* Integer VMLS scalar */
5641 case 8: /* Integer VMUL scalar */
5642 case 12: /* VQDMULH scalar */
5643 case 13: /* VQRDMULH scalar */
5644 if (u && ((rd | rn) & 1)) {
5645 return 1;
5646 }
5647 tmp = neon_get_scalar(size, rm);
5648 neon_store_scratch(0, tmp);
5649 for (pass = 0; pass < (u ? 4 : 2); pass++) {
5650 tmp = neon_load_scratch(0);
5651 tmp2 = neon_load_reg(rn, pass);
5652 if (op == 12) {
5653 if (size == 1) {
5654 gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
5655 } else {
5656 gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
5657 }
5658 } else if (op == 13) {
5659 if (size == 1) {
5660 gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
5661 } else {
5662 gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
5663 }
5664 } else if (op & 1) {
5665 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5666 gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus);
5667 tcg_temp_free_ptr(fpstatus);
5668 } else {
5669 switch (size) {
5670 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
5671 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
5672 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
5673 default: abort();
5674 }
5675 }
5676 tcg_temp_free_i32(tmp2);
5677 if (op < 8) {
5678 /* Accumulate. */
5679 tmp2 = neon_load_reg(rd, pass);
5680 switch (op) {
5681 case 0:
5682 gen_neon_add(size, tmp, tmp2);
5683 break;
5684 case 1:
5685 {
5686 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5687 gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
5688 tcg_temp_free_ptr(fpstatus);
5689 break;
5690 }
5691 case 4:
5692 gen_neon_rsb(size, tmp, tmp2);
5693 break;
5694 case 5:
5695 {
5696 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5697 gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus);
5698 tcg_temp_free_ptr(fpstatus);
5699 break;
5700 }
5701 default:
5702 abort();
5703 }
5704 tcg_temp_free_i32(tmp2);
5705 }
5706 neon_store_reg(rd, pass, tmp);
5707 }
5708 break;
5709 case 3: /* VQDMLAL scalar */
5710 case 7: /* VQDMLSL scalar */
5711 case 11: /* VQDMULL scalar */
5712 if (u == 1) {
5713 return 1;
5714 }
5715 /* fall through */
5716 case 2: /* VMLAL sclar */
5717 case 6: /* VMLSL scalar */
5718 case 10: /* VMULL scalar */
5719 if (rd & 1) {
5720 return 1;
5721 }
5722 tmp2 = neon_get_scalar(size, rm);
5723 /* We need a copy of tmp2 because gen_neon_mull
5724 * deletes it during pass 0. */
5725 tmp4 = tcg_temp_new_i32();
5726 tcg_gen_mov_i32(tmp4, tmp2);
5727 tmp3 = neon_load_reg(rn, 1);
5728
5729 for (pass = 0; pass < 2; pass++) {
5730 if (pass == 0) {
5731 tmp = neon_load_reg(rn, 0);
5732 } else {
5733 tmp = tmp3;
5734 tmp2 = tmp4;
5735 }
5736 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
5737 if (op != 11) {
5738 neon_load_reg64(cpu_V1, rd + pass);
5739 }
5740 switch (op) {
5741 case 6:
5742 gen_neon_negl(cpu_V0, size);
5743 /* Fall through */
5744 case 2:
5745 gen_neon_addl(size);
5746 break;
5747 case 3: case 7:
5748 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5749 if (op == 7) {
5750 gen_neon_negl(cpu_V0, size);
5751 }
5752 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
5753 break;
5754 case 10:
5755 /* no-op */
5756 break;
5757 case 11:
5758 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5759 break;
5760 default:
5761 abort();
5762 }
5763 neon_store_reg64(cpu_V0, rd + pass);
5764 }
5765
5766
5767 break;
5768 default: /* 14 and 15 are RESERVED */
5769 return 1;
5770 }
5771 }
5772 } else { /* size == 3 */
5773 if (!u) {
5774 /* Extract. */
5775 imm = (insn >> 8) & 0xf;
5776
5777 if (imm > 7 && !q)
5778 return 1;
5779
5780 if (q && ((rd | rn | rm) & 1)) {
5781 return 1;
5782 }
5783
5784 if (imm == 0) {
5785 neon_load_reg64(cpu_V0, rn);
5786 if (q) {
5787 neon_load_reg64(cpu_V1, rn + 1);
5788 }
5789 } else if (imm == 8) {
5790 neon_load_reg64(cpu_V0, rn + 1);
5791 if (q) {
5792 neon_load_reg64(cpu_V1, rm);
5793 }
5794 } else if (q) {
5795 tmp64 = tcg_temp_new_i64();
5796 if (imm < 8) {
5797 neon_load_reg64(cpu_V0, rn);
5798 neon_load_reg64(tmp64, rn + 1);
5799 } else {
5800 neon_load_reg64(cpu_V0, rn + 1);
5801 neon_load_reg64(tmp64, rm);
5802 }
5803 tcg_gen_shri_i64(cpu_V0, cpu_V0, (imm & 7) * 8);
5804 tcg_gen_shli_i64(cpu_V1, tmp64, 64 - ((imm & 7) * 8));
5805 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5806 if (imm < 8) {
5807 neon_load_reg64(cpu_V1, rm);
5808 } else {
5809 neon_load_reg64(cpu_V1, rm + 1);
5810 imm -= 8;
5811 }
5812 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
5813 tcg_gen_shri_i64(tmp64, tmp64, imm * 8);
5814 tcg_gen_or_i64(cpu_V1, cpu_V1, tmp64);
5815 tcg_temp_free_i64(tmp64);
5816 } else {
5817 /* BUGFIX */
5818 neon_load_reg64(cpu_V0, rn);
5819 tcg_gen_shri_i64(cpu_V0, cpu_V0, imm * 8);
5820 neon_load_reg64(cpu_V1, rm);
5821 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
5822 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5823 }
5824 neon_store_reg64(cpu_V0, rd);
5825 if (q) {
5826 neon_store_reg64(cpu_V1, rd + 1);
5827 }
5828 } else if ((insn & (1 << 11)) == 0) {
5829 /* Two register misc. */
5830 op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf);
5831 size = (insn >> 18) & 3;
5832 /* UNDEF for unknown op values and bad op-size combinations */
5833 if ((neon_2rm_sizes[op] & (1 << size)) == 0) {
5834 return 1;
5835 }
5836 if ((op != NEON_2RM_VMOVN && op != NEON_2RM_VQMOVN) &&
5837 q && ((rm | rd) & 1)) {
5838 return 1;
5839 }
5840 switch (op) {
5841 case NEON_2RM_VREV64:
5842 for (pass = 0; pass < (q ? 2 : 1); pass++) {
5843 tmp = neon_load_reg(rm, pass * 2);
5844 tmp2 = neon_load_reg(rm, pass * 2 + 1);
5845 switch (size) {
5846 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
5847 case 1: gen_swap_half(tmp); break;
5848 case 2: /* no-op */ break;
5849 default: abort();
5850 }
5851 neon_store_reg(rd, pass * 2 + 1, tmp);
5852 if (size == 2) {
5853 neon_store_reg(rd, pass * 2, tmp2);
5854 } else {
5855 switch (size) {
5856 case 0: tcg_gen_bswap32_i32(tmp2, tmp2); break;
5857 case 1: gen_swap_half(tmp2); break;
5858 default: abort();
5859 }
5860 neon_store_reg(rd, pass * 2, tmp2);
5861 }
5862 }
5863 break;
5864 case NEON_2RM_VPADDL: case NEON_2RM_VPADDL_U:
5865 case NEON_2RM_VPADAL: case NEON_2RM_VPADAL_U:
5866 for (pass = 0; pass < q + 1; pass++) {
5867 tmp = neon_load_reg(rm, pass * 2);
5868 gen_neon_widen(cpu_V0, tmp, size, op & 1);
5869 tmp = neon_load_reg(rm, pass * 2 + 1);
5870 gen_neon_widen(cpu_V1, tmp, size, op & 1);
5871 switch (size) {
5872 case 0: gen_helper_neon_paddl_u16(CPU_V001); break;
5873 case 1: gen_helper_neon_paddl_u32(CPU_V001); break;
5874 case 2: tcg_gen_add_i64(CPU_V001); break;
5875 default: abort();
5876 }
5877 if (op >= NEON_2RM_VPADAL) {
5878 /* Accumulate. */
5879 neon_load_reg64(cpu_V1, rd + pass);
5880 gen_neon_addl(size);
5881 }
5882 neon_store_reg64(cpu_V0, rd + pass);
5883 }
5884 break;
5885 case NEON_2RM_VTRN:
5886 if (size == 2) {
5887 int n;
5888 for (n = 0; n < (q ? 4 : 2); n += 2) {
5889 tmp = neon_load_reg(rm, n);
5890 tmp2 = neon_load_reg(rd, n + 1);
5891 neon_store_reg(rm, n, tmp2);
5892 neon_store_reg(rd, n + 1, tmp);
5893 }
5894 } else {
5895 goto elementwise;
5896 }
5897 break;
5898 case NEON_2RM_VUZP:
5899 if (gen_neon_unzip(rd, rm, size, q)) {
5900 return 1;
5901 }
5902 break;
5903 case NEON_2RM_VZIP:
5904 if (gen_neon_zip(rd, rm, size, q)) {
5905 return 1;
5906 }
5907 break;
5908 case NEON_2RM_VMOVN: case NEON_2RM_VQMOVN:
5909 /* also VQMOVUN; op field and mnemonics don't line up */
5910 if (rm & 1) {
5911 return 1;
5912 }
5913 TCGV_UNUSED_I32(tmp2);
5914 for (pass = 0; pass < 2; pass++) {
5915 neon_load_reg64(cpu_V0, rm + pass);
5916 tmp = tcg_temp_new_i32();
5917 gen_neon_narrow_op(op == NEON_2RM_VMOVN, q, size,
5918 tmp, cpu_V0);
5919 if (pass == 0) {
5920 tmp2 = tmp;
5921 } else {
5922 neon_store_reg(rd, 0, tmp2);
5923 neon_store_reg(rd, 1, tmp);
5924 }
5925 }
5926 break;
5927 case NEON_2RM_VSHLL:
5928 if (q || (rd & 1)) {
5929 return 1;
5930 }
5931 tmp = neon_load_reg(rm, 0);
5932 tmp2 = neon_load_reg(rm, 1);
5933 for (pass = 0; pass < 2; pass++) {
5934 if (pass == 1)
5935 tmp = tmp2;
5936 gen_neon_widen(cpu_V0, tmp, size, 1);
5937 tcg_gen_shli_i64(cpu_V0, cpu_V0, 8 << size);
5938 neon_store_reg64(cpu_V0, rd + pass);
5939 }
5940 break;
5941 case NEON_2RM_VCVT_F16_F32:
5942 if (!arm_feature(env, ARM_FEATURE_VFP_FP16) ||
5943 q || (rm & 1)) {
5944 return 1;
5945 }
5946 tmp = tcg_temp_new_i32();
5947 tmp2 = tcg_temp_new_i32();
5948 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 0));
5949 gen_helper_neon_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
5950 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 1));
5951 gen_helper_neon_fcvt_f32_to_f16(tmp2, cpu_F0s, cpu_env);
5952 tcg_gen_shli_i32(tmp2, tmp2, 16);
5953 tcg_gen_or_i32(tmp2, tmp2, tmp);
5954 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 2));
5955 gen_helper_neon_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
5956 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 3));
5957 neon_store_reg(rd, 0, tmp2);
5958 tmp2 = tcg_temp_new_i32();
5959 gen_helper_neon_fcvt_f32_to_f16(tmp2, cpu_F0s, cpu_env);
5960 tcg_gen_shli_i32(tmp2, tmp2, 16);
5961 tcg_gen_or_i32(tmp2, tmp2, tmp);
5962 neon_store_reg(rd, 1, tmp2);
5963 tcg_temp_free_i32(tmp);
5964 break;
5965 case NEON_2RM_VCVT_F32_F16:
5966 if (!arm_feature(env, ARM_FEATURE_VFP_FP16) ||
5967 q || (rd & 1)) {
5968 return 1;
5969 }
5970 tmp3 = tcg_temp_new_i32();
5971 tmp = neon_load_reg(rm, 0);
5972 tmp2 = neon_load_reg(rm, 1);
5973 tcg_gen_ext16u_i32(tmp3, tmp);
5974 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5975 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 0));
5976 tcg_gen_shri_i32(tmp3, tmp, 16);
5977 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5978 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 1));
5979 tcg_temp_free_i32(tmp);
5980 tcg_gen_ext16u_i32(tmp3, tmp2);
5981 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5982 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 2));
5983 tcg_gen_shri_i32(tmp3, tmp2, 16);
5984 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5985 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 3));
5986 tcg_temp_free_i32(tmp2);
5987 tcg_temp_free_i32(tmp3);
5988 break;
5989 default:
5990 elementwise:
5991 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5992 if (neon_2rm_is_float_op(op)) {
5993 tcg_gen_ld_f32(cpu_F0s, cpu_env,
5994 neon_reg_offset(rm, pass));
5995 TCGV_UNUSED_I32(tmp);
5996 } else {
5997 tmp = neon_load_reg(rm, pass);
5998 }
5999 switch (op) {
6000 case NEON_2RM_VREV32:
6001 switch (size) {
6002 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
6003 case 1: gen_swap_half(tmp); break;
6004 default: abort();
6005 }
6006 break;
6007 case NEON_2RM_VREV16:
6008 gen_rev16(tmp);
6009 break;
6010 case NEON_2RM_VCLS:
6011 switch (size) {
6012 case 0: gen_helper_neon_cls_s8(tmp, tmp); break;
6013 case 1: gen_helper_neon_cls_s16(tmp, tmp); break;
6014 case 2: gen_helper_neon_cls_s32(tmp, tmp); break;
6015 default: abort();
6016 }
6017 break;
6018 case NEON_2RM_VCLZ:
6019 switch (size) {
6020 case 0: gen_helper_neon_clz_u8(tmp, tmp); break;
6021 case 1: gen_helper_neon_clz_u16(tmp, tmp); break;
6022 case 2: gen_helper_clz(tmp, tmp); break;
6023 default: abort();
6024 }
6025 break;
6026 case NEON_2RM_VCNT:
6027 gen_helper_neon_cnt_u8(tmp, tmp);
6028 break;
6029 case NEON_2RM_VMVN:
6030 tcg_gen_not_i32(tmp, tmp);
6031 break;
6032 case NEON_2RM_VQABS:
6033 switch (size) {
6034 case 0:
6035 gen_helper_neon_qabs_s8(tmp, cpu_env, tmp);
6036 break;
6037 case 1:
6038 gen_helper_neon_qabs_s16(tmp, cpu_env, tmp);
6039 break;
6040 case 2:
6041 gen_helper_neon_qabs_s32(tmp, cpu_env, tmp);
6042 break;
6043 default: abort();
6044 }
6045 break;
6046 case NEON_2RM_VQNEG:
6047 switch (size) {
6048 case 0:
6049 gen_helper_neon_qneg_s8(tmp, cpu_env, tmp);
6050 break;
6051 case 1:
6052 gen_helper_neon_qneg_s16(tmp, cpu_env, tmp);
6053 break;
6054 case 2:
6055 gen_helper_neon_qneg_s32(tmp, cpu_env, tmp);
6056 break;
6057 default: abort();
6058 }
6059 break;
6060 case NEON_2RM_VCGT0: case NEON_2RM_VCLE0:
6061 tmp2 = tcg_const_i32(0);
6062 switch(size) {
6063 case 0: gen_helper_neon_cgt_s8(tmp, tmp, tmp2); break;
6064 case 1: gen_helper_neon_cgt_s16(tmp, tmp, tmp2); break;
6065 case 2: gen_helper_neon_cgt_s32(tmp, tmp, tmp2); break;
6066 default: abort();
6067 }
6068 tcg_temp_free_i32(tmp2);
6069 if (op == NEON_2RM_VCLE0) {
6070 tcg_gen_not_i32(tmp, tmp);
6071 }
6072 break;
6073 case NEON_2RM_VCGE0: case NEON_2RM_VCLT0:
6074 tmp2 = tcg_const_i32(0);
6075 switch(size) {
6076 case 0: gen_helper_neon_cge_s8(tmp, tmp, tmp2); break;
6077 case 1: gen_helper_neon_cge_s16(tmp, tmp, tmp2); break;
6078 case 2: gen_helper_neon_cge_s32(tmp, tmp, tmp2); break;
6079 default: abort();
6080 }
6081 tcg_temp_free_i32(tmp2);
6082 if (op == NEON_2RM_VCLT0) {
6083 tcg_gen_not_i32(tmp, tmp);
6084 }
6085 break;
6086 case NEON_2RM_VCEQ0:
6087 tmp2 = tcg_const_i32(0);
6088 switch(size) {
6089 case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break;
6090 case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break;
6091 case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break;
6092 default: abort();
6093 }
6094 tcg_temp_free_i32(tmp2);
6095 break;
6096 case NEON_2RM_VABS:
6097 switch(size) {
6098 case 0: gen_helper_neon_abs_s8(tmp, tmp); break;
6099 case 1: gen_helper_neon_abs_s16(tmp, tmp); break;
6100 case 2: tcg_gen_abs_i32(tmp, tmp); break;
6101 default: abort();
6102 }
6103 break;
6104 case NEON_2RM_VNEG:
6105 tmp2 = tcg_const_i32(0);
6106 gen_neon_rsb(size, tmp, tmp2);
6107 tcg_temp_free_i32(tmp2);
6108 break;
6109 case NEON_2RM_VCGT0_F:
6110 {
6111 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6112 tmp2 = tcg_const_i32(0);
6113 gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus);
6114 tcg_temp_free_i32(tmp2);
6115 tcg_temp_free_ptr(fpstatus);
6116 break;
6117 }
6118 case NEON_2RM_VCGE0_F:
6119 {
6120 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6121 tmp2 = tcg_const_i32(0);
6122 gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus);
6123 tcg_temp_free_i32(tmp2);
6124 tcg_temp_free_ptr(fpstatus);
6125 break;
6126 }
6127 case NEON_2RM_VCEQ0_F:
6128 {
6129 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6130 tmp2 = tcg_const_i32(0);
6131 gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus);
6132 tcg_temp_free_i32(tmp2);
6133 tcg_temp_free_ptr(fpstatus);
6134 break;
6135 }
6136 case NEON_2RM_VCLE0_F:
6137 {
6138 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6139 tmp2 = tcg_const_i32(0);
6140 gen_helper_neon_cge_f32(tmp, tmp2, tmp, fpstatus);
6141 tcg_temp_free_i32(tmp2);
6142 tcg_temp_free_ptr(fpstatus);
6143 break;
6144 }
6145 case NEON_2RM_VCLT0_F:
6146 {
6147 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6148 tmp2 = tcg_const_i32(0);
6149 gen_helper_neon_cgt_f32(tmp, tmp2, tmp, fpstatus);
6150 tcg_temp_free_i32(tmp2);
6151 tcg_temp_free_ptr(fpstatus);
6152 break;
6153 }
6154 case NEON_2RM_VABS_F:
6155 gen_vfp_abs(0);
6156 break;
6157 case NEON_2RM_VNEG_F:
6158 gen_vfp_neg(0);
6159 break;
6160 case NEON_2RM_VSWP:
6161 tmp2 = neon_load_reg(rd, pass);
6162 neon_store_reg(rm, pass, tmp2);
6163 break;
6164 case NEON_2RM_VTRN:
6165 tmp2 = neon_load_reg(rd, pass);
6166 switch (size) {
6167 case 0: gen_neon_trn_u8(tmp, tmp2); break;
6168 case 1: gen_neon_trn_u16(tmp, tmp2); break;
6169 default: abort();
6170 }
6171 neon_store_reg(rm, pass, tmp2);
6172 break;
6173 case NEON_2RM_VRECPE:
6174 gen_helper_recpe_u32(tmp, tmp, cpu_env);
6175 break;
6176 case NEON_2RM_VRSQRTE:
6177 gen_helper_rsqrte_u32(tmp, tmp, cpu_env);
6178 break;
6179 case NEON_2RM_VRECPE_F:
6180 gen_helper_recpe_f32(cpu_F0s, cpu_F0s, cpu_env);
6181 break;
6182 case NEON_2RM_VRSQRTE_F:
6183 gen_helper_rsqrte_f32(cpu_F0s, cpu_F0s, cpu_env);
6184 break;
6185 case NEON_2RM_VCVT_FS: /* VCVT.F32.S32 */
6186 gen_vfp_sito(0, 1);
6187 break;
6188 case NEON_2RM_VCVT_FU: /* VCVT.F32.U32 */
6189 gen_vfp_uito(0, 1);
6190 break;
6191 case NEON_2RM_VCVT_SF: /* VCVT.S32.F32 */
6192 gen_vfp_tosiz(0, 1);
6193 break;
6194 case NEON_2RM_VCVT_UF: /* VCVT.U32.F32 */
6195 gen_vfp_touiz(0, 1);
6196 break;
6197 default:
6198 /* Reserved op values were caught by the
6199 * neon_2rm_sizes[] check earlier.
6200 */
6201 abort();
6202 }
6203 if (neon_2rm_is_float_op(op)) {
6204 tcg_gen_st_f32(cpu_F0s, cpu_env,
6205 neon_reg_offset(rd, pass));
6206 } else {
6207 neon_store_reg(rd, pass, tmp);
6208 }
6209 }
6210 break;
6211 }
6212 } else if ((insn & (1 << 10)) == 0) {
6213 /* VTBL, VTBX. */
6214 int n = ((insn >> 8) & 3) + 1;
6215 if ((rn + n) > 32) {
6216 /* This is UNPREDICTABLE; we choose to UNDEF to avoid the
6217 * helper function running off the end of the register file.
