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_4T arm_feature(env, ARM_FEATURE_V4T)
38 #define ENABLE_ARCH_5 arm_feature(env, ARM_FEATURE_V5)
39 /* currently all emulated v5 cores are also v5TE, so don't bother */
40 #define ENABLE_ARCH_5TE arm_feature(env, ARM_FEATURE_V5)
41 #define ENABLE_ARCH_5J 0
42 #define ENABLE_ARCH_6 arm_feature(env, ARM_FEATURE_V6)
43 #define ENABLE_ARCH_6K arm_feature(env, ARM_FEATURE_V6K)
44 #define ENABLE_ARCH_6T2 arm_feature(env, ARM_FEATURE_THUMB2)
45 #define ENABLE_ARCH_7 arm_feature(env, ARM_FEATURE_V7)
47 #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0)
49 /* internal defines */
50 typedef struct DisasContext
{
53 /* Nonzero if this instruction has been conditionally skipped. */
55 /* The label that will be jumped to when the instruction is skipped. */
57 /* Thumb-2 condtional execution bits. */
60 struct TranslationBlock
*tb
;
61 int singlestep_enabled
;
63 #if !defined(CONFIG_USER_ONLY)
71 static uint32_t gen_opc_condexec_bits
[OPC_BUF_SIZE
];
73 #if defined(CONFIG_USER_ONLY)
76 #define IS_USER(s) (s->user)
79 /* These instructions trap after executing, so defer them until after the
80 conditional executions state has been updated. */
84 static TCGv_ptr cpu_env
;
85 /* We reuse the same 64-bit temporaries for efficiency. */
86 static TCGv_i64 cpu_V0
, cpu_V1
, cpu_M0
;
87 static TCGv_i32 cpu_R
[16];
88 static TCGv_i32 cpu_exclusive_addr
;
89 static TCGv_i32 cpu_exclusive_val
;
90 static TCGv_i32 cpu_exclusive_high
;
91 #ifdef CONFIG_USER_ONLY
92 static TCGv_i32 cpu_exclusive_test
;
93 static TCGv_i32 cpu_exclusive_info
;
96 /* FIXME: These should be removed. */
97 static TCGv cpu_F0s
, cpu_F1s
;
98 static TCGv_i64 cpu_F0d
, cpu_F1d
;
100 #include "gen-icount.h"
102 static const char *regnames
[] =
103 { "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
104 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "pc" };
106 /* initialize TCG globals. */
107 void arm_translate_init(void)
111 cpu_env
= tcg_global_reg_new_ptr(TCG_AREG0
, "env");
113 for (i
= 0; i
< 16; i
++) {
114 cpu_R
[i
] = tcg_global_mem_new_i32(TCG_AREG0
,
115 offsetof(CPUState
, regs
[i
]),
118 cpu_exclusive_addr
= tcg_global_mem_new_i32(TCG_AREG0
,
119 offsetof(CPUState
, exclusive_addr
), "exclusive_addr");
120 cpu_exclusive_val
= tcg_global_mem_new_i32(TCG_AREG0
,
121 offsetof(CPUState
, exclusive_val
), "exclusive_val");
122 cpu_exclusive_high
= tcg_global_mem_new_i32(TCG_AREG0
,
123 offsetof(CPUState
, exclusive_high
), "exclusive_high");
124 #ifdef CONFIG_USER_ONLY
125 cpu_exclusive_test
= tcg_global_mem_new_i32(TCG_AREG0
,
126 offsetof(CPUState
, exclusive_test
), "exclusive_test");
127 cpu_exclusive_info
= tcg_global_mem_new_i32(TCG_AREG0
,
128 offsetof(CPUState
, exclusive_info
), "exclusive_info");
135 static inline TCGv
load_cpu_offset(int offset
)
137 TCGv tmp
= tcg_temp_new_i32();
138 tcg_gen_ld_i32(tmp
, cpu_env
, offset
);
142 #define load_cpu_field(name) load_cpu_offset(offsetof(CPUState, name))
144 static inline void store_cpu_offset(TCGv var
, int offset
)
146 tcg_gen_st_i32(var
, cpu_env
, offset
);
147 tcg_temp_free_i32(var
);
150 #define store_cpu_field(var, name) \
151 store_cpu_offset(var, offsetof(CPUState, name))
153 /* Set a variable to the value of a CPU register. */
154 static void load_reg_var(DisasContext
*s
, TCGv var
, int reg
)
158 /* normaly, since we updated PC, we need only to add one insn */
160 addr
= (long)s
->pc
+ 2;
162 addr
= (long)s
->pc
+ 4;
163 tcg_gen_movi_i32(var
, addr
);
165 tcg_gen_mov_i32(var
, cpu_R
[reg
]);
169 /* Create a new temporary and set it to the value of a CPU register. */
170 static inline TCGv
load_reg(DisasContext
*s
, int reg
)
172 TCGv tmp
= tcg_temp_new_i32();
173 load_reg_var(s
, tmp
, reg
);
177 /* Set a CPU register. The source must be a temporary and will be
179 static void store_reg(DisasContext
*s
, int reg
, TCGv var
)
182 tcg_gen_andi_i32(var
, var
, ~1);
183 s
->is_jmp
= DISAS_JUMP
;
185 tcg_gen_mov_i32(cpu_R
[reg
], var
);
186 tcg_temp_free_i32(var
);
189 /* Value extensions. */
190 #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var)
191 #define gen_uxth(var) tcg_gen_ext16u_i32(var, var)
192 #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
193 #define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
195 #define gen_sxtb16(var) gen_helper_sxtb16(var, var)
196 #define gen_uxtb16(var) gen_helper_uxtb16(var, var)
199 static inline void gen_set_cpsr(TCGv var
, uint32_t mask
)
201 TCGv tmp_mask
= tcg_const_i32(mask
);
202 gen_helper_cpsr_write(var
, tmp_mask
);
203 tcg_temp_free_i32(tmp_mask
);
205 /* Set NZCV flags from the high 4 bits of var. */
206 #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV)
208 static void gen_exception(int excp
)
210 TCGv tmp
= tcg_temp_new_i32();
211 tcg_gen_movi_i32(tmp
, excp
);
212 gen_helper_exception(tmp
);
213 tcg_temp_free_i32(tmp
);
216 static void gen_smul_dual(TCGv a
, TCGv b
)
218 TCGv tmp1
= tcg_temp_new_i32();
219 TCGv tmp2
= tcg_temp_new_i32();
220 tcg_gen_ext16s_i32(tmp1
, a
);
221 tcg_gen_ext16s_i32(tmp2
, b
);
222 tcg_gen_mul_i32(tmp1
, tmp1
, tmp2
);
223 tcg_temp_free_i32(tmp2
);
224 tcg_gen_sari_i32(a
, a
, 16);
225 tcg_gen_sari_i32(b
, b
, 16);
226 tcg_gen_mul_i32(b
, b
, a
);
227 tcg_gen_mov_i32(a
, tmp1
);
228 tcg_temp_free_i32(tmp1
);
231 /* Byteswap each halfword. */
232 static void gen_rev16(TCGv var
)
234 TCGv tmp
= tcg_temp_new_i32();
235 tcg_gen_shri_i32(tmp
, var
, 8);
236 tcg_gen_andi_i32(tmp
, tmp
, 0x00ff00ff);
237 tcg_gen_shli_i32(var
, var
, 8);
238 tcg_gen_andi_i32(var
, var
, 0xff00ff00);
239 tcg_gen_or_i32(var
, var
, tmp
);
240 tcg_temp_free_i32(tmp
);
243 /* Byteswap low halfword and sign extend. */
244 static void gen_revsh(TCGv var
)
246 tcg_gen_ext16u_i32(var
, var
);
247 tcg_gen_bswap16_i32(var
, var
);
248 tcg_gen_ext16s_i32(var
, var
);
251 /* Unsigned bitfield extract. */
252 static void gen_ubfx(TCGv var
, int shift
, uint32_t mask
)
255 tcg_gen_shri_i32(var
, var
, shift
);
256 tcg_gen_andi_i32(var
, var
, mask
);
259 /* Signed bitfield extract. */
260 static void gen_sbfx(TCGv var
, int shift
, int width
)
265 tcg_gen_sari_i32(var
, var
, shift
);
266 if (shift
+ width
< 32) {
267 signbit
= 1u << (width
- 1);
268 tcg_gen_andi_i32(var
, var
, (1u << width
) - 1);
269 tcg_gen_xori_i32(var
, var
, signbit
);
270 tcg_gen_subi_i32(var
, var
, signbit
);
274 /* Bitfield insertion. Insert val into base. Clobbers base and val. */
275 static void gen_bfi(TCGv dest
, TCGv base
, TCGv val
, int shift
, uint32_t mask
)
277 tcg_gen_andi_i32(val
, val
, mask
);
278 tcg_gen_shli_i32(val
, val
, shift
);
279 tcg_gen_andi_i32(base
, base
, ~(mask
<< shift
));
280 tcg_gen_or_i32(dest
, base
, val
);
283 /* Return (b << 32) + a. Mark inputs as dead */
284 static TCGv_i64
gen_addq_msw(TCGv_i64 a
, TCGv b
)
286 TCGv_i64 tmp64
= tcg_temp_new_i64();
288 tcg_gen_extu_i32_i64(tmp64
, b
);
289 tcg_temp_free_i32(b
);
290 tcg_gen_shli_i64(tmp64
, tmp64
, 32);
291 tcg_gen_add_i64(a
, tmp64
, a
);
293 tcg_temp_free_i64(tmp64
);
297 /* Return (b << 32) - a. Mark inputs as dead. */
298 static TCGv_i64
gen_subq_msw(TCGv_i64 a
, TCGv b
)
300 TCGv_i64 tmp64
= tcg_temp_new_i64();
302 tcg_gen_extu_i32_i64(tmp64
, b
);
303 tcg_temp_free_i32(b
);
304 tcg_gen_shli_i64(tmp64
, tmp64
, 32);
305 tcg_gen_sub_i64(a
, tmp64
, a
);
307 tcg_temp_free_i64(tmp64
);
311 /* FIXME: Most targets have native widening multiplication.
