4 * Copyright (c) 2003 Fabrice Bellard
5 * Copyright (c) 2005-2007 CodeSourcery
6 * Copyright (c) 2007 OpenedHand, Ltd.
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.
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.
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/>.
28 #include "disas/disas.h"
31 #include "qemu/bitops.h"
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)
48 #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0)
50 #include "translate.h"
51 static uint32_t gen_opc_condexec_bits[OPC_BUF_SIZE];
53 #if defined(CONFIG_USER_ONLY)
56 #define IS_USER(s) (s->user)
60 /* We reuse the same 64-bit temporaries for efficiency. */
61 static TCGv_i64 cpu_V0, cpu_V1, cpu_M0;
62 static TCGv_i32 cpu_R[16];
63 static TCGv_i32 cpu_CF, cpu_NF, cpu_VF, cpu_ZF;
64 static TCGv_i64 cpu_exclusive_addr;
65 static TCGv_i64 cpu_exclusive_val;
66 #ifdef CONFIG_USER_ONLY
67 static TCGv_i64 cpu_exclusive_test;
68 static TCGv_i32 cpu_exclusive_info;
71 /* FIXME: These should be removed. */
72 static TCGv_i32 cpu_F0s, cpu_F1s;
73 static TCGv_i64 cpu_F0d, cpu_F1d;
75 #include "exec/gen-icount.h"
77 static const char *regnames[] =
78 { "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
79 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "pc" };
81 /* initialize TCG globals. */
82 void arm_translate_init(void)
86 cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
88 for (i = 0; i < 16; i++) {
89 cpu_R[i] = tcg_global_mem_new_i32(TCG_AREG0,
90 offsetof(CPUARMState, regs[i]),
93 cpu_CF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, CF), "CF");
94 cpu_NF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, NF), "NF");
95 cpu_VF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, VF), "VF");
96 cpu_ZF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, ZF), "ZF");
98 cpu_exclusive_addr = tcg_global_mem_new_i64(TCG_AREG0,
99 offsetof(CPUARMState, exclusive_addr), "exclusive_addr");
100 cpu_exclusive_val = tcg_global_mem_new_i64(TCG_AREG0,
101 offsetof(CPUARMState, exclusive_val), "exclusive_val");
102 #ifdef CONFIG_USER_ONLY
103 cpu_exclusive_test = tcg_global_mem_new_i64(TCG_AREG0,
104 offsetof(CPUARMState, exclusive_test), "exclusive_test");
105 cpu_exclusive_info = tcg_global_mem_new_i32(TCG_AREG0,
106 offsetof(CPUARMState, exclusive_info), "exclusive_info");
109 a64_translate_init();
112 static inline TCGv_i32 load_cpu_offset(int offset)
114 TCGv_i32 tmp = tcg_temp_new_i32();
115 tcg_gen_ld_i32(tmp, cpu_env, offset);
119 #define load_cpu_field(name) load_cpu_offset(offsetof(CPUARMState, name))
121 static inline void store_cpu_offset(TCGv_i32 var, int offset)
123 tcg_gen_st_i32(var, cpu_env, offset);
124 tcg_temp_free_i32(var);
127 #define store_cpu_field(var, name) \
128 store_cpu_offset(var, offsetof(CPUARMState, name))
130 /* Set a variable to the value of a CPU register. */
131 static void load_reg_var(DisasContext *s, TCGv_i32 var, int reg)
135 /* normally, since we updated PC, we need only to add one insn */
137 addr = (long)s->pc + 2;
139 addr = (long)s->pc + 4;
140 tcg_gen_movi_i32(var, addr);
142 tcg_gen_mov_i32(var, cpu_R[reg]);
146 /* Create a new temporary and set it to the value of a CPU register. */
147 static inline TCGv_i32 load_reg(DisasContext *s, int reg)
149 TCGv_i32 tmp = tcg_temp_new_i32();
150 load_reg_var(s, tmp, reg);
154 /* Set a CPU register. The source must be a temporary and will be
156 static void store_reg(DisasContext *s, int reg, TCGv_i32 var)
159 tcg_gen_andi_i32(var, var, ~1);
160 s->is_jmp = DISAS_JUMP;
162 tcg_gen_mov_i32(cpu_R[reg], var);
163 tcg_temp_free_i32(var);
166 /* Value extensions. */
167 #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var)
168 #define gen_uxth(var) tcg_gen_ext16u_i32(var, var)
169 #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
170 #define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
172 #define gen_sxtb16(var) gen_helper_sxtb16(var, var)
173 #define gen_uxtb16(var) gen_helper_uxtb16(var, var)
176 static inline void gen_set_cpsr(TCGv_i32 var, uint32_t mask)
178 TCGv_i32 tmp_mask = tcg_const_i32(mask);
179 gen_helper_cpsr_write(cpu_env, var, tmp_mask);
180 tcg_temp_free_i32(tmp_mask);
182 /* Set NZCV flags from the high 4 bits of var. */
183 #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV)
185 static void gen_exception(int excp)
187 TCGv_i32 tmp = tcg_temp_new_i32();
188 tcg_gen_movi_i32(tmp, excp);
189 gen_helper_exception(cpu_env, tmp);
190 tcg_temp_free_i32(tmp);
193 static void gen_smul_dual(TCGv_i32 a, TCGv_i32 b)
195 TCGv_i32 tmp1 = tcg_temp_new_i32();
196 TCGv_i32 tmp2 = tcg_temp_new_i32();
197 tcg_gen_ext16s_i32(tmp1, a);
198 tcg_gen_ext16s_i32(tmp2, b);
199 tcg_gen_mul_i32(tmp1, tmp1, tmp2);
200 tcg_temp_free_i32(tmp2);
201 tcg_gen_sari_i32(a, a, 16);
202 tcg_gen_sari_i32(b, b, 16);
203 tcg_gen_mul_i32(b, b, a);
204 tcg_gen_mov_i32(a, tmp1);
205 tcg_temp_free_i32(tmp1);
208 /* Byteswap each halfword. */
209 static void gen_rev16(TCGv_i32 var)
211 TCGv_i32 tmp = tcg_temp_new_i32();
212 tcg_gen_shri_i32(tmp, var, 8);
213 tcg_gen_andi_i32(tmp, tmp, 0x00ff00ff);
214 tcg_gen_shli_i32(var, var, 8);
215 tcg_gen_andi_i32(var, var, 0xff00ff00);
216 tcg_gen_or_i32(var, var, tmp);
217 tcg_temp_free_i32(tmp);
220 /* Byteswap low halfword and sign extend. */
221 static void gen_revsh(TCGv_i32 var)
223 tcg_gen_ext16u_i32(var, var);
224 tcg_gen_bswap16_i32(var, var);
225 tcg_gen_ext16s_i32(var, var);
228 /* Unsigned bitfield extract. */
229 static void gen_ubfx(TCGv_i32 var, int shift, uint32_t mask)
232 tcg_gen_shri_i32(var, var, shift);
233 tcg_gen_andi_i32(var, var, mask);
236 /* Signed bitfield extract. */
237 static void gen_sbfx(TCGv_i32 var, int shift, int width)
242 tcg_gen_sari_i32(var, var, shift);
243 if (shift + width < 32) {
244 signbit = 1u << (width - 1);
245 tcg_gen_andi_i32(var, var, (1u << width) - 1);
246 tcg_gen_xori_i32(var, var, signbit);
247 tcg_gen_subi_i32(var, var, signbit);
251 /* Return (b << 32) + a. Mark inputs as dead */
252 static TCGv_i64 gen_addq_msw(TCGv_i64 a, TCGv_i32 b)
254 TCGv_i64 tmp64 = tcg_temp_new_i64();
256 tcg_gen_extu_i32_i64(tmp64, b);
257 tcg_temp_free_i32(b);
258 tcg_gen_shli_i64(tmp64, tmp64, 32);
259 tcg_gen_add_i64(a, tmp64, a);
261 tcg_temp_free_i64(tmp64);
265 /* Return (b << 32) - a. Mark inputs as dead. */
266 static TCGv_i64 gen_subq_msw(TCGv_i64 a, TCGv_i32 b)
268 TCGv_i64 tmp64 = tcg_temp_new_i64();
270 tcg_gen_extu_i32_i64(tmp64, b);
271 tcg_temp_free_i32(b);
272 tcg_gen_shli_i64(tmp64, tmp64, 32);
273 tcg_gen_sub_i64(a, tmp64, a);
275 tcg_temp_free_i64(tmp64);
279 /* 32x32->64 multiply. Marks inputs as dead. */
280 static TCGv_i64 gen_mulu_i64_i32(TCGv_i32 a, TCGv_i32 b)
282 TCGv_i32 lo = tcg_temp_new_i32();
283 TCGv_i32 hi = tcg_temp_new_i32();
286 tcg_gen_mulu2_i32(lo, hi, a, b);
287 tcg_temp_free_i32(a);
288 tcg_temp_free_i32(b);
290 ret = tcg_temp_new_i64();
291 tcg_gen_concat_i32_i64(ret, lo, hi);
292 tcg_temp_free_i32(lo);
293 tcg_temp_free_i32(hi);
298 static TCGv_i64 gen_muls_i64_i32(TCGv_i32 a, TCGv_i32 b)
300 TCGv_i32 lo = tcg_temp_new_i32();
301 TCGv_i32 hi = tcg_temp_new_i32();
304 tcg_gen_muls2_i32(lo, hi, a, b);
305 tcg_temp_free_i32(a);
306 tcg_temp_free_i32(b);
308 ret = tcg_temp_new_i64();
309 tcg_gen_concat_i32_i64(ret, lo, hi);
310 tcg_temp_free_i32(lo);
311 tcg_temp_free_i32(hi);
316 /* Swap low and high halfwords. */
317 static void gen_swap_half(TCGv_i32 var)
319 TCGv_i32 tmp = tcg_temp_new_i32();
320 tcg_gen_shri_i32(tmp, var, 16);
321 tcg_gen_shli_i32(var, var, 16);
322 tcg_gen_or_i32(var, var, tmp);
323 tcg_temp_free_i32(tmp);
326 /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
327 tmp = (t0 ^ t1) & 0x8000;
330 t0 = (t0 + t1) ^ tmp;
333 static void gen_add16(TCGv_i32 t0, TCGv_i32 t1)
335 TCGv_i32 tmp = tcg_temp_new_i32();
336 tcg_gen_xor_i32(tmp, t0, t1);
337 tcg_gen_andi_i32(tmp, tmp, 0x8000);
338 tcg_gen_andi_i32(t0, t0, ~0x8000);
339 tcg_gen_andi_i32(t1, t1, ~0x8000);
340 tcg_gen_add_i32(t0, t0, t1);
341 tcg_gen_xor_i32(t0, t0, tmp);
342 tcg_temp_free_i32(tmp);
343 tcg_temp_free_i32(t1);
346 /* Set CF to the top bit of var. */
347 static void gen_set_CF_bit31(TCGv_i32 var)
349 tcg_gen_shri_i32(cpu_CF, var, 31);
352 /* Set N and Z flags from var. */
353 static inline void gen_logic_CC(TCGv_i32 var)
355 tcg_gen_mov_i32(cpu_NF, var);
356 tcg_gen_mov_i32(cpu_ZF, var);
360 static void gen_adc(TCGv_i32 t0, TCGv_i32 t1)
362 tcg_gen_add_i32(t0, t0, t1);
363 tcg_gen_add_i32(t0, t0, cpu_CF);
366 /* dest = T0 + T1 + CF. */
367 static void gen_add_carry(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
369 tcg_gen_add_i32(dest, t0, t1);
370 tcg_gen_add_i32(dest, dest, cpu_CF);
373 /* dest = T0 - T1 + CF - 1. */
374 static void gen_sub_carry(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
376 tcg_gen_sub_i32(dest, t0, t1);
377 tcg_gen_add_i32(dest, dest, cpu_CF);
378 tcg_gen_subi_i32(dest, dest, 1);
381 /* dest = T0 + T1. Compute C, N, V and Z flags */
382 static void gen_add_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
384 TCGv_i32 tmp = tcg_temp_new_i32();
385 tcg_gen_movi_i32(tmp, 0);
386 tcg_gen_add2_i32(cpu_NF, cpu_CF, t0, tmp, t1, tmp);
387 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
388 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0);
389 tcg_gen_xor_i32(tmp, t0, t1);
390 tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
391 tcg_temp_free_i32(tmp);
392 tcg_gen_mov_i32(dest, cpu_NF);
395 /* dest = T0 + T1 + CF. Compute C, N, V and Z flags */
396 static void gen_adc_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
398 TCGv_i32 tmp = tcg_temp_new_i32();
399 if (TCG_TARGET_HAS_add2_i32) {
400 tcg_gen_movi_i32(tmp, 0);
401 tcg_gen_add2_i32(cpu_NF, cpu_CF, t0, tmp, cpu_CF, tmp);
402 tcg_gen_add2_i32(cpu_NF, cpu_CF, cpu_NF, cpu_CF, t1, tmp);
404 TCGv_i64 q0 = tcg_temp_new_i64();
405 TCGv_i64 q1 = tcg_temp_new_i64();
406 tcg_gen_extu_i32_i64(q0, t0);
407 tcg_gen_extu_i32_i64(q1, t1);
408 tcg_gen_add_i64(q0, q0, q1);
409 tcg_gen_extu_i32_i64(q1, cpu_CF);
410 tcg_gen_add_i64(q0, q0, q1);
411 tcg_gen_extr_i64_i32(cpu_NF, cpu_CF, q0);
412 tcg_temp_free_i64(q0);
413 tcg_temp_free_i64(q1);
415 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
416 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0);
417 tcg_gen_xor_i32(tmp, t0, t1);
418 tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
419 tcg_temp_free_i32(tmp);
420 tcg_gen_mov_i32(dest, cpu_NF);
423 /* dest = T0 - T1. Compute C, N, V and Z flags */
424 static void gen_sub_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
427 tcg_gen_sub_i32(cpu_NF, t0, t1);
428 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
429 tcg_gen_setcond_i32(TCG_COND_GEU, cpu_CF, t0, t1);
430 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0);
431 tmp = tcg_temp_new_i32();
432 tcg_gen_xor_i32(tmp, t0, t1);
433 tcg_gen_and_i32(cpu_VF, cpu_VF, tmp);
434 tcg_temp_free_i32(tmp);
435 tcg_gen_mov_i32(dest, cpu_NF);
438 /* dest = T0 + ~T1 + CF. Compute C, N, V and Z flags */
439 static void gen_sbc_CC(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
441 TCGv_i32 tmp = tcg_temp_new_i32();
442 tcg_gen_not_i32(tmp, t1);
443 gen_adc_CC(dest, t0, tmp);
444 tcg_temp_free_i32(tmp);
447 #define GEN_SHIFT(name) \
448 static void gen_##name(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1) \
450 TCGv_i32 tmp1, tmp2, tmp3; \
451 tmp1 = tcg_temp_new_i32(); \
452 tcg_gen_andi_i32(tmp1, t1, 0xff); \
453 tmp2 = tcg_const_i32(0); \
454 tmp3 = tcg_const_i32(0x1f); \
455 tcg_gen_movcond_i32(TCG_COND_GTU, tmp2, tmp1, tmp3, tmp2, t0); \
456 tcg_temp_free_i32(tmp3); \
457 tcg_gen_andi_i32(tmp1, tmp1, 0x1f); \
458 tcg_gen_##name##_i32(dest, tmp2, tmp1); \
459 tcg_temp_free_i32(tmp2); \
460 tcg_temp_free_i32(tmp1); \
466 static void gen_sar(TCGv_i32 dest, TCGv_i32 t0, TCGv_i32 t1)
469 tmp1 = tcg_temp_new_i32();
470 tcg_gen_andi_i32(tmp1, t1, 0xff);
471 tmp2 = tcg_const_i32(0x1f);
472 tcg_gen_movcond_i32(TCG_COND_GTU, tmp1, tmp1, tmp2, tmp2, tmp1);
473 tcg_temp_free_i32(tmp2);
474 tcg_gen_sar_i32(dest, t0, tmp1);
475 tcg_temp_free_i32(tmp1);
478 static void tcg_gen_abs_i32(TCGv_i32 dest, TCGv_i32 src)
480 TCGv_i32 c0 = tcg_const_i32(0);
481 TCGv_i32 tmp = tcg_temp_new_i32();
482 tcg_gen_neg_i32(tmp, src);
483 tcg_gen_movcond_i32(TCG_COND_GT, dest, src, c0, src, tmp);
484 tcg_temp_free_i32(c0);
485 tcg_temp_free_i32(tmp);
488 static void shifter_out_im(TCGv_i32 var, int shift)
491 tcg_gen_andi_i32(cpu_CF, var, 1);
493 tcg_gen_shri_i32(cpu_CF, var, shift);
495 tcg_gen_andi_i32(cpu_CF, cpu_CF, 1);
500 /* Shift by immediate. Includes special handling for shift == 0. */
501 static inline void gen_arm_shift_im(TCGv_i32 var, int shiftop,
502 int shift, int flags)
508 shifter_out_im(var, 32 - shift);
509 tcg_gen_shli_i32(var, var, shift);
515 tcg_gen_shri_i32(cpu_CF, var, 31);
517 tcg_gen_movi_i32(var, 0);
520 shifter_out_im(var, shift - 1);
521 tcg_gen_shri_i32(var, var, shift);
528 shifter_out_im(var, shift - 1);
531 tcg_gen_sari_i32(var, var, shift);
533 case 3: /* ROR/RRX */
536 shifter_out_im(var, shift - 1);
537 tcg_gen_rotri_i32(var, var, shift); break;
539 TCGv_i32 tmp = tcg_temp_new_i32();
540 tcg_gen_shli_i32(tmp, cpu_CF, 31);
542 shifter_out_im(var, 0);
543 tcg_gen_shri_i32(var, var, 1);
544 tcg_gen_or_i32(var, var, tmp);
545 tcg_temp_free_i32(tmp);
550 static inline void gen_arm_shift_reg(TCGv_i32 var, int shiftop,
551 TCGv_i32 shift, int flags)
555 case 0: gen_helper_shl_cc(var, cpu_env, var, shift); break;
556 case 1: gen_helper_shr_cc(var, cpu_env, var, shift); break;
557 case 2: gen_helper_sar_cc(var, cpu_env, var, shift); break;
558 case 3: gen_helper_ror_cc(var, cpu_env, var, shift); break;
563 gen_shl(var, var, shift);
566 gen_shr(var, var, shift);
569 gen_sar(var, var, shift);
571 case 3: tcg_gen_andi_i32(shift, shift, 0x1f);
572 tcg_gen_rotr_i32(var, var, shift); break;
575 tcg_temp_free_i32(shift);
578 #define PAS_OP(pfx) \
580 case 0: gen_pas_helper(glue(pfx,add16)); break; \
581 case 1: gen_pas_helper(glue(pfx,addsubx)); break; \
582 case 2: gen_pas_helper(glue(pfx,subaddx)); break; \
583 case 3: gen_pas_helper(glue(pfx,sub16)); break; \
584 case 4: gen_pas_helper(glue(pfx,add8)); break; \
585 case 7: gen_pas_helper(glue(pfx,sub8)); break; \
587 static void gen_arm_parallel_addsub(int op1, int op2, TCGv_i32 a, TCGv_i32 b)
592 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
594 tmp = tcg_temp_new_ptr();
595 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
597 tcg_temp_free_ptr(tmp);
600 tmp = tcg_temp_new_ptr();
601 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
603 tcg_temp_free_ptr(tmp);
605 #undef gen_pas_helper
606 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
619 #undef gen_pas_helper
624 /* For unknown reasons Arm and Thumb-2 use arbitrarily different encodings. */
625 #define PAS_OP(pfx) \
627 case 0: gen_pas_helper(glue(pfx,add8)); break; \
628 case 1: gen_pas_helper(glue(pfx,add16)); break; \
629 case 2: gen_pas_helper(glue(pfx,addsubx)); break; \
630 case 4: gen_pas_helper(glue(pfx,sub8)); break; \
631 case 5: gen_pas_helper(glue(pfx,sub16)); break; \
632 case 6: gen_pas_helper(glue(pfx,subaddx)); break; \
634 static void gen_thumb2_parallel_addsub(int op1, int op2, TCGv_i32 a, TCGv_i32 b)
639 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
641 tmp = tcg_temp_new_ptr();
642 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
644 tcg_temp_free_ptr(tmp);
647 tmp = tcg_temp_new_ptr();
648 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUARMState, GE));
650 tcg_temp_free_ptr(tmp);
652 #undef gen_pas_helper
653 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
666 #undef gen_pas_helper
672 * generate a conditional branch based on ARM condition code cc.
673 * This is common between ARM and Aarch64 targets.
675 void arm_gen_test_cc(int cc, int label)
682 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, label);
685 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_ZF, 0, label);
688 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_CF, 0, label);
691 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_CF, 0, label);
694 tcg_gen_brcondi_i32(TCG_COND_LT, cpu_NF, 0, label);
697 tcg_gen_brcondi_i32(TCG_COND_GE, cpu_NF, 0, label);
700 tcg_gen_brcondi_i32(TCG_COND_LT, cpu_VF, 0, label);
703 tcg_gen_brcondi_i32(TCG_COND_GE, cpu_VF, 0, label);
705 case 8: /* hi: C && !Z */
706 inv = gen_new_label();
707 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_CF, 0, inv);
708 tcg_gen_brcondi_i32(TCG_COND_NE, cpu_ZF, 0, label);
711 case 9: /* ls: !C || Z */
712 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_CF, 0, label);
713 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, label);
715 case 10: /* ge: N == V -> N ^ V == 0 */
716 tmp = tcg_temp_new_i32();
717 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
718 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
719 tcg_temp_free_i32(tmp);
721 case 11: /* lt: N != V -> N ^ V != 0 */
722 tmp = tcg_temp_new_i32();
723 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
724 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
725 tcg_temp_free_i32(tmp);
727 case 12: /* gt: !Z && N == V */
728 inv = gen_new_label();
729 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, inv);
730 tmp = tcg_temp_new_i32();
731 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
732 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
733 tcg_temp_free_i32(tmp);
736 case 13: /* le: Z || N != V */
737 tcg_gen_brcondi_i32(TCG_COND_EQ, cpu_ZF, 0, label);
738 tmp = tcg_temp_new_i32();
739 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
740 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
741 tcg_temp_free_i32(tmp);
744 fprintf(stderr, "Bad condition code 0x%x\n", cc);
749 static const uint8_t table_logic_cc[16] = {
768 /* Set PC and Thumb state from an immediate address. */
769 static inline void gen_bx_im(DisasContext *s, uint32_t addr)
773 s->is_jmp = DISAS_UPDATE;
774 if (s->thumb != (addr & 1)) {
775 tmp = tcg_temp_new_i32();
776 tcg_gen_movi_i32(tmp, addr & 1);
777 tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUARMState, thumb));
778 tcg_temp_free_i32(tmp);
780 tcg_gen_movi_i32(cpu_R[15], addr & ~1);
783 /* Set PC and Thumb state from var. var is marked as dead. */
784 static inline void gen_bx(DisasContext *s, TCGv_i32 var)
786 s->is_jmp = DISAS_UPDATE;
787 tcg_gen_andi_i32(cpu_R[15], var, ~1);
788 tcg_gen_andi_i32(var, var, 1);
789 store_cpu_field(var, thumb);
792 /* Variant of store_reg which uses branch&exchange logic when storing
793 to r15 in ARM architecture v7 and above. The source must be a temporary
794 and will be marked as dead. */
795 static inline void store_reg_bx(CPUARMState *env, DisasContext *s,
796 int reg, TCGv_i32 var)
798 if (reg == 15 && ENABLE_ARCH_7) {
801 store_reg(s, reg, var);
805 /* Variant of store_reg which uses branch&exchange logic when storing
806 * to r15 in ARM architecture v5T and above. This is used for storing
807 * the results of a LDR/LDM/POP into r15, and corresponds to the cases
808 * in the ARM ARM which use the LoadWritePC() pseudocode function. */
809 static inline void store_reg_from_load(CPUARMState *env, DisasContext *s,
810 int reg, TCGv_i32 var)
812 if (reg == 15 && ENABLE_ARCH_5) {
815 store_reg(s, reg, var);
819 /* Abstractions of "generate code to do a guest load/store for
820 * AArch32", where a vaddr is always 32 bits (and is zero
821 * extended if we're a 64 bit core) and data is also
822 * 32 bits unless specifically doing a 64 bit access.
823 * These functions work like tcg_gen_qemu_{ld,st}* except
824 * that the address argument is TCGv_i32 rather than TCGv.
826 #if TARGET_LONG_BITS == 32
828 #define DO_GEN_LD(SUFF, OPC) \
829 static inline void gen_aa32_ld##SUFF(TCGv_i32 val, TCGv_i32 addr, int index) \
831 tcg_gen_qemu_ld_i32(val, addr, index, OPC); \
834 #define DO_GEN_ST(SUFF, OPC) \
835 static inline void gen_aa32_st##SUFF(TCGv_i32 val, TCGv_i32 addr, int index) \
837 tcg_gen_qemu_st_i32(val, addr, index, OPC); \
840 static inline void gen_aa32_ld64(TCGv_i64 val, TCGv_i32 addr, int index)
842 tcg_gen_qemu_ld_i64(val, addr, index, MO_TEQ);
845 static inline void gen_aa32_st64(TCGv_i64 val, TCGv_i32 addr, int index)
847 tcg_gen_qemu_st_i64(val, addr, index, MO_TEQ);
852 #define DO_GEN_LD(SUFF, OPC) \
853 static inline void gen_aa32_ld##SUFF(TCGv_i32 val, TCGv_i32 addr, int index) \
855 TCGv addr64 = tcg_temp_new(); \
856 tcg_gen_extu_i32_i64(addr64, addr); \
857 tcg_gen_qemu_ld_i32(val, addr64, index, OPC); \
858 tcg_temp_free(addr64); \
861 #define DO_GEN_ST(SUFF, OPC) \
862 static inline void gen_aa32_st##SUFF(TCGv_i32 val, TCGv_i32 addr, int index) \
864 TCGv addr64 = tcg_temp_new(); \
865 tcg_gen_extu_i32_i64(addr64, addr); \
866 tcg_gen_qemu_st_i32(val, addr64, index, OPC); \
867 tcg_temp_free(addr64); \
870 static inline void gen_aa32_ld64(TCGv_i64 val, TCGv_i32 addr, int index)
872 TCGv addr64 = tcg_temp_new();
873 tcg_gen_extu_i32_i64(addr64, addr);
874 tcg_gen_qemu_ld_i64(val, addr64, index, MO_TEQ);
875 tcg_temp_free(addr64);
878 static inline void gen_aa32_st64(TCGv_i64 val, TCGv_i32 addr, int index)
880 TCGv addr64 = tcg_temp_new();
881 tcg_gen_extu_i32_i64(addr64, addr);
882 tcg_gen_qemu_st_i64(val, addr64, index, MO_TEQ);
883 tcg_temp_free(addr64);
890 DO_GEN_LD(16s, MO_TESW)
891 DO_GEN_LD(16u, MO_TEUW)
892 DO_GEN_LD(32u, MO_TEUL)
894 DO_GEN_ST(16, MO_TEUW)
895 DO_GEN_ST(32, MO_TEUL)
897 static inline void gen_set_pc_im(DisasContext *s, target_ulong val)
899 tcg_gen_movi_i32(cpu_R[15], val);
902 /* Force a TB lookup after an instruction that changes the CPU state. */
903 static inline void gen_lookup_tb(DisasContext *s)
905 tcg_gen_movi_i32(cpu_R[15], s->pc & ~1);
906 s->is_jmp = DISAS_UPDATE;
909 static inline void gen_add_data_offset(DisasContext *s, unsigned int insn,
912 int val, rm, shift, shiftop;
915 if (!(insn & (1 << 25))) {
918 if (!(insn & (1 << 23)))
921 tcg_gen_addi_i32(var, var, val);
925 shift = (insn >> 7) & 0x1f;
926 shiftop = (insn >> 5) & 3;
927 offset = load_reg(s, rm);
928 gen_arm_shift_im(offset, shiftop, shift, 0);
929 if (!(insn & (1 << 23)))
930 tcg_gen_sub_i32(var, var, offset);
932 tcg_gen_add_i32(var, var, offset);
933 tcg_temp_free_i32(offset);
937 static inline void gen_add_datah_offset(DisasContext *s, unsigned int insn,
938 int extra, TCGv_i32 var)
943 if (insn & (1 << 22)) {
945 val = (insn & 0xf) | ((insn >> 4) & 0xf0);
946 if (!(insn & (1 << 23)))
950 tcg_gen_addi_i32(var, var, val);
954 tcg_gen_addi_i32(var, var, extra);
956 offset = load_reg(s, rm);
957 if (!(insn & (1 << 23)))
958 tcg_gen_sub_i32(var, var, offset);
960 tcg_gen_add_i32(var, var, offset);
961 tcg_temp_free_i32(offset);
965 static TCGv_ptr get_fpstatus_ptr(int neon)
967 TCGv_ptr statusptr = tcg_temp_new_ptr();
970 offset = offsetof(CPUARMState, vfp.standard_fp_status);
972 offset = offsetof(CPUARMState, vfp.fp_status);
974 tcg_gen_addi_ptr(statusptr, cpu_env, offset);
978 #define VFP_OP2(name) \
979 static inline void gen_vfp_##name(int dp) \
981 TCGv_ptr fpst = get_fpstatus_ptr(0); \
983 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, fpst); \
985 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, fpst); \
987 tcg_temp_free_ptr(fpst); \
997 static inline void gen_vfp_F1_mul(int dp)
999 /* Like gen_vfp_mul() but put result in F1 */
1000 TCGv_ptr fpst = get_fpstatus_ptr(0);
1002 gen_helper_vfp_muld(cpu_F1d, cpu_F0d, cpu_F1d, fpst);
1004 gen_helper_vfp_muls(cpu_F1s, cpu_F0s, cpu_F1s, fpst);
1006 tcg_temp_free_ptr(fpst);
1009 static inline void gen_vfp_F1_neg(int dp)
1011 /* Like gen_vfp_neg() but put result in F1 */
1013 gen_helper_vfp_negd(cpu_F1d, cpu_F0d);
1015 gen_helper_vfp_negs(cpu_F1s, cpu_F0s);
1019 static inline void gen_vfp_abs(int dp)
1022 gen_helper_vfp_absd(cpu_F0d, cpu_F0d);
1024 gen_helper_vfp_abss(cpu_F0s, cpu_F0s);
1027 static inline void gen_vfp_neg(int dp)
1030 gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
1032 gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
1035 static inline void gen_vfp_sqrt(int dp)
1038 gen_helper_vfp_sqrtd(cpu_F0d, cpu_F0d, cpu_env);
1040 gen_helper_vfp_sqrts(cpu_F0s, cpu_F0s, cpu_env);
1043 static inline void gen_vfp_cmp(int dp)
1046 gen_helper_vfp_cmpd(cpu_F0d, cpu_F1d, cpu_env);
1048 gen_helper_vfp_cmps(cpu_F0s, cpu_F1s, cpu_env);
1051 static inline void gen_vfp_cmpe(int dp)
1054 gen_helper_vfp_cmped(cpu_F0d, cpu_F1d, cpu_env);
1056 gen_helper_vfp_cmpes(cpu_F0s, cpu_F1s, cpu_env);
1059 static inline void gen_vfp_F1_ld0(int dp)
1062 tcg_gen_movi_i64(cpu_F1d, 0);
1064 tcg_gen_movi_i32(cpu_F1s, 0);
1067 #define VFP_GEN_ITOF(name) \
1068 static inline void gen_vfp_##name(int dp, int neon) \
1070 TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
1072 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0s, statusptr); \
1074 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, statusptr); \
1076 tcg_temp_free_ptr(statusptr); \
1083 #define VFP_GEN_FTOI(name) \
1084 static inline void gen_vfp_##name(int dp, int neon) \
1086 TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
1088 gen_helper_vfp_##name##d(cpu_F0s, cpu_F0d, statusptr); \
1090 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, statusptr); \
1092 tcg_temp_free_ptr(statusptr); \
1101 #define VFP_GEN_FIX(name, round) \
1102 static inline void gen_vfp_##name(int dp, int shift, int neon) \
1104 TCGv_i32 tmp_shift = tcg_const_i32(shift); \
1105 TCGv_ptr statusptr = get_fpstatus_ptr(neon); \
1107 gen_helper_vfp_##name##d##round(cpu_F0d, cpu_F0d, tmp_shift, \
1110 gen_helper_vfp_##name##s##round(cpu_F0s, cpu_F0s, tmp_shift, \
1113 tcg_temp_free_i32(tmp_shift); \
1114 tcg_temp_free_ptr(statusptr); \
1116 VFP_GEN_FIX(tosh, _round_to_zero)
1117 VFP_GEN_FIX(tosl, _round_to_zero)
1118 VFP_GEN_FIX(touh, _round_to_zero)
1119 VFP_GEN_FIX(toul, _round_to_zero)
1126 static inline void gen_vfp_ld(DisasContext *s, int dp, TCGv_i32 addr)
1129 gen_aa32_ld64(cpu_F0d, addr, IS_USER(s));
1131 gen_aa32_ld32u(cpu_F0s, addr, IS_USER(s));
1135 static inline void gen_vfp_st(DisasContext *s, int dp, TCGv_i32 addr)
1138 gen_aa32_st64(cpu_F0d, addr, IS_USER(s));
1140 gen_aa32_st32(cpu_F0s, addr, IS_USER(s));
1145 vfp_reg_offset (int dp, int reg)
1148 return offsetof(CPUARMState, vfp.regs[reg]);
1150 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1151 + offsetof(CPU_DoubleU, l.upper);
1153 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1154 + offsetof(CPU_DoubleU, l.lower);
1158 /* Return the offset of a 32-bit piece of a NEON register.