6218 */
6219 return 1;
6220 }
6221 n <<= 3;
6222 if (insn & (1 << 6)) {
6223 tmp = neon_load_reg(rd, 0);
6224 } else {
6225 tmp = tcg_temp_new_i32();
6226 tcg_gen_movi_i32(tmp, 0);
6227 }
6228 tmp2 = neon_load_reg(rm, 0);
6229 tmp4 = tcg_const_i32(rn);
6230 tmp5 = tcg_const_i32(n);
6231 gen_helper_neon_tbl(tmp2, cpu_env, tmp2, tmp, tmp4, tmp5);
6232 tcg_temp_free_i32(tmp);
6233 if (insn & (1 << 6)) {
6234 tmp = neon_load_reg(rd, 1);
6235 } else {
6236 tmp = tcg_temp_new_i32();
6237 tcg_gen_movi_i32(tmp, 0);
6238 }
6239 tmp3 = neon_load_reg(rm, 1);
6240 gen_helper_neon_tbl(tmp3, cpu_env, tmp3, tmp, tmp4, tmp5);
6241 tcg_temp_free_i32(tmp5);
6242 tcg_temp_free_i32(tmp4);
6243 neon_store_reg(rd, 0, tmp2);
6244 neon_store_reg(rd, 1, tmp3);
6245 tcg_temp_free_i32(tmp);
6246 } else if ((insn & 0x380) == 0) {
6247 /* VDUP */
6248 if ((insn & (7 << 16)) == 0 || (q && (rd & 1))) {
6249 return 1;
6250 }
6251 if (insn & (1 << 19)) {
6252 tmp = neon_load_reg(rm, 1);
6253 } else {
6254 tmp = neon_load_reg(rm, 0);
6255 }
6256 if (insn & (1 << 16)) {
6257 gen_neon_dup_u8(tmp, ((insn >> 17) & 3) * 8);
6258 } else if (insn & (1 << 17)) {
6259 if ((insn >> 18) & 1)
6260 gen_neon_dup_high16(tmp);
6261 else
6262 gen_neon_dup_low16(tmp);
6263 }
6264 for (pass = 0; pass < (q ? 4 : 2); pass++) {
6265 tmp2 = tcg_temp_new_i32();
6266 tcg_gen_mov_i32(tmp2, tmp);
6267 neon_store_reg(rd, pass, tmp2);
6268 }
6269 tcg_temp_free_i32(tmp);
6270 } else {
6271 return 1;
6272 }
6273 }
6274 }
6275 return 0;
6276 }
6277
6278 static int disas_coproc_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
6279 {
6280 int cpnum, is64, crn, crm, opc1, opc2, isread, rt, rt2;
6281 const ARMCPRegInfo *ri;
6282 ARMCPU *cpu = arm_env_get_cpu(env);
6283
6284 cpnum = (insn >> 8) & 0xf;
6285 if (arm_feature(env, ARM_FEATURE_XSCALE)
6286 && ((env->cp15.c15_cpar ^ 0x3fff) & (1 << cpnum)))
6287 return 1;
6288
6289 /* First check for coprocessor space used for actual instructions */
6290 switch (cpnum) {
6291 case 0:
6292 case 1:
6293 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
6294 return disas_iwmmxt_insn(env, s, insn);
6295 } else if (arm_feature(env, ARM_FEATURE_XSCALE)) {
6296 return disas_dsp_insn(env, s, insn);
6297 }
6298 return 1;
6299 case 10:
6300 case 11:
6301 return disas_vfp_insn (env, s, insn);
6302 default:
6303 break;
6304 }
6305
6306 /* Otherwise treat as a generic register access */
6307 is64 = (insn & (1 << 25)) == 0;
6308 if (!is64 && ((insn & (1 << 4)) == 0)) {
6309 /* cdp */
6310 return 1;
6311 }
6312
6313 crm = insn & 0xf;
6314 if (is64) {
6315 crn = 0;
6316 opc1 = (insn >> 4) & 0xf;
6317 opc2 = 0;
6318 rt2 = (insn >> 16) & 0xf;
6319 } else {
6320 crn = (insn >> 16) & 0xf;
6321 opc1 = (insn >> 21) & 7;
6322 opc2 = (insn >> 5) & 7;
6323 rt2 = 0;
6324 }
6325 isread = (insn >> 20) & 1;
6326 rt = (insn >> 12) & 0xf;
6327
6328 ri = get_arm_cp_reginfo(cpu,
6329 ENCODE_CP_REG(cpnum, is64, crn, crm, opc1, opc2));
6330 if (ri) {
6331 /* Check access permissions */
6332 if (!cp_access_ok(env, ri, isread)) {
6333 return 1;
6334 }
6335
6336 /* Handle special cases first */
6337 switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) {
6338 case ARM_CP_NOP:
6339 return 0;
6340 case ARM_CP_WFI:
6341 if (isread) {
6342 return 1;
6343 }
6344 gen_set_pc_im(s, s->pc);
6345 s->is_jmp = DISAS_WFI;
6346 return 0;
6347 default:
6348 break;
6349 }
6350
6351 if (use_icount && (ri->type & ARM_CP_IO)) {
6352 gen_io_start();
6353 }
6354
6355 if (isread) {
6356 /* Read */
6357 if (is64) {
6358 TCGv_i64 tmp64;
6359 TCGv_i32 tmp;
6360 if (ri->type & ARM_CP_CONST) {
6361 tmp64 = tcg_const_i64(ri->resetvalue);
6362 } else if (ri->readfn) {
6363 TCGv_ptr tmpptr;
6364 gen_set_pc_im(s, s->pc);
6365 tmp64 = tcg_temp_new_i64();
6366 tmpptr = tcg_const_ptr(ri);
6367 gen_helper_get_cp_reg64(tmp64, cpu_env, tmpptr);
6368 tcg_temp_free_ptr(tmpptr);
6369 } else {
6370 tmp64 = tcg_temp_new_i64();
6371 tcg_gen_ld_i64(tmp64, cpu_env, ri->fieldoffset);
6372 }
6373 tmp = tcg_temp_new_i32();
6374 tcg_gen_trunc_i64_i32(tmp, tmp64);
6375 store_reg(s, rt, tmp);
6376 tcg_gen_shri_i64(tmp64, tmp64, 32);
6377 tmp = tcg_temp_new_i32();
6378 tcg_gen_trunc_i64_i32(tmp, tmp64);
6379 tcg_temp_free_i64(tmp64);
6380 store_reg(s, rt2, tmp);
6381 } else {
6382 TCGv_i32 tmp;
6383 if (ri->type & ARM_CP_CONST) {
6384 tmp = tcg_const_i32(ri->resetvalue);
6385 } else if (ri->readfn) {
6386 TCGv_ptr tmpptr;
6387 gen_set_pc_im(s, s->pc);
6388 tmp = tcg_temp_new_i32();
6389 tmpptr = tcg_const_ptr(ri);
6390 gen_helper_get_cp_reg(tmp, cpu_env, tmpptr);
6391 tcg_temp_free_ptr(tmpptr);
6392 } else {
6393 tmp = load_cpu_offset(ri->fieldoffset);
6394 }
6395 if (rt == 15) {
6396 /* Destination register of r15 for 32 bit loads sets
6397 * the condition codes from the high 4 bits of the value
6398 */
6399 gen_set_nzcv(tmp);
6400 tcg_temp_free_i32(tmp);
6401 } else {
6402 store_reg(s, rt, tmp);
6403 }
6404 }
6405 } else {
6406 /* Write */
6407 if (ri->type & ARM_CP_CONST) {
6408 /* If not forbidden by access permissions, treat as WI */
6409 return 0;
6410 }
6411
6412 if (is64) {
6413 TCGv_i32 tmplo, tmphi;
6414 TCGv_i64 tmp64 = tcg_temp_new_i64();
6415 tmplo = load_reg(s, rt);
6416 tmphi = load_reg(s, rt2);
6417 tcg_gen_concat_i32_i64(tmp64, tmplo, tmphi);
6418 tcg_temp_free_i32(tmplo);
6419 tcg_temp_free_i32(tmphi);
6420 if (ri->writefn) {
6421 TCGv_ptr tmpptr = tcg_const_ptr(ri);
6422 gen_set_pc_im(s, s->pc);
6423 gen_helper_set_cp_reg64(cpu_env, tmpptr, tmp64);
6424 tcg_temp_free_ptr(tmpptr);
6425 } else {
6426 tcg_gen_st_i64(tmp64, cpu_env, ri->fieldoffset);
6427 }
6428 tcg_temp_free_i64(tmp64);
6429 } else {
6430 if (ri->writefn) {
6431 TCGv_i32 tmp;
6432 TCGv_ptr tmpptr;
6433 gen_set_pc_im(s, s->pc);
6434 tmp = load_reg(s, rt);
6435 tmpptr = tcg_const_ptr(ri);
6436 gen_helper_set_cp_reg(cpu_env, tmpptr, tmp);
6437 tcg_temp_free_ptr(tmpptr);
6438 tcg_temp_free_i32(tmp);
6439 } else {
6440 TCGv_i32 tmp = load_reg(s, rt);
6441 store_cpu_offset(tmp, ri->fieldoffset);
6442 }
6443 }
6444 }
6445
6446 if (use_icount && (ri->type & ARM_CP_IO)) {
6447 /* I/O operations must end the TB here (whether read or write) */
6448 gen_io_end();
6449 gen_lookup_tb(s);
6450 } else if (!isread && !(ri->type & ARM_CP_SUPPRESS_TB_END)) {
6451 /* We default to ending the TB on a coprocessor register write,
6452 * but allow this to be suppressed by the register definition
6453 * (usually only necessary to work around guest bugs).
6454 */
6455 gen_lookup_tb(s);
6456 }
6457
6458 return 0;
6459 }
6460
6461 return 1;
6462 }
6463
6464
6465 /* Store a 64-bit value to a register pair. Clobbers val. */
6466 static void gen_storeq_reg(DisasContext *s, int rlow, int rhigh, TCGv_i64 val)
6467 {
6468 TCGv_i32 tmp;
6469 tmp = tcg_temp_new_i32();
6470 tcg_gen_trunc_i64_i32(tmp, val);
6471 store_reg(s, rlow, tmp);
6472 tmp = tcg_temp_new_i32();
6473 tcg_gen_shri_i64(val, val, 32);
6474 tcg_gen_trunc_i64_i32(tmp, val);
6475 store_reg(s, rhigh, tmp);
6476 }
6477
6478 /* load a 32-bit value from a register and perform a 64-bit accumulate. */
6479 static void gen_addq_lo(DisasContext *s, TCGv_i64 val, int rlow)
6480 {
6481 TCGv_i64 tmp;
6482 TCGv_i32 tmp2;
6483
6484 /* Load value and extend to 64 bits. */
6485 tmp = tcg_temp_new_i64();
6486 tmp2 = load_reg(s, rlow);
6487 tcg_gen_extu_i32_i64(tmp, tmp2);
6488 tcg_temp_free_i32(tmp2);
6489 tcg_gen_add_i64(val, val, tmp);
6490 tcg_temp_free_i64(tmp);
6491 }
6492
6493 /* load and add a 64-bit value from a register pair. */
6494 static void gen_addq(DisasContext *s, TCGv_i64 val, int rlow, int rhigh)
6495 {
6496 TCGv_i64 tmp;
6497 TCGv_i32 tmpl;
6498 TCGv_i32 tmph;
6499
6500 /* Load 64-bit value rd:rn. */
6501 tmpl = load_reg(s, rlow);
6502 tmph = load_reg(s, rhigh);
6503 tmp = tcg_temp_new_i64();
6504 tcg_gen_concat_i32_i64(tmp, tmpl, tmph);
6505 tcg_temp_free_i32(tmpl);
6506 tcg_temp_free_i32(tmph);
6507 tcg_gen_add_i64(val, val, tmp);
6508 tcg_temp_free_i64(tmp);
6509 }
6510
6511 /* Set N and Z flags from hi|lo. */
6512 static void gen_logicq_cc(TCGv_i32 lo, TCGv_i32 hi)
6513 {
6514 tcg_gen_mov_i32(cpu_NF, hi);
6515 tcg_gen_or_i32(cpu_ZF, lo, hi);
6516 }
6517
6518 /* Load/Store exclusive instructions are implemented by remembering
6519 the value/address loaded, and seeing if these are the same
6520 when the store is performed. This should be sufficient to implement
6521 the architecturally mandated semantics, and avoids having to monitor
6522 regular stores.
6523
6524 In system emulation mode only one CPU will be running at once, so
6525 this sequence is effectively atomic. In user emulation mode we
6526 throw an exception and handle the atomic operation elsewhere. */
6527 static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
6528 TCGv_i32 addr, int size)
6529 {
6530 TCGv_i32 tmp = tcg_temp_new_i32();
6531
6532 switch (size) {
6533 case 0:
6534 gen_aa32_ld8u(tmp, addr, IS_USER(s));
6535 break;
6536 case 1:
6537 gen_aa32_ld16u(tmp, addr, IS_USER(s));
6538 break;
6539 case 2:
6540 case 3:
6541 gen_aa32_ld32u(tmp, addr, IS_USER(s));
6542 break;
6543 default:
6544 abort();
6545 }
6546 tcg_gen_mov_i32(cpu_exclusive_val, tmp);
6547 store_reg(s, rt, tmp);
6548 if (size == 3) {
6549 TCGv_i32 tmp2 = tcg_temp_new_i32();
6550 tcg_gen_addi_i32(tmp2, addr, 4);
6551 tmp = tcg_temp_new_i32();
6552 gen_aa32_ld32u(tmp, tmp2, IS_USER(s));
6553 tcg_temp_free_i32(tmp2);
6554 tcg_gen_mov_i32(cpu_exclusive_high, tmp);
6555 store_reg(s, rt2, tmp);
6556 }
6557 tcg_gen_mov_i32(cpu_exclusive_addr, addr);
6558 }
6559
6560 static void gen_clrex(DisasContext *s)
6561 {
6562 tcg_gen_movi_i32(cpu_exclusive_addr, -1);
6563 }
6564
6565 #ifdef CONFIG_USER_ONLY
6566 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
6567 TCGv_i32 addr, int size)
6568 {
6569 tcg_gen_mov_i32(cpu_exclusive_test, addr);
6570 tcg_gen_movi_i32(cpu_exclusive_info,
6571 size | (rd << 4) | (rt << 8) | (rt2 << 12));
6572 gen_exception_insn(s, 4, EXCP_STREX);
6573 }
6574 #else
6575 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
6576 TCGv_i32 addr, int size)
6577 {
6578 TCGv_i32 tmp;
6579 int done_label;
6580 int fail_label;
6581
6582 /* if (env->exclusive_addr == addr && env->exclusive_val == [addr]) {
6583 [addr] = {Rt};
6584 {Rd} = 0;
6585 } else {
6586 {Rd} = 1;
6587 } */
6588 fail_label = gen_new_label();
6589 done_label = gen_new_label();
6590 tcg_gen_brcond_i32(TCG_COND_NE, addr, cpu_exclusive_addr, fail_label);
6591 tmp = tcg_temp_new_i32();
6592 switch (size) {
6593 case 0:
6594 gen_aa32_ld8u(tmp, addr, IS_USER(s));
6595 break;
6596 case 1:
6597 gen_aa32_ld16u(tmp, addr, IS_USER(s));
6598 break;
6599 case 2:
6600 case 3:
6601 gen_aa32_ld32u(tmp, addr, IS_USER(s));
6602 break;
6603 default:
6604 abort();
6605 }
6606 tcg_gen_brcond_i32(TCG_COND_NE, tmp, cpu_exclusive_val, fail_label);
6607 tcg_temp_free_i32(tmp);
6608 if (size == 3) {
6609 TCGv_i32 tmp2 = tcg_temp_new_i32();
6610 tcg_gen_addi_i32(tmp2, addr, 4);
6611 tmp = tcg_temp_new_i32();
6612 gen_aa32_ld32u(tmp, tmp2, IS_USER(s));
6613 tcg_temp_free_i32(tmp2);
6614 tcg_gen_brcond_i32(TCG_COND_NE, tmp, cpu_exclusive_high, fail_label);
6615 tcg_temp_free_i32(tmp);
6616 }
6617 tmp = load_reg(s, rt);
6618 switch (size) {
6619 case 0:
6620 gen_aa32_st8(tmp, addr, IS_USER(s));
6621 break;
6622 case 1:
6623 gen_aa32_st16(tmp, addr, IS_USER(s));
6624 break;
6625 case 2:
6626 case 3:
6627 gen_aa32_st32(tmp, addr, IS_USER(s));
6628 break;
6629 default:
6630 abort();
6631 }
6632 tcg_temp_free_i32(tmp);
6633 if (size == 3) {
6634 tcg_gen_addi_i32(addr, addr, 4);
6635 tmp = load_reg(s, rt2);
6636 gen_aa32_st32(tmp, addr, IS_USER(s));
6637 tcg_temp_free_i32(tmp);
6638 }
6639 tcg_gen_movi_i32(cpu_R[rd], 0);
6640 tcg_gen_br(done_label);
6641 gen_set_label(fail_label);
6642 tcg_gen_movi_i32(cpu_R[rd], 1);
6643 gen_set_label(done_label);
6644 tcg_gen_movi_i32(cpu_exclusive_addr, -1);
6645 }
6646 #endif
6647
6648 /* gen_srs:
6649 * @env: CPUARMState
6650 * @s: DisasContext
6651 * @mode: mode field from insn (which stack to store to)
6652 * @amode: addressing mode (DA/IA/DB/IB), encoded as per P,U bits in ARM insn
6653 * @writeback: true if writeback bit set
6654 *
6655 * Generate code for the SRS (Store Return State) insn.
6656 */
6657 static void gen_srs(DisasContext *s,
6658 uint32_t mode, uint32_t amode, bool writeback)
6659 {
6660 int32_t offset;
6661 TCGv_i32 addr = tcg_temp_new_i32();
6662 TCGv_i32 tmp = tcg_const_i32(mode);
6663 gen_helper_get_r13_banked(addr, cpu_env, tmp);
6664 tcg_temp_free_i32(tmp);
6665 switch (amode) {
6666 case 0: /* DA */
6667 offset = -4;
6668 break;
6669 case 1: /* IA */
6670 offset = 0;
6671 break;
6672 case 2: /* DB */
6673 offset = -8;
6674 break;
6675 case 3: /* IB */
6676 offset = 4;
6677 break;
6678 default:
6679 abort();
6680 }
6681 tcg_gen_addi_i32(addr, addr, offset);
6682 tmp = load_reg(s, 14);
6683 gen_aa32_st32(tmp, addr, 0);
6684 tcg_temp_free_i32(tmp);
6685 tmp = load_cpu_field(spsr);
6686 tcg_gen_addi_i32(addr, addr, 4);
6687 gen_aa32_st32(tmp, addr, 0);
6688 tcg_temp_free_i32(tmp);
6689 if (writeback) {
6690 switch (amode) {
6691 case 0:
6692 offset = -8;
6693 break;
6694 case 1:
6695 offset = 4;
6696 break;
6697 case 2:
6698 offset = -4;
6699 break;
6700 case 3:
6701 offset = 0;
6702 break;
6703 default:
6704 abort();
6705 }
6706 tcg_gen_addi_i32(addr, addr, offset);
6707 tmp = tcg_const_i32(mode);
6708 gen_helper_set_r13_banked(cpu_env, tmp, addr);
6709 tcg_temp_free_i32(tmp);
6710 }
6711 tcg_temp_free_i32(addr);
6712 }
6713
6714 static void disas_arm_insn(CPUARMState * env, DisasContext *s)
6715 {
6716 unsigned int cond, insn, val, op1, i, shift, rm, rs, rn, rd, sh;
6717 TCGv_i32 tmp;
6718 TCGv_i32 tmp2;
6719 TCGv_i32 tmp3;
6720 TCGv_i32 addr;
6721 TCGv_i64 tmp64;
6722
6723 insn = arm_ldl_code(env, s->pc, s->bswap_code);
6724 s->pc += 4;
6725
6726 /* M variants do not implement ARM mode. */
6727 if (IS_M(env))
6728 goto illegal_op;
6729 cond = insn >> 28;
6730 if (cond == 0xf){
6731 /* In ARMv3 and v4 the NV condition is UNPREDICTABLE; we
6732 * choose to UNDEF. In ARMv5 and above the space is used
6733 * for miscellaneous unconditional instructions.