312 It would be good to use that instead of a full wide multiply. */
313 /* 32x32->64 multiply. Marks inputs as dead. */
314 static TCGv_i64
gen_mulu_i64_i32(TCGv a
, TCGv b
)
316 TCGv_i64 tmp1
= tcg_temp_new_i64();
317 TCGv_i64 tmp2
= tcg_temp_new_i64();
319 tcg_gen_extu_i32_i64(tmp1
, a
);
320 tcg_temp_free_i32(a
);
321 tcg_gen_extu_i32_i64(tmp2
, b
);
322 tcg_temp_free_i32(b
);
323 tcg_gen_mul_i64(tmp1
, tmp1
, tmp2
);
324 tcg_temp_free_i64(tmp2
);
328 static TCGv_i64
gen_muls_i64_i32(TCGv a
, TCGv b
)
330 TCGv_i64 tmp1
= tcg_temp_new_i64();
331 TCGv_i64 tmp2
= tcg_temp_new_i64();
333 tcg_gen_ext_i32_i64(tmp1
, a
);
334 tcg_temp_free_i32(a
);
335 tcg_gen_ext_i32_i64(tmp2
, b
);
336 tcg_temp_free_i32(b
);
337 tcg_gen_mul_i64(tmp1
, tmp1
, tmp2
);
338 tcg_temp_free_i64(tmp2
);
342 /* Swap low and high halfwords. */
343 static void gen_swap_half(TCGv var
)
345 TCGv tmp
= tcg_temp_new_i32();
346 tcg_gen_shri_i32(tmp
, var
, 16);
347 tcg_gen_shli_i32(var
, var
, 16);
348 tcg_gen_or_i32(var
, var
, tmp
);
349 tcg_temp_free_i32(tmp
);
352 /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
353 tmp = (t0 ^ t1) & 0x8000;
356 t0 = (t0 + t1) ^ tmp;
359 static void gen_add16(TCGv t0
, TCGv t1
)
361 TCGv tmp
= tcg_temp_new_i32();
362 tcg_gen_xor_i32(tmp
, t0
, t1
);
363 tcg_gen_andi_i32(tmp
, tmp
, 0x8000);
364 tcg_gen_andi_i32(t0
, t0
, ~0x8000);
365 tcg_gen_andi_i32(t1
, t1
, ~0x8000);
366 tcg_gen_add_i32(t0
, t0
, t1
);
367 tcg_gen_xor_i32(t0
, t0
, tmp
);
368 tcg_temp_free_i32(tmp
);
369 tcg_temp_free_i32(t1
);
372 #define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, CF))
374 /* Set CF to the top bit of var. */
375 static void gen_set_CF_bit31(TCGv var
)
377 TCGv tmp
= tcg_temp_new_i32();
378 tcg_gen_shri_i32(tmp
, var
, 31);
380 tcg_temp_free_i32(tmp
);
383 /* Set N and Z flags from var. */
384 static inline void gen_logic_CC(TCGv var
)
386 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUState
, NF
));
387 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUState
, ZF
));
391 static void gen_adc(TCGv t0
, TCGv t1
)
394 tcg_gen_add_i32(t0
, t0
, t1
);
395 tmp
= load_cpu_field(CF
);
396 tcg_gen_add_i32(t0
, t0
, tmp
);
397 tcg_temp_free_i32(tmp
);
400 /* dest = T0 + T1 + CF. */
401 static void gen_add_carry(TCGv dest
, TCGv t0
, TCGv t1
)
404 tcg_gen_add_i32(dest
, t0
, t1
);
405 tmp
= load_cpu_field(CF
);
406 tcg_gen_add_i32(dest
, dest
, tmp
);
407 tcg_temp_free_i32(tmp
);
410 /* dest = T0 - T1 + CF - 1. */
411 static void gen_sub_carry(TCGv dest
, TCGv t0
, TCGv t1
)
414 tcg_gen_sub_i32(dest
, t0
, t1
);
415 tmp
= load_cpu_field(CF
);
416 tcg_gen_add_i32(dest
, dest
, tmp
);
417 tcg_gen_subi_i32(dest
, dest
, 1);
418 tcg_temp_free_i32(tmp
);
421 /* FIXME: Implement this natively. */
422 #define tcg_gen_abs_i32(t0, t1) gen_helper_abs(t0, t1)
424 static void shifter_out_im(TCGv var
, int shift
)
426 TCGv tmp
= tcg_temp_new_i32();
428 tcg_gen_andi_i32(tmp
, var
, 1);
430 tcg_gen_shri_i32(tmp
, var
, shift
);
432 tcg_gen_andi_i32(tmp
, tmp
, 1);
435 tcg_temp_free_i32(tmp
);
438 /* Shift by immediate. Includes special handling for shift == 0. */
439 static inline void gen_arm_shift_im(TCGv var
, int shiftop
, int shift
, int flags
)
445 shifter_out_im(var
, 32 - shift
);
446 tcg_gen_shli_i32(var
, var
, shift
);
452 tcg_gen_shri_i32(var
, var
, 31);
455 tcg_gen_movi_i32(var
, 0);
458 shifter_out_im(var
, shift
- 1);
459 tcg_gen_shri_i32(var
, var
, shift
);
466 shifter_out_im(var
, shift
- 1);
469 tcg_gen_sari_i32(var
, var
, shift
);
471 case 3: /* ROR/RRX */
474 shifter_out_im(var
, shift
- 1);
475 tcg_gen_rotri_i32(var
, var
, shift
); break;
477 TCGv tmp
= load_cpu_field(CF
);
479 shifter_out_im(var
, 0);
480 tcg_gen_shri_i32(var
, var
, 1);
481 tcg_gen_shli_i32(tmp
, tmp
, 31);
482 tcg_gen_or_i32(var
, var
, tmp
);
483 tcg_temp_free_i32(tmp
);
488 static inline void gen_arm_shift_reg(TCGv var
, int shiftop
,
489 TCGv shift
, int flags
)
493 case 0: gen_helper_shl_cc(var
, var
, shift
); break;
494 case 1: gen_helper_shr_cc(var
, var
, shift
); break;
495 case 2: gen_helper_sar_cc(var
, var
, shift
); break;
496 case 3: gen_helper_ror_cc(var
, var
, shift
); break;
500 case 0: gen_helper_shl(var
, var
, shift
); break;
501 case 1: gen_helper_shr(var
, var
, shift
); break;
502 case 2: gen_helper_sar(var
, var
, shift
); break;
503 case 3: tcg_gen_andi_i32(shift
, shift
, 0x1f);
504 tcg_gen_rotr_i32(var
, var
, shift
); break;
507 tcg_temp_free_i32(shift
);
510 #define PAS_OP(pfx) \
512 case 0: gen_pas_helper(glue(pfx,add16)); break; \
513 case 1: gen_pas_helper(glue(pfx,addsubx)); break; \
514 case 2: gen_pas_helper(glue(pfx,subaddx)); break; \
515 case 3: gen_pas_helper(glue(pfx,sub16)); break; \
516 case 4: gen_pas_helper(glue(pfx,add8)); break; \
517 case 7: gen_pas_helper(glue(pfx,sub8)); break; \
519 static void gen_arm_parallel_addsub(int op1
, int op2
, TCGv a
, TCGv b
)
524 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
526 tmp
= tcg_temp_new_ptr();
527 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
529 tcg_temp_free_ptr(tmp
);
532 tmp
= tcg_temp_new_ptr();
533 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
535 tcg_temp_free_ptr(tmp
);
537 #undef gen_pas_helper
538 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
551 #undef gen_pas_helper
556 /* For unknown reasons Arm and Thumb-2 use arbitrarily different encodings. */
557 #define PAS_OP(pfx) \
559 case 0: gen_pas_helper(glue(pfx,add8)); break; \
560 case 1: gen_pas_helper(glue(pfx,add16)); break; \
561 case 2: gen_pas_helper(glue(pfx,addsubx)); break; \
562 case 4: gen_pas_helper(glue(pfx,sub8)); break; \
563 case 5: gen_pas_helper(glue(pfx,sub16)); break; \
564 case 6: gen_pas_helper(glue(pfx,subaddx)); break; \
566 static void gen_thumb2_parallel_addsub(int op1
, int op2
, TCGv a
, TCGv b
)
571 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
573 tmp
= tcg_temp_new_ptr();
574 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
576 tcg_temp_free_ptr(tmp
);
579 tmp
= tcg_temp_new_ptr();
580 tcg_gen_addi_ptr(tmp
, cpu_env
, offsetof(CPUState
, GE
));
582 tcg_temp_free_ptr(tmp
);
584 #undef gen_pas_helper
585 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
598 #undef gen_pas_helper
603 static void gen_test_cc(int cc
, int label
)
611 tmp
= load_cpu_field(ZF
);
612 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
615 tmp
= load_cpu_field(ZF
);
616 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, label
);
619 tmp
= load_cpu_field(CF
);
620 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, label
);
623 tmp
= load_cpu_field(CF
);
624 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
627 tmp
= load_cpu_field(NF
);
628 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
631 tmp
= load_cpu_field(NF
);
632 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
635 tmp
= load_cpu_field(VF
);
636 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
639 tmp
= load_cpu_field(VF
);
640 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
642 case 8: /* hi: C && !Z */
643 inv
= gen_new_label();
644 tmp
= load_cpu_field(CF
);
645 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, inv
);
646 tcg_temp_free_i32(tmp
);
647 tmp
= load_cpu_field(ZF
);
648 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, label
);
651 case 9: /* ls: !C || Z */
652 tmp
= load_cpu_field(CF
);
653 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
654 tcg_temp_free_i32(tmp
);
655 tmp
= load_cpu_field(ZF
);
656 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
658 case 10: /* ge: N == V -> N ^ V == 0 */
659 tmp
= load_cpu_field(VF
);
660 tmp2
= load_cpu_field(NF
);
661 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
662 tcg_temp_free_i32(tmp2
);
663 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
665 case 11: /* lt: N != V -> N ^ V != 0 */
666 tmp
= load_cpu_field(VF
);
667 tmp2
= load_cpu_field(NF
);
668 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
669 tcg_temp_free_i32(tmp2
);
670 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
672 case 12: /* gt: !Z && N == V */