1159 zero is the least significant end of the register. */
1161 neon_reg_offset (int reg, int n)
1165 return vfp_reg_offset(0, sreg);
1168 static TCGv_i32 neon_load_reg(int reg, int pass)
1170 TCGv_i32 tmp = tcg_temp_new_i32();
1171 tcg_gen_ld_i32(tmp, cpu_env, neon_reg_offset(reg, pass));
1175 static void neon_store_reg(int reg, int pass, TCGv_i32 var)
1177 tcg_gen_st_i32(var, cpu_env, neon_reg_offset(reg, pass));
1178 tcg_temp_free_i32(var);
1181 static inline void neon_load_reg64(TCGv_i64 var, int reg)
1183 tcg_gen_ld_i64(var, cpu_env, vfp_reg_offset(1, reg));
1186 static inline void neon_store_reg64(TCGv_i64 var, int reg)
1188 tcg_gen_st_i64(var, cpu_env, vfp_reg_offset(1, reg));
1191 #define tcg_gen_ld_f32 tcg_gen_ld_i32
1192 #define tcg_gen_ld_f64 tcg_gen_ld_i64
1193 #define tcg_gen_st_f32 tcg_gen_st_i32
1194 #define tcg_gen_st_f64 tcg_gen_st_i64
1196 static inline void gen_mov_F0_vreg(int dp, int reg)
1199 tcg_gen_ld_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1201 tcg_gen_ld_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1204 static inline void gen_mov_F1_vreg(int dp, int reg)
1207 tcg_gen_ld_f64(cpu_F1d, cpu_env, vfp_reg_offset(dp, reg));
1209 tcg_gen_ld_f32(cpu_F1s, cpu_env, vfp_reg_offset(dp, reg));
1212 static inline void gen_mov_vreg_F0(int dp, int reg)
1215 tcg_gen_st_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1217 tcg_gen_st_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1220 #define ARM_CP_RW_BIT (1 << 20)
1222 static inline void iwmmxt_load_reg(TCGv_i64 var, int reg)
1224 tcg_gen_ld_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg]));
1227 static inline void iwmmxt_store_reg(TCGv_i64 var, int reg)
1229 tcg_gen_st_i64(var, cpu_env, offsetof(CPUARMState, iwmmxt.regs[reg]));
1232 static inline TCGv_i32 iwmmxt_load_creg(int reg)
1234 TCGv_i32 var = tcg_temp_new_i32();
1235 tcg_gen_ld_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg]));
1239 static inline void iwmmxt_store_creg(int reg, TCGv_i32 var)
1241 tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, iwmmxt.cregs[reg]));
1242 tcg_temp_free_i32(var);
1245 static inline void gen_op_iwmmxt_movq_wRn_M0(int rn)
1247 iwmmxt_store_reg(cpu_M0, rn);
1250 static inline void gen_op_iwmmxt_movq_M0_wRn(int rn)
1252 iwmmxt_load_reg(cpu_M0, rn);
1255 static inline void gen_op_iwmmxt_orq_M0_wRn(int rn)
1257 iwmmxt_load_reg(cpu_V1, rn);
1258 tcg_gen_or_i64(cpu_M0, cpu_M0, cpu_V1);
1261 static inline void gen_op_iwmmxt_andq_M0_wRn(int rn)
1263 iwmmxt_load_reg(cpu_V1, rn);
1264 tcg_gen_and_i64(cpu_M0, cpu_M0, cpu_V1);
1267 static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn)
1269 iwmmxt_load_reg(cpu_V1, rn);
1270 tcg_gen_xor_i64(cpu_M0, cpu_M0, cpu_V1);
1273 #define IWMMXT_OP(name) \
1274 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1276 iwmmxt_load_reg(cpu_V1, rn); \
1277 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
1280 #define IWMMXT_OP_ENV(name) \
1281 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1283 iwmmxt_load_reg(cpu_V1, rn); \
1284 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \
1287 #define IWMMXT_OP_ENV_SIZE(name) \
1288 IWMMXT_OP_ENV(name##b) \
1289 IWMMXT_OP_ENV(name##w) \
1290 IWMMXT_OP_ENV(name##l)
1292 #define IWMMXT_OP_ENV1(name) \
1293 static inline void gen_op_iwmmxt_##name##_M0(void) \
1295 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \
1309 IWMMXT_OP_ENV_SIZE(unpackl)
1310 IWMMXT_OP_ENV_SIZE(unpackh)
1312 IWMMXT_OP_ENV1(unpacklub)
1313 IWMMXT_OP_ENV1(unpackluw)
1314 IWMMXT_OP_ENV1(unpacklul)
1315 IWMMXT_OP_ENV1(unpackhub)
1316 IWMMXT_OP_ENV1(unpackhuw)
1317 IWMMXT_OP_ENV1(unpackhul)
1318 IWMMXT_OP_ENV1(unpacklsb)
1319 IWMMXT_OP_ENV1(unpacklsw)
1320 IWMMXT_OP_ENV1(unpacklsl)
1321 IWMMXT_OP_ENV1(unpackhsb)
1322 IWMMXT_OP_ENV1(unpackhsw)
1323 IWMMXT_OP_ENV1(unpackhsl)
1325 IWMMXT_OP_ENV_SIZE(cmpeq)
1326 IWMMXT_OP_ENV_SIZE(cmpgtu)
1327 IWMMXT_OP_ENV_SIZE(cmpgts)
1329 IWMMXT_OP_ENV_SIZE(mins)
1330 IWMMXT_OP_ENV_SIZE(minu)
1331 IWMMXT_OP_ENV_SIZE(maxs)
1332 IWMMXT_OP_ENV_SIZE(maxu)
1334 IWMMXT_OP_ENV_SIZE(subn)
1335 IWMMXT_OP_ENV_SIZE(addn)
1336 IWMMXT_OP_ENV_SIZE(subu)
1337 IWMMXT_OP_ENV_SIZE(addu)
1338 IWMMXT_OP_ENV_SIZE(subs)
1339 IWMMXT_OP_ENV_SIZE(adds)
1341 IWMMXT_OP_ENV(avgb0)
1342 IWMMXT_OP_ENV(avgb1)
1343 IWMMXT_OP_ENV(avgw0)
1344 IWMMXT_OP_ENV(avgw1)
1348 IWMMXT_OP_ENV(packuw)
1349 IWMMXT_OP_ENV(packul)
1350 IWMMXT_OP_ENV(packuq)
1351 IWMMXT_OP_ENV(packsw)
1352 IWMMXT_OP_ENV(packsl)
1353 IWMMXT_OP_ENV(packsq)
1355 static void gen_op_iwmmxt_set_mup(void)
1358 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1359 tcg_gen_ori_i32(tmp, tmp, 2);
1360 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1363 static void gen_op_iwmmxt_set_cup(void)
1366 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1367 tcg_gen_ori_i32(tmp, tmp, 1);
1368 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1371 static void gen_op_iwmmxt_setpsr_nz(void)
1373 TCGv_i32 tmp = tcg_temp_new_i32();
1374 gen_helper_iwmmxt_setpsr_nz(tmp, cpu_M0);
1375 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCASF]);
1378 static inline void gen_op_iwmmxt_addl_M0_wRn(int rn)
1380 iwmmxt_load_reg(cpu_V1, rn);
1381 tcg_gen_ext32u_i64(cpu_V1, cpu_V1);
1382 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1385 static inline int gen_iwmmxt_address(DisasContext *s, uint32_t insn,
1392 rd = (insn >> 16) & 0xf;
1393 tmp = load_reg(s, rd);
1395 offset = (insn & 0xff) << ((insn >> 7) & 2);
1396 if (insn & (1 << 24)) {
1398 if (insn & (1 << 23))
1399 tcg_gen_addi_i32(tmp, tmp, offset);
1401 tcg_gen_addi_i32(tmp, tmp, -offset);
1402 tcg_gen_mov_i32(dest, tmp);
1403 if (insn & (1 << 21))
1404 store_reg(s, rd, tmp);
1406 tcg_temp_free_i32(tmp);
1407 } else if (insn & (1 << 21)) {
1409 tcg_gen_mov_i32(dest, tmp);
1410 if (insn & (1 << 23))
1411 tcg_gen_addi_i32(tmp, tmp, offset);
1413 tcg_gen_addi_i32(tmp, tmp, -offset);
1414 store_reg(s, rd, tmp);
1415 } else if (!(insn & (1 << 23)))
1420 static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask, TCGv_i32 dest)
1422 int rd = (insn >> 0) & 0xf;
1425 if (insn & (1 << 8)) {
1426 if (rd < ARM_IWMMXT_wCGR0 || rd > ARM_IWMMXT_wCGR3) {
1429 tmp = iwmmxt_load_creg(rd);
1432 tmp = tcg_temp_new_i32();
1433 iwmmxt_load_reg(cpu_V0, rd);
1434 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
1436 tcg_gen_andi_i32(tmp, tmp, mask);
1437 tcg_gen_mov_i32(dest, tmp);
1438 tcg_temp_free_i32(tmp);
1442 /* Disassemble an iwMMXt instruction. Returns nonzero if an error occurred
1443 (ie. an undefined instruction). */
1444 static int disas_iwmmxt_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
1447 int rdhi, rdlo, rd0, rd1, i;
1449 TCGv_i32 tmp, tmp2, tmp3;
1451 if ((insn & 0x0e000e00) == 0x0c000000) {
1452 if ((insn & 0x0fe00ff0) == 0x0c400000) {
1454 rdlo = (insn >> 12) & 0xf;
1455 rdhi = (insn >> 16) & 0xf;
1456 if (insn & ARM_CP_RW_BIT) { /* TMRRC */
1457 iwmmxt_load_reg(cpu_V0, wrd);
1458 tcg_gen_trunc_i64_i32(cpu_R[rdlo], cpu_V0);
1459 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
1460 tcg_gen_trunc_i64_i32(cpu_R[rdhi], cpu_V0);
1461 } else { /* TMCRR */
1462 tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]);
1463 iwmmxt_store_reg(cpu_V0, wrd);
1464 gen_op_iwmmxt_set_mup();
1469 wrd = (insn >> 12) & 0xf;
1470 addr = tcg_temp_new_i32();
1471 if (gen_iwmmxt_address(s, insn, addr)) {
1472 tcg_temp_free_i32(addr);
1475 if (insn & ARM_CP_RW_BIT) {
1476 if ((insn >> 28) == 0xf) { /* WLDRW wCx */
1477 tmp = tcg_temp_new_i32();
1478 gen_aa32_ld32u(tmp, addr, IS_USER(s));
1479 iwmmxt_store_creg(wrd, tmp);
1482 if (insn & (1 << 8)) {
1483 if (insn & (1 << 22)) { /* WLDRD */
1484 gen_aa32_ld64(cpu_M0, addr, IS_USER(s));
1486 } else { /* WLDRW wRd */
1487 tmp = tcg_temp_new_i32();
1488 gen_aa32_ld32u(tmp, addr, IS_USER(s));
1491 tmp = tcg_temp_new_i32();
1492 if (insn & (1 << 22)) { /* WLDRH */
1493 gen_aa32_ld16u(tmp, addr, IS_USER(s));
1494 } else { /* WLDRB */
1495 gen_aa32_ld8u(tmp, addr, IS_USER(s));
1499 tcg_gen_extu_i32_i64(cpu_M0, tmp);
1500 tcg_temp_free_i32(tmp);
1502 gen_op_iwmmxt_movq_wRn_M0(wrd);
1505 if ((insn >> 28) == 0xf) { /* WSTRW wCx */
1506 tmp = iwmmxt_load_creg(wrd);
1507 gen_aa32_st32(tmp, addr, IS_USER(s));
1509 gen_op_iwmmxt_movq_M0_wRn(wrd);
1510 tmp = tcg_temp_new_i32();
1511 if (insn & (1 << 8)) {
1512 if (insn & (1 << 22)) { /* WSTRD */
1513 gen_aa32_st64(cpu_M0, addr, IS_USER(s));
1514 } else { /* WSTRW wRd */
1515 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1516 gen_aa32_st32(tmp, addr, IS_USER(s));
1519 if (insn & (1 << 22)) { /* WSTRH */
1520 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1521 gen_aa32_st16(tmp, addr, IS_USER(s));
1522 } else { /* WSTRB */
1523 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1524 gen_aa32_st8(tmp, addr, IS_USER(s));
1528 tcg_temp_free_i32(tmp);
1530 tcg_temp_free_i32(addr);
1534 if ((insn & 0x0f000000) != 0x0e000000)
1537 switch (((insn >> 12) & 0xf00) | ((insn >> 4) & 0xff)) {
1538 case 0x000: /* WOR */
1539 wrd = (insn >> 12) & 0xf;
1540 rd0 = (insn >> 0) & 0xf;
1541 rd1 = (insn >> 16) & 0xf;
1542 gen_op_iwmmxt_movq_M0_wRn(rd0);
1543 gen_op_iwmmxt_orq_M0_wRn(rd1);
1544 gen_op_iwmmxt_setpsr_nz();
1545 gen_op_iwmmxt_movq_wRn_M0(wrd);
1546 gen_op_iwmmxt_set_mup();
1547 gen_op_iwmmxt_set_cup();
1549 case 0x011: /* TMCR */
1552 rd = (insn >> 12) & 0xf;
1553 wrd = (insn >> 16) & 0xf;
1555 case ARM_IWMMXT_wCID:
1556 case ARM_IWMMXT_wCASF:
1558 case ARM_IWMMXT_wCon:
1559 gen_op_iwmmxt_set_cup();
1561 case ARM_IWMMXT_wCSSF:
1562 tmp = iwmmxt_load_creg(wrd);
1563 tmp2 = load_reg(s, rd);
1564 tcg_gen_andc_i32(tmp, tmp, tmp2);
1565 tcg_temp_free_i32(tmp2);
1566 iwmmxt_store_creg(wrd, tmp);
1568 case ARM_IWMMXT_wCGR0:
1569 case ARM_IWMMXT_wCGR1:
1570 case ARM_IWMMXT_wCGR2:
1571 case ARM_IWMMXT_wCGR3:
1572 gen_op_iwmmxt_set_cup();
1573 tmp = load_reg(s, rd);
1574 iwmmxt_store_creg(wrd, tmp);
1580 case 0x100: /* WXOR */
1581 wrd = (insn >> 12) & 0xf;
1582 rd0 = (insn >> 0) & 0xf;
1583 rd1 = (insn >> 16) & 0xf;
1584 gen_op_iwmmxt_movq_M0_wRn(rd0);
1585 gen_op_iwmmxt_xorq_M0_wRn(rd1);
1586 gen_op_iwmmxt_setpsr_nz();
1587 gen_op_iwmmxt_movq_wRn_M0(wrd);
1588 gen_op_iwmmxt_set_mup();
1589 gen_op_iwmmxt_set_cup();
1591 case 0x111: /* TMRC */
1594 rd = (insn >> 12) & 0xf;
1595 wrd = (insn >> 16) & 0xf;
1596 tmp = iwmmxt_load_creg(wrd);
1597 store_reg(s, rd, tmp);
1599 case 0x300: /* WANDN */
1600 wrd = (insn >> 12) & 0xf;
1601 rd0 = (insn >> 0) & 0xf;
1602 rd1 = (insn >> 16) & 0xf;
1603 gen_op_iwmmxt_movq_M0_wRn(rd0);
1604 tcg_gen_neg_i64(cpu_M0, cpu_M0);
1605 gen_op_iwmmxt_andq_M0_wRn(rd1);
1606 gen_op_iwmmxt_setpsr_nz();
1607 gen_op_iwmmxt_movq_wRn_M0(wrd);
1608 gen_op_iwmmxt_set_mup();
1609 gen_op_iwmmxt_set_cup();
1611 case 0x200: /* WAND */
1612 wrd = (insn >> 12) & 0xf;
1613 rd0 = (insn >> 0) & 0xf;
1614 rd1 = (insn >> 16) & 0xf;
1615 gen_op_iwmmxt_movq_M0_wRn(rd0);
1616 gen_op_iwmmxt_andq_M0_wRn(rd1);
1617 gen_op_iwmmxt_setpsr_nz();
1618 gen_op_iwmmxt_movq_wRn_M0(wrd);
1619 gen_op_iwmmxt_set_mup();
1620 gen_op_iwmmxt_set_cup();
1622 case 0x810: case 0xa10: /* WMADD */
1623 wrd = (insn >> 12) & 0xf;
1624 rd0 = (insn >> 0) & 0xf;
1625 rd1 = (insn >> 16) & 0xf;
1626 gen_op_iwmmxt_movq_M0_wRn(rd0);
1627 if (insn & (1 << 21))
1628 gen_op_iwmmxt_maddsq_M0_wRn(rd1);
1630 gen_op_iwmmxt_madduq_M0_wRn(rd1);
1631 gen_op_iwmmxt_movq_wRn_M0(wrd);
1632 gen_op_iwmmxt_set_mup();
1634 case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */
1635 wrd = (insn >> 12) & 0xf;
1636 rd0 = (insn >> 16) & 0xf;
1637 rd1 = (insn >> 0) & 0xf;
1638 gen_op_iwmmxt_movq_M0_wRn(rd0);
1639 switch ((insn >> 22) & 3) {
1641 gen_op_iwmmxt_unpacklb_M0_wRn(rd1);
1644 gen_op_iwmmxt_unpacklw_M0_wRn(rd1);
1647 gen_op_iwmmxt_unpackll_M0_wRn(rd1);
1652 gen_op_iwmmxt_movq_wRn_M0(wrd);
1653 gen_op_iwmmxt_set_mup();
1654 gen_op_iwmmxt_set_cup();
1656 case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */
1657 wrd = (insn >> 12) & 0xf;
1658 rd0 = (insn >> 16) & 0xf;
1659 rd1 = (insn >> 0) & 0xf;
1660 gen_op_iwmmxt_movq_M0_wRn(rd0);
1661 switch ((insn >> 22) & 3) {
1663 gen_op_iwmmxt_unpackhb_M0_wRn(rd1);
1666 gen_op_iwmmxt_unpackhw_M0_wRn(rd1);
1669 gen_op_iwmmxt_unpackhl_M0_wRn(rd1);
1674 gen_op_iwmmxt_movq_wRn_M0(wrd);
1675 gen_op_iwmmxt_set_mup();
1676 gen_op_iwmmxt_set_cup();
1678 case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */
1679 wrd = (insn >> 12) & 0xf;
1680 rd0 = (insn >> 16) & 0xf;
1681 rd1 = (insn >> 0) & 0xf;
1682 gen_op_iwmmxt_movq_M0_wRn(rd0);
1683 if (insn & (1 << 22))
1684 gen_op_iwmmxt_sadw_M0_wRn(rd1);
1686 gen_op_iwmmxt_sadb_M0_wRn(rd1);
1687 if (!(insn & (1 << 20)))
1688 gen_op_iwmmxt_addl_M0_wRn(wrd);
1689 gen_op_iwmmxt_movq_wRn_M0(wrd);
1690 gen_op_iwmmxt_set_mup();
1692 case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */
1693 wrd = (insn >> 12) & 0xf;
1694 rd0 = (insn >> 16) & 0xf;
1695 rd1 = (insn >> 0) & 0xf;
1696 gen_op_iwmmxt_movq_M0_wRn(rd0);
1697 if (insn & (1 << 21)) {
1698 if (insn & (1 << 20))
1699 gen_op_iwmmxt_mulshw_M0_wRn(rd1);
1701 gen_op_iwmmxt_mulslw_M0_wRn(rd1);
1703 if (insn & (1 << 20))
1704 gen_op_iwmmxt_muluhw_M0_wRn(rd1);
1706 gen_op_iwmmxt_mululw_M0_wRn(rd1);
1708 gen_op_iwmmxt_movq_wRn_M0(wrd);
1709 gen_op_iwmmxt_set_mup();
1711 case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */
1712 wrd = (insn >> 12) & 0xf;
1713 rd0 = (insn >> 16) & 0xf;
1714 rd1 = (insn >> 0) & 0xf;
1715 gen_op_iwmmxt_movq_M0_wRn(rd0);
1716 if (insn & (1 << 21))
1717 gen_op_iwmmxt_macsw_M0_wRn(rd1);
1719 gen_op_iwmmxt_macuw_M0_wRn(rd1);
1720 if (!(insn & (1 << 20))) {
1721 iwmmxt_load_reg(cpu_V1, wrd);
1722 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1724 gen_op_iwmmxt_movq_wRn_M0(wrd);
1725 gen_op_iwmmxt_set_mup();
1727 case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */
1728 wrd = (insn >> 12) & 0xf;
1729 rd0 = (insn >> 16) & 0xf;
1730 rd1 = (insn >> 0) & 0xf;
1731 gen_op_iwmmxt_movq_M0_wRn(rd0);
1732 switch ((insn >> 22) & 3) {
1734 gen_op_iwmmxt_cmpeqb_M0_wRn(rd1);
1737 gen_op_iwmmxt_cmpeqw_M0_wRn(rd1);
1740 gen_op_iwmmxt_cmpeql_M0_wRn(rd1);
1745 gen_op_iwmmxt_movq_wRn_M0(wrd);
1746 gen_op_iwmmxt_set_mup();
1747 gen_op_iwmmxt_set_cup();
1749 case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */
1750 wrd = (insn >> 12) & 0xf;
1751 rd0 = (insn >> 16) & 0xf;
1752 rd1 = (insn >> 0) & 0xf;
1753 gen_op_iwmmxt_movq_M0_wRn(rd0);
1754 if (insn & (1 << 22)) {
1755 if (insn & (1 << 20))
1756 gen_op_iwmmxt_avgw1_M0_wRn(rd1);
1758 gen_op_iwmmxt_avgw0_M0_wRn(rd1);
1760 if (insn & (1 << 20))
1761 gen_op_iwmmxt_avgb1_M0_wRn(rd1);
1763 gen_op_iwmmxt_avgb0_M0_wRn(rd1);
1765 gen_op_iwmmxt_movq_wRn_M0(wrd);
1766 gen_op_iwmmxt_set_mup();
1767 gen_op_iwmmxt_set_cup();
1769 case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */
1770 wrd = (insn >> 12) & 0xf;
1771 rd0 = (insn >> 16) & 0xf;
1772 rd1 = (insn >> 0) & 0xf;
1773 gen_op_iwmmxt_movq_M0_wRn(rd0);
1774 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCGR0 + ((insn >> 20) & 3));
1775 tcg_gen_andi_i32(tmp, tmp, 7);
1776 iwmmxt_load_reg(cpu_V1, rd1);
1777 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp);
1778 tcg_temp_free_i32(tmp);
1779 gen_op_iwmmxt_movq_wRn_M0(wrd);
1780 gen_op_iwmmxt_set_mup();
1782 case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */
1783 if (((insn >> 6) & 3) == 3)
1785 rd = (insn >> 12) & 0xf;
1786 wrd = (insn >> 16) & 0xf;
1787 tmp = load_reg(s, rd);
1788 gen_op_iwmmxt_movq_M0_wRn(wrd);
1789 switch ((insn >> 6) & 3) {
1791 tmp2 = tcg_const_i32(0xff);
1792 tmp3 = tcg_const_i32((insn & 7) << 3);
1795 tmp2 = tcg_const_i32(0xffff);
1796 tmp3 = tcg_const_i32((insn & 3) << 4);
1799 tmp2 = tcg_const_i32(0xffffffff);
1800 tmp3 = tcg_const_i32((insn & 1) << 5);
1803 TCGV_UNUSED_I32(tmp2);
1804 TCGV_UNUSED_I32(tmp3);
1806 gen_helper_iwmmxt_insr(cpu_M0, cpu_M0, tmp, tmp2, tmp3);
1807 tcg_temp_free_i32(tmp3);
1808 tcg_temp_free_i32(tmp2);
1809 tcg_temp_free_i32(tmp);
1810 gen_op_iwmmxt_movq_wRn_M0(wrd);
1811 gen_op_iwmmxt_set_mup();
1813 case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */
1814 rd = (insn >> 12) & 0xf;
1815 wrd = (insn >> 16) & 0xf;
1816 if (rd == 15 || ((insn >> 22) & 3) == 3)
1818 gen_op_iwmmxt_movq_M0_wRn(wrd);
1819 tmp = tcg_temp_new_i32();
1820 switch ((insn >> 22) & 3) {
1822 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 7) << 3);
1823 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1825 tcg_gen_ext8s_i32(tmp, tmp);
1827 tcg_gen_andi_i32(tmp, tmp, 0xff);
1831 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 3) << 4);
1832 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1834 tcg_gen_ext16s_i32(tmp, tmp);
1836 tcg_gen_andi_i32(tmp, tmp, 0xffff);
1840 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 1) << 5);
1841 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1844 store_reg(s, rd, tmp);
1846 case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */
1847 if ((insn & 0x000ff008) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1849 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1850 switch ((insn >> 22) & 3) {
1852 tcg_gen_shri_i32(tmp, tmp, ((insn & 7) << 2) + 0);
1855 tcg_gen_shri_i32(tmp, tmp, ((insn & 3) << 3) + 4);
1858 tcg_gen_shri_i32(tmp, tmp, ((insn & 1) << 4) + 12);
1861 tcg_gen_shli_i32(tmp, tmp, 28);
1863 tcg_temp_free_i32(tmp);
1865 case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */
1866 if (((insn >> 6) & 3) == 3)
1868 rd = (insn >> 12) & 0xf;
1869 wrd = (insn >> 16) & 0xf;
1870 tmp = load_reg(s, rd);
1871 switch ((insn >> 6) & 3) {
1873 gen_helper_iwmmxt_bcstb(cpu_M0, tmp);
1876 gen_helper_iwmmxt_bcstw(cpu_M0, tmp);
1879 gen_helper_iwmmxt_bcstl(cpu_M0, tmp);
1882 tcg_temp_free_i32(tmp);
1883 gen_op_iwmmxt_movq_wRn_M0(wrd);
1884 gen_op_iwmmxt_set_mup();
1886 case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */
1887 if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1889 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1890 tmp2 = tcg_temp_new_i32();
1891 tcg_gen_mov_i32(tmp2, tmp);
1892 switch ((insn >> 22) & 3) {
1894 for (i = 0; i < 7; i ++) {
1895 tcg_gen_shli_i32(tmp2, tmp2, 4);
1896 tcg_gen_and_i32(tmp, tmp, tmp2);
1900 for (i = 0; i < 3; i ++) {
1901 tcg_gen_shli_i32(tmp2, tmp2, 8);
1902 tcg_gen_and_i32(tmp, tmp, tmp2);
1906 tcg_gen_shli_i32(tmp2, tmp2, 16);
1907 tcg_gen_and_i32(tmp, tmp, tmp2);
1911 tcg_temp_free_i32(tmp2);
1912 tcg_temp_free_i32(tmp);
1914 case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */
1915 wrd = (insn >> 12) & 0xf;
1916 rd0 = (insn >> 16) & 0xf;
1917 gen_op_iwmmxt_movq_M0_wRn(rd0);
1918 switch ((insn >> 22) & 3) {
1920 gen_helper_iwmmxt_addcb(cpu_M0, cpu_M0);
1923 gen_helper_iwmmxt_addcw(cpu_M0, cpu_M0);
1926 gen_helper_iwmmxt_addcl(cpu_M0, cpu_M0);
1931 gen_op_iwmmxt_movq_wRn_M0(wrd);
1932 gen_op_iwmmxt_set_mup();
1934 case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */
1935 if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1937 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1938 tmp2 = tcg_temp_new_i32();
1939 tcg_gen_mov_i32(tmp2, tmp);
1940 switch ((insn >> 22) & 3) {
1942 for (i = 0; i < 7; i ++) {
1943 tcg_gen_shli_i32(tmp2, tmp2, 4);
1944 tcg_gen_or_i32(tmp, tmp, tmp2);
1948 for (i = 0; i < 3; i ++) {
1949 tcg_gen_shli_i32(tmp2, tmp2, 8);
1950 tcg_gen_or_i32(tmp, tmp, tmp2);
1954 tcg_gen_shli_i32(tmp2, tmp2, 16);
1955 tcg_gen_or_i32(tmp, tmp, tmp2);
1959 tcg_temp_free_i32(tmp2);
1960 tcg_temp_free_i32(tmp);
1962 case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */
1963 rd = (insn >> 12) & 0xf;
1964 rd0 = (insn >> 16) & 0xf;
1965 if ((insn & 0xf) != 0 || ((insn >> 22) & 3) == 3)
1967 gen_op_iwmmxt_movq_M0_wRn(rd0);
1968 tmp = tcg_temp_new_i32();
1969 switch ((insn >> 22) & 3) {
1971 gen_helper_iwmmxt_msbb(tmp, cpu_M0);
1974 gen_helper_iwmmxt_msbw(tmp, cpu_M0);
1977 gen_helper_iwmmxt_msbl(tmp, cpu_M0);
1980 store_reg(s, rd, tmp);
1982 case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */
1983 case 0x906: case 0xb06: case 0xd06: case 0xf06:
1984 wrd = (insn >> 12) & 0xf;
1985 rd0 = (insn >> 16) & 0xf;
1986 rd1 = (insn >> 0) & 0xf;
1987 gen_op_iwmmxt_movq_M0_wRn(rd0);
1988 switch ((insn >> 22) & 3) {
1990 if (insn & (1 << 21))
1991 gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1);
1993 gen_op_iwmmxt_cmpgtub_M0_wRn(rd1);
1996 if (insn & (1 << 21))
1997 gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1);
1999 gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1);
2002 if (insn & (1 << 21))
2003 gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1);
2005 gen_op_iwmmxt_cmpgtul_M0_wRn(rd1);
2010 gen_op_iwmmxt_movq_wRn_M0(wrd);
2011 gen_op_iwmmxt_set_mup();
2012 gen_op_iwmmxt_set_cup();
2014 case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */
2015 case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e:
2016 wrd = (insn >> 12) & 0xf;
2017 rd0 = (insn >> 16) & 0xf;
2018 gen_op_iwmmxt_movq_M0_wRn(rd0);
2019 switch ((insn >> 22) & 3) {
2021 if (insn & (1 << 21))
2022 gen_op_iwmmxt_unpacklsb_M0();
2024 gen_op_iwmmxt_unpacklub_M0();
2027 if (insn & (1 << 21))
2028 gen_op_iwmmxt_unpacklsw_M0();
2030 gen_op_iwmmxt_unpackluw_M0();
2033 if (insn & (1 << 21))
2034 gen_op_iwmmxt_unpacklsl_M0();
2036 gen_op_iwmmxt_unpacklul_M0();
2041 gen_op_iwmmxt_movq_wRn_M0(wrd);
2042 gen_op_iwmmxt_set_mup();
2043 gen_op_iwmmxt_set_cup();
2045 case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */
2046 case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c:
2047 wrd = (insn >> 12) & 0xf;
2048 rd0 = (insn >> 16) & 0xf;
2049 gen_op_iwmmxt_movq_M0_wRn(rd0);
2050 switch ((insn >> 22) & 3) {
2052 if (insn & (1 << 21))
2053 gen_op_iwmmxt_unpackhsb_M0();
2055 gen_op_iwmmxt_unpackhub_M0();
2058 if (insn & (1 << 21))
2059 gen_op_iwmmxt_unpackhsw_M0();
2061 gen_op_iwmmxt_unpackhuw_M0();
2064 if (insn & (1 << 21))
2065 gen_op_iwmmxt_unpackhsl_M0();
2067 gen_op_iwmmxt_unpackhul_M0();
2072 gen_op_iwmmxt_movq_wRn_M0(wrd);
2073 gen_op_iwmmxt_set_mup();
2074 gen_op_iwmmxt_set_cup();
2076 case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */
2077 case 0x214: case 0x614: case 0xa14: case 0xe14:
2078 if (((insn >> 22) & 3) == 0)
2080 wrd = (insn >> 12) & 0xf;
2081 rd0 = (insn >> 16) & 0xf;
2082 gen_op_iwmmxt_movq_M0_wRn(rd0);
2083 tmp = tcg_temp_new_i32();
2084 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2085 tcg_temp_free_i32(tmp);
2088 switch ((insn >> 22) & 3) {
2090 gen_helper_iwmmxt_srlw(cpu_M0, cpu_env, cpu_M0, tmp);
2093 gen_helper_iwmmxt_srll(cpu_M0, cpu_env, cpu_M0, tmp);
2096 gen_helper_iwmmxt_srlq(cpu_M0, cpu_env, cpu_M0, tmp);
2099 tcg_temp_free_i32(tmp);
2100 gen_op_iwmmxt_movq_wRn_M0(wrd);
2101 gen_op_iwmmxt_set_mup();
2102 gen_op_iwmmxt_set_cup();
2104 case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */
2105 case 0x014: case 0x414: case 0x814: case 0xc14:
2106 if (((insn >> 22) & 3) == 0)
2108 wrd = (insn >> 12) & 0xf;
2109 rd0 = (insn >> 16) & 0xf;
2110 gen_op_iwmmxt_movq_M0_wRn(rd0);
2111 tmp = tcg_temp_new_i32();
2112 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2113 tcg_temp_free_i32(tmp);
2116 switch ((insn >> 22) & 3) {
2118 gen_helper_iwmmxt_sraw(cpu_M0, cpu_env, cpu_M0, tmp);
2121 gen_helper_iwmmxt_sral(cpu_M0, cpu_env, cpu_M0, tmp);
2124 gen_helper_iwmmxt_sraq(cpu_M0, cpu_env, cpu_M0, tmp);
2127 tcg_temp_free_i32(tmp);
2128 gen_op_iwmmxt_movq_wRn_M0(wrd);
2129 gen_op_iwmmxt_set_mup();
2130 gen_op_iwmmxt_set_cup();
2132 case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */
2133 case 0x114: case 0x514: case 0x914: case 0xd14:
2134 if (((insn >> 22) & 3) == 0)
2136 wrd = (insn >> 12) & 0xf;
2137 rd0 = (insn >> 16) & 0xf;
2138 gen_op_iwmmxt_movq_M0_wRn(rd0);
2139 tmp = tcg_temp_new_i32();
2140 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2141 tcg_temp_free_i32(tmp);
2144 switch ((insn >> 22) & 3) {
2146 gen_helper_iwmmxt_sllw(cpu_M0, cpu_env, cpu_M0, tmp);
2149 gen_helper_iwmmxt_slll(cpu_M0, cpu_env, cpu_M0, tmp);
2152 gen_helper_iwmmxt_sllq(cpu_M0, cpu_env, cpu_M0, tmp);
2155 tcg_temp_free_i32(tmp);
2156 gen_op_iwmmxt_movq_wRn_M0(wrd);
2157 gen_op_iwmmxt_set_mup();
2158 gen_op_iwmmxt_set_cup();
2160 case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */
2161 case 0x314: case 0x714: case 0xb14: case 0xf14:
2162 if (((insn >> 22) & 3) == 0)
2164 wrd = (insn >> 12) & 0xf;
2165 rd0 = (insn >> 16) & 0xf;
2166 gen_op_iwmmxt_movq_M0_wRn(rd0);
2167 tmp = tcg_temp_new_i32();
2168 switch ((insn >> 22) & 3) {
2170 if (gen_iwmmxt_shift(insn, 0xf, tmp)) {
2171 tcg_temp_free_i32(tmp);
2174 gen_helper_iwmmxt_rorw(cpu_M0, cpu_env, cpu_M0, tmp);
2177 if (gen_iwmmxt_shift(insn, 0x1f, tmp)) {
2178 tcg_temp_free_i32(tmp);
2181 gen_helper_iwmmxt_rorl(cpu_M0, cpu_env, cpu_M0, tmp);
2184 if (gen_iwmmxt_shift(insn, 0x3f, tmp)) {
2185 tcg_temp_free_i32(tmp);
2188 gen_helper_iwmmxt_rorq(cpu_M0, cpu_env, cpu_M0, tmp);
2191 tcg_temp_free_i32(tmp);
2192 gen_op_iwmmxt_movq_wRn_M0(wrd);
2193 gen_op_iwmmxt_set_mup();
2194 gen_op_iwmmxt_set_cup();
2196 case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */
2197 case 0x916: case 0xb16: case 0xd16: case 0xf16:
2198 wrd = (insn >> 12) & 0xf;
2199 rd0 = (insn >> 16) & 0xf;
2200 rd1 = (insn >> 0) & 0xf;
2201 gen_op_iwmmxt_movq_M0_wRn(rd0);
2202 switch ((insn >> 22) & 3) {
2204 if (insn & (1 << 21))
2205 gen_op_iwmmxt_minsb_M0_wRn(rd1);
2207 gen_op_iwmmxt_minub_M0_wRn(rd1);
2210 if (insn & (1 << 21))
2211 gen_op_iwmmxt_minsw_M0_wRn(rd1);
2213 gen_op_iwmmxt_minuw_M0_wRn(rd1);
2216 if (insn & (1 << 21))
2217 gen_op_iwmmxt_minsl_M0_wRn(rd1);
2219 gen_op_iwmmxt_minul_M0_wRn(rd1);
2224 gen_op_iwmmxt_movq_wRn_M0(wrd);
2225 gen_op_iwmmxt_set_mup();
2227 case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */
2228 case 0x816: case 0xa16: case 0xc16: case 0xe16:
2229 wrd = (insn >> 12) & 0xf;
2230 rd0 = (insn >> 16) & 0xf;
2231 rd1 = (insn >> 0) & 0xf;
2232 gen_op_iwmmxt_movq_M0_wRn(rd0);
2233 switch ((insn >> 22) & 3) {
2235 if (insn & (1 << 21))
2236 gen_op_iwmmxt_maxsb_M0_wRn(rd1);
2238 gen_op_iwmmxt_maxub_M0_wRn(rd1);
2241 if (insn & (1 << 21))
2242 gen_op_iwmmxt_maxsw_M0_wRn(rd1);
2244 gen_op_iwmmxt_maxuw_M0_wRn(rd1);
2247 if (insn & (1 << 21))
2248 gen_op_iwmmxt_maxsl_M0_wRn(rd1);
2250 gen_op_iwmmxt_maxul_M0_wRn(rd1);
2255 gen_op_iwmmxt_movq_wRn_M0(wrd);
2256 gen_op_iwmmxt_set_mup();
2258 case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */
2259 case 0x402: case 0x502: case 0x602: case 0x702:
2260 wrd = (insn >> 12) & 0xf;
2261 rd0 = (insn >> 16) & 0xf;
2262 rd1 = (insn >> 0) & 0xf;
2263 gen_op_iwmmxt_movq_M0_wRn(rd0);
2264 tmp = tcg_const_i32((insn >> 20) & 3);
2265 iwmmxt_load_reg(cpu_V1, rd1);
2266 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp);
2267 tcg_temp_free_i32(tmp);
2268 gen_op_iwmmxt_movq_wRn_M0(wrd);
2269 gen_op_iwmmxt_set_mup();
2271 case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */
2272 case 0x41a: case 0x51a: case 0x61a: case 0x71a:
2273 case 0x81a: case 0x91a: case 0xa1a: case 0xb1a:
2274 case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a:
2275 wrd = (insn >> 12) & 0xf;
2276 rd0 = (insn >> 16) & 0xf;
2277 rd1 = (insn >> 0) & 0xf;
2278 gen_op_iwmmxt_movq_M0_wRn(rd0);
2279 switch ((insn >> 20) & 0xf) {
2281 gen_op_iwmmxt_subnb_M0_wRn(rd1);
2284 gen_op_iwmmxt_subub_M0_wRn(rd1);
2287 gen_op_iwmmxt_subsb_M0_wRn(rd1);
2290 gen_op_iwmmxt_subnw_M0_wRn(rd1);
2293 gen_op_iwmmxt_subuw_M0_wRn(rd1);
2296 gen_op_iwmmxt_subsw_M0_wRn(rd1);
2299 gen_op_iwmmxt_subnl_M0_wRn(rd1);
2302 gen_op_iwmmxt_subul_M0_wRn(rd1);
2305 gen_op_iwmmxt_subsl_M0_wRn(rd1);
2310 gen_op_iwmmxt_movq_wRn_M0(wrd);
2311 gen_op_iwmmxt_set_mup();
2312 gen_op_iwmmxt_set_cup();
2314 case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */
2315 case 0x41e: case 0x51e: case 0x61e: case 0x71e:
2316 case 0x81e: case 0x91e: case 0xa1e: case 0xb1e:
2317 case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e:
2318 wrd = (insn >> 12) & 0xf;
2319 rd0 = (insn >> 16) & 0xf;
2320 gen_op_iwmmxt_movq_M0_wRn(rd0);
2321 tmp = tcg_const_i32(((insn >> 16) & 0xf0) | (insn & 0x0f));
2322 gen_helper_iwmmxt_shufh(cpu_M0, cpu_env, cpu_M0, tmp);
2323 tcg_temp_free_i32(tmp);
2324 gen_op_iwmmxt_movq_wRn_M0(wrd);
2325 gen_op_iwmmxt_set_mup();
2326 gen_op_iwmmxt_set_cup();
2328 case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */
2329 case 0x418: case 0x518: case 0x618: case 0x718:
2330 case 0x818: case 0x918: case 0xa18: case 0xb18:
2331 case 0xc18: case 0xd18: case 0xe18: case 0xf18:
2332 wrd = (insn >> 12) & 0xf;
2333 rd0 = (insn >> 16) & 0xf;
2334 rd1 = (insn >> 0) & 0xf;
2335 gen_op_iwmmxt_movq_M0_wRn(rd0);
2336 switch ((insn >> 20) & 0xf) {
2338 gen_op_iwmmxt_addnb_M0_wRn(rd1);
2341 gen_op_iwmmxt_addub_M0_wRn(rd1);
2344 gen_op_iwmmxt_addsb_M0_wRn(rd1);
2347 gen_op_iwmmxt_addnw_M0_wRn(rd1);
2350 gen_op_iwmmxt_adduw_M0_wRn(rd1);
2353 gen_op_iwmmxt_addsw_M0_wRn(rd1);
2356 gen_op_iwmmxt_addnl_M0_wRn(rd1);
2359 gen_op_iwmmxt_addul_M0_wRn(rd1);
2362 gen_op_iwmmxt_addsl_M0_wRn(rd1);
2367 gen_op_iwmmxt_movq_wRn_M0(wrd);
2368 gen_op_iwmmxt_set_mup();
2369 gen_op_iwmmxt_set_cup();
2371 case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */
2372 case 0x408: case 0x508: case 0x608: case 0x708:
2373 case 0x808: case 0x908: case 0xa08: case 0xb08:
2374 case 0xc08: case 0xd08: case 0xe08: case 0xf08:
2375 if (!(insn & (1 << 20)) || ((insn >> 22) & 3) == 0)
2377 wrd = (insn >> 12) & 0xf;
2378 rd0 = (insn >> 16) & 0xf;
2379 rd1 = (insn >> 0) & 0xf;
2380 gen_op_iwmmxt_movq_M0_wRn(rd0);
2381 switch ((insn >> 22) & 3) {
2383 if (insn & (1 << 21))
2384 gen_op_iwmmxt_packsw_M0_wRn(rd1);
2386 gen_op_iwmmxt_packuw_M0_wRn(rd1);
2389 if (insn & (1 << 21))
2390 gen_op_iwmmxt_packsl_M0_wRn(rd1);
2392 gen_op_iwmmxt_packul_M0_wRn(rd1);
2395 if (insn & (1 << 21))
2396 gen_op_iwmmxt_packsq_M0_wRn(rd1);
2398 gen_op_iwmmxt_packuq_M0_wRn(rd1);
2401 gen_op_iwmmxt_movq_wRn_M0(wrd);
2402 gen_op_iwmmxt_set_mup();
2403 gen_op_iwmmxt_set_cup();
2405 case 0x201: case 0x203: case 0x205: case 0x207:
2406 case 0x209: case 0x20b: case 0x20d: case 0x20f:
2407 case 0x211: case 0x213: case 0x215: case 0x217:
2408 case 0x219: case 0x21b: case 0x21d: case 0x21f:
2409 wrd = (insn >> 5) & 0xf;
2410 rd0 = (insn >> 12) & 0xf;
2411 rd1 = (insn >> 0) & 0xf;
2412 if (rd0 == 0xf || rd1 == 0xf)
2414 gen_op_iwmmxt_movq_M0_wRn(wrd);
2415 tmp = load_reg(s, rd0);
2416 tmp2 = load_reg(s, rd1);
2417 switch ((insn >> 16) & 0xf) {
2418 case 0x0: /* TMIA */
2419 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2);
2421 case 0x8: /* TMIAPH */
2422 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2);
2424 case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */
2425 if (insn & (1 << 16))
2426 tcg_gen_shri_i32(tmp, tmp, 16);
2427 if (insn & (1 << 17))
2428 tcg_gen_shri_i32(tmp2, tmp2, 16);
2429 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2);
2432 tcg_temp_free_i32(tmp2);
2433 tcg_temp_free_i32(tmp);
2436 tcg_temp_free_i32(tmp2);
2437 tcg_temp_free_i32(tmp);
2438 gen_op_iwmmxt_movq_wRn_M0(wrd);
2439 gen_op_iwmmxt_set_mup();
2448 /* Disassemble an XScale DSP instruction. Returns nonzero if an error occurred
2449 (ie. an undefined instruction). */
2450 static int disas_dsp_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
2452 int acc, rd0, rd1, rdhi, rdlo;
2455 if ((insn & 0x0ff00f10) == 0x0e200010) {
2456 /* Multiply with Internal Accumulate Format */
2457 rd0 = (insn >> 12) & 0xf;
2459 acc = (insn >> 5) & 7;
2464 tmp = load_reg(s, rd0);
2465 tmp2 = load_reg(s, rd1);
2466 switch ((insn >> 16) & 0xf) {
2468 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2);
2470 case 0x8: /* MIAPH */
2471 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2);
2473 case 0xc: /* MIABB */
2474 case 0xd: /* MIABT */
2475 case 0xe: /* MIATB */
2476 case 0xf: /* MIATT */
2477 if (insn & (1 << 16))
2478 tcg_gen_shri_i32(tmp, tmp, 16);
2479 if (insn & (1 << 17))
2480 tcg_gen_shri_i32(tmp2, tmp2, 16);
2481 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2);
2486 tcg_temp_free_i32(tmp2);
2487 tcg_temp_free_i32(tmp);
2489 gen_op_iwmmxt_movq_wRn_M0(acc);
2493 if ((insn & 0x0fe00ff8) == 0x0c400000) {
2494 /* Internal Accumulator Access Format */
2495 rdhi = (insn >> 16) & 0xf;
2496 rdlo = (insn >> 12) & 0xf;
2502 if (insn & ARM_CP_RW_BIT) { /* MRA */
2503 iwmmxt_load_reg(cpu_V0, acc);
2504 tcg_gen_trunc_i64_i32(cpu_R[rdlo], cpu_V0);
2505 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
2506 tcg_gen_trunc_i64_i32(cpu_R[rdhi], cpu_V0);
2507 tcg_gen_andi_i32(cpu_R[rdhi], cpu_R[rdhi], (1 << (40 - 32)) - 1);
2509 tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]);
2510 iwmmxt_store_reg(cpu_V0, acc);
2518 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
2519 #define VFP_SREG(insn, bigbit, smallbit) \
2520 ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
2521 #define VFP_DREG(reg, insn, bigbit, smallbit) do { \
2522 if (arm_feature(env, ARM_FEATURE_VFP3)) { \
2523 reg = (((insn) >> (bigbit)) & 0x0f) \
2524 | (((insn) >> ((smallbit) - 4)) & 0x10); \
2526 if (insn & (1 << (smallbit))) \
2528 reg = ((insn) >> (bigbit)) & 0x0f; \
2531 #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
2532 #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
2533 #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7)
2534 #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7)
2535 #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5)
2536 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5)
2538 /* Move between integer and VFP cores. */
2539 static TCGv_i32 gen_vfp_mrs(void)
2541 TCGv_i32 tmp = tcg_temp_new_i32();
2542 tcg_gen_mov_i32(tmp, cpu_F0s);
2546 static void gen_vfp_msr(TCGv_i32 tmp)
2548 tcg_gen_mov_i32(cpu_F0s, tmp);
2549 tcg_temp_free_i32(tmp);
2552 static void gen_neon_dup_u8(TCGv_i32 var, int shift)
2554 TCGv_i32 tmp = tcg_temp_new_i32();
2556 tcg_gen_shri_i32(var, var, shift);
2557 tcg_gen_ext8u_i32(var, var);
2558 tcg_gen_shli_i32(tmp, var, 8);
2559 tcg_gen_or_i32(var, var, tmp);
2560 tcg_gen_shli_i32(tmp, var, 16);
2561 tcg_gen_or_i32(var, var, tmp);
2562 tcg_temp_free_i32(tmp);
2565 static void gen_neon_dup_low16(TCGv_i32 var)
2567 TCGv_i32 tmp = tcg_temp_new_i32();
2568 tcg_gen_ext16u_i32(var, var);
2569 tcg_gen_shli_i32(tmp, var, 16);
2570 tcg_gen_or_i32(var, var, tmp);
2571 tcg_temp_free_i32(tmp);
2574 static void gen_neon_dup_high16(TCGv_i32 var)
2576 TCGv_i32 tmp = tcg_temp_new_i32();
2577 tcg_gen_andi_i32(var, var, 0xffff0000);
2578 tcg_gen_shri_i32(tmp, var, 16);
2579 tcg_gen_or_i32(var, var, tmp);
2580 tcg_temp_free_i32(tmp);
2583 static TCGv_i32 gen_load_and_replicate(DisasContext *s, TCGv_i32 addr, int size)
2585 /* Load a single Neon element and replicate into a 32 bit TCG reg */
2586 TCGv_i32 tmp = tcg_temp_new_i32();
2589 gen_aa32_ld8u(tmp, addr, IS_USER(s));
2590 gen_neon_dup_u8(tmp, 0);
2593 gen_aa32_ld16u(tmp, addr, IS_USER(s));
2594 gen_neon_dup_low16(tmp);
2597 gen_aa32_ld32u(tmp, addr, IS_USER(s));
2599 default: /* Avoid compiler warnings. */
2605 static int handle_vsel(uint32_t insn, uint32_t rd, uint32_t rn, uint32_t rm,
2608 uint32_t cc = extract32(insn, 20, 2);
2611 TCGv_i64 frn, frm, dest;
2612 TCGv_i64 tmp, zero, zf, nf, vf;
2614 zero = tcg_const_i64(0);
2616 frn = tcg_temp_new_i64();
2617 frm = tcg_temp_new_i64();
2618 dest = tcg_temp_new_i64();
2620 zf = tcg_temp_new_i64();
2621 nf = tcg_temp_new_i64();
2622 vf = tcg_temp_new_i64();
2624 tcg_gen_extu_i32_i64(zf, cpu_ZF);
2625 tcg_gen_ext_i32_i64(nf, cpu_NF);
2626 tcg_gen_ext_i32_i64(vf, cpu_VF);
2628 tcg_gen_ld_f64(frn, cpu_env, vfp_reg_offset(dp, rn));
2629 tcg_gen_ld_f64(frm, cpu_env, vfp_reg_offset(dp, rm));
2632 tcg_gen_movcond_i64(TCG_COND_EQ, dest, zf, zero,
2636 tcg_gen_movcond_i64(TCG_COND_LT, dest, vf, zero,
2639 case 2: /* ge: N == V -> N ^ V == 0 */
2640 tmp = tcg_temp_new_i64();
2641 tcg_gen_xor_i64(tmp, vf, nf);
2642 tcg_gen_movcond_i64(TCG_COND_GE, dest, tmp, zero,
2644 tcg_temp_free_i64(tmp);
2646 case 3: /* gt: !Z && N == V */
2647 tcg_gen_movcond_i64(TCG_COND_NE, dest, zf, zero,
2649 tmp = tcg_temp_new_i64();
2650 tcg_gen_xor_i64(tmp, vf, nf);
2651 tcg_gen_movcond_i64(TCG_COND_GE, dest, tmp, zero,
2653 tcg_temp_free_i64(tmp);
2656 tcg_gen_st_f64(dest, cpu_env, vfp_reg_offset(dp, rd));
2657 tcg_temp_free_i64(frn);
2658 tcg_temp_free_i64(frm);
2659 tcg_temp_free_i64(dest);
2661 tcg_temp_free_i64(zf);
2662 tcg_temp_free_i64(nf);
2663 tcg_temp_free_i64(vf);
2665 tcg_temp_free_i64(zero);
2667 TCGv_i32 frn, frm, dest;
2670 zero = tcg_const_i32(0);
2672 frn = tcg_temp_new_i32();
2673 frm = tcg_temp_new_i32();
2674 dest = tcg_temp_new_i32();
2675 tcg_gen_ld_f32(frn, cpu_env, vfp_reg_offset(dp, rn));
2676 tcg_gen_ld_f32(frm, cpu_env, vfp_reg_offset(dp, rm));
2679 tcg_gen_movcond_i32(TCG_COND_EQ, dest, cpu_ZF, zero,
2683 tcg_gen_movcond_i32(TCG_COND_LT, dest, cpu_VF, zero,
2686 case 2: /* ge: N == V -> N ^ V == 0 */
2687 tmp = tcg_temp_new_i32();
2688 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
2689 tcg_gen_movcond_i32(TCG_COND_GE, dest, tmp, zero,
2691 tcg_temp_free_i32(tmp);
2693 case 3: /* gt: !Z && N == V */
2694 tcg_gen_movcond_i32(TCG_COND_NE, dest, cpu_ZF, zero,
2696 tmp = tcg_temp_new_i32();
2697 tcg_gen_xor_i32(tmp, cpu_VF, cpu_NF);
2698 tcg_gen_movcond_i32(TCG_COND_GE, dest, tmp, zero,
2700 tcg_temp_free_i32(tmp);
2703 tcg_gen_st_f32(dest, cpu_env, vfp_reg_offset(dp, rd));
2704 tcg_temp_free_i32(frn);
2705 tcg_temp_free_i32(frm);
2706 tcg_temp_free_i32(dest);
2708 tcg_temp_free_i32(zero);
2714 static int handle_vminmaxnm(uint32_t insn, uint32_t rd, uint32_t rn,
2715 uint32_t rm, uint32_t dp)
2717 uint32_t vmin = extract32(insn, 6, 1);
2718 TCGv_ptr fpst = get_fpstatus_ptr(0);
2721 TCGv_i64 frn, frm, dest;
2723 frn = tcg_temp_new_i64();
2724 frm = tcg_temp_new_i64();
2725 dest = tcg_temp_new_i64();
2727 tcg_gen_ld_f64(frn, cpu_env, vfp_reg_offset(dp, rn));
2728 tcg_gen_ld_f64(frm, cpu_env, vfp_reg_offset(dp, rm));
2730 gen_helper_vfp_minnumd(dest, frn, frm, fpst);
2732 gen_helper_vfp_maxnumd(dest, frn, frm, fpst);
2734 tcg_gen_st_f64(dest, cpu_env, vfp_reg_offset(dp, rd));
2735 tcg_temp_free_i64(frn);
2736 tcg_temp_free_i64(frm);
2737 tcg_temp_free_i64(dest);
2739 TCGv_i32 frn, frm, dest;
2741 frn = tcg_temp_new_i32();
2742 frm = tcg_temp_new_i32();
2743 dest = tcg_temp_new_i32();
2745 tcg_gen_ld_f32(frn, cpu_env, vfp_reg_offset(dp, rn));
2746 tcg_gen_ld_f32(frm, cpu_env, vfp_reg_offset(dp, rm));
2748 gen_helper_vfp_minnums(dest, frn, frm, fpst);
2750 gen_helper_vfp_maxnums(dest, frn, frm, fpst);
2752 tcg_gen_st_f32(dest, cpu_env, vfp_reg_offset(dp, rd));
2753 tcg_temp_free_i32(frn);
2754 tcg_temp_free_i32(frm);
2755 tcg_temp_free_i32(dest);
2758 tcg_temp_free_ptr(fpst);
2762 static int handle_vrint(uint32_t insn, uint32_t rd, uint32_t rm, uint32_t dp,
2765 TCGv_ptr fpst = get_fpstatus_ptr(0);
2768 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rounding));
2769 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
2774 tcg_op = tcg_temp_new_i64();
2775 tcg_res = tcg_temp_new_i64();
2776 tcg_gen_ld_f64(tcg_op, cpu_env, vfp_reg_offset(dp, rm));
2777 gen_helper_rintd(tcg_res, tcg_op, fpst);
2778 tcg_gen_st_f64(tcg_res, cpu_env, vfp_reg_offset(dp, rd));
2779 tcg_temp_free_i64(tcg_op);
2780 tcg_temp_free_i64(tcg_res);
2784 tcg_op = tcg_temp_new_i32();
2785 tcg_res = tcg_temp_new_i32();
2786 tcg_gen_ld_f32(tcg_op, cpu_env, vfp_reg_offset(dp, rm));
2787 gen_helper_rints(tcg_res, tcg_op, fpst);
2788 tcg_gen_st_f32(tcg_res, cpu_env, vfp_reg_offset(dp, rd));
2789 tcg_temp_free_i32(tcg_op);
2790 tcg_temp_free_i32(tcg_res);
2793 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
2794 tcg_temp_free_i32(tcg_rmode);
2796 tcg_temp_free_ptr(fpst);
2800 static int handle_vcvt(uint32_t insn, uint32_t rd, uint32_t rm, uint32_t dp,
2803 bool is_signed = extract32(insn, 7, 1);
2804 TCGv_ptr fpst = get_fpstatus_ptr(0);
2805 TCGv_i32 tcg_rmode, tcg_shift;
2807 tcg_shift = tcg_const_i32(0);
2809 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rounding));
2810 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
2813 TCGv_i64 tcg_double, tcg_res;
2815 /* Rd is encoded as a single precision register even when the source
2816 * is double precision.