6734 */
6735 ARCH(5);
6736
6737 /* Unconditional instructions. */
6738 if (((insn >> 25) & 7) == 1) {
6739 /* NEON Data processing. */
6740 if (!arm_feature(env, ARM_FEATURE_NEON))
6741 goto illegal_op;
6742
6743 if (disas_neon_data_insn(env, s, insn))
6744 goto illegal_op;
6745 return;
6746 }
6747 if ((insn & 0x0f100000) == 0x04000000) {
6748 /* NEON load/store. */
6749 if (!arm_feature(env, ARM_FEATURE_NEON))
6750 goto illegal_op;
6751
6752 if (disas_neon_ls_insn(env, s, insn))
6753 goto illegal_op;
6754 return;
6755 }
6756 if (((insn & 0x0f30f000) == 0x0510f000) ||
6757 ((insn & 0x0f30f010) == 0x0710f000)) {
6758 if ((insn & (1 << 22)) == 0) {
6759 /* PLDW; v7MP */
6760 if (!arm_feature(env, ARM_FEATURE_V7MP)) {
6761 goto illegal_op;
6762 }
6763 }
6764 /* Otherwise PLD; v5TE+ */
6765 ARCH(5TE);
6766 return;
6767 }
6768 if (((insn & 0x0f70f000) == 0x0450f000) ||
6769 ((insn & 0x0f70f010) == 0x0650f000)) {
6770 ARCH(7);
6771 return; /* PLI; V7 */
6772 }
6773 if (((insn & 0x0f700000) == 0x04100000) ||
6774 ((insn & 0x0f700010) == 0x06100000)) {
6775 if (!arm_feature(env, ARM_FEATURE_V7MP)) {
6776 goto illegal_op;
6777 }
6778 return; /* v7MP: Unallocated memory hint: must NOP */
6779 }
6780
6781 if ((insn & 0x0ffffdff) == 0x01010000) {
6782 ARCH(6);
6783 /* setend */
6784 if (((insn >> 9) & 1) != s->bswap_code) {
6785 /* Dynamic endianness switching not implemented. */
6786 qemu_log_mask(LOG_UNIMP, "arm: unimplemented setend\n");
6787 goto illegal_op;
6788 }
6789 return;
6790 } else if ((insn & 0x0fffff00) == 0x057ff000) {
6791 switch ((insn >> 4) & 0xf) {
6792 case 1: /* clrex */
6793 ARCH(6K);
6794 gen_clrex(s);
6795 return;
6796 case 4: /* dsb */
6797 case 5: /* dmb */
6798 case 6: /* isb */
6799 ARCH(7);
6800 /* We don't emulate caches so these are a no-op. */
6801 return;
6802 default:
6803 goto illegal_op;
6804 }
6805 } else if ((insn & 0x0e5fffe0) == 0x084d0500) {
6806 /* srs */
6807 if (IS_USER(s)) {
6808 goto illegal_op;
6809 }
6810 ARCH(6);
6811 gen_srs(s, (insn & 0x1f), (insn >> 23) & 3, insn & (1 << 21));
6812 return;
6813 } else if ((insn & 0x0e50ffe0) == 0x08100a00) {
6814 /* rfe */
6815 int32_t offset;
6816 if (IS_USER(s))
6817 goto illegal_op;
6818 ARCH(6);
6819 rn = (insn >> 16) & 0xf;
6820 addr = load_reg(s, rn);
6821 i = (insn >> 23) & 3;
6822 switch (i) {
6823 case 0: offset = -4; break; /* DA */
6824 case 1: offset = 0; break; /* IA */
6825 case 2: offset = -8; break; /* DB */
6826 case 3: offset = 4; break; /* IB */
6827 default: abort();
6828 }
6829 if (offset)
6830 tcg_gen_addi_i32(addr, addr, offset);
6831 /* Load PC into tmp and CPSR into tmp2. */
6832 tmp = tcg_temp_new_i32();
6833 gen_aa32_ld32u(tmp, addr, 0);
6834 tcg_gen_addi_i32(addr, addr, 4);
6835 tmp2 = tcg_temp_new_i32();
6836 gen_aa32_ld32u(tmp2, addr, 0);
6837 if (insn & (1 << 21)) {
6838 /* Base writeback. */
6839 switch (i) {
6840 case 0: offset = -8; break;
6841 case 1: offset = 4; break;
6842 case 2: offset = -4; break;
6843 case 3: offset = 0; break;
6844 default: abort();
6845 }
6846 if (offset)
6847 tcg_gen_addi_i32(addr, addr, offset);
6848 store_reg(s, rn, addr);
6849 } else {
6850 tcg_temp_free_i32(addr);
6851 }
6852 gen_rfe(s, tmp, tmp2);
6853 return;
6854 } else if ((insn & 0x0e000000) == 0x0a000000) {
6855 /* branch link and change to thumb (blx <offset>) */
6856 int32_t offset;
6857
6858 val = (uint32_t)s->pc;
6859 tmp = tcg_temp_new_i32();
6860 tcg_gen_movi_i32(tmp, val);
6861 store_reg(s, 14, tmp);
6862 /* Sign-extend the 24-bit offset */
6863 offset = (((int32_t)insn) << 8) >> 8;
6864 /* offset * 4 + bit24 * 2 + (thumb bit) */
6865 val += (offset << 2) | ((insn >> 23) & 2) | 1;
6866 /* pipeline offset */
6867 val += 4;
6868 /* protected by ARCH(5); above, near the start of uncond block */
6869 gen_bx_im(s, val);
6870 return;
6871 } else if ((insn & 0x0e000f00) == 0x0c000100) {
6872 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
6873 /* iWMMXt register transfer. */
6874 if (env->cp15.c15_cpar & (1 << 1))
6875 if (!disas_iwmmxt_insn(env, s, insn))
6876 return;
6877 }
6878 } else if ((insn & 0x0fe00000) == 0x0c400000) {
6879 /* Coprocessor double register transfer. */
6880 ARCH(5TE);
6881 } else if ((insn & 0x0f000010) == 0x0e000010) {
6882 /* Additional coprocessor register transfer. */
6883 } else if ((insn & 0x0ff10020) == 0x01000000) {
6884 uint32_t mask;
6885 uint32_t val;
6886 /* cps (privileged) */
6887 if (IS_USER(s))
6888 return;
6889 mask = val = 0;
6890 if (insn & (1 << 19)) {
6891 if (insn & (1 << 8))
6892 mask |= CPSR_A;
6893 if (insn & (1 << 7))
6894 mask |= CPSR_I;
6895 if (insn & (1 << 6))
6896 mask |= CPSR_F;
6897 if (insn & (1 << 18))
6898 val |= mask;
6899 }
6900 if (insn & (1 << 17)) {
6901 mask |= CPSR_M;
6902 val |= (insn & 0x1f);
6903 }
6904 if (mask) {
6905 gen_set_psr_im(s, mask, 0, val);
6906 }
6907 return;
6908 }
6909 goto illegal_op;
6910 }
6911 if (cond != 0xe) {
6912 /* if not always execute, we generate a conditional jump to
6913 next instruction */
6914 s->condlabel = gen_new_label();
6915 gen_test_cc(cond ^ 1, s->condlabel);
6916 s->condjmp = 1;
6917 }
6918 if ((insn & 0x0f900000) == 0x03000000) {
6919 if ((insn & (1 << 21)) == 0) {
6920 ARCH(6T2);
6921 rd = (insn >> 12) & 0xf;
6922 val = ((insn >> 4) & 0xf000) | (insn & 0xfff);
6923 if ((insn & (1 << 22)) == 0) {
6924 /* MOVW */
6925 tmp = tcg_temp_new_i32();
6926 tcg_gen_movi_i32(tmp, val);
6927 } else {
6928 /* MOVT */
6929 tmp = load_reg(s, rd);
6930 tcg_gen_ext16u_i32(tmp, tmp);
6931 tcg_gen_ori_i32(tmp, tmp, val << 16);
6932 }
6933 store_reg(s, rd, tmp);
6934 } else {
6935 if (((insn >> 12) & 0xf) != 0xf)
6936 goto illegal_op;
6937 if (((insn >> 16) & 0xf) == 0) {
6938 gen_nop_hint(s, insn & 0xff);
6939 } else {
6940 /* CPSR = immediate */
6941 val = insn & 0xff;
6942 shift = ((insn >> 8) & 0xf) * 2;
6943 if (shift)
6944 val = (val >> shift) | (val << (32 - shift));
6945 i = ((insn & (1 << 22)) != 0);
6946 if (gen_set_psr_im(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i, val))
6947 goto illegal_op;
6948 }
6949 }
6950 } else if ((insn & 0x0f900000) == 0x01000000
6951 && (insn & 0x00000090) != 0x00000090) {
6952 /* miscellaneous instructions */
6953 op1 = (insn >> 21) & 3;
6954 sh = (insn >> 4) & 0xf;
6955 rm = insn & 0xf;
6956 switch (sh) {
6957 case 0x0: /* move program status register */
6958 if (op1 & 1) {
6959 /* PSR = reg */
6960 tmp = load_reg(s, rm);
6961 i = ((op1 & 2) != 0);
6962 if (gen_set_psr(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i, tmp))
6963 goto illegal_op;
6964 } else {
6965 /* reg = PSR */
6966 rd = (insn >> 12) & 0xf;
6967 if (op1 & 2) {
6968 if (IS_USER(s))
6969 goto illegal_op;
6970 tmp = load_cpu_field(spsr);
6971 } else {
6972 tmp = tcg_temp_new_i32();
6973 gen_helper_cpsr_read(tmp, cpu_env);
6974 }
6975 store_reg(s, rd, tmp);
6976 }
6977 break;
6978 case 0x1:
6979 if (op1 == 1) {
6980 /* branch/exchange thumb (bx). */
6981 ARCH(4T);
6982 tmp = load_reg(s, rm);
6983 gen_bx(s, tmp);
6984 } else if (op1 == 3) {
6985 /* clz */
6986 ARCH(5);
6987 rd = (insn >> 12) & 0xf;
6988 tmp = load_reg(s, rm);
6989 gen_helper_clz(tmp, tmp);
6990 store_reg(s, rd, tmp);
6991 } else {
6992 goto illegal_op;
6993 }
6994 break;
6995 case 0x2:
6996 if (op1 == 1) {
6997 ARCH(5J); /* bxj */
6998 /* Trivial implementation equivalent to bx. */
6999 tmp = load_reg(s, rm);
7000 gen_bx(s, tmp);
7001 } else {
7002 goto illegal_op;
7003 }
7004 break;
7005 case 0x3:
7006 if (op1 != 1)
7007 goto illegal_op;
7008
7009 ARCH(5);
7010 /* branch link/exchange thumb (blx) */
7011 tmp = load_reg(s, rm);
7012 tmp2 = tcg_temp_new_i32();
7013 tcg_gen_movi_i32(tmp2, s->pc);
7014 store_reg(s, 14, tmp2);
7015 gen_bx(s, tmp);
7016 break;
7017 case 0x5: /* saturating add/subtract */
7018 ARCH(5TE);
7019 rd = (insn >> 12) & 0xf;
7020 rn = (insn >> 16) & 0xf;
7021 tmp = load_reg(s, rm);
7022 tmp2 = load_reg(s, rn);
7023 if (op1 & 2)
7024 gen_helper_double_saturate(tmp2, cpu_env, tmp2);
7025 if (op1 & 1)
7026 gen_helper_sub_saturate(tmp, cpu_env, tmp, tmp2);
7027 else
7028 gen_helper_add_saturate(tmp, cpu_env, tmp, tmp2);
7029 tcg_temp_free_i32(tmp2);
7030 store_reg(s, rd, tmp);
7031 break;
7032 case 7:
7033 /* SMC instruction (op1 == 3)
7034 and undefined instructions (op1 == 0 || op1 == 2)
7035 will trap */
7036 if (op1 != 1) {
7037 goto illegal_op;
7038 }
7039 /* bkpt */
7040 ARCH(5);
7041 gen_exception_insn(s, 4, EXCP_BKPT);
7042 break;
7043 case 0x8: /* signed multiply */
7044 case 0xa:
7045 case 0xc:
7046 case 0xe:
7047 ARCH(5TE);
7048 rs = (insn >> 8) & 0xf;
7049 rn = (insn >> 12) & 0xf;
7050 rd = (insn >> 16) & 0xf;
7051 if (op1 == 1) {
7052 /* (32 * 16) >> 16 */
7053 tmp = load_reg(s, rm);
7054 tmp2 = load_reg(s, rs);
7055 if (sh & 4)
7056 tcg_gen_sari_i32(tmp2, tmp2, 16);
7057 else
7058 gen_sxth(tmp2);
7059 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7060 tcg_gen_shri_i64(tmp64, tmp64, 16);
7061 tmp = tcg_temp_new_i32();
7062 tcg_gen_trunc_i64_i32(tmp, tmp64);
7063 tcg_temp_free_i64(tmp64);
7064 if ((sh & 2) == 0) {
7065 tmp2 = load_reg(s, rn);
7066 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
7067 tcg_temp_free_i32(tmp2);
7068 }
7069 store_reg(s, rd, tmp);
7070 } else {
7071 /* 16 * 16 */
7072 tmp = load_reg(s, rm);
7073 tmp2 = load_reg(s, rs);
7074 gen_mulxy(tmp, tmp2, sh & 2, sh & 4);
7075 tcg_temp_free_i32(tmp2);
7076 if (op1 == 2) {
7077 tmp64 = tcg_temp_new_i64();
7078 tcg_gen_ext_i32_i64(tmp64, tmp);
7079 tcg_temp_free_i32(tmp);
7080 gen_addq(s, tmp64, rn, rd);
7081 gen_storeq_reg(s, rn, rd, tmp64);
7082 tcg_temp_free_i64(tmp64);
7083 } else {
7084 if (op1 == 0) {
7085 tmp2 = load_reg(s, rn);
7086 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
7087 tcg_temp_free_i32(tmp2);
7088 }
7089 store_reg(s, rd, tmp);
7090 }
7091 }
7092 break;
7093 default:
7094 goto illegal_op;
7095 }
7096 } else if (((insn & 0x0e000000) == 0 &&
7097 (insn & 0x00000090) != 0x90) ||
7098 ((insn & 0x0e000000) == (1 << 25))) {
7099 int set_cc, logic_cc, shiftop;
7100
7101 op1 = (insn >> 21) & 0xf;
7102 set_cc = (insn >> 20) & 1;
7103 logic_cc = table_logic_cc[op1] & set_cc;
7104
7105 /* data processing instruction */
7106 if (insn & (1 << 25)) {
7107 /* immediate operand */
7108 val = insn & 0xff;
7109 shift = ((insn >> 8) & 0xf) * 2;
7110 if (shift) {
7111 val = (val >> shift) | (val << (32 - shift));
7112 }
7113 tmp2 = tcg_temp_new_i32();
7114 tcg_gen_movi_i32(tmp2, val);
7115 if (logic_cc && shift) {
7116 gen_set_CF_bit31(tmp2);
7117 }
7118 } else {
7119 /* register */
7120 rm = (insn) & 0xf;
7121 tmp2 = load_reg(s, rm);
7122 shiftop = (insn >> 5) & 3;
7123 if (!(insn & (1 << 4))) {
7124 shift = (insn >> 7) & 0x1f;
7125 gen_arm_shift_im(tmp2, shiftop, shift, logic_cc);
7126 } else {
7127 rs = (insn >> 8) & 0xf;
7128 tmp = load_reg(s, rs);
7129 gen_arm_shift_reg(tmp2, shiftop, tmp, logic_cc);
7130 }
7131 }
7132 if (op1 != 0x0f && op1 != 0x0d) {
7133 rn = (insn >> 16) & 0xf;
7134 tmp = load_reg(s, rn);
7135 } else {
7136 TCGV_UNUSED_I32(tmp);
7137 }
7138 rd = (insn >> 12) & 0xf;
7139 switch(op1) {
7140 case 0x00:
7141 tcg_gen_and_i32(tmp, tmp, tmp2);
7142 if (logic_cc) {
7143 gen_logic_CC(tmp);
7144 }
7145 store_reg_bx(env, s, rd, tmp);
7146 break;
7147 case 0x01:
7148 tcg_gen_xor_i32(tmp, tmp, tmp2);
7149 if (logic_cc) {
7150 gen_logic_CC(tmp);
7151 }
7152 store_reg_bx(env, s, rd, tmp);
7153 break;
7154 case 0x02:
7155 if (set_cc && rd == 15) {
7156 /* SUBS r15, ... is used for exception return. */
7157 if (IS_USER(s)) {
7158 goto illegal_op;
7159 }
7160 gen_sub_CC(tmp, tmp, tmp2);
7161 gen_exception_return(s, tmp);
7162 } else {
7163 if (set_cc) {
7164 gen_sub_CC(tmp, tmp, tmp2);
7165 } else {
7166 tcg_gen_sub_i32(tmp, tmp, tmp2);
7167 }
7168 store_reg_bx(env, s, rd, tmp);
7169 }
7170 break;
7171 case 0x03:
7172 if (set_cc) {
7173 gen_sub_CC(tmp, tmp2, tmp);
7174 } else {
7175 tcg_gen_sub_i32(tmp, tmp2, tmp);
7176 }
7177 store_reg_bx(env, s, rd, tmp);
7178 break;
7179 case 0x04:
7180 if (set_cc) {
7181 gen_add_CC(tmp, tmp, tmp2);
7182 } else {
7183 tcg_gen_add_i32(tmp, tmp, tmp2);
7184 }
7185 store_reg_bx(env, s, rd, tmp);
7186 break;
7187 case 0x05:
7188 if (set_cc) {
7189 gen_adc_CC(tmp, tmp, tmp2);
7190 } else {
7191 gen_add_carry(tmp, tmp, tmp2);
7192 }
7193 store_reg_bx(env, s, rd, tmp);
7194 break;
7195 case 0x06:
7196 if (set_cc) {
7197 gen_sbc_CC(tmp, tmp, tmp2);
7198 } else {
7199 gen_sub_carry(tmp, tmp, tmp2);
7200 }
7201 store_reg_bx(env, s, rd, tmp);
7202 break;
7203 case 0x07:
7204 if (set_cc) {
7205 gen_sbc_CC(tmp, tmp2, tmp);
7206 } else {
7207 gen_sub_carry(tmp, tmp2, tmp);
7208 }
7209 store_reg_bx(env, s, rd, tmp);
7210 break;
7211 case 0x08:
7212 if (set_cc) {
7213 tcg_gen_and_i32(tmp, tmp, tmp2);
7214 gen_logic_CC(tmp);
7215 }
7216 tcg_temp_free_i32(tmp);
7217 break;
7218 case 0x09:
7219 if (set_cc) {
7220 tcg_gen_xor_i32(tmp, tmp, tmp2);
7221 gen_logic_CC(tmp);
7222 }
7223 tcg_temp_free_i32(tmp);
7224 break;
7225 case 0x0a:
7226 if (set_cc) {
7227 gen_sub_CC(tmp, tmp, tmp2);
7228 }
7229 tcg_temp_free_i32(tmp);
7230 break;
7231 case 0x0b:
7232 if (set_cc) {
7233 gen_add_CC(tmp, tmp, tmp2);
7234 }
7235 tcg_temp_free_i32(tmp);
7236 break;
7237 case 0x0c:
7238 tcg_gen_or_i32(tmp, tmp, tmp2);
7239 if (logic_cc) {
7240 gen_logic_CC(tmp);
7241 }
7242 store_reg_bx(env, s, rd, tmp);
7243 break;
7244 case 0x0d:
7245 if (logic_cc && rd == 15) {
7246 /* MOVS r15, ... is used for exception return. */
7247 if (IS_USER(s)) {
7248 goto illegal_op;
7249 }
7250 gen_exception_return(s, tmp2);
7251 } else {
7252 if (logic_cc) {
7253 gen_logic_CC(tmp2);
7254 }
7255 store_reg_bx(env, s, rd, tmp2);
7256 }
7257 break;
7258 case 0x0e:
7259 tcg_gen_andc_i32(tmp, tmp, tmp2);
7260 if (logic_cc) {
7261 gen_logic_CC(tmp);
7262 }
7263 store_reg_bx(env, s, rd, tmp);
7264 break;
7265 default:
7266 case 0x0f:
7267 tcg_gen_not_i32(tmp2, tmp2);
7268 if (logic_cc) {
7269 gen_logic_CC(tmp2);
7270 }
7271 store_reg_bx(env, s, rd, tmp2);
7272 break;
7273 }
7274 if (op1 != 0x0f && op1 != 0x0d) {
7275 tcg_temp_free_i32(tmp2);
7276 }
7277 } else {
7278 /* other instructions */
7279 op1 = (insn >> 24) & 0xf;
7280 switch(op1) {
7281 case 0x0:
7282 case 0x1:
7283 /* multiplies, extra load/stores */
7284 sh = (insn >> 5) & 3;
7285 if (sh == 0) {
7286 if (op1 == 0x0) {
7287 rd = (insn >> 16) & 0xf;
7288 rn = (insn >> 12) & 0xf;
7289 rs = (insn >> 8) & 0xf;
7290 rm = (insn) & 0xf;
7291 op1 = (insn >> 20) & 0xf;
7292 switch (op1) {
7293 case 0: case 1: case 2: case 3: case 6:
7294 /* 32 bit mul */
7295 tmp = load_reg(s, rs);
7296 tmp2 = load_reg(s, rm);
7297 tcg_gen_mul_i32(tmp, tmp, tmp2);
7298 tcg_temp_free_i32(tmp2);
7299 if (insn & (1 << 22)) {
7300 /* Subtract (mls) */
7301 ARCH(6T2);
7302 tmp2 = load_reg(s, rn);
7303 tcg_gen_sub_i32(tmp, tmp2, tmp);
7304 tcg_temp_free_i32(tmp2);
7305 } else if (insn & (1 << 21)) {
7306 /* Add */
7307 tmp2 = load_reg(s, rn);
7308 tcg_gen_add_i32(tmp, tmp, tmp2);
7309 tcg_temp_free_i32(tmp2);
7310 }
7311 if (insn & (1 << 20))
7312 gen_logic_CC(tmp);
7313 store_reg(s, rd, tmp);
7314 break;
7315 case 4:
7316 /* 64 bit mul double accumulate (UMAAL) */
7317 ARCH(6);
7318 tmp = load_reg(s, rs);
7319 tmp2 = load_reg(s, rm);
7320 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
7321 gen_addq_lo(s, tmp64, rn);
7322 gen_addq_lo(s, tmp64, rd);
7323 gen_storeq_reg(s, rn, rd, tmp64);
7324 tcg_temp_free_i64(tmp64);
7325 break;
7326 case 8: case 9: case 10: case 11:
7327 case 12: case 13: case 14: case 15:
7328 /* 64 bit mul: UMULL, UMLAL, SMULL, SMLAL. */
7329 tmp = load_reg(s, rs);
7330 tmp2 = load_reg(s, rm);
7331 if (insn & (1 << 22)) {
7332 tcg_gen_muls2_i32(tmp, tmp2, tmp, tmp2);
7333 } else {
7334 tcg_gen_mulu2_i32(tmp, tmp2, tmp, tmp2);
7335 }
7336 if (insn & (1 << 21)) { /* mult accumulate */
7337 TCGv_i32 al = load_reg(s, rn);
7338 TCGv_i32 ah = load_reg(s, rd);
7339 tcg_gen_add2_i32(tmp, tmp2, tmp, tmp2, al, ah);
7340 tcg_temp_free_i32(al);
7341 tcg_temp_free_i32(ah);
7342 }
7343 if (insn & (1 << 20)) {
7344 gen_logicq_cc(tmp, tmp2);
7345 }
7346 store_reg(s, rn, tmp);
7347 store_reg(s, rd, tmp2);
7348 break;
7349 default:
7350 goto illegal_op;
7351 }
7352 } else {
7353 rn = (insn >> 16) & 0xf;
7354 rd = (insn >> 12) & 0xf;
7355 if (insn & (1 << 23)) {
7356 /* load/store exclusive */
7357 int op2 = (insn >> 8) & 3;
7358 op1 = (insn >> 21) & 0x3;
7359
7360 switch (op2) {
7361 case 0: /* lda/stl */
7362 if (op1 == 1) {
7363 goto illegal_op;
7364 }
7365 ARCH(8);
7366 break;
7367 case 1: /* reserved */
7368 goto illegal_op;
7369 case 2: /* ldaex/stlex */
7370 ARCH(8);
7371 break;
7372 case 3: /* ldrex/strex */
7373 if (op1) {
7374 ARCH(6K);
7375 } else {
7376 ARCH(6);
7377 }
7378 break;
7379 }
7380
7381 addr = tcg_temp_local_new_i32();
7382 load_reg_var(s, addr, rn);
7383
7384 /* Since the emulation does not have barriers,
7385 the acquire/release semantics need no special
7386 handling */
7387 if (op2 == 0) {
7388 if (insn & (1 << 20)) {
7389 tmp = tcg_temp_new_i32();
7390 switch (op1) {
7391 case 0: /* lda */
7392 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7393 break;
7394 case 2: /* ldab */
7395 gen_aa32_ld8u(tmp, addr, IS_USER(s));
7396 break;
7397 case 3: /* ldah */
7398 gen_aa32_ld16u(tmp, addr, IS_USER(s));
7399 break;
7400 default:
7401 abort();
7402 }
7403 store_reg(s, rd, tmp);
7404 } else {
7405 rm = insn & 0xf;
7406 tmp = load_reg(s, rm);
7407 switch (op1) {
7408 case 0: /* stl */
7409 gen_aa32_st32(tmp, addr, IS_USER(s));
7410 break;
7411 case 2: /* stlb */
7412 gen_aa32_st8(tmp, addr, IS_USER(s));
7413 break;
7414 case 3: /* stlh */
7415 gen_aa32_st16(tmp, addr, IS_USER(s));
7416 break;
7417 default:
7418 abort();
7419 }
7420 tcg_temp_free_i32(tmp);
7421 }
7422 } else if (insn & (1 << 20)) {
7423 switch (op1) {
7424 case 0: /* ldrex */
7425 gen_load_exclusive(s, rd, 15, addr, 2);
7426 break;
7427 case 1: /* ldrexd */
7428 gen_load_exclusive(s, rd, rd + 1, addr, 3);
7429 break;
7430 case 2: /* ldrexb */
7431 gen_load_exclusive(s, rd, 15, addr, 0);
7432 break;
7433 case 3: /* ldrexh */
7434 gen_load_exclusive(s, rd, 15, addr, 1);
7435 break;
7436 default:
7437 abort();
7438 }
7439 } else {
7440 rm = insn & 0xf;
7441 switch (op1) {
7442 case 0: /* strex */
7443 gen_store_exclusive(s, rd, rm, 15, addr, 2);
7444 break;
7445 case 1: /* strexd */
7446 gen_store_exclusive(s, rd, rm, rm + 1, addr, 3);
7447 break;
7448 case 2: /* strexb */
7449 gen_store_exclusive(s, rd, rm, 15, addr, 0);
7450 break;
7451 case 3: /* strexh */
7452 gen_store_exclusive(s, rd, rm, 15, addr, 1);
7453 break;
7454 default:
7455 abort();
7456 }
7457 }
7458 tcg_temp_free_i32(addr);
7459 } else {
7460 /* SWP instruction */
7461 rm = (insn) & 0xf;
7462
7463 /* ??? This is not really atomic. However we know
7464 we never have multiple CPUs running in parallel,
7465 so it is good enough. */
7466 addr = load_reg(s, rn);
7467 tmp = load_reg(s, rm);
7468 tmp2 = tcg_temp_new_i32();
7469 if (insn & (1 << 22)) {
7470 gen_aa32_ld8u(tmp2, addr, IS_USER(s));
7471 gen_aa32_st8(tmp, addr, IS_USER(s));
7472 } else {
7473 gen_aa32_ld32u(tmp2, addr, IS_USER(s));
7474 gen_aa32_st32(tmp, addr, IS_USER(s));
7475 }
7476 tcg_temp_free_i32(tmp);
7477 tcg_temp_free_i32(addr);
7478 store_reg(s, rd, tmp2);
7479 }
7480 }
7481 } else {
7482 int address_offset;
7483 int load;
7484 /* Misc load/store */
7485 rn = (insn >> 16) & 0xf;
7486 rd = (insn >> 12) & 0xf;
7487 addr = load_reg(s, rn);
7488 if (insn & (1 << 24))
7489 gen_add_datah_offset(s, insn, 0, addr);
7490 address_offset = 0;
7491 if (insn & (1 << 20)) {
7492 /* load */
7493 tmp = tcg_temp_new_i32();
7494 switch(sh) {
7495 case 1:
7496 gen_aa32_ld16u(tmp, addr, IS_USER(s));
7497 break;
7498 case 2:
7499 gen_aa32_ld8s(tmp, addr, IS_USER(s));
7500 break;
7501 default:
7502 case 3:
7503 gen_aa32_ld16s(tmp, addr, IS_USER(s));
7504 break;
7505 }
7506 load = 1;
7507 } else if (sh & 2) {
7508 ARCH(5TE);
7509 /* doubleword */
7510 if (sh & 1) {
7511 /* store */
7512 tmp = load_reg(s, rd);
7513 gen_aa32_st32(tmp, addr, IS_USER(s));
7514 tcg_temp_free_i32(tmp);
7515 tcg_gen_addi_i32(addr, addr, 4);
7516 tmp = load_reg(s, rd + 1);
7517 gen_aa32_st32(tmp, addr, IS_USER(s));
7518 tcg_temp_free_i32(tmp);
7519 load = 0;
7520 } else {
7521 /* load */
7522 tmp = tcg_temp_new_i32();
7523 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7524 store_reg(s, rd, tmp);
7525 tcg_gen_addi_i32(addr, addr, 4);
7526 tmp = tcg_temp_new_i32();
7527 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7528 rd++;
7529 load = 1;
7530 }
7531 address_offset = -4;
7532 } else {
7533 /* store */
7534 tmp = load_reg(s, rd);
7535 gen_aa32_st16(tmp, addr, IS_USER(s));
7536 tcg_temp_free_i32(tmp);
7537 load = 0;
7538 }
7539 /* Perform base writeback before the loaded value to
7540 ensure correct behavior with overlapping index registers.
