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/>.
37 #define ENABLE_ARCH_5J 0
38 #define ENABLE_ARCH_6 arm_feature(env, ARM_FEATURE_V6)
39 #define ENABLE_ARCH_6K arm_feature(env, ARM_FEATURE_V6K)
40 #define ENABLE_ARCH_6T2 arm_feature(env, ARM_FEATURE_THUMB2)
41 #define ENABLE_ARCH_7 arm_feature(env, ARM_FEATURE_V7)
43 #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0)
45 /* internal defines */
46 typedef struct DisasContext
{
49 /* Nonzero if this instruction has been conditionally skipped. */
51 /* The label that will be jumped to when the instruction is skipped. */
53 /* Thumb-2 condtional execution bits. */
56 struct TranslationBlock
*tb
;
57 int singlestep_enabled
;
59 #if !defined(CONFIG_USER_ONLY)
67 static uint32_t gen_opc_condexec_bits
[OPC_BUF_SIZE
];
69 #if defined(CONFIG_USER_ONLY)
72 #define IS_USER(s) (s->user)
75 /* These instructions trap after executing, so defer them until after the
76 conditional executions state has been updated. */
80 static TCGv_ptr cpu_env
;
81 /* We reuse the same 64-bit temporaries for efficiency. */
82 static TCGv_i64 cpu_V0
, cpu_V1
, cpu_M0
;
83 static TCGv_i32 cpu_R
[16];
84 static TCGv_i32 cpu_exclusive_addr
;
85 static TCGv_i32 cpu_exclusive_val
;
86 static TCGv_i32 cpu_exclusive_high
;
87 #ifdef CONFIG_USER_ONLY
88 static TCGv_i32 cpu_exclusive_test
;
89 static TCGv_i32 cpu_exclusive_info
;
92 /* FIXME: These should be removed. */
93 static TCGv cpu_F0s
, cpu_F1s
;
94 static TCGv_i64 cpu_F0d
, cpu_F1d
;
96 #include "gen-icount.h"
98 static const char *regnames
[] =
99 { "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
100 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "pc" };
102 /* initialize TCG globals. */
103 void arm_translate_init(void)
107 cpu_env
= tcg_global_reg_new_ptr(TCG_AREG0
, "env");
109 for (i
= 0; i
< 16; i
++) {
110 cpu_R
[i
] = tcg_global_mem_new_i32(TCG_AREG0
,
111 offsetof(CPUState
, regs
[i
]),
114 cpu_exclusive_addr
= tcg_global_mem_new_i32(TCG_AREG0
,
115 offsetof(CPUState
, exclusive_addr
), "exclusive_addr");
116 cpu_exclusive_val
= tcg_global_mem_new_i32(TCG_AREG0
,
117 offsetof(CPUState
, exclusive_val
), "exclusive_val");
118 cpu_exclusive_high
= tcg_global_mem_new_i32(TCG_AREG0
,
119 offsetof(CPUState
, exclusive_high
), "exclusive_high");
120 #ifdef CONFIG_USER_ONLY
121 cpu_exclusive_test
= tcg_global_mem_new_i32(TCG_AREG0
,
122 offsetof(CPUState
, exclusive_test
), "exclusive_test");
123 cpu_exclusive_info
= tcg_global_mem_new_i32(TCG_AREG0
,
124 offsetof(CPUState
, exclusive_info
), "exclusive_info");
131 static int num_temps
;
133 /* Allocate a temporary variable. */
134 static TCGv_i32
new_tmp(void)
137 return tcg_temp_new_i32();
140 /* Release a temporary variable. */
141 static void dead_tmp(TCGv tmp
)
147 static inline TCGv
load_cpu_offset(int offset
)
149 TCGv tmp
= new_tmp();
150 tcg_gen_ld_i32(tmp
, cpu_env
, offset
);
154 #define load_cpu_field(name) load_cpu_offset(offsetof(CPUState, name))
156 static inline void store_cpu_offset(TCGv var
, int offset
)
158 tcg_gen_st_i32(var
, cpu_env
, offset
);
162 #define store_cpu_field(var, name) \
163 store_cpu_offset(var, offsetof(CPUState, name))
165 /* Set a variable to the value of a CPU register. */
166 static void load_reg_var(DisasContext
*s
, TCGv var
, int reg
)
170 /* normaly, since we updated PC, we need only to add one insn */
172 addr
= (long)s
->pc
+ 2;
174 addr
= (long)s
->pc
+ 4;
175 tcg_gen_movi_i32(var
, addr
);
177 tcg_gen_mov_i32(var
, cpu_R
[reg
]);
181 /* Create a new temporary and set it to the value of a CPU register. */
182 static inline TCGv
load_reg(DisasContext
*s
, int reg
)
184 TCGv tmp
= new_tmp();
185 load_reg_var(s
, tmp
, reg
);
189 /* Set a CPU register. The source must be a temporary and will be
191 static void store_reg(DisasContext
*s
, int reg
, TCGv var
)
194 tcg_gen_andi_i32(var
, var
, ~1);
195 s
->is_jmp
= DISAS_JUMP
;
197 tcg_gen_mov_i32(cpu_R
[reg
], var
);
201 /* Value extensions. */
202 #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var)
203 #define gen_uxth(var) tcg_gen_ext16u_i32(var, var)
204 #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
205 #define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
207 #define gen_sxtb16(var) gen_helper_sxtb16(var, var)
208 #define gen_uxtb16(var) gen_helper_uxtb16(var, var)
211 static inline void gen_set_cpsr(TCGv var
, uint32_t mask
)
213 TCGv tmp_mask
= tcg_const_i32(mask
);
214 gen_helper_cpsr_write(var
, tmp_mask
);
215 tcg_temp_free_i32(tmp_mask
);
217 /* Set NZCV flags from the high 4 bits of var. */
218 #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV)
220 static void gen_exception(int excp
)
222 TCGv tmp
= new_tmp();
223 tcg_gen_movi_i32(tmp
, excp
);
224 gen_helper_exception(tmp
);
228 static void gen_smul_dual(TCGv a
, TCGv b
)
230 TCGv tmp1
= new_tmp();
231 TCGv tmp2
= new_tmp();
232 tcg_gen_ext16s_i32(tmp1
, a
);
233 tcg_gen_ext16s_i32(tmp2
, b
);
234 tcg_gen_mul_i32(tmp1
, tmp1
, tmp2
);
236 tcg_gen_sari_i32(a
, a
, 16);
237 tcg_gen_sari_i32(b
, b
, 16);
238 tcg_gen_mul_i32(b
, b
, a
);
239 tcg_gen_mov_i32(a
, tmp1
);
243 /* Byteswap each halfword. */
244 static void gen_rev16(TCGv var
)
246 TCGv tmp
= new_tmp();
247 tcg_gen_shri_i32(tmp
, var
, 8);
248 tcg_gen_andi_i32(tmp
, tmp
, 0x00ff00ff);
249 tcg_gen_shli_i32(var
, var
, 8);
250 tcg_gen_andi_i32(var
, var
, 0xff00ff00);
251 tcg_gen_or_i32(var
, var
, tmp
);
255 /* Byteswap low halfword and sign extend. */
256 static void gen_revsh(TCGv var
)
258 tcg_gen_ext16u_i32(var
, var
);
259 tcg_gen_bswap16_i32(var
, var
);
260 tcg_gen_ext16s_i32(var
, var
);
263 /* Unsigned bitfield extract. */
264 static void gen_ubfx(TCGv var
, int shift
, uint32_t mask
)
267 tcg_gen_shri_i32(var
, var
, shift
);
268 tcg_gen_andi_i32(var
, var
, mask
);
271 /* Signed bitfield extract. */
272 static void gen_sbfx(TCGv var
, int shift
, int width
)
277 tcg_gen_sari_i32(var
, var
, shift
);
278 if (shift
+ width
< 32) {
279 signbit
= 1u << (width
- 1);
280 tcg_gen_andi_i32(var
, var
, (1u << width
) - 1);
281 tcg_gen_xori_i32(var
, var
, signbit
);
282 tcg_gen_subi_i32(var
, var
, signbit
);
286 /* Bitfield insertion. Insert val into base. Clobbers base and val. */
287 static void gen_bfi(TCGv dest
, TCGv base
, TCGv val
, int shift
, uint32_t mask
)
289 tcg_gen_andi_i32(val
, val
, mask
);
290 tcg_gen_shli_i32(val
, val
, shift
);
291 tcg_gen_andi_i32(base
, base
, ~(mask
<< shift
));
292 tcg_gen_or_i32(dest
, base
, val
);
295 /* Return (b << 32) + a. Mark inputs as dead */
296 static TCGv_i64
gen_addq_msw(TCGv_i64 a
, TCGv b
)
298 TCGv_i64 tmp64
= tcg_temp_new_i64();
300 tcg_gen_extu_i32_i64(tmp64
, b
);
302 tcg_gen_shli_i64(tmp64
, tmp64
, 32);
303 tcg_gen_add_i64(a
, tmp64
, a
);
305 tcg_temp_free_i64(tmp64
);
309 /* Return (b << 32) - a. Mark inputs as dead. */
310 static TCGv_i64
gen_subq_msw(TCGv_i64 a
, TCGv b
)
312 TCGv_i64 tmp64
= tcg_temp_new_i64();
314 tcg_gen_extu_i32_i64(tmp64
, b
);
316 tcg_gen_shli_i64(tmp64
, tmp64
, 32);
317 tcg_gen_sub_i64(a
, tmp64
, a
);
319 tcg_temp_free_i64(tmp64
);
323 /* FIXME: Most targets have native widening multiplication.
324 It would be good to use that instead of a full wide multiply. */
325 /* 32x32->64 multiply. Marks inputs as dead. */
326 static TCGv_i64
gen_mulu_i64_i32(TCGv a
, TCGv b
)
328 TCGv_i64 tmp1
= tcg_temp_new_i64();
329 TCGv_i64 tmp2
= tcg_temp_new_i64();
331 tcg_gen_extu_i32_i64(tmp1
, a
);
333 tcg_gen_extu_i32_i64(tmp2
, b
);
335 tcg_gen_mul_i64(tmp1
, tmp1
, tmp2
);
336 tcg_temp_free_i64(tmp2
);
340 static TCGv_i64
gen_muls_i64_i32(TCGv a
, TCGv b
)
342 TCGv_i64 tmp1
= tcg_temp_new_i64();
343 TCGv_i64 tmp2
= tcg_temp_new_i64();
345 tcg_gen_ext_i32_i64(tmp1
, a
);
347 tcg_gen_ext_i32_i64(tmp2
, b
);
349 tcg_gen_mul_i64(tmp1
, tmp1
, tmp2
);
350 tcg_temp_free_i64(tmp2
);
354 /* Swap low and high halfwords. */
355 static void gen_swap_half(TCGv var
)
357 TCGv tmp
= new_tmp();
358 tcg_gen_shri_i32(tmp
, var
, 16);
359 tcg_gen_shli_i32(var
, var
, 16);
360 tcg_gen_or_i32(var
, var
, tmp
);
364 /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
365 tmp = (t0 ^ t1) & 0x8000;
368 t0 = (t0 + t1) ^ tmp;
371 static void gen_add16(TCGv t0
, TCGv t1
)
373 TCGv tmp
= new_tmp();
374 tcg_gen_xor_i32(tmp
, t0
, t1
);
375 tcg_gen_andi_i32(tmp
, tmp
, 0x8000);
376 tcg_gen_andi_i32(t0
, t0
, ~0x8000);
377 tcg_gen_andi_i32(t1
, t1
, ~0x8000);
378 tcg_gen_add_i32(t0
, t0
, t1
);
379 tcg_gen_xor_i32(t0
, t0
, tmp
);
384 #define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, CF))
386 /* Set CF to the top bit of var. */
387 static void gen_set_CF_bit31(TCGv var
)
389 TCGv tmp
= new_tmp();
390 tcg_gen_shri_i32(tmp
, var
, 31);
395 /* Set N and Z flags from var. */
396 static inline void gen_logic_CC(TCGv var
)
398 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUState
, NF
));
399 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUState
, ZF
));
403 static void gen_adc(TCGv t0
, TCGv t1
)
406 tcg_gen_add_i32(t0
, t0
, t1
);
407 tmp
= load_cpu_field(CF
);
408 tcg_gen_add_i32(t0
, t0
, tmp
);
412 /* dest = T0 + T1 + CF. */
413 static void gen_add_carry(TCGv dest
, TCGv t0
, TCGv t1
)
416 tcg_gen_add_i32(dest
, t0
, t1
);
417 tmp
= load_cpu_field(CF
);
418 tcg_gen_add_i32(dest
, dest
, tmp
);
422 /* dest = T0 - T1 + CF - 1. */
423 static void gen_sub_carry(TCGv dest
, TCGv t0
, TCGv t1
)
426 tcg_gen_sub_i32(dest
, t0
, t1
);
427 tmp
= load_cpu_field(CF
);
428 tcg_gen_add_i32(dest
, dest
, tmp
);
429 tcg_gen_subi_i32(dest
, dest
, 1);
433 /* FIXME: Implement this natively. */
434 #define tcg_gen_abs_i32(t0, t1) gen_helper_abs(t0, t1)
436 static void shifter_out_im(TCGv var
, int shift
)
438 TCGv tmp
= new_tmp();
440 tcg_gen_andi_i32(tmp
, var
, 1);
442 tcg_gen_shri_i32(tmp
, var
, shift
);
444 tcg_gen_andi_i32(tmp
, tmp
, 1);
450 /* Shift by immediate. Includes special handling for shift == 0. */
451 static inline void gen_arm_shift_im(TCGv var
, int shiftop
, int shift
, int flags
)
457 shifter_out_im(var
, 32 - shift
);
458 tcg_gen_shli_i32(var
, var
, shift
);
464 tcg_gen_shri_i32(var
, var
, 31);
467 tcg_gen_movi_i32(var
, 0);
470 shifter_out_im(var
, shift
- 1);
471 tcg_gen_shri_i32(var
, var
, shift
);
478 shifter_out_im(var
, shift
- 1);
481 tcg_gen_sari_i32(var
, var
, shift
);
483 case 3: /* ROR/RRX */
486 shifter_out_im(var
, shift
- 1);
487 tcg_gen_rotri_i32(var
, var
, shift
); break;
489 TCGv tmp
= load_cpu_field(CF
);
491 shifter_out_im(var
, 0);
492 tcg_gen_shri_i32(var
, var
, 1);
493 tcg_gen_shli_i32(tmp
, tmp
, 31);
494 tcg_gen_or_i32(var
, var
, tmp
);
500 static inline void gen_arm_shift_reg(TCGv var
, int shiftop
,
501 TCGv shift
, int flags
)
505 case 0: gen_helper_shl_cc(var
, var
, shift
); break;
506 case 1: gen_helper_shr_cc(var
, var
, shift
); break;
507 case 2: gen_helper_sar_cc(var
, var
, shift
); break;
508 case 3: gen_helper_ror_cc(var
, var
, shift
); break;
512 case 0: gen_helper_shl(var
, var
, shift
); break;
513 case 1: gen_helper_shr(var
, var
, shift
); break;
514 case 2: gen_helper_sar(var
, var
, shift
); break;
515 case 3: tcg_gen_andi_i32(shift
, shift
, 0x1f);
516 tcg_gen_rotr_i32(var
, var
, shift
); break;
522 #define PAS_OP(pfx) \
524 case 0: gen_pas_helper(glue(pfx,add16)); break; \
525 case 1: gen_pas_helper(glue(pfx,addsubx)); break; \
526 case 2: gen_pas_helper(glue(pfx,subaddx)); break; \
527 case 3: gen_pas_helper(glue(pfx,sub16)); break; \
528 case 4: gen_pas_helper(glue(pfx,add8)); break; \
529 case 7: gen_pas_helper(glue(pfx,sub8)); break; \
531 static void gen_arm_parallel_addsub(int op1
, int op2
, TCGv a
, TCGv b
)
536 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
538 tmp
= tcg_temp_new_ptr();
539 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
541 tcg_temp_free_ptr(tmp
);
544 tmp
= tcg_temp_new_ptr();
545 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
547 tcg_temp_free_ptr(tmp
);
549 #undef gen_pas_helper
550 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
563 #undef gen_pas_helper
568 /* For unknown reasons Arm and Thumb-2 use arbitrarily different encodings. */
569 #define PAS_OP(pfx) \
571 case 0: gen_pas_helper(glue(pfx,add8)); break; \
572 case 1: gen_pas_helper(glue(pfx,add16)); break; \
573 case 2: gen_pas_helper(glue(pfx,addsubx)); break; \
574 case 4: gen_pas_helper(glue(pfx,sub8)); break; \
575 case 5: gen_pas_helper(glue(pfx,sub16)); break; \
576 case 6: gen_pas_helper(glue(pfx,subaddx)); break; \
578 static void gen_thumb2_parallel_addsub(int op1
, int op2
, TCGv a
, TCGv b
)
583 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
585 tmp
= tcg_temp_new_ptr();
586 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
588 tcg_temp_free_ptr(tmp
);
591 tmp
= tcg_temp_new_ptr();
592 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
594 tcg_temp_free_ptr(tmp
);
596 #undef gen_pas_helper
597 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
610 #undef gen_pas_helper
615 static void gen_test_cc(int cc
, int label
)
623 tmp
= load_cpu_field(ZF
);
624 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
627 tmp
= load_cpu_field(ZF
);
628 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, label
);
631 tmp
= load_cpu_field(CF
);
632 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, label
);
635 tmp
= load_cpu_field(CF
);
636 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
639 tmp
= load_cpu_field(NF
);
640 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
643 tmp
= load_cpu_field(NF
);
644 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
647 tmp
= load_cpu_field(VF
);
648 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
651 tmp
= load_cpu_field(VF
);
652 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
654 case 8: /* hi: C && !Z */
655 inv
= gen_new_label();
656 tmp
= load_cpu_field(CF
);
657 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, inv
);
659 tmp
= load_cpu_field(ZF
);
660 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, label
);
663 case 9: /* ls: !C || Z */
664 tmp
= load_cpu_field(CF
);
665 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
667 tmp
= load_cpu_field(ZF
);
668 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
670 case 10: /* ge: N == V -> N ^ V == 0 */
671 tmp
= load_cpu_field(VF
);
672 tmp2
= load_cpu_field(NF
);
673 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
675 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
677 case 11: /* lt: N != V -> N ^ V != 0 */
678 tmp
= load_cpu_field(VF
);
679 tmp2
= load_cpu_field(NF
);
680 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
682 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
684 case 12: /* gt: !Z && N == V */
685 inv
= gen_new_label();
686 tmp
= load_cpu_field(ZF
);
687 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, inv
);
689 tmp
= load_cpu_field(VF
);
690 tmp2
= load_cpu_field(NF
);
691 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
693 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
696 case 13: /* le: Z || N != V */
697 tmp
= load_cpu_field(ZF
);
698 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
700 tmp
= load_cpu_field(VF
);
701 tmp2
= load_cpu_field(NF
);
702 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
704 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
707 fprintf(stderr
, "Bad condition code 0x%x\n", cc
);
713 static const uint8_t table_logic_cc
[16] = {
732 /* Set PC and Thumb state from an immediate address. */
733 static inline void gen_bx_im(DisasContext
*s
, uint32_t addr
)
737 s
->is_jmp
= DISAS_UPDATE
;
738 if (s
->thumb
!= (addr
& 1)) {
740 tcg_gen_movi_i32(tmp
, addr
& 1);
741 tcg_gen_st_i32(tmp
, cpu_env
, offsetof(CPUState
, thumb
));
744 tcg_gen_movi_i32(cpu_R
[15], addr
& ~1);
747 /* Set PC and Thumb state from var. var is marked as dead. */
748 static inline void gen_bx(DisasContext
*s
, TCGv var
)
750 s
->is_jmp
= DISAS_UPDATE
;
751 tcg_gen_andi_i32(cpu_R
[15], var
, ~1);
752 tcg_gen_andi_i32(var
, var
, 1);
753 store_cpu_field(var
, thumb
);
756 /* Variant of store_reg which uses branch&exchange logic when storing
757 to r15 in ARM architecture v7 and above. The source must be a temporary
758 and will be marked as dead. */
759 static inline void store_reg_bx(CPUState
*env
, DisasContext
*s
,
762 if (reg
== 15 && ENABLE_ARCH_7
) {
765 store_reg(s
, reg
, var
);
769 static inline TCGv
gen_ld8s(TCGv addr
, int index
)
771 TCGv tmp
= new_tmp();
772 tcg_gen_qemu_ld8s(tmp
, addr
, index
);
775 static inline TCGv
gen_ld8u(TCGv addr
, int index
)
777 TCGv tmp
= new_tmp();
778 tcg_gen_qemu_ld8u(tmp
, addr
, index
);
781 static inline TCGv
gen_ld16s(TCGv addr
, int index
)
783 TCGv tmp
= new_tmp();
784 tcg_gen_qemu_ld16s(tmp
, addr
, index
);
787 static inline TCGv
gen_ld16u(TCGv addr
, int index
)
789 TCGv tmp
= new_tmp();
790 tcg_gen_qemu_ld16u(tmp
, addr
, index
);
793 static inline TCGv
gen_ld32(TCGv addr
, int index
)
795 TCGv tmp
= new_tmp();
796 tcg_gen_qemu_ld32u(tmp
, addr
, index
);
799 static inline TCGv_i64
gen_ld64(TCGv addr
, int index
)
801 TCGv_i64 tmp
= tcg_temp_new_i64();
802 tcg_gen_qemu_ld64(tmp
, addr
, index
);
805 static inline void gen_st8(TCGv val
, TCGv addr
, int index
)
807 tcg_gen_qemu_st8(val
, addr
, index
);
810 static inline void gen_st16(TCGv val
, TCGv addr
, int index
)
812 tcg_gen_qemu_st16(val
, addr
, index
);
815 static inline void gen_st32(TCGv val
, TCGv addr
, int index
)
817 tcg_gen_qemu_st32(val
, addr
, index
);
820 static inline void gen_st64(TCGv_i64 val
, TCGv addr
, int index
)
822 tcg_gen_qemu_st64(val
, addr
, index
);
823 tcg_temp_free_i64(val
);
826 static inline void gen_set_pc_im(uint32_t val
)
828 tcg_gen_movi_i32(cpu_R
[15], val
);
831 /* Force a TB lookup after an instruction that changes the CPU state. */
832 static inline void gen_lookup_tb(DisasContext
*s
)
834 tcg_gen_movi_i32(cpu_R
[15], s
->pc
& ~1);
835 s
->is_jmp
= DISAS_UPDATE
;
838 static inline void gen_add_data_offset(DisasContext
*s
, unsigned int insn
,
841 int val
, rm
, shift
, shiftop
;
844 if (!(insn
& (1 << 25))) {
847 if (!(insn
& (1 << 23)))
850 tcg_gen_addi_i32(var
, var
, val
);
854 shift
= (insn
>> 7) & 0x1f;
855 shiftop
= (insn
>> 5) & 3;
856 offset
= load_reg(s
, rm
);
857 gen_arm_shift_im(offset
, shiftop
, shift
, 0);
858 if (!(insn
& (1 << 23)))
859 tcg_gen_sub_i32(var
, var
, offset
);
861 tcg_gen_add_i32(var
, var
, offset
);
866 static inline void gen_add_datah_offset(DisasContext
*s
, unsigned int insn
,
872 if (insn
& (1 << 22)) {
874 val
= (insn
& 0xf) | ((insn
>> 4) & 0xf0);
875 if (!(insn
& (1 << 23)))
879 tcg_gen_addi_i32(var
, var
, val
);
883 tcg_gen_addi_i32(var
, var
, extra
);
885 offset
= load_reg(s
, rm
);
886 if (!(insn
& (1 << 23)))
887 tcg_gen_sub_i32(var
, var
, offset
);
889 tcg_gen_add_i32(var
, var
, offset
);
894 #define VFP_OP2(name) \
895 static inline void gen_vfp_##name(int dp) \
898 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, cpu_env); \
900 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, cpu_env); \
910 static inline void gen_vfp_abs(int dp
)
913 gen_helper_vfp_absd(cpu_F0d
, cpu_F0d
);
915 gen_helper_vfp_abss(cpu_F0s
, cpu_F0s
);
918 static inline void gen_vfp_neg(int dp
)
921 gen_helper_vfp_negd(cpu_F0d
, cpu_F0d
);
923 gen_helper_vfp_negs(cpu_F0s
, cpu_F0s
);
926 static inline void gen_vfp_sqrt(int dp
)
929 gen_helper_vfp_sqrtd(cpu_F0d
, cpu_F0d
, cpu_env
);
931 gen_helper_vfp_sqrts(cpu_F0s
, cpu_F0s
, cpu_env
);
934 static inline void gen_vfp_cmp(int dp
)
937 gen_helper_vfp_cmpd(cpu_F0d
, cpu_F1d
, cpu_env
);
939 gen_helper_vfp_cmps(cpu_F0s
, cpu_F1s
, cpu_env
);
942 static inline void gen_vfp_cmpe(int dp
)
945 gen_helper_vfp_cmped(cpu_F0d
, cpu_F1d
, cpu_env
);
947 gen_helper_vfp_cmpes(cpu_F0s
, cpu_F1s
, cpu_env
);
950 static inline void gen_vfp_F1_ld0(int dp
)
953 tcg_gen_movi_i64(cpu_F1d
, 0);
955 tcg_gen_movi_i32(cpu_F1s
, 0);
958 static inline void gen_vfp_uito(int dp
)
961 gen_helper_vfp_uitod(cpu_F0d
, cpu_F0s
, cpu_env
);
963 gen_helper_vfp_uitos(cpu_F0s
, cpu_F0s
, cpu_env
);
966 static inline void gen_vfp_sito(int dp
)
969 gen_helper_vfp_sitod(cpu_F0d
, cpu_F0s
, cpu_env
);
971 gen_helper_vfp_sitos(cpu_F0s
, cpu_F0s
, cpu_env
);
974 static inline void gen_vfp_toui(int dp
)
977 gen_helper_vfp_touid(cpu_F0s
, cpu_F0d
, cpu_env
);
979 gen_helper_vfp_touis(cpu_F0s
, cpu_F0s
, cpu_env
);
982 static inline void gen_vfp_touiz(int dp
)
985 gen_helper_vfp_touizd(cpu_F0s
, cpu_F0d
, cpu_env
);
987 gen_helper_vfp_touizs(cpu_F0s
, cpu_F0s
, cpu_env
);
990 static inline void gen_vfp_tosi(int dp
)
993 gen_helper_vfp_tosid(cpu_F0s
, cpu_F0d
, cpu_env
);
995 gen_helper_vfp_tosis(cpu_F0s
, cpu_F0s
, cpu_env
);
998 static inline void gen_vfp_tosiz(int dp
)
1001 gen_helper_vfp_tosizd(cpu_F0s
, cpu_F0d
, cpu_env
);
1003 gen_helper_vfp_tosizs(cpu_F0s
, cpu_F0s
, cpu_env
);
1006 #define VFP_GEN_FIX(name) \
1007 static inline void gen_vfp_##name(int dp, int shift) \
1009 TCGv tmp_shift = tcg_const_i32(shift); \
1011 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tmp_shift, cpu_env);\
1013 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, tmp_shift, cpu_env);\
1014 tcg_temp_free_i32(tmp_shift); \
1026 static inline void gen_vfp_ld(DisasContext
*s
, int dp
, TCGv addr
)
1029 tcg_gen_qemu_ld64(cpu_F0d
, addr
, IS_USER(s
));
1031 tcg_gen_qemu_ld32u(cpu_F0s
, addr
, IS_USER(s
));
1034 static inline void gen_vfp_st(DisasContext
*s
, int dp
, TCGv addr
)
1037 tcg_gen_qemu_st64(cpu_F0d
, addr
, IS_USER(s
));
1039 tcg_gen_qemu_st32(cpu_F0s
, addr
, IS_USER(s
));
1043 vfp_reg_offset (int dp
, int reg
)
1046 return offsetof(CPUARMState
, vfp
.regs
[reg
]);
1048 return offsetof(CPUARMState
, vfp
.regs
[reg
>> 1])
1049 + offsetof(CPU_DoubleU
, l
.upper
);
1051 return offsetof(CPUARMState
, vfp
.regs
[reg
>> 1])
1052 + offsetof(CPU_DoubleU
, l
.lower
);
1056 /* Return the offset of a 32-bit piece of a NEON register.