2818 rd = ((rd << 1) & 0x1e) | ((rd >> 4) & 0x1);
2819 tcg_double = tcg_temp_new_i64();
2820 tcg_res = tcg_temp_new_i64();
2821 tcg_tmp = tcg_temp_new_i32();
2822 tcg_gen_ld_f64(tcg_double, cpu_env, vfp_reg_offset(1, rm));
2824 gen_helper_vfp_tosld(tcg_res, tcg_double, tcg_shift, fpst);
2826 gen_helper_vfp_tould(tcg_res, tcg_double, tcg_shift, fpst);
2828 tcg_gen_trunc_i64_i32(tcg_tmp, tcg_res);
2829 tcg_gen_st_f32(tcg_tmp, cpu_env, vfp_reg_offset(0, rd));
2830 tcg_temp_free_i32(tcg_tmp);
2831 tcg_temp_free_i64(tcg_res);
2832 tcg_temp_free_i64(tcg_double);
2834 TCGv_i32 tcg_single, tcg_res;
2835 tcg_single = tcg_temp_new_i32();
2836 tcg_res = tcg_temp_new_i32();
2837 tcg_gen_ld_f32(tcg_single, cpu_env, vfp_reg_offset(0, rm));
2839 gen_helper_vfp_tosls(tcg_res, tcg_single, tcg_shift, fpst);
2841 gen_helper_vfp_touls(tcg_res, tcg_single, tcg_shift, fpst);
2843 tcg_gen_st_f32(tcg_res, cpu_env, vfp_reg_offset(0, rd));
2844 tcg_temp_free_i32(tcg_res);
2845 tcg_temp_free_i32(tcg_single);
2848 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
2849 tcg_temp_free_i32(tcg_rmode);
2851 tcg_temp_free_i32(tcg_shift);
2853 tcg_temp_free_ptr(fpst);
2858 /* Table for converting the most common AArch32 encoding of
2859 * rounding mode to arm_fprounding order (which matches the
2860 * common AArch64 order); see ARM ARM pseudocode FPDecodeRM().
2862 static const uint8_t fp_decode_rm[] = {
2869 static int disas_vfp_v8_insn(CPUARMState *env, DisasContext *s, uint32_t insn)
2871 uint32_t rd, rn, rm, dp = extract32(insn, 8, 1);
2873 if (!arm_feature(env, ARM_FEATURE_V8)) {
2878 VFP_DREG_D(rd, insn);
2879 VFP_DREG_N(rn, insn);
2880 VFP_DREG_M(rm, insn);
2882 rd = VFP_SREG_D(insn);
2883 rn = VFP_SREG_N(insn);
2884 rm = VFP_SREG_M(insn);
2887 if ((insn & 0x0f800e50) == 0x0e000a00) {
2888 return handle_vsel(insn, rd, rn, rm, dp);
2889 } else if ((insn & 0x0fb00e10) == 0x0e800a00) {
2890 return handle_vminmaxnm(insn, rd, rn, rm, dp);
2891 } else if ((insn & 0x0fbc0ed0) == 0x0eb80a40) {
2892 /* VRINTA, VRINTN, VRINTP, VRINTM */
2893 int rounding = fp_decode_rm[extract32(insn, 16, 2)];
2894 return handle_vrint(insn, rd, rm, dp, rounding);
2895 } else if ((insn & 0x0fbc0e50) == 0x0ebc0a40) {
2896 /* VCVTA, VCVTN, VCVTP, VCVTM */
2897 int rounding = fp_decode_rm[extract32(insn, 16, 2)];
2898 return handle_vcvt(insn, rd, rm, dp, rounding);
2903 /* Disassemble a VFP instruction. Returns nonzero if an error occurred
2904 (ie. an undefined instruction). */
2905 static int disas_vfp_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
2907 uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask;
2913 if (!arm_feature(env, ARM_FEATURE_VFP))
2916 if (!s->vfp_enabled) {
2917 /* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
2918 if ((insn & 0x0fe00fff) != 0x0ee00a10)
2920 rn = (insn >> 16) & 0xf;
2921 if (rn != ARM_VFP_FPSID && rn != ARM_VFP_FPEXC
2922 && rn != ARM_VFP_MVFR1 && rn != ARM_VFP_MVFR0)
2926 if (extract32(insn, 28, 4) == 0xf) {
2927 /* Encodings with T=1 (Thumb) or unconditional (ARM):
2928 * only used in v8 and above.
2930 return disas_vfp_v8_insn(env, s, insn);
2933 dp = ((insn & 0xf00) == 0xb00);
2934 switch ((insn >> 24) & 0xf) {
2936 if (insn & (1 << 4)) {
2937 /* single register transfer */
2938 rd = (insn >> 12) & 0xf;
2943 VFP_DREG_N(rn, insn);
2946 if (insn & 0x00c00060
2947 && !arm_feature(env, ARM_FEATURE_NEON))
2950 pass = (insn >> 21) & 1;
2951 if (insn & (1 << 22)) {
2953 offset = ((insn >> 5) & 3) * 8;
2954 } else if (insn & (1 << 5)) {
2956 offset = (insn & (1 << 6)) ? 16 : 0;
2961 if (insn & ARM_CP_RW_BIT) {
2963 tmp = neon_load_reg(rn, pass);
2967 tcg_gen_shri_i32(tmp, tmp, offset);
2968 if (insn & (1 << 23))
2974 if (insn & (1 << 23)) {
2976 tcg_gen_shri_i32(tmp, tmp, 16);
2982 tcg_gen_sari_i32(tmp, tmp, 16);
2991 store_reg(s, rd, tmp);
2994 tmp = load_reg(s, rd);
2995 if (insn & (1 << 23)) {
2998 gen_neon_dup_u8(tmp, 0);
2999 } else if (size == 1) {
3000 gen_neon_dup_low16(tmp);
3002 for (n = 0; n <= pass * 2; n++) {
3003 tmp2 = tcg_temp_new_i32();
3004 tcg_gen_mov_i32(tmp2, tmp);
3005 neon_store_reg(rn, n, tmp2);
3007 neon_store_reg(rn, n, tmp);
3012 tmp2 = neon_load_reg(rn, pass);
3013 tcg_gen_deposit_i32(tmp, tmp2, tmp, offset, 8);
3014 tcg_temp_free_i32(tmp2);
3017 tmp2 = neon_load_reg(rn, pass);
3018 tcg_gen_deposit_i32(tmp, tmp2, tmp, offset, 16);
3019 tcg_temp_free_i32(tmp2);
3024 neon_store_reg(rn, pass, tmp);
3028 if ((insn & 0x6f) != 0x00)
3030 rn = VFP_SREG_N(insn);
3031 if (insn & ARM_CP_RW_BIT) {
3033 if (insn & (1 << 21)) {
3034 /* system register */
3039 /* VFP2 allows access to FSID from userspace.
3040 VFP3 restricts all id registers to privileged
3043 && arm_feature(env, ARM_FEATURE_VFP3))
3045 tmp = load_cpu_field(vfp.xregs[rn]);
3050 tmp = load_cpu_field(vfp.xregs[rn]);
3052 case ARM_VFP_FPINST:
3053 case ARM_VFP_FPINST2:
3054 /* Not present in VFP3. */
3056 || arm_feature(env, ARM_FEATURE_VFP3))
3058 tmp = load_cpu_field(vfp.xregs[rn]);
3062 tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]);
3063 tcg_gen_andi_i32(tmp, tmp, 0xf0000000);
3065 tmp = tcg_temp_new_i32();
3066 gen_helper_vfp_get_fpscr(tmp, cpu_env);
3072 || !arm_feature(env, ARM_FEATURE_MVFR))
3074 tmp = load_cpu_field(vfp.xregs[rn]);
3080 gen_mov_F0_vreg(0, rn);
3081 tmp = gen_vfp_mrs();
3084 /* Set the 4 flag bits in the CPSR. */
3086 tcg_temp_free_i32(tmp);
3088 store_reg(s, rd, tmp);
3092 if (insn & (1 << 21)) {
3094 /* system register */
3099 /* Writes are ignored. */
3102 tmp = load_reg(s, rd);
3103 gen_helper_vfp_set_fpscr(cpu_env, tmp);
3104 tcg_temp_free_i32(tmp);
3110 /* TODO: VFP subarchitecture support.
3111 * For now, keep the EN bit only */
3112 tmp = load_reg(s, rd);
3113 tcg_gen_andi_i32(tmp, tmp, 1 << 30);
3114 store_cpu_field(tmp, vfp.xregs[rn]);
3117 case ARM_VFP_FPINST:
3118 case ARM_VFP_FPINST2:
3119 tmp = load_reg(s, rd);
3120 store_cpu_field(tmp, vfp.xregs[rn]);
3126 tmp = load_reg(s, rd);
3128 gen_mov_vreg_F0(0, rn);
3133 /* data processing */
3134 /* The opcode is in bits 23, 21, 20 and 6. */
3135 op = ((insn >> 20) & 8) | ((insn >> 19) & 6) | ((insn >> 6) & 1);
3139 rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
3141 /* rn is register number */
3142 VFP_DREG_N(rn, insn);
3145 if (op == 15 && (rn == 15 || ((rn & 0x1c) == 0x18) ||
3146 ((rn & 0x1e) == 0x6))) {
3147 /* Integer or single/half precision destination. */
3148 rd = VFP_SREG_D(insn);
3150 VFP_DREG_D(rd, insn);
3153 (((rn & 0x1c) == 0x10) || ((rn & 0x14) == 0x14) ||
3154 ((rn & 0x1e) == 0x4))) {
3155 /* VCVT from int or half precision is always from S reg
3156 * regardless of dp bit. VCVT with immediate frac_bits
3157 * has same format as SREG_M.
3159 rm = VFP_SREG_M(insn);
3161 VFP_DREG_M(rm, insn);
3164 rn = VFP_SREG_N(insn);
3165 if (op == 15 && rn == 15) {
3166 /* Double precision destination. */
3167 VFP_DREG_D(rd, insn);
3169 rd = VFP_SREG_D(insn);
3171 /* NB that we implicitly rely on the encoding for the frac_bits
3172 * in VCVT of fixed to float being the same as that of an SREG_M
3174 rm = VFP_SREG_M(insn);
3177 veclen = s->vec_len;
3178 if (op == 15 && rn > 3)
3181 /* Shut up compiler warnings. */
3192 /* Figure out what type of vector operation this is. */
3193 if ((rd & bank_mask) == 0) {
3198 delta_d = (s->vec_stride >> 1) + 1;
3200 delta_d = s->vec_stride + 1;
3202 if ((rm & bank_mask) == 0) {
3203 /* mixed scalar/vector */
3212 /* Load the initial operands. */
3217 /* Integer source */
3218 gen_mov_F0_vreg(0, rm);
3223 gen_mov_F0_vreg(dp, rd);
3224 gen_mov_F1_vreg(dp, rm);
3228 /* Compare with zero */
3229 gen_mov_F0_vreg(dp, rd);
3240 /* Source and destination the same. */
3241 gen_mov_F0_vreg(dp, rd);
3247 /* VCVTB, VCVTT: only present with the halfprec extension
3248 * UNPREDICTABLE if bit 8 is set prior to ARMv8
3249 * (we choose to UNDEF)
3251 if ((dp && !arm_feature(env, ARM_FEATURE_V8)) ||
3252 !arm_feature(env, ARM_FEATURE_VFP_FP16)) {
3255 if (!extract32(rn, 1, 1)) {
3256 /* Half precision source. */
3257 gen_mov_F0_vreg(0, rm);
3260 /* Otherwise fall through */
3262 /* One source operand. */
3263 gen_mov_F0_vreg(dp, rm);
3267 /* Two source operands. */
3268 gen_mov_F0_vreg(dp, rn);
3269 gen_mov_F1_vreg(dp, rm);
3273 /* Perform the calculation. */
3275 case 0: /* VMLA: fd + (fn * fm) */
3276 /* Note that order of inputs to the add matters for NaNs */
3278 gen_mov_F0_vreg(dp, rd);
3281 case 1: /* VMLS: fd + -(fn * fm) */
3284 gen_mov_F0_vreg(dp, rd);
3287 case 2: /* VNMLS: -fd + (fn * fm) */
3288 /* Note that it isn't valid to replace (-A + B) with (B - A)
3289 * or similar plausible looking simplifications
3290 * because this will give wrong results for NaNs.
3293 gen_mov_F0_vreg(dp, rd);
3297 case 3: /* VNMLA: -fd + -(fn * fm) */
3300 gen_mov_F0_vreg(dp, rd);
3304 case 4: /* mul: fn * fm */
3307 case 5: /* nmul: -(fn * fm) */
3311 case 6: /* add: fn + fm */
3314 case 7: /* sub: fn - fm */
3317 case 8: /* div: fn / fm */
3320 case 10: /* VFNMA : fd = muladd(-fd, fn, fm) */
3321 case 11: /* VFNMS : fd = muladd(-fd, -fn, fm) */
3322 case 12: /* VFMA : fd = muladd( fd, fn, fm) */
3323 case 13: /* VFMS : fd = muladd( fd, -fn, fm) */
3324 /* These are fused multiply-add, and must be done as one
3325 * floating point operation with no rounding between the
3326 * multiplication and addition steps.
3327 * NB that doing the negations here as separate steps is
3328 * correct : an input NaN should come out with its sign bit
3329 * flipped if it is a negated-input.
3331 if (!arm_feature(env, ARM_FEATURE_VFP4)) {
3339 gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
3341 frd = tcg_temp_new_i64();
3342 tcg_gen_ld_f64(frd, cpu_env, vfp_reg_offset(dp, rd));
3345 gen_helper_vfp_negd(frd, frd);
3347 fpst = get_fpstatus_ptr(0);
3348 gen_helper_vfp_muladdd(cpu_F0d, cpu_F0d,
3349 cpu_F1d, frd, fpst);
3350 tcg_temp_free_ptr(fpst);
3351 tcg_temp_free_i64(frd);
3357 gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
3359 frd = tcg_temp_new_i32();
3360 tcg_gen_ld_f32(frd, cpu_env, vfp_reg_offset(dp, rd));
3362 gen_helper_vfp_negs(frd, frd);
3364 fpst = get_fpstatus_ptr(0);
3365 gen_helper_vfp_muladds(cpu_F0s, cpu_F0s,
3366 cpu_F1s, frd, fpst);
3367 tcg_temp_free_ptr(fpst);
3368 tcg_temp_free_i32(frd);
3371 case 14: /* fconst */
3372 if (!arm_feature(env, ARM_FEATURE_VFP3))
3375 n = (insn << 12) & 0x80000000;
3376 i = ((insn >> 12) & 0x70) | (insn & 0xf);
3383 tcg_gen_movi_i64(cpu_F0d, ((uint64_t)n) << 32);
3390 tcg_gen_movi_i32(cpu_F0s, n);
3393 case 15: /* extension space */
3407 case 4: /* vcvtb.f32.f16, vcvtb.f64.f16 */
3408 tmp = gen_vfp_mrs();
3409 tcg_gen_ext16u_i32(tmp, tmp);
3411 gen_helper_vfp_fcvt_f16_to_f64(cpu_F0d, tmp,
3414 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp,
3417 tcg_temp_free_i32(tmp);
3419 case 5: /* vcvtt.f32.f16, vcvtt.f64.f16 */
3420 tmp = gen_vfp_mrs();
3421 tcg_gen_shri_i32(tmp, tmp, 16);
3423 gen_helper_vfp_fcvt_f16_to_f64(cpu_F0d, tmp,
3426 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp,
3429 tcg_temp_free_i32(tmp);
3431 case 6: /* vcvtb.f16.f32, vcvtb.f16.f64 */
3432 tmp = tcg_temp_new_i32();
3434 gen_helper_vfp_fcvt_f64_to_f16(tmp, cpu_F0d,
3437 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s,
3440 gen_mov_F0_vreg(0, rd);
3441 tmp2 = gen_vfp_mrs();
3442 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
3443 tcg_gen_or_i32(tmp, tmp, tmp2);
3444 tcg_temp_free_i32(tmp2);
3447 case 7: /* vcvtt.f16.f32, vcvtt.f16.f64 */
3448 tmp = tcg_temp_new_i32();
3450 gen_helper_vfp_fcvt_f64_to_f16(tmp, cpu_F0d,
3453 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s,
3456 tcg_gen_shli_i32(tmp, tmp, 16);
3457 gen_mov_F0_vreg(0, rd);
3458 tmp2 = gen_vfp_mrs();
3459 tcg_gen_ext16u_i32(tmp2, tmp2);
3460 tcg_gen_or_i32(tmp, tmp, tmp2);
3461 tcg_temp_free_i32(tmp2);
3473 case 11: /* cmpez */
3477 case 12: /* vrintr */
3479 TCGv_ptr fpst = get_fpstatus_ptr(0);
3481 gen_helper_rintd(cpu_F0d, cpu_F0d, fpst);
3483 gen_helper_rints(cpu_F0s, cpu_F0s, fpst);
3485 tcg_temp_free_ptr(fpst);
3488 case 13: /* vrintz */
3490 TCGv_ptr fpst = get_fpstatus_ptr(0);
3492 tcg_rmode = tcg_const_i32(float_round_to_zero);
3493 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3495 gen_helper_rintd(cpu_F0d, cpu_F0d, fpst);
3497 gen_helper_rints(cpu_F0s, cpu_F0s, fpst);
3499 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3500 tcg_temp_free_i32(tcg_rmode);
3501 tcg_temp_free_ptr(fpst);
3504 case 14: /* vrintx */
3506 TCGv_ptr fpst = get_fpstatus_ptr(0);
3508 gen_helper_rintd_exact(cpu_F0d, cpu_F0d, fpst);
3510 gen_helper_rints_exact(cpu_F0s, cpu_F0s, fpst);
3512 tcg_temp_free_ptr(fpst);
3515 case 15: /* single<->double conversion */
3517 gen_helper_vfp_fcvtsd(cpu_F0s, cpu_F0d, cpu_env);
3519 gen_helper_vfp_fcvtds(cpu_F0d, cpu_F0s, cpu_env);
3521 case 16: /* fuito */
3522 gen_vfp_uito(dp, 0);
3524 case 17: /* fsito */
3525 gen_vfp_sito(dp, 0);
3527 case 20: /* fshto */
3528 if (!arm_feature(env, ARM_FEATURE_VFP3))
3530 gen_vfp_shto(dp, 16 - rm, 0);
3532 case 21: /* fslto */
3533 if (!arm_feature(env, ARM_FEATURE_VFP3))
3535 gen_vfp_slto(dp, 32 - rm, 0);
3537 case 22: /* fuhto */
3538 if (!arm_feature(env, ARM_FEATURE_VFP3))
3540 gen_vfp_uhto(dp, 16 - rm, 0);
3542 case 23: /* fulto */
3543 if (!arm_feature(env, ARM_FEATURE_VFP3))
3545 gen_vfp_ulto(dp, 32 - rm, 0);
3547 case 24: /* ftoui */
3548 gen_vfp_toui(dp, 0);
3550 case 25: /* ftouiz */
3551 gen_vfp_touiz(dp, 0);
3553 case 26: /* ftosi */
3554 gen_vfp_tosi(dp, 0);
3556 case 27: /* ftosiz */
3557 gen_vfp_tosiz(dp, 0);
3559 case 28: /* ftosh */
3560 if (!arm_feature(env, ARM_FEATURE_VFP3))
3562 gen_vfp_tosh(dp, 16 - rm, 0);
3564 case 29: /* ftosl */
3565 if (!arm_feature(env, ARM_FEATURE_VFP3))
3567 gen_vfp_tosl(dp, 32 - rm, 0);
3569 case 30: /* ftouh */
3570 if (!arm_feature(env, ARM_FEATURE_VFP3))
3572 gen_vfp_touh(dp, 16 - rm, 0);
3574 case 31: /* ftoul */
3575 if (!arm_feature(env, ARM_FEATURE_VFP3))
3577 gen_vfp_toul(dp, 32 - rm, 0);
3579 default: /* undefined */
3583 default: /* undefined */
3587 /* Write back the result. */
3588 if (op == 15 && (rn >= 8 && rn <= 11)) {
3589 /* Comparison, do nothing. */
3590 } else if (op == 15 && dp && ((rn & 0x1c) == 0x18 ||
3591 (rn & 0x1e) == 0x6)) {
3592 /* VCVT double to int: always integer result.