7541 ldrd with base writeback is is undefined if the
7542 destination and index registers overlap. */
7543 if (!(insn & (1 << 24))) {
7544 gen_add_datah_offset(s, insn, address_offset, addr);
7545 store_reg(s, rn, addr);
7546 } else if (insn & (1 << 21)) {
7547 if (address_offset)
7548 tcg_gen_addi_i32(addr, addr, address_offset);
7549 store_reg(s, rn, addr);
7550 } else {
7551 tcg_temp_free_i32(addr);
7552 }
7553 if (load) {
7554 /* Complete the load. */
7555 store_reg(s, rd, tmp);
7556 }
7557 }
7558 break;
7559 case 0x4:
7560 case 0x5:
7561 goto do_ldst;
7562 case 0x6:
7563 case 0x7:
7564 if (insn & (1 << 4)) {
7565 ARCH(6);
7566 /* Armv6 Media instructions. */
7567 rm = insn & 0xf;
7568 rn = (insn >> 16) & 0xf;
7569 rd = (insn >> 12) & 0xf;
7570 rs = (insn >> 8) & 0xf;
7571 switch ((insn >> 23) & 3) {
7572 case 0: /* Parallel add/subtract. */
7573 op1 = (insn >> 20) & 7;
7574 tmp = load_reg(s, rn);
7575 tmp2 = load_reg(s, rm);
7576 sh = (insn >> 5) & 7;
7577 if ((op1 & 3) == 0 || sh == 5 || sh == 6)
7578 goto illegal_op;
7579 gen_arm_parallel_addsub(op1, sh, tmp, tmp2);
7580 tcg_temp_free_i32(tmp2);
7581 store_reg(s, rd, tmp);
7582 break;
7583 case 1:
7584 if ((insn & 0x00700020) == 0) {
7585 /* Halfword pack. */
7586 tmp = load_reg(s, rn);
7587 tmp2 = load_reg(s, rm);
7588 shift = (insn >> 7) & 0x1f;
7589 if (insn & (1 << 6)) {
7590 /* pkhtb */
7591 if (shift == 0)
7592 shift = 31;
7593 tcg_gen_sari_i32(tmp2, tmp2, shift);
7594 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
7595 tcg_gen_ext16u_i32(tmp2, tmp2);
7596 } else {
7597 /* pkhbt */
7598 if (shift)
7599 tcg_gen_shli_i32(tmp2, tmp2, shift);
7600 tcg_gen_ext16u_i32(tmp, tmp);
7601 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
7602 }
7603 tcg_gen_or_i32(tmp, tmp, tmp2);
7604 tcg_temp_free_i32(tmp2);
7605 store_reg(s, rd, tmp);
7606 } else if ((insn & 0x00200020) == 0x00200000) {
7607 /* [us]sat */
7608 tmp = load_reg(s, rm);
7609 shift = (insn >> 7) & 0x1f;
7610 if (insn & (1 << 6)) {
7611 if (shift == 0)
7612 shift = 31;
7613 tcg_gen_sari_i32(tmp, tmp, shift);
7614 } else {
7615 tcg_gen_shli_i32(tmp, tmp, shift);
7616 }
7617 sh = (insn >> 16) & 0x1f;
7618 tmp2 = tcg_const_i32(sh);
7619 if (insn & (1 << 22))
7620 gen_helper_usat(tmp, cpu_env, tmp, tmp2);
7621 else
7622 gen_helper_ssat(tmp, cpu_env, tmp, tmp2);
7623 tcg_temp_free_i32(tmp2);
7624 store_reg(s, rd, tmp);
7625 } else if ((insn & 0x00300fe0) == 0x00200f20) {
7626 /* [us]sat16 */
7627 tmp = load_reg(s, rm);
7628 sh = (insn >> 16) & 0x1f;
7629 tmp2 = tcg_const_i32(sh);
7630 if (insn & (1 << 22))
7631 gen_helper_usat16(tmp, cpu_env, tmp, tmp2);
7632 else
7633 gen_helper_ssat16(tmp, cpu_env, tmp, tmp2);
7634 tcg_temp_free_i32(tmp2);
7635 store_reg(s, rd, tmp);
7636 } else if ((insn & 0x00700fe0) == 0x00000fa0) {
7637 /* Select bytes. */
7638 tmp = load_reg(s, rn);
7639 tmp2 = load_reg(s, rm);
7640 tmp3 = tcg_temp_new_i32();
7641 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUARMState, GE));
7642 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
7643 tcg_temp_free_i32(tmp3);
7644 tcg_temp_free_i32(tmp2);
7645 store_reg(s, rd, tmp);
7646 } else if ((insn & 0x000003e0) == 0x00000060) {
7647 tmp = load_reg(s, rm);
7648 shift = (insn >> 10) & 3;
7649 /* ??? In many cases it's not necessary to do a
7650 rotate, a shift is sufficient. */
7651 if (shift != 0)
7652 tcg_gen_rotri_i32(tmp, tmp, shift * 8);
7653 op1 = (insn >> 20) & 7;
7654 switch (op1) {
7655 case 0: gen_sxtb16(tmp); break;
7656 case 2: gen_sxtb(tmp); break;
7657 case 3: gen_sxth(tmp); break;
7658 case 4: gen_uxtb16(tmp); break;
7659 case 6: gen_uxtb(tmp); break;
7660 case 7: gen_uxth(tmp); break;
7661 default: goto illegal_op;
7662 }
7663 if (rn != 15) {
7664 tmp2 = load_reg(s, rn);
7665 if ((op1 & 3) == 0) {
7666 gen_add16(tmp, tmp2);
7667 } else {
7668 tcg_gen_add_i32(tmp, tmp, tmp2);
7669 tcg_temp_free_i32(tmp2);
7670 }
7671 }
7672 store_reg(s, rd, tmp);
7673 } else if ((insn & 0x003f0f60) == 0x003f0f20) {
7674 /* rev */
7675 tmp = load_reg(s, rm);
7676 if (insn & (1 << 22)) {
7677 if (insn & (1 << 7)) {
7678 gen_revsh(tmp);
7679 } else {
7680 ARCH(6T2);
7681 gen_helper_rbit(tmp, tmp);
7682 }
7683 } else {
7684 if (insn & (1 << 7))
7685 gen_rev16(tmp);
7686 else
7687 tcg_gen_bswap32_i32(tmp, tmp);
7688 }
7689 store_reg(s, rd, tmp);
7690 } else {
7691 goto illegal_op;
7692 }
7693 break;
7694 case 2: /* Multiplies (Type 3). */
7695 switch ((insn >> 20) & 0x7) {
7696 case 5:
7697 if (((insn >> 6) ^ (insn >> 7)) & 1) {
7698 /* op2 not 00x or 11x : UNDEF */
7699 goto illegal_op;
7700 }
7701 /* Signed multiply most significant [accumulate].
7702 (SMMUL, SMMLA, SMMLS) */
7703 tmp = load_reg(s, rm);
7704 tmp2 = load_reg(s, rs);
7705 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7706
7707 if (rd != 15) {
7708 tmp = load_reg(s, rd);
7709 if (insn & (1 << 6)) {
7710 tmp64 = gen_subq_msw(tmp64, tmp);
7711 } else {
7712 tmp64 = gen_addq_msw(tmp64, tmp);
7713 }
7714 }
7715 if (insn & (1 << 5)) {
7716 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
7717 }
7718 tcg_gen_shri_i64(tmp64, tmp64, 32);
7719 tmp = tcg_temp_new_i32();
7720 tcg_gen_trunc_i64_i32(tmp, tmp64);
7721 tcg_temp_free_i64(tmp64);
7722 store_reg(s, rn, tmp);
7723 break;
7724 case 0:
7725 case 4:
7726 /* SMLAD, SMUAD, SMLSD, SMUSD, SMLALD, SMLSLD */
7727 if (insn & (1 << 7)) {
7728 goto illegal_op;
7729 }
7730 tmp = load_reg(s, rm);
7731 tmp2 = load_reg(s, rs);
7732 if (insn & (1 << 5))
7733 gen_swap_half(tmp2);
7734 gen_smul_dual(tmp, tmp2);
7735 if (insn & (1 << 6)) {
7736 /* This subtraction cannot overflow. */
7737 tcg_gen_sub_i32(tmp, tmp, tmp2);
7738 } else {
7739 /* This addition cannot overflow 32 bits;
7740 * however it may overflow considered as a signed
7741 * operation, in which case we must set the Q flag.
7742 */
7743 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
7744 }
7745 tcg_temp_free_i32(tmp2);
7746 if (insn & (1 << 22)) {
7747 /* smlald, smlsld */
7748 tmp64 = tcg_temp_new_i64();
7749 tcg_gen_ext_i32_i64(tmp64, tmp);
7750 tcg_temp_free_i32(tmp);
7751 gen_addq(s, tmp64, rd, rn);
7752 gen_storeq_reg(s, rd, rn, tmp64);
7753 tcg_temp_free_i64(tmp64);
7754 } else {
7755 /* smuad, smusd, smlad, smlsd */
7756 if (rd != 15)
7757 {
7758 tmp2 = load_reg(s, rd);
7759 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
7760 tcg_temp_free_i32(tmp2);
7761 }
7762 store_reg(s, rn, tmp);
7763 }
7764 break;
7765 case 1:
7766 case 3:
7767 /* SDIV, UDIV */
7768 if (!arm_feature(env, ARM_FEATURE_ARM_DIV)) {
7769 goto illegal_op;
7770 }
7771 if (((insn >> 5) & 7) || (rd != 15)) {
7772 goto illegal_op;
7773 }
7774 tmp = load_reg(s, rm);
7775 tmp2 = load_reg(s, rs);
7776 if (insn & (1 << 21)) {
7777 gen_helper_udiv(tmp, tmp, tmp2);
7778 } else {
7779 gen_helper_sdiv(tmp, tmp, tmp2);
7780 }
7781 tcg_temp_free_i32(tmp2);
7782 store_reg(s, rn, tmp);
7783 break;
7784 default:
7785 goto illegal_op;
7786 }
7787 break;
7788 case 3:
7789 op1 = ((insn >> 17) & 0x38) | ((insn >> 5) & 7);
7790 switch (op1) {
7791 case 0: /* Unsigned sum of absolute differences. */
7792 ARCH(6);
7793 tmp = load_reg(s, rm);
7794 tmp2 = load_reg(s, rs);
7795 gen_helper_usad8(tmp, tmp, tmp2);
7796 tcg_temp_free_i32(tmp2);
7797 if (rd != 15) {
7798 tmp2 = load_reg(s, rd);
7799 tcg_gen_add_i32(tmp, tmp, tmp2);
7800 tcg_temp_free_i32(tmp2);
7801 }
7802 store_reg(s, rn, tmp);
7803 break;
7804 case 0x20: case 0x24: case 0x28: case 0x2c:
7805 /* Bitfield insert/clear. */
7806 ARCH(6T2);
7807 shift = (insn >> 7) & 0x1f;
7808 i = (insn >> 16) & 0x1f;
7809 i = i + 1 - shift;
7810 if (rm == 15) {
7811 tmp = tcg_temp_new_i32();
7812 tcg_gen_movi_i32(tmp, 0);
7813 } else {
7814 tmp = load_reg(s, rm);
7815 }
7816 if (i != 32) {
7817 tmp2 = load_reg(s, rd);
7818 tcg_gen_deposit_i32(tmp, tmp2, tmp, shift, i);
7819 tcg_temp_free_i32(tmp2);
7820 }
7821 store_reg(s, rd, tmp);
7822 break;
7823 case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
7824 case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
7825 ARCH(6T2);
7826 tmp = load_reg(s, rm);
7827 shift = (insn >> 7) & 0x1f;
7828 i = ((insn >> 16) & 0x1f) + 1;
7829 if (shift + i > 32)
7830 goto illegal_op;
7831 if (i < 32) {
7832 if (op1 & 0x20) {
7833 gen_ubfx(tmp, shift, (1u << i) - 1);
7834 } else {
7835 gen_sbfx(tmp, shift, i);
7836 }
7837 }
7838 store_reg(s, rd, tmp);
7839 break;
7840 default:
7841 goto illegal_op;
7842 }
7843 break;
7844 }
7845 break;
7846 }
7847 do_ldst:
7848 /* Check for undefined extension instructions
7849 * per the ARM Bible IE:
7850 * xxxx 0111 1111 xxxx xxxx xxxx 1111 xxxx
7851 */
7852 sh = (0xf << 20) | (0xf << 4);
7853 if (op1 == 0x7 && ((insn & sh) == sh))
7854 {
7855 goto illegal_op;
7856 }
7857 /* load/store byte/word */
7858 rn = (insn >> 16) & 0xf;
7859 rd = (insn >> 12) & 0xf;
7860 tmp2 = load_reg(s, rn);
7861 i = (IS_USER(s) || (insn & 0x01200000) == 0x00200000);
7862 if (insn & (1 << 24))
7863 gen_add_data_offset(s, insn, tmp2);
7864 if (insn & (1 << 20)) {
7865 /* load */
7866 tmp = tcg_temp_new_i32();
7867 if (insn & (1 << 22)) {
7868 gen_aa32_ld8u(tmp, tmp2, i);
7869 } else {
7870 gen_aa32_ld32u(tmp, tmp2, i);
7871 }
7872 } else {
7873 /* store */
7874 tmp = load_reg(s, rd);
7875 if (insn & (1 << 22)) {
7876 gen_aa32_st8(tmp, tmp2, i);
7877 } else {
7878 gen_aa32_st32(tmp, tmp2, i);
7879 }
7880 tcg_temp_free_i32(tmp);
7881 }
7882 if (!(insn & (1 << 24))) {
7883 gen_add_data_offset(s, insn, tmp2);
7884 store_reg(s, rn, tmp2);
7885 } else if (insn & (1 << 21)) {
7886 store_reg(s, rn, tmp2);
7887 } else {
7888 tcg_temp_free_i32(tmp2);
7889 }
7890 if (insn & (1 << 20)) {
7891 /* Complete the load. */
7892 store_reg_from_load(env, s, rd, tmp);
7893 }
7894 break;
7895 case 0x08:
7896 case 0x09:
7897 {
7898 int j, n, user, loaded_base;
7899 TCGv_i32 loaded_var;
7900 /* load/store multiple words */
7901 /* XXX: store correct base if write back */
7902 user = 0;
7903 if (insn & (1 << 22)) {
7904 if (IS_USER(s))
7905 goto illegal_op; /* only usable in supervisor mode */
7906
7907 if ((insn & (1 << 15)) == 0)
7908 user = 1;
7909 }
7910 rn = (insn >> 16) & 0xf;
7911 addr = load_reg(s, rn);
7912
7913 /* compute total size */
7914 loaded_base = 0;
7915 TCGV_UNUSED_I32(loaded_var);
7916 n = 0;
7917 for(i=0;i<16;i++) {
7918 if (insn & (1 << i))
7919 n++;
7920 }
7921 /* XXX: test invalid n == 0 case ? */
7922 if (insn & (1 << 23)) {
7923 if (insn & (1 << 24)) {
7924 /* pre increment */
7925 tcg_gen_addi_i32(addr, addr, 4);
7926 } else {
7927 /* post increment */
7928 }
7929 } else {
7930 if (insn & (1 << 24)) {
7931 /* pre decrement */
7932 tcg_gen_addi_i32(addr, addr, -(n * 4));
7933 } else {
7934 /* post decrement */
7935 if (n != 1)
7936 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
7937 }
7938 }
7939 j = 0;
7940 for(i=0;i<16;i++) {
7941 if (insn & (1 << i)) {
7942 if (insn & (1 << 20)) {
7943 /* load */
7944 tmp = tcg_temp_new_i32();
7945 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7946 if (user) {
7947 tmp2 = tcg_const_i32(i);
7948 gen_helper_set_user_reg(cpu_env, tmp2, tmp);
7949 tcg_temp_free_i32(tmp2);
7950 tcg_temp_free_i32(tmp);
7951 } else if (i == rn) {
7952 loaded_var = tmp;
7953 loaded_base = 1;
7954 } else {
7955 store_reg_from_load(env, s, i, tmp);
7956 }
7957 } else {
7958 /* store */
7959 if (i == 15) {
7960 /* special case: r15 = PC + 8 */
7961 val = (long)s->pc + 4;
7962 tmp = tcg_temp_new_i32();
7963 tcg_gen_movi_i32(tmp, val);
7964 } else if (user) {
7965 tmp = tcg_temp_new_i32();
7966 tmp2 = tcg_const_i32(i);
7967 gen_helper_get_user_reg(tmp, cpu_env, tmp2);
7968 tcg_temp_free_i32(tmp2);
7969 } else {
7970 tmp = load_reg(s, i);
7971 }
7972 gen_aa32_st32(tmp, addr, IS_USER(s));
7973 tcg_temp_free_i32(tmp);
7974 }
7975 j++;
7976 /* no need to add after the last transfer */
7977 if (j != n)
7978 tcg_gen_addi_i32(addr, addr, 4);
7979 }
7980 }
7981 if (insn & (1 << 21)) {
7982 /* write back */
7983 if (insn & (1 << 23)) {
7984 if (insn & (1 << 24)) {
7985 /* pre increment */
7986 } else {
7987 /* post increment */
7988 tcg_gen_addi_i32(addr, addr, 4);
7989 }
7990 } else {
7991 if (insn & (1 << 24)) {
7992 /* pre decrement */
7993 if (n != 1)
7994 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
7995 } else {
7996 /* post decrement */
7997 tcg_gen_addi_i32(addr, addr, -(n * 4));
7998 }
7999 }
8000 store_reg(s, rn, addr);
8001 } else {
8002 tcg_temp_free_i32(addr);
8003 }
8004 if (loaded_base) {
8005 store_reg(s, rn, loaded_var);
8006 }
8007 if ((insn & (1 << 22)) && !user) {
8008 /* Restore CPSR from SPSR. */
8009 tmp = load_cpu_field(spsr);
8010 gen_set_cpsr(tmp, 0xffffffff);
8011 tcg_temp_free_i32(tmp);
8012 s->is_jmp = DISAS_UPDATE;
8013 }
8014 }
8015 break;
8016 case 0xa:
8017 case 0xb:
8018 {
8019 int32_t offset;
8020
8021 /* branch (and link) */
8022 val = (int32_t)s->pc;
8023 if (insn & (1 << 24)) {
8024 tmp = tcg_temp_new_i32();
8025 tcg_gen_movi_i32(tmp, val);
8026 store_reg(s, 14, tmp);
8027 }
8028 offset = sextract32(insn << 2, 0, 26);
8029 val += offset + 4;
8030 gen_jmp(s, val);
8031 }
8032 break;
8033 case 0xc:
8034 case 0xd:
8035 case 0xe:
8036 /* Coprocessor. */
8037 if (disas_coproc_insn(env, s, insn))
8038 goto illegal_op;
8039 break;
8040 case 0xf:
8041 /* swi */
8042 gen_set_pc_im(s, s->pc);
8043 s->is_jmp = DISAS_SWI;
8044 break;
8045 default:
8046 illegal_op:
8047 gen_exception_insn(s, 4, EXCP_UDEF);
8048 break;
8049 }
8050 }
8051 }
8052
8053 /* Return true if this is a Thumb-2 logical op. */
8054 static int
8055 thumb2_logic_op(int op)
8056 {
8057 return (op < 8);
8058 }
8059
8060 /* Generate code for a Thumb-2 data processing operation. If CONDS is nonzero
8061 then set condition code flags based on the result of the operation.