1057 zero is the least significant end of the register. */
1059 neon_reg_offset (int reg
, int n
)
1063 return vfp_reg_offset(0, sreg
);
1066 static TCGv
neon_load_reg(int reg
, int pass
)
1068 TCGv tmp
= new_tmp();
1069 tcg_gen_ld_i32(tmp
, cpu_env
, neon_reg_offset(reg
, pass
));
1073 static void neon_store_reg(int reg
, int pass
, TCGv var
)
1075 tcg_gen_st_i32(var
, cpu_env
, neon_reg_offset(reg
, pass
));
1079 static inline void neon_load_reg64(TCGv_i64 var
, int reg
)
1081 tcg_gen_ld_i64(var
, cpu_env
, vfp_reg_offset(1, reg
));
1084 static inline void neon_store_reg64(TCGv_i64 var
, int reg
)
1086 tcg_gen_st_i64(var
, cpu_env
, vfp_reg_offset(1, reg
));
1089 #define tcg_gen_ld_f32 tcg_gen_ld_i32
1090 #define tcg_gen_ld_f64 tcg_gen_ld_i64
1091 #define tcg_gen_st_f32 tcg_gen_st_i32
1092 #define tcg_gen_st_f64 tcg_gen_st_i64
1094 static inline void gen_mov_F0_vreg(int dp
, int reg
)
1097 tcg_gen_ld_f64(cpu_F0d
, cpu_env
, vfp_reg_offset(dp
, reg
));
1099 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, vfp_reg_offset(dp
, reg
));
1102 static inline void gen_mov_F1_vreg(int dp
, int reg
)
1105 tcg_gen_ld_f64(cpu_F1d
, cpu_env
, vfp_reg_offset(dp
, reg
));
1107 tcg_gen_ld_f32(cpu_F1s
, cpu_env
, vfp_reg_offset(dp
, reg
));
1110 static inline void gen_mov_vreg_F0(int dp
, int reg
)
1113 tcg_gen_st_f64(cpu_F0d
, cpu_env
, vfp_reg_offset(dp
, reg
));
1115 tcg_gen_st_f32(cpu_F0s
, cpu_env
, vfp_reg_offset(dp
, reg
));
1118 #define ARM_CP_RW_BIT (1 << 20)
1120 static inline void iwmmxt_load_reg(TCGv_i64 var
, int reg
)
1122 tcg_gen_ld_i64(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.regs
[reg
]));
1125 static inline void iwmmxt_store_reg(TCGv_i64 var
, int reg
)
1127 tcg_gen_st_i64(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.regs
[reg
]));
1130 static inline TCGv
iwmmxt_load_creg(int reg
)
1132 TCGv var
= new_tmp();
1133 tcg_gen_ld_i32(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.cregs
[reg
]));
1137 static inline void iwmmxt_store_creg(int reg
, TCGv var
)
1139 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.cregs
[reg
]));
1143 static inline void gen_op_iwmmxt_movq_wRn_M0(int rn
)
1145 iwmmxt_store_reg(cpu_M0
, rn
);
1148 static inline void gen_op_iwmmxt_movq_M0_wRn(int rn
)
1150 iwmmxt_load_reg(cpu_M0
, rn
);
1153 static inline void gen_op_iwmmxt_orq_M0_wRn(int rn
)
1155 iwmmxt_load_reg(cpu_V1
, rn
);
1156 tcg_gen_or_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1159 static inline void gen_op_iwmmxt_andq_M0_wRn(int rn
)
1161 iwmmxt_load_reg(cpu_V1
, rn
);
1162 tcg_gen_and_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1165 static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn
)
1167 iwmmxt_load_reg(cpu_V1
, rn
);
1168 tcg_gen_xor_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1171 #define IWMMXT_OP(name) \
1172 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1174 iwmmxt_load_reg(cpu_V1, rn); \
1175 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
1178 #define IWMMXT_OP_ENV(name) \
1179 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1181 iwmmxt_load_reg(cpu_V1, rn); \
1182 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \
1185 #define IWMMXT_OP_ENV_SIZE(name) \
1186 IWMMXT_OP_ENV(name##b) \
1187 IWMMXT_OP_ENV(name##w) \
1188 IWMMXT_OP_ENV(name##l)
1190 #define IWMMXT_OP_ENV1(name) \
1191 static inline void gen_op_iwmmxt_##name##_M0(void) \
1193 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \
1207 IWMMXT_OP_ENV_SIZE(unpackl
)
1208 IWMMXT_OP_ENV_SIZE(unpackh
)
1210 IWMMXT_OP_ENV1(unpacklub
)
1211 IWMMXT_OP_ENV1(unpackluw
)
1212 IWMMXT_OP_ENV1(unpacklul
)
1213 IWMMXT_OP_ENV1(unpackhub
)
1214 IWMMXT_OP_ENV1(unpackhuw
)
1215 IWMMXT_OP_ENV1(unpackhul
)
1216 IWMMXT_OP_ENV1(unpacklsb
)
1217 IWMMXT_OP_ENV1(unpacklsw
)
1218 IWMMXT_OP_ENV1(unpacklsl
)
1219 IWMMXT_OP_ENV1(unpackhsb
)
1220 IWMMXT_OP_ENV1(unpackhsw
)
1221 IWMMXT_OP_ENV1(unpackhsl
)
1223 IWMMXT_OP_ENV_SIZE(cmpeq
)
1224 IWMMXT_OP_ENV_SIZE(cmpgtu
)
1225 IWMMXT_OP_ENV_SIZE(cmpgts
)
1227 IWMMXT_OP_ENV_SIZE(mins
)
1228 IWMMXT_OP_ENV_SIZE(minu
)
1229 IWMMXT_OP_ENV_SIZE(maxs
)
1230 IWMMXT_OP_ENV_SIZE(maxu
)
1232 IWMMXT_OP_ENV_SIZE(subn
)
1233 IWMMXT_OP_ENV_SIZE(addn
)
1234 IWMMXT_OP_ENV_SIZE(subu
)
1235 IWMMXT_OP_ENV_SIZE(addu
)
1236 IWMMXT_OP_ENV_SIZE(subs
)
1237 IWMMXT_OP_ENV_SIZE(adds
)
1239 IWMMXT_OP_ENV(avgb0
)
1240 IWMMXT_OP_ENV(avgb1
)
1241 IWMMXT_OP_ENV(avgw0
)
1242 IWMMXT_OP_ENV(avgw1
)
1246 IWMMXT_OP_ENV(packuw
)
1247 IWMMXT_OP_ENV(packul
)
1248 IWMMXT_OP_ENV(packuq
)
1249 IWMMXT_OP_ENV(packsw
)
1250 IWMMXT_OP_ENV(packsl
)
1251 IWMMXT_OP_ENV(packsq
)
1253 static void gen_op_iwmmxt_set_mup(void)
1256 tmp
= load_cpu_field(iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1257 tcg_gen_ori_i32(tmp
, tmp
, 2);
1258 store_cpu_field(tmp
, iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1261 static void gen_op_iwmmxt_set_cup(void)
1264 tmp
= load_cpu_field(iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1265 tcg_gen_ori_i32(tmp
, tmp
, 1);
1266 store_cpu_field(tmp
, iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1269 static void gen_op_iwmmxt_setpsr_nz(void)
1271 TCGv tmp
= new_tmp();
1272 gen_helper_iwmmxt_setpsr_nz(tmp
, cpu_M0
);
1273 store_cpu_field(tmp
, iwmmxt
.cregs
[ARM_IWMMXT_wCASF
]);
1276 static inline void gen_op_iwmmxt_addl_M0_wRn(int rn
)
1278 iwmmxt_load_reg(cpu_V1
, rn
);
1279 tcg_gen_ext32u_i64(cpu_V1
, cpu_V1
);
1280 tcg_gen_add_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1283 static inline int gen_iwmmxt_address(DisasContext
*s
, uint32_t insn
, TCGv dest
)
1289 rd
= (insn
>> 16) & 0xf;
1290 tmp
= load_reg(s
, rd
);
1292 offset
= (insn
& 0xff) << ((insn
>> 7) & 2);
1293 if (insn
& (1 << 24)) {
1295 if (insn
& (1 << 23))
1296 tcg_gen_addi_i32(tmp
, tmp
, offset
);
1298 tcg_gen_addi_i32(tmp
, tmp
, -offset
);
1299 tcg_gen_mov_i32(dest
, tmp
);
1300 if (insn
& (1 << 21))
1301 store_reg(s
, rd
, tmp
);
1304 } else if (insn
& (1 << 21)) {
1306 tcg_gen_mov_i32(dest
, tmp
);
1307 if (insn
& (1 << 23))
1308 tcg_gen_addi_i32(tmp
, tmp
, offset
);
1310 tcg_gen_addi_i32(tmp
, tmp
, -offset
);
1311 store_reg(s
, rd
, tmp
);
1312 } else if (!(insn
& (1 << 23)))
1317 static inline int gen_iwmmxt_shift(uint32_t insn
, uint32_t mask
, TCGv dest
)
1319 int rd
= (insn
>> 0) & 0xf;
1322 if (insn
& (1 << 8)) {
1323 if (rd
< ARM_IWMMXT_wCGR0
|| rd
> ARM_IWMMXT_wCGR3
) {
1326 tmp
= iwmmxt_load_creg(rd
);
1330 iwmmxt_load_reg(cpu_V0
, rd
);
1331 tcg_gen_trunc_i64_i32(tmp
, cpu_V0
);
1333 tcg_gen_andi_i32(tmp
, tmp
, mask
);
1334 tcg_gen_mov_i32(dest
, tmp
);
1339 /* Disassemble an iwMMXt instruction. Returns nonzero if an error occured
1340 (ie. an undefined instruction). */
1341 static int disas_iwmmxt_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
1344 int rdhi
, rdlo
, rd0
, rd1
, i
;
1346 TCGv tmp
, tmp2
, tmp3
;
1348 if ((insn
& 0x0e000e00) == 0x0c000000) {
1349 if ((insn
& 0x0fe00ff0) == 0x0c400000) {
1351 rdlo
= (insn
>> 12) & 0xf;
1352 rdhi
= (insn
>> 16) & 0xf;
1353 if (insn
& ARM_CP_RW_BIT
) { /* TMRRC */
1354 iwmmxt_load_reg(cpu_V0
, wrd
);
1355 tcg_gen_trunc_i64_i32(cpu_R
[rdlo
], cpu_V0
);
1356 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
1357 tcg_gen_trunc_i64_i32(cpu_R
[rdhi
], cpu_V0
);
1358 } else { /* TMCRR */
1359 tcg_gen_concat_i32_i64(cpu_V0
, cpu_R
[rdlo
], cpu_R
[rdhi
]);
1360 iwmmxt_store_reg(cpu_V0
, wrd
);
1361 gen_op_iwmmxt_set_mup();
1366 wrd
= (insn
>> 12) & 0xf;
1368 if (gen_iwmmxt_address(s
, insn
, addr
)) {
1372 if (insn
& ARM_CP_RW_BIT
) {
1373 if ((insn
>> 28) == 0xf) { /* WLDRW wCx */
1375 tcg_gen_qemu_ld32u(tmp
, addr
, IS_USER(s
));
1376 iwmmxt_store_creg(wrd
, tmp
);
1379 if (insn
& (1 << 8)) {
1380 if (insn
& (1 << 22)) { /* WLDRD */
1381 tcg_gen_qemu_ld64(cpu_M0
, addr
, IS_USER(s
));
1383 } else { /* WLDRW wRd */
1384 tmp
= gen_ld32(addr
, IS_USER(s
));
1387 if (insn
& (1 << 22)) { /* WLDRH */
1388 tmp
= gen_ld16u(addr
, IS_USER(s
));
1389 } else { /* WLDRB */
1390 tmp
= gen_ld8u(addr
, IS_USER(s
));
1394 tcg_gen_extu_i32_i64(cpu_M0
, tmp
);
1397 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1400 if ((insn
>> 28) == 0xf) { /* WSTRW wCx */
1401 tmp
= iwmmxt_load_creg(wrd
);
1402 gen_st32(tmp
, addr
, IS_USER(s
));
1404 gen_op_iwmmxt_movq_M0_wRn(wrd
);
1406 if (insn
& (1 << 8)) {
1407 if (insn
& (1 << 22)) { /* WSTRD */
1409 tcg_gen_qemu_st64(cpu_M0
, addr
, IS_USER(s
));
1410 } else { /* WSTRW wRd */
1411 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1412 gen_st32(tmp
, addr
, IS_USER(s
));
1415 if (insn
& (1 << 22)) { /* WSTRH */
1416 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1417 gen_st16(tmp
, addr
, IS_USER(s
));
1418 } else { /* WSTRB */
1419 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1420 gen_st8(tmp
, addr
, IS_USER(s
));
1429 if ((insn
& 0x0f000000) != 0x0e000000)
1432 switch (((insn
>> 12) & 0xf00) | ((insn
>> 4) & 0xff)) {
1433 case 0x000: /* WOR */
1434 wrd
= (insn
>> 12) & 0xf;
1435 rd0
= (insn
>> 0) & 0xf;
1436 rd1
= (insn
>> 16) & 0xf;
1437 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1438 gen_op_iwmmxt_orq_M0_wRn(rd1
);
1439 gen_op_iwmmxt_setpsr_nz();
1440 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1441 gen_op_iwmmxt_set_mup();
1442 gen_op_iwmmxt_set_cup();
1444 case 0x011: /* TMCR */
1447 rd
= (insn
>> 12) & 0xf;
1448 wrd
= (insn
>> 16) & 0xf;
1450 case ARM_IWMMXT_wCID
:
1451 case ARM_IWMMXT_wCASF
:
1453 case ARM_IWMMXT_wCon
:
1454 gen_op_iwmmxt_set_cup();
1456 case ARM_IWMMXT_wCSSF
:
1457 tmp
= iwmmxt_load_creg(wrd
);
1458 tmp2
= load_reg(s
, rd
);
1459 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
1461 iwmmxt_store_creg(wrd
, tmp
);
1463 case ARM_IWMMXT_wCGR0
:
1464 case ARM_IWMMXT_wCGR1
:
1465 case ARM_IWMMXT_wCGR2
:
1466 case ARM_IWMMXT_wCGR3
:
1467 gen_op_iwmmxt_set_cup();
1468 tmp
= load_reg(s
, rd
);
1469 iwmmxt_store_creg(wrd
, tmp
);
1475 case 0x100: /* WXOR */
1476 wrd
= (insn
>> 12) & 0xf;
1477 rd0
= (insn
>> 0) & 0xf;
1478 rd1
= (insn
>> 16) & 0xf;
1479 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1480 gen_op_iwmmxt_xorq_M0_wRn(rd1
);
1481 gen_op_iwmmxt_setpsr_nz();
1482 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1483 gen_op_iwmmxt_set_mup();
1484 gen_op_iwmmxt_set_cup();
1486 case 0x111: /* TMRC */
1489 rd
= (insn
>> 12) & 0xf;
1490 wrd
= (insn
>> 16) & 0xf;
1491 tmp
= iwmmxt_load_creg(wrd
);
1492 store_reg(s
, rd
, tmp
);
1494 case 0x300: /* WANDN */
1495 wrd
= (insn
>> 12) & 0xf;
1496 rd0
= (insn
>> 0) & 0xf;
1497 rd1
= (insn
>> 16) & 0xf;
1498 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1499 tcg_gen_neg_i64(cpu_M0
, cpu_M0
);
1500 gen_op_iwmmxt_andq_M0_wRn(rd1
);
1501 gen_op_iwmmxt_setpsr_nz();
1502 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1503 gen_op_iwmmxt_set_mup();
1504 gen_op_iwmmxt_set_cup();
1506 case 0x200: /* WAND */
1507 wrd
= (insn
>> 12) & 0xf;
1508 rd0
= (insn
>> 0) & 0xf;
1509 rd1
= (insn
>> 16) & 0xf;
1510 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1511 gen_op_iwmmxt_andq_M0_wRn(rd1
);
1512 gen_op_iwmmxt_setpsr_nz();
1513 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1514 gen_op_iwmmxt_set_mup();
1515 gen_op_iwmmxt_set_cup();
1517 case 0x810: case 0xa10: /* WMADD */
1518 wrd
= (insn
>> 12) & 0xf;
1519 rd0
= (insn
>> 0) & 0xf;
1520 rd1
= (insn
>> 16) & 0xf;
1521 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1522 if (insn
& (1 << 21))
1523 gen_op_iwmmxt_maddsq_M0_wRn(rd1
);
1525 gen_op_iwmmxt_madduq_M0_wRn(rd1
);
1526 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1527 gen_op_iwmmxt_set_mup();
1529 case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */
1530 wrd
= (insn
>> 12) & 0xf;
1531 rd0
= (insn
>> 16) & 0xf;
1532 rd1
= (insn
>> 0) & 0xf;
1533 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1534 switch ((insn
>> 22) & 3) {
1536 gen_op_iwmmxt_unpacklb_M0_wRn(rd1
);
1539 gen_op_iwmmxt_unpacklw_M0_wRn(rd1
);
1542 gen_op_iwmmxt_unpackll_M0_wRn(rd1
);
1547 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1548 gen_op_iwmmxt_set_mup();
1549 gen_op_iwmmxt_set_cup();
1551 case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */
1552 wrd
= (insn
>> 12) & 0xf;
1553 rd0
= (insn
>> 16) & 0xf;
1554 rd1
= (insn
>> 0) & 0xf;
1555 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1556 switch ((insn
>> 22) & 3) {
1558 gen_op_iwmmxt_unpackhb_M0_wRn(rd1
);
1561 gen_op_iwmmxt_unpackhw_M0_wRn(rd1
);
1564 gen_op_iwmmxt_unpackhl_M0_wRn(rd1
);
1569 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1570 gen_op_iwmmxt_set_mup();
1571 gen_op_iwmmxt_set_cup();
1573 case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */
1574 wrd
= (insn
>> 12) & 0xf;
1575 rd0
= (insn
>> 16) & 0xf;
1576 rd1
= (insn
>> 0) & 0xf;
1577 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1578 if (insn
& (1 << 22))
1579 gen_op_iwmmxt_sadw_M0_wRn(rd1
);
1581 gen_op_iwmmxt_sadb_M0_wRn(rd1
);
1582 if (!(insn
& (1 << 20)))
1583 gen_op_iwmmxt_addl_M0_wRn(wrd
);
1584 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1585 gen_op_iwmmxt_set_mup();
1587 case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */
1588 wrd
= (insn
>> 12) & 0xf;
1589 rd0
= (insn
>> 16) & 0xf;
1590 rd1
= (insn
>> 0) & 0xf;
1591 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1592 if (insn
& (1 << 21)) {
1593 if (insn
& (1 << 20))
1594 gen_op_iwmmxt_mulshw_M0_wRn(rd1
);
1596 gen_op_iwmmxt_mulslw_M0_wRn(rd1
);
1598 if (insn
& (1 << 20))
1599 gen_op_iwmmxt_muluhw_M0_wRn(rd1
);
1601 gen_op_iwmmxt_mululw_M0_wRn(rd1
);
1603 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1604 gen_op_iwmmxt_set_mup();
1606 case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */
1607 wrd
= (insn
>> 12) & 0xf;
1608 rd0
= (insn
>> 16) & 0xf;
1609 rd1
= (insn
>> 0) & 0xf;
1610 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1611 if (insn
& (1 << 21))
1612 gen_op_iwmmxt_macsw_M0_wRn(rd1
);
1614 gen_op_iwmmxt_macuw_M0_wRn(rd1
);
1615 if (!(insn
& (1 << 20))) {
1616 iwmmxt_load_reg(cpu_V1
, wrd
);
1617 tcg_gen_add_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1619 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1620 gen_op_iwmmxt_set_mup();
1622 case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */
1623 wrd
= (insn
>> 12) & 0xf;
1624 rd0
= (insn
>> 16) & 0xf;
1625 rd1
= (insn
>> 0) & 0xf;
1626 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1627 switch ((insn
>> 22) & 3) {
1629 gen_op_iwmmxt_cmpeqb_M0_wRn(rd1
);
1632 gen_op_iwmmxt_cmpeqw_M0_wRn(rd1
);
1635 gen_op_iwmmxt_cmpeql_M0_wRn(rd1
);
1640 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1641 gen_op_iwmmxt_set_mup();
1642 gen_op_iwmmxt_set_cup();
1644 case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */
1645 wrd
= (insn
>> 12) & 0xf;
1646 rd0
= (insn
>> 16) & 0xf;
1647 rd1
= (insn
>> 0) & 0xf;
1648 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1649 if (insn
& (1 << 22)) {
1650 if (insn
& (1 << 20))
1651 gen_op_iwmmxt_avgw1_M0_wRn(rd1
);
1653 gen_op_iwmmxt_avgw0_M0_wRn(rd1
);
1655 if (insn
& (1 << 20))
1656 gen_op_iwmmxt_avgb1_M0_wRn(rd1
);
1658 gen_op_iwmmxt_avgb0_M0_wRn(rd1
);
1660 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1661 gen_op_iwmmxt_set_mup();
1662 gen_op_iwmmxt_set_cup();
1664 case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */
1665 wrd
= (insn
>> 12) & 0xf;
1666 rd0
= (insn
>> 16) & 0xf;
1667 rd1
= (insn
>> 0) & 0xf;
1668 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1669 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCGR0
+ ((insn
>> 20) & 3));
1670 tcg_gen_andi_i32(tmp
, tmp
, 7);
1671 iwmmxt_load_reg(cpu_V1
, rd1
);
1672 gen_helper_iwmmxt_align(cpu_M0
, cpu_M0
, cpu_V1
, tmp
);
1674 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1675 gen_op_iwmmxt_set_mup();
1677 case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */
1678 if (((insn
>> 6) & 3) == 3)
1680 rd
= (insn
>> 12) & 0xf;
1681 wrd
= (insn
>> 16) & 0xf;
1682 tmp
= load_reg(s
, rd
);
1683 gen_op_iwmmxt_movq_M0_wRn(wrd
);
1684 switch ((insn
>> 6) & 3) {
1686 tmp2
= tcg_const_i32(0xff);
1687 tmp3
= tcg_const_i32((insn
& 7) << 3);
1690 tmp2
= tcg_const_i32(0xffff);
1691 tmp3
= tcg_const_i32((insn
& 3) << 4);
1694 tmp2
= tcg_const_i32(0xffffffff);
1695 tmp3
= tcg_const_i32((insn
& 1) << 5);
1701 gen_helper_iwmmxt_insr(cpu_M0
, cpu_M0
, tmp
, tmp2
, tmp3
);
1702 tcg_temp_free(tmp3
);
1703 tcg_temp_free(tmp2
);
1705 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1706 gen_op_iwmmxt_set_mup();
1708 case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */
1709 rd
= (insn
>> 12) & 0xf;
1710 wrd
= (insn
>> 16) & 0xf;
1711 if (rd
== 15 || ((insn
>> 22) & 3) == 3)
1713 gen_op_iwmmxt_movq_M0_wRn(wrd
);
1715 switch ((insn
>> 22) & 3) {
1717 tcg_gen_shri_i64(cpu_M0
, cpu_M0
, (insn
& 7) << 3);
1718 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1720 tcg_gen_ext8s_i32(tmp
, tmp
);
1722 tcg_gen_andi_i32(tmp
, tmp
, 0xff);
1726 tcg_gen_shri_i64(cpu_M0
, cpu_M0
, (insn
& 3) << 4);
1727 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1729 tcg_gen_ext16s_i32(tmp
, tmp
);
1731 tcg_gen_andi_i32(tmp
, tmp
, 0xffff);
1735 tcg_gen_shri_i64(cpu_M0
, cpu_M0
, (insn
& 1) << 5);
1736 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1739 store_reg(s
, rd
, tmp
);
1741 case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */
1742 if ((insn
& 0x000ff008) != 0x0003f000 || ((insn
>> 22) & 3) == 3)
1744 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCASF
);
1745 switch ((insn
>> 22) & 3) {
1747 tcg_gen_shri_i32(tmp
, tmp
, ((insn
& 7) << 2) + 0);
1750 tcg_gen_shri_i32(tmp
, tmp
, ((insn
& 3) << 3) + 4);
1753 tcg_gen_shri_i32(tmp
, tmp
, ((insn
& 1) << 4) + 12);
1756 tcg_gen_shli_i32(tmp
, tmp
, 28);
1760 case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */
1761 if (((insn
>> 6) & 3) == 3)
1763 rd
= (insn
>> 12) & 0xf;
1764 wrd
= (insn
>> 16) & 0xf;
1765 tmp
= load_reg(s
, rd
);
1766 switch ((insn
>> 6) & 3) {
1768 gen_helper_iwmmxt_bcstb(cpu_M0
, tmp
);
1771 gen_helper_iwmmxt_bcstw(cpu_M0
, tmp
);
1774 gen_helper_iwmmxt_bcstl(cpu_M0
, tmp
);
1778 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1779 gen_op_iwmmxt_set_mup();
1781 case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */
1782 if ((insn
& 0x000ff00f) != 0x0003f000 || ((insn
>> 22) & 3) == 3)
1784 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCASF
);
1786 tcg_gen_mov_i32(tmp2
, tmp
);
1787 switch ((insn
>> 22) & 3) {
1789 for (i
= 0; i
< 7; i
++) {
1790 tcg_gen_shli_i32(tmp2
, tmp2
, 4);
1791 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
1795 for (i
= 0; i
< 3; i
++) {
1796 tcg_gen_shli_i32(tmp2
, tmp2
, 8);
1797 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
1801 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
1802 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
1809 case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */
1810 wrd
= (insn
>> 12) & 0xf;
1811 rd0
= (insn
>> 16) & 0xf;
1812 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1813 switch ((insn
>> 22) & 3) {
1815 gen_helper_iwmmxt_addcb(cpu_M0
, cpu_M0
);
1818 gen_helper_iwmmxt_addcw(cpu_M0
, cpu_M0
);
1821 gen_helper_iwmmxt_addcl(cpu_M0
, cpu_M0
);
1826 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1827 gen_op_iwmmxt_set_mup();
1829 case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */
1830 if ((insn
& 0x000ff00f) != 0x0003f000 || ((insn
>> 22) & 3) == 3)
1832 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCASF
);
1834 tcg_gen_mov_i32(tmp2
, tmp
);
1835 switch ((insn
>> 22) & 3) {
1837 for (i
= 0; i
< 7; i
++) {
1838 tcg_gen_shli_i32(tmp2
, tmp2
, 4);
1839 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
1843 for (i
= 0; i
< 3; i
++) {
1844 tcg_gen_shli_i32(tmp2
, tmp2
, 8);
1845 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
1849 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
1850 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
1857 case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */
1858 rd
= (insn
>> 12) & 0xf;
1859 rd0
= (insn
>> 16) & 0xf;
1860 if ((insn
& 0xf) != 0 || ((insn
>> 22) & 3) == 3)
1862 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1864 switch ((insn
>> 22) & 3) {
1866 gen_helper_iwmmxt_msbb(tmp
, cpu_M0
);
1869 gen_helper_iwmmxt_msbw(tmp
, cpu_M0
);
1872 gen_helper_iwmmxt_msbl(tmp
, cpu_M0
);
1875 store_reg(s
, rd
, tmp
);
1877 case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */
1878 case 0x906: case 0xb06: case 0xd06: case 0xf06:
1879 wrd
= (insn
>> 12) & 0xf;
1880 rd0
= (insn
>> 16) & 0xf;
1881 rd1
= (insn
>> 0) & 0xf;
1882 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1883 switch ((insn
>> 22) & 3) {
1885 if (insn
& (1 << 21))
1886 gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1
);
1888 gen_op_iwmmxt_cmpgtub_M0_wRn(rd1
);
1891 if (insn
& (1 << 21))
1892 gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1
);
1894 gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1
);
1897 if (insn
& (1 << 21))
1898 gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1
);
1900 gen_op_iwmmxt_cmpgtul_M0_wRn(rd1
);
1905 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1906 gen_op_iwmmxt_set_mup();
1907 gen_op_iwmmxt_set_cup();
1909 case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */
1910 case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e:
1911 wrd
= (insn
>> 12) & 0xf;
1912 rd0
= (insn
>> 16) & 0xf;
1913 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1914 switch ((insn
>> 22) & 3) {
1916 if (insn
& (1 << 21))
1917 gen_op_iwmmxt_unpacklsb_M0();
1919 gen_op_iwmmxt_unpacklub_M0();
1922 if (insn
& (1 << 21))
1923 gen_op_iwmmxt_unpacklsw_M0();
1925 gen_op_iwmmxt_unpackluw_M0();
1928 if (insn
& (1 << 21))
1929 gen_op_iwmmxt_unpacklsl_M0();
1931 gen_op_iwmmxt_unpacklul_M0();
1936 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1937 gen_op_iwmmxt_set_mup();
1938 gen_op_iwmmxt_set_cup();
1940 case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */
1941 case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c:
1942 wrd
= (insn
>> 12) & 0xf;
1943 rd0
= (insn
>> 16) & 0xf;
1944 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1945 switch ((insn
>> 22) & 3) {
1947 if (insn
& (1 << 21))
1948 gen_op_iwmmxt_unpackhsb_M0();
1950 gen_op_iwmmxt_unpackhub_M0();
1953 if (insn
& (1 << 21))
1954 gen_op_iwmmxt_unpackhsw_M0();
1956 gen_op_iwmmxt_unpackhuw_M0();
1959 if (insn
& (1 << 21))
1960 gen_op_iwmmxt_unpackhsl_M0();
1962 gen_op_iwmmxt_unpackhul_M0();
1967 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1968 gen_op_iwmmxt_set_mup();
1969 gen_op_iwmmxt_set_cup();
1971 case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */
1972 case 0x214: case 0x614: case 0xa14: case 0xe14:
1973 if (((insn
>> 22) & 3) == 0)
1975 wrd
= (insn
>> 12) & 0xf;
1976 rd0
= (insn
>> 16) & 0xf;
1977 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1979 if (gen_iwmmxt_shift(insn
, 0xff, tmp
)) {
1983 switch ((insn
>> 22) & 3) {
1985 gen_helper_iwmmxt_srlw(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
1988 gen_helper_iwmmxt_srll(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
1991 gen_helper_iwmmxt_srlq(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
1995 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1996 gen_op_iwmmxt_set_mup();
1997 gen_op_iwmmxt_set_cup();
1999 case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */
2000 case 0x014: case 0x414: case 0x814: case 0xc14:
2001 if (((insn
>> 22) & 3) == 0)
2003 wrd
= (insn
>> 12) & 0xf;
2004 rd0
= (insn
>> 16) & 0xf;
2005 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2007 if (gen_iwmmxt_shift(insn
, 0xff, tmp
)) {
2011 switch ((insn
>> 22) & 3) {
2013 gen_helper_iwmmxt_sraw(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2016 gen_helper_iwmmxt_sral(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2019 gen_helper_iwmmxt_sraq(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2023 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2024 gen_op_iwmmxt_set_mup();
2025 gen_op_iwmmxt_set_cup();
2027 case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */
2028 case 0x114: case 0x514: case 0x914: case 0xd14:
2029 if (((insn
>> 22) & 3) == 0)
2031 wrd
= (insn
>> 12) & 0xf;
2032 rd0
= (insn
>> 16) & 0xf;
2033 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2035 if (gen_iwmmxt_shift(insn
, 0xff, tmp
)) {
2039 switch ((insn
>> 22) & 3) {
2041 gen_helper_iwmmxt_sllw(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2044 gen_helper_iwmmxt_slll(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2047 gen_helper_iwmmxt_sllq(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2051 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2052 gen_op_iwmmxt_set_mup();
2053 gen_op_iwmmxt_set_cup();
2055 case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */
2056 case 0x314: case 0x714: case 0xb14: case 0xf14:
2057 if (((insn
>> 22) & 3) == 0)
2059 wrd
= (insn
>> 12) & 0xf;
2060 rd0
= (insn
>> 16) & 0xf;
2061 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2063 switch ((insn
>> 22) & 3) {
2065 if (gen_iwmmxt_shift(insn
, 0xf, tmp
)) {
2069 gen_helper_iwmmxt_rorw(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2072 if (gen_iwmmxt_shift(insn
, 0x1f, tmp
)) {
2076 gen_helper_iwmmxt_rorl(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2079 if (gen_iwmmxt_shift(insn
, 0x3f, tmp
)) {
2083 gen_helper_iwmmxt_rorq(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2087 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2088 gen_op_iwmmxt_set_mup();
2089 gen_op_iwmmxt_set_cup();
2091 case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */
2092 case 0x916: case 0xb16: case 0xd16: case 0xf16:
2093 wrd
= (insn
>> 12) & 0xf;
2094 rd0
= (insn
>> 16) & 0xf;
2095 rd1
= (insn
>> 0) & 0xf;
2096 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2097 switch ((insn
>> 22) & 3) {
2099 if (insn
& (1 << 21))
2100 gen_op_iwmmxt_minsb_M0_wRn(rd1
);
2102 gen_op_iwmmxt_minub_M0_wRn(rd1
);
2105 if (insn
& (1 << 21))
2106 gen_op_iwmmxt_minsw_M0_wRn(rd1
);
2108 gen_op_iwmmxt_minuw_M0_wRn(rd1
);
2111 if (insn
& (1 << 21))
2112 gen_op_iwmmxt_minsl_M0_wRn(rd1
);
2114 gen_op_iwmmxt_minul_M0_wRn(rd1
);
2119 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2120 gen_op_iwmmxt_set_mup();