3593 * VCVT double to half precision is always a single
3596 gen_mov_vreg_F0(0, rd);
3597 } else if (op == 15 && rn == 15) {
3599 gen_mov_vreg_F0(!dp, rd);
3601 gen_mov_vreg_F0(dp, rd);
3604 /* break out of the loop if we have finished */
3608 if (op == 15 && delta_m == 0) {
3609 /* single source one-many */
3611 rd = ((rd + delta_d) & (bank_mask - 1))
3613 gen_mov_vreg_F0(dp, rd);
3617 /* Setup the next operands. */
3619 rd = ((rd + delta_d) & (bank_mask - 1))
3623 /* One source operand. */
3624 rm = ((rm + delta_m) & (bank_mask - 1))
3626 gen_mov_F0_vreg(dp, rm);
3628 /* Two source operands. */
3629 rn = ((rn + delta_d) & (bank_mask - 1))
3631 gen_mov_F0_vreg(dp, rn);
3633 rm = ((rm + delta_m) & (bank_mask - 1))
3635 gen_mov_F1_vreg(dp, rm);
3643 if ((insn & 0x03e00000) == 0x00400000) {
3644 /* two-register transfer */
3645 rn = (insn >> 16) & 0xf;
3646 rd = (insn >> 12) & 0xf;
3648 VFP_DREG_M(rm, insn);
3650 rm = VFP_SREG_M(insn);
3653 if (insn & ARM_CP_RW_BIT) {
3656 gen_mov_F0_vreg(0, rm * 2);
3657 tmp = gen_vfp_mrs();
3658 store_reg(s, rd, tmp);
3659 gen_mov_F0_vreg(0, rm * 2 + 1);
3660 tmp = gen_vfp_mrs();
3661 store_reg(s, rn, tmp);
3663 gen_mov_F0_vreg(0, rm);
3664 tmp = gen_vfp_mrs();
3665 store_reg(s, rd, tmp);
3666 gen_mov_F0_vreg(0, rm + 1);
3667 tmp = gen_vfp_mrs();
3668 store_reg(s, rn, tmp);
3673 tmp = load_reg(s, rd);
3675 gen_mov_vreg_F0(0, rm * 2);
3676 tmp = load_reg(s, rn);
3678 gen_mov_vreg_F0(0, rm * 2 + 1);
3680 tmp = load_reg(s, rd);
3682 gen_mov_vreg_F0(0, rm);
3683 tmp = load_reg(s, rn);
3685 gen_mov_vreg_F0(0, rm + 1);
3690 rn = (insn >> 16) & 0xf;
3692 VFP_DREG_D(rd, insn);
3694 rd = VFP_SREG_D(insn);
3695 if ((insn & 0x01200000) == 0x01000000) {
3696 /* Single load/store */
3697 offset = (insn & 0xff) << 2;
3698 if ((insn & (1 << 23)) == 0)
3700 if (s->thumb && rn == 15) {
3701 /* This is actually UNPREDICTABLE */
3702 addr = tcg_temp_new_i32();
3703 tcg_gen_movi_i32(addr, s->pc & ~2);
3705 addr = load_reg(s, rn);
3707 tcg_gen_addi_i32(addr, addr, offset);
3708 if (insn & (1 << 20)) {
3709 gen_vfp_ld(s, dp, addr);
3710 gen_mov_vreg_F0(dp, rd);
3712 gen_mov_F0_vreg(dp, rd);
3713 gen_vfp_st(s, dp, addr);
3715 tcg_temp_free_i32(addr);
3717 /* load/store multiple */
3718 int w = insn & (1 << 21);
3720 n = (insn >> 1) & 0x7f;
3724 if (w && !(((insn >> 23) ^ (insn >> 24)) & 1)) {
3725 /* P == U , W == 1 => UNDEF */
3728 if (n == 0 || (rd + n) > 32 || (dp && n > 16)) {
3729 /* UNPREDICTABLE cases for bad immediates: we choose to
3730 * UNDEF to avoid generating huge numbers of TCG ops
3734 if (rn == 15 && w) {
3735 /* writeback to PC is UNPREDICTABLE, we choose to UNDEF */
3739 if (s->thumb && rn == 15) {
3740 /* This is actually UNPREDICTABLE */
3741 addr = tcg_temp_new_i32();
3742 tcg_gen_movi_i32(addr, s->pc & ~2);
3744 addr = load_reg(s, rn);
3746 if (insn & (1 << 24)) /* pre-decrement */
3747 tcg_gen_addi_i32(addr, addr, -((insn & 0xff) << 2));
3753 for (i = 0; i < n; i++) {
3754 if (insn & ARM_CP_RW_BIT) {
3756 gen_vfp_ld(s, dp, addr);
3757 gen_mov_vreg_F0(dp, rd + i);
3760 gen_mov_F0_vreg(dp, rd + i);
3761 gen_vfp_st(s, dp, addr);
3763 tcg_gen_addi_i32(addr, addr, offset);
3767 if (insn & (1 << 24))
3768 offset = -offset * n;
3769 else if (dp && (insn & 1))
3775 tcg_gen_addi_i32(addr, addr, offset);
3776 store_reg(s, rn, addr);
3778 tcg_temp_free_i32(addr);
3784 /* Should never happen. */
3790 static inline void gen_goto_tb(DisasContext *s, int n, target_ulong dest)
3792 TranslationBlock *tb;
3795 if ((tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK)) {
3797 gen_set_pc_im(s, dest);
3798 tcg_gen_exit_tb((uintptr_t)tb + n);
3800 gen_set_pc_im(s, dest);
3805 static inline void gen_jmp (DisasContext *s, uint32_t dest)
3807 if (unlikely(s->singlestep_enabled)) {
3808 /* An indirect jump so that we still trigger the debug exception. */
3813 gen_goto_tb(s, 0, dest);
3814 s->is_jmp = DISAS_TB_JUMP;
3818 static inline void gen_mulxy(TCGv_i32 t0, TCGv_i32 t1, int x, int y)
3821 tcg_gen_sari_i32(t0, t0, 16);
3825 tcg_gen_sari_i32(t1, t1, 16);
3828 tcg_gen_mul_i32(t0, t0, t1);
3831 /* Return the mask of PSR bits set by a MSR instruction. */
3832 static uint32_t msr_mask(CPUARMState *env, DisasContext *s, int flags, int spsr) {
3836 if (flags & (1 << 0))
3838 if (flags & (1 << 1))
3840 if (flags & (1 << 2))
3842 if (flags & (1 << 3))
3845 /* Mask out undefined bits. */
3846 mask &= ~CPSR_RESERVED;
3847 if (!arm_feature(env, ARM_FEATURE_V4T))
3849 if (!arm_feature(env, ARM_FEATURE_V5))
3850 mask &= ~CPSR_Q; /* V5TE in reality*/
3851 if (!arm_feature(env, ARM_FEATURE_V6))
3852 mask &= ~(CPSR_E | CPSR_GE);
3853 if (!arm_feature(env, ARM_FEATURE_THUMB2))
3855 /* Mask out execution state bits. */
3858 /* Mask out privileged bits. */
3864 /* Returns nonzero if access to the PSR is not permitted. Marks t0 as dead. */
3865 static int gen_set_psr(DisasContext *s, uint32_t mask, int spsr, TCGv_i32 t0)
3869 /* ??? This is also undefined in system mode. */
3873 tmp = load_cpu_field(spsr);
3874 tcg_gen_andi_i32(tmp, tmp, ~mask);
3875 tcg_gen_andi_i32(t0, t0, mask);
3876 tcg_gen_or_i32(tmp, tmp, t0);
3877 store_cpu_field(tmp, spsr);
3879 gen_set_cpsr(t0, mask);
3881 tcg_temp_free_i32(t0);
3886 /* Returns nonzero if access to the PSR is not permitted. */
3887 static int gen_set_psr_im(DisasContext *s, uint32_t mask, int spsr, uint32_t val)
3890 tmp = tcg_temp_new_i32();
3891 tcg_gen_movi_i32(tmp, val);
3892 return gen_set_psr(s, mask, spsr, tmp);
3895 /* Generate an old-style exception return. Marks pc as dead. */
3896 static void gen_exception_return(DisasContext *s, TCGv_i32 pc)
3899 store_reg(s, 15, pc);
3900 tmp = load_cpu_field(spsr);
3901 gen_set_cpsr(tmp, 0xffffffff);
3902 tcg_temp_free_i32(tmp);
3903 s->is_jmp = DISAS_UPDATE;
3906 /* Generate a v6 exception return. Marks both values as dead. */
3907 static void gen_rfe(DisasContext *s, TCGv_i32 pc, TCGv_i32 cpsr)
3909 gen_set_cpsr(cpsr, 0xffffffff);
3910 tcg_temp_free_i32(cpsr);
3911 store_reg(s, 15, pc);
3912 s->is_jmp = DISAS_UPDATE;
3916 gen_set_condexec (DisasContext *s)
3918 if (s->condexec_mask) {
3919 uint32_t val = (s->condexec_cond << 4) | (s->condexec_mask >> 1);
3920 TCGv_i32 tmp = tcg_temp_new_i32();
3921 tcg_gen_movi_i32(tmp, val);
3922 store_cpu_field(tmp, condexec_bits);
3926 static void gen_exception_insn(DisasContext *s, int offset, int excp)
3928 gen_set_condexec(s);
3929 gen_set_pc_im(s, s->pc - offset);
3930 gen_exception(excp);
3931 s->is_jmp = DISAS_JUMP;
3934 static void gen_nop_hint(DisasContext *s, int val)
3938 gen_set_pc_im(s, s->pc);
3939 s->is_jmp = DISAS_WFI;
3944 /* TODO: Implement SEV, SEVL and WFE. May help SMP performance. */
3950 #define CPU_V001 cpu_V0, cpu_V0, cpu_V1
3952 static inline void gen_neon_add(int size, TCGv_i32 t0, TCGv_i32 t1)
3955 case 0: gen_helper_neon_add_u8(t0, t0, t1); break;
3956 case 1: gen_helper_neon_add_u16(t0, t0, t1); break;
3957 case 2: tcg_gen_add_i32(t0, t0, t1); break;
3962 static inline void gen_neon_rsb(int size, TCGv_i32 t0, TCGv_i32 t1)
3965 case 0: gen_helper_neon_sub_u8(t0, t1, t0); break;
3966 case 1: gen_helper_neon_sub_u16(t0, t1, t0); break;
3967 case 2: tcg_gen_sub_i32(t0, t1, t0); break;
3972 /* 32-bit pairwise ops end up the same as the elementwise versions. */
3973 #define gen_helper_neon_pmax_s32 gen_helper_neon_max_s32
3974 #define gen_helper_neon_pmax_u32 gen_helper_neon_max_u32
3975 #define gen_helper_neon_pmin_s32 gen_helper_neon_min_s32
3976 #define gen_helper_neon_pmin_u32 gen_helper_neon_min_u32
3978 #define GEN_NEON_INTEGER_OP_ENV(name) do { \
3979 switch ((size << 1) | u) { \
3981 gen_helper_neon_##name##_s8(tmp, cpu_env, tmp, tmp2); \
3984 gen_helper_neon_##name##_u8(tmp, cpu_env, tmp, tmp2); \
3987 gen_helper_neon_##name##_s16(tmp, cpu_env, tmp, tmp2); \
3990 gen_helper_neon_##name##_u16(tmp, cpu_env, tmp, tmp2); \
3993 gen_helper_neon_##name##_s32(tmp, cpu_env, tmp, tmp2); \
3996 gen_helper_neon_##name##_u32(tmp, cpu_env, tmp, tmp2); \
3998 default: return 1; \
4001 #define GEN_NEON_INTEGER_OP(name) do { \
4002 switch ((size << 1) | u) { \
4004 gen_helper_neon_##name##_s8(tmp, tmp, tmp2); \
4007 gen_helper_neon_##name##_u8(tmp, tmp, tmp2); \
4010 gen_helper_neon_##name##_s16(tmp, tmp, tmp2); \
4013 gen_helper_neon_##name##_u16(tmp, tmp, tmp2); \
4016 gen_helper_neon_##name##_s32(tmp, tmp, tmp2); \
4019 gen_helper_neon_##name##_u32(tmp, tmp, tmp2); \
4021 default: return 1; \
4024 static TCGv_i32 neon_load_scratch(int scratch)
4026 TCGv_i32 tmp = tcg_temp_new_i32();
4027 tcg_gen_ld_i32(tmp, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch]));
4031 static void neon_store_scratch(int scratch, TCGv_i32 var)
4033 tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch]));
4034 tcg_temp_free_i32(var);
4037 static inline TCGv_i32 neon_get_scalar(int size, int reg)
4041 tmp = neon_load_reg(reg & 7, reg >> 4);
4043 gen_neon_dup_high16(tmp);
4045 gen_neon_dup_low16(tmp);
4048 tmp = neon_load_reg(reg & 15, reg >> 4);
4053 static int gen_neon_unzip(int rd, int rm, int size, int q)
4056 if (!q && size == 2) {
4059 tmp = tcg_const_i32(rd);
4060 tmp2 = tcg_const_i32(rm);
4064 gen_helper_neon_qunzip8(cpu_env, tmp, tmp2);
4067 gen_helper_neon_qunzip16(cpu_env, tmp, tmp2);
4070 gen_helper_neon_qunzip32(cpu_env, tmp, tmp2);
4078 gen_helper_neon_unzip8(cpu_env, tmp, tmp2);
4081 gen_helper_neon_unzip16(cpu_env, tmp, tmp2);
4087 tcg_temp_free_i32(tmp);
4088 tcg_temp_free_i32(tmp2);
4092 static int gen_neon_zip(int rd, int rm, int size, int q)
4095 if (!q && size == 2) {
4098 tmp = tcg_const_i32(rd);
4099 tmp2 = tcg_const_i32(rm);
4103 gen_helper_neon_qzip8(cpu_env, tmp, tmp2);
4106 gen_helper_neon_qzip16(cpu_env, tmp, tmp2);
4109 gen_helper_neon_qzip32(cpu_env, tmp, tmp2);
4117 gen_helper_neon_zip8(cpu_env, tmp, tmp2);
4120 gen_helper_neon_zip16(cpu_env, tmp, tmp2);
4126 tcg_temp_free_i32(tmp);
4127 tcg_temp_free_i32(tmp2);
4131 static void gen_neon_trn_u8(TCGv_i32 t0, TCGv_i32 t1)
4135 rd = tcg_temp_new_i32();
4136 tmp = tcg_temp_new_i32();
4138 tcg_gen_shli_i32(rd, t0, 8);
4139 tcg_gen_andi_i32(rd, rd, 0xff00ff00);
4140 tcg_gen_andi_i32(tmp, t1, 0x00ff00ff);
4141 tcg_gen_or_i32(rd, rd, tmp);
4143 tcg_gen_shri_i32(t1, t1, 8);
4144 tcg_gen_andi_i32(t1, t1, 0x00ff00ff);
4145 tcg_gen_andi_i32(tmp, t0, 0xff00ff00);
4146 tcg_gen_or_i32(t1, t1, tmp);
4147 tcg_gen_mov_i32(t0, rd);
4149 tcg_temp_free_i32(tmp);
4150 tcg_temp_free_i32(rd);
4153 static void gen_neon_trn_u16(TCGv_i32 t0, TCGv_i32 t1)
4157 rd = tcg_temp_new_i32();
4158 tmp = tcg_temp_new_i32();
4160 tcg_gen_shli_i32(rd, t0, 16);
4161 tcg_gen_andi_i32(tmp, t1, 0xffff);
4162 tcg_gen_or_i32(rd, rd, tmp);
4163 tcg_gen_shri_i32(t1, t1, 16);
4164 tcg_gen_andi_i32(tmp, t0, 0xffff0000);
4165 tcg_gen_or_i32(t1, t1, tmp);
4166 tcg_gen_mov_i32(t0, rd);
4168 tcg_temp_free_i32(tmp);
4169 tcg_temp_free_i32(rd);
4177 } neon_ls_element_type[11] = {
4191 /* Translate a NEON load/store element instruction. Return nonzero if the
4192 instruction is invalid. */
4193 static int disas_neon_ls_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
4212 if (!s->vfp_enabled)
4214 VFP_DREG_D(rd, insn);
4215 rn = (insn >> 16) & 0xf;
4217 load = (insn & (1 << 21)) != 0;
4218 if ((insn & (1 << 23)) == 0) {
4219 /* Load store all elements. */
4220 op = (insn >> 8) & 0xf;
4221 size = (insn >> 6) & 3;
4224 /* Catch UNDEF cases for bad values of align field */
4227 if (((insn >> 5) & 1) == 1) {
4232 if (((insn >> 4) & 3) == 3) {
4239 nregs = neon_ls_element_type[op].nregs;
4240 interleave = neon_ls_element_type[op].interleave;
4241 spacing = neon_ls_element_type[op].spacing;
4242 if (size == 3 && (interleave | spacing) != 1)
4244 addr = tcg_temp_new_i32();
4245 load_reg_var(s, addr, rn);
4246 stride = (1 << size) * interleave;
4247 for (reg = 0; reg < nregs; reg++) {
4248 if (interleave > 2 || (interleave == 2 && nregs == 2)) {
4249 load_reg_var(s, addr, rn);
4250 tcg_gen_addi_i32(addr, addr, (1 << size) * reg);
4251 } else if (interleave == 2 && nregs == 4 && reg == 2) {
4252 load_reg_var(s, addr, rn);
4253 tcg_gen_addi_i32(addr, addr, 1 << size);
4256 tmp64 = tcg_temp_new_i64();
4258 gen_aa32_ld64(tmp64, addr, IS_USER(s));
4259 neon_store_reg64(tmp64, rd);
4261 neon_load_reg64(tmp64, rd);
4262 gen_aa32_st64(tmp64, addr, IS_USER(s));
4264 tcg_temp_free_i64(tmp64);
4265 tcg_gen_addi_i32(addr, addr, stride);
4267 for (pass = 0; pass < 2; pass++) {
4270 tmp = tcg_temp_new_i32();
4271 gen_aa32_ld32u(tmp, addr, IS_USER(s));
4272 neon_store_reg(rd, pass, tmp);
4274 tmp = neon_load_reg(rd, pass);
4275 gen_aa32_st32(tmp, addr, IS_USER(s));
4276 tcg_temp_free_i32(tmp);
4278 tcg_gen_addi_i32(addr, addr, stride);
4279 } else if (size == 1) {
4281 tmp = tcg_temp_new_i32();
4282 gen_aa32_ld16u(tmp, addr, IS_USER(s));
4283 tcg_gen_addi_i32(addr, addr, stride);
4284 tmp2 = tcg_temp_new_i32();
4285 gen_aa32_ld16u(tmp2, addr, IS_USER(s));
4286 tcg_gen_addi_i32(addr, addr, stride);
4287 tcg_gen_shli_i32(tmp2, tmp2, 16);
4288 tcg_gen_or_i32(tmp, tmp, tmp2);
4289 tcg_temp_free_i32(tmp2);
4290 neon_store_reg(rd, pass, tmp);
4292 tmp = neon_load_reg(rd, pass);
4293 tmp2 = tcg_temp_new_i32();
4294 tcg_gen_shri_i32(tmp2, tmp, 16);
4295 gen_aa32_st16(tmp, addr, IS_USER(s));
4296 tcg_temp_free_i32(tmp);
4297 tcg_gen_addi_i32(addr, addr, stride);
4298 gen_aa32_st16(tmp2, addr, IS_USER(s));
4299 tcg_temp_free_i32(tmp2);
4300 tcg_gen_addi_i32(addr, addr, stride);
4302 } else /* size == 0 */ {
4304 TCGV_UNUSED_I32(tmp2);
4305 for (n = 0; n < 4; n++) {
4306 tmp = tcg_temp_new_i32();
4307 gen_aa32_ld8u(tmp, addr, IS_USER(s));
4308 tcg_gen_addi_i32(addr, addr, stride);
4312 tcg_gen_shli_i32(tmp, tmp, n * 8);
4313 tcg_gen_or_i32(tmp2, tmp2, tmp);
4314 tcg_temp_free_i32(tmp);
4317 neon_store_reg(rd, pass, tmp2);
4319 tmp2 = neon_load_reg(rd, pass);
4320 for (n = 0; n < 4; n++) {
4321 tmp = tcg_temp_new_i32();
4323 tcg_gen_mov_i32(tmp, tmp2);
4325 tcg_gen_shri_i32(tmp, tmp2, n * 8);
4327 gen_aa32_st8(tmp, addr, IS_USER(s));
4328 tcg_temp_free_i32(tmp);
4329 tcg_gen_addi_i32(addr, addr, stride);
4331 tcg_temp_free_i32(tmp2);
4338 tcg_temp_free_i32(addr);
4341 size = (insn >> 10) & 3;
4343 /* Load single element to all lanes. */
4344 int a = (insn >> 4) & 1;
4348 size = (insn >> 6) & 3;
4349 nregs = ((insn >> 8) & 3) + 1;
4352 if (nregs != 4 || a == 0) {
4355 /* For VLD4 size==3 a == 1 means 32 bits at 16 byte alignment */
4358 if (nregs == 1 && a == 1 && size == 0) {
4361 if (nregs == 3 && a == 1) {
4364 addr = tcg_temp_new_i32();
4365 load_reg_var(s, addr, rn);
4367 /* VLD1 to all lanes: bit 5 indicates how many Dregs to write */
4368 tmp = gen_load_and_replicate(s, addr, size);
4369 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 0));
4370 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 1));
4371 if (insn & (1 << 5)) {
4372 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd + 1, 0));
4373 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd + 1, 1));
4375 tcg_temp_free_i32(tmp);
4377 /* VLD2/3/4 to all lanes: bit 5 indicates register stride */
4378 stride = (insn & (1 << 5)) ? 2 : 1;
4379 for (reg = 0; reg < nregs; reg++) {
4380 tmp = gen_load_and_replicate(s, addr, size);
4381 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 0));
4382 tcg_gen_st_i32(tmp, cpu_env, neon_reg_offset(rd, 1));
4383 tcg_temp_free_i32(tmp);
4384 tcg_gen_addi_i32(addr, addr, 1 << size);
4388 tcg_temp_free_i32(addr);
4389 stride = (1 << size) * nregs;
4391 /* Single element. */
4392 int idx = (insn >> 4) & 0xf;
4393 pass = (insn >> 7) & 1;
4396 shift = ((insn >> 5) & 3) * 8;
4400 shift = ((insn >> 6) & 1) * 16;
4401 stride = (insn & (1 << 5)) ? 2 : 1;
4405 stride = (insn & (1 << 6)) ? 2 : 1;
4410 nregs = ((insn >> 8) & 3) + 1;
4411 /* Catch the UNDEF cases. This is unavoidably a bit messy. */
4414 if (((idx & (1 << size)) != 0) ||
4415 (size == 2 && ((idx & 3) == 1 || (idx & 3) == 2))) {
4420 if ((idx & 1) != 0) {
4425 if (size == 2 && (idx & 2) != 0) {
4430 if ((size == 2) && ((idx & 3) == 3)) {
4437 if ((rd + stride * (nregs - 1)) > 31) {
4438 /* Attempts to write off the end of the register file
4439 * are UNPREDICTABLE; we choose to UNDEF because otherwise
4440 * the neon_load_reg() would write off the end of the array.
4444 addr = tcg_temp_new_i32();
4445 load_reg_var(s, addr, rn);
4446 for (reg = 0; reg < nregs; reg++) {
4448 tmp = tcg_temp_new_i32();
4451 gen_aa32_ld8u(tmp, addr, IS_USER(s));
4454 gen_aa32_ld16u(tmp, addr, IS_USER(s));
4457 gen_aa32_ld32u(tmp, addr, IS_USER(s));
4459 default: /* Avoid compiler warnings. */
4463 tmp2 = neon_load_reg(rd, pass);
4464 tcg_gen_deposit_i32(tmp, tmp2, tmp,
4465 shift, size ? 16 : 8);
4466 tcg_temp_free_i32(tmp2);
4468 neon_store_reg(rd, pass, tmp);
4469 } else { /* Store */
4470 tmp = neon_load_reg(rd, pass);
4472 tcg_gen_shri_i32(tmp, tmp, shift);
4475 gen_aa32_st8(tmp, addr, IS_USER(s));
4478 gen_aa32_st16(tmp, addr, IS_USER(s));
4481 gen_aa32_st32(tmp, addr, IS_USER(s));
4484 tcg_temp_free_i32(tmp);
4487 tcg_gen_addi_i32(addr, addr, 1 << size);
4489 tcg_temp_free_i32(addr);
4490 stride = nregs * (1 << size);
4496 base = load_reg(s, rn);
4498 tcg_gen_addi_i32(base, base, stride);
4501 index = load_reg(s, rm);
4502 tcg_gen_add_i32(base, base, index);
4503 tcg_temp_free_i32(index);
4505 store_reg(s, rn, base);
4510 /* Bitwise select. dest = c ? t : f. Clobbers T and F. */
4511 static void gen_neon_bsl(TCGv_i32 dest, TCGv_i32 t, TCGv_i32 f, TCGv_i32 c)
4513 tcg_gen_and_i32(t, t, c);
4514 tcg_gen_andc_i32(f, f, c);
4515 tcg_gen_or_i32(dest, t, f);
4518 static inline void gen_neon_narrow(int size, TCGv_i32 dest, TCGv_i64 src)
4521 case 0: gen_helper_neon_narrow_u8(dest, src); break;
4522 case 1: gen_helper_neon_narrow_u16(dest, src); break;
4523 case 2: tcg_gen_trunc_i64_i32(dest, src); break;
4528 static inline void gen_neon_narrow_sats(int size, TCGv_i32 dest, TCGv_i64 src)
4531 case 0: gen_helper_neon_narrow_sat_s8(dest, cpu_env, src); break;
4532 case 1: gen_helper_neon_narrow_sat_s16(dest, cpu_env, src); break;
4533 case 2: gen_helper_neon_narrow_sat_s32(dest, cpu_env, src); break;
4538 static inline void gen_neon_narrow_satu(int size, TCGv_i32 dest, TCGv_i64 src)
4541 case 0: gen_helper_neon_narrow_sat_u8(dest, cpu_env, src); break;
4542 case 1: gen_helper_neon_narrow_sat_u16(dest, cpu_env, src); break;
4543 case 2: gen_helper_neon_narrow_sat_u32(dest, cpu_env, src); break;
4548 static inline void gen_neon_unarrow_sats(int size, TCGv_i32 dest, TCGv_i64 src)
4551 case 0: gen_helper_neon_unarrow_sat8(dest, cpu_env, src); break;
4552 case 1: gen_helper_neon_unarrow_sat16(dest, cpu_env, src); break;
4553 case 2: gen_helper_neon_unarrow_sat32(dest, cpu_env, src); break;
4558 static inline void gen_neon_shift_narrow(int size, TCGv_i32 var, TCGv_i32 shift,
4564 case 1: gen_helper_neon_rshl_u16(var, var, shift); break;
4565 case 2: gen_helper_neon_rshl_u32(var, var, shift); break;
4570 case 1: gen_helper_neon_rshl_s16(var, var, shift); break;
4571 case 2: gen_helper_neon_rshl_s32(var, var, shift); break;
4578 case 1: gen_helper_neon_shl_u16(var, var, shift); break;
4579 case 2: gen_helper_neon_shl_u32(var, var, shift); break;
4584 case 1: gen_helper_neon_shl_s16(var, var, shift); break;
4585 case 2: gen_helper_neon_shl_s32(var, var, shift); break;
4592 static inline void gen_neon_widen(TCGv_i64 dest, TCGv_i32 src, int size, int u)
4596 case 0: gen_helper_neon_widen_u8(dest, src); break;
4597 case 1: gen_helper_neon_widen_u16(dest, src); break;
4598 case 2: tcg_gen_extu_i32_i64(dest, src); break;
4603 case 0: gen_helper_neon_widen_s8(dest, src); break;
4604 case 1: gen_helper_neon_widen_s16(dest, src); break;
4605 case 2: tcg_gen_ext_i32_i64(dest, src); break;
4609 tcg_temp_free_i32(src);
4612 static inline void gen_neon_addl(int size)
4615 case 0: gen_helper_neon_addl_u16(CPU_V001); break;
4616 case 1: gen_helper_neon_addl_u32(CPU_V001); break;
4617 case 2: tcg_gen_add_i64(CPU_V001); break;
4622 static inline void gen_neon_subl(int size)
4625 case 0: gen_helper_neon_subl_u16(CPU_V001); break;
4626 case 1: gen_helper_neon_subl_u32(CPU_V001); break;
4627 case 2: tcg_gen_sub_i64(CPU_V001); break;
4632 static inline void gen_neon_negl(TCGv_i64 var, int size)
4635 case 0: gen_helper_neon_negl_u16(var, var); break;
4636 case 1: gen_helper_neon_negl_u32(var, var); break;
4638 tcg_gen_neg_i64(var, var);
4644 static inline void gen_neon_addl_saturate(TCGv_i64 op0, TCGv_i64 op1, int size)
4647 case 1: gen_helper_neon_addl_saturate_s32(op0, cpu_env, op0, op1); break;
4648 case 2: gen_helper_neon_addl_saturate_s64(op0, cpu_env, op0, op1); break;
4653 static inline void gen_neon_mull(TCGv_i64 dest, TCGv_i32 a, TCGv_i32 b,
4658 switch ((size << 1) | u) {
4659 case 0: gen_helper_neon_mull_s8(dest, a, b); break;
4660 case 1: gen_helper_neon_mull_u8(dest, a, b); break;
4661 case 2: gen_helper_neon_mull_s16(dest, a, b); break;
4662 case 3: gen_helper_neon_mull_u16(dest, a, b); break;
4664 tmp = gen_muls_i64_i32(a, b);
4665 tcg_gen_mov_i64(dest, tmp);
4666 tcg_temp_free_i64(tmp);
4669 tmp = gen_mulu_i64_i32(a, b);
4670 tcg_gen_mov_i64(dest, tmp);
4671 tcg_temp_free_i64(tmp);
4676 /* gen_helper_neon_mull_[su]{8|16} do not free their parameters.
4677 Don't forget to clean them now. */
4679 tcg_temp_free_i32(a);
4680 tcg_temp_free_i32(b);
4684 static void gen_neon_narrow_op(int op, int u, int size,
4685 TCGv_i32 dest, TCGv_i64 src)
4689 gen_neon_unarrow_sats(size, dest, src);
4691 gen_neon_narrow(size, dest, src);
4695 gen_neon_narrow_satu(size, dest, src);
4697 gen_neon_narrow_sats(size, dest, src);
4702 /* Symbolic constants for op fields for Neon 3-register same-length.
4703 * The values correspond to bits [11:8,4]; see the ARM ARM DDI0406B
4706 #define NEON_3R_VHADD 0
4707 #define NEON_3R_VQADD 1
4708 #define NEON_3R_VRHADD 2
4709 #define NEON_3R_LOGIC 3 /* VAND,VBIC,VORR,VMOV,VORN,VEOR,VBIF,VBIT,VBSL */
4710 #define NEON_3R_VHSUB 4
4711 #define NEON_3R_VQSUB 5
4712 #define NEON_3R_VCGT 6
4713 #define NEON_3R_VCGE 7
4714 #define NEON_3R_VSHL 8
4715 #define NEON_3R_VQSHL 9
4716 #define NEON_3R_VRSHL 10
4717 #define NEON_3R_VQRSHL 11
4718 #define NEON_3R_VMAX 12
4719 #define NEON_3R_VMIN 13
4720 #define NEON_3R_VABD 14
4721 #define NEON_3R_VABA 15
4722 #define NEON_3R_VADD_VSUB 16
4723 #define NEON_3R_VTST_VCEQ 17
4724 #define NEON_3R_VML 18 /* VMLA, VMLAL, VMLS, VMLSL */
4725 #define NEON_3R_VMUL 19
4726 #define NEON_3R_VPMAX 20
4727 #define NEON_3R_VPMIN 21
4728 #define NEON_3R_VQDMULH_VQRDMULH 22
4729 #define NEON_3R_VPADD 23
4730 #define NEON_3R_VFM 25 /* VFMA, VFMS : float fused multiply-add */
4731 #define NEON_3R_FLOAT_ARITH 26 /* float VADD, VSUB, VPADD, VABD */
4732 #define NEON_3R_FLOAT_MULTIPLY 27 /* float VMLA, VMLS, VMUL */
4733 #define NEON_3R_FLOAT_CMP 28 /* float VCEQ, VCGE, VCGT */
4734 #define NEON_3R_FLOAT_ACMP 29 /* float VACGE, VACGT, VACLE, VACLT */
4735 #define NEON_3R_FLOAT_MINMAX 30 /* float VMIN, VMAX */
4736 #define NEON_3R_FLOAT_MISC 31 /* float VRECPS, VRSQRTS, VMAXNM/MINNM */
4738 static const uint8_t neon_3r_sizes[] = {
4739 [NEON_3R_VHADD] = 0x7,
4740 [NEON_3R_VQADD] = 0xf,
4741 [NEON_3R_VRHADD] = 0x7,
4742 [NEON_3R_LOGIC] = 0xf, /* size field encodes op type */
4743 [NEON_3R_VHSUB] = 0x7,
4744 [NEON_3R_VQSUB] = 0xf,
4745 [NEON_3R_VCGT] = 0x7,
4746 [NEON_3R_VCGE] = 0x7,
4747 [NEON_3R_VSHL] = 0xf,
4748 [NEON_3R_VQSHL] = 0xf,
4749 [NEON_3R_VRSHL] = 0xf,
4750 [NEON_3R_VQRSHL] = 0xf,
4751 [NEON_3R_VMAX] = 0x7,
4752 [NEON_3R_VMIN] = 0x7,
4753 [NEON_3R_VABD] = 0x7,
4754 [NEON_3R_VABA] = 0x7,
4755 [NEON_3R_VADD_VSUB] = 0xf,
4756 [NEON_3R_VTST_VCEQ] = 0x7,
4757 [NEON_3R_VML] = 0x7,
4758 [NEON_3R_VMUL] = 0x7,
4759 [NEON_3R_VPMAX] = 0x7,
4760 [NEON_3R_VPMIN] = 0x7,
4761 [NEON_3R_VQDMULH_VQRDMULH] = 0x6,
4762 [NEON_3R_VPADD] = 0x7,
4763 [NEON_3R_VFM] = 0x5, /* size bit 1 encodes op */
4764 [NEON_3R_FLOAT_ARITH] = 0x5, /* size bit 1 encodes op */
4765 [NEON_3R_FLOAT_MULTIPLY] = 0x5, /* size bit 1 encodes op */
4766 [NEON_3R_FLOAT_CMP] = 0x5, /* size bit 1 encodes op */
4767 [NEON_3R_FLOAT_ACMP] = 0x5, /* size bit 1 encodes op */
4768 [NEON_3R_FLOAT_MINMAX] = 0x5, /* size bit 1 encodes op */
4769 [NEON_3R_FLOAT_MISC] = 0x5, /* size bit 1 encodes op */
4772 /* Symbolic constants for op fields for Neon 2-register miscellaneous.
4773 * The values correspond to bits [17:16,10:7]; see the ARM ARM DDI0406B
4776 #define NEON_2RM_VREV64 0
4777 #define NEON_2RM_VREV32 1
4778 #define NEON_2RM_VREV16 2
4779 #define NEON_2RM_VPADDL 4
4780 #define NEON_2RM_VPADDL_U 5
4781 #define NEON_2RM_AESE 6 /* Includes AESD */
4782 #define NEON_2RM_AESMC 7 /* Includes AESIMC */
4783 #define NEON_2RM_VCLS 8
4784 #define NEON_2RM_VCLZ 9
4785 #define NEON_2RM_VCNT 10
4786 #define NEON_2RM_VMVN 11
4787 #define NEON_2RM_VPADAL 12
4788 #define NEON_2RM_VPADAL_U 13
4789 #define NEON_2RM_VQABS 14
4790 #define NEON_2RM_VQNEG 15
4791 #define NEON_2RM_VCGT0 16
4792 #define NEON_2RM_VCGE0 17
4793 #define NEON_2RM_VCEQ0 18
4794 #define NEON_2RM_VCLE0 19
4795 #define NEON_2RM_VCLT0 20
4796 #define NEON_2RM_VABS 22
4797 #define NEON_2RM_VNEG 23
4798 #define NEON_2RM_VCGT0_F 24
4799 #define NEON_2RM_VCGE0_F 25
4800 #define NEON_2RM_VCEQ0_F 26
4801 #define NEON_2RM_VCLE0_F 27
4802 #define NEON_2RM_VCLT0_F 28
4803 #define NEON_2RM_VABS_F 30
4804 #define NEON_2RM_VNEG_F 31
4805 #define NEON_2RM_VSWP 32
4806 #define NEON_2RM_VTRN 33
4807 #define NEON_2RM_VUZP 34
4808 #define NEON_2RM_VZIP 35
4809 #define NEON_2RM_VMOVN 36 /* Includes VQMOVN, VQMOVUN */
4810 #define NEON_2RM_VQMOVN 37 /* Includes VQMOVUN */
4811 #define NEON_2RM_VSHLL 38
4812 #define NEON_2RM_VRINTN 40
4813 #define NEON_2RM_VRINTX 41
4814 #define NEON_2RM_VRINTA 42
4815 #define NEON_2RM_VRINTZ 43
4816 #define NEON_2RM_VCVT_F16_F32 44
4817 #define NEON_2RM_VRINTM 45
4818 #define NEON_2RM_VCVT_F32_F16 46
4819 #define NEON_2RM_VRINTP 47
4820 #define NEON_2RM_VCVTAU 48
4821 #define NEON_2RM_VCVTAS 49
4822 #define NEON_2RM_VCVTNU 50
4823 #define NEON_2RM_VCVTNS 51
4824 #define NEON_2RM_VCVTPU 52
4825 #define NEON_2RM_VCVTPS 53
4826 #define NEON_2RM_VCVTMU 54
4827 #define NEON_2RM_VCVTMS 55
4828 #define NEON_2RM_VRECPE 56
4829 #define NEON_2RM_VRSQRTE 57
4830 #define NEON_2RM_VRECPE_F 58
4831 #define NEON_2RM_VRSQRTE_F 59
4832 #define NEON_2RM_VCVT_FS 60
4833 #define NEON_2RM_VCVT_FU 61
4834 #define NEON_2RM_VCVT_SF 62
4835 #define NEON_2RM_VCVT_UF 63
4837 static int neon_2rm_is_float_op(int op)
4839 /* Return true if this neon 2reg-misc op is float-to-float */
4840 return (op == NEON_2RM_VABS_F || op == NEON_2RM_VNEG_F ||
4841 (op >= NEON_2RM_VRINTN && op <= NEON_2RM_VRINTZ) ||
4842 op == NEON_2RM_VRINTM ||
4843 (op >= NEON_2RM_VRINTP && op <= NEON_2RM_VCVTMS) ||
4844 op >= NEON_2RM_VRECPE_F);
4847 /* Each entry in this array has bit n set if the insn allows
4848 * size value n (otherwise it will UNDEF). Since unallocated
4849 * op values will have no bits set they always UNDEF.
4851 static const uint8_t neon_2rm_sizes[] = {
4852 [NEON_2RM_VREV64] = 0x7,
4853 [NEON_2RM_VREV32] = 0x3,
4854 [NEON_2RM_VREV16] = 0x1,
4855 [NEON_2RM_VPADDL] = 0x7,
4856 [NEON_2RM_VPADDL_U] = 0x7,
4857 [NEON_2RM_AESE] = 0x1,
4858 [NEON_2RM_AESMC] = 0x1,
4859 [NEON_2RM_VCLS] = 0x7,
4860 [NEON_2RM_VCLZ] = 0x7,
4861 [NEON_2RM_VCNT] = 0x1,
4862 [NEON_2RM_VMVN] = 0x1,
4863 [NEON_2RM_VPADAL] = 0x7,
4864 [NEON_2RM_VPADAL_U] = 0x7,
4865 [NEON_2RM_VQABS] = 0x7,
4866 [NEON_2RM_VQNEG] = 0x7,
4867 [NEON_2RM_VCGT0] = 0x7,
4868 [NEON_2RM_VCGE0] = 0x7,
4869 [NEON_2RM_VCEQ0] = 0x7,
4870 [NEON_2RM_VCLE0] = 0x7,
4871 [NEON_2RM_VCLT0] = 0x7,
4872 [NEON_2RM_VABS] = 0x7,
4873 [NEON_2RM_VNEG] = 0x7,
4874 [NEON_2RM_VCGT0_F] = 0x4,
4875 [NEON_2RM_VCGE0_F] = 0x4,
4876 [NEON_2RM_VCEQ0_F] = 0x4,
4877 [NEON_2RM_VCLE0_F] = 0x4,
4878 [NEON_2RM_VCLT0_F] = 0x4,
4879 [NEON_2RM_VABS_F] = 0x4,
4880 [NEON_2RM_VNEG_F] = 0x4,
4881 [NEON_2RM_VSWP] = 0x1,
4882 [NEON_2RM_VTRN] = 0x7,
4883 [NEON_2RM_VUZP] = 0x7,
4884 [NEON_2RM_VZIP] = 0x7,
4885 [NEON_2RM_VMOVN] = 0x7,
4886 [NEON_2RM_VQMOVN] = 0x7,
4887 [NEON_2RM_VSHLL] = 0x7,
4888 [NEON_2RM_VRINTN] = 0x4,
4889 [NEON_2RM_VRINTX] = 0x4,
4890 [NEON_2RM_VRINTA] = 0x4,
4891 [NEON_2RM_VRINTZ] = 0x4,
4892 [NEON_2RM_VCVT_F16_F32] = 0x2,
4893 [NEON_2RM_VRINTM] = 0x4,
4894 [NEON_2RM_VCVT_F32_F16] = 0x2,
4895 [NEON_2RM_VRINTP] = 0x4,
4896 [NEON_2RM_VCVTAU] = 0x4,
4897 [NEON_2RM_VCVTAS] = 0x4,
4898 [NEON_2RM_VCVTNU] = 0x4,
4899 [NEON_2RM_VCVTNS] = 0x4,
4900 [NEON_2RM_VCVTPU] = 0x4,
4901 [NEON_2RM_VCVTPS] = 0x4,
4902 [NEON_2RM_VCVTMU] = 0x4,
4903 [NEON_2RM_VCVTMS] = 0x4,
4904 [NEON_2RM_VRECPE] = 0x4,
4905 [NEON_2RM_VRSQRTE] = 0x4,
4906 [NEON_2RM_VRECPE_F] = 0x4,
4907 [NEON_2RM_VRSQRTE_F] = 0x4,
4908 [NEON_2RM_VCVT_FS] = 0x4,
4909 [NEON_2RM_VCVT_FU] = 0x4,
4910 [NEON_2RM_VCVT_SF] = 0x4,
4911 [NEON_2RM_VCVT_UF] = 0x4,
4914 /* Translate a NEON data processing instruction. Return nonzero if the
4915 instruction is invalid.
4916 We process data in a mixture of 32-bit and 64-bit chunks.
4917 Mostly we use 32-bit chunks so we can use normal scalar instructions. */
4919 static int disas_neon_data_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
4931 TCGv_i32 tmp, tmp2, tmp3, tmp4, tmp5;
4934 if (!s->vfp_enabled)
4936 q = (insn & (1 << 6)) != 0;
4937 u = (insn >> 24) & 1;
4938 VFP_DREG_D(rd, insn);
4939 VFP_DREG_N(rn, insn);
4940 VFP_DREG_M(rm, insn);
4941 size = (insn >> 20) & 3;
4942 if ((insn & (1 << 23)) == 0) {
4943 /* Three register same length. */
4944 op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1);
4945 /* Catch invalid op and bad size combinations: UNDEF */
4946 if ((neon_3r_sizes[op] & (1 << size)) == 0) {
4949 /* All insns of this form UNDEF for either this condition or the
4950 * superset of cases "Q==1"; we catch the latter later.