8062 If SHIFTER_OUT is nonzero then set the carry flag for logical operations
8063 to the high bit of T1.
8064 Returns zero if the opcode is valid. */
8065
8066 static int
8067 gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out,
8068 TCGv_i32 t0, TCGv_i32 t1)
8069 {
8070 int logic_cc;
8071
8072 logic_cc = 0;
8073 switch (op) {
8074 case 0: /* and */
8075 tcg_gen_and_i32(t0, t0, t1);
8076 logic_cc = conds;
8077 break;
8078 case 1: /* bic */
8079 tcg_gen_andc_i32(t0, t0, t1);
8080 logic_cc = conds;
8081 break;
8082 case 2: /* orr */
8083 tcg_gen_or_i32(t0, t0, t1);
8084 logic_cc = conds;
8085 break;
8086 case 3: /* orn */
8087 tcg_gen_orc_i32(t0, t0, t1);
8088 logic_cc = conds;
8089 break;
8090 case 4: /* eor */
8091 tcg_gen_xor_i32(t0, t0, t1);
8092 logic_cc = conds;
8093 break;
8094 case 8: /* add */
8095 if (conds)
8096 gen_add_CC(t0, t0, t1);
8097 else
8098 tcg_gen_add_i32(t0, t0, t1);
8099 break;
8100 case 10: /* adc */
8101 if (conds)
8102 gen_adc_CC(t0, t0, t1);
8103 else
8104 gen_adc(t0, t1);
8105 break;
8106 case 11: /* sbc */
8107 if (conds) {
8108 gen_sbc_CC(t0, t0, t1);
8109 } else {
8110 gen_sub_carry(t0, t0, t1);
8111 }
8112 break;
8113 case 13: /* sub */
8114 if (conds)
8115 gen_sub_CC(t0, t0, t1);
8116 else
8117 tcg_gen_sub_i32(t0, t0, t1);
8118 break;
8119 case 14: /* rsb */
8120 if (conds)
8121 gen_sub_CC(t0, t1, t0);
8122 else
8123 tcg_gen_sub_i32(t0, t1, t0);
8124 break;
8125 default: /* 5, 6, 7, 9, 12, 15. */
8126 return 1;
8127 }
8128 if (logic_cc) {
8129 gen_logic_CC(t0);
8130 if (shifter_out)
8131 gen_set_CF_bit31(t1);
8132 }
8133 return 0;
8134 }
8135
8136 /* Translate a 32-bit thumb instruction. Returns nonzero if the instruction
8137 is not legal. */
8138 static int disas_thumb2_insn(CPUARMState *env, DisasContext *s, uint16_t insn_hw1)
8139 {
8140 uint32_t insn, imm, shift, offset;
8141 uint32_t rd, rn, rm, rs;
8142 TCGv_i32 tmp;
8143 TCGv_i32 tmp2;
8144 TCGv_i32 tmp3;
8145 TCGv_i32 addr;
8146 TCGv_i64 tmp64;
8147 int op;
8148 int shiftop;
8149 int conds;
8150 int logic_cc;
8151
8152 if (!(arm_feature(env, ARM_FEATURE_THUMB2)
8153 || arm_feature (env, ARM_FEATURE_M))) {
8154 /* Thumb-1 cores may need to treat bl and blx as a pair of
8155 16-bit instructions to get correct prefetch abort behavior. */
8156 insn = insn_hw1;
8157 if ((insn & (1 << 12)) == 0) {
8158 ARCH(5);
8159 /* Second half of blx. */
8160 offset = ((insn & 0x7ff) << 1);
8161 tmp = load_reg(s, 14);
8162 tcg_gen_addi_i32(tmp, tmp, offset);
8163 tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
8164
8165 tmp2 = tcg_temp_new_i32();
8166 tcg_gen_movi_i32(tmp2, s->pc | 1);
8167 store_reg(s, 14, tmp2);
8168 gen_bx(s, tmp);
8169 return 0;
8170 }
8171 if (insn & (1 << 11)) {
8172 /* Second half of bl. */
8173 offset = ((insn & 0x7ff) << 1) | 1;
8174 tmp = load_reg(s, 14);
8175 tcg_gen_addi_i32(tmp, tmp, offset);
8176
8177 tmp2 = tcg_temp_new_i32();
8178 tcg_gen_movi_i32(tmp2, s->pc | 1);
8179 store_reg(s, 14, tmp2);
8180 gen_bx(s, tmp);
8181 return 0;
8182 }
8183 if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
8184 /* Instruction spans a page boundary. Implement it as two
8185 16-bit instructions in case the second half causes an
8186 prefetch abort. */
8187 offset = ((int32_t)insn << 21) >> 9;
8188 tcg_gen_movi_i32(cpu_R[14], s->pc + 2 + offset);
8189 return 0;
8190 }
8191 /* Fall through to 32-bit decode. */
8192 }
8193
8194 insn = arm_lduw_code(env, s->pc, s->bswap_code);
8195 s->pc += 2;
8196 insn |= (uint32_t)insn_hw1 << 16;
8197
8198 if ((insn & 0xf800e800) != 0xf000e800) {
8199 ARCH(6T2);
8200 }
8201
8202 rn = (insn >> 16) & 0xf;
8203 rs = (insn >> 12) & 0xf;
8204 rd = (insn >> 8) & 0xf;
8205 rm = insn & 0xf;
8206 switch ((insn >> 25) & 0xf) {
8207 case 0: case 1: case 2: case 3:
8208 /* 16-bit instructions. Should never happen. */
8209 abort();
8210 case 4:
8211 if (insn & (1 << 22)) {
8212 /* Other load/store, table branch. */
8213 if (insn & 0x01200000) {
8214 /* Load/store doubleword. */
8215 if (rn == 15) {
8216 addr = tcg_temp_new_i32();
8217 tcg_gen_movi_i32(addr, s->pc & ~3);
8218 } else {
8219 addr = load_reg(s, rn);
8220 }
8221 offset = (insn & 0xff) * 4;
8222 if ((insn & (1 << 23)) == 0)
8223 offset = -offset;
8224 if (insn & (1 << 24)) {
8225 tcg_gen_addi_i32(addr, addr, offset);
8226 offset = 0;
8227 }
8228 if (insn & (1 << 20)) {
8229 /* ldrd */
8230 tmp = tcg_temp_new_i32();
8231 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8232 store_reg(s, rs, tmp);
8233 tcg_gen_addi_i32(addr, addr, 4);
8234 tmp = tcg_temp_new_i32();
8235 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8236 store_reg(s, rd, tmp);
8237 } else {
8238 /* strd */
8239 tmp = load_reg(s, rs);
8240 gen_aa32_st32(tmp, addr, IS_USER(s));
8241 tcg_temp_free_i32(tmp);
8242 tcg_gen_addi_i32(addr, addr, 4);
8243 tmp = load_reg(s, rd);
8244 gen_aa32_st32(tmp, addr, IS_USER(s));
8245 tcg_temp_free_i32(tmp);
8246 }
8247 if (insn & (1 << 21)) {
8248 /* Base writeback. */
8249 if (rn == 15)
8250 goto illegal_op;
8251 tcg_gen_addi_i32(addr, addr, offset - 4);
8252 store_reg(s, rn, addr);
8253 } else {
8254 tcg_temp_free_i32(addr);
8255 }
8256 } else if ((insn & (1 << 23)) == 0) {
8257 /* Load/store exclusive word. */
8258 addr = tcg_temp_local_new_i32();
8259 load_reg_var(s, addr, rn);
8260 tcg_gen_addi_i32(addr, addr, (insn & 0xff) << 2);
8261 if (insn & (1 << 20)) {
8262 gen_load_exclusive(s, rs, 15, addr, 2);
8263 } else {
8264 gen_store_exclusive(s, rd, rs, 15, addr, 2);
8265 }
8266 tcg_temp_free_i32(addr);
8267 } else if ((insn & (7 << 5)) == 0) {
8268 /* Table Branch. */
8269 if (rn == 15) {
8270 addr = tcg_temp_new_i32();
8271 tcg_gen_movi_i32(addr, s->pc);
8272 } else {
8273 addr = load_reg(s, rn);
8274 }
8275 tmp = load_reg(s, rm);
8276 tcg_gen_add_i32(addr, addr, tmp);
8277 if (insn & (1 << 4)) {
8278 /* tbh */
8279 tcg_gen_add_i32(addr, addr, tmp);
8280 tcg_temp_free_i32(tmp);
8281 tmp = tcg_temp_new_i32();
8282 gen_aa32_ld16u(tmp, addr, IS_USER(s));
8283 } else { /* tbb */
8284 tcg_temp_free_i32(tmp);
8285 tmp = tcg_temp_new_i32();
8286 gen_aa32_ld8u(tmp, addr, IS_USER(s));
8287 }
8288 tcg_temp_free_i32(addr);
8289 tcg_gen_shli_i32(tmp, tmp, 1);
8290 tcg_gen_addi_i32(tmp, tmp, s->pc);
8291 store_reg(s, 15, tmp);
8292 } else {
8293 int op2 = (insn >> 6) & 0x3;
8294 op = (insn >> 4) & 0x3;
8295 switch (op2) {
8296 case 0:
8297 goto illegal_op;
8298 case 1:
8299 /* Load/store exclusive byte/halfword/doubleword */
8300 if (op == 2) {
8301 goto illegal_op;
8302 }
8303 ARCH(7);
8304 break;
8305 case 2:
8306 /* Load-acquire/store-release */
8307 if (op == 3) {
8308 goto illegal_op;
8309 }
8310 /* Fall through */
8311 case 3:
8312 /* Load-acquire/store-release exclusive */
8313 ARCH(8);
8314 break;
8315 }
8316 addr = tcg_temp_local_new_i32();
8317 load_reg_var(s, addr, rn);
8318 if (!(op2 & 1)) {
8319 if (insn & (1 << 20)) {
8320 tmp = tcg_temp_new_i32();
8321 switch (op) {
8322 case 0: /* ldab */
8323 gen_aa32_ld8u(tmp, addr, IS_USER(s));
8324 break;
8325 case 1: /* ldah */
8326 gen_aa32_ld16u(tmp, addr, IS_USER(s));
8327 break;
8328 case 2: /* lda */
8329 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8330 break;
8331 default:
8332 abort();
8333 }
8334 store_reg(s, rs, tmp);
8335 } else {
8336 tmp = load_reg(s, rs);
8337 switch (op) {
8338 case 0: /* stlb */
8339 gen_aa32_st8(tmp, addr, IS_USER(s));
8340 break;
8341 case 1: /* stlh */
8342 gen_aa32_st16(tmp, addr, IS_USER(s));
8343 break;
8344 case 2: /* stl */
8345 gen_aa32_st32(tmp, addr, IS_USER(s));
8346 break;
8347 default:
8348 abort();
8349 }
8350 tcg_temp_free_i32(tmp);
8351 }
8352 } else if (insn & (1 << 20)) {
8353 gen_load_exclusive(s, rs, rd, addr, op);
8354 } else {
8355 gen_store_exclusive(s, rm, rs, rd, addr, op);
8356 }
8357 tcg_temp_free_i32(addr);
8358 }
8359 } else {
8360 /* Load/store multiple, RFE, SRS. */
8361 if (((insn >> 23) & 1) == ((insn >> 24) & 1)) {
8362 /* RFE, SRS: not available in user mode or on M profile */
8363 if (IS_USER(s) || IS_M(env)) {
8364 goto illegal_op;
8365 }
8366 if (insn & (1 << 20)) {
8367 /* rfe */
8368 addr = load_reg(s, rn);
8369 if ((insn & (1 << 24)) == 0)
8370 tcg_gen_addi_i32(addr, addr, -8);
8371 /* Load PC into tmp and CPSR into tmp2. */
8372 tmp = tcg_temp_new_i32();
8373 gen_aa32_ld32u(tmp, addr, 0);
8374 tcg_gen_addi_i32(addr, addr, 4);
8375 tmp2 = tcg_temp_new_i32();
8376 gen_aa32_ld32u(tmp2, addr, 0);
8377 if (insn & (1 << 21)) {
8378 /* Base writeback. */
8379 if (insn & (1 << 24)) {
8380 tcg_gen_addi_i32(addr, addr, 4);
8381 } else {
8382 tcg_gen_addi_i32(addr, addr, -4);
8383 }
8384 store_reg(s, rn, addr);
8385 } else {
8386 tcg_temp_free_i32(addr);
8387 }
8388 gen_rfe(s, tmp, tmp2);
8389 } else {
8390 /* srs */
8391 gen_srs(s, (insn & 0x1f), (insn & (1 << 24)) ? 1 : 2,
8392 insn & (1 << 21));
8393 }
8394 } else {
8395 int i, loaded_base = 0;
8396 TCGv_i32 loaded_var;
8397 /* Load/store multiple. */
8398 addr = load_reg(s, rn);
8399 offset = 0;
8400 for (i = 0; i < 16; i++) {
8401 if (insn & (1 << i))
8402 offset += 4;
8403 }
8404 if (insn & (1 << 24)) {
8405 tcg_gen_addi_i32(addr, addr, -offset);
8406 }
8407
8408 TCGV_UNUSED_I32(loaded_var);
8409 for (i = 0; i < 16; i++) {
8410 if ((insn & (1 << i)) == 0)
8411 continue;
8412 if (insn & (1 << 20)) {
8413 /* Load. */
8414 tmp = tcg_temp_new_i32();
8415 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8416 if (i == 15) {
8417 gen_bx(s, tmp);
8418 } else if (i == rn) {
8419 loaded_var = tmp;
8420 loaded_base = 1;
8421 } else {
8422 store_reg(s, i, tmp);
8423 }
8424 } else {
8425 /* Store. */
8426 tmp = load_reg(s, i);
8427 gen_aa32_st32(tmp, addr, IS_USER(s));
8428 tcg_temp_free_i32(tmp);
8429 }
8430 tcg_gen_addi_i32(addr, addr, 4);
8431 }
8432 if (loaded_base) {
8433 store_reg(s, rn, loaded_var);
8434 }
8435 if (insn & (1 << 21)) {
8436 /* Base register writeback. */
8437 if (insn & (1 << 24)) {
8438 tcg_gen_addi_i32(addr, addr, -offset);
8439 }
8440 /* Fault if writeback register is in register list. */
8441 if (insn & (1 << rn))
8442 goto illegal_op;
8443 store_reg(s, rn, addr);
8444 } else {
8445 tcg_temp_free_i32(addr);
8446 }
8447 }
8448 }
8449 break;
8450 case 5:
8451
8452 op = (insn >> 21) & 0xf;
8453 if (op == 6) {
8454 /* Halfword pack. */
8455 tmp = load_reg(s, rn);
8456 tmp2 = load_reg(s, rm);
8457 shift = ((insn >> 10) & 0x1c) | ((insn >> 6) & 0x3);
8458 if (insn & (1 << 5)) {
8459 /* pkhtb */
8460 if (shift == 0)
8461 shift = 31;
8462 tcg_gen_sari_i32(tmp2, tmp2, shift);
8463 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
8464 tcg_gen_ext16u_i32(tmp2, tmp2);
8465 } else {
8466 /* pkhbt */
8467 if (shift)
8468 tcg_gen_shli_i32(tmp2, tmp2, shift);
8469 tcg_gen_ext16u_i32(tmp, tmp);
8470 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
8471 }
8472 tcg_gen_or_i32(tmp, tmp, tmp2);
8473 tcg_temp_free_i32(tmp2);
8474 store_reg(s, rd, tmp);
8475 } else {
8476 /* Data processing register constant shift. */
8477 if (rn == 15) {
8478 tmp = tcg_temp_new_i32();
8479 tcg_gen_movi_i32(tmp, 0);
8480 } else {
8481 tmp = load_reg(s, rn);
8482 }
8483 tmp2 = load_reg(s, rm);
8484
8485 shiftop = (insn >> 4) & 3;
8486 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
8487 conds = (insn & (1 << 20)) != 0;
8488 logic_cc = (conds && thumb2_logic_op(op));
8489 gen_arm_shift_im(tmp2, shiftop, shift, logic_cc);
8490 if (gen_thumb2_data_op(s, op, conds, 0, tmp, tmp2))
8491 goto illegal_op;
8492 tcg_temp_free_i32(tmp2);
8493 if (rd != 15) {
8494 store_reg(s, rd, tmp);
8495 } else {
8496 tcg_temp_free_i32(tmp);
8497 }
8498 }
8499 break;
8500 case 13: /* Misc data processing. */
8501 op = ((insn >> 22) & 6) | ((insn >> 7) & 1);
8502 if (op < 4 && (insn & 0xf000) != 0xf000)
8503 goto illegal_op;
8504 switch (op) {
8505 case 0: /* Register controlled shift. */
8506 tmp = load_reg(s, rn);
8507 tmp2 = load_reg(s, rm);
8508 if ((insn & 0x70) != 0)
8509 goto illegal_op;
8510 op = (insn >> 21) & 3;
8511 logic_cc = (insn & (1 << 20)) != 0;
8512 gen_arm_shift_reg(tmp, op, tmp2, logic_cc);
8513 if (logic_cc)
8514 gen_logic_CC(tmp);
8515 store_reg_bx(env, s, rd, tmp);
8516 break;
8517 case 1: /* Sign/zero extend. */
8518 tmp = load_reg(s, rm);
8519 shift = (insn >> 4) & 3;
8520 /* ??? In many cases it's not necessary to do a
8521 rotate, a shift is sufficient. */
8522 if (shift != 0)
8523 tcg_gen_rotri_i32(tmp, tmp, shift * 8);
8524 op = (insn >> 20) & 7;
8525 switch (op) {
8526 case 0: gen_sxth(tmp); break;
8527 case 1: gen_uxth(tmp); break;
8528 case 2: gen_sxtb16(tmp); break;
8529 case 3: gen_uxtb16(tmp); break;
8530 case 4: gen_sxtb(tmp); break;
8531 case 5: gen_uxtb(tmp); break;
8532 default: goto illegal_op;
8533 }
8534 if (rn != 15) {
8535 tmp2 = load_reg(s, rn);
8536 if ((op >> 1) == 1) {
8537 gen_add16(tmp, tmp2);
8538 } else {
8539 tcg_gen_add_i32(tmp, tmp, tmp2);
8540 tcg_temp_free_i32(tmp2);
8541 }
8542 }
8543 store_reg(s, rd, tmp);
8544 break;
8545 case 2: /* SIMD add/subtract. */
8546 op = (insn >> 20) & 7;
8547 shift = (insn >> 4) & 7;
8548 if ((op & 3) == 3 || (shift & 3) == 3)
8549 goto illegal_op;
8550 tmp = load_reg(s, rn);
8551 tmp2 = load_reg(s, rm);
8552 gen_thumb2_parallel_addsub(op, shift, tmp, tmp2);
8553 tcg_temp_free_i32(tmp2);
8554 store_reg(s, rd, tmp);
8555 break;
8556 case 3: /* Other data processing. */
8557 op = ((insn >> 17) & 0x38) | ((insn >> 4) & 7);
8558 if (op < 4) {
8559 /* Saturating add/subtract. */
8560 tmp = load_reg(s, rn);
8561 tmp2 = load_reg(s, rm);
8562 if (op & 1)
8563 gen_helper_double_saturate(tmp, cpu_env, tmp);
8564 if (op & 2)
8565 gen_helper_sub_saturate(tmp, cpu_env, tmp2, tmp);
8566 else
8567 gen_helper_add_saturate(tmp, cpu_env, tmp, tmp2);
8568 tcg_temp_free_i32(tmp2);
8569 } else {
8570 tmp = load_reg(s, rn);
8571 switch (op) {
8572 case 0x0a: /* rbit */
8573 gen_helper_rbit(tmp, tmp);
8574 break;
8575 case 0x08: /* rev */
8576 tcg_gen_bswap32_i32(tmp, tmp);
8577 break;
8578 case 0x09: /* rev16 */
8579 gen_rev16(tmp);
8580 break;
8581 case 0x0b: /* revsh */
8582 gen_revsh(tmp);
8583 break;
8584 case 0x10: /* sel */
8585 tmp2 = load_reg(s, rm);
8586 tmp3 = tcg_temp_new_i32();
8587 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUARMState, GE));
8588 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
8589 tcg_temp_free_i32(tmp3);
8590 tcg_temp_free_i32(tmp2);
8591 break;
8592 case 0x18: /* clz */
8593 gen_helper_clz(tmp, tmp);
8594 break;
8595 default:
8596 goto illegal_op;
8597 }
8598 }
8599 store_reg(s, rd, tmp);
8600 break;
8601 case 4: case 5: /* 32-bit multiply. Sum of absolute differences. */
8602 op = (insn >> 4) & 0xf;
8603 tmp = load_reg(s, rn);
8604 tmp2 = load_reg(s, rm);
8605 switch ((insn >> 20) & 7) {
8606 case 0: /* 32 x 32 -> 32 */
8607 tcg_gen_mul_i32(tmp, tmp, tmp2);
8608 tcg_temp_free_i32(tmp2);
8609 if (rs != 15) {
8610 tmp2 = load_reg(s, rs);
8611 if (op)
8612 tcg_gen_sub_i32(tmp, tmp2, tmp);
8613 else
8614 tcg_gen_add_i32(tmp, tmp, tmp2);
8615 tcg_temp_free_i32(tmp2);
8616 }
8617 break;
8618 case 1: /* 16 x 16 -> 32 */
8619 gen_mulxy(tmp, tmp2, op & 2, op & 1);
8620 tcg_temp_free_i32(tmp2);
8621 if (rs != 15) {
8622 tmp2 = load_reg(s, rs);
8623 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
8624 tcg_temp_free_i32(tmp2);
8625 }
8626 break;
8627 case 2: /* Dual multiply add. */
8628 case 4: /* Dual multiply subtract. */
8629 if (op)
8630 gen_swap_half(tmp2);
8631 gen_smul_dual(tmp, tmp2);
8632 if (insn & (1 << 22)) {
8633 /* This subtraction cannot overflow. */
8634 tcg_gen_sub_i32(tmp, tmp, tmp2);
8635 } else {
8636 /* This addition cannot overflow 32 bits;
8637 * however it may overflow considered as a signed
8638 * operation, in which case we must set the Q flag.