673 inv
= gen_new_label();
674 tmp
= load_cpu_field(ZF
);
675 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, inv
);
676 tcg_temp_free_i32(tmp
);
677 tmp
= load_cpu_field(VF
);
678 tmp2
= load_cpu_field(NF
);
679 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
680 tcg_temp_free_i32(tmp2
);
681 tcg_gen_brcondi_i32(TCG_COND_GE
, tmp
, 0, label
);
684 case 13: /* le: Z || N != V */
685 tmp
= load_cpu_field(ZF
);
686 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, label
);
687 tcg_temp_free_i32(tmp
);
688 tmp
= load_cpu_field(VF
);
689 tmp2
= load_cpu_field(NF
);
690 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
691 tcg_temp_free_i32(tmp2
);
692 tcg_gen_brcondi_i32(TCG_COND_LT
, tmp
, 0, label
);
695 fprintf(stderr
, "Bad condition code 0x%x\n", cc
);
698 tcg_temp_free_i32(tmp
);
701 static const uint8_t table_logic_cc
[16] = {
720 /* Set PC and Thumb state from an immediate address. */
721 static inline void gen_bx_im(DisasContext
*s
, uint32_t addr
)
725 s
->is_jmp
= DISAS_UPDATE
;
726 if (s
->thumb
!= (addr
& 1)) {
727 tmp
= tcg_temp_new_i32();
728 tcg_gen_movi_i32(tmp
, addr
& 1);
729 tcg_gen_st_i32(tmp
, cpu_env
, offsetof(CPUState
, thumb
));
730 tcg_temp_free_i32(tmp
);
732 tcg_gen_movi_i32(cpu_R
[15], addr
& ~1);
735 /* Set PC and Thumb state from var. var is marked as dead. */
736 static inline void gen_bx(DisasContext
*s
, TCGv var
)
738 s
->is_jmp
= DISAS_UPDATE
;
739 tcg_gen_andi_i32(cpu_R
[15], var
, ~1);
740 tcg_gen_andi_i32(var
, var
, 1);
741 store_cpu_field(var
, thumb
);
744 /* Variant of store_reg which uses branch&exchange logic when storing
745 to r15 in ARM architecture v7 and above. The source must be a temporary
746 and will be marked as dead. */
747 static inline void store_reg_bx(CPUState
*env
, DisasContext
*s
,
750 if (reg
== 15 && ENABLE_ARCH_7
) {
753 store_reg(s
, reg
, var
);
757 /* Variant of store_reg which uses branch&exchange logic when storing
758 * to r15 in ARM architecture v5T and above. This is used for storing
759 * the results of a LDR/LDM/POP into r15, and corresponds to the cases
760 * in the ARM ARM which use the LoadWritePC() pseudocode function. */
761 static inline void store_reg_from_load(CPUState
*env
, DisasContext
*s
,
764 if (reg
== 15 && ENABLE_ARCH_5
) {
767 store_reg(s
, reg
, var
);
771 static inline TCGv
gen_ld8s(TCGv addr
, int index
)
773 TCGv tmp
= tcg_temp_new_i32();
774 tcg_gen_qemu_ld8s(tmp
, addr
, index
);
777 static inline TCGv
gen_ld8u(TCGv addr
, int index
)
779 TCGv tmp
= tcg_temp_new_i32();
780 tcg_gen_qemu_ld8u(tmp
, addr
, index
);
783 static inline TCGv
gen_ld16s(TCGv addr
, int index
)
785 TCGv tmp
= tcg_temp_new_i32();
786 tcg_gen_qemu_ld16s(tmp
, addr
, index
);
789 static inline TCGv
gen_ld16u(TCGv addr
, int index
)
791 TCGv tmp
= tcg_temp_new_i32();
792 tcg_gen_qemu_ld16u(tmp
, addr
, index
);
795 static inline TCGv
gen_ld32(TCGv addr
, int index
)
797 TCGv tmp
= tcg_temp_new_i32();
798 tcg_gen_qemu_ld32u(tmp
, addr
, index
);
801 static inline TCGv_i64
gen_ld64(TCGv addr
, int index
)
803 TCGv_i64 tmp
= tcg_temp_new_i64();
804 tcg_gen_qemu_ld64(tmp
, addr
, index
);
807 static inline void gen_st8(TCGv val
, TCGv addr
, int index
)
809 tcg_gen_qemu_st8(val
, addr
, index
);
810 tcg_temp_free_i32(val
);
812 static inline void gen_st16(TCGv val
, TCGv addr
, int index
)
814 tcg_gen_qemu_st16(val
, addr
, index
);
815 tcg_temp_free_i32(val
);
817 static inline void gen_st32(TCGv val
, TCGv addr
, int index
)
819 tcg_gen_qemu_st32(val
, addr
, index
);
820 tcg_temp_free_i32(val
);
822 static inline void gen_st64(TCGv_i64 val
, TCGv addr
, int index
)
824 tcg_gen_qemu_st64(val
, addr
, index
);
825 tcg_temp_free_i64(val
);
828 static inline void gen_set_pc_im(uint32_t val
)
830 tcg_gen_movi_i32(cpu_R
[15], val
);
833 /* Force a TB lookup after an instruction that changes the CPU state. */
834 static inline void gen_lookup_tb(DisasContext
*s
)
836 tcg_gen_movi_i32(cpu_R
[15], s
->pc
& ~1);
837 s
->is_jmp
= DISAS_UPDATE
;
840 static inline void gen_add_data_offset(DisasContext
*s
, unsigned int insn
,
843 int val
, rm
, shift
, shiftop
;
846 if (!(insn
& (1 << 25))) {
849 if (!(insn
& (1 << 23)))
852 tcg_gen_addi_i32(var
, var
, val
);
856 shift
= (insn
>> 7) & 0x1f;
857 shiftop
= (insn
>> 5) & 3;
858 offset
= load_reg(s
, rm
);
859 gen_arm_shift_im(offset
, shiftop
, shift
, 0);
860 if (!(insn
& (1 << 23)))
861 tcg_gen_sub_i32(var
, var
, offset
);
863 tcg_gen_add_i32(var
, var
, offset
);
864 tcg_temp_free_i32(offset
);
868 static inline void gen_add_datah_offset(DisasContext
*s
, unsigned int insn
,
874 if (insn
& (1 << 22)) {
876 val
= (insn
& 0xf) | ((insn
>> 4) & 0xf0);
877 if (!(insn
& (1 << 23)))
881 tcg_gen_addi_i32(var
, var
, val
);
885 tcg_gen_addi_i32(var
, var
, extra
);
887 offset
= load_reg(s
, rm
);
888 if (!(insn
& (1 << 23)))
889 tcg_gen_sub_i32(var
, var
, offset
);
891 tcg_gen_add_i32(var
, var
, offset
);
892 tcg_temp_free_i32(offset
);
896 #define VFP_OP2(name) \
897 static inline void gen_vfp_##name(int dp) \
900 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, cpu_env); \
902 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, cpu_env); \
912 static inline void gen_vfp_abs(int dp
)
915 gen_helper_vfp_absd(cpu_F0d
, cpu_F0d
);
917 gen_helper_vfp_abss(cpu_F0s
, cpu_F0s
);
920 static inline void gen_vfp_neg(int dp
)
923 gen_helper_vfp_negd(cpu_F0d
, cpu_F0d
);
925 gen_helper_vfp_negs(cpu_F0s
, cpu_F0s
);
928 static inline void gen_vfp_sqrt(int dp
)
931 gen_helper_vfp_sqrtd(cpu_F0d
, cpu_F0d
, cpu_env
);
933 gen_helper_vfp_sqrts(cpu_F0s
, cpu_F0s
, cpu_env
);
936 static inline void gen_vfp_cmp(int dp
)
939 gen_helper_vfp_cmpd(cpu_F0d
, cpu_F1d
, cpu_env
);
941 gen_helper_vfp_cmps(cpu_F0s
, cpu_F1s
, cpu_env
);
944 static inline void gen_vfp_cmpe(int dp
)
947 gen_helper_vfp_cmped(cpu_F0d
, cpu_F1d
, cpu_env
);
949 gen_helper_vfp_cmpes(cpu_F0s
, cpu_F1s
, cpu_env
);
952 static inline void gen_vfp_F1_ld0(int dp
)
955 tcg_gen_movi_i64(cpu_F1d
, 0);
957 tcg_gen_movi_i32(cpu_F1s
, 0);
960 static inline void gen_vfp_uito(int dp
)
963 gen_helper_vfp_uitod(cpu_F0d
, cpu_F0s
, cpu_env
);
965 gen_helper_vfp_uitos(cpu_F0s
, cpu_F0s
, cpu_env
);
968 static inline void gen_vfp_sito(int dp
)
971 gen_helper_vfp_sitod(cpu_F0d
, cpu_F0s
, cpu_env
);
973 gen_helper_vfp_sitos(cpu_F0s
, cpu_F0s
, cpu_env
);
976 static inline void gen_vfp_toui(int dp
)
979 gen_helper_vfp_touid(cpu_F0s
, cpu_F0d
, cpu_env
);
981 gen_helper_vfp_touis(cpu_F0s
, cpu_F0s
, cpu_env
);
984 static inline void gen_vfp_touiz(int dp
)
987 gen_helper_vfp_touizd(cpu_F0s
, cpu_F0d
, cpu_env
);
989 gen_helper_vfp_touizs(cpu_F0s
, cpu_F0s
, cpu_env
);
992 static inline void gen_vfp_tosi(int dp
)
995 gen_helper_vfp_tosid(cpu_F0s
, cpu_F0d
, cpu_env
);
997 gen_helper_vfp_tosis(cpu_F0s
, cpu_F0s
, cpu_env
);
1000 static inline void gen_vfp_tosiz(int dp
)
1003 gen_helper_vfp_tosizd(cpu_F0s
, cpu_F0d
, cpu_env
);
1005 gen_helper_vfp_tosizs(cpu_F0s
, cpu_F0s
, cpu_env
);
1008 #define VFP_GEN_FIX(name) \
1009 static inline void gen_vfp_##name(int dp, int shift) \
1011 TCGv tmp_shift = tcg_const_i32(shift); \
1013 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tmp_shift, cpu_env);\
1015 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, tmp_shift, cpu_env);\
1016 tcg_temp_free_i32(tmp_shift); \
1028 static inline void gen_vfp_ld(DisasContext
*s
, int dp
, TCGv addr
)
1031 tcg_gen_qemu_ld64(cpu_F0d
, addr
, IS_USER(s
));
1033 tcg_gen_qemu_ld32u(cpu_F0s
, addr
, IS_USER(s
));
1036 static inline void gen_vfp_st(DisasContext
*s
, int dp
, TCGv addr
)
1039 tcg_gen_qemu_st64(cpu_F0d
, addr
, IS_USER(s
));
1041 tcg_gen_qemu_st32(cpu_F0s
, addr
, IS_USER(s
));
1045 vfp_reg_offset (int dp
, int reg
)
1048 return offsetof(CPUARMState
, vfp
.regs
[reg
]);
1050 return offsetof(CPUARMState
, vfp
.regs
[reg
>> 1])
1051 + offsetof(CPU_DoubleU
, l
.upper
);
1053 return offsetof(CPUARMState
, vfp
.regs
[reg
>> 1])
1054 + offsetof(CPU_DoubleU
, l
.lower
);
1058 /* Return the offset of a 32-bit piece of a NEON register.