4952 if (q && ((rd | rn | rm) & 1)) {
4955 if (size == 3 && op != NEON_3R_LOGIC) {
4956 /* 64-bit element instructions. */
4957 for (pass = 0; pass < (q ? 2 : 1); pass++) {
4958 neon_load_reg64(cpu_V0, rn + pass);
4959 neon_load_reg64(cpu_V1, rm + pass);
4963 gen_helper_neon_qadd_u64(cpu_V0, cpu_env,
4966 gen_helper_neon_qadd_s64(cpu_V0, cpu_env,
4972 gen_helper_neon_qsub_u64(cpu_V0, cpu_env,
4975 gen_helper_neon_qsub_s64(cpu_V0, cpu_env,
4981 gen_helper_neon_shl_u64(cpu_V0, cpu_V1, cpu_V0);
4983 gen_helper_neon_shl_s64(cpu_V0, cpu_V1, cpu_V0);
4988 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
4991 gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
4997 gen_helper_neon_rshl_u64(cpu_V0, cpu_V1, cpu_V0);
4999 gen_helper_neon_rshl_s64(cpu_V0, cpu_V1, cpu_V0);
5002 case NEON_3R_VQRSHL:
5004 gen_helper_neon_qrshl_u64(cpu_V0, cpu_env,
5007 gen_helper_neon_qrshl_s64(cpu_V0, cpu_env,
5011 case NEON_3R_VADD_VSUB:
5013 tcg_gen_sub_i64(CPU_V001);
5015 tcg_gen_add_i64(CPU_V001);
5021 neon_store_reg64(cpu_V0, rd + pass);
5030 case NEON_3R_VQRSHL:
5033 /* Shift instruction operands are reversed. */
5048 case NEON_3R_FLOAT_ARITH:
5049 pairwise = (u && size < 2); /* if VPADD (float) */
5051 case NEON_3R_FLOAT_MINMAX:
5052 pairwise = u; /* if VPMIN/VPMAX (float) */
5054 case NEON_3R_FLOAT_CMP:
5056 /* no encoding for U=0 C=1x */
5060 case NEON_3R_FLOAT_ACMP:
5065 case NEON_3R_FLOAT_MISC:
5066 /* VMAXNM/VMINNM in ARMv8 */
5067 if (u && !arm_feature(env, ARM_FEATURE_V8)) {
5072 if (u && (size != 0)) {
5073 /* UNDEF on invalid size for polynomial subcase */
5078 if (!arm_feature(env, ARM_FEATURE_VFP4) || u) {
5086 if (pairwise && q) {
5087 /* All the pairwise insns UNDEF if Q is set */
5091 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5096 tmp = neon_load_reg(rn, 0);
5097 tmp2 = neon_load_reg(rn, 1);
5099 tmp = neon_load_reg(rm, 0);
5100 tmp2 = neon_load_reg(rm, 1);
5104 tmp = neon_load_reg(rn, pass);
5105 tmp2 = neon_load_reg(rm, pass);
5109 GEN_NEON_INTEGER_OP(hadd);
5112 GEN_NEON_INTEGER_OP_ENV(qadd);
5114 case NEON_3R_VRHADD:
5115 GEN_NEON_INTEGER_OP(rhadd);
5117 case NEON_3R_LOGIC: /* Logic ops. */
5118 switch ((u << 2) | size) {
5120 tcg_gen_and_i32(tmp, tmp, tmp2);
5123 tcg_gen_andc_i32(tmp, tmp, tmp2);
5126 tcg_gen_or_i32(tmp, tmp, tmp2);
5129 tcg_gen_orc_i32(tmp, tmp, tmp2);
5132 tcg_gen_xor_i32(tmp, tmp, tmp2);
5135 tmp3 = neon_load_reg(rd, pass);
5136 gen_neon_bsl(tmp, tmp, tmp2, tmp3);
5137 tcg_temp_free_i32(tmp3);
5140 tmp3 = neon_load_reg(rd, pass);
5141 gen_neon_bsl(tmp, tmp, tmp3, tmp2);
5142 tcg_temp_free_i32(tmp3);
5145 tmp3 = neon_load_reg(rd, pass);
5146 gen_neon_bsl(tmp, tmp3, tmp, tmp2);
5147 tcg_temp_free_i32(tmp3);
5152 GEN_NEON_INTEGER_OP(hsub);
5155 GEN_NEON_INTEGER_OP_ENV(qsub);
5158 GEN_NEON_INTEGER_OP(cgt);
5161 GEN_NEON_INTEGER_OP(cge);
5164 GEN_NEON_INTEGER_OP(shl);
5167 GEN_NEON_INTEGER_OP_ENV(qshl);
5170 GEN_NEON_INTEGER_OP(rshl);
5172 case NEON_3R_VQRSHL:
5173 GEN_NEON_INTEGER_OP_ENV(qrshl);
5176 GEN_NEON_INTEGER_OP(max);
5179 GEN_NEON_INTEGER_OP(min);
5182 GEN_NEON_INTEGER_OP(abd);
5185 GEN_NEON_INTEGER_OP(abd);
5186 tcg_temp_free_i32(tmp2);
5187 tmp2 = neon_load_reg(rd, pass);
5188 gen_neon_add(size, tmp, tmp2);
5190 case NEON_3R_VADD_VSUB:
5191 if (!u) { /* VADD */
5192 gen_neon_add(size, tmp, tmp2);
5195 case 0: gen_helper_neon_sub_u8(tmp, tmp, tmp2); break;
5196 case 1: gen_helper_neon_sub_u16(tmp, tmp, tmp2); break;
5197 case 2: tcg_gen_sub_i32(tmp, tmp, tmp2); break;
5202 case NEON_3R_VTST_VCEQ:
5203 if (!u) { /* VTST */
5205 case 0: gen_helper_neon_tst_u8(tmp, tmp, tmp2); break;
5206 case 1: gen_helper_neon_tst_u16(tmp, tmp, tmp2); break;
5207 case 2: gen_helper_neon_tst_u32(tmp, tmp, tmp2); break;
5212 case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break;
5213 case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break;
5214 case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break;
5219 case NEON_3R_VML: /* VMLA, VMLAL, VMLS,VMLSL */
5221 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
5222 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
5223 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
5226 tcg_temp_free_i32(tmp2);
5227 tmp2 = neon_load_reg(rd, pass);
5229 gen_neon_rsb(size, tmp, tmp2);
5231 gen_neon_add(size, tmp, tmp2);
5235 if (u) { /* polynomial */
5236 gen_helper_neon_mul_p8(tmp, tmp, tmp2);
5237 } else { /* Integer */
5239 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
5240 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
5241 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
5247 GEN_NEON_INTEGER_OP(pmax);
5250 GEN_NEON_INTEGER_OP(pmin);
5252 case NEON_3R_VQDMULH_VQRDMULH: /* Multiply high. */
5253 if (!u) { /* VQDMULH */
5256 gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
5259 gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
5263 } else { /* VQRDMULH */
5266 gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
5269 gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
5277 case 0: gen_helper_neon_padd_u8(tmp, tmp, tmp2); break;
5278 case 1: gen_helper_neon_padd_u16(tmp, tmp, tmp2); break;
5279 case 2: tcg_gen_add_i32(tmp, tmp, tmp2); break;
5283 case NEON_3R_FLOAT_ARITH: /* Floating point arithmetic. */
5285 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5286 switch ((u << 2) | size) {
5289 gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
5292 gen_helper_vfp_subs(tmp, tmp, tmp2, fpstatus);
5295 gen_helper_neon_abd_f32(tmp, tmp, tmp2, fpstatus);
5300 tcg_temp_free_ptr(fpstatus);
5303 case NEON_3R_FLOAT_MULTIPLY:
5305 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5306 gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus);
5308 tcg_temp_free_i32(tmp2);
5309 tmp2 = neon_load_reg(rd, pass);
5311 gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
5313 gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus);
5316 tcg_temp_free_ptr(fpstatus);
5319 case NEON_3R_FLOAT_CMP:
5321 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5323 gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus);
5326 gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus);
5328 gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus);
5331 tcg_temp_free_ptr(fpstatus);
5334 case NEON_3R_FLOAT_ACMP:
5336 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5338 gen_helper_neon_acge_f32(tmp, tmp, tmp2, fpstatus);
5340 gen_helper_neon_acgt_f32(tmp, tmp, tmp2, fpstatus);
5342 tcg_temp_free_ptr(fpstatus);
5345 case NEON_3R_FLOAT_MINMAX:
5347 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5349 gen_helper_vfp_maxs(tmp, tmp, tmp2, fpstatus);
5351 gen_helper_vfp_mins(tmp, tmp, tmp2, fpstatus);
5353 tcg_temp_free_ptr(fpstatus);
5356 case NEON_3R_FLOAT_MISC:
5359 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5361 gen_helper_vfp_maxnums(tmp, tmp, tmp2, fpstatus);
5363 gen_helper_vfp_minnums(tmp, tmp, tmp2, fpstatus);
5365 tcg_temp_free_ptr(fpstatus);
5368 gen_helper_recps_f32(tmp, tmp, tmp2, cpu_env);
5370 gen_helper_rsqrts_f32(tmp, tmp, tmp2, cpu_env);
5376 /* VFMA, VFMS: fused multiply-add */
5377 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
5378 TCGv_i32 tmp3 = neon_load_reg(rd, pass);
5381 gen_helper_vfp_negs(tmp, tmp);
5383 gen_helper_vfp_muladds(tmp, tmp, tmp2, tmp3, fpstatus);
5384 tcg_temp_free_i32(tmp3);
5385 tcg_temp_free_ptr(fpstatus);
5391 tcg_temp_free_i32(tmp2);
5393 /* Save the result. For elementwise operations we can put it
5394 straight into the destination register. For pairwise operations
5395 we have to be careful to avoid clobbering the source operands. */
5396 if (pairwise && rd == rm) {
5397 neon_store_scratch(pass, tmp);
5399 neon_store_reg(rd, pass, tmp);
5403 if (pairwise && rd == rm) {
5404 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5405 tmp = neon_load_scratch(pass);
5406 neon_store_reg(rd, pass, tmp);
5409 /* End of 3 register same size operations. */
5410 } else if (insn & (1 << 4)) {
5411 if ((insn & 0x00380080) != 0) {
5412 /* Two registers and shift. */
5413 op = (insn >> 8) & 0xf;
5414 if (insn & (1 << 7)) {
5422 while ((insn & (1 << (size + 19))) == 0)
5425 shift = (insn >> 16) & ((1 << (3 + size)) - 1);
5426 /* To avoid excessive duplication of ops we implement shift
5427 by immediate using the variable shift operations. */
5429 /* Shift by immediate:
5430 VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
5431 if (q && ((rd | rm) & 1)) {
5434 if (!u && (op == 4 || op == 6)) {
5437 /* Right shifts are encoded as N - shift, where N is the
5438 element size in bits. */
5440 shift = shift - (1 << (size + 3));
5448 imm = (uint8_t) shift;
5453 imm = (uint16_t) shift;
5464 for (pass = 0; pass < count; pass++) {
5466 neon_load_reg64(cpu_V0, rm + pass);
5467 tcg_gen_movi_i64(cpu_V1, imm);
5472 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
5474 gen_helper_neon_shl_s64(cpu_V0, cpu_V0, cpu_V1);
5479 gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, cpu_V1);
5481 gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, cpu_V1);
5484 case 5: /* VSHL, VSLI */
5485 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
5487 case 6: /* VQSHLU */
5488 gen_helper_neon_qshlu_s64(cpu_V0, cpu_env,
5493 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
5496 gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
5501 if (op == 1 || op == 3) {
5503 neon_load_reg64(cpu_V1, rd + pass);
5504 tcg_gen_add_i64(cpu_V0, cpu_V0, cpu_V1);
5505 } else if (op == 4 || (op == 5 && u)) {
5507 neon_load_reg64(cpu_V1, rd + pass);
5509 if (shift < -63 || shift > 63) {
5513 mask = 0xffffffffffffffffull >> -shift;
5515 mask = 0xffffffffffffffffull << shift;
5518 tcg_gen_andi_i64(cpu_V1, cpu_V1, ~mask);
5519 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5521 neon_store_reg64(cpu_V0, rd + pass);
5522 } else { /* size < 3 */
5523 /* Operands in T0 and T1. */
5524 tmp = neon_load_reg(rm, pass);
5525 tmp2 = tcg_temp_new_i32();
5526 tcg_gen_movi_i32(tmp2, imm);
5530 GEN_NEON_INTEGER_OP(shl);
5534 GEN_NEON_INTEGER_OP(rshl);
5537 case 5: /* VSHL, VSLI */
5539 case 0: gen_helper_neon_shl_u8(tmp, tmp, tmp2); break;
5540 case 1: gen_helper_neon_shl_u16(tmp, tmp, tmp2); break;
5541 case 2: gen_helper_neon_shl_u32(tmp, tmp, tmp2); break;
5545 case 6: /* VQSHLU */
5548 gen_helper_neon_qshlu_s8(tmp, cpu_env,
5552 gen_helper_neon_qshlu_s16(tmp, cpu_env,
5556 gen_helper_neon_qshlu_s32(tmp, cpu_env,
5564 GEN_NEON_INTEGER_OP_ENV(qshl);
5567 tcg_temp_free_i32(tmp2);
5569 if (op == 1 || op == 3) {
5571 tmp2 = neon_load_reg(rd, pass);
5572 gen_neon_add(size, tmp, tmp2);
5573 tcg_temp_free_i32(tmp2);
5574 } else if (op == 4 || (op == 5 && u)) {
5579 mask = 0xff >> -shift;
5581 mask = (uint8_t)(0xff << shift);
5587 mask = 0xffff >> -shift;
5589 mask = (uint16_t)(0xffff << shift);
5593 if (shift < -31 || shift > 31) {
5597 mask = 0xffffffffu >> -shift;
5599 mask = 0xffffffffu << shift;
5605 tmp2 = neon_load_reg(rd, pass);
5606 tcg_gen_andi_i32(tmp, tmp, mask);
5607 tcg_gen_andi_i32(tmp2, tmp2, ~mask);
5608 tcg_gen_or_i32(tmp, tmp, tmp2);
5609 tcg_temp_free_i32(tmp2);
5611 neon_store_reg(rd, pass, tmp);
5614 } else if (op < 10) {
5615 /* Shift by immediate and narrow:
5616 VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
5617 int input_unsigned = (op == 8) ? !u : u;
5621 shift = shift - (1 << (size + 3));
5624 tmp64 = tcg_const_i64(shift);
5625 neon_load_reg64(cpu_V0, rm);
5626 neon_load_reg64(cpu_V1, rm + 1);
5627 for (pass = 0; pass < 2; pass++) {
5635 if (input_unsigned) {
5636 gen_helper_neon_rshl_u64(cpu_V0, in, tmp64);
5638 gen_helper_neon_rshl_s64(cpu_V0, in, tmp64);
5641 if (input_unsigned) {
5642 gen_helper_neon_shl_u64(cpu_V0, in, tmp64);
5644 gen_helper_neon_shl_s64(cpu_V0, in, tmp64);
5647 tmp = tcg_temp_new_i32();
5648 gen_neon_narrow_op(op == 8, u, size - 1, tmp, cpu_V0);
5649 neon_store_reg(rd, pass, tmp);
5651 tcg_temp_free_i64(tmp64);
5654 imm = (uint16_t)shift;
5658 imm = (uint32_t)shift;
5660 tmp2 = tcg_const_i32(imm);
5661 tmp4 = neon_load_reg(rm + 1, 0);
5662 tmp5 = neon_load_reg(rm + 1, 1);
5663 for (pass = 0; pass < 2; pass++) {
5665 tmp = neon_load_reg(rm, 0);
5669 gen_neon_shift_narrow(size, tmp, tmp2, q,
5672 tmp3 = neon_load_reg(rm, 1);
5676 gen_neon_shift_narrow(size, tmp3, tmp2, q,
5678 tcg_gen_concat_i32_i64(cpu_V0, tmp, tmp3);
5679 tcg_temp_free_i32(tmp);
5680 tcg_temp_free_i32(tmp3);
5681 tmp = tcg_temp_new_i32();
5682 gen_neon_narrow_op(op == 8, u, size - 1, tmp, cpu_V0);
5683 neon_store_reg(rd, pass, tmp);
5685 tcg_temp_free_i32(tmp2);
5687 } else if (op == 10) {
5689 if (q || (rd & 1)) {
5692 tmp = neon_load_reg(rm, 0);
5693 tmp2 = neon_load_reg(rm, 1);
5694 for (pass = 0; pass < 2; pass++) {
5698 gen_neon_widen(cpu_V0, tmp, size, u);
5701 /* The shift is less than the width of the source
5702 type, so we can just shift the whole register. */
5703 tcg_gen_shli_i64(cpu_V0, cpu_V0, shift);
5704 /* Widen the result of shift: we need to clear
5705 * the potential overflow bits resulting from
5706 * left bits of the narrow input appearing as
5707 * right bits of left the neighbour narrow
5709 if (size < 2 || !u) {
5712 imm = (0xffu >> (8 - shift));
5714 } else if (size == 1) {
5715 imm = 0xffff >> (16 - shift);
5718 imm = 0xffffffff >> (32 - shift);
5721 imm64 = imm | (((uint64_t)imm) << 32);
5725 tcg_gen_andi_i64(cpu_V0, cpu_V0, ~imm64);
5728 neon_store_reg64(cpu_V0, rd + pass);
5730 } else if (op >= 14) {
5731 /* VCVT fixed-point. */
5732 if (!(insn & (1 << 21)) || (q && ((rd | rm) & 1))) {
5735 /* We have already masked out the must-be-1 top bit of imm6,
5736 * hence this 32-shift where the ARM ARM has 64-imm6.
5739 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5740 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, pass));
5743 gen_vfp_ulto(0, shift, 1);
5745 gen_vfp_slto(0, shift, 1);
5748 gen_vfp_toul(0, shift, 1);
5750 gen_vfp_tosl(0, shift, 1);
5752 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, pass));
5757 } else { /* (insn & 0x00380080) == 0 */
5759 if (q && (rd & 1)) {
5763 op = (insn >> 8) & 0xf;
5764 /* One register and immediate. */
5765 imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf);
5766 invert = (insn & (1 << 5)) != 0;
5767 /* Note that op = 2,3,4,5,6,7,10,11,12,13 imm=0 is UNPREDICTABLE.
5768 * We choose to not special-case this and will behave as if a
5769 * valid constant encoding of 0 had been given.
5788 imm = (imm << 8) | (imm << 24);
5791 imm = (imm << 8) | 0xff;
5794 imm = (imm << 16) | 0xffff;
5797 imm |= (imm << 8) | (imm << 16) | (imm << 24);
5805 imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19)
5806 | ((imm & 0x40) ? (0x1f << 25) : (1 << 30));
5812 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5813 if (op & 1 && op < 12) {
5814 tmp = neon_load_reg(rd, pass);
5816 /* The immediate value has already been inverted, so
5818 tcg_gen_andi_i32(tmp, tmp, imm);
5820 tcg_gen_ori_i32(tmp, tmp, imm);
5824 tmp = tcg_temp_new_i32();
5825 if (op == 14 && invert) {
5829 for (n = 0; n < 4; n++) {
5830 if (imm & (1 << (n + (pass & 1) * 4)))
5831 val |= 0xff << (n * 8);
5833 tcg_gen_movi_i32(tmp, val);
5835 tcg_gen_movi_i32(tmp, imm);
5838 neon_store_reg(rd, pass, tmp);
5841 } else { /* (insn & 0x00800010 == 0x00800000) */
5843 op = (insn >> 8) & 0xf;
5844 if ((insn & (1 << 6)) == 0) {
5845 /* Three registers of different lengths. */
5849 /* undefreq: bit 0 : UNDEF if size != 0
5850 * bit 1 : UNDEF if size == 0
5851 * bit 2 : UNDEF if U == 1
5852 * Note that [1:0] set implies 'always UNDEF'
5855 /* prewiden, src1_wide, src2_wide, undefreq */
5856 static const int neon_3reg_wide[16][4] = {
5857 {1, 0, 0, 0}, /* VADDL */
5858 {1, 1, 0, 0}, /* VADDW */
5859 {1, 0, 0, 0}, /* VSUBL */
5860 {1, 1, 0, 0}, /* VSUBW */
5861 {0, 1, 1, 0}, /* VADDHN */
5862 {0, 0, 0, 0}, /* VABAL */
5863 {0, 1, 1, 0}, /* VSUBHN */
5864 {0, 0, 0, 0}, /* VABDL */
5865 {0, 0, 0, 0}, /* VMLAL */
5866 {0, 0, 0, 6}, /* VQDMLAL */
5867 {0, 0, 0, 0}, /* VMLSL */
5868 {0, 0, 0, 6}, /* VQDMLSL */
5869 {0, 0, 0, 0}, /* Integer VMULL */
5870 {0, 0, 0, 2}, /* VQDMULL */
5871 {0, 0, 0, 5}, /* Polynomial VMULL */
5872 {0, 0, 0, 3}, /* Reserved: always UNDEF */
5875 prewiden = neon_3reg_wide[op][0];
5876 src1_wide = neon_3reg_wide[op][1];
5877 src2_wide = neon_3reg_wide[op][2];
5878 undefreq = neon_3reg_wide[op][3];
5880 if (((undefreq & 1) && (size != 0)) ||
5881 ((undefreq & 2) && (size == 0)) ||
5882 ((undefreq & 4) && u)) {
5885 if ((src1_wide && (rn & 1)) ||
5886 (src2_wide && (rm & 1)) ||
5887 (!src2_wide && (rd & 1))) {
5891 /* Avoid overlapping operands. Wide source operands are
5892 always aligned so will never overlap with wide
5893 destinations in problematic ways. */
5894 if (rd == rm && !src2_wide) {
5895 tmp = neon_load_reg(rm, 1);
5896 neon_store_scratch(2, tmp);
5897 } else if (rd == rn && !src1_wide) {
5898 tmp = neon_load_reg(rn, 1);
5899 neon_store_scratch(2, tmp);
5901 TCGV_UNUSED_I32(tmp3);
5902 for (pass = 0; pass < 2; pass++) {
5904 neon_load_reg64(cpu_V0, rn + pass);
5905 TCGV_UNUSED_I32(tmp);
5907 if (pass == 1 && rd == rn) {
5908 tmp = neon_load_scratch(2);
5910 tmp = neon_load_reg(rn, pass);
5913 gen_neon_widen(cpu_V0, tmp, size, u);
5917 neon_load_reg64(cpu_V1, rm + pass);
5918 TCGV_UNUSED_I32(tmp2);
5920 if (pass == 1 && rd == rm) {
5921 tmp2 = neon_load_scratch(2);
5923 tmp2 = neon_load_reg(rm, pass);
5926 gen_neon_widen(cpu_V1, tmp2, size, u);
5930 case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
5931 gen_neon_addl(size);
5933 case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */
5934 gen_neon_subl(size);
5936 case 5: case 7: /* VABAL, VABDL */
5937 switch ((size << 1) | u) {
5939 gen_helper_neon_abdl_s16(cpu_V0, tmp, tmp2);
5942 gen_helper_neon_abdl_u16(cpu_V0, tmp, tmp2);
5945 gen_helper_neon_abdl_s32(cpu_V0, tmp, tmp2);
5948 gen_helper_neon_abdl_u32(cpu_V0, tmp, tmp2);
5951 gen_helper_neon_abdl_s64(cpu_V0, tmp, tmp2);
5954 gen_helper_neon_abdl_u64(cpu_V0, tmp, tmp2);
5958 tcg_temp_free_i32(tmp2);
5959 tcg_temp_free_i32(tmp);
5961 case 8: case 9: case 10: case 11: case 12: case 13:
5962 /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
5963 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
5965 case 14: /* Polynomial VMULL */
5966 gen_helper_neon_mull_p8(cpu_V0, tmp, tmp2);
5967 tcg_temp_free_i32(tmp2);
5968 tcg_temp_free_i32(tmp);
5970 default: /* 15 is RESERVED: caught earlier */
5975 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5976 neon_store_reg64(cpu_V0, rd + pass);
5977 } else if (op == 5 || (op >= 8 && op <= 11)) {
5979 neon_load_reg64(cpu_V1, rd + pass);
5981 case 10: /* VMLSL */
5982 gen_neon_negl(cpu_V0, size);
5984 case 5: case 8: /* VABAL, VMLAL */
5985 gen_neon_addl(size);
5987 case 9: case 11: /* VQDMLAL, VQDMLSL */
5988 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5990 gen_neon_negl(cpu_V0, size);
5992 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
5997 neon_store_reg64(cpu_V0, rd + pass);
5998 } else if (op == 4 || op == 6) {
5999 /* Narrowing operation. */
6000 tmp = tcg_temp_new_i32();
6004 gen_helper_neon_narrow_high_u8(tmp, cpu_V0);
6007 gen_helper_neon_narrow_high_u16(tmp, cpu_V0);
6010 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
6011 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
6018 gen_helper_neon_narrow_round_high_u8(tmp, cpu_V0);
6021 gen_helper_neon_narrow_round_high_u16(tmp, cpu_V0);
6024 tcg_gen_addi_i64(cpu_V0, cpu_V0, 1u << 31);
6025 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
6026 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
6034 neon_store_reg(rd, 0, tmp3);
6035 neon_store_reg(rd, 1, tmp);
6038 /* Write back the result. */
6039 neon_store_reg64(cpu_V0, rd + pass);
6043 /* Two registers and a scalar. NB that for ops of this form
6044 * the ARM ARM labels bit 24 as Q, but it is in our variable
6051 case 1: /* Float VMLA scalar */
6052 case 5: /* Floating point VMLS scalar */
6053 case 9: /* Floating point VMUL scalar */
6058 case 0: /* Integer VMLA scalar */
6059 case 4: /* Integer VMLS scalar */
6060 case 8: /* Integer VMUL scalar */
6061 case 12: /* VQDMULH scalar */
6062 case 13: /* VQRDMULH scalar */
6063 if (u && ((rd | rn) & 1)) {
6066 tmp = neon_get_scalar(size, rm);
6067 neon_store_scratch(0, tmp);
6068 for (pass = 0; pass < (u ? 4 : 2); pass++) {
6069 tmp = neon_load_scratch(0);
6070 tmp2 = neon_load_reg(rn, pass);
6073 gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
6075 gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
6077 } else if (op == 13) {
6079 gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
6081 gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
6083 } else if (op & 1) {
6084 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6085 gen_helper_vfp_muls(tmp, tmp, tmp2, fpstatus);
6086 tcg_temp_free_ptr(fpstatus);
6089 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
6090 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
6091 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
6095 tcg_temp_free_i32(tmp2);
6098 tmp2 = neon_load_reg(rd, pass);
6101 gen_neon_add(size, tmp, tmp2);
6105 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6106 gen_helper_vfp_adds(tmp, tmp, tmp2, fpstatus);
6107 tcg_temp_free_ptr(fpstatus);
6111 gen_neon_rsb(size, tmp, tmp2);
6115 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6116 gen_helper_vfp_subs(tmp, tmp2, tmp, fpstatus);
6117 tcg_temp_free_ptr(fpstatus);
6123 tcg_temp_free_i32(tmp2);
6125 neon_store_reg(rd, pass, tmp);
6128 case 3: /* VQDMLAL scalar */
6129 case 7: /* VQDMLSL scalar */
6130 case 11: /* VQDMULL scalar */
6135 case 2: /* VMLAL sclar */
6136 case 6: /* VMLSL scalar */
6137 case 10: /* VMULL scalar */
6141 tmp2 = neon_get_scalar(size, rm);
6142 /* We need a copy of tmp2 because gen_neon_mull
6143 * deletes it during pass 0. */
6144 tmp4 = tcg_temp_new_i32();
6145 tcg_gen_mov_i32(tmp4, tmp2);
6146 tmp3 = neon_load_reg(rn, 1);
6148 for (pass = 0; pass < 2; pass++) {
6150 tmp = neon_load_reg(rn, 0);
6155 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
6157 neon_load_reg64(cpu_V1, rd + pass);
6161 gen_neon_negl(cpu_V0, size);
6164 gen_neon_addl(size);
6167 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
6169 gen_neon_negl(cpu_V0, size);
6171 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
6177 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
6182 neon_store_reg64(cpu_V0, rd + pass);
6187 default: /* 14 and 15 are RESERVED */
6191 } else { /* size == 3 */
6194 imm = (insn >> 8) & 0xf;
6199 if (q && ((rd | rn | rm) & 1)) {
6204 neon_load_reg64(cpu_V0, rn);
6206 neon_load_reg64(cpu_V1, rn + 1);
6208 } else if (imm == 8) {
6209 neon_load_reg64(cpu_V0, rn + 1);
6211 neon_load_reg64(cpu_V1, rm);
6214 tmp64 = tcg_temp_new_i64();
6216 neon_load_reg64(cpu_V0, rn);
6217 neon_load_reg64(tmp64, rn + 1);
6219 neon_load_reg64(cpu_V0, rn + 1);
6220 neon_load_reg64(tmp64, rm);
6222 tcg_gen_shri_i64(cpu_V0, cpu_V0, (imm & 7) * 8);
6223 tcg_gen_shli_i64(cpu_V1, tmp64, 64 - ((imm & 7) * 8));
6224 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
6226 neon_load_reg64(cpu_V1, rm);
6228 neon_load_reg64(cpu_V1, rm + 1);
6231 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
6232 tcg_gen_shri_i64(tmp64, tmp64, imm * 8);
6233 tcg_gen_or_i64(cpu_V1, cpu_V1, tmp64);
6234 tcg_temp_free_i64(tmp64);
6237 neon_load_reg64(cpu_V0, rn);
6238 tcg_gen_shri_i64(cpu_V0, cpu_V0, imm * 8);
6239 neon_load_reg64(cpu_V1, rm);
6240 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
6241 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
6243 neon_store_reg64(cpu_V0, rd);
6245 neon_store_reg64(cpu_V1, rd + 1);
6247 } else if ((insn & (1 << 11)) == 0) {
6248 /* Two register misc. */
6249 op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf);
6250 size = (insn >> 18) & 3;
6251 /* UNDEF for unknown op values and bad op-size combinations */
6252 if ((neon_2rm_sizes[op] & (1 << size)) == 0) {
6255 if ((op != NEON_2RM_VMOVN && op != NEON_2RM_VQMOVN) &&
6256 q && ((rm | rd) & 1)) {
6260 case NEON_2RM_VREV64:
6261 for (pass = 0; pass < (q ? 2 : 1); pass++) {
6262 tmp = neon_load_reg(rm, pass * 2);
6263 tmp2 = neon_load_reg(rm, pass * 2 + 1);
6265 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
6266 case 1: gen_swap_half(tmp); break;
6267 case 2: /* no-op */ break;
6270 neon_store_reg(rd, pass * 2 + 1, tmp);
6272 neon_store_reg(rd, pass * 2, tmp2);
6275 case 0: tcg_gen_bswap32_i32(tmp2, tmp2); break;
6276 case 1: gen_swap_half(tmp2); break;
6279 neon_store_reg(rd, pass * 2, tmp2);
6283 case NEON_2RM_VPADDL: case NEON_2RM_VPADDL_U:
6284 case NEON_2RM_VPADAL: case NEON_2RM_VPADAL_U:
6285 for (pass = 0; pass < q + 1; pass++) {
6286 tmp = neon_load_reg(rm, pass * 2);
6287 gen_neon_widen(cpu_V0, tmp, size, op & 1);
6288 tmp = neon_load_reg(rm, pass * 2 + 1);
6289 gen_neon_widen(cpu_V1, tmp, size, op & 1);
6291 case 0: gen_helper_neon_paddl_u16(CPU_V001); break;
6292 case 1: gen_helper_neon_paddl_u32(CPU_V001); break;
6293 case 2: tcg_gen_add_i64(CPU_V001); break;
6296 if (op >= NEON_2RM_VPADAL) {
6298 neon_load_reg64(cpu_V1, rd + pass);
6299 gen_neon_addl(size);
6301 neon_store_reg64(cpu_V0, rd + pass);
6307 for (n = 0; n < (q ? 4 : 2); n += 2) {
6308 tmp = neon_load_reg(rm, n);
6309 tmp2 = neon_load_reg(rd, n + 1);
6310 neon_store_reg(rm, n, tmp2);
6311 neon_store_reg(rd, n + 1, tmp);
6318 if (gen_neon_unzip(rd, rm, size, q)) {
6323 if (gen_neon_zip(rd, rm, size, q)) {
6327 case NEON_2RM_VMOVN: case NEON_2RM_VQMOVN:
6328 /* also VQMOVUN; op field and mnemonics don't line up */
6332 TCGV_UNUSED_I32(tmp2);
6333 for (pass = 0; pass < 2; pass++) {
6334 neon_load_reg64(cpu_V0, rm + pass);
6335 tmp = tcg_temp_new_i32();
6336 gen_neon_narrow_op(op == NEON_2RM_VMOVN, q, size,
6341 neon_store_reg(rd, 0, tmp2);
6342 neon_store_reg(rd, 1, tmp);
6346 case NEON_2RM_VSHLL:
6347 if (q || (rd & 1)) {
6350 tmp = neon_load_reg(rm, 0);
6351 tmp2 = neon_load_reg(rm, 1);
6352 for (pass = 0; pass < 2; pass++) {
6355 gen_neon_widen(cpu_V0, tmp, size, 1);
6356 tcg_gen_shli_i64(cpu_V0, cpu_V0, 8 << size);
6357 neon_store_reg64(cpu_V0, rd + pass);
6360 case NEON_2RM_VCVT_F16_F32:
6361 if (!arm_feature(env, ARM_FEATURE_VFP_FP16) ||
6365 tmp = tcg_temp_new_i32();
6366 tmp2 = tcg_temp_new_i32();
6367 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 0));
6368 gen_helper_neon_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
6369 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 1));
6370 gen_helper_neon_fcvt_f32_to_f16(tmp2, cpu_F0s, cpu_env);
6371 tcg_gen_shli_i32(tmp2, tmp2, 16);
6372 tcg_gen_or_i32(tmp2, tmp2, tmp);
6373 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 2));
6374 gen_helper_neon_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
6375 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 3));
6376 neon_store_reg(rd, 0, tmp2);
6377 tmp2 = tcg_temp_new_i32();
6378 gen_helper_neon_fcvt_f32_to_f16(tmp2, cpu_F0s, cpu_env);
6379 tcg_gen_shli_i32(tmp2, tmp2, 16);
6380 tcg_gen_or_i32(tmp2, tmp2, tmp);
6381 neon_store_reg(rd, 1, tmp2);
6382 tcg_temp_free_i32(tmp);
6384 case NEON_2RM_VCVT_F32_F16:
6385 if (!arm_feature(env, ARM_FEATURE_VFP_FP16) ||
6389 tmp3 = tcg_temp_new_i32();
6390 tmp = neon_load_reg(rm, 0);
6391 tmp2 = neon_load_reg(rm, 1);
6392 tcg_gen_ext16u_i32(tmp3, tmp);
6393 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
6394 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 0));
6395 tcg_gen_shri_i32(tmp3, tmp, 16);
6396 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
6397 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 1));
6398 tcg_temp_free_i32(tmp);
6399 tcg_gen_ext16u_i32(tmp3, tmp2);
6400 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
6401 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 2));
6402 tcg_gen_shri_i32(tmp3, tmp2, 16);
6403 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
6404 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 3));
6405 tcg_temp_free_i32(tmp2);
6406 tcg_temp_free_i32(tmp3);
6408 case NEON_2RM_AESE: case NEON_2RM_AESMC:
6409 if (!arm_feature(env, ARM_FEATURE_V8_AES)
6410 || ((rm | rd) & 1)) {
6413 tmp = tcg_const_i32(rd);
6414 tmp2 = tcg_const_i32(rm);
6416 /* Bit 6 is the lowest opcode bit; it distinguishes between
6417 * encryption (AESE/AESMC) and decryption (AESD/AESIMC)
6419 tmp3 = tcg_const_i32(extract32(insn, 6, 1));
6421 if (op == NEON_2RM_AESE) {
6422 gen_helper_crypto_aese(cpu_env, tmp, tmp2, tmp3);
6424 gen_helper_crypto_aesmc(cpu_env, tmp, tmp2, tmp3);
6426 tcg_temp_free_i32(tmp);
6427 tcg_temp_free_i32(tmp2);
6428 tcg_temp_free_i32(tmp3);
6432 for (pass = 0; pass < (q ? 4 : 2); pass++) {
6433 if (neon_2rm_is_float_op(op)) {
6434 tcg_gen_ld_f32(cpu_F0s, cpu_env,
6435 neon_reg_offset(rm, pass));
6436 TCGV_UNUSED_I32(tmp);
6438 tmp = neon_load_reg(rm, pass);
6441 case NEON_2RM_VREV32:
6443 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
6444 case 1: gen_swap_half(tmp); break;
6448 case NEON_2RM_VREV16:
6453 case 0: gen_helper_neon_cls_s8(tmp, tmp); break;
6454 case 1: gen_helper_neon_cls_s16(tmp, tmp); break;
6455 case 2: gen_helper_neon_cls_s32(tmp, tmp); break;
6461 case 0: gen_helper_neon_clz_u8(tmp, tmp); break;
6462 case 1: gen_helper_neon_clz_u16(tmp, tmp); break;
6463 case 2: gen_helper_clz(tmp, tmp); break;
6468 gen_helper_neon_cnt_u8(tmp, tmp);
6471 tcg_gen_not_i32(tmp, tmp);
6473 case NEON_2RM_VQABS:
6476 gen_helper_neon_qabs_s8(tmp, cpu_env, tmp);
6479 gen_helper_neon_qabs_s16(tmp, cpu_env, tmp);
6482 gen_helper_neon_qabs_s32(tmp, cpu_env, tmp);
6487 case NEON_2RM_VQNEG:
6490 gen_helper_neon_qneg_s8(tmp, cpu_env, tmp);
6493 gen_helper_neon_qneg_s16(tmp, cpu_env, tmp);
6496 gen_helper_neon_qneg_s32(tmp, cpu_env, tmp);
6501 case NEON_2RM_VCGT0: case NEON_2RM_VCLE0:
6502 tmp2 = tcg_const_i32(0);
6504 case 0: gen_helper_neon_cgt_s8(tmp, tmp, tmp2); break;
6505 case 1: gen_helper_neon_cgt_s16(tmp, tmp, tmp2); break;
6506 case 2: gen_helper_neon_cgt_s32(tmp, tmp, tmp2); break;
6509 tcg_temp_free_i32(tmp2);
6510 if (op == NEON_2RM_VCLE0) {
6511 tcg_gen_not_i32(tmp, tmp);
6514 case NEON_2RM_VCGE0: case NEON_2RM_VCLT0:
6515 tmp2 = tcg_const_i32(0);
6517 case 0: gen_helper_neon_cge_s8(tmp, tmp, tmp2); break;
6518 case 1: gen_helper_neon_cge_s16(tmp, tmp, tmp2); break;
6519 case 2: gen_helper_neon_cge_s32(tmp, tmp, tmp2); break;
6522 tcg_temp_free_i32(tmp2);
6523 if (op == NEON_2RM_VCLT0) {
6524 tcg_gen_not_i32(tmp, tmp);
6527 case NEON_2RM_VCEQ0:
6528 tmp2 = tcg_const_i32(0);
6530 case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break;
6531 case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break;
6532 case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break;
6535 tcg_temp_free_i32(tmp2);
6539 case 0: gen_helper_neon_abs_s8(tmp, tmp); break;
6540 case 1: gen_helper_neon_abs_s16(tmp, tmp); break;
6541 case 2: tcg_gen_abs_i32(tmp, tmp); break;
6546 tmp2 = tcg_const_i32(0);
6547 gen_neon_rsb(size, tmp, tmp2);
6548 tcg_temp_free_i32(tmp2);
6550 case NEON_2RM_VCGT0_F:
6552 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6553 tmp2 = tcg_const_i32(0);
6554 gen_helper_neon_cgt_f32(tmp, tmp, tmp2, fpstatus);
6555 tcg_temp_free_i32(tmp2);
6556 tcg_temp_free_ptr(fpstatus);
6559 case NEON_2RM_VCGE0_F:
6561 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6562 tmp2 = tcg_const_i32(0);
6563 gen_helper_neon_cge_f32(tmp, tmp, tmp2, fpstatus);
6564 tcg_temp_free_i32(tmp2);
6565 tcg_temp_free_ptr(fpstatus);
6568 case NEON_2RM_VCEQ0_F:
6570 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6571 tmp2 = tcg_const_i32(0);
6572 gen_helper_neon_ceq_f32(tmp, tmp, tmp2, fpstatus);
6573 tcg_temp_free_i32(tmp2);
6574 tcg_temp_free_ptr(fpstatus);
6577 case NEON_2RM_VCLE0_F:
6579 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6580 tmp2 = tcg_const_i32(0);
6581 gen_helper_neon_cge_f32(tmp, tmp2, tmp, fpstatus);
6582 tcg_temp_free_i32(tmp2);
6583 tcg_temp_free_ptr(fpstatus);
6586 case NEON_2RM_VCLT0_F:
6588 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6589 tmp2 = tcg_const_i32(0);
6590 gen_helper_neon_cgt_f32(tmp, tmp2, tmp, fpstatus);
6591 tcg_temp_free_i32(tmp2);
6592 tcg_temp_free_ptr(fpstatus);
6595 case NEON_2RM_VABS_F:
6598 case NEON_2RM_VNEG_F:
6602 tmp2 = neon_load_reg(rd, pass);
6603 neon_store_reg(rm, pass, tmp2);
6606 tmp2 = neon_load_reg(rd, pass);
6608 case 0: gen_neon_trn_u8(tmp, tmp2); break;
6609 case 1: gen_neon_trn_u16(tmp, tmp2); break;
6612 neon_store_reg(rm, pass, tmp2);
6614 case NEON_2RM_VRINTN:
6615 case NEON_2RM_VRINTA:
6616 case NEON_2RM_VRINTM:
6617 case NEON_2RM_VRINTP:
6618 case NEON_2RM_VRINTZ:
6621 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6624 if (op == NEON_2RM_VRINTZ) {
6625 rmode = FPROUNDING_ZERO;
6627 rmode = fp_decode_rm[((op & 0x6) >> 1) ^ 1];
6630 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode));
6631 gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode,
6633 gen_helper_rints(cpu_F0s, cpu_F0s, fpstatus);
6634 gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode,
6636 tcg_temp_free_ptr(fpstatus);
6637 tcg_temp_free_i32(tcg_rmode);
6640 case NEON_2RM_VRINTX:
6642 TCGv_ptr fpstatus = get_fpstatus_ptr(1);
6643 gen_helper_rints_exact(cpu_F0s, cpu_F0s, fpstatus);
6644 tcg_temp_free_ptr(fpstatus);
6647 case NEON_2RM_VCVTAU:
6648 case NEON_2RM_VCVTAS:
6649 case NEON_2RM_VCVTNU:
6650 case NEON_2RM_VCVTNS:
6651 case NEON_2RM_VCVTPU:
6652 case NEON_2RM_VCVTPS:
6653 case NEON_2RM_VCVTMU:
6654 case NEON_2RM_VCVTMS:
6656 bool is_signed = !extract32(insn, 7, 1);
6657 TCGv_ptr fpst = get_fpstatus_ptr(1);
6658 TCGv_i32 tcg_rmode, tcg_shift;
6659 int rmode = fp_decode_rm[extract32(insn, 8, 2)];
6661 tcg_shift = tcg_const_i32(0);
6662 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode));
6663 gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode,
6667 gen_helper_vfp_tosls(cpu_F0s, cpu_F0s,
6670 gen_helper_vfp_touls(cpu_F0s, cpu_F0s,
6674 gen_helper_set_neon_rmode(tcg_rmode, tcg_rmode,
6676 tcg_temp_free_i32(tcg_rmode);
6677 tcg_temp_free_i32(tcg_shift);
6678 tcg_temp_free_ptr(fpst);
6681 case NEON_2RM_VRECPE:
6682 gen_helper_recpe_u32(tmp, tmp, cpu_env);
6684 case NEON_2RM_VRSQRTE:
6685 gen_helper_rsqrte_u32(tmp, tmp, cpu_env);
6687 case NEON_2RM_VRECPE_F:
6688 gen_helper_recpe_f32(cpu_F0s, cpu_F0s, cpu_env);
6690 case NEON_2RM_VRSQRTE_F:
6691 gen_helper_rsqrte_f32(cpu_F0s, cpu_F0s, cpu_env);
6693 case NEON_2RM_VCVT_FS: /* VCVT.F32.S32 */
6696 case NEON_2RM_VCVT_FU: /* VCVT.F32.U32 */
6699 case NEON_2RM_VCVT_SF: /* VCVT.S32.F32 */
6700 gen_vfp_tosiz(0, 1);
6702 case NEON_2RM_VCVT_UF: /* VCVT.U32.F32 */
6703 gen_vfp_touiz(0, 1);
6706 /* Reserved op values were caught by the
6707 * neon_2rm_sizes[] check earlier.