2122 case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */
2123 case 0x816: case 0xa16: case 0xc16: case 0xe16:
2124 wrd
= (insn
>> 12) & 0xf;
2125 rd0
= (insn
>> 16) & 0xf;
2126 rd1
= (insn
>> 0) & 0xf;
2127 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2128 switch ((insn
>> 22) & 3) {
2130 if (insn
& (1 << 21))
2131 gen_op_iwmmxt_maxsb_M0_wRn(rd1
);
2133 gen_op_iwmmxt_maxub_M0_wRn(rd1
);
2136 if (insn
& (1 << 21))
2137 gen_op_iwmmxt_maxsw_M0_wRn(rd1
);
2139 gen_op_iwmmxt_maxuw_M0_wRn(rd1
);
2142 if (insn
& (1 << 21))
2143 gen_op_iwmmxt_maxsl_M0_wRn(rd1
);
2145 gen_op_iwmmxt_maxul_M0_wRn(rd1
);
2150 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2151 gen_op_iwmmxt_set_mup();
2153 case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */
2154 case 0x402: case 0x502: case 0x602: case 0x702:
2155 wrd
= (insn
>> 12) & 0xf;
2156 rd0
= (insn
>> 16) & 0xf;
2157 rd1
= (insn
>> 0) & 0xf;
2158 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2159 tmp
= tcg_const_i32((insn
>> 20) & 3);
2160 iwmmxt_load_reg(cpu_V1
, rd1
);
2161 gen_helper_iwmmxt_align(cpu_M0
, cpu_M0
, cpu_V1
, tmp
);
2163 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2164 gen_op_iwmmxt_set_mup();
2166 case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */
2167 case 0x41a: case 0x51a: case 0x61a: case 0x71a:
2168 case 0x81a: case 0x91a: case 0xa1a: case 0xb1a:
2169 case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a:
2170 wrd
= (insn
>> 12) & 0xf;
2171 rd0
= (insn
>> 16) & 0xf;
2172 rd1
= (insn
>> 0) & 0xf;
2173 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2174 switch ((insn
>> 20) & 0xf) {
2176 gen_op_iwmmxt_subnb_M0_wRn(rd1
);
2179 gen_op_iwmmxt_subub_M0_wRn(rd1
);
2182 gen_op_iwmmxt_subsb_M0_wRn(rd1
);
2185 gen_op_iwmmxt_subnw_M0_wRn(rd1
);
2188 gen_op_iwmmxt_subuw_M0_wRn(rd1
);
2191 gen_op_iwmmxt_subsw_M0_wRn(rd1
);
2194 gen_op_iwmmxt_subnl_M0_wRn(rd1
);
2197 gen_op_iwmmxt_subul_M0_wRn(rd1
);
2200 gen_op_iwmmxt_subsl_M0_wRn(rd1
);
2205 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2206 gen_op_iwmmxt_set_mup();
2207 gen_op_iwmmxt_set_cup();
2209 case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */
2210 case 0x41e: case 0x51e: case 0x61e: case 0x71e:
2211 case 0x81e: case 0x91e: case 0xa1e: case 0xb1e:
2212 case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e:
2213 wrd
= (insn
>> 12) & 0xf;
2214 rd0
= (insn
>> 16) & 0xf;
2215 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2216 tmp
= tcg_const_i32(((insn
>> 16) & 0xf0) | (insn
& 0x0f));
2217 gen_helper_iwmmxt_shufh(cpu_M0
, cpu_env
, cpu_M0
, tmp
);
2219 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2220 gen_op_iwmmxt_set_mup();
2221 gen_op_iwmmxt_set_cup();
2223 case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */
2224 case 0x418: case 0x518: case 0x618: case 0x718:
2225 case 0x818: case 0x918: case 0xa18: case 0xb18:
2226 case 0xc18: case 0xd18: case 0xe18: case 0xf18:
2227 wrd
= (insn
>> 12) & 0xf;
2228 rd0
= (insn
>> 16) & 0xf;
2229 rd1
= (insn
>> 0) & 0xf;
2230 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2231 switch ((insn
>> 20) & 0xf) {
2233 gen_op_iwmmxt_addnb_M0_wRn(rd1
);
2236 gen_op_iwmmxt_addub_M0_wRn(rd1
);
2239 gen_op_iwmmxt_addsb_M0_wRn(rd1
);
2242 gen_op_iwmmxt_addnw_M0_wRn(rd1
);
2245 gen_op_iwmmxt_adduw_M0_wRn(rd1
);
2248 gen_op_iwmmxt_addsw_M0_wRn(rd1
);
2251 gen_op_iwmmxt_addnl_M0_wRn(rd1
);
2254 gen_op_iwmmxt_addul_M0_wRn(rd1
);
2257 gen_op_iwmmxt_addsl_M0_wRn(rd1
);
2262 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2263 gen_op_iwmmxt_set_mup();
2264 gen_op_iwmmxt_set_cup();
2266 case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */
2267 case 0x408: case 0x508: case 0x608: case 0x708:
2268 case 0x808: case 0x908: case 0xa08: case 0xb08:
2269 case 0xc08: case 0xd08: case 0xe08: case 0xf08:
2270 if (!(insn
& (1 << 20)) || ((insn
>> 22) & 3) == 0)
2272 wrd
= (insn
>> 12) & 0xf;
2273 rd0
= (insn
>> 16) & 0xf;
2274 rd1
= (insn
>> 0) & 0xf;
2275 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2276 switch ((insn
>> 22) & 3) {
2278 if (insn
& (1 << 21))
2279 gen_op_iwmmxt_packsw_M0_wRn(rd1
);
2281 gen_op_iwmmxt_packuw_M0_wRn(rd1
);
2284 if (insn
& (1 << 21))
2285 gen_op_iwmmxt_packsl_M0_wRn(rd1
);
2287 gen_op_iwmmxt_packul_M0_wRn(rd1
);
2290 if (insn
& (1 << 21))
2291 gen_op_iwmmxt_packsq_M0_wRn(rd1
);
2293 gen_op_iwmmxt_packuq_M0_wRn(rd1
);
2296 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2297 gen_op_iwmmxt_set_mup();
2298 gen_op_iwmmxt_set_cup();
2300 case 0x201: case 0x203: case 0x205: case 0x207:
2301 case 0x209: case 0x20b: case 0x20d: case 0x20f:
2302 case 0x211: case 0x213: case 0x215: case 0x217:
2303 case 0x219: case 0x21b: case 0x21d: case 0x21f:
2304 wrd
= (insn
>> 5) & 0xf;
2305 rd0
= (insn
>> 12) & 0xf;
2306 rd1
= (insn
>> 0) & 0xf;
2307 if (rd0
== 0xf || rd1
== 0xf)
2309 gen_op_iwmmxt_movq_M0_wRn(wrd
);
2310 tmp
= load_reg(s
, rd0
);
2311 tmp2
= load_reg(s
, rd1
);
2312 switch ((insn
>> 16) & 0xf) {
2313 case 0x0: /* TMIA */
2314 gen_helper_iwmmxt_muladdsl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2316 case 0x8: /* TMIAPH */
2317 gen_helper_iwmmxt_muladdsw(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2319 case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */
2320 if (insn
& (1 << 16))
2321 tcg_gen_shri_i32(tmp
, tmp
, 16);
2322 if (insn
& (1 << 17))
2323 tcg_gen_shri_i32(tmp2
, tmp2
, 16);
2324 gen_helper_iwmmxt_muladdswl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2333 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2334 gen_op_iwmmxt_set_mup();
2343 /* Disassemble an XScale DSP instruction. Returns nonzero if an error occured
2344 (ie. an undefined instruction). */
2345 static int disas_dsp_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
2347 int acc
, rd0
, rd1
, rdhi
, rdlo
;
2350 if ((insn
& 0x0ff00f10) == 0x0e200010) {
2351 /* Multiply with Internal Accumulate Format */
2352 rd0
= (insn
>> 12) & 0xf;
2354 acc
= (insn
>> 5) & 7;
2359 tmp
= load_reg(s
, rd0
);
2360 tmp2
= load_reg(s
, rd1
);
2361 switch ((insn
>> 16) & 0xf) {
2363 gen_helper_iwmmxt_muladdsl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2365 case 0x8: /* MIAPH */
2366 gen_helper_iwmmxt_muladdsw(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2368 case 0xc: /* MIABB */
2369 case 0xd: /* MIABT */
2370 case 0xe: /* MIATB */
2371 case 0xf: /* MIATT */
2372 if (insn
& (1 << 16))
2373 tcg_gen_shri_i32(tmp
, tmp
, 16);
2374 if (insn
& (1 << 17))
2375 tcg_gen_shri_i32(tmp2
, tmp2
, 16);
2376 gen_helper_iwmmxt_muladdswl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2384 gen_op_iwmmxt_movq_wRn_M0(acc
);
2388 if ((insn
& 0x0fe00ff8) == 0x0c400000) {
2389 /* Internal Accumulator Access Format */
2390 rdhi
= (insn
>> 16) & 0xf;
2391 rdlo
= (insn
>> 12) & 0xf;
2397 if (insn
& ARM_CP_RW_BIT
) { /* MRA */
2398 iwmmxt_load_reg(cpu_V0
, acc
);
2399 tcg_gen_trunc_i64_i32(cpu_R
[rdlo
], cpu_V0
);
2400 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
2401 tcg_gen_trunc_i64_i32(cpu_R
[rdhi
], cpu_V0
);
2402 tcg_gen_andi_i32(cpu_R
[rdhi
], cpu_R
[rdhi
], (1 << (40 - 32)) - 1);
2404 tcg_gen_concat_i32_i64(cpu_V0
, cpu_R
[rdlo
], cpu_R
[rdhi
]);
2405 iwmmxt_store_reg(cpu_V0
, acc
);
2413 /* Disassemble system coprocessor instruction. Return nonzero if
2414 instruction is not defined. */
2415 static int disas_cp_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
2418 uint32_t rd
= (insn
>> 12) & 0xf;
2419 uint32_t cp
= (insn
>> 8) & 0xf;
2424 if (insn
& ARM_CP_RW_BIT
) {
2425 if (!env
->cp
[cp
].cp_read
)
2427 gen_set_pc_im(s
->pc
);
2429 tmp2
= tcg_const_i32(insn
);
2430 gen_helper_get_cp(tmp
, cpu_env
, tmp2
);
2431 tcg_temp_free(tmp2
);
2432 store_reg(s
, rd
, tmp
);
2434 if (!env
->cp
[cp
].cp_write
)
2436 gen_set_pc_im(s
->pc
);
2437 tmp
= load_reg(s
, rd
);
2438 tmp2
= tcg_const_i32(insn
);
2439 gen_helper_set_cp(cpu_env
, tmp2
, tmp
);
2440 tcg_temp_free(tmp2
);
2446 static int cp15_user_ok(uint32_t insn
)
2448 int cpn
= (insn
>> 16) & 0xf;
2449 int cpm
= insn
& 0xf;
2450 int op
= ((insn
>> 5) & 7) | ((insn
>> 18) & 0x38);
2452 if (cpn
== 13 && cpm
== 0) {
2454 if (op
== 2 || (op
== 3 && (insn
& ARM_CP_RW_BIT
)))
2458 /* ISB, DSB, DMB. */
2459 if ((cpm
== 5 && op
== 4)
2460 || (cpm
== 10 && (op
== 4 || op
== 5)))
2466 static int cp15_tls_load_store(CPUState
*env
, DisasContext
*s
, uint32_t insn
, uint32_t rd
)
2469 int cpn
= (insn
>> 16) & 0xf;
2470 int cpm
= insn
& 0xf;
2471 int op
= ((insn
>> 5) & 7) | ((insn
>> 18) & 0x38);
2473 if (!arm_feature(env
, ARM_FEATURE_V6K
))
2476 if (!(cpn
== 13 && cpm
== 0))
2479 if (insn
& ARM_CP_RW_BIT
) {
2482 tmp
= load_cpu_field(cp15
.c13_tls1
);
2485 tmp
= load_cpu_field(cp15
.c13_tls2
);
2488 tmp
= load_cpu_field(cp15
.c13_tls3
);
2493 store_reg(s
, rd
, tmp
);
2496 tmp
= load_reg(s
, rd
);
2499 store_cpu_field(tmp
, cp15
.c13_tls1
);
2502 store_cpu_field(tmp
, cp15
.c13_tls2
);
2505 store_cpu_field(tmp
, cp15
.c13_tls3
);
2515 /* Disassemble system coprocessor (cp15) instruction. Return nonzero if
2516 instruction is not defined. */
2517 static int disas_cp15_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
2522 /* M profile cores use memory mapped registers instead of cp15. */
2523 if (arm_feature(env
, ARM_FEATURE_M
))
2526 if ((insn
& (1 << 25)) == 0) {
2527 if (insn
& (1 << 20)) {
2531 /* mcrr. Used for block cache operations, so implement as no-op. */
2534 if ((insn
& (1 << 4)) == 0) {
2538 if (IS_USER(s
) && !cp15_user_ok(insn
)) {
2541 if ((insn
& 0x0fff0fff) == 0x0e070f90
2542 || (insn
& 0x0fff0fff) == 0x0e070f58) {
2543 /* Wait for interrupt. */
2544 gen_set_pc_im(s
->pc
);
2545 s
->is_jmp
= DISAS_WFI
;
2548 rd
= (insn
>> 12) & 0xf;
2550 if (cp15_tls_load_store(env
, s
, insn
, rd
))
2553 tmp2
= tcg_const_i32(insn
);
2554 if (insn
& ARM_CP_RW_BIT
) {
2556 gen_helper_get_cp15(tmp
, cpu_env
, tmp2
);
2557 /* If the destination register is r15 then sets condition codes. */
2559 store_reg(s
, rd
, tmp
);
2563 tmp
= load_reg(s
, rd
);
2564 gen_helper_set_cp15(cpu_env
, tmp2
, tmp
);
2566 /* Normally we would always end the TB here, but Linux
2567 * arch/arm/mach-pxa/sleep.S expects two instructions following
2568 * an MMU enable to execute from cache. Imitate this behaviour. */
2569 if (!arm_feature(env
, ARM_FEATURE_XSCALE
) ||
2570 (insn
& 0x0fff0fff) != 0x0e010f10)
2573 tcg_temp_free_i32(tmp2
);
2577 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
2578 #define VFP_SREG(insn, bigbit, smallbit) \
2579 ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
2580 #define VFP_DREG(reg, insn, bigbit, smallbit) do { \
2581 if (arm_feature(env, ARM_FEATURE_VFP3)) { \
2582 reg = (((insn) >> (bigbit)) & 0x0f) \
2583 | (((insn) >> ((smallbit) - 4)) & 0x10); \
2585 if (insn & (1 << (smallbit))) \
2587 reg = ((insn) >> (bigbit)) & 0x0f; \
2590 #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
2591 #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
2592 #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7)
2593 #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7)
2594 #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5)
2595 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5)
2597 /* Move between integer and VFP cores. */
2598 static TCGv
gen_vfp_mrs(void)
2600 TCGv tmp
= new_tmp();
2601 tcg_gen_mov_i32(tmp
, cpu_F0s
);
2605 static void gen_vfp_msr(TCGv tmp
)
2607 tcg_gen_mov_i32(cpu_F0s
, tmp
);
2611 static void gen_neon_dup_u8(TCGv var
, int shift
)
2613 TCGv tmp
= new_tmp();
2615 tcg_gen_shri_i32(var
, var
, shift
);
2616 tcg_gen_ext8u_i32(var
, var
);
2617 tcg_gen_shli_i32(tmp
, var
, 8);
2618 tcg_gen_or_i32(var
, var
, tmp
);
2619 tcg_gen_shli_i32(tmp
, var
, 16);
2620 tcg_gen_or_i32(var
, var
, tmp
);
2624 static void gen_neon_dup_low16(TCGv var
)
2626 TCGv tmp
= new_tmp();
2627 tcg_gen_ext16u_i32(var
, var
);
2628 tcg_gen_shli_i32(tmp
, var
, 16);
2629 tcg_gen_or_i32(var
, var
, tmp
);
2633 static void gen_neon_dup_high16(TCGv var
)
2635 TCGv tmp
= new_tmp();
2636 tcg_gen_andi_i32(var
, var
, 0xffff0000);
2637 tcg_gen_shri_i32(tmp
, var
, 16);
2638 tcg_gen_or_i32(var
, var
, tmp
);
2642 /* Disassemble a VFP instruction. Returns nonzero if an error occured
2643 (ie. an undefined instruction). */
2644 static int disas_vfp_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
2646 uint32_t rd
, rn
, rm
, op
, i
, n
, offset
, delta_d
, delta_m
, bank_mask
;
2652 if (!arm_feature(env
, ARM_FEATURE_VFP
))
2655 if (!s
->vfp_enabled
) {
2656 /* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
2657 if ((insn
& 0x0fe00fff) != 0x0ee00a10)
2659 rn
= (insn
>> 16) & 0xf;
2660 if (rn
!= ARM_VFP_FPSID
&& rn
!= ARM_VFP_FPEXC
2661 && rn
!= ARM_VFP_MVFR1
&& rn
!= ARM_VFP_MVFR0
)
2664 dp
= ((insn
& 0xf00) == 0xb00);
2665 switch ((insn
>> 24) & 0xf) {
2667 if (insn
& (1 << 4)) {
2668 /* single register transfer */
2669 rd
= (insn
>> 12) & 0xf;
2674 VFP_DREG_N(rn
, insn
);
2677 if (insn
& 0x00c00060
2678 && !arm_feature(env
, ARM_FEATURE_NEON
))
2681 pass
= (insn
>> 21) & 1;
2682 if (insn
& (1 << 22)) {
2684 offset
= ((insn
>> 5) & 3) * 8;
2685 } else if (insn
& (1 << 5)) {
2687 offset
= (insn
& (1 << 6)) ? 16 : 0;
2692 if (insn
& ARM_CP_RW_BIT
) {
2694 tmp
= neon_load_reg(rn
, pass
);
2698 tcg_gen_shri_i32(tmp
, tmp
, offset
);
2699 if (insn
& (1 << 23))
2705 if (insn
& (1 << 23)) {
2707 tcg_gen_shri_i32(tmp
, tmp
, 16);
2713 tcg_gen_sari_i32(tmp
, tmp
, 16);
2722 store_reg(s
, rd
, tmp
);
2725 tmp
= load_reg(s
, rd
);
2726 if (insn
& (1 << 23)) {
2729 gen_neon_dup_u8(tmp
, 0);
2730 } else if (size
== 1) {
2731 gen_neon_dup_low16(tmp
);
2733 for (n
= 0; n
<= pass
* 2; n
++) {
2735 tcg_gen_mov_i32(tmp2
, tmp
);
2736 neon_store_reg(rn
, n
, tmp2
);
2738 neon_store_reg(rn
, n
, tmp
);
2743 tmp2
= neon_load_reg(rn
, pass
);
2744 gen_bfi(tmp
, tmp2
, tmp
, offset
, 0xff);
2748 tmp2
= neon_load_reg(rn
, pass
);
2749 gen_bfi(tmp
, tmp2
, tmp
, offset
, 0xffff);
2755 neon_store_reg(rn
, pass
, tmp
);
2759 if ((insn
& 0x6f) != 0x00)
2761 rn
= VFP_SREG_N(insn
);
2762 if (insn
& ARM_CP_RW_BIT
) {
2764 if (insn
& (1 << 21)) {
2765 /* system register */
2770 /* VFP2 allows access to FSID from userspace.
2771 VFP3 restricts all id registers to privileged
2774 && arm_feature(env
, ARM_FEATURE_VFP3
))
2776 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2781 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2783 case ARM_VFP_FPINST
:
2784 case ARM_VFP_FPINST2
:
2785 /* Not present in VFP3. */
2787 || arm_feature(env
, ARM_FEATURE_VFP3
))
2789 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2793 tmp
= load_cpu_field(vfp
.xregs
[ARM_VFP_FPSCR
]);
2794 tcg_gen_andi_i32(tmp
, tmp
, 0xf0000000);
2797 gen_helper_vfp_get_fpscr(tmp
, cpu_env
);
2803 || !arm_feature(env
, ARM_FEATURE_VFP3
))
2805 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2811 gen_mov_F0_vreg(0, rn
);
2812 tmp
= gen_vfp_mrs();
2815 /* Set the 4 flag bits in the CPSR. */
2819 store_reg(s
, rd
, tmp
);
2823 tmp
= load_reg(s
, rd
);
2824 if (insn
& (1 << 21)) {
2826 /* system register */
2831 /* Writes are ignored. */
2834 gen_helper_vfp_set_fpscr(cpu_env
, tmp
);
2841 /* TODO: VFP subarchitecture support.
2842 * For now, keep the EN bit only */
2843 tcg_gen_andi_i32(tmp
, tmp
, 1 << 30);
2844 store_cpu_field(tmp
, vfp
.xregs
[rn
]);
2847 case ARM_VFP_FPINST
:
2848 case ARM_VFP_FPINST2
:
2849 store_cpu_field(tmp
, vfp
.xregs
[rn
]);
2856 gen_mov_vreg_F0(0, rn
);
2861 /* data processing */
2862 /* The opcode is in bits 23, 21, 20 and 6. */
2863 op
= ((insn
>> 20) & 8) | ((insn
>> 19) & 6) | ((insn
>> 6) & 1);
2867 rn
= ((insn
>> 15) & 0x1e) | ((insn
>> 7) & 1);
2869 /* rn is register number */
2870 VFP_DREG_N(rn
, insn
);
2873 if (op
== 15 && (rn
== 15 || ((rn
& 0x1c) == 0x18))) {
2874 /* Integer or single precision destination. */
2875 rd
= VFP_SREG_D(insn
);
2877 VFP_DREG_D(rd
, insn
);
2880 (((rn
& 0x1c) == 0x10) || ((rn
& 0x14) == 0x14))) {
2881 /* VCVT from int is always from S reg regardless of dp bit.
2882 * VCVT with immediate frac_bits has same format as SREG_M
2884 rm
= VFP_SREG_M(insn
);
2886 VFP_DREG_M(rm
, insn
);
2889 rn
= VFP_SREG_N(insn
);
2890 if (op
== 15 && rn
== 15) {
2891 /* Double precision destination. */
2892 VFP_DREG_D(rd
, insn
);
2894 rd
= VFP_SREG_D(insn
);
2896 /* NB that we implicitly rely on the encoding for the frac_bits
2897 * in VCVT of fixed to float being the same as that of an SREG_M
2899 rm
= VFP_SREG_M(insn
);
2902 veclen
= s
->vec_len
;
2903 if (op
== 15 && rn
> 3)
2906 /* Shut up compiler warnings. */
2917 /* Figure out what type of vector operation this is. */
2918 if ((rd
& bank_mask
) == 0) {
2923 delta_d
= (s
->vec_stride
>> 1) + 1;
2925 delta_d
= s
->vec_stride
+ 1;
2927 if ((rm
& bank_mask
) == 0) {
2928 /* mixed scalar/vector */
2937 /* Load the initial operands. */
2942 /* Integer source */
2943 gen_mov_F0_vreg(0, rm
);
2948 gen_mov_F0_vreg(dp
, rd
);
2949 gen_mov_F1_vreg(dp
, rm
);
2953 /* Compare with zero */
2954 gen_mov_F0_vreg(dp
, rd
);
2965 /* Source and destination the same. */
2966 gen_mov_F0_vreg(dp
, rd
);
2969 /* One source operand. */
2970 gen_mov_F0_vreg(dp
, rm
);
2974 /* Two source operands. */
2975 gen_mov_F0_vreg(dp
, rn
);
2976 gen_mov_F1_vreg(dp
, rm
);
2980 /* Perform the calculation. */
2982 case 0: /* mac: fd + (fn * fm) */
2984 gen_mov_F1_vreg(dp
, rd
);
2987 case 1: /* nmac: fd - (fn * fm) */
2990 gen_mov_F1_vreg(dp
, rd
);
2993 case 2: /* msc: -fd + (fn * fm) */
2995 gen_mov_F1_vreg(dp
, rd
);
2998 case 3: /* nmsc: -fd - (fn * fm) */
3001 gen_mov_F1_vreg(dp
, rd
);
3004 case 4: /* mul: fn * fm */
3007 case 5: /* nmul: -(fn * fm) */
3011 case 6: /* add: fn + fm */
3014 case 7: /* sub: fn - fm */
3017 case 8: /* div: fn / fm */
3020 case 14: /* fconst */
3021 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3024 n
= (insn
<< 12) & 0x80000000;
3025 i
= ((insn
>> 12) & 0x70) | (insn
& 0xf);
3032 tcg_gen_movi_i64(cpu_F0d
, ((uint64_t)n
) << 32);
3039 tcg_gen_movi_i32(cpu_F0s
, n
);
3042 case 15: /* extension space */
3056 case 4: /* vcvtb.f32.f16 */
3057 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3059 tmp
= gen_vfp_mrs();
3060 tcg_gen_ext16u_i32(tmp
, tmp
);
3061 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s
, tmp
, cpu_env
);
3064 case 5: /* vcvtt.f32.f16 */
3065 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3067 tmp
= gen_vfp_mrs();
3068 tcg_gen_shri_i32(tmp
, tmp
, 16);
3069 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s
, tmp
, cpu_env
);
3072 case 6: /* vcvtb.f16.f32 */
3073 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3076 gen_helper_vfp_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
3077 gen_mov_F0_vreg(0, rd
);
3078 tmp2
= gen_vfp_mrs();
3079 tcg_gen_andi_i32(tmp2
, tmp2
, 0xffff0000);
3080 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
3084 case 7: /* vcvtt.f16.f32 */
3085 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3088 gen_helper_vfp_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
3089 tcg_gen_shli_i32(tmp
, tmp
, 16);
3090 gen_mov_F0_vreg(0, rd
);
3091 tmp2
= gen_vfp_mrs();
3092 tcg_gen_ext16u_i32(tmp2
, tmp2
);
3093 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
3106 case 11: /* cmpez */
3110 case 15: /* single<->double conversion */
3112 gen_helper_vfp_fcvtsd(cpu_F0s
, cpu_F0d
, cpu_env
);
3114 gen_helper_vfp_fcvtds(cpu_F0d
, cpu_F0s
, cpu_env
);
3116 case 16: /* fuito */
3119 case 17: /* fsito */
3122 case 20: /* fshto */
3123 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3125 gen_vfp_shto(dp
, 16 - rm
);
3127 case 21: /* fslto */
3128 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3130 gen_vfp_slto(dp
, 32 - rm
);
3132 case 22: /* fuhto */
3133 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3135 gen_vfp_uhto(dp
, 16 - rm
);
3137 case 23: /* fulto */
3138 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3140 gen_vfp_ulto(dp
, 32 - rm
);
3142 case 24: /* ftoui */
3145 case 25: /* ftouiz */
3148 case 26: /* ftosi */
3151 case 27: /* ftosiz */
3154 case 28: /* ftosh */
3155 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3157 gen_vfp_tosh(dp
, 16 - rm
);
3159 case 29: /* ftosl */
3160 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3162 gen_vfp_tosl(dp
, 32 - rm
);
3164 case 30: /* ftouh */
3165 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3167 gen_vfp_touh(dp
, 16 - rm
);
3169 case 31: /* ftoul */
3170 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3172 gen_vfp_toul(dp
, 32 - rm
);
3174 default: /* undefined */
3175 printf ("rn:%d\n", rn
);
3179 default: /* undefined */
3180 printf ("op:%d\n", op
);
3184 /* Write back the result. */
3185 if (op
== 15 && (rn
>= 8 && rn
<= 11))
3186 ; /* Comparison, do nothing. */
3187 else if (op
== 15 && dp
&& ((rn
& 0x1c) == 0x18))
3188 /* VCVT double to int: always integer result. */
3189 gen_mov_vreg_F0(0, rd
);
3190 else if (op
== 15 && rn
== 15)
3192 gen_mov_vreg_F0(!dp
, rd
);
3194 gen_mov_vreg_F0(dp
, rd
);
3196 /* break out of the loop if we have finished */
3200 if (op
== 15 && delta_m
== 0) {
3201 /* single source one-many */
3203 rd
= ((rd
+ delta_d
) & (bank_mask
- 1))
3205 gen_mov_vreg_F0(dp
, rd
);
3209 /* Setup the next operands. */
3211 rd
= ((rd
+ delta_d
) & (bank_mask
- 1))
3215 /* One source operand. */
3216 rm
= ((rm
+ delta_m
) & (bank_mask
- 1))
3218 gen_mov_F0_vreg(dp
, rm
);
3220 /* Two source operands. */
3221 rn
= ((rn
+ delta_d
) & (bank_mask
- 1))
3223 gen_mov_F0_vreg(dp
, rn
);
3225 rm
= ((rm
+ delta_m
) & (bank_mask
- 1))
3227 gen_mov_F1_vreg(dp
, rm
);
3235 if (dp
&& (insn
& 0x03e00000) == 0x00400000) {
3236 /* two-register transfer */
3237 rn
= (insn
>> 16) & 0xf;
3238 rd
= (insn
>> 12) & 0xf;
3240 VFP_DREG_M(rm
, insn
);
3242 rm
= VFP_SREG_M(insn
);
3245 if (insn
& ARM_CP_RW_BIT
) {
3248 gen_mov_F0_vreg(0, rm
* 2);
3249 tmp
= gen_vfp_mrs();
3250 store_reg(s
, rd
, tmp
);
3251 gen_mov_F0_vreg(0, rm
* 2 + 1);
3252 tmp
= gen_vfp_mrs();
3253 store_reg(s
, rn
, tmp
);
3255 gen_mov_F0_vreg(0, rm
);
3256 tmp
= gen_vfp_mrs();
3257 store_reg(s
, rn
, tmp
);
3258 gen_mov_F0_vreg(0, rm
+ 1);
3259 tmp
= gen_vfp_mrs();
3260 store_reg(s
, rd
, tmp
);
3265 tmp
= load_reg(s
, rd
);
3267 gen_mov_vreg_F0(0, rm
* 2);
3268 tmp
= load_reg(s
, rn
);
3270 gen_mov_vreg_F0(0, rm
* 2 + 1);
3272 tmp
= load_reg(s
, rn
);
3274 gen_mov_vreg_F0(0, rm
);
3275 tmp
= load_reg(s
, rd
);
3277 gen_mov_vreg_F0(0, rm
+ 1);
3282 rn
= (insn
>> 16) & 0xf;
3284 VFP_DREG_D(rd
, insn
);
3286 rd
= VFP_SREG_D(insn
);
3287 if (s
->thumb
&& rn
== 15) {
3289 tcg_gen_movi_i32(addr
, s
->pc
& ~2);
3291 addr
= load_reg(s
, rn
);
3293 if ((insn
& 0x01200000) == 0x01000000) {
3294 /* Single load/store */
3295 offset
= (insn
& 0xff) << 2;
3296 if ((insn
& (1 << 23)) == 0)
3298 tcg_gen_addi_i32(addr
, addr
, offset
);
3299 if (insn
& (1 << 20)) {
3300 gen_vfp_ld(s
, dp
, addr
);
3301 gen_mov_vreg_F0(dp
, rd
);
3303 gen_mov_F0_vreg(dp
, rd
);
3304 gen_vfp_st(s
, dp
, addr
);
3308 /* load/store multiple */
3310 n
= (insn
>> 1) & 0x7f;
3314 if (insn
& (1 << 24)) /* pre-decrement */
3315 tcg_gen_addi_i32(addr
, addr
, -((insn
& 0xff) << 2));
3321 for (i
= 0; i
< n
; i
++) {
3322 if (insn
& ARM_CP_RW_BIT
) {
3324 gen_vfp_ld(s
, dp
, addr
);
3325 gen_mov_vreg_F0(dp
, rd
+ i
);
3328 gen_mov_F0_vreg(dp
, rd
+ i
);
3329 gen_vfp_st(s
, dp
, addr
);
3331 tcg_gen_addi_i32(addr
, addr
, offset
);
3333 if (insn
& (1 << 21)) {
3335 if (insn
& (1 << 24))
3336 offset
= -offset
* n
;
3337 else if (dp
&& (insn
& 1))
3343 tcg_gen_addi_i32(addr
, addr
, offset
);
3344 store_reg(s
, rn
, addr
);
3352 /* Should never happen. */
3358 static inline void gen_goto_tb(DisasContext
*s
, int n
, uint32_t dest
)
3360 TranslationBlock
*tb
;
3363 if ((tb
->pc
& TARGET_PAGE_MASK
) == (dest
& TARGET_PAGE_MASK
)) {
3365 gen_set_pc_im(dest
);
3366 tcg_gen_exit_tb((long)tb
+ n
);
3368 gen_set_pc_im(dest
);
3373 static inline void gen_jmp (DisasContext
*s
, uint32_t dest
)
3375 if (unlikely(s
->singlestep_enabled
)) {
3376 /* An indirect jump so that we still trigger the debug exception. */
3381 gen_goto_tb(s
, 0, dest
);
3382 s
->is_jmp
= DISAS_TB_JUMP
;
3386 static inline void gen_mulxy(TCGv t0
, TCGv t1
, int x
, int y
)
3389 tcg_gen_sari_i32(t0
, t0
, 16);
3393 tcg_gen_sari_i32(t1
, t1
, 16);
3396 tcg_gen_mul_i32(t0
, t0
, t1
);
3399 /* Return the mask of PSR bits set by a MSR instruction. */
3400 static uint32_t msr_mask(CPUState
*env
, DisasContext
*s
, int flags
, int spsr
) {
3404 if (flags
& (1 << 0))
3406 if (flags
& (1 << 1))
3408 if (flags
& (1 << 2))
3410 if (flags
& (1 << 3))
3413 /* Mask out undefined bits. */
3414 mask
&= ~CPSR_RESERVED
;
3415 if (!arm_feature(env
, ARM_FEATURE_V6
))
3416 mask
&= ~(CPSR_E
| CPSR_GE
);
3417 if (!arm_feature(env
, ARM_FEATURE_THUMB2
))
3419 /* Mask out execution state bits. */
3422 /* Mask out privileged bits. */
3428 /* Returns nonzero if access to the PSR is not permitted. Marks t0 as dead. */
3429 static int gen_set_psr(DisasContext
*s
, uint32_t mask
, int spsr
, TCGv t0
)
3433 /* ??? This is also undefined in system mode. */
3437 tmp
= load_cpu_field(spsr
);
3438 tcg_gen_andi_i32(tmp
, tmp
, ~mask
);
3439 tcg_gen_andi_i32(t0
, t0
, mask
);
3440 tcg_gen_or_i32(tmp
, tmp
, t0
);
3441 store_cpu_field(tmp
, spsr
);
3443 gen_set_cpsr(t0
, mask
);
3450 /* Returns nonzero if access to the PSR is not permitted. */
3451 static int gen_set_psr_im(DisasContext
*s
, uint32_t mask
, int spsr
, uint32_t val
)
3455 tcg_gen_movi_i32(tmp
, val
);
3456 return gen_set_psr(s
, mask
, spsr
, tmp
);
3459 /* Generate an old-style exception return. Marks pc as dead. */
3460 static void gen_exception_return(DisasContext
*s
, TCGv pc
)
3463 store_reg(s
, 15, pc
);
3464 tmp
= load_cpu_field(spsr
);
3465 gen_set_cpsr(tmp
, 0xffffffff);
3467 s
->is_jmp
= DISAS_UPDATE
;
3470 /* Generate a v6 exception return. Marks both values as dead. */
3471 static void gen_rfe(DisasContext
*s
, TCGv pc
, TCGv cpsr
)
3473 gen_set_cpsr(cpsr
, 0xffffffff);
3475 store_reg(s
, 15, pc
);
3476 s
->is_jmp
= DISAS_UPDATE
;
3480 gen_set_condexec (DisasContext
*s
)
3482 if (s
->condexec_mask
) {
3483 uint32_t val
= (s
->condexec_cond
<< 4) | (s