8639 */
8640 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
8641 }
8642 tcg_temp_free_i32(tmp2);
8643 if (rs != 15)
8644 {
8645 tmp2 = load_reg(s, rs);
8646 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
8647 tcg_temp_free_i32(tmp2);
8648 }
8649 break;
8650 case 3: /* 32 * 16 -> 32msb */
8651 if (op)
8652 tcg_gen_sari_i32(tmp2, tmp2, 16);
8653 else
8654 gen_sxth(tmp2);
8655 tmp64 = gen_muls_i64_i32(tmp, tmp2);
8656 tcg_gen_shri_i64(tmp64, tmp64, 16);
8657 tmp = tcg_temp_new_i32();
8658 tcg_gen_trunc_i64_i32(tmp, tmp64);
8659 tcg_temp_free_i64(tmp64);
8660 if (rs != 15)
8661 {
8662 tmp2 = load_reg(s, rs);
8663 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
8664 tcg_temp_free_i32(tmp2);
8665 }
8666 break;
8667 case 5: case 6: /* 32 * 32 -> 32msb (SMMUL, SMMLA, SMMLS) */
8668 tmp64 = gen_muls_i64_i32(tmp, tmp2);
8669 if (rs != 15) {
8670 tmp = load_reg(s, rs);
8671 if (insn & (1 << 20)) {
8672 tmp64 = gen_addq_msw(tmp64, tmp);
8673 } else {
8674 tmp64 = gen_subq_msw(tmp64, tmp);
8675 }
8676 }
8677 if (insn & (1 << 4)) {
8678 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
8679 }
8680 tcg_gen_shri_i64(tmp64, tmp64, 32);
8681 tmp = tcg_temp_new_i32();
8682 tcg_gen_trunc_i64_i32(tmp, tmp64);
8683 tcg_temp_free_i64(tmp64);
8684 break;
8685 case 7: /* Unsigned sum of absolute differences. */
8686 gen_helper_usad8(tmp, tmp, tmp2);
8687 tcg_temp_free_i32(tmp2);
8688 if (rs != 15) {
8689 tmp2 = load_reg(s, rs);
8690 tcg_gen_add_i32(tmp, tmp, tmp2);
8691 tcg_temp_free_i32(tmp2);
8692 }
8693 break;
8694 }
8695 store_reg(s, rd, tmp);
8696 break;
8697 case 6: case 7: /* 64-bit multiply, Divide. */
8698 op = ((insn >> 4) & 0xf) | ((insn >> 16) & 0x70);
8699 tmp = load_reg(s, rn);
8700 tmp2 = load_reg(s, rm);
8701 if ((op & 0x50) == 0x10) {
8702 /* sdiv, udiv */
8703 if (!arm_feature(env, ARM_FEATURE_THUMB_DIV)) {
8704 goto illegal_op;
8705 }
8706 if (op & 0x20)
8707 gen_helper_udiv(tmp, tmp, tmp2);
8708 else
8709 gen_helper_sdiv(tmp, tmp, tmp2);
8710 tcg_temp_free_i32(tmp2);
8711 store_reg(s, rd, tmp);
8712 } else if ((op & 0xe) == 0xc) {
8713 /* Dual multiply accumulate long. */
8714 if (op & 1)
8715 gen_swap_half(tmp2);
8716 gen_smul_dual(tmp, tmp2);
8717 if (op & 0x10) {
8718 tcg_gen_sub_i32(tmp, tmp, tmp2);
8719 } else {
8720 tcg_gen_add_i32(tmp, tmp, tmp2);
8721 }
8722 tcg_temp_free_i32(tmp2);
8723 /* BUGFIX */
8724 tmp64 = tcg_temp_new_i64();
8725 tcg_gen_ext_i32_i64(tmp64, tmp);
8726 tcg_temp_free_i32(tmp);
8727 gen_addq(s, tmp64, rs, rd);
8728 gen_storeq_reg(s, rs, rd, tmp64);
8729 tcg_temp_free_i64(tmp64);
8730 } else {
8731 if (op & 0x20) {
8732 /* Unsigned 64-bit multiply */
8733 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
8734 } else {
8735 if (op & 8) {
8736 /* smlalxy */
8737 gen_mulxy(tmp, tmp2, op & 2, op & 1);
8738 tcg_temp_free_i32(tmp2);
8739 tmp64 = tcg_temp_new_i64();
8740 tcg_gen_ext_i32_i64(tmp64, tmp);
8741 tcg_temp_free_i32(tmp);
8742 } else {
8743 /* Signed 64-bit multiply */
8744 tmp64 = gen_muls_i64_i32(tmp, tmp2);
8745 }
8746 }
8747 if (op & 4) {
8748 /* umaal */
8749 gen_addq_lo(s, tmp64, rs);
8750 gen_addq_lo(s, tmp64, rd);
8751 } else if (op & 0x40) {
8752 /* 64-bit accumulate. */
8753 gen_addq(s, tmp64, rs, rd);
8754 }
8755 gen_storeq_reg(s, rs, rd, tmp64);
8756 tcg_temp_free_i64(tmp64);
8757 }
8758 break;
8759 }
8760 break;
8761 case 6: case 7: case 14: case 15:
8762 /* Coprocessor. */
8763 if (((insn >> 24) & 3) == 3) {
8764 /* Translate into the equivalent ARM encoding. */
8765 insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4) | (1 << 28);
8766 if (disas_neon_data_insn(env, s, insn))
8767 goto illegal_op;
8768 } else {
8769 if (insn & (1 << 28))
8770 goto illegal_op;
8771 if (disas_coproc_insn (env, s, insn))
8772 goto illegal_op;
8773 }
8774 break;
8775 case 8: case 9: case 10: case 11:
8776 if (insn & (1 << 15)) {
8777 /* Branches, misc control. */
8778 if (insn & 0x5000) {
8779 /* Unconditional branch. */
8780 /* signextend(hw1[10:0]) -> offset[:12]. */
8781 offset = ((int32_t)insn << 5) >> 9 & ~(int32_t)0xfff;
8782 /* hw1[10:0] -> offset[11:1]. */
8783 offset |= (insn & 0x7ff) << 1;
8784 /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
8785 offset[24:22] already have the same value because of the
8786 sign extension above. */
8787 offset ^= ((~insn) & (1 << 13)) << 10;
8788 offset ^= ((~insn) & (1 << 11)) << 11;
8789
8790 if (insn & (1 << 14)) {
8791 /* Branch and link. */
8792 tcg_gen_movi_i32(cpu_R[14], s->pc | 1);
8793 }
8794
8795 offset += s->pc;
8796 if (insn & (1 << 12)) {
8797 /* b/bl */
8798 gen_jmp(s, offset);
8799 } else {
8800 /* blx */
8801 offset &= ~(uint32_t)2;
8802 /* thumb2 bx, no need to check */
8803 gen_bx_im(s, offset);
8804 }
8805 } else if (((insn >> 23) & 7) == 7) {
8806 /* Misc control */
8807 if (insn & (1 << 13))
8808 goto illegal_op;
8809
8810 if (insn & (1 << 26)) {
8811 /* Secure monitor call (v6Z) */
8812 qemu_log_mask(LOG_UNIMP,
8813 "arm: unimplemented secure monitor call\n");
8814 goto illegal_op; /* not implemented. */
8815 } else {
8816 op = (insn >> 20) & 7;
8817 switch (op) {
8818 case 0: /* msr cpsr. */
8819 if (IS_M(env)) {
8820 tmp = load_reg(s, rn);
8821 addr = tcg_const_i32(insn & 0xff);
8822 gen_helper_v7m_msr(cpu_env, addr, tmp);
8823 tcg_temp_free_i32(addr);
8824 tcg_temp_free_i32(tmp);
8825 gen_lookup_tb(s);
8826 break;
8827 }
8828 /* fall through */
8829 case 1: /* msr spsr. */
8830 if (IS_M(env))
8831 goto illegal_op;
8832 tmp = load_reg(s, rn);
8833 if (gen_set_psr(s,
8834 msr_mask(env, s, (insn >> 8) & 0xf, op == 1),
8835 op == 1, tmp))
8836 goto illegal_op;
8837 break;
8838 case 2: /* cps, nop-hint. */
8839 if (((insn >> 8) & 7) == 0) {
8840 gen_nop_hint(s, insn & 0xff);
8841 }
8842 /* Implemented as NOP in user mode. */
8843 if (IS_USER(s))
8844 break;
8845 offset = 0;
8846 imm = 0;
8847 if (insn & (1 << 10)) {
8848 if (insn & (1 << 7))
8849 offset |= CPSR_A;
8850 if (insn & (1 << 6))
8851 offset |= CPSR_I;
8852 if (insn & (1 << 5))
8853 offset |= CPSR_F;
8854 if (insn & (1 << 9))
8855 imm = CPSR_A | CPSR_I | CPSR_F;
8856 }
8857 if (insn & (1 << 8)) {
8858 offset |= 0x1f;
8859 imm |= (insn & 0x1f);
8860 }
8861 if (offset) {
8862 gen_set_psr_im(s, offset, 0, imm);
8863 }
8864 break;
8865 case 3: /* Special control operations. */
8866 ARCH(7);
8867 op = (insn >> 4) & 0xf;
8868 switch (op) {
8869 case 2: /* clrex */
8870 gen_clrex(s);
8871 break;
8872 case 4: /* dsb */
8873 case 5: /* dmb */
8874 case 6: /* isb */
8875 /* These execute as NOPs. */
8876 break;
8877 default:
8878 goto illegal_op;
8879 }
8880 break;
8881 case 4: /* bxj */
8882 /* Trivial implementation equivalent to bx. */
8883 tmp = load_reg(s, rn);
8884 gen_bx(s, tmp);
8885 break;
8886 case 5: /* Exception return. */
8887 if (IS_USER(s)) {
8888 goto illegal_op;
8889 }
8890 if (rn != 14 || rd != 15) {
8891 goto illegal_op;
8892 }
8893 tmp = load_reg(s, rn);
8894 tcg_gen_subi_i32(tmp, tmp, insn & 0xff);
8895 gen_exception_return(s, tmp);
8896 break;
8897 case 6: /* mrs cpsr. */
8898 tmp = tcg_temp_new_i32();
8899 if (IS_M(env)) {
8900 addr = tcg_const_i32(insn & 0xff);
8901 gen_helper_v7m_mrs(tmp, cpu_env, addr);
8902 tcg_temp_free_i32(addr);
8903 } else {
8904 gen_helper_cpsr_read(tmp, cpu_env);
8905 }
8906 store_reg(s, rd, tmp);
8907 break;
8908 case 7: /* mrs spsr. */
8909 /* Not accessible in user mode. */
8910 if (IS_USER(s) || IS_M(env))
8911 goto illegal_op;
8912 tmp = load_cpu_field(spsr);
8913 store_reg(s, rd, tmp);
8914 break;
8915 }
8916 }
8917 } else {
8918 /* Conditional branch. */
8919 op = (insn >> 22) & 0xf;
8920 /* Generate a conditional jump to next instruction. */
8921 s->condlabel = gen_new_label();
8922 gen_test_cc(op ^ 1, s->condlabel);
8923 s->condjmp = 1;
8924
8925 /* offset[11:1] = insn[10:0] */
8926 offset = (insn & 0x7ff) << 1;
8927 /* offset[17:12] = insn[21:16]. */
8928 offset |= (insn & 0x003f0000) >> 4;
8929 /* offset[31:20] = insn[26]. */
8930 offset |= ((int32_t)((insn << 5) & 0x80000000)) >> 11;
8931 /* offset[18] = insn[13]. */
8932 offset |= (insn & (1 << 13)) << 5;
8933 /* offset[19] = insn[11]. */
8934 offset |= (insn & (1 << 11)) << 8;
8935
8936 /* jump to the offset */
8937 gen_jmp(s, s->pc + offset);
8938 }
8939 } else {
8940 /* Data processing immediate. */
8941 if (insn & (1 << 25)) {
8942 if (insn & (1 << 24)) {
8943 if (insn & (1 << 20))
8944 goto illegal_op;
8945 /* Bitfield/Saturate. */
8946 op = (insn >> 21) & 7;
8947 imm = insn & 0x1f;
8948 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
8949 if (rn == 15) {
8950 tmp = tcg_temp_new_i32();
8951 tcg_gen_movi_i32(tmp, 0);
8952 } else {
8953 tmp = load_reg(s, rn);
8954 }
8955 switch (op) {
8956 case 2: /* Signed bitfield extract. */
8957 imm++;
8958 if (shift + imm > 32)
8959 goto illegal_op;
8960 if (imm < 32)
8961 gen_sbfx(tmp, shift, imm);
8962 break;
8963 case 6: /* Unsigned bitfield extract. */
8964 imm++;
8965 if (shift + imm > 32)
8966 goto illegal_op;
8967 if (imm < 32)
8968 gen_ubfx(tmp, shift, (1u << imm) - 1);
8969 break;
8970 case 3: /* Bitfield insert/clear. */
8971 if (imm < shift)
8972 goto illegal_op;
8973 imm = imm + 1 - shift;
8974 if (imm != 32) {
8975 tmp2 = load_reg(s, rd);
8976 tcg_gen_deposit_i32(tmp, tmp2, tmp, shift, imm);
8977 tcg_temp_free_i32(tmp2);
8978 }
8979 break;
8980 case 7:
8981 goto illegal_op;
8982 default: /* Saturate. */
8983 if (shift) {
8984 if (op & 1)
8985 tcg_gen_sari_i32(tmp, tmp, shift);
8986 else
8987 tcg_gen_shli_i32(tmp, tmp, shift);
8988 }
8989 tmp2 = tcg_const_i32(imm);
8990 if (op & 4) {
8991 /* Unsigned. */
8992 if ((op & 1) && shift == 0)
8993 gen_helper_usat16(tmp, cpu_env, tmp, tmp2);
8994 else
8995 gen_helper_usat(tmp, cpu_env, tmp, tmp2);
8996 } else {
8997 /* Signed. */
8998 if ((op & 1) && shift == 0)
8999 gen_helper_ssat16(tmp, cpu_env, tmp, tmp2);
9000 else
9001 gen_helper_ssat(tmp, cpu_env, tmp, tmp2);
9002 }
9003 tcg_temp_free_i32(tmp2);
9004 break;
9005 }
9006 store_reg(s, rd, tmp);
9007 } else {
9008 imm = ((insn & 0x04000000) >> 15)
9009 | ((insn & 0x7000) >> 4) | (insn & 0xff);
9010 if (insn & (1 << 22)) {
9011 /* 16-bit immediate. */
9012 imm |= (insn >> 4) & 0xf000;
9013 if (insn & (1 << 23)) {
9014 /* movt */
9015 tmp = load_reg(s, rd);
9016 tcg_gen_ext16u_i32(tmp, tmp);
9017 tcg_gen_ori_i32(tmp, tmp, imm << 16);
9018 } else {
9019 /* movw */
9020 tmp = tcg_temp_new_i32();
9021 tcg_gen_movi_i32(tmp, imm);
9022 }
9023 } else {
9024 /* Add/sub 12-bit immediate. */
9025 if (rn == 15) {
9026 offset = s->pc & ~(uint32_t)3;
9027 if (insn & (1 << 23))
9028 offset -= imm;
9029 else
9030 offset += imm;
9031 tmp = tcg_temp_new_i32();
9032 tcg_gen_movi_i32(tmp, offset);
9033 } else {
9034 tmp = load_reg(s, rn);
9035 if (insn & (1 << 23))
9036 tcg_gen_subi_i32(tmp, tmp, imm);
9037 else
9038 tcg_gen_addi_i32(tmp, tmp, imm);
9039 }
9040 }
9041 store_reg(s, rd, tmp);
9042 }
9043 } else {
9044 int shifter_out = 0;
9045 /* modified 12-bit immediate. */
9046 shift = ((insn & 0x04000000) >> 23) | ((insn & 0x7000) >> 12);
9047 imm = (insn & 0xff);
9048 switch (shift) {
9049 case 0: /* XY */
9050 /* Nothing to do. */
9051 break;
9052 case 1: /* 00XY00XY */
9053 imm |= imm << 16;
9054 break;
9055 case 2: /* XY00XY00 */
9056 imm |= imm << 16;
9057 imm <<= 8;
9058 break;
9059 case 3: /* XYXYXYXY */
9060 imm |= imm << 16;
9061 imm |= imm << 8;
9062 break;
9063 default: /* Rotated constant. */
9064 shift = (shift << 1) | (imm >> 7);
9065 imm |= 0x80;
9066 imm = imm << (32 - shift);
9067 shifter_out = 1;
9068 break;
9069 }
9070 tmp2 = tcg_temp_new_i32();
9071 tcg_gen_movi_i32(tmp2, imm);
9072 rn = (insn >> 16) & 0xf;
9073 if (rn == 15) {
9074 tmp = tcg_temp_new_i32();
9075 tcg_gen_movi_i32(tmp, 0);
9076 } else {
9077 tmp = load_reg(s, rn);
9078 }
9079 op = (insn >> 21) & 0xf;
9080 if (gen_thumb2_data_op(s, op, (insn & (1 << 20)) != 0,
9081 shifter_out, tmp, tmp2))
9082 goto illegal_op;
9083 tcg_temp_free_i32(tmp2);
9084 rd = (insn >> 8) & 0xf;
9085 if (rd != 15) {
9086 store_reg(s, rd, tmp);
9087 } else {
9088 tcg_temp_free_i32(tmp);
9089 }
9090 }
9091 }
9092 break;
9093 case 12: /* Load/store single data item. */
9094 {
9095 int postinc = 0;
9096 int writeback = 0;
9097 int user;
9098 if ((insn & 0x01100000) == 0x01000000) {
9099 if (disas_neon_ls_insn(env, s, insn))
9100 goto illegal_op;
9101 break;
9102 }
9103 op = ((insn >> 21) & 3) | ((insn >> 22) & 4);
9104 if (rs == 15) {
9105 if (!(insn & (1 << 20))) {
9106 goto illegal_op;
9107 }
9108 if (op != 2) {
9109 /* Byte or halfword load space with dest == r15 : memory hints.
9110 * Catch them early so we don't emit pointless addressing code.