6711 if (neon_2rm_is_float_op(op)) {
6712 tcg_gen_st_f32(cpu_F0s, cpu_env,
6713 neon_reg_offset(rd, pass));
6715 neon_store_reg(rd, pass, tmp);
6720 } else if ((insn & (1 << 10)) == 0) {
6722 int n = ((insn >> 8) & 3) + 1;
6723 if ((rn + n) > 32) {
6724 /* This is UNPREDICTABLE; we choose to UNDEF to avoid the
6725 * helper function running off the end of the register file.
6730 if (insn & (1 << 6)) {
6731 tmp = neon_load_reg(rd, 0);
6733 tmp = tcg_temp_new_i32();
6734 tcg_gen_movi_i32(tmp, 0);
6736 tmp2 = neon_load_reg(rm, 0);
6737 tmp4 = tcg_const_i32(rn);
6738 tmp5 = tcg_const_i32(n);
6739 gen_helper_neon_tbl(tmp2, cpu_env, tmp2, tmp, tmp4, tmp5);
6740 tcg_temp_free_i32(tmp);
6741 if (insn & (1 << 6)) {
6742 tmp = neon_load_reg(rd, 1);
6744 tmp = tcg_temp_new_i32();
6745 tcg_gen_movi_i32(tmp, 0);
6747 tmp3 = neon_load_reg(rm, 1);
6748 gen_helper_neon_tbl(tmp3, cpu_env, tmp3, tmp, tmp4, tmp5);
6749 tcg_temp_free_i32(tmp5);
6750 tcg_temp_free_i32(tmp4);
6751 neon_store_reg(rd, 0, tmp2);
6752 neon_store_reg(rd, 1, tmp3);
6753 tcg_temp_free_i32(tmp);
6754 } else if ((insn & 0x380) == 0) {
6756 if ((insn & (7 << 16)) == 0 || (q && (rd & 1))) {
6759 if (insn & (1 << 19)) {
6760 tmp = neon_load_reg(rm, 1);
6762 tmp = neon_load_reg(rm, 0);
6764 if (insn & (1 << 16)) {
6765 gen_neon_dup_u8(tmp, ((insn >> 17) & 3) * 8);
6766 } else if (insn & (1 << 17)) {
6767 if ((insn >> 18) & 1)
6768 gen_neon_dup_high16(tmp);
6770 gen_neon_dup_low16(tmp);
6772 for (pass = 0; pass < (q ? 4 : 2); pass++) {
6773 tmp2 = tcg_temp_new_i32();
6774 tcg_gen_mov_i32(tmp2, tmp);
6775 neon_store_reg(rd, pass, tmp2);
6777 tcg_temp_free_i32(tmp);
6786 static int disas_coproc_insn(CPUARMState * env, DisasContext *s, uint32_t insn)
6788 int cpnum, is64, crn, crm, opc1, opc2, isread, rt, rt2;
6789 const ARMCPRegInfo *ri;
6791 cpnum = (insn >> 8) & 0xf;
6792 if (arm_feature(env, ARM_FEATURE_XSCALE)
6793 && ((env->cp15.c15_cpar ^ 0x3fff) & (1 << cpnum)))
6796 /* First check for coprocessor space used for actual instructions */
6800 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
6801 return disas_iwmmxt_insn(env, s, insn);
6802 } else if (arm_feature(env, ARM_FEATURE_XSCALE)) {
6803 return disas_dsp_insn(env, s, insn);
6810 /* Otherwise treat as a generic register access */
6811 is64 = (insn & (1 << 25)) == 0;
6812 if (!is64 && ((insn & (1 << 4)) == 0)) {
6820 opc1 = (insn >> 4) & 0xf;
6822 rt2 = (insn >> 16) & 0xf;
6824 crn = (insn >> 16) & 0xf;
6825 opc1 = (insn >> 21) & 7;
6826 opc2 = (insn >> 5) & 7;
6829 isread = (insn >> 20) & 1;
6830 rt = (insn >> 12) & 0xf;
6832 ri = get_arm_cp_reginfo(s->cp_regs,
6833 ENCODE_CP_REG(cpnum, is64, crn, crm, opc1, opc2));
6835 /* Check access permissions */
6836 if (!cp_access_ok(s->current_pl, ri, isread)) {
6841 /* Emit code to perform further access permissions checks at
6842 * runtime; this may result in an exception.
6845 gen_set_pc_im(s, s->pc);
6846 tmpptr = tcg_const_ptr(ri);
6847 gen_helper_access_check_cp_reg(cpu_env, tmpptr);
6848 tcg_temp_free_ptr(tmpptr);
6851 /* Handle special cases first */
6852 switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) {
6859 gen_set_pc_im(s, s->pc);
6860 s->is_jmp = DISAS_WFI;
6866 if (use_icount && (ri->type & ARM_CP_IO)) {
6875 if (ri->type & ARM_CP_CONST) {
6876 tmp64 = tcg_const_i64(ri->resetvalue);
6877 } else if (ri->readfn) {
6879 tmp64 = tcg_temp_new_i64();
6880 tmpptr = tcg_const_ptr(ri);
6881 gen_helper_get_cp_reg64(tmp64, cpu_env, tmpptr);
6882 tcg_temp_free_ptr(tmpptr);
6884 tmp64 = tcg_temp_new_i64();
6885 tcg_gen_ld_i64(tmp64, cpu_env, ri->fieldoffset);
6887 tmp = tcg_temp_new_i32();
6888 tcg_gen_trunc_i64_i32(tmp, tmp64);
6889 store_reg(s, rt, tmp);
6890 tcg_gen_shri_i64(tmp64, tmp64, 32);
6891 tmp = tcg_temp_new_i32();
6892 tcg_gen_trunc_i64_i32(tmp, tmp64);
6893 tcg_temp_free_i64(tmp64);
6894 store_reg(s, rt2, tmp);
6897 if (ri->type & ARM_CP_CONST) {
6898 tmp = tcg_const_i32(ri->resetvalue);
6899 } else if (ri->readfn) {
6901 tmp = tcg_temp_new_i32();
6902 tmpptr = tcg_const_ptr(ri);
6903 gen_helper_get_cp_reg(tmp, cpu_env, tmpptr);
6904 tcg_temp_free_ptr(tmpptr);
6906 tmp = load_cpu_offset(ri->fieldoffset);
6909 /* Destination register of r15 for 32 bit loads sets
6910 * the condition codes from the high 4 bits of the value
6913 tcg_temp_free_i32(tmp);
6915 store_reg(s, rt, tmp);
6920 if (ri->type & ARM_CP_CONST) {
6921 /* If not forbidden by access permissions, treat as WI */
6926 TCGv_i32 tmplo, tmphi;
6927 TCGv_i64 tmp64 = tcg_temp_new_i64();
6928 tmplo = load_reg(s, rt);
6929 tmphi = load_reg(s, rt2);
6930 tcg_gen_concat_i32_i64(tmp64, tmplo, tmphi);
6931 tcg_temp_free_i32(tmplo);
6932 tcg_temp_free_i32(tmphi);
6934 TCGv_ptr tmpptr = tcg_const_ptr(ri);
6935 gen_helper_set_cp_reg64(cpu_env, tmpptr, tmp64);
6936 tcg_temp_free_ptr(tmpptr);
6938 tcg_gen_st_i64(tmp64, cpu_env, ri->fieldoffset);
6940 tcg_temp_free_i64(tmp64);
6945 tmp = load_reg(s, rt);
6946 tmpptr = tcg_const_ptr(ri);
6947 gen_helper_set_cp_reg(cpu_env, tmpptr, tmp);
6948 tcg_temp_free_ptr(tmpptr);
6949 tcg_temp_free_i32(tmp);
6951 TCGv_i32 tmp = load_reg(s, rt);
6952 store_cpu_offset(tmp, ri->fieldoffset);
6957 if (use_icount && (ri->type & ARM_CP_IO)) {
6958 /* I/O operations must end the TB here (whether read or write) */
6961 } else if (!isread && !(ri->type & ARM_CP_SUPPRESS_TB_END)) {
6962 /* We default to ending the TB on a coprocessor register write,
6963 * but allow this to be suppressed by the register definition
6964 * (usually only necessary to work around guest bugs).
6972 /* Unknown register; this might be a guest error or a QEMU
6973 * unimplemented feature.
6976 qemu_log_mask(LOG_UNIMP, "%s access to unsupported AArch32 "
6977 "64 bit system register cp:%d opc1: %d crm:%d\n",
6978 isread ? "read" : "write", cpnum, opc1, crm);
6980 qemu_log_mask(LOG_UNIMP, "%s access to unsupported AArch32 "
6981 "system register cp:%d opc1:%d crn:%d crm:%d opc2:%d\n",
6982 isread ? "read" : "write", cpnum, opc1, crn, crm, opc2);
6989 /* Store a 64-bit value to a register pair. Clobbers val. */
6990 static void gen_storeq_reg(DisasContext *s, int rlow, int rhigh, TCGv_i64 val)
6993 tmp = tcg_temp_new_i32();
6994 tcg_gen_trunc_i64_i32(tmp, val);
6995 store_reg(s, rlow, tmp);
6996 tmp = tcg_temp_new_i32();
6997 tcg_gen_shri_i64(val, val, 32);
6998 tcg_gen_trunc_i64_i32(tmp, val);
6999 store_reg(s, rhigh, tmp);
7002 /* load a 32-bit value from a register and perform a 64-bit accumulate. */
7003 static void gen_addq_lo(DisasContext *s, TCGv_i64 val, int rlow)
7008 /* Load value and extend to 64 bits. */
7009 tmp = tcg_temp_new_i64();
7010 tmp2 = load_reg(s, rlow);
7011 tcg_gen_extu_i32_i64(tmp, tmp2);
7012 tcg_temp_free_i32(tmp2);
7013 tcg_gen_add_i64(val, val, tmp);
7014 tcg_temp_free_i64(tmp);
7017 /* load and add a 64-bit value from a register pair. */
7018 static void gen_addq(DisasContext *s, TCGv_i64 val, int rlow, int rhigh)
7024 /* Load 64-bit value rd:rn. */
7025 tmpl = load_reg(s, rlow);
7026 tmph = load_reg(s, rhigh);
7027 tmp = tcg_temp_new_i64();
7028 tcg_gen_concat_i32_i64(tmp, tmpl, tmph);
7029 tcg_temp_free_i32(tmpl);
7030 tcg_temp_free_i32(tmph);
7031 tcg_gen_add_i64(val, val, tmp);
7032 tcg_temp_free_i64(tmp);
7035 /* Set N and Z flags from hi|lo. */
7036 static void gen_logicq_cc(TCGv_i32 lo, TCGv_i32 hi)
7038 tcg_gen_mov_i32(cpu_NF, hi);
7039 tcg_gen_or_i32(cpu_ZF, lo, hi);
7042 /* Load/Store exclusive instructions are implemented by remembering
7043 the value/address loaded, and seeing if these are the same
7044 when the store is performed. This should be sufficient to implement
7045 the architecturally mandated semantics, and avoids having to monitor
7048 In system emulation mode only one CPU will be running at once, so
7049 this sequence is effectively atomic. In user emulation mode we
7050 throw an exception and handle the atomic operation elsewhere. */
7051 static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
7052 TCGv_i32 addr, int size)
7054 TCGv_i32 tmp = tcg_temp_new_i32();
7058 gen_aa32_ld8u(tmp, addr, IS_USER(s));
7061 gen_aa32_ld16u(tmp, addr, IS_USER(s));
7065 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7072 TCGv_i32 tmp2 = tcg_temp_new_i32();
7073 TCGv_i32 tmp3 = tcg_temp_new_i32();
7075 tcg_gen_addi_i32(tmp2, addr, 4);
7076 gen_aa32_ld32u(tmp3, tmp2, IS_USER(s));
7077 tcg_temp_free_i32(tmp2);
7078 tcg_gen_concat_i32_i64(cpu_exclusive_val, tmp, tmp3);
7079 store_reg(s, rt2, tmp3);
7081 tcg_gen_extu_i32_i64(cpu_exclusive_val, tmp);
7084 store_reg(s, rt, tmp);
7085 tcg_gen_extu_i32_i64(cpu_exclusive_addr, addr);
7088 static void gen_clrex(DisasContext *s)
7090 tcg_gen_movi_i64(cpu_exclusive_addr, -1);
7093 #ifdef CONFIG_USER_ONLY
7094 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
7095 TCGv_i32 addr, int size)
7097 tcg_gen_extu_i32_i64(cpu_exclusive_test, addr);
7098 tcg_gen_movi_i32(cpu_exclusive_info,
7099 size | (rd << 4) | (rt << 8) | (rt2 << 12));
7100 gen_exception_insn(s, 4, EXCP_STREX);
7103 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
7104 TCGv_i32 addr, int size)
7107 TCGv_i64 val64, extaddr;
7111 /* if (env->exclusive_addr == addr && env->exclusive_val == [addr]) {
7117 fail_label = gen_new_label();
7118 done_label = gen_new_label();
7119 extaddr = tcg_temp_new_i64();
7120 tcg_gen_extu_i32_i64(extaddr, addr);
7121 tcg_gen_brcond_i64(TCG_COND_NE, extaddr, cpu_exclusive_addr, fail_label);
7122 tcg_temp_free_i64(extaddr);
7124 tmp = tcg_temp_new_i32();
7127 gen_aa32_ld8u(tmp, addr, IS_USER(s));
7130 gen_aa32_ld16u(tmp, addr, IS_USER(s));
7134 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7140 val64 = tcg_temp_new_i64();
7142 TCGv_i32 tmp2 = tcg_temp_new_i32();
7143 TCGv_i32 tmp3 = tcg_temp_new_i32();
7144 tcg_gen_addi_i32(tmp2, addr, 4);
7145 gen_aa32_ld32u(tmp3, tmp2, IS_USER(s));
7146 tcg_temp_free_i32(tmp2);
7147 tcg_gen_concat_i32_i64(val64, tmp, tmp3);
7148 tcg_temp_free_i32(tmp3);
7150 tcg_gen_extu_i32_i64(val64, tmp);
7152 tcg_temp_free_i32(tmp);
7154 tcg_gen_brcond_i64(TCG_COND_NE, val64, cpu_exclusive_val, fail_label);
7155 tcg_temp_free_i64(val64);
7157 tmp = load_reg(s, rt);
7160 gen_aa32_st8(tmp, addr, IS_USER(s));
7163 gen_aa32_st16(tmp, addr, IS_USER(s));
7167 gen_aa32_st32(tmp, addr, IS_USER(s));
7172 tcg_temp_free_i32(tmp);
7174 tcg_gen_addi_i32(addr, addr, 4);
7175 tmp = load_reg(s, rt2);
7176 gen_aa32_st32(tmp, addr, IS_USER(s));
7177 tcg_temp_free_i32(tmp);
7179 tcg_gen_movi_i32(cpu_R[rd], 0);
7180 tcg_gen_br(done_label);
7181 gen_set_label(fail_label);
7182 tcg_gen_movi_i32(cpu_R[rd], 1);
7183 gen_set_label(done_label);
7184 tcg_gen_movi_i64(cpu_exclusive_addr, -1);
7191 * @mode: mode field from insn (which stack to store to)
7192 * @amode: addressing mode (DA/IA/DB/IB), encoded as per P,U bits in ARM insn
7193 * @writeback: true if writeback bit set
7195 * Generate code for the SRS (Store Return State) insn.
7197 static void gen_srs(DisasContext *s,
7198 uint32_t mode, uint32_t amode, bool writeback)
7201 TCGv_i32 addr = tcg_temp_new_i32();
7202 TCGv_i32 tmp = tcg_const_i32(mode);
7203 gen_helper_get_r13_banked(addr, cpu_env, tmp);
7204 tcg_temp_free_i32(tmp);
7221 tcg_gen_addi_i32(addr, addr, offset);
7222 tmp = load_reg(s, 14);
7223 gen_aa32_st32(tmp, addr, 0);
7224 tcg_temp_free_i32(tmp);
7225 tmp = load_cpu_field(spsr);
7226 tcg_gen_addi_i32(addr, addr, 4);
7227 gen_aa32_st32(tmp, addr, 0);
7228 tcg_temp_free_i32(tmp);
7246 tcg_gen_addi_i32(addr, addr, offset);
7247 tmp = tcg_const_i32(mode);
7248 gen_helper_set_r13_banked(cpu_env, tmp, addr);
7249 tcg_temp_free_i32(tmp);
7251 tcg_temp_free_i32(addr);
7254 static void disas_arm_insn(CPUARMState * env, DisasContext *s)
7256 unsigned int cond, insn, val, op1, i, shift, rm, rs, rn, rd, sh;
7263 insn = arm_ldl_code(env, s->pc, s->bswap_code);
7266 /* M variants do not implement ARM mode. */
7271 /* In ARMv3 and v4 the NV condition is UNPREDICTABLE; we
7272 * choose to UNDEF. In ARMv5 and above the space is used
7273 * for miscellaneous unconditional instructions.
7277 /* Unconditional instructions. */
7278 if (((insn >> 25) & 7) == 1) {
7279 /* NEON Data processing. */
7280 if (!arm_feature(env, ARM_FEATURE_NEON))
7283 if (disas_neon_data_insn(env, s, insn))
7287 if ((insn & 0x0f100000) == 0x04000000) {
7288 /* NEON load/store. */
7289 if (!arm_feature(env, ARM_FEATURE_NEON))
7292 if (disas_neon_ls_insn(env, s, insn))
7296 if ((insn & 0x0f000e10) == 0x0e000a00) {
7298 if (disas_vfp_insn(env, s, insn)) {
7303 if (((insn & 0x0f30f000) == 0x0510f000) ||
7304 ((insn & 0x0f30f010) == 0x0710f000)) {
7305 if ((insn & (1 << 22)) == 0) {
7307 if (!arm_feature(env, ARM_FEATURE_V7MP)) {
7311 /* Otherwise PLD; v5TE+ */
7315 if (((insn & 0x0f70f000) == 0x0450f000) ||
7316 ((insn & 0x0f70f010) == 0x0650f000)) {
7318 return; /* PLI; V7 */
7320 if (((insn & 0x0f700000) == 0x04100000) ||
7321 ((insn & 0x0f700010) == 0x06100000)) {
7322 if (!arm_feature(env, ARM_FEATURE_V7MP)) {
7325 return; /* v7MP: Unallocated memory hint: must NOP */
7328 if ((insn & 0x0ffffdff) == 0x01010000) {
7331 if (((insn >> 9) & 1) != s->bswap_code) {
7332 /* Dynamic endianness switching not implemented. */
7333 qemu_log_mask(LOG_UNIMP, "arm: unimplemented setend\n");
7337 } else if ((insn & 0x0fffff00) == 0x057ff000) {
7338 switch ((insn >> 4) & 0xf) {
7347 /* We don't emulate caches so these are a no-op. */
7352 } else if ((insn & 0x0e5fffe0) == 0x084d0500) {
7358 gen_srs(s, (insn & 0x1f), (insn >> 23) & 3, insn & (1 << 21));
7360 } else if ((insn & 0x0e50ffe0) == 0x08100a00) {
7366 rn = (insn >> 16) & 0xf;
7367 addr = load_reg(s, rn);
7368 i = (insn >> 23) & 3;
7370 case 0: offset = -4; break; /* DA */
7371 case 1: offset = 0; break; /* IA */
7372 case 2: offset = -8; break; /* DB */
7373 case 3: offset = 4; break; /* IB */
7377 tcg_gen_addi_i32(addr, addr, offset);
7378 /* Load PC into tmp and CPSR into tmp2. */
7379 tmp = tcg_temp_new_i32();
7380 gen_aa32_ld32u(tmp, addr, 0);
7381 tcg_gen_addi_i32(addr, addr, 4);
7382 tmp2 = tcg_temp_new_i32();
7383 gen_aa32_ld32u(tmp2, addr, 0);
7384 if (insn & (1 << 21)) {
7385 /* Base writeback. */
7387 case 0: offset = -8; break;
7388 case 1: offset = 4; break;
7389 case 2: offset = -4; break;
7390 case 3: offset = 0; break;
7394 tcg_gen_addi_i32(addr, addr, offset);
7395 store_reg(s, rn, addr);
7397 tcg_temp_free_i32(addr);
7399 gen_rfe(s, tmp, tmp2);
7401 } else if ((insn & 0x0e000000) == 0x0a000000) {
7402 /* branch link and change to thumb (blx <offset>) */
7405 val = (uint32_t)s->pc;
7406 tmp = tcg_temp_new_i32();
7407 tcg_gen_movi_i32(tmp, val);
7408 store_reg(s, 14, tmp);
7409 /* Sign-extend the 24-bit offset */
7410 offset = (((int32_t)insn) << 8) >> 8;
7411 /* offset * 4 + bit24 * 2 + (thumb bit) */
7412 val += (offset << 2) | ((insn >> 23) & 2) | 1;
7413 /* pipeline offset */
7415 /* protected by ARCH(5); above, near the start of uncond block */
7418 } else if ((insn & 0x0e000f00) == 0x0c000100) {
7419 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
7420 /* iWMMXt register transfer. */
7421 if (env->cp15.c15_cpar & (1 << 1))
7422 if (!disas_iwmmxt_insn(env, s, insn))
7425 } else if ((insn & 0x0fe00000) == 0x0c400000) {
7426 /* Coprocessor double register transfer. */
7428 } else if ((insn & 0x0f000010) == 0x0e000010) {
7429 /* Additional coprocessor register transfer. */
7430 } else if ((insn & 0x0ff10020) == 0x01000000) {
7433 /* cps (privileged) */
7437 if (insn & (1 << 19)) {
7438 if (insn & (1 << 8))
7440 if (insn & (1 << 7))
7442 if (insn & (1 << 6))
7444 if (insn & (1 << 18))
7447 if (insn & (1 << 17)) {
7449 val |= (insn & 0x1f);
7452 gen_set_psr_im(s, mask, 0, val);
7459 /* if not always execute, we generate a conditional jump to
7461 s->condlabel = gen_new_label();
7462 arm_gen_test_cc(cond ^ 1, s->condlabel);
7465 if ((insn & 0x0f900000) == 0x03000000) {
7466 if ((insn & (1 << 21)) == 0) {
7468 rd = (insn >> 12) & 0xf;
7469 val = ((insn >> 4) & 0xf000) | (insn & 0xfff);
7470 if ((insn & (1 << 22)) == 0) {
7472 tmp = tcg_temp_new_i32();
7473 tcg_gen_movi_i32(tmp, val);
7476 tmp = load_reg(s, rd);
7477 tcg_gen_ext16u_i32(tmp, tmp);
7478 tcg_gen_ori_i32(tmp, tmp, val << 16);
7480 store_reg(s, rd, tmp);
7482 if (((insn >> 12) & 0xf) != 0xf)
7484 if (((insn >> 16) & 0xf) == 0) {
7485 gen_nop_hint(s, insn & 0xff);
7487 /* CPSR = immediate */
7489 shift = ((insn >> 8) & 0xf) * 2;
7491 val = (val >> shift) | (val << (32 - shift));
7492 i = ((insn & (1 << 22)) != 0);
7493 if (gen_set_psr_im(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i, val))
7497 } else if ((insn & 0x0f900000) == 0x01000000
7498 && (insn & 0x00000090) != 0x00000090) {
7499 /* miscellaneous instructions */
7500 op1 = (insn >> 21) & 3;
7501 sh = (insn >> 4) & 0xf;
7504 case 0x0: /* move program status register */
7507 tmp = load_reg(s, rm);
7508 i = ((op1 & 2) != 0);
7509 if (gen_set_psr(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i, tmp))
7513 rd = (insn >> 12) & 0xf;
7517 tmp = load_cpu_field(spsr);
7519 tmp = tcg_temp_new_i32();
7520 gen_helper_cpsr_read(tmp, cpu_env);
7522 store_reg(s, rd, tmp);
7527 /* branch/exchange thumb (bx). */
7529 tmp = load_reg(s, rm);
7531 } else if (op1 == 3) {
7534 rd = (insn >> 12) & 0xf;
7535 tmp = load_reg(s, rm);
7536 gen_helper_clz(tmp, tmp);
7537 store_reg(s, rd, tmp);
7545 /* Trivial implementation equivalent to bx. */
7546 tmp = load_reg(s, rm);
7557 /* branch link/exchange thumb (blx) */
7558 tmp = load_reg(s, rm);
7559 tmp2 = tcg_temp_new_i32();
7560 tcg_gen_movi_i32(tmp2, s->pc);
7561 store_reg(s, 14, tmp2);
7564 case 0x5: /* saturating add/subtract */
7566 rd = (insn >> 12) & 0xf;
7567 rn = (insn >> 16) & 0xf;
7568 tmp = load_reg(s, rm);
7569 tmp2 = load_reg(s, rn);
7571 gen_helper_double_saturate(tmp2, cpu_env, tmp2);
7573 gen_helper_sub_saturate(tmp, cpu_env, tmp, tmp2);
7575 gen_helper_add_saturate(tmp, cpu_env, tmp, tmp2);
7576 tcg_temp_free_i32(tmp2);
7577 store_reg(s, rd, tmp);
7580 /* SMC instruction (op1 == 3)
7581 and undefined instructions (op1 == 0 || op1 == 2)
7588 gen_exception_insn(s, 4, EXCP_BKPT);
7590 case 0x8: /* signed multiply */
7595 rs = (insn >> 8) & 0xf;
7596 rn = (insn >> 12) & 0xf;
7597 rd = (insn >> 16) & 0xf;
7599 /* (32 * 16) >> 16 */
7600 tmp = load_reg(s, rm);
7601 tmp2 = load_reg(s, rs);
7603 tcg_gen_sari_i32(tmp2, tmp2, 16);
7606 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7607 tcg_gen_shri_i64(tmp64, tmp64, 16);
7608 tmp = tcg_temp_new_i32();
7609 tcg_gen_trunc_i64_i32(tmp, tmp64);
7610 tcg_temp_free_i64(tmp64);
7611 if ((sh & 2) == 0) {
7612 tmp2 = load_reg(s, rn);
7613 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
7614 tcg_temp_free_i32(tmp2);
7616 store_reg(s, rd, tmp);
7619 tmp = load_reg(s, rm);
7620 tmp2 = load_reg(s, rs);
7621 gen_mulxy(tmp, tmp2, sh & 2, sh & 4);
7622 tcg_temp_free_i32(tmp2);
7624 tmp64 = tcg_temp_new_i64();
7625 tcg_gen_ext_i32_i64(tmp64, tmp);
7626 tcg_temp_free_i32(tmp);
7627 gen_addq(s, tmp64, rn, rd);
7628 gen_storeq_reg(s, rn, rd, tmp64);
7629 tcg_temp_free_i64(tmp64);
7632 tmp2 = load_reg(s, rn);
7633 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
7634 tcg_temp_free_i32(tmp2);
7636 store_reg(s, rd, tmp);
7643 } else if (((insn & 0x0e000000) == 0 &&
7644 (insn & 0x00000090) != 0x90) ||
7645 ((insn & 0x0e000000) == (1 << 25))) {
7646 int set_cc, logic_cc, shiftop;
7648 op1 = (insn >> 21) & 0xf;
7649 set_cc = (insn >> 20) & 1;
7650 logic_cc = table_logic_cc[op1] & set_cc;
7652 /* data processing instruction */
7653 if (insn & (1 << 25)) {
7654 /* immediate operand */
7656 shift = ((insn >> 8) & 0xf) * 2;
7658 val = (val >> shift) | (val << (32 - shift));
7660 tmp2 = tcg_temp_new_i32();
7661 tcg_gen_movi_i32(tmp2, val);
7662 if (logic_cc && shift) {
7663 gen_set_CF_bit31(tmp2);
7668 tmp2 = load_reg(s, rm);
7669 shiftop = (insn >> 5) & 3;
7670 if (!(insn & (1 << 4))) {
7671 shift = (insn >> 7) & 0x1f;
7672 gen_arm_shift_im(tmp2, shiftop, shift, logic_cc);
7674 rs = (insn >> 8) & 0xf;
7675 tmp = load_reg(s, rs);
7676 gen_arm_shift_reg(tmp2, shiftop, tmp, logic_cc);
7679 if (op1 != 0x0f && op1 != 0x0d) {
7680 rn = (insn >> 16) & 0xf;
7681 tmp = load_reg(s, rn);
7683 TCGV_UNUSED_I32(tmp);
7685 rd = (insn >> 12) & 0xf;
7688 tcg_gen_and_i32(tmp, tmp, tmp2);
7692 store_reg_bx(env, s, rd, tmp);
7695 tcg_gen_xor_i32(tmp, tmp, tmp2);
7699 store_reg_bx(env, s, rd, tmp);
7702 if (set_cc && rd == 15) {
7703 /* SUBS r15, ... is used for exception return. */
7707 gen_sub_CC(tmp, tmp, tmp2);
7708 gen_exception_return(s, tmp);
7711 gen_sub_CC(tmp, tmp, tmp2);
7713 tcg_gen_sub_i32(tmp, tmp, tmp2);
7715 store_reg_bx(env, s, rd, tmp);
7720 gen_sub_CC(tmp, tmp2, tmp);
7722 tcg_gen_sub_i32(tmp, tmp2, tmp);
7724 store_reg_bx(env, s, rd, tmp);
7728 gen_add_CC(tmp, tmp, tmp2);
7730 tcg_gen_add_i32(tmp, tmp, tmp2);
7732 store_reg_bx(env, s, rd, tmp);
7736 gen_adc_CC(tmp, tmp, tmp2);
7738 gen_add_carry(tmp, tmp, tmp2);
7740 store_reg_bx(env, s, rd, tmp);
7744 gen_sbc_CC(tmp, tmp, tmp2);
7746 gen_sub_carry(tmp, tmp, tmp2);
7748 store_reg_bx(env, s, rd, tmp);
7752 gen_sbc_CC(tmp, tmp2, tmp);
7754 gen_sub_carry(tmp, tmp2, tmp);
7756 store_reg_bx(env, s, rd, tmp);
7760 tcg_gen_and_i32(tmp, tmp, tmp2);
7763 tcg_temp_free_i32(tmp);
7767 tcg_gen_xor_i32(tmp, tmp, tmp2);
7770 tcg_temp_free_i32(tmp);
7774 gen_sub_CC(tmp, tmp, tmp2);
7776 tcg_temp_free_i32(tmp);
7780 gen_add_CC(tmp, tmp, tmp2);
7782 tcg_temp_free_i32(tmp);
7785 tcg_gen_or_i32(tmp, tmp, tmp2);
7789 store_reg_bx(env, s, rd, tmp);
7792 if (logic_cc && rd == 15) {
7793 /* MOVS r15, ... is used for exception return. */
7797 gen_exception_return(s, tmp2);
7802 store_reg_bx(env, s, rd, tmp2);
7806 tcg_gen_andc_i32(tmp, tmp, tmp2);
7810 store_reg_bx(env, s, rd, tmp);
7814 tcg_gen_not_i32(tmp2, tmp2);
7818 store_reg_bx(env, s, rd, tmp2);
7821 if (op1 != 0x0f && op1 != 0x0d) {
7822 tcg_temp_free_i32(tmp2);
7825 /* other instructions */
7826 op1 = (insn >> 24) & 0xf;
7830 /* multiplies, extra load/stores */
7831 sh = (insn >> 5) & 3;
7834 rd = (insn >> 16) & 0xf;
7835 rn = (insn >> 12) & 0xf;
7836 rs = (insn >> 8) & 0xf;
7838 op1 = (insn >> 20) & 0xf;
7840 case 0: case 1: case 2: case 3: case 6:
7842 tmp = load_reg(s, rs);
7843 tmp2 = load_reg(s, rm);
7844 tcg_gen_mul_i32(tmp, tmp, tmp2);
7845 tcg_temp_free_i32(tmp2);
7846 if (insn & (1 << 22)) {
7847 /* Subtract (mls) */
7849 tmp2 = load_reg(s, rn);
7850 tcg_gen_sub_i32(tmp, tmp2, tmp);
7851 tcg_temp_free_i32(tmp2);
7852 } else if (insn & (1 << 21)) {
7854 tmp2 = load_reg(s, rn);
7855 tcg_gen_add_i32(tmp, tmp, tmp2);
7856 tcg_temp_free_i32(tmp2);
7858 if (insn & (1 << 20))
7860 store_reg(s, rd, tmp);
7863 /* 64 bit mul double accumulate (UMAAL) */
7865 tmp = load_reg(s, rs);
7866 tmp2 = load_reg(s, rm);
7867 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
7868 gen_addq_lo(s, tmp64, rn);
7869 gen_addq_lo(s, tmp64, rd);
7870 gen_storeq_reg(s, rn, rd, tmp64);
7871 tcg_temp_free_i64(tmp64);
7873 case 8: case 9: case 10: case 11:
7874 case 12: case 13: case 14: case 15:
7875 /* 64 bit mul: UMULL, UMLAL, SMULL, SMLAL. */
7876 tmp = load_reg(s, rs);
7877 tmp2 = load_reg(s, rm);
7878 if (insn & (1 << 22)) {
7879 tcg_gen_muls2_i32(tmp, tmp2, tmp, tmp2);
7881 tcg_gen_mulu2_i32(tmp, tmp2, tmp, tmp2);
7883 if (insn & (1 << 21)) { /* mult accumulate */
7884 TCGv_i32 al = load_reg(s, rn);
7885 TCGv_i32 ah = load_reg(s, rd);
7886 tcg_gen_add2_i32(tmp, tmp2, tmp, tmp2, al, ah);
7887 tcg_temp_free_i32(al);
7888 tcg_temp_free_i32(ah);
7890 if (insn & (1 << 20)) {
7891 gen_logicq_cc(tmp, tmp2);
7893 store_reg(s, rn, tmp);
7894 store_reg(s, rd, tmp2);
7900 rn = (insn >> 16) & 0xf;
7901 rd = (insn >> 12) & 0xf;
7902 if (insn & (1 << 23)) {
7903 /* load/store exclusive */
7904 int op2 = (insn >> 8) & 3;
7905 op1 = (insn >> 21) & 0x3;
7908 case 0: /* lda/stl */
7914 case 1: /* reserved */
7916 case 2: /* ldaex/stlex */
7919 case 3: /* ldrex/strex */
7928 addr = tcg_temp_local_new_i32();
7929 load_reg_var(s, addr, rn);
7931 /* Since the emulation does not have barriers,
7932 the acquire/release semantics need no special
7935 if (insn & (1 << 20)) {
7936 tmp = tcg_temp_new_i32();
7939 gen_aa32_ld32u(tmp, addr, IS_USER(s));
7942 gen_aa32_ld8u(tmp, addr, IS_USER(s));
7945 gen_aa32_ld16u(tmp, addr, IS_USER(s));
7950 store_reg(s, rd, tmp);
7953 tmp = load_reg(s, rm);
7956 gen_aa32_st32(tmp, addr, IS_USER(s));
7959 gen_aa32_st8(tmp, addr, IS_USER(s));
7962 gen_aa32_st16(tmp, addr, IS_USER(s));
7967 tcg_temp_free_i32(tmp);
7969 } else if (insn & (1 << 20)) {
7972 gen_load_exclusive(s, rd, 15, addr, 2);
7974 case 1: /* ldrexd */
7975 gen_load_exclusive(s, rd, rd + 1, addr, 3);
7977 case 2: /* ldrexb */
7978 gen_load_exclusive(s, rd, 15, addr, 0);
7980 case 3: /* ldrexh */
7981 gen_load_exclusive(s, rd, 15, addr, 1);
7990 gen_store_exclusive(s, rd, rm, 15, addr, 2);
7992 case 1: /* strexd */
7993 gen_store_exclusive(s, rd, rm, rm + 1, addr, 3);
7995 case 2: /* strexb */
7996 gen_store_exclusive(s, rd, rm, 15, addr, 0);
7998 case 3: /* strexh */
7999 gen_store_exclusive(s, rd, rm, 15, addr, 1);
8005 tcg_temp_free_i32(addr);
8007 /* SWP instruction */
8010 /* ??? This is not really atomic. However we know
8011 we never have multiple CPUs running in parallel,
8012 so it is good enough. */
8013 addr = load_reg(s, rn);
8014 tmp = load_reg(s, rm);
8015 tmp2 = tcg_temp_new_i32();
8016 if (insn & (1 << 22)) {
8017 gen_aa32_ld8u(tmp2, addr, IS_USER(s));
8018 gen_aa32_st8(tmp, addr, IS_USER(s));
8020 gen_aa32_ld32u(tmp2, addr, IS_USER(s));
8021 gen_aa32_st32(tmp, addr, IS_USER(s));
8023 tcg_temp_free_i32(tmp);
8024 tcg_temp_free_i32(addr);
8025 store_reg(s, rd, tmp2);
8031 /* Misc load/store */
8032 rn = (insn >> 16) & 0xf;
8033 rd = (insn >> 12) & 0xf;
8034 addr = load_reg(s, rn);
8035 if (insn & (1 << 24))
8036 gen_add_datah_offset(s, insn, 0, addr);
8038 if (insn & (1 << 20)) {
8040 tmp = tcg_temp_new_i32();
8043 gen_aa32_ld16u(tmp, addr, IS_USER(s));
8046 gen_aa32_ld8s(tmp, addr, IS_USER(s));
8050 gen_aa32_ld16s(tmp, addr, IS_USER(s));
8054 } else if (sh & 2) {
8059 tmp = load_reg(s, rd);
8060 gen_aa32_st32(tmp, addr, IS_USER(s));
8061 tcg_temp_free_i32(tmp);
8062 tcg_gen_addi_i32(addr, addr, 4);
8063 tmp = load_reg(s, rd + 1);
8064 gen_aa32_st32(tmp, addr, IS_USER(s));
8065 tcg_temp_free_i32(tmp);
8069 tmp = tcg_temp_new_i32();
8070 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8071 store_reg(s, rd, tmp);
8072 tcg_gen_addi_i32(addr, addr, 4);
8073 tmp = tcg_temp_new_i32();
8074 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8078 address_offset = -4;
8081 tmp = load_reg(s, rd);
8082 gen_aa32_st16(tmp, addr, IS_USER(s));
8083 tcg_temp_free_i32(tmp);
8086 /* Perform base writeback before the loaded value to
8087 ensure correct behavior with overlapping index registers.