->condexec_mask
>> 1);
3484 TCGv tmp
= new_tmp();
3485 tcg_gen_movi_i32(tmp
, val
);
3486 store_cpu_field(tmp
, condexec_bits
);
3490 static void gen_exception_insn(DisasContext
*s
, int offset
, int excp
)
3492 gen_set_condexec(s
);
3493 gen_set_pc_im(s
->pc
- offset
);
3494 gen_exception(excp
);
3495 s
->is_jmp
= DISAS_JUMP
;
3498 static void gen_nop_hint(DisasContext
*s
, int val
)
3502 gen_set_pc_im(s
->pc
);
3503 s
->is_jmp
= DISAS_WFI
;
3507 /* TODO: Implement SEV and WFE. May help SMP performance. */
3513 #define CPU_V001 cpu_V0, cpu_V0, cpu_V1
3515 static inline int gen_neon_add(int size
, TCGv t0
, TCGv t1
)
3518 case 0: gen_helper_neon_add_u8(t0
, t0
, t1
); break;
3519 case 1: gen_helper_neon_add_u16(t0
, t0
, t1
); break;
3520 case 2: tcg_gen_add_i32(t0
, t0
, t1
); break;
3526 static inline void gen_neon_rsb(int size
, TCGv t0
, TCGv t1
)
3529 case 0: gen_helper_neon_sub_u8(t0
, t1
, t0
); break;
3530 case 1: gen_helper_neon_sub_u16(t0
, t1
, t0
); break;
3531 case 2: tcg_gen_sub_i32(t0
, t1
, t0
); break;
3536 /* 32-bit pairwise ops end up the same as the elementwise versions. */
3537 #define gen_helper_neon_pmax_s32 gen_helper_neon_max_s32
3538 #define gen_helper_neon_pmax_u32 gen_helper_neon_max_u32
3539 #define gen_helper_neon_pmin_s32 gen_helper_neon_min_s32
3540 #define gen_helper_neon_pmin_u32 gen_helper_neon_min_u32
3542 #define GEN_NEON_INTEGER_OP_ENV(name) do { \
3543 switch ((size << 1) | u) { \
3545 gen_helper_neon_##name##_s8(tmp, cpu_env, tmp, tmp2); \
3548 gen_helper_neon_##name##_u8(tmp, cpu_env, tmp, tmp2); \
3551 gen_helper_neon_##name##_s16(tmp, cpu_env, tmp, tmp2); \
3554 gen_helper_neon_##name##_u16(tmp, cpu_env, tmp, tmp2); \
3557 gen_helper_neon_##name##_s32(tmp, cpu_env, tmp, tmp2); \
3560 gen_helper_neon_##name##_u32(tmp, cpu_env, tmp, tmp2); \
3562 default: return 1; \
3565 #define GEN_NEON_INTEGER_OP(name) do { \
3566 switch ((size << 1) | u) { \
3568 gen_helper_neon_##name##_s8(tmp, tmp, tmp2); \
3571 gen_helper_neon_##name##_u8(tmp, tmp, tmp2); \
3574 gen_helper_neon_##name##_s16(tmp, tmp, tmp2); \
3577 gen_helper_neon_##name##_u16(tmp, tmp, tmp2); \
3580 gen_helper_neon_##name##_s32(tmp, tmp, tmp2); \
3583 gen_helper_neon_##name##_u32(tmp, tmp, tmp2); \
3585 default: return 1; \
3588 static TCGv
neon_load_scratch(int scratch
)
3590 TCGv tmp
= new_tmp();
3591 tcg_gen_ld_i32(tmp
, cpu_env
, offsetof(CPUARMState
, vfp
.scratch
[scratch
]));
3595 static void neon_store_scratch(int scratch
, TCGv var
)
3597 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUARMState
, vfp
.scratch
[scratch
]));
3601 static inline TCGv
neon_get_scalar(int size
, int reg
)
3605 tmp
= neon_load_reg(reg
& 7, reg
>> 4);
3607 gen_neon_dup_high16(tmp
);
3609 gen_neon_dup_low16(tmp
);
3612 tmp
= neon_load_reg(reg
& 15, reg
>> 4);
3617 static int gen_neon_unzip(int rd
, int rm
, int size
, int q
)
3620 if (size
== 3 || (!q
&& size
== 2)) {
3623 tmp
= tcg_const_i32(rd
);
3624 tmp2
= tcg_const_i32(rm
);
3628 gen_helper_neon_qunzip8(cpu_env
, tmp
, tmp2
);
3631 gen_helper_neon_qunzip16(cpu_env
, tmp
, tmp2
);
3634 gen_helper_neon_qunzip32(cpu_env
, tmp
, tmp2
);
3642 gen_helper_neon_unzip8(cpu_env
, tmp
, tmp2
);
3645 gen_helper_neon_unzip16(cpu_env
, tmp
, tmp2
);
3651 tcg_temp_free_i32(tmp
);
3652 tcg_temp_free_i32(tmp2
);
3656 static int gen_neon_zip(int rd
, int rm
, int size
, int q
)
3659 if (size
== 3 || (!q
&& size
== 2)) {
3662 tmp
= tcg_const_i32(rd
);
3663 tmp2
= tcg_const_i32(rm
);
3667 gen_helper_neon_qzip8(cpu_env
, tmp
, tmp2
);
3670 gen_helper_neon_qzip16(cpu_env
, tmp
, tmp2
);
3673 gen_helper_neon_qzip32(cpu_env
, tmp
, tmp2
);
3681 gen_helper_neon_zip8(cpu_env
, tmp
, tmp2
);
3684 gen_helper_neon_zip16(cpu_env
, tmp
, tmp2
);
3690 tcg_temp_free_i32(tmp
);
3691 tcg_temp_free_i32(tmp2
);
3695 static void gen_neon_trn_u8(TCGv t0
, TCGv t1
)
3702 tcg_gen_shli_i32(rd
, t0
, 8);
3703 tcg_gen_andi_i32(rd
, rd
, 0xff00ff00);
3704 tcg_gen_andi_i32(tmp
, t1
, 0x00ff00ff);
3705 tcg_gen_or_i32(rd
, rd
, tmp
);
3707 tcg_gen_shri_i32(t1
, t1
, 8);
3708 tcg_gen_andi_i32(t1
, t1
, 0x00ff00ff);
3709 tcg_gen_andi_i32(tmp
, t0
, 0xff00ff00);
3710 tcg_gen_or_i32(t1
, t1
, tmp
);
3711 tcg_gen_mov_i32(t0
, rd
);
3717 static void gen_neon_trn_u16(TCGv t0
, TCGv t1
)
3724 tcg_gen_shli_i32(rd
, t0
, 16);
3725 tcg_gen_andi_i32(tmp
, t1
, 0xffff);
3726 tcg_gen_or_i32(rd
, rd
, tmp
);
3727 tcg_gen_shri_i32(t1
, t1
, 16);
3728 tcg_gen_andi_i32(tmp
, t0
, 0xffff0000);
3729 tcg_gen_or_i32(t1
, t1
, tmp
);
3730 tcg_gen_mov_i32(t0
, rd
);
3741 } neon_ls_element_type
[11] = {
3755 /* Translate a NEON load/store element instruction. Return nonzero if the
3756 instruction is invalid. */
3757 static int disas_neon_ls_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
3776 if (!s
->vfp_enabled
)
3778 VFP_DREG_D(rd
, insn
);
3779 rn
= (insn
>> 16) & 0xf;
3781 load
= (insn
& (1 << 21)) != 0;
3783 if ((insn
& (1 << 23)) == 0) {
3784 /* Load store all elements. */
3785 op
= (insn
>> 8) & 0xf;
3786 size
= (insn
>> 6) & 3;
3789 nregs
= neon_ls_element_type
[op
].nregs
;
3790 interleave
= neon_ls_element_type
[op
].interleave
;
3791 spacing
= neon_ls_element_type
[op
].spacing
;
3792 if (size
== 3 && (interleave
| spacing
) != 1)
3794 load_reg_var(s
, addr
, rn
);
3795 stride
= (1 << size
) * interleave
;
3796 for (reg
= 0; reg
< nregs
; reg
++) {
3797 if (interleave
> 2 || (interleave
== 2 && nregs
== 2)) {
3798 load_reg_var(s
, addr
, rn
);
3799 tcg_gen_addi_i32(addr
, addr
, (1 << size
) * reg
);
3800 } else if (interleave
== 2 && nregs
== 4 && reg
== 2) {
3801 load_reg_var(s
, addr
, rn
);
3802 tcg_gen_addi_i32(addr
, addr
, 1 << size
);
3806 tmp64
= gen_ld64(addr
, IS_USER(s
));
3807 neon_store_reg64(tmp64
, rd
);
3808 tcg_temp_free_i64(tmp64
);
3810 tmp64
= tcg_temp_new_i64();
3811 neon_load_reg64(tmp64
, rd
);
3812 gen_st64(tmp64
, addr
, IS_USER(s
));
3814 tcg_gen_addi_i32(addr
, addr
, stride
);
3816 for (pass
= 0; pass
< 2; pass
++) {
3819 tmp
= gen_ld32(addr
, IS_USER(s
));
3820 neon_store_reg(rd
, pass
, tmp
);
3822 tmp
= neon_load_reg(rd
, pass
);
3823 gen_st32(tmp
, addr
, IS_USER(s
));
3825 tcg_gen_addi_i32(addr
, addr
, stride
);
3826 } else if (size
== 1) {
3828 tmp
= gen_ld16u(addr
, IS_USER(s
));
3829 tcg_gen_addi_i32(addr
, addr
, stride
);
3830 tmp2
= gen_ld16u(addr
, IS_USER(s
));
3831 tcg_gen_addi_i32(addr
, addr
, stride
);
3832 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
3833 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
3835 neon_store_reg(rd
, pass
, tmp
);
3837 tmp
= neon_load_reg(rd
, pass
);
3839 tcg_gen_shri_i32(tmp2
, tmp
, 16);
3840 gen_st16(tmp
, addr
, IS_USER(s
));
3841 tcg_gen_addi_i32(addr
, addr
, stride
);
3842 gen_st16(tmp2
, addr
, IS_USER(s
));
3843 tcg_gen_addi_i32(addr
, addr
, stride
);
3845 } else /* size == 0 */ {
3848 for (n
= 0; n
< 4; n
++) {
3849 tmp
= gen_ld8u(addr
, IS_USER(s
));
3850 tcg_gen_addi_i32(addr
, addr
, stride
);
3854 tcg_gen_shli_i32(tmp
, tmp
, n
* 8);
3855 tcg_gen_or_i32(tmp2
, tmp2
, tmp
);
3859 neon_store_reg(rd
, pass
, tmp2
);
3861 tmp2
= neon_load_reg(rd
, pass
);
3862 for (n
= 0; n
< 4; n
++) {
3865 tcg_gen_mov_i32(tmp
, tmp2
);
3867 tcg_gen_shri_i32(tmp
, tmp2
, n
* 8);
3869 gen_st8(tmp
, addr
, IS_USER(s
));
3870 tcg_gen_addi_i32(addr
, addr
, stride
);
3881 size
= (insn
>> 10) & 3;
3883 /* Load single element to all lanes. */
3886 size
= (insn
>> 6) & 3;
3887 nregs
= ((insn
>> 8) & 3) + 1;
3888 stride
= (insn
& (1 << 5)) ? 2 : 1;
3889 load_reg_var(s
, addr
, rn
);
3890 for (reg
= 0; reg
< nregs
; reg
++) {
3893 tmp
= gen_ld8u(addr
, IS_USER(s
));
3894 gen_neon_dup_u8(tmp
, 0);
3897 tmp
= gen_ld16u(addr
, IS_USER(s
));
3898 gen_neon_dup_low16(tmp
);
3901 tmp
= gen_ld32(addr
, IS_USER(s
));
3905 default: /* Avoid compiler warnings. */
3908 tcg_gen_addi_i32(addr
, addr
, 1 << size
);
3910 tcg_gen_mov_i32(tmp2
, tmp
);
3911 neon_store_reg(rd
, 0, tmp2
);
3912 neon_store_reg(rd
, 1, tmp
);
3915 stride
= (1 << size
) * nregs
;
3917 /* Single element. */
3918 pass
= (insn
>> 7) & 1;
3921 shift
= ((insn
>> 5) & 3) * 8;
3925 shift
= ((insn
>> 6) & 1) * 16;
3926 stride
= (insn
& (1 << 5)) ? 2 : 1;
3930 stride
= (insn
& (1 << 6)) ? 2 : 1;
3935 nregs
= ((insn
>> 8) & 3) + 1;
3936 load_reg_var(s
, addr
, rn
);
3937 for (reg
= 0; reg
< nregs
; reg
++) {
3941 tmp
= gen_ld8u(addr
, IS_USER(s
));
3944 tmp
= gen_ld16u(addr
, IS_USER(s
));
3947 tmp
= gen_ld32(addr
, IS_USER(s
));
3949 default: /* Avoid compiler warnings. */
3953 tmp2
= neon_load_reg(rd
, pass
);
3954 gen_bfi(tmp
, tmp2
, tmp
, shift
, size
? 0xffff : 0xff);
3957 neon_store_reg(rd
, pass
, tmp
);
3958 } else { /* Store */
3959 tmp
= neon_load_reg(rd
, pass
);
3961 tcg_gen_shri_i32(tmp
, tmp
, shift
);
3964 gen_st8(tmp
, addr
, IS_USER(s
));
3967 gen_st16(tmp
, addr
, IS_USER(s
));
3970 gen_st32(tmp
, addr
, IS_USER(s
));
3975 tcg_gen_addi_i32(addr
, addr
, 1 << size
);
3977 stride
= nregs
* (1 << size
);
3984 base
= load_reg(s
, rn
);
3986 tcg_gen_addi_i32(base
, base
, stride
);
3989 index
= load_reg(s
, rm
);
3990 tcg_gen_add_i32(base
, base
, index
);
3993 store_reg(s
, rn
, base
);
3998 /* Bitwise select. dest = c ? t : f. Clobbers T and F. */
3999 static void gen_neon_bsl(TCGv dest
, TCGv t
, TCGv f
, TCGv c
)
4001 tcg_gen_and_i32(t
, t
, c
);
4002 tcg_gen_andc_i32(f
, f
, c
);
4003 tcg_gen_or_i32(dest
, t
, f
);
4006 static inline void gen_neon_narrow(int size
, TCGv dest
, TCGv_i64 src
)
4009 case 0: gen_helper_neon_narrow_u8(dest
, src
); break;
4010 case 1: gen_helper_neon_narrow_u16(dest
, src
); break;
4011 case 2: tcg_gen_trunc_i64_i32(dest
, src
); break;
4016 static inline void gen_neon_narrow_sats(int size
, TCGv dest
, TCGv_i64 src
)
4019 case 0: gen_helper_neon_narrow_sat_s8(dest
, cpu_env
, src
); break;
4020 case 1: gen_helper_neon_narrow_sat_s16(dest
, cpu_env
, src
); break;
4021 case 2: gen_helper_neon_narrow_sat_s32(dest
, cpu_env
, src
); break;
4026 static inline void gen_neon_narrow_satu(int size
, TCGv dest
, TCGv_i64 src
)
4029 case 0: gen_helper_neon_narrow_sat_u8(dest
, cpu_env
, src
); break;
4030 case 1: gen_helper_neon_narrow_sat_u16(dest
, cpu_env
, src
); break;
4031 case 2: gen_helper_neon_narrow_sat_u32(dest
, cpu_env
, src
); break;
4036 static inline void gen_neon_unarrow_sats(int size
, TCGv dest
, TCGv_i64 src
)
4039 case 0: gen_helper_neon_unarrow_sat8(dest
, cpu_env
, src
); break;
4040 case 1: gen_helper_neon_unarrow_sat16(dest
, cpu_env
, src
); break;
4041 case 2: gen_helper_neon_unarrow_sat32(dest
, cpu_env
, src
); break;
4046 static inline void gen_neon_shift_narrow(int size
, TCGv var
, TCGv shift
,
4052 case 1: gen_helper_neon_rshl_u16(var
, var
, shift
); break;
4053 case 2: gen_helper_neon_rshl_u32(var
, var
, shift
); break;
4058 case 1: gen_helper_neon_rshl_s16(var
, var
, shift
); break;
4059 case 2: gen_helper_neon_rshl_s32(var
, var
, shift
); break;
4066 case 1: gen_helper_neon_shl_u16(var
, var
, shift
); break;
4067 case 2: gen_helper_neon_shl_u32(var
, var
, shift
); break;
4072 case 1: gen_helper_neon_shl_s16(var
, var
, shift
); break;
4073 case 2: gen_helper_neon_shl_s32(var
, var
, shift
); break;
4080 static inline void gen_neon_widen(TCGv_i64 dest
, TCGv src
, int size
, int u
)
4084 case 0: gen_helper_neon_widen_u8(dest
, src
); break;
4085 case 1: gen_helper_neon_widen_u16(dest
, src
); break;
4086 case 2: tcg_gen_extu_i32_i64(dest
, src
); break;
4091 case 0: gen_helper_neon_widen_s8(dest
, src
); break;
4092 case 1: gen_helper_neon_widen_s16(dest
, src
); break;
4093 case 2: tcg_gen_ext_i32_i64(dest
, src
); break;
4100 static inline void gen_neon_addl(int size
)
4103 case 0: gen_helper_neon_addl_u16(CPU_V001
); break;
4104 case 1: gen_helper_neon_addl_u32(CPU_V001
); break;
4105 case 2: tcg_gen_add_i64(CPU_V001
); break;
4110 static inline void gen_neon_subl(int size
)
4113 case 0: gen_helper_neon_subl_u16(CPU_V001
); break;
4114 case 1: gen_helper_neon_subl_u32(CPU_V001
); break;
4115 case 2: tcg_gen_sub_i64(CPU_V001
); break;
4120 static inline void gen_neon_negl(TCGv_i64 var
, int size
)
4123 case 0: gen_helper_neon_negl_u16(var
, var
); break;
4124 case 1: gen_helper_neon_negl_u32(var
, var
); break;
4125 case 2: gen_helper_neon_negl_u64(var
, var
); break;
4130 static inline void gen_neon_addl_saturate(TCGv_i64 op0
, TCGv_i64 op1
, int size
)
4133 case 1: gen_helper_neon_addl_saturate_s32(op0
, cpu_env
, op0
, op1
); break;
4134 case 2: gen_helper_neon_addl_saturate_s64(op0
, cpu_env
, op0
, op1
); break;
4139 static inline void gen_neon_mull(TCGv_i64 dest
, TCGv a
, TCGv b
, int size
, int u
)
4143 switch ((size
<< 1) | u
) {
4144 case 0: gen_helper_neon_mull_s8(dest
, a
, b
); break;
4145 case 1: gen_helper_neon_mull_u8(dest
, a
, b
); break;
4146 case 2: gen_helper_neon_mull_s16(dest
, a
, b
); break;
4147 case 3: gen_helper_neon_mull_u16(dest
, a
, b
); break;
4149 tmp
= gen_muls_i64_i32(a
, b
);
4150 tcg_gen_mov_i64(dest
, tmp
);
4153 tmp
= gen_mulu_i64_i32(a
, b
);
4154 tcg_gen_mov_i64(dest
, tmp
);
4159 /* gen_helper_neon_mull_[su]{8|16} do not free their parameters.
4160 Don't forget to clean them now. */
4167 /* Translate a NEON data processing instruction. Return nonzero if the
4168 instruction is invalid.
4169 We process data in a mixture of 32-bit and 64-bit chunks.
4170 Mostly we use 32-bit chunks so we can use normal scalar instructions. */
4172 static int disas_neon_data_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
4185 TCGv tmp
, tmp2
, tmp3
, tmp4
, tmp5
;
4188 if (!s
->vfp_enabled
)
4190 q
= (insn
& (1 << 6)) != 0;
4191 u
= (insn
>> 24) & 1;
4192 VFP_DREG_D(rd
, insn
);
4193 VFP_DREG_N(rn
, insn
);
4194 VFP_DREG_M(rm
, insn
);
4195 size
= (insn
>> 20) & 3;
4196 if ((insn
& (1 << 23)) == 0) {
4197 /* Three register same length. */
4198 op
= ((insn
>> 7) & 0x1e) | ((insn
>> 4) & 1);
4199 if (size
== 3 && (op
== 1 || op
== 5 || op
== 8 || op
== 9
4200 || op
== 10 || op
== 11 || op
== 16)) {
4201 /* 64-bit element instructions. */
4202 for (pass
= 0; pass
< (q
? 2 : 1); pass
++) {
4203 neon_load_reg64(cpu_V0
, rn
+ pass
);
4204 neon_load_reg64(cpu_V1
, rm
+ pass
);
4208 gen_helper_neon_qadd_u64(cpu_V0
, cpu_env
,
4211 gen_helper_neon_qadd_s64(cpu_V0
, cpu_env
,
4217 gen_helper_neon_qsub_u64(cpu_V0
, cpu_env
,
4220 gen_helper_neon_qsub_s64(cpu_V0
, cpu_env
,
4226 gen_helper_neon_shl_u64(cpu_V0
, cpu_V1
, cpu_V0
);
4228 gen_helper_neon_shl_s64(cpu_V0
, cpu_V1
, cpu_V0
);
4233 gen_helper_neon_qshl_u64(cpu_V0
, cpu_env
,
4236 gen_helper_neon_qshl_s64(cpu_V0
, cpu_env
,
4240 case 10: /* VRSHL */
4242 gen_helper_neon_rshl_u64(cpu_V0
, cpu_V1
, cpu_V0
);
4244 gen_helper_neon_rshl_s64(cpu_V0
, cpu_V1
, cpu_V0
);
4247 case 11: /* VQRSHL */
4249 gen_helper_neon_qrshl_u64(cpu_V0
, cpu_env
,
4252 gen_helper_neon_qrshl_s64(cpu_V0
, cpu_env
,
4258 tcg_gen_sub_i64(CPU_V001
);
4260 tcg_gen_add_i64(CPU_V001
);
4266 neon_store_reg64(cpu_V0
, rd
+ pass
);
4273 case 10: /* VRSHL */
4274 case 11: /* VQRSHL */
4277 /* Shift instruction operands are reversed. */
4284 case 20: /* VPMAX */
4285 case 21: /* VPMIN */
4286 case 23: /* VPADD */
4289 case 26: /* VPADD (float) */
4290 pairwise
= (u
&& size
< 2);
4292 case 30: /* VPMIN/VPMAX (float) */
4300 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
4309 tmp
= neon_load_reg(rn
, n
);
4310 tmp2
= neon_load_reg(rn
, n
+ 1);
4312 tmp
= neon_load_reg(rm
, n
);
4313 tmp2
= neon_load_reg(rm
, n
+ 1);
4317 tmp
= neon_load_reg(rn
, pass
);
4318 tmp2
= neon_load_reg(rm
, pass
);
4322 GEN_NEON_INTEGER_OP(hadd
);
4325 GEN_NEON_INTEGER_OP_ENV(qadd
);
4327 case 2: /* VRHADD */
4328 GEN_NEON_INTEGER_OP(rhadd
);
4330 case 3: /* Logic ops. */
4331 switch ((u
<< 2) | size
) {
4333 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
4336 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
4339 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
4342 tcg_gen_orc_i32(tmp
, tmp
, tmp2
);
4345 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
4348 tmp3
= neon_load_reg(rd
, pass
);
4349 gen_neon_bsl(tmp
, tmp
, tmp2
, tmp3
);
4353 tmp3
= neon_load_reg(rd
, pass
);
4354 gen_neon_bsl(tmp
, tmp
, tmp3
, tmp2
);
4358 tmp3
= neon_load_reg(rd
, pass
);
4359 gen_neon_bsl(tmp
, tmp3
, tmp
, tmp2
);
4365 GEN_NEON_INTEGER_OP(hsub
);
4368 GEN_NEON_INTEGER_OP_ENV(qsub
);
4371 GEN_NEON_INTEGER_OP(cgt
);
4374 GEN_NEON_INTEGER_OP(cge
);
4377 GEN_NEON_INTEGER_OP(shl
);
4380 GEN_NEON_INTEGER_OP_ENV(qshl
);
4382 case 10: /* VRSHL */
4383 GEN_NEON_INTEGER_OP(rshl
);
4385 case 11: /* VQRSHL */
4386 GEN_NEON_INTEGER_OP_ENV(qrshl
);
4389 GEN_NEON_INTEGER_OP(max
);
4392 GEN_NEON_INTEGER_OP(min
);
4395 GEN_NEON_INTEGER_OP(abd
);
4398 GEN_NEON_INTEGER_OP(abd
);
4400 tmp2
= neon_load_reg(rd
, pass
);
4401 gen_neon_add(size
, tmp
, tmp2
);
4404 if (!u
) { /* VADD */
4405 if (gen_neon_add(size
, tmp
, tmp2
))
4409 case 0: gen_helper_neon_sub_u8(tmp
, tmp
, tmp2
); break;
4410 case 1: gen_helper_neon_sub_u16(tmp
, tmp
, tmp2
); break;
4411 case 2: tcg_gen_sub_i32(tmp
, tmp
, tmp2
); break;
4417 if (!u
) { /* VTST */
4419 case 0: gen_helper_neon_tst_u8(tmp
, tmp
, tmp2
); break;
4420 case 1: gen_helper_neon_tst_u16(tmp
, tmp
, tmp2
); break;
4421 case 2: gen_helper_neon_tst_u32(tmp
, tmp
, tmp2
); break;
4426 case 0: gen_helper_neon_ceq_u8(tmp
, tmp
, tmp2
); break;
4427 case 1: gen_helper_neon_ceq_u16(tmp
, tmp
, tmp2
); break;
4428 case 2: gen_helper_neon_ceq_u32(tmp
, tmp
, tmp2
); break;
4433 case 18: /* Multiply. */
4435 case 0: gen_helper_neon_mul_u8(tmp
, tmp
, tmp2
); break;
4436 case 1: gen_helper_neon_mul_u16(tmp
, tmp
, tmp2
); break;
4437 case 2: tcg_gen_mul_i32(tmp
, tmp
, tmp2
); break;
4441 tmp2
= neon_load_reg(rd
, pass
);
4443 gen_neon_rsb(size
, tmp
, tmp2
);
4445 gen_neon_add(size
, tmp
, tmp2
);
4449 if (u
) { /* polynomial */
4450 gen_helper_neon_mul_p8(tmp
, tmp
, tmp2
);
4451 } else { /* Integer */
4453 case 0: gen_helper_neon_mul_u8(tmp
, tmp
, tmp2
); break;
4454 case 1: gen_helper_neon_mul_u16(tmp
, tmp
, tmp2
); break;
4455 case 2: tcg_gen_mul_i32(tmp
, tmp
, tmp2
); break;
4460 case 20: /* VPMAX */
4461 GEN_NEON_INTEGER_OP(pmax
);
4463 case 21: /* VPMIN */
4464 GEN_NEON_INTEGER_OP(pmin
);
4466 case 22: /* Hultiply high. */
4467 if (!u
) { /* VQDMULH */
4469 case 1: gen_helper_neon_qdmulh_s16(tmp
, cpu_env
, tmp
, tmp2
); break;
4470 case 2: gen_helper_neon_qdmulh_s32(tmp
, cpu_env
, tmp
, tmp2
); break;
4473 } else { /* VQRDHMUL */
4475 case 1: gen_helper_neon_qrdmulh_s16(tmp
, cpu_env
, tmp
, tmp2
); break;
4476 case 2: gen_helper_neon_qrdmulh_s32(tmp
, cpu_env
, tmp
, tmp2
); break;
4481 case 23: /* VPADD */
4485 case 0: gen_helper_neon_padd_u8(tmp
, tmp
, tmp2
); break;
4486 case 1: gen_helper_neon_padd_u16(tmp
, tmp
, tmp2
); break;
4487 case 2: tcg_gen_add_i32(tmp
, tmp
, tmp2
); break;
4491 case 26: /* Floating point arithnetic. */
4492 switch ((u
<< 2) | size
) {
4494 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
4497 gen_helper_neon_sub_f32(tmp
, tmp
, tmp2
);
4500 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
4503 gen_helper_neon_abd_f32(tmp
, tmp
, tmp2
);
4509 case 27: /* Float multiply. */
4510 gen_helper_neon_mul_f32(tmp
, tmp
, tmp2
);
4513 tmp2
= neon_load_reg(rd
, pass
);
4515 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
4517 gen_helper_neon_sub_f32(tmp
, tmp2
, tmp
);
4521 case 28: /* Float compare. */
4523 gen_helper_neon_ceq_f32(tmp
, tmp
, tmp2
);
4526 gen_helper_neon_cge_f32(tmp
, tmp
, tmp2
);
4528 gen_helper_neon_cgt_f32(tmp
, tmp
, tmp2
);
4531 case 29: /* Float compare absolute. */
4535 gen_helper_neon_acge_f32(tmp
, tmp
, tmp2
);
4537 gen_helper_neon_acgt_f32(tmp
, tmp
, tmp2
);
4539 case 30: /* Float min/max. */
4541 gen_helper_neon_max_f32(tmp
, tmp
, tmp2
);
4543 gen_helper_neon_min_f32(tmp
, tmp
, tmp2
);
4547 gen_helper_recps_f32(tmp
, tmp
, tmp2
, cpu_env
);
4549 gen_helper_rsqrts_f32(tmp
, tmp
, tmp2
, cpu_env
);
4556 /* Save the result. For elementwise operations we can put it
4557 straight into the destination register. For pairwise operations
4558 we have to be careful to avoid clobbering the source operands. */
4559 if (pairwise
&& rd
== rm
) {
4560 neon_store_scratch(pass
, tmp
);
4562 neon_store_reg(rd
, pass
, tmp
);
4566 if (pairwise
&& rd
== rm
) {
4567 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
4568 tmp
= neon_load_scratch(pass
);
4569 neon_store_reg(rd
, pass
, tmp
);
4572 /* End of 3 register same size operations. */
4573 } else if (insn
& (1 << 4)) {
4574 if ((insn
& 0x00380080) != 0) {
4575 /* Two registers and shift. */
4576 op
= (insn
>> 8) & 0xf;
4577 if (insn
& (1 << 7)) {
4582 while ((insn
& (1 << (size
+ 19))) == 0)
4585 shift
= (insn
>> 16) & ((1 << (3 + size
)) - 1);
4586 /* To avoid excessive dumplication of ops we implement shift
4587 by immediate using the variable shift operations. */
4589 /* Shift by immediate:
4590 VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
4591 /* Right shifts are encoded as N - shift, where N is the
4592 element size in bits. */
4594 shift
= shift
- (1 << (size
+ 3));
4602 imm
= (uint8_t) shift
;
4607 imm
= (uint16_t) shift
;
4618 for (pass
= 0; pass
< count
; pass
++) {
4620 neon_load_reg64(cpu_V0
, rm
+ pass
);
4621 tcg_gen_movi_i64(cpu_V1
, imm
);
4626 gen_helper_neon_shl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4628 gen_helper_neon_shl_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4633 gen_helper_neon_rshl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4635 gen_helper_neon_rshl_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4640 gen_helper_neon_shl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4642 case 5: /* VSHL, VSLI */
4643 gen_helper_neon_shl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4645 case 6: /* VQSHLU */
4647 gen_helper_neon_qshlu_s64(cpu_V0
, cpu_env
,
4655 gen_helper_neon_qshl_u64(cpu_V0
, cpu_env
,
4658 gen_helper_neon_qshl_s64(cpu_V0
, cpu_env
,
4663 if (op
== 1 || op
== 3) {
4665 neon_load_reg64(cpu_V1
, rd
+ pass
);
4666 tcg_gen_add_i64(cpu_V0
, cpu_V0
, cpu_V1
);
4667 } else if (op
== 4 || (op
== 5 && u
)) {
4669 neon_load_reg64(cpu_V1
, rd
+ pass
);
4671 if (shift
< -63 || shift
> 63) {
4675 mask
= 0xffffffffffffffffull
>> -shift
;
4677 mask
= 0xffffffffffffffffull
<< shift
;
4680 tcg_gen_andi_i64(cpu_V1
, cpu_V1
, ~mask
);
4681 tcg_gen_or_i64(cpu_V0
, cpu_V0
, cpu_V1
);
4683 neon_store_reg64(cpu_V0
, rd
+ pass
);
4684 } else { /* size < 3 */
4685 /* Operands in T0 and T1. */
4686 tmp
= neon_load_reg(rm
, pass
);
4688 tcg_gen_movi_i32(tmp2
, imm
);
4692 GEN_NEON_INTEGER_OP(shl
);
4696 GEN_NEON_INTEGER_OP(rshl
);
4701 GEN_NEON_INTEGER_OP(shl
);
4703 case 5: /* VSHL, VSLI */
4705 case 0: gen_helper_neon_shl_u8(tmp
, tmp
, tmp2
); break;
4706 case 1: gen_helper_neon_shl_u16(tmp
, tmp
, tmp2
); break;
4707 case 2: gen_helper_neon_shl_u32(tmp
, tmp
, tmp2
); break;
4711 case 6: /* VQSHLU */
4717 gen_helper_neon_qshlu_s8(tmp
, cpu_env
,
4721 gen_helper_neon_qshlu_s16(tmp
, cpu_env
,
4725 gen_helper_neon_qshlu_s32(tmp
, cpu_env
,
4733 GEN_NEON_INTEGER_OP_ENV(qshl
);
4738 if (op
== 1 || op
== 3) {
4740 tmp2
= neon_load_reg(rd
, pass
);
4741 gen_neon_add(size
, tmp
, tmp2
);
4743 } else if (op
== 4 || (op
== 5 && u
)) {
4748 mask
= 0xff >> -shift
;
4750 mask
= (uint8_t)(0xff << shift
);
4756 mask
= 0xffff >> -shift
;
4758 mask
= (uint16_t)(0xffff << shift
);
4762 if (shift
< -31 || shift
> 31) {
4766 mask
= 0xffffffffu
>> -shift
;
4768 mask
= 0xffffffffu
<< shift
;
4774 tmp2
= neon_load_reg(rd
, pass
);
4775 tcg_gen_andi_i32(tmp
, tmp
, mask
);
4776 tcg_gen_andi_i32(tmp2
, tmp2
, ~mask
);
4777 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
4780 neon_store_reg(rd
, pass
, tmp
);
4783 } else if (op
< 10) {
4784 /* Shift by immediate and narrow:
4785 VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
4786 int input_unsigned
= (op
== 8) ? !u
: u
;
4788 shift
= shift
- (1 << (size
+ 3));
4792 imm
= (uint16_t)shift
;
4794 tmp2
= tcg_const_i32(imm
);
4795 TCGV_UNUSED_I64(tmp64
);
4798 imm
= (uint32_t)shift
;
4799 tmp2
= tcg_const_i32(imm
);
4800 TCGV_UNUSED_I64(tmp64
);
4803 tmp64
= tcg_const_i64(shift
);
4810 for (pass
= 0; pass
< 2; pass
++) {
4812 neon_load_reg64(cpu_V0
, rm
+ pass
);
4814 if (input_unsigned
) {
4815 gen_helper_neon_rshl_u64(cpu_V0
, cpu_V0
,
4818 gen_helper_neon_rshl_s64(cpu_V0
, cpu_V0
,
4822 if (input_unsigned
) {
4823 gen_helper_neon_shl_u64(cpu_V0
, cpu_V0
,
4826 gen_helper_neon_shl_s64(cpu_V0
, cpu_V0
,
4831 tmp
= neon_load_reg(rm
+ pass
, 0);
4832 gen_neon_shift_narrow(size
, tmp
, tmp2
, q
,
4834 tmp3
= neon_load_reg(rm
+ pass
, 1);
4835 gen_neon_shift_narrow(size
, tmp3
, tmp2
, q
,
4837 tcg_gen_concat_i32_i64(cpu_V0
, tmp
, tmp3
);
4843 if (u
) { /* VQSHRUN / VQRSHRUN */
4844 gen_neon_unarrow_sats(size
- 1, tmp
, cpu_V0
);
4845 } else { /* VSHRN / VRSHRN */
4846 gen_neon_narrow(size
- 1, tmp
, cpu_V0
);
4849 if (u
) { /* VQSHRN / VQRSHRN */
4850 gen_neon_narrow_satu(size
- 1, tmp
, cpu_V0
);
4851 } else { /* VQSHRN / VQRSHRN */
4852 gen_neon_narrow_sats(size
- 1, tmp
, cpu_V0
);
4855 neon_store_reg(rd
, pass
, tmp
);
4858 tcg_temp_free_i64(tmp64
);
4860 tcg_temp_free_i32(tmp2
);
4862 } else if (op
== 10) {
4866 tmp
= neon_load_reg(rm
, 0);
4867 tmp2
= neon_load_reg(rm
, 1);
4868 for (pass
= 0; pass
< 2; pass
++) {
4872 gen_neon_widen(cpu_V0
, tmp
, size
, u
);
4875 /* The shift is less than the width of the source
4876 type, so we can just shift the whole register. */
4877 tcg_gen_shli_i64(cpu_V0
, cpu_V0
, shift
);
4878 /* Widen the result of shift: we need to clear
4879 * the potential overflow bits resulting from
4880 * left bits of the narrow input appearing as
4881 * right bits of left the neighbour narrow
4883 if (size
< 2 || !u
) {
4886 imm
= (0xffu
>> (8 - shift
));
4888 } else if (size
== 1) {
4889 imm
= 0xffff >> (16 - shift
);
4892 imm
= 0xffffffff >> (32 - shift
);
4895 imm64
= imm
| (((uint64_t)imm
) << 32);
4899 tcg_gen_andi_i64(cpu_V0
, cpu_V0
, ~imm64
);
4902 neon_store_reg64(cpu_V0
, rd
+ pass
);
4904 } else if (op
>= 14) {
4905 /* VCVT fixed-point. */
4906 /* We have already masked out the must-be-1 top bit of imm6,
4907 * hence this 32-shift where the ARM ARM has 64-imm6.