9111 * This space is a mix of:
9112 * PLD/PLDW/PLI, which we implement as NOPs (note that unlike
9113 * the ARM encodings, PLDW space doesn't UNDEF for non-v7MP
9114 * cores)
9115 * unallocated hints, which must be treated as NOPs
9116 * UNPREDICTABLE space, which we NOP or UNDEF depending on
9117 * which is easiest for the decoding logic
9118 * Some space which must UNDEF
9119 */
9120 int op1 = (insn >> 23) & 3;
9121 int op2 = (insn >> 6) & 0x3f;
9122 if (op & 2) {
9123 goto illegal_op;
9124 }
9125 if (rn == 15) {
9126 /* UNPREDICTABLE, unallocated hint or
9127 * PLD/PLDW/PLI (literal)
9128 */
9129 return 0;
9130 }
9131 if (op1 & 1) {
9132 return 0; /* PLD/PLDW/PLI or unallocated hint */
9133 }
9134 if ((op2 == 0) || ((op2 & 0x3c) == 0x30)) {
9135 return 0; /* PLD/PLDW/PLI or unallocated hint */
9136 }
9137 /* UNDEF space, or an UNPREDICTABLE */
9138 return 1;
9139 }
9140 }
9141 user = IS_USER(s);
9142 if (rn == 15) {
9143 addr = tcg_temp_new_i32();
9144 /* PC relative. */
9145 /* s->pc has already been incremented by 4. */
9146 imm = s->pc & 0xfffffffc;
9147 if (insn & (1 << 23))
9148 imm += insn & 0xfff;
9149 else
9150 imm -= insn & 0xfff;
9151 tcg_gen_movi_i32(addr, imm);
9152 } else {
9153 addr = load_reg(s, rn);
9154 if (insn & (1 << 23)) {
9155 /* Positive offset. */
9156 imm = insn & 0xfff;
9157 tcg_gen_addi_i32(addr, addr, imm);
9158 } else {
9159 imm = insn & 0xff;
9160 switch ((insn >> 8) & 0xf) {
9161 case 0x0: /* Shifted Register. */
9162 shift = (insn >> 4) & 0xf;
9163 if (shift > 3) {
9164 tcg_temp_free_i32(addr);
9165 goto illegal_op;
9166 }
9167 tmp = load_reg(s, rm);
9168 if (shift)
9169 tcg_gen_shli_i32(tmp, tmp, shift);
9170 tcg_gen_add_i32(addr, addr, tmp);
9171 tcg_temp_free_i32(tmp);
9172 break;
9173 case 0xc: /* Negative offset. */
9174 tcg_gen_addi_i32(addr, addr, -imm);
9175 break;
9176 case 0xe: /* User privilege. */
9177 tcg_gen_addi_i32(addr, addr, imm);
9178 user = 1;
9179 break;
9180 case 0x9: /* Post-decrement. */
9181 imm = -imm;
9182 /* Fall through. */
9183 case 0xb: /* Post-increment. */
9184 postinc = 1;
9185 writeback = 1;
9186 break;
9187 case 0xd: /* Pre-decrement. */
9188 imm = -imm;
9189 /* Fall through. */
9190 case 0xf: /* Pre-increment. */
9191 tcg_gen_addi_i32(addr, addr, imm);
9192 writeback = 1;
9193 break;
9194 default:
9195 tcg_temp_free_i32(addr);
9196 goto illegal_op;
9197 }
9198 }
9199 }
9200 if (insn & (1 << 20)) {
9201 /* Load. */
9202 tmp = tcg_temp_new_i32();
9203 switch (op) {
9204 case 0:
9205 gen_aa32_ld8u(tmp, addr, user);
9206 break;
9207 case 4:
9208 gen_aa32_ld8s(tmp, addr, user);
9209 break;
9210 case 1:
9211 gen_aa32_ld16u(tmp, addr, user);
9212 break;
9213 case 5:
9214 gen_aa32_ld16s(tmp, addr, user);
9215 break;
9216 case 2:
9217 gen_aa32_ld32u(tmp, addr, user);
9218 break;
9219 default:
9220 tcg_temp_free_i32(tmp);
9221 tcg_temp_free_i32(addr);
9222 goto illegal_op;
9223 }
9224 if (rs == 15) {
9225 gen_bx(s, tmp);
9226 } else {
9227 store_reg(s, rs, tmp);
9228 }
9229 } else {
9230 /* Store. */
9231 tmp = load_reg(s, rs);
9232 switch (op) {
9233 case 0:
9234 gen_aa32_st8(tmp, addr, user);
9235 break;
9236 case 1:
9237 gen_aa32_st16(tmp, addr, user);
9238 break;
9239 case 2:
9240 gen_aa32_st32(tmp, addr, user);
9241 break;
9242 default:
9243 tcg_temp_free_i32(tmp);
9244 tcg_temp_free_i32(addr);
9245 goto illegal_op;
9246 }
9247 tcg_temp_free_i32(tmp);
9248 }
9249 if (postinc)
9250 tcg_gen_addi_i32(addr, addr, imm);
9251 if (writeback) {
9252 store_reg(s, rn, addr);
9253 } else {
9254 tcg_temp_free_i32(addr);
9255 }
9256 }
9257 break;
9258 default:
9259 goto illegal_op;
9260 }
9261 return 0;
9262 illegal_op:
9263 return 1;
9264 }
9265
9266 static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
9267 {
9268 uint32_t val, insn, op, rm, rn, rd, shift, cond;
9269 int32_t offset;
9270 int i;
9271 TCGv_i32 tmp;
9272 TCGv_i32 tmp2;
9273 TCGv_i32 addr;
9274
9275 if (s->condexec_mask) {
9276 cond = s->condexec_cond;
9277 if (cond != 0x0e) { /* Skip conditional when condition is AL. */
9278 s->condlabel = gen_new_label();
9279 gen_test_cc(cond ^ 1, s->condlabel);
9280 s->condjmp = 1;
9281 }
9282 }
9283
9284 insn = arm_lduw_code(env, s->pc, s->bswap_code);
9285 s->pc += 2;
9286
9287 switch (insn >> 12) {
9288 case 0: case 1:
9289
9290 rd = insn & 7;
9291 op = (insn >> 11) & 3;
9292 if (op == 3) {
9293 /* add/subtract */
9294 rn = (insn >> 3) & 7;
9295 tmp = load_reg(s, rn);
9296 if (insn & (1 << 10)) {
9297 /* immediate */
9298 tmp2 = tcg_temp_new_i32();
9299 tcg_gen_movi_i32(tmp2, (insn >> 6) & 7);
9300 } else {
9301 /* reg */
9302 rm = (insn >> 6) & 7;
9303 tmp2 = load_reg(s, rm);
9304 }
9305 if (insn & (1 << 9)) {
9306 if (s->condexec_mask)
9307 tcg_gen_sub_i32(tmp, tmp, tmp2);
9308 else
9309 gen_sub_CC(tmp, tmp, tmp2);
9310 } else {
9311 if (s->condexec_mask)
9312 tcg_gen_add_i32(tmp, tmp, tmp2);
9313 else
9314 gen_add_CC(tmp, tmp, tmp2);
9315 }
9316 tcg_temp_free_i32(tmp2);
9317 store_reg(s, rd, tmp);
9318 } else {
9319 /* shift immediate */
9320 rm = (insn >> 3) & 7;
9321 shift = (insn >> 6) & 0x1f;
9322 tmp = load_reg(s, rm);
9323 gen_arm_shift_im(tmp, op, shift, s->condexec_mask == 0);
9324 if (!s->condexec_mask)
9325 gen_logic_CC(tmp);
9326 store_reg(s, rd, tmp);
9327 }
9328 break;
9329 case 2: case 3:
9330 /* arithmetic large immediate */
9331 op = (insn >> 11) & 3;
9332 rd = (insn >> 8) & 0x7;
9333 if (op == 0) { /* mov */
9334 tmp = tcg_temp_new_i32();
9335 tcg_gen_movi_i32(tmp, insn & 0xff);
9336 if (!s->condexec_mask)
9337 gen_logic_CC(tmp);
9338 store_reg(s, rd, tmp);
9339 } else {
9340 tmp = load_reg(s, rd);
9341 tmp2 = tcg_temp_new_i32();
9342 tcg_gen_movi_i32(tmp2, insn & 0xff);
9343 switch (op) {
9344 case 1: /* cmp */
9345 gen_sub_CC(tmp, tmp, tmp2);
9346 tcg_temp_free_i32(tmp);
9347 tcg_temp_free_i32(tmp2);
9348 break;
9349 case 2: /* add */
9350 if (s->condexec_mask)
9351 tcg_gen_add_i32(tmp, tmp, tmp2);
9352 else
9353 gen_add_CC(tmp, tmp, tmp2);
9354 tcg_temp_free_i32(tmp2);
9355 store_reg(s, rd, tmp);
9356 break;
9357 case 3: /* sub */
9358 if (s->condexec_mask)
9359 tcg_gen_sub_i32(tmp, tmp, tmp2);
9360 else
9361 gen_sub_CC(tmp, tmp, tmp2);
9362 tcg_temp_free_i32(tmp2);
9363 store_reg(s, rd, tmp);
9364 break;
9365 }
9366 }
9367 break;
9368 case 4:
9369 if (insn & (1 << 11)) {
9370 rd = (insn >> 8) & 7;
9371 /* load pc-relative. Bit 1 of PC is ignored. */
9372 val = s->pc + 2 + ((insn & 0xff) * 4);
9373 val &= ~(uint32_t)2;
9374 addr = tcg_temp_new_i32();
9375 tcg_gen_movi_i32(addr, val);
9376 tmp = tcg_temp_new_i32();
9377 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9378 tcg_temp_free_i32(addr);
9379 store_reg(s, rd, tmp);
9380 break;
9381 }
9382 if (insn & (1 << 10)) {
9383 /* data processing extended or blx */
9384 rd = (insn & 7) | ((insn >> 4) & 8);
9385 rm = (insn >> 3) & 0xf;
9386 op = (insn >> 8) & 3;
9387 switch (op) {
9388 case 0: /* add */
9389 tmp = load_reg(s, rd);
9390 tmp2 = load_reg(s, rm);
9391 tcg_gen_add_i32(tmp, tmp, tmp2);
9392 tcg_temp_free_i32(tmp2);
9393 store_reg(s, rd, tmp);
9394 break;
9395 case 1: /* cmp */
9396 tmp = load_reg(s, rd);
9397 tmp2 = load_reg(s, rm);
9398 gen_sub_CC(tmp, tmp, tmp2);
9399 tcg_temp_free_i32(tmp2);
9400 tcg_temp_free_i32(tmp);
9401 break;
9402 case 2: /* mov/cpy */
9403 tmp = load_reg(s, rm);
9404 store_reg(s, rd, tmp);
9405 break;
9406 case 3:/* branch [and link] exchange thumb register */
9407 tmp = load_reg(s, rm);
9408 if (insn & (1 << 7)) {
9409 ARCH(5);
9410 val = (uint32_t)s->pc | 1;
9411 tmp2 = tcg_temp_new_i32();
9412 tcg_gen_movi_i32(tmp2, val);
9413 store_reg(s, 14, tmp2);
9414 }
9415 /* already thumb, no need to check */
9416 gen_bx(s, tmp);
9417 break;
9418 }
9419 break;
9420 }
9421
9422 /* data processing register */
9423 rd = insn & 7;
9424 rm = (insn >> 3) & 7;
9425 op = (insn >> 6) & 0xf;
9426 if (op == 2 || op == 3 || op == 4 || op == 7) {
9427 /* the shift/rotate ops want the operands backwards */
9428 val = rm;
9429 rm = rd;
9430 rd = val;
9431 val = 1;
9432 } else {
9433 val = 0;
9434 }
9435
9436 if (op == 9) { /* neg */
9437 tmp = tcg_temp_new_i32();
9438 tcg_gen_movi_i32(tmp, 0);
9439 } else if (op != 0xf) { /* mvn doesn't read its first operand */
9440 tmp = load_reg(s, rd);
9441 } else {
9442 TCGV_UNUSED_I32(tmp);
9443 }
9444
9445 tmp2 = load_reg(s, rm);
9446 switch (op) {
9447 case 0x0: /* and */
9448 tcg_gen_and_i32(tmp, tmp, tmp2);
9449 if (!s->condexec_mask)
9450 gen_logic_CC(tmp);
9451 break;
9452 case 0x1: /* eor */
9453 tcg_gen_xor_i32(tmp, tmp, tmp2);
9454 if (!s->condexec_mask)
9455 gen_logic_CC(tmp);
9456 break;
9457 case 0x2: /* lsl */
9458 if (s->condexec_mask) {
9459 gen_shl(tmp2, tmp2, tmp);
9460 } else {
9461 gen_helper_shl_cc(tmp2, cpu_env, tmp2, tmp);
9462 gen_logic_CC(tmp2);
9463 }
9464 break;
9465 case 0x3: /* lsr */
9466 if (s->condexec_mask) {
9467 gen_shr(tmp2, tmp2, tmp);
9468 } else {
9469 gen_helper_shr_cc(tmp2, cpu_env, tmp2, tmp);
9470 gen_logic_CC(tmp2);
9471 }
9472 break;
9473 case 0x4: /* asr */
9474 if (s->condexec_mask) {
9475 gen_sar(tmp2, tmp2, tmp);
9476 } else {
9477 gen_helper_sar_cc(tmp2, cpu_env, tmp2, tmp);
9478 gen_logic_CC(tmp2);
9479 }
9480 break;
9481 case 0x5: /* adc */
9482 if (s->condexec_mask) {
9483 gen_adc(tmp, tmp2);
9484 } else {
9485 gen_adc_CC(tmp, tmp, tmp2);
9486 }
9487 break;
9488 case 0x6: /* sbc */
9489 if (s->condexec_mask) {
9490 gen_sub_carry(tmp, tmp, tmp2);
9491 } else {
9492 gen_sbc_CC(tmp, tmp, tmp2);
9493 }
9494 break;
9495 case 0x7: /* ror */
9496 if (s->condexec_mask) {
9497 tcg_gen_andi_i32(tmp, tmp, 0x1f);
9498 tcg_gen_rotr_i32(tmp2, tmp2, tmp);
9499 } else {
9500 gen_helper_ror_cc(tmp2, cpu_env, tmp2, tmp);
9501 gen_logic_CC(tmp2);
9502 }
9503 break;
9504 case 0x8: /* tst */
9505 tcg_gen_and_i32(tmp, tmp, tmp2);
9506 gen_logic_CC(tmp);
9507 rd = 16;
9508 break;
9509 case 0x9: /* neg */
9510 if (s->condexec_mask)
9511 tcg_gen_neg_i32(tmp, tmp2);
9512 else
9513 gen_sub_CC(tmp, tmp, tmp2);
9514 break;
9515 case 0xa: /* cmp */
9516 gen_sub_CC(tmp, tmp, tmp2);
9517 rd = 16;
9518 break;
9519 case 0xb: /* cmn */
9520 gen_add_CC(tmp, tmp, tmp2);
9521 rd = 16;
9522 break;
9523 case 0xc: /* orr */
9524 tcg_gen_or_i32(tmp, tmp, tmp2);
9525 if (!s->condexec_mask)
9526 gen_logic_CC(tmp);
9527 break;
9528 case 0xd: /* mul */
9529 tcg_gen_mul_i32(tmp, tmp, tmp2);
9530 if (!s->condexec_mask)
9531 gen_logic_CC(tmp);
9532 break;
9533 case 0xe: /* bic */
9534 tcg_gen_andc_i32(tmp, tmp, tmp2);
9535 if (!s->condexec_mask)
9536 gen_logic_CC(tmp);
9537 break;
9538 case 0xf: /* mvn */
9539 tcg_gen_not_i32(tmp2, tmp2);
9540 if (!s->condexec_mask)
9541 gen_logic_CC(tmp2);
9542 val = 1;
9543 rm = rd;
9544 break;
9545 }
9546 if (rd != 16) {
9547 if (val) {
9548 store_reg(s, rm, tmp2);
9549 if (op != 0xf)
9550 tcg_temp_free_i32(tmp);
9551 } else {
9552 store_reg(s, rd, tmp);
9553 tcg_temp_free_i32(tmp2);
9554 }
9555 } else {
9556 tcg_temp_free_i32(tmp);
9557 tcg_temp_free_i32(tmp2);
9558 }
9559 break;
9560
9561 case 5:
9562 /* load/store register offset. */
9563 rd = insn & 7;
9564 rn = (insn >> 3) & 7;
9565 rm = (insn >> 6) & 7;
9566 op = (insn >> 9) & 7;
9567 addr = load_reg(s, rn);
9568 tmp = load_reg(s, rm);
9569 tcg_gen_add_i32(addr, addr, tmp);
9570 tcg_temp_free_i32(tmp);
9571
9572 if (op < 3) { /* store */
9573 tmp = load_reg(s, rd);
9574 } else {
9575 tmp = tcg_temp_new_i32();
9576 }
9577
9578 switch (op) {
9579 case 0: /* str */
9580 gen_aa32_st32(tmp, addr, IS_USER(s));
9581 break;
9582 case 1: /* strh */
9583 gen_aa32_st16(tmp, addr, IS_USER(s));
9584 break;
9585 case 2: /* strb */
9586 gen_aa32_st8(tmp, addr, IS_USER(s));
9587 break;
9588 case 3: /* ldrsb */
9589 gen_aa32_ld8s(tmp, addr, IS_USER(s));
9590 break;
9591 case 4: /* ldr */
9592 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9593 break;
9594 case 5: /* ldrh */
9595 gen_aa32_ld16u(tmp, addr, IS_USER(s));
9596 break;
9597 case 6: /* ldrb */
9598 gen_aa32_ld8u(tmp, addr, IS_USER(s));
9599 break;
9600 case 7: /* ldrsh */
9601 gen_aa32_ld16s(tmp, addr, IS_USER(s));
9602 break;
9603 }
9604 if (op >= 3) { /* load */
9605 store_reg(s, rd, tmp);
9606 } else {
9607 tcg_temp_free_i32(tmp);
9608 }
9609 tcg_temp_free_i32(addr);
9610 break;
9611
9612 case 6:
9613 /* load/store word immediate offset */
9614 rd = insn & 7;
9615 rn = (insn >> 3) & 7;
9616 addr = load_reg(s, rn);
9617 val = (insn >> 4) & 0x7c;
9618 tcg_gen_addi_i32(addr, addr, val);
9619
9620 if (insn & (1 << 11)) {
9621 /* load */
9622 tmp = tcg_temp_new_i32();
9623 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9624 store_reg(s, rd, tmp);
9625 } else {
9626 /* store */
9627 tmp = load_reg(s, rd);
9628 gen_aa32_st32(tmp, addr, IS_USER(s));
9629 tcg_temp_free_i32(tmp);
9630 }
9631 tcg_temp_free_i32(addr);
9632 break;
9633
9634 case 7:
9635 /* load/store byte immediate offset */
9636 rd = insn & 7;
9637 rn = (insn >> 3) & 7;
9638 addr = load_reg(s, rn);
9639 val = (insn >> 6) & 0x1f;
9640 tcg_gen_addi_i32(addr, addr, val);
9641
9642 if (insn & (1 << 11)) {
9643 /* load */
9644 tmp = tcg_temp_new_i32();
9645 gen_aa32_ld8u(tmp, addr, IS_USER(s));
9646 store_reg(s, rd, tmp);
9647 } else {
9648 /* store */
9649 tmp = load_reg(s, rd);
9650 gen_aa32_st8(tmp, addr, IS_USER(s));
9651 tcg_temp_free_i32(tmp);
9652 }
9653 tcg_temp_free_i32(addr);
9654 break;
9655
9656 case 8:
9657 /* load/store halfword immediate offset */
9658 rd = insn & 7;
9659 rn = (insn >> 3) & 7;
9660 addr = load_reg(s, rn);
9661 val = (insn >> 5) & 0x3e;
9662 tcg_gen_addi_i32(addr, addr, val);
9663
9664 if (insn & (1 << 11)) {
9665 /* load */
9666 tmp = tcg_temp_new_i32();
9667 gen_aa32_ld16u(tmp, addr, IS_USER(s));
9668 store_reg(s, rd, tmp);
9669 } else {
9670 /* store */
9671 tmp = load_reg(s, rd);
9672 gen_aa32_st16(tmp, addr, IS_USER(s));
9673 tcg_temp_free_i32(tmp);
9674 }
9675 tcg_temp_free_i32(addr);
9676 break;
9677
9678 case 9:
9679 /* load/store from stack */
9680 rd = (insn >> 8) & 7;
9681 addr = load_reg(s, 13);
9682 val = (insn & 0xff) * 4;
9683 tcg_gen_addi_i32(addr, addr, val);
9684
9685 if (insn & (1 << 11)) {
9686 /* load */
9687 tmp = tcg_temp_new_i32();
9688 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9689 store_reg(s, rd, tmp);
9690 } else {
9691 /* store */
9692 tmp = load_reg(s, rd);
9693 gen_aa32_st32(tmp, addr, IS_USER(s));
9694 tcg_temp_free_i32(tmp);
9695 }
9696 tcg_temp_free_i32(addr);
9697 break;
9698
9699 case 10:
9700 /* add to high reg */
9701 rd = (insn >> 8) & 7;
9702 if (insn & (1 << 11)) {
9703 /* SP */
9704 tmp = load_reg(s, 13);
9705 } else {
9706 /* PC. bit 1 is ignored. */
9707 tmp = tcg_temp_new_i32();
9708 tcg_gen_movi_i32(tmp, (s->pc + 2) & ~(uint32_t)2);
9709 }
9710 val = (insn & 0xff) * 4;
9711 tcg_gen_addi_i32(tmp, tmp, val);
9712 store_reg(s, rd, tmp);
9713 break;
9714
9715 case 11:
9716 /* misc */
9717 op = (insn >> 8) & 0xf;
9718 switch (op) {
9719 case 0:
9720 /* adjust stack pointer */
9721 tmp = load_reg(s, 13);
9722 val = (insn & 0x7f) * 4;
9723 if (insn & (1 << 7))
9724 val = -(int32_t)val;
9725 tcg_gen_addi_i32(tmp, tmp, val);
9726 store_reg(s, 13, tmp);
9727 break;
9728
9729 case 2: /* sign/zero extend. */
9730 ARCH(6);
9731 rd = insn & 7;
9732 rm = (insn >> 3) & 7;
9733 tmp = load_reg(s, rm);
9734 switch ((insn >> 6) & 3) {
9735 case 0: gen_sxth(tmp); break;
9736 case 1: gen_sxtb(tmp); break;
9737 case 2: gen_uxth(tmp); break;
9738 case 3: gen_uxtb(tmp); break;
9739 }
9740 store_reg(s, rd, tmp);
9741 break;
9742 case 4: case 5: case 0xc: case 0xd:
9743 /* push/pop */
9744 addr = load_reg(s, 13);
9745 if (insn & (1 << 8))
9746 offset = 4;
9747 else
9748 offset = 0;
9749 for (i = 0; i < 8; i++) {
9750 if (insn & (1 << i))
9751 offset += 4;
9752 }
9753 if ((insn & (1 << 11)) == 0) {
9754 tcg_gen_addi_i32(addr, addr, -offset);
9755 }
9756 for (i = 0; i < 8; i++) {
9757 if (insn & (1 << i)) {
9758 if (insn & (1 << 11)) {
9759 /* pop */
9760 tmp = tcg_temp_new_i32();
9761 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9762 store_reg(s, i, tmp);
9763 } else {
9764 /* push */
9765 tmp = load_reg(s, i);
9766 gen_aa32_st32(tmp, addr, IS_USER(s));
9767 tcg_temp_free_i32(tmp);
9768 }
9769 /* advance to the next address. */
9770 tcg_gen_addi_i32(addr, addr, 4);
9771 }
9772 }
9773 TCGV_UNUSED_I32(tmp);
9774 if (insn & (1 << 8)) {
9775 if (insn & (1 << 11)) {
9776 /* pop pc */
9777 tmp = tcg_temp_new_i32();
9778 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9779 /* don't set the pc until the rest of the instruction
9780 has completed */
9781 } else {
9782 /* push lr */
9783 tmp = load_reg(s, 14);
9784 gen_aa32_st32(tmp, addr, IS_USER(s));
9785 tcg_temp_free_i32(tmp);
9786 }
9787 tcg_gen_addi_i32(addr, addr, 4);
9788 }
9789 if ((insn & (1 << 11)) == 0) {
9790 tcg_gen_addi_i32(addr, addr, -offset);
9791 }
9792 /* write back the new stack pointer */
9793 store_reg(s, 13, addr);
9794 /* set the new PC value */
9795 if ((insn & 0x0900) == 0x0900) {