8088 ldrd with base writeback is is undefined if the
8089 destination and index registers overlap. */
8090 if (!(insn & (1 << 24))) {
8091 gen_add_datah_offset(s, insn, address_offset, addr);
8092 store_reg(s, rn, addr);
8093 } else if (insn & (1 << 21)) {
8095 tcg_gen_addi_i32(addr, addr, address_offset);
8096 store_reg(s, rn, addr);
8098 tcg_temp_free_i32(addr);
8101 /* Complete the load. */
8102 store_reg(s, rd, tmp);
8111 if (insn & (1 << 4)) {
8113 /* Armv6 Media instructions. */
8115 rn = (insn >> 16) & 0xf;
8116 rd = (insn >> 12) & 0xf;
8117 rs = (insn >> 8) & 0xf;
8118 switch ((insn >> 23) & 3) {
8119 case 0: /* Parallel add/subtract. */
8120 op1 = (insn >> 20) & 7;
8121 tmp = load_reg(s, rn);
8122 tmp2 = load_reg(s, rm);
8123 sh = (insn >> 5) & 7;
8124 if ((op1 & 3) == 0 || sh == 5 || sh == 6)
8126 gen_arm_parallel_addsub(op1, sh, tmp, tmp2);
8127 tcg_temp_free_i32(tmp2);
8128 store_reg(s, rd, tmp);
8131 if ((insn & 0x00700020) == 0) {
8132 /* Halfword pack. */
8133 tmp = load_reg(s, rn);
8134 tmp2 = load_reg(s, rm);
8135 shift = (insn >> 7) & 0x1f;
8136 if (insn & (1 << 6)) {
8140 tcg_gen_sari_i32(tmp2, tmp2, shift);
8141 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
8142 tcg_gen_ext16u_i32(tmp2, tmp2);
8146 tcg_gen_shli_i32(tmp2, tmp2, shift);
8147 tcg_gen_ext16u_i32(tmp, tmp);
8148 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
8150 tcg_gen_or_i32(tmp, tmp, tmp2);
8151 tcg_temp_free_i32(tmp2);
8152 store_reg(s, rd, tmp);
8153 } else if ((insn & 0x00200020) == 0x00200000) {
8155 tmp = load_reg(s, rm);
8156 shift = (insn >> 7) & 0x1f;
8157 if (insn & (1 << 6)) {
8160 tcg_gen_sari_i32(tmp, tmp, shift);
8162 tcg_gen_shli_i32(tmp, tmp, shift);
8164 sh = (insn >> 16) & 0x1f;
8165 tmp2 = tcg_const_i32(sh);
8166 if (insn & (1 << 22))
8167 gen_helper_usat(tmp, cpu_env, tmp, tmp2);
8169 gen_helper_ssat(tmp, cpu_env, tmp, tmp2);
8170 tcg_temp_free_i32(tmp2);
8171 store_reg(s, rd, tmp);
8172 } else if ((insn & 0x00300fe0) == 0x00200f20) {
8174 tmp = load_reg(s, rm);
8175 sh = (insn >> 16) & 0x1f;
8176 tmp2 = tcg_const_i32(sh);
8177 if (insn & (1 << 22))
8178 gen_helper_usat16(tmp, cpu_env, tmp, tmp2);
8180 gen_helper_ssat16(tmp, cpu_env, tmp, tmp2);
8181 tcg_temp_free_i32(tmp2);
8182 store_reg(s, rd, tmp);
8183 } else if ((insn & 0x00700fe0) == 0x00000fa0) {
8185 tmp = load_reg(s, rn);
8186 tmp2 = load_reg(s, rm);
8187 tmp3 = tcg_temp_new_i32();
8188 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUARMState, GE));
8189 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
8190 tcg_temp_free_i32(tmp3);
8191 tcg_temp_free_i32(tmp2);
8192 store_reg(s, rd, tmp);
8193 } else if ((insn & 0x000003e0) == 0x00000060) {
8194 tmp = load_reg(s, rm);
8195 shift = (insn >> 10) & 3;
8196 /* ??? In many cases it's not necessary to do a
8197 rotate, a shift is sufficient. */
8199 tcg_gen_rotri_i32(tmp, tmp, shift * 8);
8200 op1 = (insn >> 20) & 7;
8202 case 0: gen_sxtb16(tmp); break;
8203 case 2: gen_sxtb(tmp); break;
8204 case 3: gen_sxth(tmp); break;
8205 case 4: gen_uxtb16(tmp); break;
8206 case 6: gen_uxtb(tmp); break;
8207 case 7: gen_uxth(tmp); break;
8208 default: goto illegal_op;
8211 tmp2 = load_reg(s, rn);
8212 if ((op1 & 3) == 0) {
8213 gen_add16(tmp, tmp2);
8215 tcg_gen_add_i32(tmp, tmp, tmp2);
8216 tcg_temp_free_i32(tmp2);
8219 store_reg(s, rd, tmp);
8220 } else if ((insn & 0x003f0f60) == 0x003f0f20) {
8222 tmp = load_reg(s, rm);
8223 if (insn & (1 << 22)) {
8224 if (insn & (1 << 7)) {
8228 gen_helper_rbit(tmp, tmp);
8231 if (insn & (1 << 7))
8234 tcg_gen_bswap32_i32(tmp, tmp);
8236 store_reg(s, rd, tmp);
8241 case 2: /* Multiplies (Type 3). */
8242 switch ((insn >> 20) & 0x7) {
8244 if (((insn >> 6) ^ (insn >> 7)) & 1) {
8245 /* op2 not 00x or 11x : UNDEF */
8248 /* Signed multiply most significant [accumulate].
8249 (SMMUL, SMMLA, SMMLS) */
8250 tmp = load_reg(s, rm);
8251 tmp2 = load_reg(s, rs);
8252 tmp64 = gen_muls_i64_i32(tmp, tmp2);
8255 tmp = load_reg(s, rd);
8256 if (insn & (1 << 6)) {
8257 tmp64 = gen_subq_msw(tmp64, tmp);
8259 tmp64 = gen_addq_msw(tmp64, tmp);
8262 if (insn & (1 << 5)) {
8263 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
8265 tcg_gen_shri_i64(tmp64, tmp64, 32);
8266 tmp = tcg_temp_new_i32();
8267 tcg_gen_trunc_i64_i32(tmp, tmp64);
8268 tcg_temp_free_i64(tmp64);
8269 store_reg(s, rn, tmp);
8273 /* SMLAD, SMUAD, SMLSD, SMUSD, SMLALD, SMLSLD */
8274 if (insn & (1 << 7)) {
8277 tmp = load_reg(s, rm);
8278 tmp2 = load_reg(s, rs);
8279 if (insn & (1 << 5))
8280 gen_swap_half(tmp2);
8281 gen_smul_dual(tmp, tmp2);
8282 if (insn & (1 << 6)) {
8283 /* This subtraction cannot overflow. */
8284 tcg_gen_sub_i32(tmp, tmp, tmp2);
8286 /* This addition cannot overflow 32 bits;
8287 * however it may overflow considered as a signed
8288 * operation, in which case we must set the Q flag.
8290 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
8292 tcg_temp_free_i32(tmp2);
8293 if (insn & (1 << 22)) {
8294 /* smlald, smlsld */
8295 tmp64 = tcg_temp_new_i64();
8296 tcg_gen_ext_i32_i64(tmp64, tmp);
8297 tcg_temp_free_i32(tmp);
8298 gen_addq(s, tmp64, rd, rn);
8299 gen_storeq_reg(s, rd, rn, tmp64);
8300 tcg_temp_free_i64(tmp64);
8302 /* smuad, smusd, smlad, smlsd */
8305 tmp2 = load_reg(s, rd);
8306 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
8307 tcg_temp_free_i32(tmp2);
8309 store_reg(s, rn, tmp);
8315 if (!arm_feature(env, ARM_FEATURE_ARM_DIV)) {
8318 if (((insn >> 5) & 7) || (rd != 15)) {
8321 tmp = load_reg(s, rm);
8322 tmp2 = load_reg(s, rs);
8323 if (insn & (1 << 21)) {
8324 gen_helper_udiv(tmp, tmp, tmp2);
8326 gen_helper_sdiv(tmp, tmp, tmp2);
8328 tcg_temp_free_i32(tmp2);
8329 store_reg(s, rn, tmp);
8336 op1 = ((insn >> 17) & 0x38) | ((insn >> 5) & 7);
8338 case 0: /* Unsigned sum of absolute differences. */
8340 tmp = load_reg(s, rm);
8341 tmp2 = load_reg(s, rs);
8342 gen_helper_usad8(tmp, tmp, tmp2);
8343 tcg_temp_free_i32(tmp2);
8345 tmp2 = load_reg(s, rd);
8346 tcg_gen_add_i32(tmp, tmp, tmp2);
8347 tcg_temp_free_i32(tmp2);
8349 store_reg(s, rn, tmp);
8351 case 0x20: case 0x24: case 0x28: case 0x2c:
8352 /* Bitfield insert/clear. */
8354 shift = (insn >> 7) & 0x1f;
8355 i = (insn >> 16) & 0x1f;
8358 tmp = tcg_temp_new_i32();
8359 tcg_gen_movi_i32(tmp, 0);
8361 tmp = load_reg(s, rm);
8364 tmp2 = load_reg(s, rd);
8365 tcg_gen_deposit_i32(tmp, tmp2, tmp, shift, i);
8366 tcg_temp_free_i32(tmp2);
8368 store_reg(s, rd, tmp);
8370 case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
8371 case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
8373 tmp = load_reg(s, rm);
8374 shift = (insn >> 7) & 0x1f;
8375 i = ((insn >> 16) & 0x1f) + 1;
8380 gen_ubfx(tmp, shift, (1u << i) - 1);
8382 gen_sbfx(tmp, shift, i);
8385 store_reg(s, rd, tmp);
8395 /* Check for undefined extension instructions
8396 * per the ARM Bible IE:
8397 * xxxx 0111 1111 xxxx xxxx xxxx 1111 xxxx
8399 sh = (0xf << 20) | (0xf << 4);
8400 if (op1 == 0x7 && ((insn & sh) == sh))
8404 /* load/store byte/word */
8405 rn = (insn >> 16) & 0xf;
8406 rd = (insn >> 12) & 0xf;
8407 tmp2 = load_reg(s, rn);
8408 i = (IS_USER(s) || (insn & 0x01200000) == 0x00200000);
8409 if (insn & (1 << 24))
8410 gen_add_data_offset(s, insn, tmp2);
8411 if (insn & (1 << 20)) {
8413 tmp = tcg_temp_new_i32();
8414 if (insn & (1 << 22)) {
8415 gen_aa32_ld8u(tmp, tmp2, i);
8417 gen_aa32_ld32u(tmp, tmp2, i);
8421 tmp = load_reg(s, rd);
8422 if (insn & (1 << 22)) {
8423 gen_aa32_st8(tmp, tmp2, i);
8425 gen_aa32_st32(tmp, tmp2, i);
8427 tcg_temp_free_i32(tmp);
8429 if (!(insn & (1 << 24))) {
8430 gen_add_data_offset(s, insn, tmp2);
8431 store_reg(s, rn, tmp2);
8432 } else if (insn & (1 << 21)) {
8433 store_reg(s, rn, tmp2);
8435 tcg_temp_free_i32(tmp2);
8437 if (insn & (1 << 20)) {
8438 /* Complete the load. */
8439 store_reg_from_load(env, s, rd, tmp);
8445 int j, n, user, loaded_base;
8446 TCGv_i32 loaded_var;
8447 /* load/store multiple words */
8448 /* XXX: store correct base if write back */
8450 if (insn & (1 << 22)) {
8452 goto illegal_op; /* only usable in supervisor mode */
8454 if ((insn & (1 << 15)) == 0)
8457 rn = (insn >> 16) & 0xf;
8458 addr = load_reg(s, rn);
8460 /* compute total size */
8462 TCGV_UNUSED_I32(loaded_var);
8465 if (insn & (1 << i))
8468 /* XXX: test invalid n == 0 case ? */
8469 if (insn & (1 << 23)) {
8470 if (insn & (1 << 24)) {
8472 tcg_gen_addi_i32(addr, addr, 4);
8474 /* post increment */
8477 if (insn & (1 << 24)) {
8479 tcg_gen_addi_i32(addr, addr, -(n * 4));
8481 /* post decrement */
8483 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
8488 if (insn & (1 << i)) {
8489 if (insn & (1 << 20)) {
8491 tmp = tcg_temp_new_i32();
8492 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8494 tmp2 = tcg_const_i32(i);
8495 gen_helper_set_user_reg(cpu_env, tmp2, tmp);
8496 tcg_temp_free_i32(tmp2);
8497 tcg_temp_free_i32(tmp);
8498 } else if (i == rn) {
8502 store_reg_from_load(env, s, i, tmp);
8507 /* special case: r15 = PC + 8 */
8508 val = (long)s->pc + 4;
8509 tmp = tcg_temp_new_i32();
8510 tcg_gen_movi_i32(tmp, val);
8512 tmp = tcg_temp_new_i32();
8513 tmp2 = tcg_const_i32(i);
8514 gen_helper_get_user_reg(tmp, cpu_env, tmp2);
8515 tcg_temp_free_i32(tmp2);
8517 tmp = load_reg(s, i);
8519 gen_aa32_st32(tmp, addr, IS_USER(s));
8520 tcg_temp_free_i32(tmp);
8523 /* no need to add after the last transfer */
8525 tcg_gen_addi_i32(addr, addr, 4);
8528 if (insn & (1 << 21)) {
8530 if (insn & (1 << 23)) {
8531 if (insn & (1 << 24)) {
8534 /* post increment */
8535 tcg_gen_addi_i32(addr, addr, 4);
8538 if (insn & (1 << 24)) {
8541 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
8543 /* post decrement */
8544 tcg_gen_addi_i32(addr, addr, -(n * 4));
8547 store_reg(s, rn, addr);
8549 tcg_temp_free_i32(addr);
8552 store_reg(s, rn, loaded_var);
8554 if ((insn & (1 << 22)) && !user) {
8555 /* Restore CPSR from SPSR. */
8556 tmp = load_cpu_field(spsr);
8557 gen_set_cpsr(tmp, 0xffffffff);
8558 tcg_temp_free_i32(tmp);
8559 s->is_jmp = DISAS_UPDATE;
8568 /* branch (and link) */
8569 val = (int32_t)s->pc;
8570 if (insn & (1 << 24)) {
8571 tmp = tcg_temp_new_i32();
8572 tcg_gen_movi_i32(tmp, val);
8573 store_reg(s, 14, tmp);
8575 offset = sextract32(insn << 2, 0, 26);
8583 if (((insn >> 8) & 0xe) == 10) {
8585 if (disas_vfp_insn(env, s, insn)) {
8588 } else if (disas_coproc_insn(env, s, insn)) {
8595 gen_set_pc_im(s, s->pc);
8596 s->is_jmp = DISAS_SWI;
8600 gen_exception_insn(s, 4, EXCP_UDEF);
8606 /* Return true if this is a Thumb-2 logical op. */
8608 thumb2_logic_op(int op)
8613 /* Generate code for a Thumb-2 data processing operation. If CONDS is nonzero
8614 then set condition code flags based on the result of the operation.
8615 If SHIFTER_OUT is nonzero then set the carry flag for logical operations
8616 to the high bit of T1.
8617 Returns zero if the opcode is valid. */
8620 gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out,
8621 TCGv_i32 t0, TCGv_i32 t1)
8628 tcg_gen_and_i32(t0, t0, t1);
8632 tcg_gen_andc_i32(t0, t0, t1);
8636 tcg_gen_or_i32(t0, t0, t1);
8640 tcg_gen_orc_i32(t0, t0, t1);
8644 tcg_gen_xor_i32(t0, t0, t1);
8649 gen_add_CC(t0, t0, t1);
8651 tcg_gen_add_i32(t0, t0, t1);
8655 gen_adc_CC(t0, t0, t1);
8661 gen_sbc_CC(t0, t0, t1);
8663 gen_sub_carry(t0, t0, t1);
8668 gen_sub_CC(t0, t0, t1);
8670 tcg_gen_sub_i32(t0, t0, t1);
8674 gen_sub_CC(t0, t1, t0);
8676 tcg_gen_sub_i32(t0, t1, t0);
8678 default: /* 5, 6, 7, 9, 12, 15. */
8684 gen_set_CF_bit31(t1);
8689 /* Translate a 32-bit thumb instruction. Returns nonzero if the instruction
8691 static int disas_thumb2_insn(CPUARMState *env, DisasContext *s, uint16_t insn_hw1)
8693 uint32_t insn, imm, shift, offset;
8694 uint32_t rd, rn, rm, rs;
8705 if (!(arm_feature(env, ARM_FEATURE_THUMB2)
8706 || arm_feature (env, ARM_FEATURE_M))) {
8707 /* Thumb-1 cores may need to treat bl and blx as a pair of
8708 16-bit instructions to get correct prefetch abort behavior. */
8710 if ((insn & (1 << 12)) == 0) {
8712 /* Second half of blx. */
8713 offset = ((insn & 0x7ff) << 1);
8714 tmp = load_reg(s, 14);
8715 tcg_gen_addi_i32(tmp, tmp, offset);
8716 tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
8718 tmp2 = tcg_temp_new_i32();
8719 tcg_gen_movi_i32(tmp2, s->pc | 1);
8720 store_reg(s, 14, tmp2);
8724 if (insn & (1 << 11)) {
8725 /* Second half of bl. */
8726 offset = ((insn & 0x7ff) << 1) | 1;
8727 tmp = load_reg(s, 14);
8728 tcg_gen_addi_i32(tmp, tmp, offset);
8730 tmp2 = tcg_temp_new_i32();
8731 tcg_gen_movi_i32(tmp2, s->pc | 1);
8732 store_reg(s, 14, tmp2);
8736 if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
8737 /* Instruction spans a page boundary. Implement it as two
8738 16-bit instructions in case the second half causes an
8740 offset = ((int32_t)insn << 21) >> 9;
8741 tcg_gen_movi_i32(cpu_R[14], s->pc + 2 + offset);
8744 /* Fall through to 32-bit decode. */
8747 insn = arm_lduw_code(env, s->pc, s->bswap_code);
8749 insn |= (uint32_t)insn_hw1 << 16;
8751 if ((insn & 0xf800e800) != 0xf000e800) {
8755 rn = (insn >> 16) & 0xf;
8756 rs = (insn >> 12) & 0xf;
8757 rd = (insn >> 8) & 0xf;
8759 switch ((insn >> 25) & 0xf) {
8760 case 0: case 1: case 2: case 3:
8761 /* 16-bit instructions. Should never happen. */
8764 if (insn & (1 << 22)) {
8765 /* Other load/store, table branch. */
8766 if (insn & 0x01200000) {
8767 /* Load/store doubleword. */
8769 addr = tcg_temp_new_i32();
8770 tcg_gen_movi_i32(addr, s->pc & ~3);
8772 addr = load_reg(s, rn);
8774 offset = (insn & 0xff) * 4;
8775 if ((insn & (1 << 23)) == 0)
8777 if (insn & (1 << 24)) {
8778 tcg_gen_addi_i32(addr, addr, offset);
8781 if (insn & (1 << 20)) {
8783 tmp = tcg_temp_new_i32();
8784 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8785 store_reg(s, rs, tmp);
8786 tcg_gen_addi_i32(addr, addr, 4);
8787 tmp = tcg_temp_new_i32();
8788 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8789 store_reg(s, rd, tmp);
8792 tmp = load_reg(s, rs);
8793 gen_aa32_st32(tmp, addr, IS_USER(s));
8794 tcg_temp_free_i32(tmp);
8795 tcg_gen_addi_i32(addr, addr, 4);
8796 tmp = load_reg(s, rd);
8797 gen_aa32_st32(tmp, addr, IS_USER(s));
8798 tcg_temp_free_i32(tmp);
8800 if (insn & (1 << 21)) {
8801 /* Base writeback. */
8804 tcg_gen_addi_i32(addr, addr, offset - 4);
8805 store_reg(s, rn, addr);
8807 tcg_temp_free_i32(addr);
8809 } else if ((insn & (1 << 23)) == 0) {
8810 /* Load/store exclusive word. */
8811 addr = tcg_temp_local_new_i32();
8812 load_reg_var(s, addr, rn);
8813 tcg_gen_addi_i32(addr, addr, (insn & 0xff) << 2);
8814 if (insn & (1 << 20)) {
8815 gen_load_exclusive(s, rs, 15, addr, 2);
8817 gen_store_exclusive(s, rd, rs, 15, addr, 2);
8819 tcg_temp_free_i32(addr);
8820 } else if ((insn & (7 << 5)) == 0) {
8823 addr = tcg_temp_new_i32();
8824 tcg_gen_movi_i32(addr, s->pc);
8826 addr = load_reg(s, rn);
8828 tmp = load_reg(s, rm);
8829 tcg_gen_add_i32(addr, addr, tmp);
8830 if (insn & (1 << 4)) {
8832 tcg_gen_add_i32(addr, addr, tmp);
8833 tcg_temp_free_i32(tmp);
8834 tmp = tcg_temp_new_i32();
8835 gen_aa32_ld16u(tmp, addr, IS_USER(s));
8837 tcg_temp_free_i32(tmp);
8838 tmp = tcg_temp_new_i32();
8839 gen_aa32_ld8u(tmp, addr, IS_USER(s));
8841 tcg_temp_free_i32(addr);
8842 tcg_gen_shli_i32(tmp, tmp, 1);
8843 tcg_gen_addi_i32(tmp, tmp, s->pc);
8844 store_reg(s, 15, tmp);
8846 int op2 = (insn >> 6) & 0x3;
8847 op = (insn >> 4) & 0x3;
8852 /* Load/store exclusive byte/halfword/doubleword */
8859 /* Load-acquire/store-release */
8865 /* Load-acquire/store-release exclusive */
8869 addr = tcg_temp_local_new_i32();
8870 load_reg_var(s, addr, rn);
8872 if (insn & (1 << 20)) {
8873 tmp = tcg_temp_new_i32();
8876 gen_aa32_ld8u(tmp, addr, IS_USER(s));
8879 gen_aa32_ld16u(tmp, addr, IS_USER(s));
8882 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8887 store_reg(s, rs, tmp);
8889 tmp = load_reg(s, rs);
8892 gen_aa32_st8(tmp, addr, IS_USER(s));
8895 gen_aa32_st16(tmp, addr, IS_USER(s));
8898 gen_aa32_st32(tmp, addr, IS_USER(s));
8903 tcg_temp_free_i32(tmp);
8905 } else if (insn & (1 << 20)) {
8906 gen_load_exclusive(s, rs, rd, addr, op);
8908 gen_store_exclusive(s, rm, rs, rd, addr, op);
8910 tcg_temp_free_i32(addr);
8913 /* Load/store multiple, RFE, SRS. */
8914 if (((insn >> 23) & 1) == ((insn >> 24) & 1)) {
8915 /* RFE, SRS: not available in user mode or on M profile */
8916 if (IS_USER(s) || IS_M(env)) {
8919 if (insn & (1 << 20)) {
8921 addr = load_reg(s, rn);
8922 if ((insn & (1 << 24)) == 0)
8923 tcg_gen_addi_i32(addr, addr, -8);
8924 /* Load PC into tmp and CPSR into tmp2. */
8925 tmp = tcg_temp_new_i32();
8926 gen_aa32_ld32u(tmp, addr, 0);
8927 tcg_gen_addi_i32(addr, addr, 4);
8928 tmp2 = tcg_temp_new_i32();
8929 gen_aa32_ld32u(tmp2, addr, 0);
8930 if (insn & (1 << 21)) {
8931 /* Base writeback. */
8932 if (insn & (1 << 24)) {
8933 tcg_gen_addi_i32(addr, addr, 4);
8935 tcg_gen_addi_i32(addr, addr, -4);
8937 store_reg(s, rn, addr);
8939 tcg_temp_free_i32(addr);
8941 gen_rfe(s, tmp, tmp2);
8944 gen_srs(s, (insn & 0x1f), (insn & (1 << 24)) ? 1 : 2,
8948 int i, loaded_base = 0;
8949 TCGv_i32 loaded_var;
8950 /* Load/store multiple. */
8951 addr = load_reg(s, rn);
8953 for (i = 0; i < 16; i++) {
8954 if (insn & (1 << i))
8957 if (insn & (1 << 24)) {
8958 tcg_gen_addi_i32(addr, addr, -offset);
8961 TCGV_UNUSED_I32(loaded_var);
8962 for (i = 0; i < 16; i++) {
8963 if ((insn & (1 << i)) == 0)
8965 if (insn & (1 << 20)) {
8967 tmp = tcg_temp_new_i32();
8968 gen_aa32_ld32u(tmp, addr, IS_USER(s));
8971 } else if (i == rn) {
8975 store_reg(s, i, tmp);
8979 tmp = load_reg(s, i);
8980 gen_aa32_st32(tmp, addr, IS_USER(s));
8981 tcg_temp_free_i32(tmp);
8983 tcg_gen_addi_i32(addr, addr, 4);
8986 store_reg(s, rn, loaded_var);
8988 if (insn & (1 << 21)) {
8989 /* Base register writeback. */
8990 if (insn & (1 << 24)) {
8991 tcg_gen_addi_i32(addr, addr, -offset);
8993 /* Fault if writeback register is in register list. */
8994 if (insn & (1 << rn))
8996 store_reg(s, rn, addr);
8998 tcg_temp_free_i32(addr);
9005 op = (insn >> 21) & 0xf;
9007 /* Halfword pack. */
9008 tmp = load_reg(s, rn);
9009 tmp2 = load_reg(s, rm);
9010 shift = ((insn >> 10) & 0x1c) | ((insn >> 6) & 0x3);
9011 if (insn & (1 << 5)) {
9015 tcg_gen_sari_i32(tmp2, tmp2, shift);
9016 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
9017 tcg_gen_ext16u_i32(tmp2, tmp2);
9021 tcg_gen_shli_i32(tmp2, tmp2, shift);
9022 tcg_gen_ext16u_i32(tmp, tmp);
9023 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
9025 tcg_gen_or_i32(tmp, tmp, tmp2);
9026 tcg_temp_free_i32(tmp2);
9027 store_reg(s, rd, tmp);
9029 /* Data processing register constant shift. */
9031 tmp = tcg_temp_new_i32();
9032 tcg_gen_movi_i32(tmp, 0);
9034 tmp = load_reg(s, rn);
9036 tmp2 = load_reg(s, rm);
9038 shiftop = (insn >> 4) & 3;
9039 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
9040 conds = (insn & (1 << 20)) != 0;
9041 logic_cc = (conds && thumb2_logic_op(op));
9042 gen_arm_shift_im(tmp2, shiftop, shift, logic_cc);
9043 if (gen_thumb2_data_op(s, op, conds, 0, tmp, tmp2))
9045 tcg_temp_free_i32(tmp2);
9047 store_reg(s, rd, tmp);
9049 tcg_temp_free_i32(tmp);
9053 case 13: /* Misc data processing. */
9054 op = ((insn >> 22) & 6) | ((insn >> 7) & 1);
9055 if (op < 4 && (insn & 0xf000) != 0xf000)
9058 case 0: /* Register controlled shift. */
9059 tmp = load_reg(s, rn);
9060 tmp2 = load_reg(s, rm);
9061 if ((insn & 0x70) != 0)
9063 op = (insn >> 21) & 3;
9064 logic_cc = (insn & (1 << 20)) != 0;
9065 gen_arm_shift_reg(tmp, op, tmp2, logic_cc);
9068 store_reg_bx(env, s, rd, tmp);
9070 case 1: /* Sign/zero extend. */
9071 tmp = load_reg(s, rm);
9072 shift = (insn >> 4) & 3;
9073 /* ??? In many cases it's not necessary to do a
9074 rotate, a shift is sufficient. */
9076 tcg_gen_rotri_i32(tmp, tmp, shift * 8);
9077 op = (insn >> 20) & 7;
9079 case 0: gen_sxth(tmp); break;
9080 case 1: gen_uxth(tmp); break;
9081 case 2: gen_sxtb16(tmp); break;
9082 case 3: gen_uxtb16(tmp); break;
9083 case 4: gen_sxtb(tmp); break;
9084 case 5: gen_uxtb(tmp); break;
9085 default: goto illegal_op;
9088 tmp2 = load_reg(s, rn);
9089 if ((op >> 1) == 1) {
9090 gen_add16(tmp, tmp2);
9092 tcg_gen_add_i32(tmp, tmp, tmp2);
9093 tcg_temp_free_i32(tmp2);
9096 store_reg(s, rd, tmp);
9098 case 2: /* SIMD add/subtract. */
9099 op = (insn >> 20) & 7;
9100 shift = (insn >> 4) & 7;
9101 if ((op & 3) == 3 || (shift & 3) == 3)
9103 tmp = load_reg(s, rn);
9104 tmp2 = load_reg(s, rm);
9105 gen_thumb2_parallel_addsub(op, shift, tmp, tmp2);
9106 tcg_temp_free_i32(tmp2);
9107 store_reg(s, rd, tmp);
9109 case 3: /* Other data processing. */
9110 op = ((insn >> 17) & 0x38) | ((insn >> 4) & 7);
9112 /* Saturating add/subtract. */
9113 tmp = load_reg(s, rn);
9114 tmp2 = load_reg(s, rm);
9116 gen_helper_double_saturate(tmp, cpu_env, tmp);
9118 gen_helper_sub_saturate(tmp, cpu_env, tmp2, tmp);
9120 gen_helper_add_saturate(tmp, cpu_env, tmp, tmp2);
9121 tcg_temp_free_i32(tmp2);
9123 tmp = load_reg(s, rn);
9125 case 0x0a: /* rbit */
9126 gen_helper_rbit(tmp, tmp);
9128 case 0x08: /* rev */
9129 tcg_gen_bswap32_i32(tmp, tmp);
9131 case 0x09: /* rev16 */
9134 case 0x0b: /* revsh */
9137 case 0x10: /* sel */
9138 tmp2 = load_reg(s, rm);
9139 tmp3 = tcg_temp_new_i32();
9140 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUARMState, GE));
9141 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
9142 tcg_temp_free_i32(tmp3);
9143 tcg_temp_free_i32(tmp2);
9145 case 0x18: /* clz */
9146 gen_helper_clz(tmp, tmp);
9152 store_reg(s, rd, tmp);
9154 case 4: case 5: /* 32-bit multiply. Sum of absolute differences. */
9155 op = (insn >> 4) & 0xf;
9156 tmp = load_reg(s, rn);
9157 tmp2 = load_reg(s, rm);
9158 switch ((insn >> 20) & 7) {
9159 case 0: /* 32 x 32 -> 32 */
9160 tcg_gen_mul_i32(tmp, tmp, tmp2);
9161 tcg_temp_free_i32(tmp2);
9163 tmp2 = load_reg(s, rs);
9165 tcg_gen_sub_i32(tmp, tmp2, tmp);
9167 tcg_gen_add_i32(tmp, tmp, tmp2);
9168 tcg_temp_free_i32(tmp2);
9171 case 1: /* 16 x 16 -> 32 */
9172 gen_mulxy(tmp, tmp2, op & 2, op & 1);
9173 tcg_temp_free_i32(tmp2);
9175 tmp2 = load_reg(s, rs);
9176 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
9177 tcg_temp_free_i32(tmp2);
9180 case 2: /* Dual multiply add. */
9181 case 4: /* Dual multiply subtract. */
9183 gen_swap_half(tmp2);
9184 gen_smul_dual(tmp, tmp2);
9185 if (insn & (1 << 22)) {
9186 /* This subtraction cannot overflow. */
9187 tcg_gen_sub_i32(tmp, tmp, tmp2);
9189 /* This addition cannot overflow 32 bits;
9190 * however it may overflow considered as a signed
9191 * operation, in which case we must set the Q flag.
9193 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
9195 tcg_temp_free_i32(tmp2);
9198 tmp2 = load_reg(s, rs);
9199 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
9200 tcg_temp_free_i32(tmp2);
9203 case 3: /* 32 * 16 -> 32msb */
9205 tcg_gen_sari_i32(tmp2, tmp2, 16);
9208 tmp64 = gen_muls_i64_i32(tmp, tmp2);
9209 tcg_gen_shri_i64(tmp64, tmp64, 16);
9210 tmp = tcg_temp_new_i32();
9211 tcg_gen_trunc_i64_i32(tmp, tmp64);
9212 tcg_temp_free_i64(tmp64);
9215 tmp2 = load_reg(s, rs);
9216 gen_helper_add_setq(tmp, cpu_env, tmp, tmp2);
9217 tcg_temp_free_i32(tmp2);
9220 case 5: case 6: /* 32 * 32 -> 32msb (SMMUL, SMMLA, SMMLS) */
9221 tmp64 = gen_muls_i64_i32(tmp, tmp2);
9223 tmp = load_reg(s, rs);
9224 if (insn & (1 << 20)) {
9225 tmp64 = gen_addq_msw(tmp64, tmp);
9227 tmp64 = gen_subq_msw(tmp64, tmp);
9230 if (insn & (1 << 4)) {
9231 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
9233 tcg_gen_shri_i64(tmp64, tmp64, 32);
9234 tmp = tcg_temp_new_i32();
9235 tcg_gen_trunc_i64_i32(tmp, tmp64);
9236 tcg_temp_free_i64(tmp64);
9238 case 7: /* Unsigned sum of absolute differences. */
9239 gen_helper_usad8(tmp, tmp, tmp2);
9240 tcg_temp_free_i32(tmp2);
9242 tmp2 = load_reg(s, rs);
9243 tcg_gen_add_i32(tmp, tmp, tmp2);
9244 tcg_temp_free_i32(tmp2);
9248 store_reg(s, rd, tmp);
9250 case 6: case 7: /* 64-bit multiply, Divide. */
9251 op = ((insn >> 4) & 0xf) | ((insn >> 16) & 0x70);
9252 tmp = load_reg(s, rn);
9253 tmp2 = load_reg(s, rm);
9254 if ((op & 0x50) == 0x10) {
9256 if (!arm_feature(env, ARM_FEATURE_THUMB_DIV)) {
9260 gen_helper_udiv(tmp, tmp, tmp2);
9262 gen_helper_sdiv(tmp, tmp, tmp2);
9263 tcg_temp_free_i32(tmp2);
9264 store_reg(s, rd, tmp);
9265 } else if ((op & 0xe) == 0xc) {
9266 /* Dual multiply accumulate long. */
9268 gen_swap_half(tmp2);
9269 gen_smul_dual(tmp, tmp2);
9271 tcg_gen_sub_i32(tmp, tmp, tmp2);
9273 tcg_gen_add_i32(tmp, tmp, tmp2);
9275 tcg_temp_free_i32(tmp2);
9277 tmp64 = tcg_temp_new_i64();
9278 tcg_gen_ext_i32_i64(tmp64, tmp);
9279 tcg_temp_free_i32(tmp);
9280 gen_addq(s, tmp64, rs, rd);
9281 gen_storeq_reg(s, rs, rd, tmp64);
9282 tcg_temp_free_i64(tmp64);
9285 /* Unsigned 64-bit multiply */
9286 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
9290 gen_mulxy(tmp, tmp2, op & 2, op & 1);
9291 tcg_temp_free_i32(tmp2);
9292 tmp64 = tcg_temp_new_i64();
9293 tcg_gen_ext_i32_i64(tmp64, tmp);
9294 tcg_temp_free_i32(tmp);
9296 /* Signed 64-bit multiply */
9297 tmp64 = gen_muls_i64_i32(tmp, tmp2);
9302 gen_addq_lo(s, tmp64, rs);
9303 gen_addq_lo(s, tmp64, rd);
9304 } else if (op & 0x40) {
9305 /* 64-bit accumulate. */
9306 gen_addq(s, tmp64, rs, rd);
9308 gen_storeq_reg(s, rs, rd, tmp64);
9309 tcg_temp_free_i64(tmp64);
9314 case 6: case 7: case 14: case 15:
9316 if (((insn >> 24) & 3) == 3) {
9317 /* Translate into the equivalent ARM encoding. */
9318 insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4) | (1 << 28);
9319 if (disas_neon_data_insn(env, s, insn))
9321 } else if (((insn >> 8) & 0xe) == 10) {
9322 if (disas_vfp_insn(env, s, insn)) {
9326 if (insn & (1 << 28))
9328 if (disas_coproc_insn (env, s, insn))
9332 case 8: case 9: case 10: case 11:
9333 if (insn & (1 << 15)) {
9334 /* Branches, misc control. */
9335 if (insn & 0x5000) {
9336 /* Unconditional branch. */
9337 /* signextend(hw1[10:0]) -> offset[:12]. */
9338 offset = ((int32_t)insn << 5) >> 9 & ~(int32_t)0xfff;
9339 /* hw1[10:0] -> offset[11:1]. */
9340 offset |= (insn & 0x7ff) << 1;
9341 /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
9342 offset[24:22] already have the same value because of the
9343 sign extension above. */
9344 offset ^= ((~insn) & (1 << 13)) << 10;
9345 offset ^= ((~insn) & (1 << 11)) << 11;
9347 if (insn & (1 << 14)) {
9348 /* Branch and link. */
9349 tcg_gen_movi_i32(cpu_R[14], s->pc | 1);
9353 if (insn & (1 << 12)) {
9358 offset &= ~(uint32_t)2;
9359 /* thumb2 bx, no need to check */
9360 gen_bx_im(s, offset);
9362 } else if (((insn >> 23) & 7) == 7) {
9364 if (insn & (1 << 13))
9367 if (insn & (1 << 26)) {
9368 /* Secure monitor call (v6Z) */
9369 qemu_log_mask(LOG_UNIMP,
9370 "arm: unimplemented secure monitor call\n");
9371 goto illegal_op; /* not implemented. */
9373 op = (insn >> 20) & 7;
9375 case 0: /* msr cpsr. */
9377 tmp = load_reg(s, rn);
9378 addr = tcg_const_i32(insn & 0xff);
9379 gen_helper_v7m_msr(cpu_env, addr, tmp);
9380 tcg_temp_free_i32(addr);
9381 tcg_temp_free_i32(tmp);
9386 case 1: /* msr spsr. */
9389 tmp = load_reg(s, rn);
9391 msr_mask(env, s, (insn >> 8) & 0xf, op == 1),
9395 case 2: /* cps, nop-hint. */
9396 if (((insn >> 8) & 7) == 0) {
9397 gen_nop_hint(s, insn & 0xff);
9399 /* Implemented as NOP in user mode. */
9404 if (insn & (1 << 10)) {
9405 if (insn & (1 << 7))
9407 if (insn & (1 << 6))
9409 if (insn & (1 << 5))
9411 if (insn & (1 << 9))
9412 imm = CPSR_A | CPSR_I | CPSR_F;
9414 if (insn & (1 << 8)) {
9416 imm |= (insn & 0x1f);
9419 gen_set_psr_im(s, offset, 0, imm);
9422 case 3: /* Special control operations. */
9424 op = (insn >> 4) & 0xf;
9432 /* These execute as NOPs. */
9439 /* Trivial implementation equivalent to bx. */
9440 tmp = load_reg(s, rn);
9443 case 5: /* Exception return. */
9447 if (rn != 14 || rd != 15) {
9450 tmp = load_reg(s, rn);
9451 tcg_gen_subi_i32(tmp, tmp, insn & 0xff);
9452 gen_exception_return(s, tmp);
9454 case 6: /* mrs cpsr. */
9455 tmp = tcg_temp_new_i32();
9457 addr = tcg_const_i32(insn & 0xff);
9458 gen_helper_v7m_mrs(tmp, cpu_env, addr);
9459 tcg_temp_free_i32(addr);
9461 gen_helper_cpsr_read(tmp, cpu_env);
9463 store_reg(s, rd, tmp);
9465 case 7: /* mrs spsr. */
9466 /* Not accessible in user mode. */
9467 if (IS_USER(s) || IS_M(env))
9469 tmp = load_cpu_field(spsr);
9470 store_reg(s, rd, tmp);
9475 /* Conditional branch. */
9476 op = (insn >> 22) & 0xf;
9477 /* Generate a conditional jump to next instruction. */
9478 s->condlabel = gen_new_label();
9479 arm_gen_test_cc(op ^ 1, s->condlabel);
9482 /* offset[11:1] = insn[10:0] */
9483 offset = (insn & 0x7ff) << 1;
9484 /* offset[17:12] = insn[21:16]. */
9485 offset |= (insn & 0x003f0000) >> 4;
9486 /* offset[31:20] = insn[26]. */
9487 offset |= ((int32_t)((insn << 5) & 0x80000000)) >> 11;
9488 /* offset[18] = insn[13]. */
9489 offset |= (insn & (1 << 13)) << 5;
9490 /* offset[19] = insn[11]. */
9491 offset |= (insn & (1 << 11)) << 8;
9493 /* jump to the offset */
9494 gen_jmp(s, s->pc + offset);
9497 /* Data processing immediate. */
9498 if (insn & (1 << 25)) {
9499 if (insn & (1 << 24)) {
9500 if (insn & (1 << 20))
9502 /* Bitfield/Saturate. */
9503 op = (insn >> 21) & 7;
9505 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
9507 tmp = tcg_temp_new_i32();
9508 tcg_gen_movi_i32(tmp, 0);
9510 tmp = load_reg(s, rn);
9513 case 2: /* Signed bitfield extract. */
9515 if (shift + imm > 32)
9518 gen_sbfx(tmp, shift, imm);
9520 case 6: /* Unsigned bitfield extract. */
9522 if (shift + imm > 32)
9525 gen_ubfx(tmp, shift, (1u << imm) - 1);
9527 case 3: /* Bitfield insert/clear. */
9530 imm = imm + 1 - shift;
9532 tmp2 = load_reg(s, rd);
9533 tcg_gen_deposit_i32(tmp, tmp2, tmp, shift, imm);
9534 tcg_temp_free_i32(tmp2);
9539 default: /* Saturate. */
9542 tcg_gen_sari_i32(tmp, tmp, shift);
9544 tcg_gen_shli_i32(tmp, tmp, shift);
9546 tmp2 = tcg_const_i32(imm);
9549 if ((op & 1) && shift == 0)
9550 gen_helper_usat16(tmp, cpu_env, tmp, tmp2);
9552 gen_helper_usat(tmp, cpu_env, tmp, tmp2);
9555 if ((op & 1) && shift == 0)
9556 gen_helper_ssat16(tmp, cpu_env, tmp, tmp2);
9558 gen_helper_ssat(tmp, cpu_env, tmp, tmp2);
9560 tcg_temp_free_i32(tmp2);
9563 store_reg(s, rd, tmp);
9565 imm = ((insn & 0x04000000) >> 15)
9566 | ((insn & 0x7000) >> 4) | (insn & 0xff);
9567 if (insn & (1 << 22)) {
9568 /* 16-bit immediate. */
9569 imm |= (insn >> 4) & 0xf000;
9570 if (insn & (1 << 23)) {
9572 tmp = load_reg(s, rd);
9573 tcg_gen_ext16u_i32(tmp, tmp);
9574 tcg_gen_ori_i32(tmp, tmp, imm << 16);
9577 tmp = tcg_temp_new_i32();
9578 tcg_gen_movi_i32(tmp, imm);
9581 /* Add/sub 12-bit immediate. */
9583 offset = s->pc & ~(uint32_t)3;
9584 if (insn & (1 << 23))
9588 tmp = tcg_temp_new_i32();
9589 tcg_gen_movi_i32(tmp, offset);
9591 tmp = load_reg(s, rn);
9592 if (insn & (1 << 23))
9593 tcg_gen_subi_i32(tmp, tmp, imm);
9595 tcg_gen_addi_i32(tmp, tmp, imm);
9598 store_reg(s, rd, tmp);
9601 int shifter_out = 0;
9602 /* modified 12-bit immediate. */
9603 shift = ((insn & 0x04000000) >> 23) | ((insn & 0x7000) >> 12);
9604 imm = (insn & 0xff);
9607 /* Nothing to do. */
9609 case 1: /* 00XY00XY */
9612 case 2: /* XY00XY00 */
9616 case 3: /* XYXYXYXY */
9620 default: /* Rotated constant. */
9621 shift = (shift << 1) | (imm >> 7);
9623 imm = imm << (32 - shift);
9627 tmp2 = tcg_temp_new_i32();
9628 tcg_gen_movi_i32(tmp2, imm);
9629 rn = (insn >> 16) & 0xf;
9631 tmp = tcg_temp_new_i32();
9632 tcg_gen_movi_i32(tmp, 0);
9634 tmp = load_reg(s, rn);
9636 op = (insn >> 21) & 0xf;
9637 if (gen_thumb2_data_op(s, op, (insn & (1 << 20)) != 0,
9638 shifter_out, tmp, tmp2))
9640 tcg_temp_free_i32(tmp2);
9641 rd = (insn >> 8) & 0xf;
9643 store_reg(s, rd, tmp);
9645 tcg_temp_free_i32(tmp);
9650 case 12: /* Load/store single data item. */
9655 if ((insn & 0x01100000) == 0x01000000) {
9656 if (disas_neon_ls_insn(env, s, insn))
9660 op = ((insn >> 21) & 3) | ((insn >> 22) & 4);
9662 if (!(insn & (1 << 20))) {
9666 /* Byte or halfword load space with dest == r15 : memory hints.