4910 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
4911 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, pass
));
4914 gen_vfp_ulto(0, shift
);
4916 gen_vfp_slto(0, shift
);
4919 gen_vfp_toul(0, shift
);
4921 gen_vfp_tosl(0, shift
);
4923 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, pass
));
4928 } else { /* (insn & 0x00380080) == 0 */
4931 op
= (insn
>> 8) & 0xf;
4932 /* One register and immediate. */
4933 imm
= (u
<< 7) | ((insn
>> 12) & 0x70) | (insn
& 0xf);
4934 invert
= (insn
& (1 << 5)) != 0;
4952 imm
= (imm
<< 8) | (imm
<< 24);
4955 imm
= (imm
<< 8) | 0xff;
4958 imm
= (imm
<< 16) | 0xffff;
4961 imm
|= (imm
<< 8) | (imm
<< 16) | (imm
<< 24);
4966 imm
= ((imm
& 0x80) << 24) | ((imm
& 0x3f) << 19)
4967 | ((imm
& 0x40) ? (0x1f << 25) : (1 << 30));
4973 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
4974 if (op
& 1 && op
< 12) {
4975 tmp
= neon_load_reg(rd
, pass
);
4977 /* The immediate value has already been inverted, so
4979 tcg_gen_andi_i32(tmp
, tmp
, imm
);
4981 tcg_gen_ori_i32(tmp
, tmp
, imm
);
4986 if (op
== 14 && invert
) {
4989 for (n
= 0; n
< 4; n
++) {
4990 if (imm
& (1 << (n
+ (pass
& 1) * 4)))
4991 val
|= 0xff << (n
* 8);
4993 tcg_gen_movi_i32(tmp
, val
);
4995 tcg_gen_movi_i32(tmp
, imm
);
4998 neon_store_reg(rd
, pass
, tmp
);
5001 } else { /* (insn & 0x00800010 == 0x00800000) */
5003 op
= (insn
>> 8) & 0xf;
5004 if ((insn
& (1 << 6)) == 0) {
5005 /* Three registers of different lengths. */
5009 /* prewiden, src1_wide, src2_wide */
5010 static const int neon_3reg_wide
[16][3] = {
5011 {1, 0, 0}, /* VADDL */
5012 {1, 1, 0}, /* VADDW */
5013 {1, 0, 0}, /* VSUBL */
5014 {1, 1, 0}, /* VSUBW */
5015 {0, 1, 1}, /* VADDHN */
5016 {0, 0, 0}, /* VABAL */
5017 {0, 1, 1}, /* VSUBHN */
5018 {0, 0, 0}, /* VABDL */
5019 {0, 0, 0}, /* VMLAL */
5020 {0, 0, 0}, /* VQDMLAL */
5021 {0, 0, 0}, /* VMLSL */
5022 {0, 0, 0}, /* VQDMLSL */
5023 {0, 0, 0}, /* Integer VMULL */
5024 {0, 0, 0}, /* VQDMULL */
5025 {0, 0, 0} /* Polynomial VMULL */
5028 prewiden
= neon_3reg_wide
[op
][0];
5029 src1_wide
= neon_3reg_wide
[op
][1];
5030 src2_wide
= neon_3reg_wide
[op
][2];
5032 if (size
== 0 && (op
== 9 || op
== 11 || op
== 13))
5035 /* Avoid overlapping operands. Wide source operands are
5036 always aligned so will never overlap with wide
5037 destinations in problematic ways. */
5038 if (rd
== rm
&& !src2_wide
) {
5039 tmp
= neon_load_reg(rm
, 1);
5040 neon_store_scratch(2, tmp
);
5041 } else if (rd
== rn
&& !src1_wide
) {
5042 tmp
= neon_load_reg(rn
, 1);
5043 neon_store_scratch(2, tmp
);
5046 for (pass
= 0; pass
< 2; pass
++) {
5048 neon_load_reg64(cpu_V0
, rn
+ pass
);
5051 if (pass
== 1 && rd
== rn
) {
5052 tmp
= neon_load_scratch(2);
5054 tmp
= neon_load_reg(rn
, pass
);
5057 gen_neon_widen(cpu_V0
, tmp
, size
, u
);
5061 neon_load_reg64(cpu_V1
, rm
+ pass
);
5064 if (pass
== 1 && rd
== rm
) {
5065 tmp2
= neon_load_scratch(2);
5067 tmp2
= neon_load_reg(rm
, pass
);
5070 gen_neon_widen(cpu_V1
, tmp2
, size
, u
);
5074 case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
5075 gen_neon_addl(size
);
5077 case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */
5078 gen_neon_subl(size
);
5080 case 5: case 7: /* VABAL, VABDL */
5081 switch ((size
<< 1) | u
) {
5083 gen_helper_neon_abdl_s16(cpu_V0
, tmp
, tmp2
);
5086 gen_helper_neon_abdl_u16(cpu_V0
, tmp
, tmp2
);
5089 gen_helper_neon_abdl_s32(cpu_V0
, tmp
, tmp2
);
5092 gen_helper_neon_abdl_u32(cpu_V0
, tmp
, tmp2
);
5095 gen_helper_neon_abdl_s64(cpu_V0
, tmp
, tmp2
);
5098 gen_helper_neon_abdl_u64(cpu_V0
, tmp
, tmp2
);
5105 case 8: case 9: case 10: case 11: case 12: case 13:
5106 /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
5107 gen_neon_mull(cpu_V0
, tmp
, tmp2
, size
, u
);
5109 case 14: /* Polynomial VMULL */
5110 gen_helper_neon_mull_p8(cpu_V0
, tmp
, tmp2
);
5114 default: /* 15 is RESERVED. */
5119 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5120 neon_store_reg64(cpu_V0
, rd
+ pass
);
5121 } else if (op
== 5 || (op
>= 8 && op
<= 11)) {
5123 neon_load_reg64(cpu_V1
, rd
+ pass
);
5125 case 10: /* VMLSL */
5126 gen_neon_negl(cpu_V0
, size
);
5128 case 5: case 8: /* VABAL, VMLAL */
5129 gen_neon_addl(size
);
5131 case 9: case 11: /* VQDMLAL, VQDMLSL */
5132 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5134 gen_neon_negl(cpu_V0
, size
);
5136 gen_neon_addl_saturate(cpu_V0
, cpu_V1
, size
);
5141 neon_store_reg64(cpu_V0
, rd
+ pass
);
5142 } else if (op
== 4 || op
== 6) {
5143 /* Narrowing operation. */
5148 gen_helper_neon_narrow_high_u8(tmp
, cpu_V0
);
5151 gen_helper_neon_narrow_high_u16(tmp
, cpu_V0
);
5154 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
5155 tcg_gen_trunc_i64_i32(tmp
, cpu_V0
);
5162 gen_helper_neon_narrow_round_high_u8(tmp
, cpu_V0
);
5165 gen_helper_neon_narrow_round_high_u16(tmp
, cpu_V0
);
5168 tcg_gen_addi_i64(cpu_V0
, cpu_V0
, 1u << 31);
5169 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
5170 tcg_gen_trunc_i64_i32(tmp
, cpu_V0
);
5178 neon_store_reg(rd
, 0, tmp3
);
5179 neon_store_reg(rd
, 1, tmp
);
5182 /* Write back the result. */
5183 neon_store_reg64(cpu_V0
, rd
+ pass
);
5187 /* Two registers and a scalar. */
5189 case 0: /* Integer VMLA scalar */
5190 case 1: /* Float VMLA scalar */
5191 case 4: /* Integer VMLS scalar */
5192 case 5: /* Floating point VMLS scalar */
5193 case 8: /* Integer VMUL scalar */
5194 case 9: /* Floating point VMUL scalar */
5195 case 12: /* VQDMULH scalar */
5196 case 13: /* VQRDMULH scalar */
5197 tmp
= neon_get_scalar(size
, rm
);
5198 neon_store_scratch(0, tmp
);
5199 for (pass
= 0; pass
< (u
? 4 : 2); pass
++) {
5200 tmp
= neon_load_scratch(0);
5201 tmp2
= neon_load_reg(rn
, pass
);
5204 gen_helper_neon_qdmulh_s16(tmp
, cpu_env
, tmp
, tmp2
);
5206 gen_helper_neon_qdmulh_s32(tmp
, cpu_env
, tmp
, tmp2
);
5208 } else if (op
== 13) {
5210 gen_helper_neon_qrdmulh_s16(tmp
, cpu_env
, tmp
, tmp2
);
5212 gen_helper_neon_qrdmulh_s32(tmp
, cpu_env
, tmp
, tmp2
);
5214 } else if (op
& 1) {
5215 gen_helper_neon_mul_f32(tmp
, tmp
, tmp2
);
5218 case 0: gen_helper_neon_mul_u8(tmp
, tmp
, tmp2
); break;
5219 case 1: gen_helper_neon_mul_u16(tmp
, tmp
, tmp2
); break;
5220 case 2: tcg_gen_mul_i32(tmp
, tmp
, tmp2
); break;
5227 tmp2
= neon_load_reg(rd
, pass
);
5230 gen_neon_add(size
, tmp
, tmp2
);
5233 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
5236 gen_neon_rsb(size
, tmp
, tmp2
);
5239 gen_helper_neon_sub_f32(tmp
, tmp2
, tmp
);
5246 neon_store_reg(rd
, pass
, tmp
);
5249 case 2: /* VMLAL sclar */
5250 case 3: /* VQDMLAL scalar */
5251 case 6: /* VMLSL scalar */
5252 case 7: /* VQDMLSL scalar */
5253 case 10: /* VMULL scalar */
5254 case 11: /* VQDMULL scalar */
5255 if (size
== 0 && (op
== 3 || op
== 7 || op
== 11))
5258 tmp2
= neon_get_scalar(size
, rm
);
5259 /* We need a copy of tmp2 because gen_neon_mull
5260 * deletes it during pass 0. */
5262 tcg_gen_mov_i32(tmp4
, tmp2
);
5263 tmp3
= neon_load_reg(rn
, 1);
5265 for (pass
= 0; pass
< 2; pass
++) {
5267 tmp
= neon_load_reg(rn
, 0);
5272 gen_neon_mull(cpu_V0
, tmp
, tmp2
, size
, u
);
5274 neon_load_reg64(cpu_V1
, rd
+ pass
);
5278 gen_neon_negl(cpu_V0
, size
);
5281 gen_neon_addl(size
);
5284 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5286 gen_neon_negl(cpu_V0
, size
);
5288 gen_neon_addl_saturate(cpu_V0
, cpu_V1
, size
);
5294 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5299 neon_store_reg64(cpu_V0
, rd
+ pass
);
5304 default: /* 14 and 15 are RESERVED */
5308 } else { /* size == 3 */
5311 imm
= (insn
>> 8) & 0xf;
5317 neon_load_reg64(cpu_V0
, rn
);
5319 neon_load_reg64(cpu_V1
, rn
+ 1);
5321 } else if (imm
== 8) {
5322 neon_load_reg64(cpu_V0
, rn
+ 1);
5324 neon_load_reg64(cpu_V1
, rm
);
5327 tmp64
= tcg_temp_new_i64();
5329 neon_load_reg64(cpu_V0
, rn
);
5330 neon_load_reg64(tmp64
, rn
+ 1);
5332 neon_load_reg64(cpu_V0
, rn
+ 1);
5333 neon_load_reg64(tmp64
, rm
);
5335 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, (imm
& 7) * 8);
5336 tcg_gen_shli_i64(cpu_V1
, tmp64
, 64 - ((imm
& 7) * 8));
5337 tcg_gen_or_i64(cpu_V0
, cpu_V0
, cpu_V1
);
5339 neon_load_reg64(cpu_V1
, rm
);
5341 neon_load_reg64(cpu_V1
, rm
+ 1);
5344 tcg_gen_shli_i64(cpu_V1
, cpu_V1
, 64 - (imm
* 8));
5345 tcg_gen_shri_i64(tmp64
, tmp64
, imm
* 8);
5346 tcg_gen_or_i64(cpu_V1
, cpu_V1
, tmp64
);
5347 tcg_temp_free_i64(tmp64
);
5350 neon_load_reg64(cpu_V0
, rn
);
5351 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, imm
* 8);
5352 neon_load_reg64(cpu_V1
, rm
);
5353 tcg_gen_shli_i64(cpu_V1
, cpu_V1
, 64 - (imm
* 8));
5354 tcg_gen_or_i64(cpu_V0
, cpu_V0
, cpu_V1
);
5356 neon_store_reg64(cpu_V0
, rd
);
5358 neon_store_reg64(cpu_V1
, rd
+ 1);
5360 } else if ((insn
& (1 << 11)) == 0) {
5361 /* Two register misc. */
5362 op
= ((insn
>> 12) & 0x30) | ((insn
>> 7) & 0xf);
5363 size
= (insn
>> 18) & 3;
5365 case 0: /* VREV64 */
5368 for (pass
= 0; pass
< (q
? 2 : 1); pass
++) {
5369 tmp
= neon_load_reg(rm
, pass
* 2);
5370 tmp2
= neon_load_reg(rm
, pass
* 2 + 1);
5372 case 0: tcg_gen_bswap32_i32(tmp
, tmp
); break;
5373 case 1: gen_swap_half(tmp
); break;
5374 case 2: /* no-op */ break;
5377 neon_store_reg(rd
, pass
* 2 + 1, tmp
);
5379 neon_store_reg(rd
, pass
* 2, tmp2
);
5382 case 0: tcg_gen_bswap32_i32(tmp2
, tmp2
); break;
5383 case 1: gen_swap_half(tmp2
); break;
5386 neon_store_reg(rd
, pass
* 2, tmp2
);
5390 case 4: case 5: /* VPADDL */
5391 case 12: case 13: /* VPADAL */
5394 for (pass
= 0; pass
< q
+ 1; pass
++) {
5395 tmp
= neon_load_reg(rm
, pass
* 2);
5396 gen_neon_widen(cpu_V0
, tmp
, size
, op
& 1);
5397 tmp
= neon_load_reg(rm
, pass
* 2 + 1);
5398 gen_neon_widen(cpu_V1
, tmp
, size
, op
& 1);
5400 case 0: gen_helper_neon_paddl_u16(CPU_V001
); break;
5401 case 1: gen_helper_neon_paddl_u32(CPU_V001
); break;
5402 case 2: tcg_gen_add_i64(CPU_V001
); break;
5407 neon_load_reg64(cpu_V1
, rd
+ pass
);
5408 gen_neon_addl(size
);
5410 neon_store_reg64(cpu_V0
, rd
+ pass
);
5415 for (n
= 0; n
< (q
? 4 : 2); n
+= 2) {
5416 tmp
= neon_load_reg(rm
, n
);
5417 tmp2
= neon_load_reg(rd
, n
+ 1);
5418 neon_store_reg(rm
, n
, tmp2
);
5419 neon_store_reg(rd
, n
+ 1, tmp
);
5426 if (gen_neon_unzip(rd
, rm
, size
, q
)) {
5431 if (gen_neon_zip(rd
, rm
, size
, q
)) {
5435 case 36: case 37: /* VMOVN, VQMOVUN, VQMOVN */
5439 for (pass
= 0; pass
< 2; pass
++) {
5440 neon_load_reg64(cpu_V0
, rm
+ pass
);
5443 if (q
) { /* VQMOVUN */
5444 gen_neon_unarrow_sats(size
, tmp
, cpu_V0
);
5445 } else { /* VMOVN */
5446 gen_neon_narrow(size
, tmp
, cpu_V0
);
5448 } else { /* VQMOVN */
5450 gen_neon_narrow_satu(size
, tmp
, cpu_V0
);
5452 gen_neon_narrow_sats(size
, tmp
, cpu_V0
);
5458 neon_store_reg(rd
, 0, tmp2
);
5459 neon_store_reg(rd
, 1, tmp
);
5463 case 38: /* VSHLL */
5466 tmp
= neon_load_reg(rm
, 0);
5467 tmp2
= neon_load_reg(rm
, 1);
5468 for (pass
= 0; pass
< 2; pass
++) {
5471 gen_neon_widen(cpu_V0
, tmp
, size
, 1);
5472 tcg_gen_shli_i64(cpu_V0
, cpu_V0
, 8 << size
);
5473 neon_store_reg64(cpu_V0
, rd
+ pass
);
5476 case 44: /* VCVT.F16.F32 */
5477 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
5481 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 0));
5482 gen_helper_neon_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
5483 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 1));
5484 gen_helper_neon_fcvt_f32_to_f16(tmp2
, cpu_F0s
, cpu_env
);
5485 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
5486 tcg_gen_or_i32(tmp2
, tmp2
, tmp
);
5487 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 2));
5488 gen_helper_neon_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
5489 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 3));
5490 neon_store_reg(rd
, 0, tmp2
);
5492 gen_helper_neon_fcvt_f32_to_f16(tmp2
, cpu_F0s
, cpu_env
);
5493 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
5494 tcg_gen_or_i32(tmp2
, tmp2
, tmp
);
5495 neon_store_reg(rd
, 1, tmp2
);
5498 case 46: /* VCVT.F32.F16 */
5499 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
5502 tmp
= neon_load_reg(rm
, 0);
5503 tmp2
= neon_load_reg(rm
, 1);
5504 tcg_gen_ext16u_i32(tmp3
, tmp
);
5505 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5506 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 0));
5507 tcg_gen_shri_i32(tmp3
, tmp
, 16);
5508 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5509 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 1));
5511 tcg_gen_ext16u_i32(tmp3
, tmp2
);
5512 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5513 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 2));
5514 tcg_gen_shri_i32(tmp3
, tmp2
, 16);
5515 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5516 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 3));
5522 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
5523 if (op
== 30 || op
== 31 || op
>= 58) {
5524 tcg_gen_ld_f32(cpu_F0s
, cpu_env
,
5525 neon_reg_offset(rm
, pass
));
5528 tmp
= neon_load_reg(rm
, pass
);
5531 case 1: /* VREV32 */
5533 case 0: tcg_gen_bswap32_i32(tmp
, tmp
); break;
5534 case 1: gen_swap_half(tmp
); break;
5538 case 2: /* VREV16 */
5545 case 0: gen_helper_neon_cls_s8(tmp
, tmp
); break;
5546 case 1: gen_helper_neon_cls_s16(tmp
, tmp
); break;
5547 case 2: gen_helper_neon_cls_s32(tmp
, tmp
); break;
5553 case 0: gen_helper_neon_clz_u8(tmp
, tmp
); break;
5554 case 1: gen_helper_neon_clz_u16(tmp
, tmp
); break;
5555 case 2: gen_helper_clz(tmp
, tmp
); break;
5562 gen_helper_neon_cnt_u8(tmp
, tmp
);
5567 tcg_gen_not_i32(tmp
, tmp
);
5569 case 14: /* VQABS */
5571 case 0: gen_helper_neon_qabs_s8(tmp
, cpu_env
, tmp
); break;
5572 case 1: gen_helper_neon_qabs_s16(tmp
, cpu_env
, tmp
); break;
5573 case 2: gen_helper_neon_qabs_s32(tmp
, cpu_env
, tmp
); break;
5577 case 15: /* VQNEG */
5579 case 0: gen_helper_neon_qneg_s8(tmp
, cpu_env
, tmp
); break;
5580 case 1: gen_helper_neon_qneg_s16(tmp
, cpu_env
, tmp
); break;
5581 case 2: gen_helper_neon_qneg_s32(tmp
, cpu_env
, tmp
); break;
5585 case 16: case 19: /* VCGT #0, VCLE #0 */
5586 tmp2
= tcg_const_i32(0);
5588 case 0: gen_helper_neon_cgt_s8(tmp
, tmp
, tmp2
); break;
5589 case 1: gen_helper_neon_cgt_s16(tmp
, tmp
, tmp2
); break;
5590 case 2: gen_helper_neon_cgt_s32(tmp
, tmp
, tmp2
); break;
5593 tcg_temp_free(tmp2
);
5595 tcg_gen_not_i32(tmp
, tmp
);
5597 case 17: case 20: /* VCGE #0, VCLT #0 */
5598 tmp2
= tcg_const_i32(0);
5600 case 0: gen_helper_neon_cge_s8(tmp
, tmp
, tmp2
); break;
5601 case 1: gen_helper_neon_cge_s16(tmp
, tmp
, tmp2
); break;
5602 case 2: gen_helper_neon_cge_s32(tmp
, tmp
, tmp2
); break;
5605 tcg_temp_free(tmp2
);
5607 tcg_gen_not_i32(tmp
, tmp
);
5609 case 18: /* VCEQ #0 */
5610 tmp2
= tcg_const_i32(0);
5612 case 0: gen_helper_neon_ceq_u8(tmp
, tmp
, tmp2
); break;
5613 case 1: gen_helper_neon_ceq_u16(tmp
, tmp
, tmp2
); break;
5614 case 2: gen_helper_neon_ceq_u32(tmp
, tmp
, tmp2
); break;
5617 tcg_temp_free(tmp2
);
5621 case 0: gen_helper_neon_abs_s8(tmp
, tmp
); break;
5622 case 1: gen_helper_neon_abs_s16(tmp
, tmp
); break;
5623 case 2: tcg_gen_abs_i32(tmp
, tmp
); break;
5630 tmp2
= tcg_const_i32(0);
5631 gen_neon_rsb(size
, tmp
, tmp2
);
5632 tcg_temp_free(tmp2
);
5634 case 24: case 27: /* Float VCGT #0, Float VCLE #0 */
5635 tmp2
= tcg_const_i32(0);
5636 gen_helper_neon_cgt_f32(tmp
, tmp
, tmp2
);
5637 tcg_temp_free(tmp2
);
5639 tcg_gen_not_i32(tmp
, tmp
);
5641 case 25: case 28: /* Float VCGE #0, Float VCLT #0 */
5642 tmp2
= tcg_const_i32(0);
5643 gen_helper_neon_cge_f32(tmp
, tmp
, tmp2
);
5644 tcg_temp_free(tmp2
);
5646 tcg_gen_not_i32(tmp
, tmp
);
5648 case 26: /* Float VCEQ #0 */
5649 tmp2
= tcg_const_i32(0);
5650 gen_helper_neon_ceq_f32(tmp
, tmp
, tmp2
);
5651 tcg_temp_free(tmp2
);
5653 case 30: /* Float VABS */
5656 case 31: /* Float VNEG */
5660 tmp2
= neon_load_reg(rd
, pass
);
5661 neon_store_reg(rm
, pass
, tmp2
);
5664 tmp2
= neon_load_reg(rd
, pass
);
5666 case 0: gen_neon_trn_u8(tmp
, tmp2
); break;
5667 case 1: gen_neon_trn_u16(tmp
, tmp2
); break;
5671 neon_store_reg(rm
, pass
, tmp2
);
5673 case 56: /* Integer VRECPE */
5674 gen_helper_recpe_u32(tmp
, tmp
, cpu_env
);
5676 case 57: /* Integer VRSQRTE */
5677 gen_helper_rsqrte_u32(tmp
, tmp
, cpu_env
);
5679 case 58: /* Float VRECPE */
5680 gen_helper_recpe_f32(cpu_F0s
, cpu_F0s
, cpu_env
);
5682 case 59: /* Float VRSQRTE */
5683 gen_helper_rsqrte_f32(cpu_F0s
, cpu_F0s
, cpu_env
);
5685 case 60: /* VCVT.F32.S32 */
5688 case 61: /* VCVT.F32.U32 */
5691 case 62: /* VCVT.S32.F32 */
5694 case 63: /* VCVT.U32.F32 */
5698 /* Reserved: 21, 29, 39-56 */
5701 if (op
== 30 || op
== 31 || op
>= 58) {
5702 tcg_gen_st_f32(cpu_F0s
, cpu_env
,
5703 neon_reg_offset(rd
, pass
));
5705 neon_store_reg(rd
, pass
, tmp
);
5710 } else if ((insn
& (1 << 10)) == 0) {
5712 n
= ((insn
>> 5) & 0x18) + 8;
5713 if (insn
& (1 << 6)) {
5714 tmp
= neon_load_reg(rd
, 0);
5717 tcg_gen_movi_i32(tmp
, 0);
5719 tmp2
= neon_load_reg(rm
, 0);
5720 tmp4
= tcg_const_i32(rn
);
5721 tmp5
= tcg_const_i32(n
);
5722 gen_helper_neon_tbl(tmp2
, tmp2
, tmp
, tmp4
, tmp5
);
5724 if (insn
& (1 << 6)) {
5725 tmp
= neon_load_reg(rd
, 1);
5728 tcg_gen_movi_i32(tmp
, 0);
5730 tmp3
= neon_load_reg(rm
, 1);
5731 gen_helper_neon_tbl(tmp3
, tmp3
, tmp
, tmp4
, tmp5
);
5732 tcg_temp_free_i32(tmp5
);
5733 tcg_temp_free_i32(tmp4
);
5734 neon_store_reg(rd
, 0, tmp2
);
5735 neon_store_reg(rd
, 1, tmp3
);
5737 } else if ((insn
& 0x380) == 0) {
5739 if (insn
& (1 << 19)) {
5740 tmp
= neon_load_reg(rm
, 1);
5742 tmp
= neon_load_reg(rm
, 0);
5744 if (insn
& (1 << 16)) {
5745 gen_neon_dup_u8(tmp
, ((insn
>> 17) & 3) * 8);
5746 } else if (insn
& (1 << 17)) {
5747 if ((insn
>> 18) & 1)
5748 gen_neon_dup_high16(tmp
);
5750 gen_neon_dup_low16(tmp
);
5752 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
5754 tcg_gen_mov_i32(tmp2
, tmp
);
5755 neon_store_reg(rd
, pass
, tmp2
);
5766 static int disas_cp14_read(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
5768 int crn
= (insn
>> 16) & 0xf;
5769 int crm
= insn
& 0xf;
5770 int op1
= (insn
>> 21) & 7;
5771 int op2
= (insn
>> 5) & 7;
5772 int rt
= (insn
>> 12) & 0xf;
5775 if (arm_feature(env
, ARM_FEATURE_THUMB2EE
)) {
5776 if (op1
== 6 && crn
== 0 && crm
== 0 && op2
== 0) {
5780 tmp
= load_cpu_field(teecr
);
5781 store_reg(s
, rt
, tmp
);
5784 if (op1
== 6 && crn
== 1 && crm
== 0 && op2
== 0) {
5786 if (IS_USER(s
) && (env
->teecr
& 1))
5788 tmp
= load_cpu_field(teehbr
);
5789 store_reg(s
, rt
, tmp
);
5793 fprintf(stderr
, "Unknown cp14 read op1:%d crn:%d crm:%d op2:%d\n",
5794 op1
, crn
, crm
, op2
);
5798 static int disas_cp14_write(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
5800 int crn
= (insn
>> 16) & 0xf;
5801 int crm
= insn
& 0xf;
5802 int op1
= (insn
>> 21) & 7;
5803 int op2
= (insn
>> 5) & 7;
5804 int rt
= (insn
>> 12) & 0xf;
5807 if (arm_feature(env
, ARM_FEATURE_THUMB2EE
)) {
5808 if (op1
== 6 && crn
== 0 && crm
== 0 && op2
== 0) {
5812 tmp
= load_reg(s
, rt
);
5813 gen_helper_set_teecr(cpu_env
, tmp
);
5817 if (op1
== 6 && crn
== 1 && crm
== 0 && op2
== 0) {
5819 if (IS_USER(s
) && (env
->teecr
& 1))
5821 tmp
= load_reg(s
, rt
);
5822 store_cpu_field(tmp
, teehbr
);
5826 fprintf(stderr
, "Unknown cp14 write op1:%d crn:%d crm:%d op2:%d\n",
5827 op1
, crn
, crm
, op2
);
5831 static int disas_coproc_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
5835 cpnum
= (insn
>> 8) & 0xf;
5836 if (arm_feature(env
, ARM_FEATURE_XSCALE
)
5837 && ((env
->cp15
.c15_cpar
^ 0x3fff) & (1 << cpnum
)))
5843 if (arm_feature(env
, ARM_FEATURE_IWMMXT
)) {
5844 return disas_iwmmxt_insn(env
, s
, insn
);
5845 } else if (arm_feature(env
, ARM_FEATURE_XSCALE
)) {
5846 return disas_dsp_insn(env
, s
, insn
);
5851 return disas_vfp_insn (env
, s
, insn
);
5853 /* Coprocessors 7-15 are architecturally reserved by ARM.