1059 zero is the least significant end of the register. */
1061 neon_reg_offset (int reg
, int n
)
1065 return vfp_reg_offset(0, sreg
);
1068 static TCGv
neon_load_reg(int reg
, int pass
)
1070 TCGv tmp
= tcg_temp_new_i32();
1071 tcg_gen_ld_i32(tmp
, cpu_env
, neon_reg_offset(reg
, pass
));
1075 static void neon_store_reg(int reg
, int pass
, TCGv var
)
1077 tcg_gen_st_i32(var
, cpu_env
, neon_reg_offset(reg
, pass
));
1078 tcg_temp_free_i32(var
);
1081 static inline void neon_load_reg64(TCGv_i64 var
, int reg
)
1083 tcg_gen_ld_i64(var
, cpu_env
, vfp_reg_offset(1, reg
));
1086 static inline void neon_store_reg64(TCGv_i64 var
, int reg
)
1088 tcg_gen_st_i64(var
, cpu_env
, vfp_reg_offset(1, reg
));
1091 #define tcg_gen_ld_f32 tcg_gen_ld_i32
1092 #define tcg_gen_ld_f64 tcg_gen_ld_i64
1093 #define tcg_gen_st_f32 tcg_gen_st_i32
1094 #define tcg_gen_st_f64 tcg_gen_st_i64
1096 static inline void gen_mov_F0_vreg(int dp
, int reg
)
1099 tcg_gen_ld_f64(cpu_F0d
, cpu_env
, vfp_reg_offset(dp
, reg
));
1101 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, vfp_reg_offset(dp
, reg
));
1104 static inline void gen_mov_F1_vreg(int dp
, int reg
)
1107 tcg_gen_ld_f64(cpu_F1d
, cpu_env
, vfp_reg_offset(dp
, reg
));
1109 tcg_gen_ld_f32(cpu_F1s
, cpu_env
, vfp_reg_offset(dp
, reg
));
1112 static inline void gen_mov_vreg_F0(int dp
, int reg
)
1115 tcg_gen_st_f64(cpu_F0d
, cpu_env
, vfp_reg_offset(dp
, reg
));
1117 tcg_gen_st_f32(cpu_F0s
, cpu_env
, vfp_reg_offset(dp
, reg
));
1120 #define ARM_CP_RW_BIT (1 << 20)
1122 static inline void iwmmxt_load_reg(TCGv_i64 var
, int reg
)
1124 tcg_gen_ld_i64(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.regs
[reg
]));
1127 static inline void iwmmxt_store_reg(TCGv_i64 var
, int reg
)
1129 tcg_gen_st_i64(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.regs
[reg
]));
1132 static inline TCGv
iwmmxt_load_creg(int reg
)
1134 TCGv var
= tcg_temp_new_i32();
1135 tcg_gen_ld_i32(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.cregs
[reg
]));
1139 static inline void iwmmxt_store_creg(int reg
, TCGv var
)
1141 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUState
, iwmmxt
.cregs
[reg
]));
1142 tcg_temp_free_i32(var
);
1145 static inline void gen_op_iwmmxt_movq_wRn_M0(int rn
)
1147 iwmmxt_store_reg(cpu_M0
, rn
);
1150 static inline void gen_op_iwmmxt_movq_M0_wRn(int rn
)
1152 iwmmxt_load_reg(cpu_M0
, rn
);
1155 static inline void gen_op_iwmmxt_orq_M0_wRn(int rn
)
1157 iwmmxt_load_reg(cpu_V1
, rn
);
1158 tcg_gen_or_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1161 static inline void gen_op_iwmmxt_andq_M0_wRn(int rn
)
1163 iwmmxt_load_reg(cpu_V1
, rn
);
1164 tcg_gen_and_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1167 static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn
)
1169 iwmmxt_load_reg(cpu_V1
, rn
);
1170 tcg_gen_xor_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1173 #define IWMMXT_OP(name) \
1174 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1176 iwmmxt_load_reg(cpu_V1, rn); \
1177 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
1180 #define IWMMXT_OP_SIZE(name) \
1181 IWMMXT_OP(name##b) \
1182 IWMMXT_OP(name##w) \
1185 #define IWMMXT_OP_1(name) \
1186 static inline void gen_op_iwmmxt_##name##_M0(void) \
1188 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0); \
1202 IWMMXT_OP_SIZE(unpackl
)
1203 IWMMXT_OP_SIZE(unpackh
)
1205 IWMMXT_OP_1(unpacklub
)
1206 IWMMXT_OP_1(unpackluw
)
1207 IWMMXT_OP_1(unpacklul
)
1208 IWMMXT_OP_1(unpackhub
)
1209 IWMMXT_OP_1(unpackhuw
)
1210 IWMMXT_OP_1(unpackhul
)
1211 IWMMXT_OP_1(unpacklsb
)
1212 IWMMXT_OP_1(unpacklsw
)
1213 IWMMXT_OP_1(unpacklsl
)
1214 IWMMXT_OP_1(unpackhsb
)
1215 IWMMXT_OP_1(unpackhsw
)
1216 IWMMXT_OP_1(unpackhsl
)
1218 IWMMXT_OP_SIZE(cmpeq
)
1219 IWMMXT_OP_SIZE(cmpgtu
)
1220 IWMMXT_OP_SIZE(cmpgts
)
1222 IWMMXT_OP_SIZE(mins
)
1223 IWMMXT_OP_SIZE(minu
)
1224 IWMMXT_OP_SIZE(maxs
)
1225 IWMMXT_OP_SIZE(maxu
)
1227 IWMMXT_OP_SIZE(subn
)
1228 IWMMXT_OP_SIZE(addn
)
1229 IWMMXT_OP_SIZE(subu
)
1230 IWMMXT_OP_SIZE(addu
)
1231 IWMMXT_OP_SIZE(subs
)
1232 IWMMXT_OP_SIZE(adds
)
1248 static void gen_op_iwmmxt_set_mup(void)
1251 tmp
= load_cpu_field(iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1252 tcg_gen_ori_i32(tmp
, tmp
, 2);
1253 store_cpu_field(tmp
, iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1256 static void gen_op_iwmmxt_set_cup(void)
1259 tmp
= load_cpu_field(iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1260 tcg_gen_ori_i32(tmp
, tmp
, 1);
1261 store_cpu_field(tmp
, iwmmxt
.cregs
[ARM_IWMMXT_wCon
]);
1264 static void gen_op_iwmmxt_setpsr_nz(void)
1266 TCGv tmp
= tcg_temp_new_i32();
1267 gen_helper_iwmmxt_setpsr_nz(tmp
, cpu_M0
);
1268 store_cpu_field(tmp
, iwmmxt
.cregs
[ARM_IWMMXT_wCASF
]);
1271 static inline void gen_op_iwmmxt_addl_M0_wRn(int rn
)
1273 iwmmxt_load_reg(cpu_V1
, rn
);
1274 tcg_gen_ext32u_i64(cpu_V1
, cpu_V1
);
1275 tcg_gen_add_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1278 static inline int gen_iwmmxt_address(DisasContext
*s
, uint32_t insn
, TCGv dest
)
1284 rd
= (insn
>> 16) & 0xf;
1285 tmp
= load_reg(s
, rd
);
1287 offset
= (insn
& 0xff) << ((insn
>> 7) & 2);
1288 if (insn
& (1 << 24)) {
1290 if (insn
& (1 << 23))
1291 tcg_gen_addi_i32(tmp
, tmp
, offset
);
1293 tcg_gen_addi_i32(tmp
, tmp
, -offset
);
1294 tcg_gen_mov_i32(dest
, tmp
);
1295 if (insn
& (1 << 21))
1296 store_reg(s
, rd
, tmp
);
1298 tcg_temp_free_i32(tmp
);
1299 } else if (insn
& (1 << 21)) {
1301 tcg_gen_mov_i32(dest
, tmp
);
1302 if (insn
& (1 << 23))
1303 tcg_gen_addi_i32(tmp
, tmp
, offset
);
1305 tcg_gen_addi_i32(tmp
, tmp
, -offset
);
1306 store_reg(s
, rd
, tmp
);
1307 } else if (!(insn
& (1 << 23)))
1312 static inline int gen_iwmmxt_shift(uint32_t insn
, uint32_t mask
, TCGv dest
)
1314 int rd
= (insn
>> 0) & 0xf;
1317 if (insn
& (1 << 8)) {
1318 if (rd
< ARM_IWMMXT_wCGR0
|| rd
> ARM_IWMMXT_wCGR3
) {
1321 tmp
= iwmmxt_load_creg(rd
);
1324 tmp
= tcg_temp_new_i32();
1325 iwmmxt_load_reg(cpu_V0
, rd
);
1326 tcg_gen_trunc_i64_i32(tmp
, cpu_V0
);
1328 tcg_gen_andi_i32(tmp
, tmp
, mask
);
1329 tcg_gen_mov_i32(dest
, tmp
);
1330 tcg_temp_free_i32(tmp
);
1334 /* Disassemble an iwMMXt instruction. Returns nonzero if an error occured
1335 (ie. an undefined instruction). */
1336 static int disas_iwmmxt_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
1339 int rdhi
, rdlo
, rd0
, rd1
, i
;
1341 TCGv tmp
, tmp2
, tmp3
;
1343 if ((insn
& 0x0e000e00) == 0x0c000000) {
1344 if ((insn
& 0x0fe00ff0) == 0x0c400000) {
1346 rdlo
= (insn
>> 12) & 0xf;
1347 rdhi
= (insn
>> 16) & 0xf;
1348 if (insn
& ARM_CP_RW_BIT
) { /* TMRRC */
1349 iwmmxt_load_reg(cpu_V0
, wrd
);
1350 tcg_gen_trunc_i64_i32(cpu_R
[rdlo
], cpu_V0
);
1351 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
1352 tcg_gen_trunc_i64_i32(cpu_R
[rdhi
], cpu_V0
);
1353 } else { /* TMCRR */
1354 tcg_gen_concat_i32_i64(cpu_V0
, cpu_R
[rdlo
], cpu_R
[rdhi
]);
1355 iwmmxt_store_reg(cpu_V0
, wrd
);
1356 gen_op_iwmmxt_set_mup();
1361 wrd
= (insn
>> 12) & 0xf;
1362 addr
= tcg_temp_new_i32();
1363 if (gen_iwmmxt_address(s
, insn
, addr
)) {