9667 * Catch them early so we don't emit pointless addressing code.
9668 * This space is a mix of:
9669 * PLD/PLDW/PLI, which we implement as NOPs (note that unlike
9670 * the ARM encodings, PLDW space doesn't UNDEF for non-v7MP
9672 * unallocated hints, which must be treated as NOPs
9673 * UNPREDICTABLE space, which we NOP or UNDEF depending on
9674 * which is easiest for the decoding logic
9675 * Some space which must UNDEF
9677 int op1 = (insn >> 23) & 3;
9678 int op2 = (insn >> 6) & 0x3f;
9683 /* UNPREDICTABLE, unallocated hint or
9684 * PLD/PLDW/PLI (literal)
9689 return 0; /* PLD/PLDW/PLI or unallocated hint */
9691 if ((op2 == 0) || ((op2 & 0x3c) == 0x30)) {
9692 return 0; /* PLD/PLDW/PLI or unallocated hint */
9694 /* UNDEF space, or an UNPREDICTABLE */
9700 addr = tcg_temp_new_i32();
9702 /* s->pc has already been incremented by 4. */
9703 imm = s->pc & 0xfffffffc;
9704 if (insn & (1 << 23))
9705 imm += insn & 0xfff;
9707 imm -= insn & 0xfff;
9708 tcg_gen_movi_i32(addr, imm);
9710 addr = load_reg(s, rn);
9711 if (insn & (1 << 23)) {
9712 /* Positive offset. */
9714 tcg_gen_addi_i32(addr, addr, imm);
9717 switch ((insn >> 8) & 0xf) {
9718 case 0x0: /* Shifted Register. */
9719 shift = (insn >> 4) & 0xf;
9721 tcg_temp_free_i32(addr);
9724 tmp = load_reg(s, rm);
9726 tcg_gen_shli_i32(tmp, tmp, shift);
9727 tcg_gen_add_i32(addr, addr, tmp);
9728 tcg_temp_free_i32(tmp);
9730 case 0xc: /* Negative offset. */
9731 tcg_gen_addi_i32(addr, addr, -imm);
9733 case 0xe: /* User privilege. */
9734 tcg_gen_addi_i32(addr, addr, imm);
9737 case 0x9: /* Post-decrement. */
9740 case 0xb: /* Post-increment. */
9744 case 0xd: /* Pre-decrement. */
9747 case 0xf: /* Pre-increment. */
9748 tcg_gen_addi_i32(addr, addr, imm);
9752 tcg_temp_free_i32(addr);
9757 if (insn & (1 << 20)) {
9759 tmp = tcg_temp_new_i32();
9762 gen_aa32_ld8u(tmp, addr, user);
9765 gen_aa32_ld8s(tmp, addr, user);
9768 gen_aa32_ld16u(tmp, addr, user);
9771 gen_aa32_ld16s(tmp, addr, user);
9774 gen_aa32_ld32u(tmp, addr, user);
9777 tcg_temp_free_i32(tmp);
9778 tcg_temp_free_i32(addr);
9784 store_reg(s, rs, tmp);
9788 tmp = load_reg(s, rs);
9791 gen_aa32_st8(tmp, addr, user);
9794 gen_aa32_st16(tmp, addr, user);
9797 gen_aa32_st32(tmp, addr, user);
9800 tcg_temp_free_i32(tmp);
9801 tcg_temp_free_i32(addr);
9804 tcg_temp_free_i32(tmp);
9807 tcg_gen_addi_i32(addr, addr, imm);
9809 store_reg(s, rn, addr);
9811 tcg_temp_free_i32(addr);
9823 static void disas_thumb_insn(CPUARMState *env, DisasContext *s)
9825 uint32_t val, insn, op, rm, rn, rd, shift, cond;
9832 if (s->condexec_mask) {
9833 cond = s->condexec_cond;
9834 if (cond != 0x0e) { /* Skip conditional when condition is AL. */
9835 s->condlabel = gen_new_label();
9836 arm_gen_test_cc(cond ^ 1, s->condlabel);
9841 insn = arm_lduw_code(env, s->pc, s->bswap_code);
9844 switch (insn >> 12) {
9848 op = (insn >> 11) & 3;
9851 rn = (insn >> 3) & 7;
9852 tmp = load_reg(s, rn);
9853 if (insn & (1 << 10)) {
9855 tmp2 = tcg_temp_new_i32();
9856 tcg_gen_movi_i32(tmp2, (insn >> 6) & 7);
9859 rm = (insn >> 6) & 7;
9860 tmp2 = load_reg(s, rm);
9862 if (insn & (1 << 9)) {
9863 if (s->condexec_mask)
9864 tcg_gen_sub_i32(tmp, tmp, tmp2);
9866 gen_sub_CC(tmp, tmp, tmp2);
9868 if (s->condexec_mask)
9869 tcg_gen_add_i32(tmp, tmp, tmp2);
9871 gen_add_CC(tmp, tmp, tmp2);
9873 tcg_temp_free_i32(tmp2);
9874 store_reg(s, rd, tmp);
9876 /* shift immediate */
9877 rm = (insn >> 3) & 7;
9878 shift = (insn >> 6) & 0x1f;
9879 tmp = load_reg(s, rm);
9880 gen_arm_shift_im(tmp, op, shift, s->condexec_mask == 0);
9881 if (!s->condexec_mask)
9883 store_reg(s, rd, tmp);
9887 /* arithmetic large immediate */
9888 op = (insn >> 11) & 3;
9889 rd = (insn >> 8) & 0x7;
9890 if (op == 0) { /* mov */
9891 tmp = tcg_temp_new_i32();
9892 tcg_gen_movi_i32(tmp, insn & 0xff);
9893 if (!s->condexec_mask)
9895 store_reg(s, rd, tmp);
9897 tmp = load_reg(s, rd);
9898 tmp2 = tcg_temp_new_i32();
9899 tcg_gen_movi_i32(tmp2, insn & 0xff);
9902 gen_sub_CC(tmp, tmp, tmp2);
9903 tcg_temp_free_i32(tmp);
9904 tcg_temp_free_i32(tmp2);
9907 if (s->condexec_mask)
9908 tcg_gen_add_i32(tmp, tmp, tmp2);
9910 gen_add_CC(tmp, tmp, tmp2);
9911 tcg_temp_free_i32(tmp2);
9912 store_reg(s, rd, tmp);
9915 if (s->condexec_mask)
9916 tcg_gen_sub_i32(tmp, tmp, tmp2);
9918 gen_sub_CC(tmp, tmp, tmp2);
9919 tcg_temp_free_i32(tmp2);
9920 store_reg(s, rd, tmp);
9926 if (insn & (1 << 11)) {
9927 rd = (insn >> 8) & 7;
9928 /* load pc-relative. Bit 1 of PC is ignored. */
9929 val = s->pc + 2 + ((insn & 0xff) * 4);
9930 val &= ~(uint32_t)2;
9931 addr = tcg_temp_new_i32();
9932 tcg_gen_movi_i32(addr, val);
9933 tmp = tcg_temp_new_i32();
9934 gen_aa32_ld32u(tmp, addr, IS_USER(s));
9935 tcg_temp_free_i32(addr);
9936 store_reg(s, rd, tmp);
9939 if (insn & (1 << 10)) {
9940 /* data processing extended or blx */
9941 rd = (insn & 7) | ((insn >> 4) & 8);
9942 rm = (insn >> 3) & 0xf;
9943 op = (insn >> 8) & 3;
9946 tmp = load_reg(s, rd);
9947 tmp2 = load_reg(s, rm);
9948 tcg_gen_add_i32(tmp, tmp, tmp2);
9949 tcg_temp_free_i32(tmp2);
9950 store_reg(s, rd, tmp);
9953 tmp = load_reg(s, rd);
9954 tmp2 = load_reg(s, rm);
9955 gen_sub_CC(tmp, tmp, tmp2);
9956 tcg_temp_free_i32(tmp2);
9957 tcg_temp_free_i32(tmp);
9959 case 2: /* mov/cpy */
9960 tmp = load_reg(s, rm);
9961 store_reg(s, rd, tmp);
9963 case 3:/* branch [and link] exchange thumb register */
9964 tmp = load_reg(s, rm);
9965 if (insn & (1 << 7)) {
9967 val = (uint32_t)s->pc | 1;
9968 tmp2 = tcg_temp_new_i32();
9969 tcg_gen_movi_i32(tmp2, val);
9970 store_reg(s, 14, tmp2);
9972 /* already thumb, no need to check */
9979 /* data processing register */
9981 rm = (insn >> 3) & 7;
9982 op = (insn >> 6) & 0xf;
9983 if (op == 2 || op == 3 || op == 4 || op == 7) {
9984 /* the shift/rotate ops want the operands backwards */
9993 if (op == 9) { /* neg */
9994 tmp = tcg_temp_new_i32();
9995 tcg_gen_movi_i32(tmp, 0);
9996 } else if (op != 0xf) { /* mvn doesn't read its first operand */
9997 tmp = load_reg(s, rd);
9999 TCGV_UNUSED_I32(tmp);
10002 tmp2 = load_reg(s, rm);
10004 case 0x0: /* and */
10005 tcg_gen_and_i32(tmp, tmp, tmp2);
10006 if (!s->condexec_mask)
10009 case 0x1: /* eor */
10010 tcg_gen_xor_i32(tmp, tmp, tmp2);
10011 if (!s->condexec_mask)
10014 case 0x2: /* lsl */
10015 if (s->condexec_mask) {
10016 gen_shl(tmp2, tmp2, tmp);
10018 gen_helper_shl_cc(tmp2, cpu_env, tmp2, tmp);
10019 gen_logic_CC(tmp2);
10022 case 0x3: /* lsr */
10023 if (s->condexec_mask) {
10024 gen_shr(tmp2, tmp2, tmp);
10026 gen_helper_shr_cc(tmp2, cpu_env, tmp2, tmp);
10027 gen_logic_CC(tmp2);
10030 case 0x4: /* asr */
10031 if (s->condexec_mask) {
10032 gen_sar(tmp2, tmp2, tmp);
10034 gen_helper_sar_cc(tmp2, cpu_env, tmp2, tmp);
10035 gen_logic_CC(tmp2);
10038 case 0x5: /* adc */
10039 if (s->condexec_mask) {
10040 gen_adc(tmp, tmp2);
10042 gen_adc_CC(tmp, tmp, tmp2);
10045 case 0x6: /* sbc */
10046 if (s->condexec_mask) {
10047 gen_sub_carry(tmp, tmp, tmp2);
10049 gen_sbc_CC(tmp, tmp, tmp2);
10052 case 0x7: /* ror */
10053 if (s->condexec_mask) {
10054 tcg_gen_andi_i32(tmp, tmp, 0x1f);
10055 tcg_gen_rotr_i32(tmp2, tmp2, tmp);
10057 gen_helper_ror_cc(tmp2, cpu_env, tmp2, tmp);
10058 gen_logic_CC(tmp2);
10061 case 0x8: /* tst */
10062 tcg_gen_and_i32(tmp, tmp, tmp2);
10066 case 0x9: /* neg */
10067 if (s->condexec_mask)
10068 tcg_gen_neg_i32(tmp, tmp2);
10070 gen_sub_CC(tmp, tmp, tmp2);
10072 case 0xa: /* cmp */
10073 gen_sub_CC(tmp, tmp, tmp2);
10076 case 0xb: /* cmn */
10077 gen_add_CC(tmp, tmp, tmp2);
10080 case 0xc: /* orr */
10081 tcg_gen_or_i32(tmp, tmp, tmp2);
10082 if (!s->condexec_mask)
10085 case 0xd: /* mul */
10086 tcg_gen_mul_i32(tmp, tmp, tmp2);
10087 if (!s->condexec_mask)
10090 case 0xe: /* bic */
10091 tcg_gen_andc_i32(tmp, tmp, tmp2);
10092 if (!s->condexec_mask)
10095 case 0xf: /* mvn */
10096 tcg_gen_not_i32(tmp2, tmp2);
10097 if (!s->condexec_mask)
10098 gen_logic_CC(tmp2);
10105 store_reg(s, rm, tmp2);
10107 tcg_temp_free_i32(tmp);
10109 store_reg(s, rd, tmp);
10110 tcg_temp_free_i32(tmp2);
10113 tcg_temp_free_i32(tmp);
10114 tcg_temp_free_i32(tmp2);
10119 /* load/store register offset. */
10121 rn = (insn >> 3) & 7;
10122 rm = (insn >> 6) & 7;
10123 op = (insn >> 9) & 7;
10124 addr = load_reg(s, rn);
10125 tmp = load_reg(s, rm);
10126 tcg_gen_add_i32(addr, addr, tmp);
10127 tcg_temp_free_i32(tmp);
10129 if (op < 3) { /* store */
10130 tmp = load_reg(s, rd);
10132 tmp = tcg_temp_new_i32();
10137 gen_aa32_st32(tmp, addr, IS_USER(s));
10140 gen_aa32_st16(tmp, addr, IS_USER(s));
10143 gen_aa32_st8(tmp, addr, IS_USER(s));
10145 case 3: /* ldrsb */
10146 gen_aa32_ld8s(tmp, addr, IS_USER(s));
10149 gen_aa32_ld32u(tmp, addr, IS_USER(s));
10152 gen_aa32_ld16u(tmp, addr, IS_USER(s));
10155 gen_aa32_ld8u(tmp, addr, IS_USER(s));
10157 case 7: /* ldrsh */
10158 gen_aa32_ld16s(tmp, addr, IS_USER(s));
10161 if (op >= 3) { /* load */
10162 store_reg(s, rd, tmp);
10164 tcg_temp_free_i32(tmp);
10166 tcg_temp_free_i32(addr);
10170 /* load/store word immediate offset */
10172 rn = (insn >> 3) & 7;
10173 addr = load_reg(s, rn);
10174 val = (insn >> 4) & 0x7c;
10175 tcg_gen_addi_i32(addr, addr, val);
10177 if (insn & (1 << 11)) {
10179 tmp = tcg_temp_new_i32();
10180 gen_aa32_ld32u(tmp, addr, IS_USER(s));
10181 store_reg(s, rd, tmp);
10184 tmp = load_reg(s, rd);
10185 gen_aa32_st32(tmp, addr, IS_USER(s));
10186 tcg_temp_free_i32(tmp);
10188 tcg_temp_free_i32(addr);
10192 /* load/store byte immediate offset */
10194 rn = (insn >> 3) & 7;
10195 addr = load_reg(s, rn);
10196 val = (insn >> 6) & 0x1f;
10197 tcg_gen_addi_i32(addr, addr, val);
10199 if (insn & (1 << 11)) {
10201 tmp = tcg_temp_new_i32();
10202 gen_aa32_ld8u(tmp, addr, IS_USER(s));
10203 store_reg(s, rd, tmp);
10206 tmp = load_reg(s, rd);
10207 gen_aa32_st8(tmp, addr, IS_USER(s));
10208 tcg_temp_free_i32(tmp);
10210 tcg_temp_free_i32(addr);
10214 /* load/store halfword immediate offset */
10216 rn = (insn >> 3) & 7;
10217 addr = load_reg(s, rn);
10218 val = (insn >> 5) & 0x3e;
10219 tcg_gen_addi_i32(addr, addr, val);
10221 if (insn & (1 << 11)) {
10223 tmp = tcg_temp_new_i32();
10224 gen_aa32_ld16u(tmp, addr, IS_USER(s));
10225 store_reg(s, rd, tmp);
10228 tmp = load_reg(s, rd);
10229 gen_aa32_st16(tmp, addr, IS_USER(s));
10230 tcg_temp_free_i32(tmp);
10232 tcg_temp_free_i32(addr);
10236 /* load/store from stack */
10237 rd = (insn >> 8) & 7;
10238 addr = load_reg(s, 13);
10239 val = (insn & 0xff) * 4;
10240 tcg_gen_addi_i32(addr, addr, val);
10242 if (insn & (1 << 11)) {
10244 tmp = tcg_temp_new_i32();
10245 gen_aa32_ld32u(tmp, addr, IS_USER(s));
10246 store_reg(s, rd, tmp);
10249 tmp = load_reg(s, rd);
10250 gen_aa32_st32(tmp, addr, IS_USER(s));
10251 tcg_temp_free_i32(tmp);
10253 tcg_temp_free_i32(addr);
10257 /* add to high reg */
10258 rd = (insn >> 8) & 7;
10259 if (insn & (1 << 11)) {
10261 tmp = load_reg(s, 13);
10263 /* PC. bit 1 is ignored. */
10264 tmp = tcg_temp_new_i32();
10265 tcg_gen_movi_i32(tmp, (s->pc + 2) & ~(uint32_t)2);
10267 val = (insn & 0xff) * 4;
10268 tcg_gen_addi_i32(tmp, tmp, val);
10269 store_reg(s, rd, tmp);
10274 op = (insn >> 8) & 0xf;
10277 /* adjust stack pointer */
10278 tmp = load_reg(s, 13);
10279 val = (insn & 0x7f) * 4;
10280 if (insn & (1 << 7))
10281 val = -(int32_t)val;
10282 tcg_gen_addi_i32(tmp, tmp, val);
10283 store_reg(s, 13, tmp);
10286 case 2: /* sign/zero extend. */
10289 rm = (insn >> 3) & 7;
10290 tmp = load_reg(s, rm);
10291 switch ((insn >> 6) & 3) {
10292 case 0: gen_sxth(tmp); break;
10293 case 1: gen_sxtb(tmp); break;
10294 case 2: gen_uxth(tmp); break;
10295 case 3: gen_uxtb(tmp); break;
10297 store_reg(s, rd, tmp);
10299 case 4: case 5: case 0xc: case 0xd:
10301 addr = load_reg(s, 13);
10302 if (insn & (1 << 8))
10306 for (i = 0; i < 8; i++) {
10307 if (insn & (1 << i))
10310 if ((insn & (1 << 11)) == 0) {
10311 tcg_gen_addi_i32(addr, addr, -offset);
10313 for (i = 0; i < 8; i++) {
10314 if (insn & (1 << i)) {
10315 if (insn & (1 << 11)) {
10317 tmp = tcg_temp_new_i32();
10318 gen_aa32_ld32u(tmp, addr, IS_USER(s));
10319 store_reg(s, i, tmp);
10322 tmp = load_reg(s, i);
10323 gen_aa32_st32(tmp, addr, IS_USER(s));
10324 tcg_temp_free_i32(tmp);
10326 /* advance to the next address. */
10327 tcg_gen_addi_i32(addr, addr, 4);
10330 TCGV_UNUSED_I32(tmp);
10331 if (insn & (1 << 8)) {
10332 if (insn & (1 << 11)) {
10334 tmp = tcg_temp_new_i32();
10335 gen_aa32_ld32u(tmp, addr, IS_USER(s));
10336 /* don't set the pc until the rest of the instruction
10340 tmp = load_reg(s, 14);
10341 gen_aa32_st32(tmp, addr, IS_USER(s));
10342 tcg_temp_free_i32(tmp);
10344 tcg_gen_addi_i32(addr, addr, 4);
10346 if ((insn & (1 << 11)) == 0) {
10347 tcg_gen_addi_i32(addr, addr, -offset);
10349 /* write back the new stack pointer */
10350 store_reg(s, 13, addr);
10351 /* set the new PC value */
10352 if ((insn & 0x0900) == 0x0900) {
10353 store_reg_from_load(env, s, 15, tmp);
10357 case 1: case 3: case 9: case 11: /* czb */
10359 tmp = load_reg(s, rm);
10360 s->condlabel = gen_new_label();
10362 if (insn & (1 << 11))
10363 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, s->condlabel);
10365 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, s->condlabel);
10366 tcg_temp_free_i32(tmp);
10367 offset = ((insn & 0xf8) >> 2) | (insn & 0x200) >> 3;
10368 val = (uint32_t)s->pc + 2;
10373 case 15: /* IT, nop-hint. */
10374 if ((insn & 0xf) == 0) {
10375 gen_nop_hint(s, (insn >> 4) & 0xf);
10379 s->condexec_cond = (insn >> 4) & 0xe;
10380 s->condexec_mask = insn & 0x1f;
10381 /* No actual code generated for this insn, just setup state. */
10384 case 0xe: /* bkpt */
10386 gen_exception_insn(s, 2, EXCP_BKPT);
10389 case 0xa: /* rev */
10391 rn = (insn >> 3) & 0x7;
10393 tmp = load_reg(s, rn);
10394 switch ((insn >> 6) & 3) {
10395 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
10396 case 1: gen_rev16(tmp); break;
10397 case 3: gen_revsh(tmp); break;
10398 default: goto illegal_op;
10400 store_reg(s, rd, tmp);
10404 switch ((insn >> 5) & 7) {
10408 if (((insn >> 3) & 1) != s->bswap_code) {
10409 /* Dynamic endianness switching not implemented. */
10410 qemu_log_mask(LOG_UNIMP, "arm: unimplemented setend\n");
10421 tmp = tcg_const_i32((insn & (1 << 4)) != 0);
10424 addr = tcg_const_i32(19);
10425 gen_helper_v7m_msr(cpu_env, addr, tmp);
10426 tcg_temp_free_i32(addr);
10430 addr = tcg_const_i32(16);
10431 gen_helper_v7m_msr(cpu_env, addr, tmp);
10432 tcg_temp_free_i32(addr);
10434 tcg_temp_free_i32(tmp);
10437 if (insn & (1 << 4)) {
10438 shift = CPSR_A | CPSR_I | CPSR_F;
10442 gen_set_psr_im(s, ((insn & 7) << 6), 0, shift);
10457 /* load/store multiple */
10458 TCGv_i32 loaded_var;
10459 TCGV_UNUSED_I32(loaded_var);
10460 rn = (insn >> 8) & 0x7;
10461 addr = load_reg(s, rn);
10462 for (i = 0; i < 8; i++) {
10463 if (insn & (1 << i)) {
10464 if (insn & (1 << 11)) {
10466 tmp = tcg_temp_new_i32();
10467 gen_aa32_ld32u(tmp, addr, IS_USER(s));
10471 store_reg(s, i, tmp);
10475 tmp = load_reg(s, i);
10476 gen_aa32_st32(tmp, addr, IS_USER(s));
10477 tcg_temp_free_i32(tmp);
10479 /* advance to the next address */
10480 tcg_gen_addi_i32(addr, addr, 4);
10483 if ((insn & (1 << rn)) == 0) {
10484 /* base reg not in list: base register writeback */
10485 store_reg(s, rn, addr);
10487 /* base reg in list: if load, complete it now */
10488 if (insn & (1 << 11)) {
10489 store_reg(s, rn, loaded_var);
10491 tcg_temp_free_i32(addr);
10496 /* conditional branch or swi */
10497 cond = (insn >> 8) & 0xf;
10503 gen_set_pc_im(s, s->pc);
10504 s->is_jmp = DISAS_SWI;
10507 /* generate a conditional jump to next instruction */
10508 s->condlabel = gen_new_label();
10509 arm_gen_test_cc(cond ^ 1, s->condlabel);
10512 /* jump to the offset */
10513 val = (uint32_t)s->pc + 2;
10514 offset = ((int32_t)insn << 24) >> 24;
10515 val += offset << 1;
10520 if (insn & (1 << 11)) {
10521 if (disas_thumb2_insn(env, s, insn))
10525 /* unconditional branch */
10526 val = (uint32_t)s->pc;
10527 offset = ((int32_t)insn << 21) >> 21;
10528 val += (offset << 1) + 2;
10533 if (disas_thumb2_insn(env, s, insn))
10539 gen_exception_insn(s, 4, EXCP_UDEF);
10543 gen_exception_insn(s, 2, EXCP_UDEF);
10546 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
10547 basic block 'tb'. If search_pc is TRUE, also generate PC
10548 information for each intermediate instruction. */
10549 static inline void gen_intermediate_code_internal(ARMCPU *cpu,
10550 TranslationBlock *tb,
10553 CPUState *cs = CPU(cpu);
10554 CPUARMState *env = &cpu->env;
10555 DisasContext dc1, *dc = &dc1;
10557 uint16_t *gen_opc_end;
10559 target_ulong pc_start;
10560 target_ulong next_page_start;
10564 /* generate intermediate code */
10566 /* The A64 decoder has its own top level loop, because it doesn't need
10567 * the A32/T32 complexity to do with conditional execution/IT blocks/etc.
10569 if (ARM_TBFLAG_AARCH64_STATE(tb->flags)) {
10570 gen_intermediate_code_internal_a64(cpu, tb, search_pc);
10578 gen_opc_end = tcg_ctx.gen_opc_buf + OPC_MAX_SIZE;
10580 dc->is_jmp = DISAS_NEXT;
10582 dc->singlestep_enabled = cs->singlestep_enabled;
10586 dc->thumb = ARM_TBFLAG_THUMB(tb->flags);
10587 dc->bswap_code = ARM_TBFLAG_BSWAP_CODE(tb->flags);
10588 dc->condexec_mask = (ARM_TBFLAG_CONDEXEC(tb->flags) & 0xf) << 1;
10589 dc->condexec_cond = ARM_TBFLAG_CONDEXEC(tb->flags) >> 4;
10590 #if !defined(CONFIG_USER_ONLY)
10591 dc->user = (ARM_TBFLAG_PRIV(tb->flags) == 0);
10593 dc->vfp_enabled = ARM_TBFLAG_VFPEN(tb->flags);
10594 dc->vec_len = ARM_TBFLAG_VECLEN(tb->flags);
10595 dc->vec_stride = ARM_TBFLAG_VECSTRIDE(tb->flags);
10596 dc->cp_regs = cpu->cp_regs;
10597 dc->current_pl = arm_current_pl(env);
10599 cpu_F0s = tcg_temp_new_i32();
10600 cpu_F1s = tcg_temp_new_i32();
10601 cpu_F0d = tcg_temp_new_i64();
10602 cpu_F1d = tcg_temp_new_i64();
10605 /* FIXME: cpu_M0 can probably be the same as cpu_V0. */
10606 cpu_M0 = tcg_temp_new_i64();
10607 next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
10610 max_insns = tb->cflags & CF_COUNT_MASK;
10611 if (max_insns == 0)
10612 max_insns = CF_COUNT_MASK;
10616 tcg_clear_temp_count();
10618 /* A note on handling of the condexec (IT) bits:
10620 * We want to avoid the overhead of having to write the updated condexec
10621 * bits back to the CPUARMState for every instruction in an IT block. So:
10622 * (1) if the condexec bits are not already zero then we write
10623 * zero back into the CPUARMState now. This avoids complications trying
10624 * to do it at the end of the block. (For example if we don't do this
10625 * it's hard to identify whether we can safely skip writing condexec
10626 * at the end of the TB, which we definitely want to do for the case
10627 * where a TB doesn't do anything with the IT state at all.)
10628 * (2) if we are going to leave the TB then we call gen_set_condexec()
10629 * which will write the correct value into CPUARMState if zero is wrong.
10630 * This is done both for leaving the TB at the end, and for leaving
10631 * it because of an exception we know will happen, which is done in
10632 * gen_exception_insn(). The latter is necessary because we need to
10633 * leave the TB with the PC/IT state just prior to execution of the
10634 * instruction which caused the exception.
10635 * (3) if we leave the TB unexpectedly (eg a data abort on a load)
10636 * then the CPUARMState will be wrong and we need to reset it.
10637 * This is handled in the same way as restoration of the
10638 * PC in these situations: we will be called again with search_pc=1
10639 * and generate a mapping of the condexec bits for each PC in
10640 * gen_opc_condexec_bits[]. restore_state_to_opc() then uses
10641 * this to restore the condexec bits.
10643 * Note that there are no instructions which can read the condexec
10644 * bits, and none which can write non-static values to them, so
10645 * we don't need to care about whether CPUARMState is correct in the
10649 /* Reset the conditional execution bits immediately. This avoids
10650 complications trying to do it at the end of the block. */
10651 if (dc->condexec_mask || dc->condexec_cond)
10653 TCGv_i32 tmp = tcg_temp_new_i32();
10654 tcg_gen_movi_i32(tmp, 0);
10655 store_cpu_field(tmp, condexec_bits);
10658 #ifdef CONFIG_USER_ONLY
10659 /* Intercept jump to the magic kernel page. */
10660 if (dc->pc >= 0xffff0000) {
10661 /* We always get here via a jump, so know we are not in a
10662 conditional execution block. */
10663 gen_exception(EXCP_KERNEL_TRAP);
10664 dc->is_jmp = DISAS_UPDATE;
10668 if (dc->pc >= 0xfffffff0 && IS_M(env)) {
10669 /* We always get here via a jump, so know we are not in a
10670 conditional execution block. */
10671 gen_exception(EXCP_EXCEPTION_EXIT);
10672 dc->is_jmp = DISAS_UPDATE;
10677 if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
10678 QTAILQ_FOREACH(bp, &env->breakpoints, entry) {
10679 if (bp->pc == dc->pc) {
10680 gen_exception_insn(dc, 0, EXCP_DEBUG);
10681 /* Advance PC so that clearing the breakpoint will
10682 invalidate this TB. */
10684 goto done_generating;
10689 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
10693 tcg_ctx.gen_opc_instr_start[lj++] = 0;
10695 tcg_ctx.gen_opc_pc[lj] = dc->pc;
10696 gen_opc_condexec_bits[lj] = (dc->condexec_cond << 4) | (dc->condexec_mask >> 1);
10697 tcg_ctx.gen_opc_instr_start[lj] = 1;
10698 tcg_ctx.gen_opc_icount[lj] = num_insns;
10701 if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO))
10704 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP | CPU_LOG_TB_OP_OPT))) {
10705 tcg_gen_debug_insn_start(dc->pc);
10709 disas_thumb_insn(env, dc);
10710 if (dc->condexec_mask) {
10711 dc->condexec_cond = (dc->condexec_cond & 0xe)
10712 | ((dc->condexec_mask >> 4) & 1);
10713 dc->condexec_mask = (dc->condexec_mask << 1) & 0x1f;
10714 if (dc->condexec_mask == 0) {
10715 dc->condexec_cond = 0;
10719 disas_arm_insn(env, dc);
10722 if (dc->condjmp && !dc->is_jmp) {
10723 gen_set_label(dc->condlabel);
10727 if (tcg_check_temp_count()) {
10728 fprintf(stderr, "TCG temporary leak before "TARGET_FMT_lx"\n",
10732 /* Translation stops when a conditional branch is encountered.
10733 * Otherwise the subsequent code could get translated several times.
10734 * Also stop translation when a page boundary is reached. This
10735 * ensures prefetch aborts occur at the right place. */
10737 } while (!dc->is_jmp && tcg_ctx.gen_opc_ptr < gen_opc_end &&
10738 !cs->singlestep_enabled &&
10740 dc->pc < next_page_start &&
10741 num_insns < max_insns);
10743 if (tb->cflags & CF_LAST_IO) {
10745 /* FIXME: This can theoretically happen with self-modifying
10747 cpu_abort(env, "IO on conditional branch instruction");
10752 /* At this stage dc->condjmp will only be set when the skipped
10753 instruction was a conditional branch or trap, and the PC has
10754 already been written. */
10755 if (unlikely(cs->singlestep_enabled)) {
10756 /* Make sure the pc is updated, and raise a debug exception. */
10758 gen_set_condexec(dc);
10759 if (dc->is_jmp == DISAS_SWI) {
10760 gen_exception(EXCP_SWI);
10762 gen_exception(EXCP_DEBUG);
10764 gen_set_label(dc->condlabel);
10766 if (dc->condjmp || !dc->is_jmp) {
10767 gen_set_pc_im(dc, dc->pc);
10770 gen_set_condexec(dc);
10771 if (dc->is_jmp == DISAS_SWI && !dc->condjmp) {
10772 gen_exception(EXCP_SWI);
10774 /* FIXME: Single stepping a WFI insn will not halt
10776 gen_exception(EXCP_DEBUG);
10779 /* While branches must always occur at the end of an IT block,
10780 there are a few other things that can cause us to terminate
10781 the TB in the middle of an IT block:
10782 - Exception generating instructions (bkpt, swi, undefined).
10784 - Hardware watchpoints.
10785 Hardware breakpoints have already been handled and skip this code.
10787 gen_set_condexec(dc);
10788 switch(dc->is_jmp) {
10790 gen_goto_tb(dc, 1, dc->pc);
10795 /* indicate that the hash table must be used to find the next TB */
10796 tcg_gen_exit_tb(0);
10798 case DISAS_TB_JUMP:
10799 /* nothing more to generate */
10802 gen_helper_wfi(cpu_env);
10805 gen_exception(EXCP_SWI);
10809 gen_set_label(dc->condlabel);
10810 gen_set_condexec(dc);
10811 gen_goto_tb(dc, 1, dc->pc);
10817 gen_tb_end(tb, num_insns);
10818 *tcg_ctx.gen_opc_ptr = INDEX_op_end;
10821 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
10822 qemu_log("----------------\n");
10823 qemu_log("IN: %s\n", lookup_symbol(pc_start));
10824 log_target_disas(env, pc_start, dc->pc - pc_start,
10825 dc->thumb | (dc->bswap_code << 1));
10830 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
10833 tcg_ctx.gen_opc_instr_start[lj++] = 0;
10835 tb->size = dc->pc - pc_start;
10836 tb->icount = num_insns;
10840 void gen_intermediate_code(CPUARMState *env, TranslationBlock *tb)
10842 gen_intermediate_code_internal(arm_env_get_cpu(env), tb, false);
10845 void gen_intermediate_code_pc(CPUARMState *env, TranslationBlock *tb)
10847 gen_intermediate_code_internal(arm_env_get_cpu(env), tb, true);
10850 static const char *cpu_mode_names[16] = {
10851 "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
10852 "???", "???", "???", "und", "???", "???", "???", "sys"
10855 void arm_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
10858 ARMCPU *cpu = ARM_CPU(cs);
10859 CPUARMState *env = &cpu->env;
10863 for(i=0;i<16;i++) {
10864 cpu_fprintf(f, "R%02d=%08x", i, env->regs[i]);
10866 cpu_fprintf(f, "\n");
10868 cpu_fprintf(f, " ");
10870 psr = cpsr_read(env);
10871 cpu_fprintf(f, "PSR=%08x %c%c%c%c %c %s%d\n",
10873 psr & (1 << 31) ? 'N' : '-',
10874 psr & (1 << 30) ? 'Z' : '-',
10875 psr & (1 << 29) ? 'C' : '-',
10876 psr & (1 << 28) ? 'V' : '-',
10877 psr & CPSR_T ? 'T' : 'A',
10878 cpu_mode_names[psr & 0xf], (psr & 0x10) ? 32 : 26);
10880 if (flags & CPU_DUMP_FPU) {
10881 int numvfpregs = 0;
10882 if (arm_feature(env, ARM_FEATURE_VFP)) {
10885 if (arm_feature(env, ARM_FEATURE_VFP3)) {
10888 for (i = 0; i < numvfpregs; i++) {
10889 uint64_t v = float64_val(env->vfp.regs[i]);
10890 cpu_fprintf(f, "s%02d=%08x s%02d=%08x d%02d=%016" PRIx64 "\n",
10891 i * 2, (uint32_t)v,
10892 i * 2 + 1, (uint32_t)(v >> 32),
10895 cpu_fprintf(f, "FPSCR: %08x\n", (int)env->vfp.xregs[ARM_VFP_FPSCR]);
10899 void restore_state_to_opc(CPUARMState *env, TranslationBlock *tb, int pc_pos)
10902 env->pc = tcg_ctx.gen_opc_pc[pc_pos];
10903 env->condexec_bits = 0;
10905 env->regs[15] = tcg_ctx.gen_opc_pc[pc_pos];
10906 env->condexec_bits = gen_opc_condexec_bits[pc_pos];