5854 Unfortunately Intel decided to ignore this. */
5855 if (arm_feature(env
, ARM_FEATURE_XSCALE
))
5857 if (insn
& (1 << 20))
5858 return disas_cp14_read(env
, s
, insn
);
5860 return disas_cp14_write(env
, s
, insn
);
5862 return disas_cp15_insn (env
, s
, insn
);
5865 /* Unknown coprocessor. See if the board has hooked it. */
5866 return disas_cp_insn (env
, s
, insn
);
5871 /* Store a 64-bit value to a register pair. Clobbers val. */
5872 static void gen_storeq_reg(DisasContext
*s
, int rlow
, int rhigh
, TCGv_i64 val
)
5876 tcg_gen_trunc_i64_i32(tmp
, val
);
5877 store_reg(s
, rlow
, tmp
);
5879 tcg_gen_shri_i64(val
, val
, 32);
5880 tcg_gen_trunc_i64_i32(tmp
, val
);
5881 store_reg(s
, rhigh
, tmp
);
5884 /* load a 32-bit value from a register and perform a 64-bit accumulate. */
5885 static void gen_addq_lo(DisasContext
*s
, TCGv_i64 val
, int rlow
)
5890 /* Load value and extend to 64 bits. */
5891 tmp
= tcg_temp_new_i64();
5892 tmp2
= load_reg(s
, rlow
);
5893 tcg_gen_extu_i32_i64(tmp
, tmp2
);
5895 tcg_gen_add_i64(val
, val
, tmp
);
5896 tcg_temp_free_i64(tmp
);
5899 /* load and add a 64-bit value from a register pair. */
5900 static void gen_addq(DisasContext
*s
, TCGv_i64 val
, int rlow
, int rhigh
)
5906 /* Load 64-bit value rd:rn. */
5907 tmpl
= load_reg(s
, rlow
);
5908 tmph
= load_reg(s
, rhigh
);
5909 tmp
= tcg_temp_new_i64();
5910 tcg_gen_concat_i32_i64(tmp
, tmpl
, tmph
);
5913 tcg_gen_add_i64(val
, val
, tmp
);
5914 tcg_temp_free_i64(tmp
);
5917 /* Set N and Z flags from a 64-bit value. */
5918 static void gen_logicq_cc(TCGv_i64 val
)
5920 TCGv tmp
= new_tmp();
5921 gen_helper_logicq_cc(tmp
, val
);
5926 /* Load/Store exclusive instructions are implemented by remembering
5927 the value/address loaded, and seeing if these are the same
5928 when the store is performed. This should be is sufficient to implement
5929 the architecturally mandated semantics, and avoids having to monitor
5932 In system emulation mode only one CPU will be running at once, so
5933 this sequence is effectively atomic. In user emulation mode we
5934 throw an exception and handle the atomic operation elsewhere. */
5935 static void gen_load_exclusive(DisasContext
*s
, int rt
, int rt2
,
5936 TCGv addr
, int size
)
5942 tmp
= gen_ld8u(addr
, IS_USER(s
));
5945 tmp
= gen_ld16u(addr
, IS_USER(s
));
5949 tmp
= gen_ld32(addr
, IS_USER(s
));
5954 tcg_gen_mov_i32(cpu_exclusive_val
, tmp
);
5955 store_reg(s
, rt
, tmp
);
5957 TCGv tmp2
= new_tmp();
5958 tcg_gen_addi_i32(tmp2
, addr
, 4);
5959 tmp
= gen_ld32(tmp2
, IS_USER(s
));
5961 tcg_gen_mov_i32(cpu_exclusive_high
, tmp
);
5962 store_reg(s
, rt2
, tmp
);
5964 tcg_gen_mov_i32(cpu_exclusive_addr
, addr
);
5967 static void gen_clrex(DisasContext
*s
)
5969 tcg_gen_movi_i32(cpu_exclusive_addr
, -1);
5972 #ifdef CONFIG_USER_ONLY
5973 static void gen_store_exclusive(DisasContext
*s
, int rd
, int rt
, int rt2
,
5974 TCGv addr
, int size
)
5976 tcg_gen_mov_i32(cpu_exclusive_test
, addr
);
5977 tcg_gen_movi_i32(cpu_exclusive_info
,
5978 size
| (rd
<< 4) | (rt
<< 8) | (rt2
<< 12));
5979 gen_exception_insn(s
, 4, EXCP_STREX
);
5982 static void gen_store_exclusive(DisasContext
*s
, int rd
, int rt
, int rt2
,
5983 TCGv addr
, int size
)
5989 /* if (env->exclusive_addr == addr && env->exclusive_val == [addr]) {
5995 fail_label
= gen_new_label();
5996 done_label
= gen_new_label();
5997 tcg_gen_brcond_i32(TCG_COND_NE
, addr
, cpu_exclusive_addr
, fail_label
);
6000 tmp
= gen_ld8u(addr
, IS_USER(s
));
6003 tmp
= gen_ld16u(addr
, IS_USER(s
));
6007 tmp
= gen_ld32(addr
, IS_USER(s
));
6012 tcg_gen_brcond_i32(TCG_COND_NE
, tmp
, cpu_exclusive_val
, fail_label
);
6015 TCGv tmp2
= new_tmp();
6016 tcg_gen_addi_i32(tmp2
, addr
, 4);
6017 tmp
= gen_ld32(tmp2
, IS_USER(s
));
6019 tcg_gen_brcond_i32(TCG_COND_NE
, tmp
, cpu_exclusive_high
, fail_label
);
6022 tmp
= load_reg(s
, rt
);
6025 gen_st8(tmp
, addr
, IS_USER(s
));
6028 gen_st16(tmp
, addr
, IS_USER(s
));
6032 gen_st32(tmp
, addr
, IS_USER(s
));
6038 tcg_gen_addi_i32(addr
, addr
, 4);
6039 tmp
= load_reg(s
, rt2
);
6040 gen_st32(tmp
, addr
, IS_USER(s
));
6042 tcg_gen_movi_i32(cpu_R
[rd
], 0);
6043 tcg_gen_br(done_label
);
6044 gen_set_label(fail_label
);
6045 tcg_gen_movi_i32(cpu_R
[rd
], 1);
6046 gen_set_label(done_label
);
6047 tcg_gen_movi_i32(cpu_exclusive_addr
, -1);
6051 static void disas_arm_insn(CPUState
* env
, DisasContext
*s
)
6053 unsigned int cond
, insn
, val
, op1
, i
, shift
, rm
, rs
, rn
, rd
, sh
;
6060 insn
= ldl_code(s
->pc
);
6063 /* M variants do not implement ARM mode. */
6068 /* Unconditional instructions. */
6069 if (((insn
>> 25) & 7) == 1) {
6070 /* NEON Data processing. */
6071 if (!arm_feature(env
, ARM_FEATURE_NEON
))
6074 if (disas_neon_data_insn(env
, s
, insn
))
6078 if ((insn
& 0x0f100000) == 0x04000000) {
6079 /* NEON load/store. */
6080 if (!arm_feature(env
, ARM_FEATURE_NEON
))
6083 if (disas_neon_ls_insn(env
, s
, insn
))
6087 if (((insn
& 0x0f30f000) == 0x0510f000) ||
6088 ((insn
& 0x0f30f010) == 0x0710f000)) {
6089 if ((insn
& (1 << 22)) == 0) {
6091 if (!arm_feature(env
, ARM_FEATURE_V7MP
)) {
6095 /* Otherwise PLD; v5TE+ */
6098 if (((insn
& 0x0f70f000) == 0x0450f000) ||
6099 ((insn
& 0x0f70f010) == 0x0650f000)) {
6101 return; /* PLI; V7 */
6103 if (((insn
& 0x0f700000) == 0x04100000) ||
6104 ((insn
& 0x0f700010) == 0x06100000)) {
6105 if (!arm_feature(env
, ARM_FEATURE_V7MP
)) {
6108 return; /* v7MP: Unallocated memory hint: must NOP */
6111 if ((insn
& 0x0ffffdff) == 0x01010000) {
6114 if (insn
& (1 << 9)) {
6115 /* BE8 mode not implemented. */
6119 } else if ((insn
& 0x0fffff00) == 0x057ff000) {
6120 switch ((insn
>> 4) & 0xf) {
6129 /* We don't emulate caches so these are a no-op. */
6134 } else if ((insn
& 0x0e5fffe0) == 0x084d0500) {
6140 op1
= (insn
& 0x1f);
6142 tmp
= tcg_const_i32(op1
);
6143 gen_helper_get_r13_banked(addr
, cpu_env
, tmp
);
6144 tcg_temp_free_i32(tmp
);
6145 i
= (insn
>> 23) & 3;
6147 case 0: offset
= -4; break; /* DA */
6148 case 1: offset
= 0; break; /* IA */
6149 case 2: offset
= -8; break; /* DB */
6150 case 3: offset
= 4; break; /* IB */
6154 tcg_gen_addi_i32(addr
, addr
, offset
);
6155 tmp
= load_reg(s
, 14);
6156 gen_st32(tmp
, addr
, 0);
6157 tmp
= load_cpu_field(spsr
);
6158 tcg_gen_addi_i32(addr
, addr
, 4);
6159 gen_st32(tmp
, addr
, 0);
6160 if (insn
& (1 << 21)) {
6161 /* Base writeback. */
6163 case 0: offset
= -8; break;
6164 case 1: offset
= 4; break;
6165 case 2: offset
= -4; break;
6166 case 3: offset
= 0; break;
6170 tcg_gen_addi_i32(addr
, addr
, offset
);
6171 tmp
= tcg_const_i32(op1
);
6172 gen_helper_set_r13_banked(cpu_env
, tmp
, addr
);
6173 tcg_temp_free_i32(tmp
);
6179 } else if ((insn
& 0x0e50ffe0) == 0x08100a00) {
6185 rn
= (insn
>> 16) & 0xf;
6186 addr
= load_reg(s
, rn
);
6187 i
= (insn
>> 23) & 3;
6189 case 0: offset
= -4; break; /* DA */
6190 case 1: offset
= 0; break; /* IA */
6191 case 2: offset
= -8; break; /* DB */
6192 case 3: offset
= 4; break; /* IB */
6196 tcg_gen_addi_i32(addr
, addr
, offset
);
6197 /* Load PC into tmp and CPSR into tmp2. */
6198 tmp
= gen_ld32(addr
, 0);
6199 tcg_gen_addi_i32(addr
, addr
, 4);
6200 tmp2
= gen_ld32(addr
, 0);
6201 if (insn
& (1 << 21)) {
6202 /* Base writeback. */
6204 case 0: offset
= -8; break;
6205 case 1: offset
= 4; break;
6206 case 2: offset
= -4; break;
6207 case 3: offset
= 0; break;
6211 tcg_gen_addi_i32(addr
, addr
, offset
);
6212 store_reg(s
, rn
, addr
);
6216 gen_rfe(s
, tmp
, tmp2
);
6218 } else if ((insn
& 0x0e000000) == 0x0a000000) {
6219 /* branch link and change to thumb (blx <offset>) */
6222 val
= (uint32_t)s
->pc
;
6224 tcg_gen_movi_i32(tmp
, val
);
6225 store_reg(s
, 14, tmp
);
6226 /* Sign-extend the 24-bit offset */
6227 offset
= (((int32_t)insn
) << 8) >> 8;
6228 /* offset * 4 + bit24 * 2 + (thumb bit) */
6229 val
+= (offset
<< 2) | ((insn
>> 23) & 2) | 1;
6230 /* pipeline offset */
6234 } else if ((insn
& 0x0e000f00) == 0x0c000100) {
6235 if (arm_feature(env
, ARM_FEATURE_IWMMXT
)) {
6236 /* iWMMXt register transfer. */
6237 if (env
->cp15
.c15_cpar
& (1 << 1))
6238 if (!disas_iwmmxt_insn(env
, s
, insn
))
6241 } else if ((insn
& 0x0fe00000) == 0x0c400000) {
6242 /* Coprocessor double register transfer. */
6243 } else if ((insn
& 0x0f000010) == 0x0e000010) {
6244 /* Additional coprocessor register transfer. */
6245 } else if ((insn
& 0x0ff10020) == 0x01000000) {
6248 /* cps (privileged) */
6252 if (insn
& (1 << 19)) {
6253 if (insn
& (1 << 8))
6255 if (insn
& (1 << 7))
6257 if (insn
& (1 << 6))
6259 if (insn
& (1 << 18))
6262 if (insn
& (1 << 17)) {
6264 val
|= (insn
& 0x1f);
6267 gen_set_psr_im(s
, mask
, 0, val
);
6274 /* if not always execute, we generate a conditional jump to
6276 s
->condlabel
= gen_new_label();
6277 gen_test_cc(cond
^ 1, s
->condlabel
);
6280 if ((insn
& 0x0f900000) == 0x03000000) {
6281 if ((insn
& (1 << 21)) == 0) {
6283 rd
= (insn
>> 12) & 0xf;
6284 val
= ((insn
>> 4) & 0xf000) | (insn
& 0xfff);
6285 if ((insn
& (1 << 22)) == 0) {
6288 tcg_gen_movi_i32(tmp
, val
);
6291 tmp
= load_reg(s
, rd
);
6292 tcg_gen_ext16u_i32(tmp
, tmp
);
6293 tcg_gen_ori_i32(tmp
, tmp
, val
<< 16);
6295 store_reg(s
, rd
, tmp
);
6297 if (((insn
>> 12) & 0xf) != 0xf)
6299 if (((insn
>> 16) & 0xf) == 0) {
6300 gen_nop_hint(s
, insn
& 0xff);
6302 /* CPSR = immediate */
6304 shift
= ((insn
>> 8) & 0xf) * 2;
6306 val
= (val
>> shift
) | (val
<< (32 - shift
));
6307 i
= ((insn
& (1 << 22)) != 0);
6308 if (gen_set_psr_im(s
, msr_mask(env
, s
, (insn
>> 16) & 0xf, i
), i
, val
))
6312 } else if ((insn
& 0x0f900000) == 0x01000000
6313 && (insn
& 0x00000090) != 0x00000090) {
6314 /* miscellaneous instructions */
6315 op1
= (insn
>> 21) & 3;
6316 sh
= (insn
>> 4) & 0xf;
6319 case 0x0: /* move program status register */
6322 tmp
= load_reg(s
, rm
);
6323 i
= ((op1
& 2) != 0);
6324 if (gen_set_psr(s
, msr_mask(env
, s
, (insn
>> 16) & 0xf, i
), i
, tmp
))
6328 rd
= (insn
>> 12) & 0xf;
6332 tmp
= load_cpu_field(spsr
);
6335 gen_helper_cpsr_read(tmp
);
6337 store_reg(s
, rd
, tmp
);
6342 /* branch/exchange thumb (bx). */
6343 tmp
= load_reg(s
, rm
);
6345 } else if (op1
== 3) {
6347 rd
= (insn
>> 12) & 0xf;
6348 tmp
= load_reg(s
, rm
);
6349 gen_helper_clz(tmp
, tmp
);
6350 store_reg(s
, rd
, tmp
);
6358 /* Trivial implementation equivalent to bx. */
6359 tmp
= load_reg(s
, rm
);
6369 /* branch link/exchange thumb (blx) */
6370 tmp
= load_reg(s
, rm
);
6372 tcg_gen_movi_i32(tmp2
, s
->pc
);
6373 store_reg(s
, 14, tmp2
);
6376 case 0x5: /* saturating add/subtract */
6377 rd
= (insn
>> 12) & 0xf;
6378 rn
= (insn
>> 16) & 0xf;
6379 tmp
= load_reg(s
, rm
);
6380 tmp2
= load_reg(s
, rn
);
6382 gen_helper_double_saturate(tmp2
, tmp2
);
6384 gen_helper_sub_saturate(tmp
, tmp
, tmp2
);
6386 gen_helper_add_saturate(tmp
, tmp
, tmp2
);
6388 store_reg(s
, rd
, tmp
);
6391 /* SMC instruction (op1 == 3)
6392 and undefined instructions (op1 == 0 || op1 == 2)
6398 gen_exception_insn(s
, 4, EXCP_BKPT
);
6400 case 0x8: /* signed multiply */
6404 rs
= (insn
>> 8) & 0xf;
6405 rn
= (insn
>> 12) & 0xf;
6406 rd
= (insn
>> 16) & 0xf;
6408 /* (32 * 16) >> 16 */
6409 tmp
= load_reg(s
, rm
);
6410 tmp2
= load_reg(s
, rs
);
6412 tcg_gen_sari_i32(tmp2
, tmp2
, 16);
6415 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
6416 tcg_gen_shri_i64(tmp64
, tmp64
, 16);
6418 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
6419 tcg_temp_free_i64(tmp64
);
6420 if ((sh
& 2) == 0) {
6421 tmp2
= load_reg(s
, rn
);
6422 gen_helper_add_setq(tmp
, tmp
, tmp2
);
6425 store_reg(s
, rd
, tmp
);
6428 tmp
= load_reg(s
, rm
);
6429 tmp2
= load_reg(s
, rs
);
6430 gen_mulxy(tmp
, tmp2
, sh
& 2, sh
& 4);
6433 tmp64
= tcg_temp_new_i64();
6434 tcg_gen_ext_i32_i64(tmp64
, tmp
);
6436 gen_addq(s
, tmp64
, rn
, rd
);
6437 gen_storeq_reg(s
, rn
, rd
, tmp64
);
6438 tcg_temp_free_i64(tmp64
);
6441 tmp2
= load_reg(s
, rn
);
6442 gen_helper_add_setq(tmp
, tmp
, tmp2
);
6445 store_reg(s
, rd
, tmp
);
6452 } else if (((insn
& 0x0e000000) == 0 &&
6453 (insn
& 0x00000090) != 0x90) ||
6454 ((insn
& 0x0e000000) == (1 << 25))) {
6455 int set_cc
, logic_cc
, shiftop
;
6457 op1
= (insn
>> 21) & 0xf;
6458 set_cc
= (insn
>> 20) & 1;
6459 logic_cc
= table_logic_cc
[op1
] & set_cc
;
6461 /* data processing instruction */
6462 if (insn
& (1 << 25)) {
6463 /* immediate operand */
6465 shift
= ((insn
>> 8) & 0xf) * 2;
6467 val
= (val
>> shift
) | (val
<< (32 - shift
));
6470 tcg_gen_movi_i32(tmp2
, val
);
6471 if (logic_cc
&& shift
) {
6472 gen_set_CF_bit31(tmp2
);
6477 tmp2
= load_reg(s
, rm
);
6478 shiftop
= (insn
>> 5) & 3;
6479 if (!(insn
& (1 << 4))) {
6480 shift
= (insn
>> 7) & 0x1f;
6481 gen_arm_shift_im(tmp2
, shiftop
, shift
, logic_cc
);
6483 rs
= (insn
>> 8) & 0xf;
6484 tmp
= load_reg(s
, rs
);
6485 gen_arm_shift_reg(tmp2
, shiftop
, tmp
, logic_cc
);
6488 if (op1
!= 0x0f && op1
!= 0x0d) {
6489 rn
= (insn
>> 16) & 0xf;
6490 tmp
= load_reg(s
, rn
);
6494 rd
= (insn
>> 12) & 0xf;
6497 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
6501 store_reg_bx(env
, s
, rd
, tmp
);
6504 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
6508 store_reg_bx(env
, s
, rd
, tmp
);
6511 if (set_cc
&& rd
== 15) {
6512 /* SUBS r15, ... is used for exception return. */
6516 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
6517 gen_exception_return(s
, tmp
);
6520 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
6522 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
6524 store_reg_bx(env
, s
, rd
, tmp
);
6529 gen_helper_sub_cc(tmp
, tmp2
, tmp
);
6531 tcg_gen_sub_i32(tmp
, tmp2
, tmp
);
6533 store_reg_bx(env
, s
, rd
, tmp
);
6537 gen_helper_add_cc(tmp
, tmp
, tmp2
);
6539 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
6541 store_reg_bx(env
, s
, rd
, tmp
);
6545 gen_helper_adc_cc(tmp
, tmp
, tmp2
);
6547 gen_add_carry(tmp
, tmp
, tmp2
);
6549 store_reg_bx(env
, s
, rd
, tmp
);
6553 gen_helper_sbc_cc(tmp
, tmp
, tmp2
);
6555 gen_sub_carry(tmp
, tmp
, tmp2
);
6557 store_reg_bx(env
, s
, rd
, tmp
);
6561 gen_helper_sbc_cc(tmp
, tmp2
, tmp
);
6563 gen_sub_carry(tmp
, tmp2
, tmp
);
6565 store_reg_bx(env
, s
, rd
, tmp
);
6569 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
6576 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
6583 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
6589 gen_helper_add_cc(tmp
, tmp
, tmp2
);
6594 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
6598 store_reg_bx(env
, s
, rd
, tmp
);
6601 if (logic_cc
&& rd
== 15) {
6602 /* MOVS r15, ... is used for exception return. */
6606 gen_exception_return(s
, tmp2
);
6611 store_reg_bx(env
, s
, rd
, tmp2
);
6615 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
6619 store_reg_bx(env
, s
, rd
, tmp
);
6623 tcg_gen_not_i32(tmp2
, tmp2
);
6627 store_reg_bx(env
, s
, rd
, tmp2
);
6630 if (op1
!= 0x0f && op1
!= 0x0d) {
6634 /* other instructions */
6635 op1
= (insn
>> 24) & 0xf;
6639 /* multiplies, extra load/stores */
6640 sh
= (insn
>> 5) & 3;
6643 rd
= (insn
>> 16) & 0xf;
6644 rn
= (insn
>> 12) & 0xf;
6645 rs
= (insn
>> 8) & 0xf;
6647 op1
= (insn
>> 20) & 0xf;
6649 case 0: case 1: case 2: case 3: case 6:
6651 tmp
= load_reg(s
, rs
);
6652 tmp2
= load_reg(s
, rm
);
6653 tcg_gen_mul_i32(tmp
, tmp
, tmp2
);
6655 if (insn
& (1 << 22)) {
6656 /* Subtract (mls) */
6658 tmp2
= load_reg(s
, rn
);
6659 tcg_gen_sub_i32(tmp
, tmp2
, tmp
);
6661 } else if (insn
& (1 << 21)) {
6663 tmp2
= load_reg(s
, rn
);
6664 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
6667 if (insn
& (1 << 20))
6669 store_reg(s
, rd
, tmp
);
6672 /* 64 bit mul double accumulate (UMAAL) */
6674 tmp
= load_reg(s
, rs
);
6675 tmp2
= load_reg(s
, rm
);
6676 tmp64
= gen_mulu_i64_i32(tmp
, tmp2
);
6677 gen_addq_lo(s
, tmp64
, rn
);
6678 gen_addq_lo(s
, tmp64
, rd
);
6679 gen_storeq_reg(s
, rn
, rd
, tmp64
);
6680 tcg_temp_free_i64(tmp64
);
6682 case 8: case 9: case 10: case 11:
6683 case 12: case 13: case 14: case 15:
6684 /* 64 bit mul: UMULL, UMLAL, SMULL, SMLAL. */
6685 tmp
= load_reg(s
, rs
);
6686 tmp2
= load_reg(s
, rm
);
6687 if (insn
& (1 << 22)) {
6688 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
6690 tmp64
= gen_mulu_i64_i32(tmp
, tmp2
);
6692 if (insn
& (1 << 21)) { /* mult accumulate */
6693 gen_addq(s
, tmp64
, rn
, rd
);
6695 if (insn
& (1 << 20)) {
6696 gen_logicq_cc(tmp64
);
6698 gen_storeq_reg(s
, rn
, rd
, tmp64
);
6699 tcg_temp_free_i64(tmp64
);
6705 rn
= (insn
>> 16) & 0xf;
6706 rd
= (insn
>> 12) & 0xf;
6707 if (insn
& (1 << 23)) {
6708 /* load/store exclusive */
6709 op1
= (insn
>> 21) & 0x3;
6714 addr
= tcg_temp_local_new_i32();
6715 load_reg_var(s
, addr
, rn
);
6716 if (insn
& (1 << 20)) {
6719 gen_load_exclusive(s
, rd
, 15, addr
, 2);
6721 case 1: /* ldrexd */
6722 gen_load_exclusive(s
, rd
, rd
+ 1, addr
, 3);
6724 case 2: /* ldrexb */
6725 gen_load_exclusive(s
, rd
, 15, addr
, 0);
6727 case 3: /* ldrexh */
6728 gen_load_exclusive(s
, rd
, 15, addr
, 1);
6737 gen_store_exclusive(s
, rd
, rm
, 15, addr
, 2);
6739 case 1: /* strexd */
6740 gen_store_exclusive(s
, rd
, rm
, rm
+ 1, addr
, 3);
6742 case 2: /* strexb */
6743 gen_store_exclusive(s
, rd
, rm
, 15, addr
, 0);
6745 case 3: /* strexh */
6746 gen_store_exclusive(s
, rd
, rm
, 15, addr
, 1);
6752 tcg_temp_free(addr
);
6754 /* SWP instruction */
6757 /* ??? This is not really atomic. However we know
6758 we never have multiple CPUs running in parallel,
6759 so it is good enough. */
6760 addr
= load_reg(s
, rn
);
6761 tmp
= load_reg(s
, rm
);
6762 if (insn
& (1 << 22)) {
6763 tmp2
= gen_ld8u(addr
, IS_USER(s
));
6764 gen_st8(tmp
, addr
, IS_USER(s
));
6766 tmp2
= gen_ld32(addr
, IS_USER(s
));
6767 gen_st32(tmp
, addr
, IS_USER(s
));
6770 store_reg(s
, rd
, tmp2
);
6776 /* Misc load/store */
6777 rn
= (insn
>> 16) & 0xf;
6778 rd
= (insn
>> 12) & 0xf;
6779 addr
= load_reg(s
, rn
);
6780 if (insn
& (1 << 24))
6781 gen_add_datah_offset(s
, insn
, 0, addr
);
6783 if (insn
& (1 << 20)) {
6787 tmp
= gen_ld16u(addr
, IS_USER(s
));
6790 tmp
= gen_ld8s(addr
, IS_USER(s
));
6794 tmp
= gen_ld16s(addr
, IS_USER(s
));
6798 } else if (sh
& 2) {
6802 tmp
= load_reg(s
, rd
);
6803 gen_st32(tmp
, addr
, IS_USER(s
));
6804 tcg_gen_addi_i32(addr
, addr
, 4);
6805 tmp
= load_reg(s
, rd
+ 1);
6806 gen_st32(tmp
, addr
, IS_USER(s
));
6810 tmp
= gen_ld32(addr
, IS_USER(s
));
6811 store_reg(s
, rd
, tmp
);
6812 tcg_gen_addi_i32(addr
, addr
, 4);
6813 tmp
= gen_ld32(addr
, IS_USER(s
));
6817 address_offset
= -4;
6820 tmp
= load_reg(s
, rd
);
6821 gen_st16(tmp
, addr
, IS_USER(s
));
6824 /* Perform base writeback before the loaded value to
6825 ensure correct behavior with overlapping index registers.
6826 ldrd with base writeback is is undefined if the
6827 destination and index registers overlap. */
6828 if (!(insn
& (1 << 24))) {
6829 gen_add_datah_offset(s
, insn
, address_offset
, addr
);
6830 store_reg(s
, rn
, addr
);
6831 } else if (insn
& (1 << 21)) {
6833 tcg_gen_addi_i32(addr
, addr
, address_offset
);
6834 store_reg(s
, rn
, addr
);
6839 /* Complete the load. */
6840 store_reg(s
, rd
, tmp
);
6849 if (insn
& (1 << 4)) {
6851 /* Armv6 Media instructions. */
6853 rn
= (insn
>> 16) & 0xf;
6854 rd
= (insn
>> 12) & 0xf;
6855 rs
= (insn
>> 8) & 0xf;
6856 switch ((insn
>> 23) & 3) {
6857 case 0: /* Parallel add/subtract. */
6858 op1
= (insn
>> 20) & 7;
6859 tmp
= load_reg(s
, rn
);
6860 tmp2
= load_reg(s
, rm
);
6861 sh
= (insn
>> 5) & 7;
6862 if ((op1
& 3) == 0 || sh
== 5 || sh
== 6)
6864 gen_arm_parallel_addsub(op1
, sh
, tmp
, tmp2
);
6866 store_reg(s
, rd
, tmp
);
6869 if ((insn
& 0x00700020) == 0) {
6870 /* Halfword pack. */
6871 tmp
= load_reg(s
, rn
);
6872 tmp2
= load_reg(s
, rm
);
6873 shift
= (insn
>> 7) & 0x1f;
6874 if (insn
& (1 << 6)) {
6878 tcg_gen_sari_i32(tmp2
, tmp2
, shift
);
6879 tcg_gen_andi_i32(tmp
, tmp
, 0xffff0000);
6880 tcg_gen_ext16u_i32(tmp2
, tmp2
);
6884 tcg_gen_shli_i32(tmp2
, tmp2
, shift
);
6885 tcg_gen_ext16u_i32(tmp
, tmp
);
6886 tcg_gen_andi_i32(tmp2
, tmp2
, 0xffff0000);
6888 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
6890 store_reg(s
, rd
, tmp
);
6891 } else if ((insn
& 0x00200020) == 0x00200000) {
6893 tmp
= load_reg(s
, rm
);
6894 shift
= (insn
>> 7) & 0x1f;
6895 if (insn
& (1 << 6)) {
6898 tcg_gen_sari_i32(tmp
, tmp
, shift
);
6900 tcg_gen_shli_i32(tmp
, tmp
, shift
);
6902 sh
= (insn
>> 16) & 0x1f;
6903 tmp2
= tcg_const_i32(sh
);
6904 if (insn
& (1 << 22))
6905 gen_helper_usat(tmp
, tmp
, tmp2
);
6907 gen_helper_ssat(tmp
, tmp
, tmp2
);
6908 tcg_temp_free_i32(tmp2
);
6909 store_reg(s
, rd
, tmp
);
6910 } else if ((insn
& 0x00300fe0) == 0x00200f20) {
6912 tmp
= load_reg(s
, rm
);
6913 sh
= (insn
>> 16) & 0x1f;
6914 tmp2
= tcg_const_i32(sh
);
6915 if (insn
& (1 << 22))
6916 gen_helper_usat16(tmp
, tmp
, tmp2
);
6918 gen_helper_ssat16(tmp
, tmp
, tmp2
);
6919 tcg_temp_free_i32(tmp2
);
6920 store_reg(s
, rd
, tmp
);
6921 } else if ((insn
& 0x00700fe0) == 0x00000fa0) {
6923 tmp
= load_reg(s
, rn
);
6924 tmp2
= load_reg(s
, rm
);
6926 tcg_gen_ld_i32(tmp3
, cpu_env
, offsetof(CPUState
, GE
));
6927 gen_helper_sel_flags(tmp
, tmp3
, tmp
, tmp2
);
6930 store_reg(s
, rd
, tmp
);
6931 } else if ((insn
& 0x000003e0) == 0x00000060) {
6932 tmp
= load_reg(s
, rm
);
6933 shift
= (insn
>> 10) & 3;
6934 /* ??? In many cases it's not neccessary to do a
6935 rotate, a shift is sufficient. */
6937 tcg_gen_rotri_i32(tmp
, tmp
, shift
* 8);
6938 op1
= (insn
>> 20) & 7;
6940 case 0: gen_sxtb16(tmp
); break;
6941 case 2: gen_sxtb(tmp
); break;
6942 case 3: gen_sxth(tmp
); break;
6943 case 4: gen_uxtb16(tmp
); break;
6944 case 6: gen_uxtb(tmp
); break;
6945 case 7: gen_uxth(tmp
); break;
6946 default: goto illegal_op
;
6949 tmp2
= load_reg(s
, rn
);
6950 if ((op1
& 3) == 0) {
6951 gen_add16(tmp
, tmp2
);
6953 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
6957 store_reg(s
, rd
, tmp
);
6958 } else if ((insn
& 0x003f0f60) == 0x003f0f20) {
6960 tmp
= load_reg(s
, rm
);
6961 if (insn
& (1 << 22)) {
6962 if (insn
& (1 << 7)) {
6966 gen_helper_rbit(tmp
, tmp
);
6969 if (insn
& (1 << 7))
6972 tcg_gen_bswap32_i32(tmp
, tmp
);
6974 store_reg(s
, rd
, tmp
);
6979 case 2: /* Multiplies (Type 3). */
6980 tmp
= load_reg(s
, rm
);
6981 tmp2
= load_reg(s
, rs
);
6982 if (insn
& (1 << 20)) {
6983 /* Signed multiply most significant [accumulate].