9796 store_reg_from_load(env, s, 15, tmp);
9797 }
9798 break;
9799
9800 case 1: case 3: case 9: case 11: /* czb */
9801 rm = insn & 7;
9802 tmp = load_reg(s, rm);
9803 s->condlabel = gen_new_label();
9804 s->condjmp = 1;
9805 if (insn & (1 << 11))
9806 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, s->condlabel);
9807 else
9808 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, s->condlabel);
9809 tcg_temp_free_i32(tmp);
9810 offset = ((insn & 0xf8) >> 2) | (insn & 0x200) >> 3;
9811 val = (uint32_t)s->pc + 2;
9812 val += offset;
9813 gen_jmp(s, val);
9814 break;
9815
9816 case 15: /* IT, nop-hint. */
9817 if ((insn & 0xf) == 0) {
9818 gen_nop_hint(s, (insn >> 4) & 0xf);
9819 break;
9820 }
9821 /* If Then. */
9822 s->condexec_cond = (insn >> 4) & 0xe;
9823 s->condexec_mask = insn & 0x1f;
9824 /* No actual code generated for this insn, just setup state. */
9825 break;
9826
9827 case 0xe: /* bkpt */
9828 ARCH(5);
9829 gen_exception_insn(s, 2, EXCP_BKPT);
9830 break;
9831
9832 case 0xa: /* rev */
9833 ARCH(6);
9834 rn = (insn >> 3) & 0x7;
9835 rd = insn & 0x7;
9836 tmp = load_reg(s, rn);
9837 switch ((insn >> 6) & 3) {
9838 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
9839 case 1: gen_rev16(tmp); break;
9840 case 3: gen_revsh(tmp); break;
9841 default: goto illegal_op;
9842 }
9843 store_reg(s, rd, tmp);
9844 break;
9845
9846 case 6:
9847 switch ((insn >> 5) & 7) {
9848 case 2:
9849 /* setend */
9850 ARCH(6);
9851 if (((insn >> 3) & 1) != s->bswap_code) {
9852 /* Dynamic endianness switching not implemented. */
9853 qemu_log_mask(LOG_UNIMP, "arm: unimplemented setend\n");
9854 goto illegal_op;
9855 }
9856 break;
9857 case 3:
9858 /* cps */
9859 ARCH(6);
9860 if (IS_USER(s)) {
9861 break;
9862 }
9863 if (IS_M(env)) {
9864 tmp = tcg_const_i32((insn & (1 << 4)) != 0);
9865 /* FAULTMASK */
9866 if (insn & 1) {
9867 addr = tcg_const_i32(19);
9868 gen_helper_v7m_msr(cpu_env, addr, tmp);
9869 tcg_temp_free_i32(addr);
9870 }
9871 /* PRIMASK */
9872 if (insn & 2) {
9873 addr = tcg_const_i32(16);
9874 gen_helper_v7m_msr(cpu_env, addr, tmp);
9875 tcg_temp_free_i32(addr);
9876 }
9877 tcg_temp_free_i32(tmp);
9878 gen_lookup_tb(s);
9879 } else {
9880 if (insn & (1 << 4)) {
9881 shift = CPSR_A | CPSR_I | CPSR_F;
9882 } else {
9883 shift = 0;
9884 }
9885 gen_set_psr_im(s, ((insn & 7) << 6), 0, shift);
9886 }
9887 break;
9888 default:
9889 goto undef;
9890 }
9891 break;
9892
9893 default:
9894 goto undef;
9895 }
9896 break;
9897
9898 case 12:
9899 {
9900 /* load/store multiple */
9901 TCGv_i32 loaded_var;
9902 TCGV_UNUSED_I32(loaded_var);
9903 rn = (insn >> 8) & 0x7;
9904 addr = load_reg(s, rn);
9905 for (i = 0; i < 8; i++) {
9906 if (insn & (1 << i)) {
9907 if (insn & (1 << 11)) {
9908 /* load */
9909 tmp = tcg_temp_new_i32();
9910 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9911 if (i == rn) {
9912 loaded_var = tmp;
9913 } else {
9914 store_reg(s, i, tmp);
9915 }
9916 } else {
9917 /* store */
9918 tmp = load_reg(s, i);
9919 gen_aa32_st32(tmp, addr, IS_USER(s));
9920 tcg_temp_free_i32(tmp);
9921 }
9922 /* advance to the next address */
9923 tcg_gen_addi_i32(addr, addr, 4);
9924 }
9925 }
9926 if ((insn & (1 << rn)) == 0) {
9927 /* base reg not in list: base register writeback */
9928 store_reg(s, rn, addr);
9929 } else {
9930 /* base reg in list: if load, complete it now */
9931 if (insn & (1 << 11)) {
9932 store_reg(s, rn, loaded_var);
9933 }
9934 tcg_temp_free_i32(addr);
9935 }
9936 break;
9937 }
9938 case 13:
9939 /* conditional branch or swi */
9940 cond = (insn >> 8) & 0xf;
9941 if (cond == 0xe)
9942 goto undef;
9943
9944 if (cond == 0xf) {
9945 /* swi */
9946 gen_set_pc_im(s, s->pc);
9947 s->is_jmp = DISAS_SWI;
9948 break;
9949 }
9950 /* generate a conditional jump to next instruction */
9951 s->condlabel = gen_new_label();
9952 gen_test_cc(cond ^ 1, s->condlabel);
9953 s->condjmp = 1;
9954
9955 /* jump to the offset */
9956 val = (uint32_t)s->pc + 2;
9957 offset = ((int32_t)insn << 24) >> 24;
9958 val += offset << 1;
9959 gen_jmp(s, val);
9960 break;
9961
9962 case 14:
9963 if (insn & (1 << 11)) {
9964 if (disas_thumb2_insn(env, s, insn))
9965 goto undef32;
9966 break;
9967 }
9968 /* unconditional branch */
9969 val = (uint32_t)s->pc;
9970 offset = ((int32_t)insn << 21) >> 21;
9971 val += (offset << 1) + 2;
9972 gen_jmp(s, val);
9973 break;
9974
9975 case 15:
9976 if (disas_thumb2_insn(env, s, insn))
9977 goto undef32;
9978 break;
9979 }
9980 return;
9981 undef32:
9982 gen_exception_insn(s, 4, EXCP_UDEF);
9983 return;
9984 illegal_op:
9985 undef:
9986 gen_exception_insn(s, 2, EXCP_UDEF);
9987 }
9988
9989 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
9990 basic block 'tb'. If search_pc is TRUE, also generate PC
9991 information for each intermediate instruction. */
9992 static inline void gen_intermediate_code_internal(ARMCPU *cpu,
9993 TranslationBlock *tb,
9994 bool search_pc)
9995 {
9996 CPUState *cs = CPU(cpu);
9997 CPUARMState *env = &cpu->env;
9998 DisasContext dc1, *dc = &dc1;
9999 CPUBreakpoint *bp;
10000 uint16_t *gen_opc_end;
10001 int j, lj;
10002 target_ulong pc_start;
10003 target_ulong next_page_start;
10004 int num_insns;
10005 int max_insns;
10006
10007 /* generate intermediate code */
10008 pc_start = tb->pc;
10009
10010 dc->tb = tb;
10011
10012 gen_opc_end = tcg_ctx.gen_opc_buf + OPC_MAX_SIZE;
10013
10014 dc->is_jmp = DISAS_NEXT;
10015 dc->pc = pc_start;
10016 dc->singlestep_enabled = cs->singlestep_enabled;
10017 dc->condjmp = 0;
10018
10019 if (ARM_TBFLAG_AARCH64_STATE(tb->flags)) {
10020 dc->aarch64 = 1;
10021 dc->thumb = 0;
10022 dc->bswap_code = 0;
10023 dc->condexec_mask = 0;
10024 dc->condexec_cond = 0;
10025 #if !defined(CONFIG_USER_ONLY)
10026 dc->user = 0;
10027 #endif
10028 dc->vfp_enabled = 0;
10029 dc->vec_len = 0;
10030 dc->vec_stride = 0;
10031 } else {
10032 dc->aarch64 = 0;
10033 dc->thumb = ARM_TBFLAG_THUMB(tb->flags);
10034 dc->bswap_code = ARM_TBFLAG_BSWAP_CODE(tb->flags);
10035 dc->condexec_mask = (ARM_TBFLAG_CONDEXEC(tb->flags) & 0xf) << 1;
10036 dc->condexec_cond = ARM_TBFLAG_CONDEXEC(tb->flags) >> 4;
10037 #if !defined(CONFIG_USER_ONLY)
10038 dc->user = (ARM_TBFLAG_PRIV(tb->flags) == 0);
10039 #endif
10040 dc->vfp_enabled = ARM_TBFLAG_VFPEN(tb->flags);
10041 dc->vec_len = ARM_TBFLAG_VECLEN(tb->flags);
10042 dc->vec_stride = ARM_TBFLAG_VECSTRIDE(tb->flags);
10043 }
10044 cpu_F0s = tcg_temp_new_i32();
10045 cpu_F1s = tcg_temp_new_i32();
10046 cpu_F0d = tcg_temp_new_i64();
10047 cpu_F1d = tcg_temp_new_i64();
10048 cpu_V0 = cpu_F0d;
10049 cpu_V1 = cpu_F1d;
10050 /* FIXME: cpu_M0 can probably be the same as cpu_V0. */
10051 cpu_M0 = tcg_temp_new_i64();
10052 next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
10053 lj = -1;
10054 num_insns = 0;
10055 max_insns = tb->cflags & CF_COUNT_MASK;
10056 if (max_insns == 0)
10057 max_insns = CF_COUNT_MASK;
10058
10059 gen_tb_start();
10060
10061 tcg_clear_temp_count();
10062
10063 /* A note on handling of the condexec (IT) bits:
10064 *
10065 * We want to avoid the overhead of having to write the updated condexec
10066 * bits back to the CPUARMState for every instruction in an IT block. So:
10067 * (1) if the condexec bits are not already zero then we write
10068 * zero back into the CPUARMState now. This avoids complications trying
10069 * to do it at the end of the block. (For example if we don't do this
10070 * it's hard to identify whether we can safely skip writing condexec
10071 * at the end of the TB, which we definitely want to do for the case
10072 * where a TB doesn't do anything with the IT state at all.)
10073 * (2) if we are going to leave the TB then we call gen_set_condexec()
10074 * which will write the correct value into CPUARMState if zero is wrong.
10075 * This is done both for leaving the TB at the end, and for leaving
10076 * it because of an exception we know will happen, which is done in
10077 * gen_exception_insn(). The latter is necessary because we need to
10078 * leave the TB with the PC/IT state just prior to execution of the
10079 * instruction which caused the exception.
10080 * (3) if we leave the TB unexpectedly (eg a data abort on a load)
10081 * then the CPUARMState will be wrong and we need to reset it.
10082 * This is handled in the same way as restoration of the
10083 * PC in these situations: we will be called again with search_pc=1
10084 * and generate a mapping of the condexec bits for each PC in
10085 * gen_opc_condexec_bits[]. restore_state_to_opc() then uses
10086 * this to restore the condexec bits.
10087 *
10088 * Note that there are no instructions which can read the condexec
10089 * bits, and none which can write non-static values to them, so
10090 * we don't need to care about whether CPUARMState is correct in the
10091 * middle of a TB.
10092 */
10093
10094 /* Reset the conditional execution bits immediately. This avoids
10095 complications trying to do it at the end of the block. */
10096 if (dc->condexec_mask || dc->condexec_cond)
10097 {
10098 TCGv_i32 tmp = tcg_temp_new_i32();
10099 tcg_gen_movi_i32(tmp, 0);
10100 store_cpu_field(tmp, condexec_bits);
10101 }
10102 do {
10103 #ifdef CONFIG_USER_ONLY
10104 /* Intercept jump to the magic kernel page. */
10105 if (!dc->aarch64 && dc->pc >= 0xffff0000) {
10106 /* We always get here via a jump, so know we are not in a
10107 conditional execution block. */
10108 gen_exception(EXCP_KERNEL_TRAP);
10109 dc->is_jmp = DISAS_UPDATE;
10110 break;
10111 }
10112 #else
10113 if (dc->pc >= 0xfffffff0 && IS_M(env)) {
10114 /* We always get here via a jump, so know we are not in a
10115 conditional execution block. */
10116 gen_exception(EXCP_EXCEPTION_EXIT);
10117 dc->is_jmp = DISAS_UPDATE;
10118 break;
10119 }
10120 #endif
10121
10122 if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
10123 QTAILQ_FOREACH(bp, &env->breakpoints, entry) {
10124 if (bp->pc == dc->pc) {
10125 gen_exception_insn(dc, 0, EXCP_DEBUG);
10126 /* Advance PC so that clearing the breakpoint will
10127 invalidate this TB. */
10128 dc->pc += 2;
10129 goto done_generating;
10130 }
10131 }
10132 }
10133 if (search_pc) {
10134 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
10135 if (lj < j) {
10136 lj++;
10137 while (lj < j)
10138 tcg_ctx.gen_opc_instr_start[lj++] = 0;
10139 }
10140 tcg_ctx.gen_opc_pc[lj] = dc->pc;
10141 gen_opc_condexec_bits[lj] = (dc->condexec_cond << 4) | (dc->condexec_mask >> 1);
10142 tcg_ctx.gen_opc_instr_start[lj] = 1;
10143 tcg_ctx.gen_opc_icount[lj] = num_insns;
10144 }
10145
10146 if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO))
10147 gen_io_start();
10148
10149 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP | CPU_LOG_TB_OP_OPT))) {
10150 tcg_gen_debug_insn_start(dc->pc);
10151 }
10152
10153 if (dc->aarch64) {
10154 disas_a64_insn(env, dc);
10155 } else if (dc->thumb) {
10156 disas_thumb_insn(env, dc);
10157 if (dc->condexec_mask) {
10158 dc->condexec_cond = (dc->condexec_cond & 0xe)
10159 | ((dc->condexec_mask >> 4) & 1);
10160 dc->condexec_mask = (dc->condexec_mask << 1) & 0x1f;
10161 if (dc->condexec_mask == 0) {
10162 dc->condexec_cond = 0;
10163 }
10164 }
10165 } else {
10166 disas_arm_insn(env, dc);
10167 }
10168
10169 if (dc->condjmp && !dc->is_jmp) {
10170 gen_set_label(dc->condlabel);
10171 dc->condjmp = 0;
10172 }
10173
10174 if (tcg_check_temp_count()) {
10175 fprintf(stderr, "TCG temporary leak before "TARGET_FMT_lx"\n",
10176 dc->pc);
10177 }
10178
10179 /* Translation stops when a conditional branch is encountered.
10180 * Otherwise the subsequent code could get translated several times.
10181 * Also stop translation when a page boundary is reached. This
10182 * ensures prefetch aborts occur at the right place. */
10183 num_insns ++;
10184 } while (!dc->is_jmp && tcg_ctx.gen_opc_ptr < gen_opc_end &&
10185 !cs->singlestep_enabled &&
10186 !singlestep &&
10187 dc->pc < next_page_start &&
10188 num_insns < max_insns);
10189
10190 if (tb->cflags & CF_LAST_IO) {
10191 if (dc->condjmp) {
10192 /* FIXME: This can theoretically happen with self-modifying
10193 code. */
10194 cpu_abort(env, "IO on conditional branch instruction");
10195 }
10196 gen_io_end();
10197 }
10198
10199 /* At this stage dc->condjmp will only be set when the skipped
10200 instruction was a conditional branch or trap, and the PC has
10201 already been written. */
10202 if (unlikely(cs->singlestep_enabled)) {
10203 /* Make sure the pc is updated, and raise a debug exception. */
10204 if (dc->condjmp) {
10205 gen_set_condexec(dc);
10206 if (dc->is_jmp == DISAS_SWI) {
10207 gen_exception(EXCP_SWI);
10208 } else {
10209 gen_exception(EXCP_DEBUG);
10210 }
10211 gen_set_label(dc->condlabel);
10212 }
10213 if (dc->condjmp || !dc->is_jmp) {
10214 gen_set_pc_im(dc, dc->pc);
10215 dc->condjmp = 0;
10216 }
10217 gen_set_condexec(dc);
10218 if (dc->is_jmp == DISAS_SWI && !dc->condjmp) {
10219 gen_exception(EXCP_SWI);
10220 } else {
10221 /* FIXME: Single stepping a WFI insn will not halt
10222 the CPU. */
10223 gen_exception(EXCP_DEBUG);
10224 }
10225 } else {
10226 /* While branches must always occur at the end of an IT block,
10227 there are a few other things that can cause us to terminate
10228 the TB in the middle of an IT block:
10229 - Exception generating instructions (bkpt, swi, undefined).
10230 - Page boundaries.
10231 - Hardware watchpoints.
10232 Hardware breakpoints have already been handled and skip this code.
10233 */
10234 gen_set_condexec(dc);
10235 switch(dc->is_jmp) {
10236 case DISAS_NEXT:
10237 gen_goto_tb(dc, 1, dc->pc);
10238 break;
10239 default:
10240 case DISAS_JUMP:
10241 case DISAS_UPDATE:
10242 /* indicate that the hash table must be used to find the next TB */
10243 tcg_gen_exit_tb(0);
10244 break;
10245 case DISAS_TB_JUMP:
10246 /* nothing more to generate */
10247 break;
10248 case DISAS_WFI:
10249 gen_helper_wfi(cpu_env);
10250 break;
10251 case DISAS_SWI:
10252 gen_exception(EXCP_SWI);
10253 break;
10254 }
10255 if (dc->condjmp) {
10256 gen_set_label(dc->condlabel);
10257 gen_set_condexec(dc);
10258 gen_goto_tb(dc, 1, dc->pc);
10259 dc->condjmp = 0;
10260 }
10261 }
10262
10263 done_generating:
10264 gen_tb_end(tb, num_insns);
10265 *tcg_ctx.gen_opc_ptr = INDEX_op_end;
10266
10267 #ifdef DEBUG_DISAS
10268 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
10269 qemu_log("----------------\n");
10270 qemu_log("IN: %s\n", lookup_symbol(pc_start));
10271 log_target_disas(env, pc_start, dc->pc - pc_start,
10272 dc->thumb | (dc->bswap_code << 1));
10273 qemu_log("\n");
10274 }
10275 #endif
10276 if (search_pc) {
10277 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
10278 lj++;
10279 while (lj <= j)
10280 tcg_ctx.gen_opc_instr_start[lj++] = 0;
10281 } else {
10282 tb->size = dc->pc - pc_start;
10283 tb->icount = num_insns;
10284 }
10285 }
10286
10287 void gen_intermediate_code(CPUARMState *env, TranslationBlock *tb)
10288 {
10289 gen_intermediate_code_internal(arm_env_get_cpu(env), tb, false);
10290 }
10291
10292 void gen_intermediate_code_pc(CPUARMState *env, TranslationBlock *tb)
10293 {
10294 gen_intermediate_code_internal(arm_env_get_cpu(env), tb, true);
10295 }
10296
10297 static const char *cpu_mode_names[16] = {
10298 "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
10299 "???", "???", "???", "und", "???", "???", "???", "sys"
10300 };
10301
10302 void arm_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
10303 int flags)
10304 {
10305 ARMCPU *cpu = ARM_CPU(cs);
10306 CPUARMState *env = &cpu->env;
10307 int i;
10308 uint32_t psr;
10309
10310 for(i=0;i<16;i++) {
10311 cpu_fprintf(f, "R%02d=%08x", i, env->regs[i]);
10312 if ((i % 4) == 3)
10313 cpu_fprintf(f, "\n");
10314 else
10315 cpu_fprintf(f, " ");
10316 }
10317 psr = cpsr_read(env);
10318 cpu_fprintf(f, "PSR=%08x %c%c%c%c %c %s%d\n",
10319 psr,
10320 psr & (1 << 31) ? 'N' : '-',
10321 psr & (1 << 30) ? 'Z' : '-',
10322 psr & (1 << 29) ? 'C' : '-',
10323 psr & (1 << 28) ? 'V' : '-',
10324 psr & CPSR_T ? 'T' : 'A',
10325 cpu_mode_names[psr & 0xf], (psr & 0x10) ? 32 : 26);
10326
10327 if (flags & CPU_DUMP_FPU) {
10328 int numvfpregs = 0;
10329 if (arm_feature(env, ARM_FEATURE_VFP)) {
10330 numvfpregs += 16;
10331 }
10332 if (arm_feature(env, ARM_FEATURE_VFP3)) {
10333 numvfpregs += 16;
10334 }
10335 for (i = 0; i < numvfpregs; i++) {
10336 uint64_t v = float64_val(env->vfp.regs[i]);
10337 cpu_fprintf(f, "s%02d=%08x s%02d=%08x d%02d=%016" PRIx64 "\n",
10338 i * 2, (uint32_t)v,
10339 i * 2 + 1, (uint32_t)(v >> 32),
10340 i, v);
10341 }
10342 cpu_fprintf(f, "FPSCR: %08x\n", (int)env->vfp.xregs[ARM_VFP_FPSCR]);
10343 }
10344 }
10345
10346 void restore_state_to_opc(CPUARMState *env, TranslationBlock *tb, int pc_pos)
10347 {
10348 if (is_a64(env)) {
10349 env->pc = tcg_ctx.gen_opc_pc[pc_pos];
10350 } else {
10351 env->regs[15] = tcg_ctx.gen_opc_pc[pc_pos];
10352 }
10353 env->condexec_bits = gen_opc_condexec_bits[pc_pos];
10354 }
AR7 emulation for QEMU