1364 tcg_temp_free_i32(addr
);
1367 if (insn
& ARM_CP_RW_BIT
) {
1368 if ((insn
>> 28) == 0xf) { /* WLDRW wCx */
1369 tmp
= tcg_temp_new_i32();
1370 tcg_gen_qemu_ld32u(tmp
, addr
, IS_USER(s
));
1371 iwmmxt_store_creg(wrd
, tmp
);
1374 if (insn
& (1 << 8)) {
1375 if (insn
& (1 << 22)) { /* WLDRD */
1376 tcg_gen_qemu_ld64(cpu_M0
, addr
, IS_USER(s
));
1378 } else { /* WLDRW wRd */
1379 tmp
= gen_ld32(addr
, IS_USER(s
));
1382 if (insn
& (1 << 22)) { /* WLDRH */
1383 tmp
= gen_ld16u(addr
, IS_USER(s
));
1384 } else { /* WLDRB */
1385 tmp
= gen_ld8u(addr
, IS_USER(s
));
1389 tcg_gen_extu_i32_i64(cpu_M0
, tmp
);
1390 tcg_temp_free_i32(tmp
);
1392 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1395 if ((insn
>> 28) == 0xf) { /* WSTRW wCx */
1396 tmp
= iwmmxt_load_creg(wrd
);
1397 gen_st32(tmp
, addr
, IS_USER(s
));
1399 gen_op_iwmmxt_movq_M0_wRn(wrd
);
1400 tmp
= tcg_temp_new_i32();
1401 if (insn
& (1 << 8)) {
1402 if (insn
& (1 << 22)) { /* WSTRD */
1403 tcg_temp_free_i32(tmp
);
1404 tcg_gen_qemu_st64(cpu_M0
, addr
, IS_USER(s
));
1405 } else { /* WSTRW wRd */
1406 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1407 gen_st32(tmp
, addr
, IS_USER(s
));
1410 if (insn
& (1 << 22)) { /* WSTRH */
1411 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1412 gen_st16(tmp
, addr
, IS_USER(s
));
1413 } else { /* WSTRB */
1414 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1415 gen_st8(tmp
, addr
, IS_USER(s
));
1420 tcg_temp_free_i32(addr
);
1424 if ((insn
& 0x0f000000) != 0x0e000000)
1427 switch (((insn
>> 12) & 0xf00) | ((insn
>> 4) & 0xff)) {
1428 case 0x000: /* WOR */
1429 wrd
= (insn
>> 12) & 0xf;
1430 rd0
= (insn
>> 0) & 0xf;
1431 rd1
= (insn
>> 16) & 0xf;
1432 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1433 gen_op_iwmmxt_orq_M0_wRn(rd1
);
1434 gen_op_iwmmxt_setpsr_nz();
1435 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1436 gen_op_iwmmxt_set_mup();
1437 gen_op_iwmmxt_set_cup();
1439 case 0x011: /* TMCR */
1442 rd
= (insn
>> 12) & 0xf;
1443 wrd
= (insn
>> 16) & 0xf;
1445 case ARM_IWMMXT_wCID
:
1446 case ARM_IWMMXT_wCASF
:
1448 case ARM_IWMMXT_wCon
:
1449 gen_op_iwmmxt_set_cup();
1451 case ARM_IWMMXT_wCSSF
:
1452 tmp
= iwmmxt_load_creg(wrd
);
1453 tmp2
= load_reg(s
, rd
);
1454 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
1455 tcg_temp_free_i32(tmp2
);
1456 iwmmxt_store_creg(wrd
, tmp
);
1458 case ARM_IWMMXT_wCGR0
:
1459 case ARM_IWMMXT_wCGR1
:
1460 case ARM_IWMMXT_wCGR2
:
1461 case ARM_IWMMXT_wCGR3
:
1462 gen_op_iwmmxt_set_cup();
1463 tmp
= load_reg(s
, rd
);
1464 iwmmxt_store_creg(wrd
, tmp
);
1470 case 0x100: /* WXOR */
1471 wrd
= (insn
>> 12) & 0xf;
1472 rd0
= (insn
>> 0) & 0xf;
1473 rd1
= (insn
>> 16) & 0xf;
1474 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1475 gen_op_iwmmxt_xorq_M0_wRn(rd1
);
1476 gen_op_iwmmxt_setpsr_nz();
1477 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1478 gen_op_iwmmxt_set_mup();
1479 gen_op_iwmmxt_set_cup();
1481 case 0x111: /* TMRC */
1484 rd
= (insn
>> 12) & 0xf;
1485 wrd
= (insn
>> 16) & 0xf;
1486 tmp
= iwmmxt_load_creg(wrd
);
1487 store_reg(s
, rd
, tmp
);
1489 case 0x300: /* WANDN */
1490 wrd
= (insn
>> 12) & 0xf;
1491 rd0
= (insn
>> 0) & 0xf;
1492 rd1
= (insn
>> 16) & 0xf;
1493 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1494 tcg_gen_neg_i64(cpu_M0
, cpu_M0
);
1495 gen_op_iwmmxt_andq_M0_wRn(rd1
);
1496 gen_op_iwmmxt_setpsr_nz();
1497 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1498 gen_op_iwmmxt_set_mup();
1499 gen_op_iwmmxt_set_cup();
1501 case 0x200: /* WAND */
1502 wrd
= (insn
>> 12) & 0xf;
1503 rd0
= (insn
>> 0) & 0xf;
1504 rd1
= (insn
>> 16) & 0xf;
1505 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1506 gen_op_iwmmxt_andq_M0_wRn(rd1
);
1507 gen_op_iwmmxt_setpsr_nz();
1508 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1509 gen_op_iwmmxt_set_mup();
1510 gen_op_iwmmxt_set_cup();
1512 case 0x810: case 0xa10: /* WMADD */
1513 wrd
= (insn
>> 12) & 0xf;
1514 rd0
= (insn
>> 0) & 0xf;
1515 rd1
= (insn
>> 16) & 0xf;
1516 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1517 if (insn
& (1 << 21))
1518 gen_op_iwmmxt_maddsq_M0_wRn(rd1
);
1520 gen_op_iwmmxt_madduq_M0_wRn(rd1
);
1521 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1522 gen_op_iwmmxt_set_mup();
1524 case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */
1525 wrd
= (insn
>> 12) & 0xf;
1526 rd0
= (insn
>> 16) & 0xf;
1527 rd1
= (insn
>> 0) & 0xf;
1528 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1529 switch ((insn
>> 22) & 3) {
1531 gen_op_iwmmxt_unpacklb_M0_wRn(rd1
);
1534 gen_op_iwmmxt_unpacklw_M0_wRn(rd1
);
1537 gen_op_iwmmxt_unpackll_M0_wRn(rd1
);
1542 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1543 gen_op_iwmmxt_set_mup();
1544 gen_op_iwmmxt_set_cup();
1546 case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */
1547 wrd
= (insn
>> 12) & 0xf;
1548 rd0
= (insn
>> 16) & 0xf;
1549 rd1
= (insn
>> 0) & 0xf;
1550 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1551 switch ((insn
>> 22) & 3) {
1553 gen_op_iwmmxt_unpackhb_M0_wRn(rd1
);
1556 gen_op_iwmmxt_unpackhw_M0_wRn(rd1
);
1559 gen_op_iwmmxt_unpackhl_M0_wRn(rd1
);
1564 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1565 gen_op_iwmmxt_set_mup();
1566 gen_op_iwmmxt_set_cup();
1568 case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */
1569 wrd
= (insn
>> 12) & 0xf;
1570 rd0
= (insn
>> 16) & 0xf;
1571 rd1
= (insn
>> 0) & 0xf;
1572 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1573 if (insn
& (1 << 22))
1574 gen_op_iwmmxt_sadw_M0_wRn(rd1
);
1576 gen_op_iwmmxt_sadb_M0_wRn(rd1
);
1577 if (!(insn
& (1 << 20)))
1578 gen_op_iwmmxt_addl_M0_wRn(wrd
);
1579 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1580 gen_op_iwmmxt_set_mup();
1582 case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */
1583 wrd
= (insn
>> 12) & 0xf;
1584 rd0
= (insn
>> 16) & 0xf;
1585 rd1
= (insn
>> 0) & 0xf;
1586 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1587 if (insn
& (1 << 21)) {
1588 if (insn
& (1 << 20))
1589 gen_op_iwmmxt_mulshw_M0_wRn(rd1
);
1591 gen_op_iwmmxt_mulslw_M0_wRn(rd1
);
1593 if (insn
& (1 << 20))
1594 gen_op_iwmmxt_muluhw_M0_wRn(rd1
);
1596 gen_op_iwmmxt_mululw_M0_wRn(rd1
);
1598 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1599 gen_op_iwmmxt_set_mup();
1601 case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */
1602 wrd
= (insn
>> 12) & 0xf;
1603 rd0
= (insn
>> 16) & 0xf;
1604 rd1
= (insn
>> 0) & 0xf;
1605 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1606 if (insn
& (1 << 21))
1607 gen_op_iwmmxt_macsw_M0_wRn(rd1
);
1609 gen_op_iwmmxt_macuw_M0_wRn(rd1
);
1610 if (!(insn
& (1 << 20))) {
1611 iwmmxt_load_reg(cpu_V1
, wrd
);
1612 tcg_gen_add_i64(cpu_M0
, cpu_M0
, cpu_V1
);
1614 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1615 gen_op_iwmmxt_set_mup();
1617 case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */
1618 wrd
= (insn
>> 12) & 0xf;
1619 rd0
= (insn
>> 16) & 0xf;
1620 rd1
= (insn
>> 0) & 0xf;
1621 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1622 switch ((insn
>> 22) & 3) {
1624 gen_op_iwmmxt_cmpeqb_M0_wRn(rd1
);