6984 (SMMUL, SMMLA, SMMLS) */
6985 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
6988 tmp
= load_reg(s
, rd
);
6989 if (insn
& (1 << 6)) {
6990 tmp64
= gen_subq_msw(tmp64
, tmp
);
6992 tmp64
= gen_addq_msw(tmp64
, tmp
);
6995 if (insn
& (1 << 5)) {
6996 tcg_gen_addi_i64(tmp64
, tmp64
, 0x80000000u
);
6998 tcg_gen_shri_i64(tmp64
, tmp64
, 32);
7000 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
7001 tcg_temp_free_i64(tmp64
);
7002 store_reg(s
, rn
, tmp
);
7004 if (insn
& (1 << 5))
7005 gen_swap_half(tmp2
);
7006 gen_smul_dual(tmp
, tmp2
);
7007 /* This addition cannot overflow. */
7008 if (insn
& (1 << 6)) {
7009 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
7011 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7014 if (insn
& (1 << 22)) {
7015 /* smlald, smlsld */
7016 tmp64
= tcg_temp_new_i64();
7017 tcg_gen_ext_i32_i64(tmp64
, tmp
);
7019 gen_addq(s
, tmp64
, rd
, rn
);
7020 gen_storeq_reg(s
, rd
, rn
, tmp64
);
7021 tcg_temp_free_i64(tmp64
);
7023 /* smuad, smusd, smlad, smlsd */
7026 tmp2
= load_reg(s
, rd
);
7027 gen_helper_add_setq(tmp
, tmp
, tmp2
);
7030 store_reg(s
, rn
, tmp
);
7035 op1
= ((insn
>> 17) & 0x38) | ((insn
>> 5) & 7);
7037 case 0: /* Unsigned sum of absolute differences. */
7039 tmp
= load_reg(s
, rm
);
7040 tmp2
= load_reg(s
, rs
);
7041 gen_helper_usad8(tmp
, tmp
, tmp2
);
7044 tmp2
= load_reg(s
, rd
);
7045 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7048 store_reg(s
, rn
, tmp
);
7050 case 0x20: case 0x24: case 0x28: case 0x2c:
7051 /* Bitfield insert/clear. */
7053 shift
= (insn
>> 7) & 0x1f;
7054 i
= (insn
>> 16) & 0x1f;
7058 tcg_gen_movi_i32(tmp
, 0);
7060 tmp
= load_reg(s
, rm
);
7063 tmp2
= load_reg(s
, rd
);
7064 gen_bfi(tmp
, tmp2
, tmp
, shift
, (1u << i
) - 1);
7067 store_reg(s
, rd
, tmp
);
7069 case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
7070 case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
7072 tmp
= load_reg(s
, rm
);
7073 shift
= (insn
>> 7) & 0x1f;
7074 i
= ((insn
>> 16) & 0x1f) + 1;
7079 gen_ubfx(tmp
, shift
, (1u << i
) - 1);
7081 gen_sbfx(tmp
, shift
, i
);
7084 store_reg(s
, rd
, tmp
);
7094 /* Check for undefined extension instructions
7095 * per the ARM Bible IE:
7096 * xxxx 0111 1111 xxxx xxxx xxxx 1111 xxxx
7098 sh
= (0xf << 20) | (0xf << 4);
7099 if (op1
== 0x7 && ((insn
& sh
) == sh
))
7103 /* load/store byte/word */
7104 rn
= (insn
>> 16) & 0xf;
7105 rd
= (insn
>> 12) & 0xf;
7106 tmp2
= load_reg(s
, rn
);
7107 i
= (IS_USER(s
) || (insn
& 0x01200000) == 0x00200000);
7108 if (insn
& (1 << 24))
7109 gen_add_data_offset(s
, insn
, tmp2
);
7110 if (insn
& (1 << 20)) {
7112 if (insn
& (1 << 22)) {
7113 tmp
= gen_ld8u(tmp2
, i
);
7115 tmp
= gen_ld32(tmp2
, i
);
7119 tmp
= load_reg(s
, rd
);
7120 if (insn
& (1 << 22))
7121 gen_st8(tmp
, tmp2
, i
);
7123 gen_st32(tmp
, tmp2
, i
);
7125 if (!(insn
& (1 << 24))) {
7126 gen_add_data_offset(s
, insn
, tmp2
);
7127 store_reg(s
, rn
, tmp2
);
7128 } else if (insn
& (1 << 21)) {
7129 store_reg(s
, rn
, tmp2
);
7133 if (insn
& (1 << 20)) {
7134 /* Complete the load. */
7138 store_reg(s
, rd
, tmp
);
7144 int j
, n
, user
, loaded_base
;
7146 /* load/store multiple words */
7147 /* XXX: store correct base if write back */
7149 if (insn
& (1 << 22)) {
7151 goto illegal_op
; /* only usable in supervisor mode */
7153 if ((insn
& (1 << 15)) == 0)
7156 rn
= (insn
>> 16) & 0xf;
7157 addr
= load_reg(s
, rn
);
7159 /* compute total size */
7161 TCGV_UNUSED(loaded_var
);
7164 if (insn
& (1 << i
))
7167 /* XXX: test invalid n == 0 case ? */
7168 if (insn
& (1 << 23)) {
7169 if (insn
& (1 << 24)) {
7171 tcg_gen_addi_i32(addr
, addr
, 4);
7173 /* post increment */
7176 if (insn
& (1 << 24)) {
7178 tcg_gen_addi_i32(addr
, addr
, -(n
* 4));
7180 /* post decrement */
7182 tcg_gen_addi_i32(addr
, addr
, -((n
- 1) * 4));
7187 if (insn
& (1 << i
)) {
7188 if (insn
& (1 << 20)) {
7190 tmp
= gen_ld32(addr
, IS_USER(s
));
7194 tmp2
= tcg_const_i32(i
);
7195 gen_helper_set_user_reg(tmp2
, tmp
);
7196 tcg_temp_free_i32(tmp2
);
7198 } else if (i
== rn
) {
7202 store_reg(s
, i
, tmp
);
7207 /* special case: r15 = PC + 8 */
7208 val
= (long)s
->pc
+ 4;
7210 tcg_gen_movi_i32(tmp
, val
);
7213 tmp2
= tcg_const_i32(i
);
7214 gen_helper_get_user_reg(tmp
, tmp2
);
7215 tcg_temp_free_i32(tmp2
);
7217 tmp
= load_reg(s
, i
);
7219 gen_st32(tmp
, addr
, IS_USER(s
));
7222 /* no need to add after the last transfer */
7224 tcg_gen_addi_i32(addr
, addr
, 4);
7227 if (insn
& (1 << 21)) {
7229 if (insn
& (1 << 23)) {
7230 if (insn
& (1 << 24)) {
7233 /* post increment */
7234 tcg_gen_addi_i32(addr
, addr
, 4);
7237 if (insn
& (1 << 24)) {
7240 tcg_gen_addi_i32(addr
, addr
, -((n
- 1) * 4));
7242 /* post decrement */
7243 tcg_gen_addi_i32(addr
, addr
, -(n
* 4));
7246 store_reg(s
, rn
, addr
);
7251 store_reg(s
, rn
, loaded_var
);
7253 if ((insn
& (1 << 22)) && !user
) {
7254 /* Restore CPSR from SPSR. */
7255 tmp
= load_cpu_field(spsr
);
7256 gen_set_cpsr(tmp
, 0xffffffff);
7258 s
->is_jmp
= DISAS_UPDATE
;
7267 /* branch (and link) */
7268 val
= (int32_t)s
->pc
;
7269 if (insn
& (1 << 24)) {
7271 tcg_gen_movi_i32(tmp
, val
);
7272 store_reg(s
, 14, tmp
);
7274 offset
= (((int32_t)insn
<< 8) >> 8);
7275 val
+= (offset
<< 2) + 4;
7283 if (disas_coproc_insn(env
, s
, insn
))
7288 gen_set_pc_im(s
->pc
);
7289 s
->is_jmp
= DISAS_SWI
;
7293 gen_exception_insn(s
, 4, EXCP_UDEF
);
7299 /* Return true if this is a Thumb-2 logical op. */
7301 thumb2_logic_op(int op
)
7306 /* Generate code for a Thumb-2 data processing operation. If CONDS is nonzero
7307 then set condition code flags based on the result of the operation.
7308 If SHIFTER_OUT is nonzero then set the carry flag for logical operations
7309 to the high bit of T1.
7310 Returns zero if the opcode is valid. */
7313 gen_thumb2_data_op(DisasContext
*s
, int op
, int conds
, uint32_t shifter_out
, TCGv t0
, TCGv t1
)
7320 tcg_gen_and_i32(t0
, t0
, t1
);
7324 tcg_gen_andc_i32(t0
, t0
, t1
);
7328 tcg_gen_or_i32(t0
, t0
, t1
);
7332 tcg_gen_not_i32(t1
, t1
);
7333 tcg_gen_or_i32(t0
, t0
, t1
);
7337 tcg_gen_xor_i32(t0
, t0
, t1
);
7342 gen_helper_add_cc(t0
, t0
, t1
);
7344 tcg_gen_add_i32(t0
, t0
, t1
);
7348 gen_helper_adc_cc(t0
, t0
, t1
);
7354 gen_helper_sbc_cc(t0
, t0
, t1
);
7356 gen_sub_carry(t0
, t0
, t1
);
7360 gen_helper_sub_cc(t0
, t0
, t1
);
7362 tcg_gen_sub_i32(t0
, t0
, t1
);
7366 gen_helper_sub_cc(t0
, t1
, t0
);
7368 tcg_gen_sub_i32(t0
, t1
, t0
);
7370 default: /* 5, 6, 7, 9, 12, 15. */
7376 gen_set_CF_bit31(t1
);
7381 /* Translate a 32-bit thumb instruction. Returns nonzero if the instruction
7383 static int disas_thumb2_insn(CPUState
*env
, DisasContext
*s
, uint16_t insn_hw1
)
7385 uint32_t insn
, imm
, shift
, offset
;
7386 uint32_t rd
, rn
, rm
, rs
;
7397 if (!(arm_feature(env
, ARM_FEATURE_THUMB2
)
7398 || arm_feature (env
, ARM_FEATURE_M
))) {
7399 /* Thumb-1 cores may need to treat bl and blx as a pair of
7400 16-bit instructions to get correct prefetch abort behavior. */
7402 if ((insn
& (1 << 12)) == 0) {
7403 /* Second half of blx. */
7404 offset
= ((insn
& 0x7ff) << 1);
7405 tmp
= load_reg(s
, 14);
7406 tcg_gen_addi_i32(tmp
, tmp
, offset
);
7407 tcg_gen_andi_i32(tmp
, tmp
, 0xfffffffc);
7410 tcg_gen_movi_i32(tmp2
, s
->pc
| 1);
7411 store_reg(s
, 14, tmp2
);
7415 if (insn
& (1 << 11)) {
7416 /* Second half of bl. */
7417 offset
= ((insn
& 0x7ff) << 1) | 1;
7418 tmp
= load_reg(s
, 14);
7419 tcg_gen_addi_i32(tmp
, tmp
, offset
);
7422 tcg_gen_movi_i32(tmp2
, s
->pc
| 1);
7423 store_reg(s
, 14, tmp2
);
7427 if ((s
->pc
& ~TARGET_PAGE_MASK
) == 0) {
7428 /* Instruction spans a page boundary. Implement it as two
7429 16-bit instructions in case the second half causes an
7431 offset
= ((int32_t)insn
<< 21) >> 9;
7432 tcg_gen_movi_i32(cpu_R
[14], s
->pc
+ 2 + offset
);
7435 /* Fall through to 32-bit decode. */
7438 insn
= lduw_code(s
->pc
);
7440 insn
|= (uint32_t)insn_hw1
<< 16;
7442 if ((insn
& 0xf800e800) != 0xf000e800) {
7446 rn
= (insn
>> 16) & 0xf;
7447 rs
= (insn
>> 12) & 0xf;
7448 rd
= (insn
>> 8) & 0xf;
7450 switch ((insn
>> 25) & 0xf) {
7451 case 0: case 1: case 2: case 3:
7452 /* 16-bit instructions. Should never happen. */
7455 if (insn
& (1 << 22)) {
7456 /* Other load/store, table branch. */
7457 if (insn
& 0x01200000) {
7458 /* Load/store doubleword. */
7461 tcg_gen_movi_i32(addr
, s
->pc
& ~3);
7463 addr
= load_reg(s
, rn
);
7465 offset
= (insn
& 0xff) * 4;
7466 if ((insn
& (1 << 23)) == 0)
7468 if (insn
& (1 << 24)) {
7469 tcg_gen_addi_i32(addr
, addr
, offset
);
7472 if (insn
& (1 << 20)) {
7474 tmp
= gen_ld32(addr
, IS_USER(s
));
7475 store_reg(s
, rs
, tmp
);
7476 tcg_gen_addi_i32(addr
, addr
, 4);
7477 tmp
= gen_ld32(addr
, IS_USER(s
));
7478 store_reg(s
, rd
, tmp
);
7481 tmp
= load_reg(s
, rs
);
7482 gen_st32(tmp
, addr
, IS_USER(s
));
7483 tcg_gen_addi_i32(addr
, addr
, 4);
7484 tmp
= load_reg(s
, rd
);
7485 gen_st32(tmp
, addr
, IS_USER(s
));
7487 if (insn
& (1 << 21)) {
7488 /* Base writeback. */
7491 tcg_gen_addi_i32(addr
, addr
, offset
- 4);
7492 store_reg(s
, rn
, addr
);
7496 } else if ((insn
& (1 << 23)) == 0) {
7497 /* Load/store exclusive word. */
7498 addr
= tcg_temp_local_new();
7499 load_reg_var(s
, addr
, rn
);
7500 tcg_gen_addi_i32(addr
, addr
, (insn
& 0xff) << 2);
7501 if (insn
& (1 << 20)) {
7502 gen_load_exclusive(s
, rs
, 15, addr
, 2);
7504 gen_store_exclusive(s
, rd
, rs
, 15, addr
, 2);
7506 tcg_temp_free(addr
);
7507 } else if ((insn
& (1 << 6)) == 0) {
7511 tcg_gen_movi_i32(addr
, s
->pc
);
7513 addr
= load_reg(s
, rn
);
7515 tmp
= load_reg(s
, rm
);
7516 tcg_gen_add_i32(addr
, addr
, tmp
);
7517 if (insn
& (1 << 4)) {
7519 tcg_gen_add_i32(addr
, addr
, tmp
);
7521 tmp
= gen_ld16u(addr
, IS_USER(s
));
7524 tmp
= gen_ld8u(addr
, IS_USER(s
));
7527 tcg_gen_shli_i32(tmp
, tmp
, 1);
7528 tcg_gen_addi_i32(tmp
, tmp
, s
->pc
);
7529 store_reg(s
, 15, tmp
);
7531 /* Load/store exclusive byte/halfword/doubleword. */
7533 op
= (insn
>> 4) & 0x3;
7537 addr
= tcg_temp_local_new();
7538 load_reg_var(s
, addr
, rn
);
7539 if (insn
& (1 << 20)) {
7540 gen_load_exclusive(s
, rs
, rd
, addr
, op
);
7542 gen_store_exclusive(s
, rm
, rs
, rd
, addr
, op
);
7544 tcg_temp_free(addr
);
7547 /* Load/store multiple, RFE, SRS. */
7548 if (((insn
>> 23) & 1) == ((insn
>> 24) & 1)) {
7549 /* Not available in user mode. */
7552 if (insn
& (1 << 20)) {
7554 addr
= load_reg(s
, rn
);
7555 if ((insn
& (1 << 24)) == 0)
7556 tcg_gen_addi_i32(addr
, addr
, -8);
7557 /* Load PC into tmp and CPSR into tmp2. */
7558 tmp
= gen_ld32(addr
, 0);
7559 tcg_gen_addi_i32(addr
, addr
, 4);
7560 tmp2
= gen_ld32(addr
, 0);
7561 if (insn
& (1 << 21)) {
7562 /* Base writeback. */
7563 if (insn
& (1 << 24)) {
7564 tcg_gen_addi_i32(addr
, addr
, 4);
7566 tcg_gen_addi_i32(addr
, addr
, -4);
7568 store_reg(s
, rn
, addr
);
7572 gen_rfe(s
, tmp
, tmp2
);
7577 tmp
= tcg_const_i32(op
);
7578 gen_helper_get_r13_banked(addr
, cpu_env
, tmp
);
7579 tcg_temp_free_i32(tmp
);
7580 if ((insn
& (1 << 24)) == 0) {
7581 tcg_gen_addi_i32(addr
, addr
, -8);
7583 tmp
= load_reg(s
, 14);
7584 gen_st32(tmp
, addr
, 0);
7585 tcg_gen_addi_i32(addr
, addr
, 4);
7587 gen_helper_cpsr_read(tmp
);
7588 gen_st32(tmp
, addr
, 0);
7589 if (insn
& (1 << 21)) {
7590 if ((insn
& (1 << 24)) == 0) {
7591 tcg_gen_addi_i32(addr
, addr
, -4);
7593 tcg_gen_addi_i32(addr
, addr
, 4);
7595 tmp
= tcg_const_i32(op
);
7596 gen_helper_set_r13_banked(cpu_env
, tmp
, addr
);
7597 tcg_temp_free_i32(tmp
);
7604 /* Load/store multiple. */
7605 addr
= load_reg(s
, rn
);
7607 for (i
= 0; i
< 16; i
++) {
7608 if (insn
& (1 << i
))
7611 if (insn
& (1 << 24)) {
7612 tcg_gen_addi_i32(addr
, addr
, -offset
);
7615 for (i
= 0; i
< 16; i
++) {
7616 if ((insn
& (1 << i
)) == 0)
7618 if (insn
& (1 << 20)) {
7620 tmp
= gen_ld32(addr
, IS_USER(s
));
7624 store_reg(s
, i
, tmp
);
7628 tmp
= load_reg(s
, i
);
7629 gen_st32(tmp
, addr
, IS_USER(s
));
7631 tcg_gen_addi_i32(addr
, addr
, 4);
7633 if (insn
& (1 << 21)) {
7634 /* Base register writeback. */
7635 if (insn
& (1 << 24)) {
7636 tcg_gen_addi_i32(addr
, addr
, -offset
);
7638 /* Fault if writeback register is in register list. */
7639 if (insn
& (1 << rn
))
7641 store_reg(s
, rn
, addr
);
7650 op
= (insn
>> 21) & 0xf;
7652 /* Halfword pack. */
7653 tmp
= load_reg(s
, rn
);
7654 tmp2
= load_reg(s
, rm
);
7655 shift
= ((insn
>> 10) & 0x1c) | ((insn
>> 6) & 0x3);
7656 if (insn
& (1 << 5)) {
7660 tcg_gen_sari_i32(tmp2
, tmp2
, shift
);
7661 tcg_gen_andi_i32(tmp
, tmp
, 0xffff0000);
7662 tcg_gen_ext16u_i32(tmp2
, tmp2
);
7666 tcg_gen_shli_i32(tmp2
, tmp2
, shift
);
7667 tcg_gen_ext16u_i32(tmp
, tmp
);
7668 tcg_gen_andi_i32(tmp2
, tmp2
, 0xffff0000);
7670 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
7672 store_reg(s
, rd
, tmp
);
7674 /* Data processing register constant shift. */
7677 tcg_gen_movi_i32(tmp
, 0);
7679 tmp
= load_reg(s
, rn
);
7681 tmp2
= load_reg(s
, rm
);
7683 shiftop
= (insn
>> 4) & 3;
7684 shift
= ((insn
>> 6) & 3) | ((insn
>> 10) & 0x1c);
7685 conds
= (insn
& (1 << 20)) != 0;
7686 logic_cc
= (conds
&& thumb2_logic_op(op
));
7687 gen_arm_shift_im(tmp2
, shiftop
, shift
, logic_cc
);
7688 if (gen_thumb2_data_op(s
, op
, conds
, 0, tmp
, tmp2
))
7692 store_reg(s
, rd
, tmp
);
7698 case 13: /* Misc data processing. */
7699 op
= ((insn
>> 22) & 6) | ((insn
>> 7) & 1);
7700 if (op
< 4 && (insn
& 0xf000) != 0xf000)
7703 case 0: /* Register controlled shift. */
7704 tmp
= load_reg(s
, rn
);
7705 tmp2
= load_reg(s
, rm
);
7706 if ((insn
& 0x70) != 0)
7708 op
= (insn
>> 21) & 3;
7709 logic_cc
= (insn
& (1 << 20)) != 0;
7710 gen_arm_shift_reg(tmp
, op
, tmp2
, logic_cc
);
7713 store_reg_bx(env
, s
, rd
, tmp
);
7715 case 1: /* Sign/zero extend. */
7716 tmp
= load_reg(s
, rm
);
7717 shift
= (insn
>> 4) & 3;
7718 /* ??? In many cases it's not neccessary to do a
7719 rotate, a shift is sufficient. */
7721 tcg_gen_rotri_i32(tmp
, tmp
, shift
* 8);
7722 op
= (insn
>> 20) & 7;
7724 case 0: gen_sxth(tmp
); break;
7725 case 1: gen_uxth(tmp
); break;
7726 case 2: gen_sxtb16(tmp
); break;
7727 case 3: gen_uxtb16(tmp
); break;
7728 case 4: gen_sxtb(tmp
); break;
7729 case 5: gen_uxtb(tmp
); break;
7730 default: goto illegal_op
;
7733 tmp2
= load_reg(s
, rn
);
7734 if ((op
>> 1) == 1) {
7735 gen_add16(tmp
, tmp2
);
7737 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7741 store_reg(s
, rd
, tmp
);
7743 case 2: /* SIMD add/subtract. */
7744 op
= (insn
>> 20) & 7;
7745 shift
= (insn
>> 4) & 7;
7746 if ((op
& 3) == 3 || (shift
& 3) == 3)
7748 tmp
= load_reg(s
, rn
);
7749 tmp2
= load_reg(s
, rm
);
7750 gen_thumb2_parallel_addsub(op
, shift
, tmp
, tmp2
);
7752 store_reg(s
, rd
, tmp
);
7754 case 3: /* Other data processing. */
7755 op
= ((insn
>> 17) & 0x38) | ((insn
>> 4) & 7);
7757 /* Saturating add/subtract. */
7758 tmp
= load_reg(s
, rn
);
7759 tmp2
= load_reg(s
, rm
);
7761 gen_helper_double_saturate(tmp
, tmp
);
7763 gen_helper_sub_saturate(tmp
, tmp2
, tmp
);
7765 gen_helper_add_saturate(tmp
, tmp
, tmp2
);
7768 tmp
= load_reg(s
, rn
);
7770 case 0x0a: /* rbit */
7771 gen_helper_rbit(tmp
, tmp
);
7773 case 0x08: /* rev */
7774 tcg_gen_bswap32_i32(tmp
, tmp
);
7776 case 0x09: /* rev16 */
7779 case 0x0b: /* revsh */
7782 case 0x10: /* sel */
7783 tmp2
= load_reg(s
, rm
);
7785 tcg_gen_ld_i32(tmp3
, cpu_env
, offsetof(CPUState
, GE
));
7786 gen_helper_sel_flags(tmp
, tmp3
, tmp
, tmp2
);
7790 case 0x18: /* clz */
7791 gen_helper_clz(tmp
, tmp
);
7797 store_reg(s
, rd
, tmp
);
7799 case 4: case 5: /* 32-bit multiply. Sum of absolute differences. */
7800 op
= (insn
>> 4) & 0xf;
7801 tmp
= load_reg(s
, rn
);
7802 tmp2
= load_reg(s
, rm
);
7803 switch ((insn
>> 20) & 7) {
7804 case 0: /* 32 x 32 -> 32 */
7805 tcg_gen_mul_i32(tmp
, tmp
, tmp2
);
7808 tmp2
= load_reg(s
, rs
);
7810 tcg_gen_sub_i32(tmp
, tmp2
, tmp
);
7812 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7816 case 1: /* 16 x 16 -> 32 */
7817 gen_mulxy(tmp
, tmp2
, op
& 2, op
& 1);
7820 tmp2
= load_reg(s
, rs
);
7821 gen_helper_add_setq(tmp
, tmp
, tmp2
);
7825 case 2: /* Dual multiply add. */
7826 case 4: /* Dual multiply subtract. */
7828 gen_swap_half(tmp2
);
7829 gen_smul_dual(tmp
, tmp2
);
7830 /* This addition cannot overflow. */
7831 if (insn
& (1 << 22)) {
7832 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
7834 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7839 tmp2
= load_reg(s
, rs
);
7840 gen_helper_add_setq(tmp
, tmp
, tmp2
);
7844 case 3: /* 32 * 16 -> 32msb */
7846 tcg_gen_sari_i32(tmp2
, tmp2
, 16);
7849 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
7850 tcg_gen_shri_i64(tmp64
, tmp64
, 16);
7852 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
7853 tcg_temp_free_i64(tmp64
);
7856 tmp2
= load_reg(s
, rs
);
7857 gen_helper_add_setq(tmp
, tmp
, tmp2
);
7861 case 5: case 6: /* 32 * 32 -> 32msb (SMMUL, SMMLA, SMMLS) */
7862 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
7864 tmp
= load_reg(s
, rs
);
7865 if (insn
& (1 << 20)) {
7866 tmp64
= gen_addq_msw(tmp64
, tmp
);
7868 tmp64
= gen_subq_msw(tmp64
, tmp
);
7871 if (insn
& (1 << 4)) {
7872 tcg_gen_addi_i64(tmp64
, tmp64
, 0x80000000u
);
7874 tcg_gen_shri_i64(tmp64
, tmp64
, 32);
7876 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
7877 tcg_temp_free_i64(tmp64
);
7879 case 7: /* Unsigned sum of absolute differences. */
7880 gen_helper_usad8(tmp
, tmp
, tmp2
);
7883 tmp2
= load_reg(s
, rs
);
7884 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7889 store_reg(s
, rd
, tmp
);
7891 case 6: case 7: /* 64-bit multiply, Divide. */
7892 op
= ((insn
>> 4) & 0xf) | ((insn
>> 16) & 0x70);
7893 tmp
= load_reg(s
, rn
);
7894 tmp2
= load_reg(s
, rm
);
7895 if ((op
& 0x50) == 0x10) {
7897 if (!arm_feature(env
, ARM_FEATURE_DIV
))
7900 gen_helper_udiv(tmp
, tmp
, tmp2
);
7902 gen_helper_sdiv(tmp
, tmp
, tmp2
);
7904 store_reg(s
, rd
, tmp
);
7905 } else if ((op
& 0xe) == 0xc) {
7906 /* Dual multiply accumulate long. */
7908 gen_swap_half(tmp2
);
7909 gen_smul_dual(tmp
, tmp2
);
7911 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
7913 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7917 tmp64
= tcg_temp_new_i64();
7918 tcg_gen_ext_i32_i64(tmp64
, tmp
);
7920 gen_addq(s
, tmp64
, rs
, rd
);
7921 gen_storeq_reg(s
, rs
, rd
, tmp64
);
7922 tcg_temp_free_i64(tmp64
);
7925 /* Unsigned 64-bit multiply */
7926 tmp64
= gen_mulu_i64_i32(tmp
, tmp2
);
7930 gen_mulxy(tmp
, tmp2
, op
& 2, op
& 1);
7932 tmp64
= tcg_temp_new_i64();
7933 tcg_gen_ext_i32_i64(tmp64
, tmp
);
7936 /* Signed 64-bit multiply */
7937 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
7942 gen_addq_lo(s
, tmp64
, rs
);
7943 gen_addq_lo(s
, tmp64
, rd
);
7944 } else if (op
& 0x40) {
7945 /* 64-bit accumulate. */
7946 gen_addq(s
, tmp64
, rs
, rd
);
7948 gen_storeq_reg(s
, rs
, rd
, tmp64
);
7949 tcg_temp_free_i64(tmp64
);
7954 case 6: case 7: case 14: case 15:
7956 if (((insn
>> 24) & 3) == 3) {
7957 /* Translate into the equivalent ARM encoding. */
7958 insn
= (insn
& 0xe2ffffff) | ((insn
& (1 << 28)) >> 4) | (1 << 28);
7959 if (disas_neon_data_insn(env
, s
, insn
))
7962 if (insn
& (1 << 28))
7964 if (disas_coproc_insn (env
, s
, insn
))
7968 case 8: case 9: case 10: case 11:
7969 if (insn
& (1 << 15)) {
7970 /* Branches, misc control. */
7971 if (insn
& 0x5000) {
7972 /* Unconditional branch. */
7973 /* signextend(hw1[10:0]) -> offset[:12]. */
7974 offset
= ((int32_t)insn
<< 5) >> 9 & ~(int32_t)0xfff;
7975 /* hw1[10:0] -> offset[11:1]. */
7976 offset
|= (insn
& 0x7ff) << 1;
7977 /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
7978 offset[24:22] already have the same value because of the
7979 sign extension above. */
7980 offset
^= ((~insn
) & (1 << 13)) << 10;
7981 offset
^= ((~insn
) & (1 << 11)) << 11;
7983 if (insn
& (1 << 14)) {
7984 /* Branch and link. */
7985 tcg_gen_movi_i32(cpu_R
[14], s
->pc
| 1);
7989 if (insn
& (1 << 12)) {
7994 offset
&= ~(uint32_t)2;
7995 gen_bx_im(s
, offset
);
7997 } else if (((insn
>> 23) & 7) == 7) {
7999 if (insn
& (1 << 13))
8002 if (insn
& (1 << 26)) {
8003 /* Secure monitor call (v6Z) */
8004 goto illegal_op
; /* not implemented. */
8006 op
= (insn
>> 20) & 7;
8008 case 0: /* msr cpsr. */
8010 tmp
= load_reg(s
, rn
);
8011 addr
= tcg_const_i32(insn
& 0xff);
8012 gen_helper_v7m_msr(cpu_env
, addr
, tmp
);
8013 tcg_temp_free_i32(addr
);
8019 case 1: /* msr spsr. */
8022 tmp
= load_reg(s
, rn
);
8024 msr_mask(env
, s
, (insn
>> 8) & 0xf, op
== 1),
8028 case 2: /* cps, nop-hint. */
8029 if (((insn
>> 8) & 7) == 0) {
8030 gen_nop_hint(s
, insn
& 0xff);
8032 /* Implemented as NOP in user mode. */
8037 if (insn
& (1 << 10)) {
8038 if (insn
& (1 << 7))
8040 if (insn
& (1 << 6))
8042 if (insn
& (1 << 5))
8044 if (insn
& (1 << 9))
8045 imm
= CPSR_A
| CPSR_I
| CPSR_F
;
8047 if (insn
& (1 << 8)) {
8049 imm
|= (insn
& 0x1f);
8052 gen_set_psr_im(s
, offset
, 0, imm
);
8055 case 3: /* Special control operations. */
8057 op
= (insn
>> 4) & 0xf;
8065 /* These execute as NOPs. */
8072 /* Trivial implementation equivalent to bx. */
8073 tmp
= load_reg(s
, rn
);
8076 case 5: /* Exception return. */
8080 if (rn
!= 14 || rd
!= 15) {
8083 tmp
= load_reg(s
, rn
);
8084 tcg_gen_subi_i32(tmp
, tmp
, insn
& 0xff);
8085 gen_exception_return(s
, tmp
);
8087 case 6: /* mrs cpsr. */
8090 addr
= tcg_const_i32(insn
& 0xff);
8091 gen_helper_v7m_mrs(tmp
, cpu_env
, addr
);
8092 tcg_temp_free_i32(addr
);
8094 gen_helper_cpsr_read(tmp
);
8096 store_reg(s
, rd
, tmp
);
8098 case 7: /* mrs spsr. */
8099 /* Not accessible in user mode. */
8100 if (IS_USER(s
) || IS_M(env
))
8102 tmp
= load_cpu_field(spsr
);
8103 store_reg(s
, rd
, tmp
);
8108 /* Conditional branch. */
8109 op
= (insn
>> 22) & 0xf;
8110 /* Generate a conditional jump to next instruction. */
8111 s
->condlabel
= gen_new_label();
8112 gen_test_cc(op
^ 1, s
->condlabel
);
8115 /* offset[11:1] = insn[10:0] */
8116 offset
= (insn
& 0x7ff) << 1;
8117 /* offset[17:12] = insn[21:16]. */
8118 offset
|= (insn
& 0x003f0000) >> 4;
8119 /* offset[31:20] = insn[26]. */
8120 offset
|= ((int32_t)((insn
<< 5) & 0x80000000)) >> 11;
8121 /* offset[18] = insn[13]. */
8122 offset
|= (insn
& (1 << 13)) << 5;
8123 /* offset[19] = insn[11]. */
8124 offset
|= (insn
& (1 << 11)) << 8;
8126 /* jump to the offset */
8127 gen_jmp(s
, s
->pc
+ offset
);
8130 /* Data processing immediate. */
8131 if (insn
& (1 << 25)) {
8132 if (insn
& (1 << 24)) {
8133 if (insn
& (1 << 20))
8135 /* Bitfield/Saturate. */
8136 op
= (insn
>> 21) & 7;
8138 shift
= ((insn
>> 6) & 3) | ((insn
>> 10) & 0x1c);
8141 tcg_gen_movi_i32(tmp
, 0);
8143 tmp
= load_reg(s
, rn
);
8146 case 2: /* Signed bitfield extract. */
8148 if (shift
+ imm
> 32)
8151 gen_sbfx(tmp
, shift
, imm
);
8153 case 6: /* Unsigned bitfield extract. */
8155 if (shift
+ imm
> 32)
8158 gen_ubfx(tmp
, shift
, (1u << imm
) - 1);
8160 case 3: /* Bitfield insert/clear. */
8163 imm
= imm
+ 1 - shift
;
8165 tmp2
= load_reg(s
, rd
);
8166 gen_bfi(tmp
, tmp2
, tmp
, shift
, (1u << imm
) - 1);
8172 default: /* Saturate. */
8175 tcg_gen_sari_i32(tmp
, tmp
, shift
);
8177 tcg_gen_shli_i32(tmp
, tmp
, shift
);
8179 tmp2
= tcg_const_i32(imm
);
8182 if ((op
& 1) && shift
== 0)
8183 gen_helper_usat16(tmp
, tmp
, tmp2
);
8185 gen_helper_usat(tmp
, tmp
, tmp2
);
8188 if ((op
& 1) && shift
== 0)
8189 gen_helper_ssat16(tmp
, tmp
, tmp2
);
8191 gen_helper_ssat(tmp
, tmp
, tmp2
);
8193 tcg_temp_free_i32(tmp2
);
8196 store_reg(s
, rd
, tmp
);
8198 imm
= ((insn
& 0x04000000) >> 15)
8199 | ((insn
& 0x7000) >> 4) | (insn
& 0xff);
8200 if (insn
& (1 << 22)) {
8201 /* 16-bit immediate. */
8202 imm
|= (insn
>> 4) & 0xf000;
8203 if (insn
& (1 << 23)) {
8205 tmp
= load_reg(s
, rd
);
8206 tcg_gen_ext16u_i32(tmp
, tmp
);
8207 tcg_gen_ori_i32(tmp
, tmp
, imm
<< 16);
8211 tcg_gen_movi_i32(tmp
, imm
);
8214 /* Add/sub 12-bit immediate. */
8216 offset
= s
->pc
& ~(uint32_t)3;
8217 if (insn
& (1 << 23))
8222 tcg_gen_movi_i32(tmp
, offset
);
8224 tmp
= load_reg(s
, rn
);
8225 if (insn
& (1 << 23))
8226 tcg_gen_subi_i32(tmp
, tmp
, imm
);
8228 tcg_gen_addi_i32(tmp
, tmp
, imm
);
8231 store_reg(s
, rd
, tmp
);
8234 int shifter_out
= 0;
8235 /* modified 12-bit immediate. */
8236 shift
= ((insn
& 0x04000000) >> 23) | ((insn
& 0x7000) >> 12);
8237 imm
= (insn
& 0xff);
8240 /* Nothing to do. */
8242 case 1: /* 00XY00XY */
8245 case 2: /* XY00XY00 */
8249 case 3: /* XYXYXYXY */
8253 default: /* Rotated constant. */
8254 shift
= (shift
<< 1) | (imm
>> 7);
8256 imm
= imm
<< (32 - shift
);
8261 tcg_gen_movi_i32(tmp2
, imm
);
8262 rn
= (insn
>> 16) & 0xf;
8265 tcg_gen_movi_i32(tmp
, 0);
8267 tmp
= load_reg(s
, rn
);
8269 op
= (insn
>> 21) & 0xf;
8270 if (gen_thumb2_data_op(s
, op
, (insn
& (1 << 20)) != 0,
8271 shifter_out
, tmp
, tmp2
))
8274 rd
= (insn
>> 8) & 0xf;
8276 store_reg(s
, rd
, tmp
);
8283 case 12: /* Load/store single data item. */
8288 if ((insn
& 0x01100000) == 0x01000000) {
8289 if (disas_neon_ls_insn(env
, s
, insn
))
8293 op
= ((insn
>> 21) & 3) | ((insn
>> 22) & 4);
8295 if (!(insn
& (1 << 20))) {
8299 /* Byte or halfword load space with dest == r15 : memory hints.
8300 * Catch them early so we don't emit pointless addressing code.
8301 * This space is a mix of:
8302 * PLD/PLDW/PLI, which we implement as NOPs (note that unlike
8303 * the ARM encodings, PLDW space doesn't UNDEF for non-v7MP
8305 * unallocated hints, which must be treated as NOPs
8306 * UNPREDICTABLE space, which we NOP or UNDEF depending on
8307 * which is easiest for the decoding logic
8308 * Some space which must UNDEF
8310 int op1
= (insn
>> 23) & 3;
8311 int op2
= (insn
>> 6) & 0x3f;
8316 /* UNPREDICTABLE or unallocated hint */
8320 return 0; /* PLD* or unallocated hint */
8322 if ((op2
== 0) || ((op2
& 0x3c) == 0x30)) {
8323 return 0; /* PLD* or unallocated hint */
8325 /* UNDEF space, or an UNPREDICTABLE */
8333 /* s->pc has already been incremented by 4. */
8334 imm
= s
->pc
& 0xfffffffc;
8335 if (insn
& (1 << 23))
8336 imm
+= insn
& 0xfff;
8338 imm
-= insn
& 0xfff;
8339 tcg_gen_movi_i32(addr
, imm
);
8341 addr
= load_reg(s
, rn
);
8342 if (insn
& (1 << 23)) {
8343 /* Positive offset. */
8345 tcg_gen_addi_i32(addr
, addr
, imm
);
8348 switch ((insn
>> 8) & 7) {
8349 case 0: case 8: /* Shifted Register. */
8350 shift
= (insn
>> 4) & 0xf;
8353 tmp
= load_reg(s
, rm
);
8355 tcg_gen_shli_i32(tmp
, tmp
, shift
);
8356 tcg_gen_add_i32(addr
, addr
, tmp
);
8359 case 4: /* Negative offset. */
8360 tcg_gen_addi_i32(addr
, addr
, -imm
);
8362 case 6: /* User privilege. */
8363 tcg_gen_addi_i32(addr
, addr
, imm
);
8366 case 1: /* Post-decrement. */
8369 case 3: /* Post-increment. */
8373 case 5: /* Pre-decrement. */
8376 case 7: /* Pre-increment. */
8377 tcg_gen_addi_i32(addr
, addr
, imm
);
8385 if (insn
& (1 << 20)) {
8388 case 0: tmp
= gen_ld8u(addr
, user
); break;
8389 case 4: tmp
= gen_ld8s(addr
, user
); break;
8390 case 1: tmp
= gen_ld16u(addr
, user
); break;
8391 case 5: tmp
= gen_ld16s(addr
, user
); break;
8392 case 2: tmp
= gen_ld32(addr
, user
); break;
8393 default: goto illegal_op
;
8398 store_reg(s
, rs
, tmp
);
8402 tmp
= load_reg(s
, rs
);
8404 case 0: gen_st8(tmp
, addr
, user
); break;
8405 case 1: gen_st16(tmp
, addr
, user
); break;
8406 case 2: gen_st32(tmp
, addr
, user
); break;
8407 default: goto illegal_op
;
8411 tcg_gen_addi_i32(addr
, addr
, imm
);
8413 store_reg(s
, rn
, addr
);
8427 static void disas_thumb_insn(CPUState
*env
, DisasContext
*s
)
8429 uint32_t val
, insn
, op
, rm
, rn
, rd
, shift
, cond
;
8436 if (s
->condexec_mask
) {
8437 cond
= s
->condexec_cond
;
8438 if (cond
!= 0x0e) { /* Skip conditional when condition is AL. */
8439 s
->condlabel
= gen_new_label();
8440 gen_test_cc(cond
^ 1, s
->condlabel
);
8445 insn
= lduw_code(s
->pc
);
8448 switch (insn
>> 12) {
8452 op
= (insn
>> 11) & 3;
8455 rn
= (insn
>> 3) & 7;
8456 tmp
= load_reg(s
, rn
);
8457 if (insn
& (1 << 10)) {
8460 tcg_gen_movi_i32(tmp2
, (insn
>> 6) & 7);
8463 rm
= (insn
>> 6) & 7;
8464 tmp2
= load_reg(s
, rm
);
8466 if (insn
& (1 << 9)) {
8467 if (s
->condexec_mask
)
8468 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
8470 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8472 if (s
->condexec_mask
)
8473 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8475 gen_helper_add_cc(tmp
, tmp
, tmp2
);
8478 store_reg(s
, rd
, tmp
);
8480 /* shift immediate */
8481 rm
= (insn
>> 3) & 7;
8482 shift
= (insn
>> 6) & 0x1f;
8483 tmp
= load_reg(s
, rm
);
8484 gen_arm_shift_im(tmp
, op
, shift
, s
->condexec_mask
== 0);
8485 if (!s
->condexec_mask
)
8487 store_reg(s
, rd
, tmp
);
8491 /* arithmetic large immediate */
8492 op
= (insn
>> 11) & 3;
8493 rd
= (insn
>> 8) & 0x7;
8494 if (op
== 0) { /* mov */
8496 tcg_gen_movi_i32(tmp
, insn
& 0xff);
8497 if (!s
->condexec_mask
)
8499 store_reg(s
, rd
, tmp
);
8501 tmp
= load_reg(s
, rd
);
8503 tcg_gen_movi_i32(tmp2
, insn
& 0xff);
8506 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8511 if (s
->condexec_mask
)
8512 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8514 gen_helper_add_cc(tmp
, tmp
, tmp2
);
8516 store_reg(s
, rd
, tmp
);
8519 if (s
->condexec_mask
)
8520 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
8522 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8524 store_reg(s
, rd
, tmp
);
8530 if (insn
& (1 << 11)) {
8531 rd
= (insn
>> 8) & 7;
8532 /* load pc-relative. Bit 1 of PC is ignored. */
8533 val
= s
->pc
+ 2 + ((insn
& 0xff) * 4);
8534 val
&= ~(uint32_t)2;
8536 tcg_gen_movi_i32(addr
, val
);
8537 tmp
= gen_ld32(addr
, IS_USER(s
));
8539 store_reg(s
, rd
, tmp
);
8542 if (insn
& (1 << 10)) {
8543 /* data processing extended or blx */
8544 rd
= (insn
& 7) | ((insn
>> 4) & 8);
8545 rm
= (insn
>> 3) & 0xf;
8546 op
= (insn
>> 8) & 3;
8549 tmp
= load_reg(s
, rd
);
8550 tmp2
= load_reg(s
, rm
);
8551 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8553 store_reg(s
, rd
, tmp
);
8556 tmp
= load_reg(s
, rd
);
8557 tmp2
= load_reg(s
, rm
);
8558 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8562 case 2: /* mov/cpy */
8563 tmp
= load_reg(s
, rm
);
8564 store_reg(s
, rd
, tmp
);
8566 case 3:/* branch [and link] exchange thumb register */
8567 tmp
= load_reg(s
, rm
);
8568 if (insn
& (1 << 7)) {
8569 val
= (uint32_t)s
->pc
| 1;
8571 tcg_gen_movi_i32(tmp2
, val
);
8572 store_reg(s
, 14, tmp2
);
8580 /* data processing register */
8582 rm
= (insn
>> 3) & 7;
8583 op
= (insn
>> 6) & 0xf;
8584 if (op
== 2 || op
== 3 || op
== 4 || op
== 7) {
8585 /* the shift/rotate ops want the operands backwards */
8594 if (op
== 9) { /* neg */
8596 tcg_gen_movi_i32(tmp
, 0);
8597 } else if (op
!= 0xf) { /* mvn doesn't read its first operand */
8598 tmp
= load_reg(s
, rd
);
8603 tmp2
= load_reg(s
, rm
);
8606 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
8607 if (!s
->condexec_mask
)
8611 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
8612 if (!s
->condexec_mask
)
8616 if (s
->condexec_mask
) {
8617 gen_helper_shl(tmp2
, tmp2
, tmp
);
8619 gen_helper_shl_cc(tmp2
, tmp2
, tmp
);
8624 if (s
->condexec_mask
) {
8625 gen_helper_shr(tmp2
, tmp2
, tmp
);
8627 gen_helper_shr_cc(tmp2
, tmp2
, tmp
);
8632 if (s
->condexec_mask
) {
8633 gen_helper_sar(tmp2
, tmp2
, tmp
);
8635 gen_helper_sar_cc(tmp2
, tmp2
, tmp
);
8640 if (s
->condexec_mask
)
8643 gen_helper_adc_cc(tmp
, tmp
, tmp2
);
8646 if (s
->condexec_mask
)
8647 gen_sub_carry(tmp
, tmp
, tmp2
);
8649 gen_helper_sbc_cc(tmp
, tmp
, tmp2
);
8652 if (s
->condexec_mask
) {
8653 tcg_gen_andi_i32(tmp
, tmp
, 0x1f);
8654 tcg_gen_rotr_i32(tmp2
, tmp2
, tmp
);
8656 gen_helper_ror_cc(tmp2
, tmp2
, tmp
);
8661 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
8666 if (s
->condexec_mask
)
8667 tcg_gen_neg_i32(tmp
, tmp2
);
8669 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8672 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8676 gen_helper_add_cc(tmp
, tmp
, tmp2
);
8680 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
8681 if (!s
->condexec_mask
)
8685 tcg_gen_mul_i32(tmp
, tmp
, tmp2
);
8686 if (!s
->condexec_mask
)
8690 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
8691 if (!s
->condexec_mask
)
8695 tcg_gen_not_i32(tmp2
, tmp2
);
8696 if (!s
->condexec_mask
)
8704 store_reg(s
, rm
, tmp2
);
8708 store_reg(s
, rd
, tmp
);
8718 /* load/store register offset. */
8720 rn
= (insn
>> 3) & 7;
8721 rm
= (insn
>> 6) & 7;
8722 op
= (insn
>> 9) & 7;
8723 addr
= load_reg(s
, rn
);
8724 tmp
= load_reg(s
, rm
);
8725 tcg_gen_add_i32(addr
, addr
, tmp
);
8728 if (op
< 3) /* store */
8729 tmp
= load_reg(s
, rd
);
8733 gen_st32(tmp
, addr
, IS_USER(s
));
8736 gen_st16(tmp
, addr
, IS_USER(s
));
8739 gen_st8(tmp
, addr
, IS_USER(s
));
8742 tmp
= gen_ld8s(addr
, IS_USER(s
));
8745 tmp
= gen_ld32(addr
, IS_USER(s
));
8748 tmp
= gen_ld16u(addr
, IS_USER(s
));
8751 tmp
= gen_ld8u(addr
, IS_USER(s
));
8754 tmp
= gen_ld16s(addr
, IS_USER(s
));
8757 if (op
>= 3) /* load */
8758 store_reg(s
, rd
, tmp
);
8763 /* load/store word immediate offset */
8765 rn
= (insn
>> 3) & 7;
8766 addr
= load_reg(s
, rn
);
8767 val
= (insn
>> 4) & 0x7c;
8768 tcg_gen_addi_i32(addr
, addr
, val
);
8770 if (insn
& (1 << 11)) {
8772 tmp
= gen_ld32(addr
, IS_USER(s
));
8773 store_reg(s
, rd
, tmp
);
8776 tmp
= load_reg(s
, rd
);
8777 gen_st32(tmp
, addr
, IS_USER(s
));
8783 /* load/store byte immediate offset */
8785 rn
= (insn
>> 3) & 7;
8786 addr
= load_reg(s
, rn
);
8787 val
= (insn
>> 6) & 0x1f;
8788 tcg_gen_addi_i32(addr
, addr
, val
);
8790 if (insn
& (1 << 11)) {
8792 tmp
= gen_ld8u(addr
, IS_USER(s
));
8793 store_reg(s
, rd
, tmp
);
8796 tmp
= load_reg(s
, rd
);
8797 gen_st8(tmp
, addr
, IS_USER(s
));
8803 /* load/store halfword immediate offset */
8805 rn
= (insn
>> 3) & 7;
8806 addr
= load_reg(s
, rn
);
8807 val
= (insn
>> 5) & 0x3e;
8808 tcg_gen_addi_i32(addr
, addr
, val
);
8810 if (insn
& (1 << 11)) {
8812 tmp
= gen_ld16u(addr
, IS_USER(s
));
8813 store_reg(s
, rd
, tmp
);
8816 tmp
= load_reg(s
, rd
);
8817 gen_st16(tmp
, addr
, IS_USER(s
));
8823 /* load/store from stack */
8824 rd
= (insn
>> 8) & 7;
8825 addr
= load_reg(s
, 13);
8826 val
= (insn
& 0xff) * 4;
8827 tcg_gen_addi_i32(addr
, addr
, val
);
8829 if (insn
& (1 << 11)) {
8831 tmp
= gen_ld32(addr
, IS_USER(s
));
8832 store_reg(s
, rd
, tmp
);
8835 tmp
= load_reg(s
, rd
);
8836 gen_st32(tmp
, addr
, IS_USER(s
));
8842 /* add to high reg */
8843 rd
= (insn
>> 8) & 7;
8844 if (insn
& (1 << 11)) {
8846 tmp
= load_reg(s
, 13);
8848 /* PC. bit 1 is ignored. */
8850 tcg_gen_movi_i32(tmp
, (s
->pc
+ 2) & ~(uint32_t)2);
8852 val
= (insn
& 0xff) * 4;
8853 tcg_gen_addi_i32(tmp
, tmp
, val
);
8854 store_reg(s
, rd
, tmp
);
8859 op
= (insn
>> 8) & 0xf;
8862 /* adjust stack pointer */
8863 tmp
= load_reg(s
, 13);
8864 val
= (insn
& 0x7f) * 4;
8865 if (insn
& (1 << 7))
8866 val
= -(int32_t)val
;
8867 tcg_gen_addi_i32(tmp
, tmp
, val
);
8868 store_reg(s
, 13, tmp
);
8871 case 2: /* sign/zero extend. */
8874 rm
= (insn
>> 3) & 7;
8875 tmp
= load_reg(s
, rm
);
8876 switch ((insn
>> 6) & 3) {
8877 case 0: gen_sxth(tmp
); break;
8878 case 1: gen_sxtb(tmp
); break;
8879 case 2: gen_uxth(tmp
); break;
8880 case 3: gen_uxtb(tmp
); break;
8882 store_reg(s
, rd
, tmp
);
8884 case 4: case 5: case 0xc: case 0xd:
8886 addr
= load_reg(s
, 13);
8887 if (insn
& (1 << 8))
8891 for (i
= 0; i
< 8; i
++) {
8892 if (insn
& (1 << i
))
8895 if ((insn
& (1 << 11)) == 0) {
8896 tcg_gen_addi_i32(addr
, addr
, -offset
);
8898 for (i
= 0; i
< 8; i
++) {
8899 if (insn
& (1 << i
)) {
8900 if (insn
& (1 << 11)) {
8902 tmp
= gen_ld32(addr
, IS_USER(s
));
8903 store_reg(s
, i
, tmp
);
8906 tmp
= load_reg(s
, i
);
8907 gen_st32(tmp
, addr
, IS_USER(s
));
8909 /* advance to the next address. */
8910 tcg_gen_addi_i32(addr
, addr
, 4);
8914 if (insn
& (1 << 8)) {
8915 if (insn
& (1 << 11)) {
8917 tmp
= gen_ld32(addr
, IS_USER(s
));
8918 /* don't set the pc until the rest of the instruction
8922 tmp
= load_reg(s
, 14);
8923 gen_st32(tmp
, addr
, IS_USER(s
));
8925 tcg_gen_addi_i32(addr
, addr
, 4);
8927 if ((insn
& (1 << 11)) == 0) {
8928 tcg_gen_addi_i32(addr
, addr
, -offset
);
8930 /* write back the new stack pointer */
8931 store_reg(s
, 13, addr
);
8932 /* set the new PC value */
8933 if ((insn
& 0x0900) == 0x0900)
8937 case 1: case 3: case 9: case 11: /* czb */
8939 tmp
= load_reg(s
, rm
);
8940 s
->condlabel
= gen_new_label();
8942 if (insn
& (1 << 11))
8943 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, s
->condlabel
);
8945 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, s
->condlabel
);
8947 offset
= ((insn
& 0xf8) >> 2) | (insn
& 0x200) >> 3;
8948 val
= (uint32_t)s
->pc
+ 2;
8953 case 15: /* IT, nop-hint. */
8954 if ((insn
& 0xf) == 0) {
8955 gen_nop_hint(s
, (insn
>> 4) & 0xf);
8959 s
->condexec_cond
= (insn
>> 4) & 0xe;
8960 s
->condexec_mask
= insn
& 0x1f;
8961 /* No actual code generated for this insn, just setup state. */
8964 case 0xe: /* bkpt */
8965 gen_exception_insn(s
, 2, EXCP_BKPT
);
8970 rn
= (insn
>> 3) & 0x7;
8972 tmp
= load_reg(s
, rn
);
8973 switch ((insn
>> 6) & 3) {
8974 case 0: tcg_gen_bswap32_i32(tmp
, tmp
); break;
8975 case 1: gen_rev16(tmp
); break;
8976 case 3: gen_revsh(tmp
); break;
8977 default: goto illegal_op
;
8979 store_reg(s
, rd
, tmp
);
8987 tmp
= tcg_const_i32((insn
& (1 << 4)) != 0);
8990 addr
= tcg_const_i32(16);
8991 gen_helper_v7m_msr(cpu_env
, addr
, tmp
);
8992 tcg_temp_free_i32(addr
);
8996 addr
= tcg_const_i32(17);
8997 gen_helper_v7m_msr(cpu_env
, addr
, tmp
);
8998 tcg_temp_free_i32(addr
);
9000 tcg_temp_free_i32(tmp
);
9003 if (insn
& (1 << 4))
9004 shift
= CPSR_A
| CPSR_I
| CPSR_F
;
9007 gen_set_psr_im(s
, ((insn
& 7) << 6), 0, shift
);
9017 /* load/store multiple */
9018 rn
= (insn
>> 8) & 0x7;
9019 addr
= load_reg(s
, rn
);
9020 for (i
= 0; i
< 8; i
++) {
9021 if (insn
& (1 << i
)) {
9022 if (insn
& (1 << 11)) {
9024 tmp
= gen_ld32(addr
, IS_USER(s
));
9025 store_reg(s
, i
, tmp
);
9028 tmp
= load_reg(s
, i
);
9029 gen_st32(tmp
, addr
, IS_USER(s
));
9031 /* advance to the next address */
9032 tcg_gen_addi_i32(addr
, addr
, 4);
9035 /* Base register writeback. */
9036 if ((insn
& (1 << rn
)) == 0) {
9037 store_reg(s
, rn
, addr
);
9044 /* conditional branch or swi */
9045 cond
= (insn
>> 8) & 0xf;
9051 gen_set_pc_im(s
->pc
);
9052 s
->is_jmp
= DISAS_SWI
;
9055 /* generate a conditional jump to next instruction */
9056 s
->condlabel
= gen_new_label();
9057 gen_test_cc(cond
^ 1, s
->condlabel
);
9060 /* jump to the offset */
9061 val
= (uint32_t)s
->pc
+ 2;
9062 offset
= ((int32_t)insn
<< 24) >> 24;
9068 if (insn
& (1 << 11)) {
9069 if (disas_thumb2_insn(env
, s
, insn
))
9073 /* unconditional branch */
9074 val
= (uint32_t)s
->pc
;
9075 offset
= ((int32_t)insn
<< 21) >> 21;
9076 val
+= (offset
<< 1) + 2;
9081 if (disas_thumb2_insn(env
, s
, insn
))
9087 gen_exception_insn(s
, 4, EXCP_UDEF
);
9091 gen_exception_insn(s
, 2, EXCP_UDEF
);
9094 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
9095 basic block 'tb'. If search_pc is TRUE, also generate PC
9096 information for each intermediate instruction. */
9097 static inline void gen_intermediate_code_internal(CPUState
*env
,
9098 TranslationBlock
*tb
,
9101 DisasContext dc1
, *dc
= &dc1
;
9103 uint16_t *gen_opc_end
;
9105 target_ulong pc_start
;
9106 uint32_t next_page_start
;
9110 /* generate intermediate code */
9117 gen_opc_end
= gen_opc_buf
+ OPC_MAX_SIZE
;
9119 dc
->is_jmp
= DISAS_NEXT
;
9121 dc
->singlestep_enabled
= env
->singlestep_enabled
;
9123 dc
->thumb
= ARM_TBFLAG_THUMB(tb
->flags
);
9124 dc
->condexec_mask
= (ARM_TBFLAG_CONDEXEC(tb
->flags
) & 0xf) << 1;
9125 dc
->condexec_cond
= ARM_TBFLAG_CONDEXEC(tb
->flags
) >> 4;
9126 #if !defined(CONFIG_USER_ONLY)
9127 dc
->user
= (ARM_TBFLAG_PRIV(tb
->flags
) == 0);
9129 dc
->vfp_enabled
= ARM_TBFLAG_VFPEN(tb
->flags
);
9130 dc
->vec_len
= ARM_TBFLAG_VECLEN(tb
->flags
);
9131 dc
->vec_stride
= ARM_TBFLAG_VECSTRIDE(tb
->flags
);
9132 cpu_F0s
= tcg_temp_new_i32();
9133 cpu_F1s
= tcg_temp_new_i32();
9134 cpu_F0d
= tcg_temp_new_i64();
9135 cpu_F1d
= tcg_temp_new_i64();
9138 /* FIXME: cpu_M0 can probably be the same as cpu_V0. */
9139 cpu_M0
= tcg_temp_new_i64();
9140 next_page_start
= (pc_start
& TARGET_PAGE_MASK
) + TARGET_PAGE_SIZE
;
9143 max_insns
= tb
->cflags
& CF_COUNT_MASK
;
9145 max_insns
= CF_COUNT_MASK
;
9149 /* A note on handling of the condexec (IT) bits:
9151 * We want to avoid the overhead of having to write the updated condexec
9152 * bits back to the CPUState for every instruction in an IT block. So:
9153 * (1) if the condexec bits are not already zero then we write
9154 * zero back into the CPUState now. This avoids complications trying
9155 * to do it at the end of the block. (For example if we don't do this
9156 * it's hard to identify whether we can safely skip writing condexec
9157 * at the end of the TB, which we definitely want to do for the case
9158 * where a TB doesn't do anything with the IT state at all.)
9159 * (2) if we are going to leave the TB then we call gen_set_condexec()
9160 * which will write the correct value into CPUState if zero is wrong.
9161 * This is done both for leaving the TB at the end, and for leaving
9162 * it because of an exception we know will happen, which is done in
9163 * gen_exception_insn(). The latter is necessary because we need to
9164 * leave the TB with the PC/IT state just prior to execution of the
9165 * instruction which caused the exception.
9166 * (3) if we leave the TB unexpectedly (eg a data abort on a load)
9167 * then the CPUState will be wrong and we need to reset it.
9168 * This is handled in the same way as restoration of the
9169 * PC in these situations: we will be called again with search_pc=1
9170 * and generate a mapping of the condexec bits for each PC in
9171 * gen_opc_condexec_bits[]. gen_pc_load[] then uses this to restore
9172 * the condexec bits.
9174 * Note that there are no instructions which can read the condexec
9175 * bits, and none which can write non-static values to them, so
9176 * we don't need to care about whether CPUState is correct in the
9180 /* Reset the conditional execution bits immediately. This avoids
9181 complications trying to do it at the end of the block. */
9182 if (dc
->condexec_mask
|| dc
->condexec_cond
)
9184 TCGv tmp
= new_tmp();
9185 tcg_gen_movi_i32(tmp
, 0);
9186 store_cpu_field(tmp
, condexec_bits
);
9189 #ifdef CONFIG_USER_ONLY
9190 /* Intercept jump to the magic kernel page. */
9191 if (dc
->pc
>= 0xffff0000) {
9192 /* We always get here via a jump, so know we are not in a
9193 conditional execution block. */
9194 gen_exception(EXCP_KERNEL_TRAP
);
9195 dc
->is_jmp
= DISAS_UPDATE
;
9199 if (dc
->pc
>= 0xfffffff0 && IS_M(env
)) {
9200 /* We always get here via a jump, so know we are not in a
9201 conditional execution block. */
9202 gen_exception(EXCP_EXCEPTION_EXIT
);
9203 dc
->is_jmp
= DISAS_UPDATE
;
9208 if (unlikely(!QTAILQ_EMPTY(&env
->breakpoints
))) {
9209 QTAILQ_FOREACH(bp
, &env
->breakpoints
, entry
) {
9210 if (bp
->pc
== dc
->pc
) {
9211 gen_exception_insn(dc
, 0, EXCP_DEBUG
);
9212 /* Advance PC so that clearing the breakpoint will
9213 invalidate this TB. */
9215 goto done_generating
;
9221 j
= gen_opc_ptr
- gen_opc_buf
;
9225 gen_opc_instr_start
[lj
++] = 0;
9227 gen_opc_pc
[lj
] = dc
->pc
;
9228 gen_opc_condexec_bits
[lj
] = (dc
->condexec_cond
<< 4) | (dc
->condexec_mask
>> 1);
9229 gen_opc_instr_start
[lj
] = 1;
9230 gen_opc_icount
[lj
] = num_insns
;
9233 if (num_insns
+ 1 == max_insns
&& (tb
->cflags
& CF_LAST_IO
))
9236 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP
))) {
9237 tcg_gen_debug_insn_start(dc
->pc
);
9241 disas_thumb_insn(env
, dc
);
9242 if (dc
->condexec_mask
) {
9243 dc
->condexec_cond
= (dc
->condexec_cond
& 0xe)
9244 | ((dc
->condexec_mask
>> 4) & 1);
9245 dc
->condexec_mask
= (dc
->condexec_mask
<< 1) & 0x1f;
9246 if (dc
->condexec_mask
== 0) {
9247 dc
->condexec_cond
= 0;
9251 disas_arm_insn(env
, dc
);
9254 fprintf(stderr
, "Internal resource leak before %08x\n", dc
->pc
);
9258 if (dc
->condjmp
&& !dc
->is_jmp
) {
9259 gen_set_label(dc
->condlabel
);
9262 /* Translation stops when a conditional branch is encountered.
9263 * Otherwise the subsequent code could get translated several times.
9264 * Also stop translation when a page boundary is reached. This
9265 * ensures prefetch aborts occur at the right place. */
9267 } while (!dc
->is_jmp
&& gen_opc_ptr
< gen_opc_end
&&
9268 !env
->singlestep_enabled
&&
9270 dc
->pc
< next_page_start
&&
9271 num_insns
< max_insns
);
9273 if (tb
->cflags
& CF_LAST_IO
) {
9275 /* FIXME: This can theoretically happen with self-modifying
9277 cpu_abort(env
, "IO on conditional branch instruction");
9282 /* At this stage dc->condjmp will only be set when the skipped
9283 instruction was a conditional branch or trap, and the PC has
9284 already been written. */
9285 if (unlikely(env
->singlestep_enabled
)) {
9286 /* Make sure the pc is updated, and raise a debug exception. */
9288 gen_set_condexec(dc
);
9289 if (dc
->is_jmp
== DISAS_SWI
) {
9290 gen_exception(EXCP_SWI
);
9292 gen_exception(EXCP_DEBUG
);
9294 gen_set_label(dc
->condlabel
);
9296 if (dc
->condjmp
|| !dc
->is_jmp
) {
9297 gen_set_pc_im(dc
->pc
);
9300 gen_set_condexec(dc
);
9301 if (dc
->is_jmp
== DISAS_SWI
&& !dc
->condjmp
) {
9302 gen_exception(EXCP_SWI
);
9304 /* FIXME: Single stepping a WFI insn will not halt
9306 gen_exception(EXCP_DEBUG
);
9309 /* While branches must always occur at the end of an IT block,
9310 there are a few other things that can cause us to terminate
9311 the TB in the middel of an IT block:
9312 - Exception generating instructions (bkpt, swi, undefined).
9314 - Hardware watchpoints.
9315 Hardware breakpoints have already been handled and skip this code.
9317 gen_set_condexec(dc
);
9318 switch(dc
->is_jmp
) {
9320 gen_goto_tb(dc
, 1, dc
->pc
);
9325 /* indicate that the hash table must be used to find the next TB */
9329 /* nothing more to generate */
9335 gen_exception(EXCP_SWI
);
9339 gen_set_label(dc
->condlabel
);
9340 gen_set_condexec(dc
);
9341 gen_goto_tb(dc
, 1, dc
->pc
);
9347 gen_icount_end(tb
, num_insns
);
9348 *gen_opc_ptr
= INDEX_op_end
;
9351 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM
)) {
9352 qemu_log("----------------\n");
9353 qemu_log("IN: %s\n", lookup_symbol(pc_start
));
9354 log_target_disas(pc_start
, dc
->pc
- pc_start
, dc
->thumb
);
9359 j
= gen_opc_ptr
- gen_opc_buf
;
9362 gen_opc_instr_start
[lj
++] = 0;
9364 tb
->size
= dc
->pc
- pc_start
;
9365 tb
->icount
= num_insns
;
9369 void gen_intermediate_code(CPUState
*env
, TranslationBlock
*tb
)
9371 gen_intermediate_code_internal(env
, tb
, 0);
9374 void gen_intermediate_code_pc(CPUState
*env
, TranslationBlock
*tb
)
9376 gen_intermediate_code_internal(env
, tb
, 1);
9379 static const char *cpu_mode_names
[16] = {
9380 "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
9381 "???", "???", "???", "und", "???", "???", "???", "sys"
9384 void cpu_dump_state(CPUState
*env
, FILE *f
, fprintf_function cpu_fprintf
,
9394 /* ??? This assumes float64 and double have the same layout.
9395 Oh well, it's only debug dumps. */
9404 cpu_fprintf(f
, "R%02d=%08x", i
, env
->regs
[i
]);
9406 cpu_fprintf(f
, "\n");
9408 cpu_fprintf(f
, " ");
9410 psr
= cpsr_read(env
);
9411 cpu_fprintf(f
, "PSR=%08x %c%c%c%c %c %s%d\n",
9413 psr
& (1 << 31) ? 'N' : '-',
9414 psr
& (1 << 30) ? 'Z' : '-',
9415 psr
& (1 << 29) ? 'C' : '-',
9416 psr
& (1 << 28) ? 'V' : '-',
9417 psr
& CPSR_T
? 'T' : 'A',
9418 cpu_mode_names
[psr
& 0xf], (psr
& 0x10) ? 32 : 26);
9421 for (i
= 0; i
< 16; i
++) {
9422 d
.d
= env
->vfp
.regs
[i
];
9426 cpu_fprintf(f
, "s%02d=%08x(%8g) s%02d=%08x(%8g) d%02d=%08x%08x(%8g)\n",
9427 i
* 2, (int)s0
.i
, s0
.s
,
9428 i
* 2 + 1, (int)s1
.i
, s1
.s
,
9429 i
, (int)(uint32_t)d
.l
.upper
, (int)(uint32_t)d
.l
.lower
,
9432 cpu_fprintf(f
, "FPSCR: %08x\n", (int)env
->vfp
.xregs
[ARM_VFP_FPSCR
]);
9436 void gen_pc_load(CPUState
*env
, TranslationBlock
*tb
,
9437 unsigned long searched_pc
, int pc_pos
, void *puc
)
9439 env
->regs
[15] = gen_opc_pc
[pc_pos
];
9440 env
->condexec_bits
= gen_opc_condexec_bits
[pc_pos
];