1627 gen_op_iwmmxt_cmpeqw_M0_wRn(rd1
);
1630 gen_op_iwmmxt_cmpeql_M0_wRn(rd1
);
1635 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1636 gen_op_iwmmxt_set_mup();
1637 gen_op_iwmmxt_set_cup();
1639 case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */
1640 wrd
= (insn
>> 12) & 0xf;
1641 rd0
= (insn
>> 16) & 0xf;
1642 rd1
= (insn
>> 0) & 0xf;
1643 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1644 if (insn
& (1 << 22)) {
1645 if (insn
& (1 << 20))
1646 gen_op_iwmmxt_avgw1_M0_wRn(rd1
);
1648 gen_op_iwmmxt_avgw0_M0_wRn(rd1
);
1650 if (insn
& (1 << 20))
1651 gen_op_iwmmxt_avgb1_M0_wRn(rd1
);
1653 gen_op_iwmmxt_avgb0_M0_wRn(rd1
);
1655 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1656 gen_op_iwmmxt_set_mup();
1657 gen_op_iwmmxt_set_cup();
1659 case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */
1660 wrd
= (insn
>> 12) & 0xf;
1661 rd0
= (insn
>> 16) & 0xf;
1662 rd1
= (insn
>> 0) & 0xf;
1663 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1664 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCGR0
+ ((insn
>> 20) & 3));
1665 tcg_gen_andi_i32(tmp
, tmp
, 7);
1666 iwmmxt_load_reg(cpu_V1
, rd1
);
1667 gen_helper_iwmmxt_align(cpu_M0
, cpu_M0
, cpu_V1
, tmp
);
1668 tcg_temp_free_i32(tmp
);
1669 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1670 gen_op_iwmmxt_set_mup();
1672 case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */
1673 if (((insn
>> 6) & 3) == 3)
1675 rd
= (insn
>> 12) & 0xf;
1676 wrd
= (insn
>> 16) & 0xf;
1677 tmp
= load_reg(s
, rd
);
1678 gen_op_iwmmxt_movq_M0_wRn(wrd
);
1679 switch ((insn
>> 6) & 3) {
1681 tmp2
= tcg_const_i32(0xff);
1682 tmp3
= tcg_const_i32((insn
& 7) << 3);
1685 tmp2
= tcg_const_i32(0xffff);
1686 tmp3
= tcg_const_i32((insn
& 3) << 4);
1689 tmp2
= tcg_const_i32(0xffffffff);
1690 tmp3
= tcg_const_i32((insn
& 1) << 5);
1696 gen_helper_iwmmxt_insr(cpu_M0
, cpu_M0
, tmp
, tmp2
, tmp3
);
1697 tcg_temp_free(tmp3
);
1698 tcg_temp_free(tmp2
);
1699 tcg_temp_free_i32(tmp
);
1700 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1701 gen_op_iwmmxt_set_mup();
1703 case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */
1704 rd
= (insn
>> 12) & 0xf;
1705 wrd
= (insn
>> 16) & 0xf;
1706 if (rd
== 15 || ((insn
>> 22) & 3) == 3)
1708 gen_op_iwmmxt_movq_M0_wRn(wrd
);
1709 tmp
= tcg_temp_new_i32();
1710 switch ((insn
>> 22) & 3) {
1712 tcg_gen_shri_i64(cpu_M0
, cpu_M0
, (insn
& 7) << 3);
1713 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1715 tcg_gen_ext8s_i32(tmp
, tmp
);
1717 tcg_gen_andi_i32(tmp
, tmp
, 0xff);
1721 tcg_gen_shri_i64(cpu_M0
, cpu_M0
, (insn
& 3) << 4);
1722 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1724 tcg_gen_ext16s_i32(tmp
, tmp
);
1726 tcg_gen_andi_i32(tmp
, tmp
, 0xffff);
1730 tcg_gen_shri_i64(cpu_M0
, cpu_M0
, (insn
& 1) << 5);
1731 tcg_gen_trunc_i64_i32(tmp
, cpu_M0
);
1734 store_reg(s
, rd
, tmp
);
1736 case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */
1737 if ((insn
& 0x000ff008) != 0x0003f000 || ((insn
>> 22) & 3) == 3)
1739 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCASF
);
1740 switch ((insn
>> 22) & 3) {
1742 tcg_gen_shri_i32(tmp
, tmp
, ((insn
& 7) << 2) + 0);
1745 tcg_gen_shri_i32(tmp
, tmp
, ((insn
& 3) << 3) + 4);
1748 tcg_gen_shri_i32(tmp
, tmp
, ((insn
& 1) << 4) + 12);
1751 tcg_gen_shli_i32(tmp
, tmp
, 28);
1753 tcg_temp_free_i32(tmp
);
1755 case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */
1756 if (((insn
>> 6) & 3) == 3)
1758 rd
= (insn
>> 12) & 0xf;
1759 wrd
= (insn
>> 16) & 0xf;
1760 tmp
= load_reg(s
, rd
);
1761 switch ((insn
>> 6) & 3) {
1763 gen_helper_iwmmxt_bcstb(cpu_M0
, tmp
);
1766 gen_helper_iwmmxt_bcstw(cpu_M0
, tmp
);
1769 gen_helper_iwmmxt_bcstl(cpu_M0
, tmp
);
1772 tcg_temp_free_i32(tmp
);
1773 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1774 gen_op_iwmmxt_set_mup();
1776 case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */
1777 if ((insn
& 0x000ff00f) != 0x0003f000 || ((insn
>> 22) & 3) == 3)
1779 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCASF
);
1780 tmp2
= tcg_temp_new_i32();
1781 tcg_gen_mov_i32(tmp2
, tmp
);
1782 switch ((insn
>> 22) & 3) {
1784 for (i
= 0; i
< 7; i
++) {
1785 tcg_gen_shli_i32(tmp2
, tmp2
, 4);
1786 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
1790 for (i
= 0; i
< 3; i
++) {
1791 tcg_gen_shli_i32(tmp2
, tmp2
, 8);
1792 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
1796 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
1797 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
1801 tcg_temp_free_i32(tmp2
);
1802 tcg_temp_free_i32(tmp
);
1804 case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */
1805 wrd
= (insn
>> 12) & 0xf;
1806 rd0
= (insn
>> 16) & 0xf;
1807 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1808 switch ((insn
>> 22) & 3) {
1810 gen_helper_iwmmxt_addcb(cpu_M0
, cpu_M0
);
1813 gen_helper_iwmmxt_addcw(cpu_M0
, cpu_M0
);
1816 gen_helper_iwmmxt_addcl(cpu_M0
, cpu_M0
);
1821 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1822 gen_op_iwmmxt_set_mup();
1824 case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */
1825 if ((insn
& 0x000ff00f) != 0x0003f000 || ((insn
>> 22) & 3) == 3)
1827 tmp
= iwmmxt_load_creg(ARM_IWMMXT_wCASF
);
1828 tmp2
= tcg_temp_new_i32();
1829 tcg_gen_mov_i32(tmp2
, tmp
);
1830 switch ((insn
>> 22) & 3) {
1832 for (i
= 0; i
< 7; i
++) {
1833 tcg_gen_shli_i32(tmp2
, tmp2
, 4);
1834 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
1838 for (i
= 0; i
< 3; i
++) {
1839 tcg_gen_shli_i32(tmp2
, tmp2
, 8);
1840 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
1844 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
1845 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
1849 tcg_temp_free_i32(tmp2
);
1850 tcg_temp_free_i32(tmp
);
1852 case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */
1853 rd
= (insn
>> 12) & 0xf;
1854 rd0
= (insn
>> 16) & 0xf;
1855 if ((insn
& 0xf) != 0 || ((insn
>> 22) & 3) == 3)
1857 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1858 tmp
= tcg_temp_new_i32();
1859 switch ((insn
>> 22) & 3) {
1861 gen_helper_iwmmxt_msbb(tmp
, cpu_M0
);
1864 gen_helper_iwmmxt_msbw(tmp
, cpu_M0
);
1867 gen_helper_iwmmxt_msbl(tmp
, cpu_M0
);
1870 store_reg(s
, rd
, tmp
);
1872 case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */
1873 case 0x906: case 0xb06: case 0xd06: case 0xf06:
1874 wrd
= (insn
>> 12) & 0xf;
1875 rd0
= (insn
>> 16) & 0xf;
1876 rd1
= (insn
>> 0) & 0xf;
1877 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1878 switch ((insn
>> 22) & 3) {
1880 if (insn
& (1 << 21))
1881 gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1
);
1883 gen_op_iwmmxt_cmpgtub_M0_wRn(rd1
);
1886 if (insn
& (1 << 21))
1887 gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1
);
1889 gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1
);
1892 if (insn
& (1 << 21))
1893 gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1
);
1895 gen_op_iwmmxt_cmpgtul_M0_wRn(rd1
);
1900 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1901 gen_op_iwmmxt_set_mup();
1902 gen_op_iwmmxt_set_cup();
1904 case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */
1905 case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e:
1906 wrd
= (insn
>> 12) & 0xf;
1907 rd0
= (insn
>> 16) & 0xf;
1908 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1909 switch ((insn
>> 22) & 3) {
1911 if (insn
& (1 << 21))
1912 gen_op_iwmmxt_unpacklsb_M0();
1914 gen_op_iwmmxt_unpacklub_M0();
1917 if (insn
& (1 << 21))
1918 gen_op_iwmmxt_unpacklsw_M0();
1920 gen_op_iwmmxt_unpackluw_M0();
1923 if (insn
& (1 << 21))
1924 gen_op_iwmmxt_unpacklsl_M0();
1926 gen_op_iwmmxt_unpacklul_M0();
1931 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1932 gen_op_iwmmxt_set_mup();
1933 gen_op_iwmmxt_set_cup();
1935 case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */
1936 case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c:
1937 wrd
= (insn
>> 12) & 0xf;
1938 rd0
= (insn
>> 16) & 0xf;
1939 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1940 switch ((insn
>> 22) & 3) {
1942 if (insn
& (1 << 21))
1943 gen_op_iwmmxt_unpackhsb_M0();
1945 gen_op_iwmmxt_unpackhub_M0();
1948 if (insn
& (1 << 21))
1949 gen_op_iwmmxt_unpackhsw_M0();
1951 gen_op_iwmmxt_unpackhuw_M0();
1954 if (insn
& (1 << 21))
1955 gen_op_iwmmxt_unpackhsl_M0();
1957 gen_op_iwmmxt_unpackhul_M0();
1962 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1963 gen_op_iwmmxt_set_mup();
1964 gen_op_iwmmxt_set_cup();
1966 case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */
1967 case 0x214: case 0x614: case 0xa14: case 0xe14:
1968 if (((insn
>> 22) & 3) == 0)
1970 wrd
= (insn
>> 12) & 0xf;
1971 rd0
= (insn
>> 16) & 0xf;
1972 gen_op_iwmmxt_movq_M0_wRn(rd0
);
1973 tmp
= tcg_temp_new_i32();
1974 if (gen_iwmmxt_shift(insn
, 0xff, tmp
)) {
1975 tcg_temp_free_i32(tmp
);
1978 switch ((insn
>> 22) & 3) {
1980 gen_helper_iwmmxt_srlw(cpu_M0
, cpu_M0
, tmp
);
1983 gen_helper_iwmmxt_srll(cpu_M0
, cpu_M0
, tmp
);
1986 gen_helper_iwmmxt_srlq(cpu_M0
, cpu_M0
, tmp
);
1989 tcg_temp_free_i32(tmp
);
1990 gen_op_iwmmxt_movq_wRn_M0(wrd
);
1991 gen_op_iwmmxt_set_mup();
1992 gen_op_iwmmxt_set_cup();
1994 case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */
1995 case 0x014: case 0x414: case 0x814: case 0xc14:
1996 if (((insn
>> 22) & 3) == 0)
1998 wrd
= (insn
>> 12) & 0xf;
1999 rd0
= (insn
>> 16) & 0xf;
2000 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2001 tmp
= tcg_temp_new_i32();
2002 if (gen_iwmmxt_shift(insn
, 0xff, tmp
)) {
2003 tcg_temp_free_i32(tmp
);
2006 switch ((insn
>> 22) & 3) {
2008 gen_helper_iwmmxt_sraw(cpu_M0
, cpu_M0
, tmp
);
2011 gen_helper_iwmmxt_sral(cpu_M0
, cpu_M0
, tmp
);
2014 gen_helper_iwmmxt_sraq(cpu_M0
, cpu_M0
, tmp
);
2017 tcg_temp_free_i32(tmp
);
2018 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2019 gen_op_iwmmxt_set_mup();
2020 gen_op_iwmmxt_set_cup();
2022 case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */
2023 case 0x114: case 0x514: case 0x914: case 0xd14:
2024 if (((insn
>> 22) & 3) == 0)
2026 wrd
= (insn
>> 12) & 0xf;
2027 rd0
= (insn
>> 16) & 0xf;
2028 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2029 tmp
= tcg_temp_new_i32();
2030 if (gen_iwmmxt_shift(insn
, 0xff, tmp
)) {
2031 tcg_temp_free_i32(tmp
);
2034 switch ((insn
>> 22) & 3) {
2036 gen_helper_iwmmxt_sllw(cpu_M0
, cpu_M0
, tmp
);
2039 gen_helper_iwmmxt_slll(cpu_M0
, cpu_M0
, tmp
);
2042 gen_helper_iwmmxt_sllq(cpu_M0
, cpu_M0
, tmp
);
2045 tcg_temp_free_i32(tmp
);
2046 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2047 gen_op_iwmmxt_set_mup();
2048 gen_op_iwmmxt_set_cup();
2050 case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */
2051 case 0x314: case 0x714: case 0xb14: case 0xf14:
2052 if (((insn
>> 22) & 3) == 0)
2054 wrd
= (insn
>> 12) & 0xf;
2055 rd0
= (insn
>> 16) & 0xf;
2056 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2057 tmp
= tcg_temp_new_i32();
2058 switch ((insn
>> 22) & 3) {
2060 if (gen_iwmmxt_shift(insn
, 0xf, tmp
)) {
2061 tcg_temp_free_i32(tmp
);
2064 gen_helper_iwmmxt_rorw(cpu_M0
, cpu_M0
, tmp
);
2067 if (gen_iwmmxt_shift(insn
, 0x1f, tmp
)) {
2068 tcg_temp_free_i32(tmp
);
2071 gen_helper_iwmmxt_rorl(cpu_M0
, cpu_M0
, tmp
);
2074 if (gen_iwmmxt_shift(insn
, 0x3f, tmp
)) {
2075 tcg_temp_free_i32(tmp
);
2078 gen_helper_iwmmxt_rorq(cpu_M0
, cpu_M0
, tmp
);
2081 tcg_temp_free_i32(tmp
);
2082 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2083 gen_op_iwmmxt_set_mup();
2084 gen_op_iwmmxt_set_cup();
2086 case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */
2087 case 0x916: case 0xb16: case 0xd16: case 0xf16:
2088 wrd
= (insn
>> 12) & 0xf;
2089 rd0
= (insn
>> 16) & 0xf;
2090 rd1
= (insn
>> 0) & 0xf;
2091 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2092 switch ((insn
>> 22) & 3) {
2094 if (insn
& (1 << 21))
2095 gen_op_iwmmxt_minsb_M0_wRn(rd1
);
2097 gen_op_iwmmxt_minub_M0_wRn(rd1
);
2100 if (insn
& (1 << 21))
2101 gen_op_iwmmxt_minsw_M0_wRn(rd1
);
2103 gen_op_iwmmxt_minuw_M0_wRn(rd1
);
2106 if (insn
& (1 << 21))
2107 gen_op_iwmmxt_minsl_M0_wRn(rd1
);
2109 gen_op_iwmmxt_minul_M0_wRn(rd1
);
2114 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2115 gen_op_iwmmxt_set_mup();
2117 case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */
2118 case 0x816: case 0xa16: case 0xc16: case 0xe16:
2119 wrd
= (insn
>> 12) & 0xf;
2120 rd0
= (insn
>> 16) & 0xf;
2121 rd1
= (insn
>> 0) & 0xf;
2122 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2123 switch ((insn
>> 22) & 3) {
2125 if (insn
& (1 << 21))
2126 gen_op_iwmmxt_maxsb_M0_wRn(rd1
);
2128 gen_op_iwmmxt_maxub_M0_wRn(rd1
);
2131 if (insn
& (1 << 21))
2132 gen_op_iwmmxt_maxsw_M0_wRn(rd1
);
2134 gen_op_iwmmxt_maxuw_M0_wRn(rd1
);
2137 if (insn
& (1 << 21))
2138 gen_op_iwmmxt_maxsl_M0_wRn(rd1
);
2140 gen_op_iwmmxt_maxul_M0_wRn(rd1
);
2145 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2146 gen_op_iwmmxt_set_mup();
2148 case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */
2149 case 0x402: case 0x502: case 0x602: case 0x702:
2150 wrd
= (insn
>> 12) & 0xf;
2151 rd0
= (insn
>> 16) & 0xf;
2152 rd1
= (insn
>> 0) & 0xf;
2153 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2154 tmp
= tcg_const_i32((insn
>> 20) & 3);
2155 iwmmxt_load_reg(cpu_V1
, rd1
);
2156 gen_helper_iwmmxt_align(cpu_M0
, cpu_M0
, cpu_V1
, tmp
);
2158 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2159 gen_op_iwmmxt_set_mup();
2161 case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */
2162 case 0x41a: case 0x51a: case 0x61a: case 0x71a:
2163 case 0x81a: case 0x91a: case 0xa1a: case 0xb1a:
2164 case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a:
2165 wrd
= (insn
>> 12) & 0xf;
2166 rd0
= (insn
>> 16) & 0xf;
2167 rd1
= (insn
>> 0) & 0xf;
2168 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2169 switch ((insn
>> 20) & 0xf) {
2171 gen_op_iwmmxt_subnb_M0_wRn(rd1
);
2174 gen_op_iwmmxt_subub_M0_wRn(rd1
);
2177 gen_op_iwmmxt_subsb_M0_wRn(rd1
);
2180 gen_op_iwmmxt_subnw_M0_wRn(rd1
);
2183 gen_op_iwmmxt_subuw_M0_wRn(rd1
);
2186 gen_op_iwmmxt_subsw_M0_wRn(rd1
);
2189 gen_op_iwmmxt_subnl_M0_wRn(rd1
);
2192 gen_op_iwmmxt_subul_M0_wRn(rd1
);
2195 gen_op_iwmmxt_subsl_M0_wRn(rd1
);
2200 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2201 gen_op_iwmmxt_set_mup();
2202 gen_op_iwmmxt_set_cup();
2204 case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */
2205 case 0x41e: case 0x51e: case 0x61e: case 0x71e:
2206 case 0x81e: case 0x91e: case 0xa1e: case 0xb1e:
2207 case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e:
2208 wrd
= (insn
>> 12) & 0xf;
2209 rd0
= (insn
>> 16) & 0xf;
2210 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2211 tmp
= tcg_const_i32(((insn
>> 16) & 0xf0) | (insn
& 0x0f));
2212 gen_helper_iwmmxt_shufh(cpu_M0
, cpu_M0
, tmp
);
2214 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2215 gen_op_iwmmxt_set_mup();
2216 gen_op_iwmmxt_set_cup();
2218 case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */
2219 case 0x418: case 0x518: case 0x618: case 0x718:
2220 case 0x818: case 0x918: case 0xa18: case 0xb18:
2221 case 0xc18: case 0xd18: case 0xe18: case 0xf18:
2222 wrd
= (insn
>> 12) & 0xf;
2223 rd0
= (insn
>> 16) & 0xf;
2224 rd1
= (insn
>> 0) & 0xf;
2225 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2226 switch ((insn
>> 20) & 0xf) {
2228 gen_op_iwmmxt_addnb_M0_wRn(rd1
);
2231 gen_op_iwmmxt_addub_M0_wRn(rd1
);
2234 gen_op_iwmmxt_addsb_M0_wRn(rd1
);
2237 gen_op_iwmmxt_addnw_M0_wRn(rd1
);
2240 gen_op_iwmmxt_adduw_M0_wRn(rd1
);
2243 gen_op_iwmmxt_addsw_M0_wRn(rd1
);
2246 gen_op_iwmmxt_addnl_M0_wRn(rd1
);
2249 gen_op_iwmmxt_addul_M0_wRn(rd1
);
2252 gen_op_iwmmxt_addsl_M0_wRn(rd1
);
2257 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2258 gen_op_iwmmxt_set_mup();
2259 gen_op_iwmmxt_set_cup();
2261 case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */
2262 case 0x408: case 0x508: case 0x608: case 0x708:
2263 case 0x808: case 0x908: case 0xa08: case 0xb08:
2264 case 0xc08: case 0xd08: case 0xe08: case 0xf08:
2265 if (!(insn
& (1 << 20)) || ((insn
>> 22) & 3) == 0)
2267 wrd
= (insn
>> 12) & 0xf;
2268 rd0
= (insn
>> 16) & 0xf;
2269 rd1
= (insn
>> 0) & 0xf;
2270 gen_op_iwmmxt_movq_M0_wRn(rd0
);
2271 switch ((insn
>> 22) & 3) {
2273 if (insn
& (1 << 21))
2274 gen_op_iwmmxt_packsw_M0_wRn(rd1
);
2276 gen_op_iwmmxt_packuw_M0_wRn(rd1
);
2279 if (insn
& (1 << 21))
2280 gen_op_iwmmxt_packsl_M0_wRn(rd1
);
2282 gen_op_iwmmxt_packul_M0_wRn(rd1
);
2285 if (insn
& (1 << 21))
2286 gen_op_iwmmxt_packsq_M0_wRn(rd1
);
2288 gen_op_iwmmxt_packuq_M0_wRn(rd1
);
2291 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2292 gen_op_iwmmxt_set_mup();
2293 gen_op_iwmmxt_set_cup();
2295 case 0x201: case 0x203: case 0x205: case 0x207:
2296 case 0x209: case 0x20b: case 0x20d: case 0x20f:
2297 case 0x211: case 0x213: case 0x215: case 0x217:
2298 case 0x219: case 0x21b: case 0x21d: case 0x21f:
2299 wrd
= (insn
>> 5) & 0xf;
2300 rd0
= (insn
>> 12) & 0xf;
2301 rd1
= (insn
>> 0) & 0xf;
2302 if (rd0
== 0xf || rd1
== 0xf)
2304 gen_op_iwmmxt_movq_M0_wRn(wrd
);
2305 tmp
= load_reg(s
, rd0
);
2306 tmp2
= load_reg(s
, rd1
);
2307 switch ((insn
>> 16) & 0xf) {
2308 case 0x0: /* TMIA */
2309 gen_helper_iwmmxt_muladdsl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2311 case 0x8: /* TMIAPH */
2312 gen_helper_iwmmxt_muladdsw(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2314 case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */
2315 if (insn
& (1 << 16))
2316 tcg_gen_shri_i32(tmp
, tmp
, 16);
2317 if (insn
& (1 << 17))
2318 tcg_gen_shri_i32(tmp2
, tmp2
, 16);
2319 gen_helper_iwmmxt_muladdswl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2322 tcg_temp_free_i32(tmp2
);
2323 tcg_temp_free_i32(tmp
);
2326 tcg_temp_free_i32(tmp2
);
2327 tcg_temp_free_i32(tmp
);
2328 gen_op_iwmmxt_movq_wRn_M0(wrd
);
2329 gen_op_iwmmxt_set_mup();
2338 /* Disassemble an XScale DSP instruction. Returns nonzero if an error occured
2339 (ie. an undefined instruction). */
2340 static int disas_dsp_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
2342 int acc
, rd0
, rd1
, rdhi
, rdlo
;
2345 if ((insn
& 0x0ff00f10) == 0x0e200010) {
2346 /* Multiply with Internal Accumulate Format */
2347 rd0
= (insn
>> 12) & 0xf;
2349 acc
= (insn
>> 5) & 7;
2354 tmp
= load_reg(s
, rd0
);
2355 tmp2
= load_reg(s
, rd1
);
2356 switch ((insn
>> 16) & 0xf) {
2358 gen_helper_iwmmxt_muladdsl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2360 case 0x8: /* MIAPH */
2361 gen_helper_iwmmxt_muladdsw(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2363 case 0xc: /* MIABB */
2364 case 0xd: /* MIABT */
2365 case 0xe: /* MIATB */
2366 case 0xf: /* MIATT */
2367 if (insn
& (1 << 16))
2368 tcg_gen_shri_i32(tmp
, tmp
, 16);
2369 if (insn
& (1 << 17))
2370 tcg_gen_shri_i32(tmp2
, tmp2
, 16);
2371 gen_helper_iwmmxt_muladdswl(cpu_M0
, cpu_M0
, tmp
, tmp2
);
2376 tcg_temp_free_i32(tmp2
);
2377 tcg_temp_free_i32(tmp
);
2379 gen_op_iwmmxt_movq_wRn_M0(acc
);
2383 if ((insn
& 0x0fe00ff8) == 0x0c400000) {
2384 /* Internal Accumulator Access Format */
2385 rdhi
= (insn
>> 16) & 0xf;
2386 rdlo
= (insn
>> 12) & 0xf;
2392 if (insn
& ARM_CP_RW_BIT
) { /* MRA */
2393 iwmmxt_load_reg(cpu_V0
, acc
);
2394 tcg_gen_trunc_i64_i32(cpu_R
[rdlo
], cpu_V0
);
2395 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
2396 tcg_gen_trunc_i64_i32(cpu_R
[rdhi
], cpu_V0
);
2397 tcg_gen_andi_i32(cpu_R
[rdhi
], cpu_R
[rdhi
], (1 << (40 - 32)) - 1);
2399 tcg_gen_concat_i32_i64(cpu_V0
, cpu_R
[rdlo
], cpu_R
[rdhi
]);
2400 iwmmxt_store_reg(cpu_V0
, acc
);
2408 /* Disassemble system coprocessor instruction. Return nonzero if
2409 instruction is not defined. */
2410 static int disas_cp_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
2413 uint32_t rd
= (insn
>> 12) & 0xf;
2414 uint32_t cp
= (insn
>> 8) & 0xf;
2419 if (insn
& ARM_CP_RW_BIT
) {
2420 if (!env
->cp
[cp
].cp_read
)
2422 gen_set_pc_im(s
->pc
);
2423 tmp
= tcg_temp_new_i32();
2424 tmp2
= tcg_const_i32(insn
);
2425 gen_helper_get_cp(tmp
, cpu_env
, tmp2
);
2426 tcg_temp_free(tmp2
);
2427 store_reg(s
, rd
, tmp
);
2429 if (!env
->cp
[cp
].cp_write
)
2431 gen_set_pc_im(s
->pc
);
2432 tmp
= load_reg(s
, rd
);
2433 tmp2
= tcg_const_i32(insn
);
2434 gen_helper_set_cp(cpu_env
, tmp2
, tmp
);
2435 tcg_temp_free(tmp2
);
2436 tcg_temp_free_i32(tmp
);
2441 static int cp15_user_ok(uint32_t insn
)
2443 int cpn
= (insn
>> 16) & 0xf;
2444 int cpm
= insn
& 0xf;
2445 int op
= ((insn
>> 5) & 7) | ((insn
>> 18) & 0x38);
2447 if (cpn
== 13 && cpm
== 0) {
2449 if (op
== 2 || (op
== 3 && (insn
& ARM_CP_RW_BIT
)))
2453 /* ISB, DSB, DMB. */
2454 if ((cpm
== 5 && op
== 4)
2455 || (cpm
== 10 && (op
== 4 || op
== 5)))
2461 static int cp15_tls_load_store(CPUState
*env
, DisasContext
*s
, uint32_t insn
, uint32_t rd
)
2464 int cpn
= (insn
>> 16) & 0xf;
2465 int cpm
= insn
& 0xf;
2466 int op
= ((insn
>> 5) & 7) | ((insn
>> 18) & 0x38);
2468 if (!arm_feature(env
, ARM_FEATURE_V6K
))
2471 if (!(cpn
== 13 && cpm
== 0))
2474 if (insn
& ARM_CP_RW_BIT
) {
2477 tmp
= load_cpu_field(cp15
.c13_tls1
);
2480 tmp
= load_cpu_field(cp15
.c13_tls2
);
2483 tmp
= load_cpu_field(cp15
.c13_tls3
);
2488 store_reg(s
, rd
, tmp
);
2491 tmp
= load_reg(s
, rd
);
2494 store_cpu_field(tmp
, cp15
.c13_tls1
);
2497 store_cpu_field(tmp
, cp15
.c13_tls2
);
2500 store_cpu_field(tmp
, cp15
.c13_tls3
);
2503 tcg_temp_free_i32(tmp
);
2510 /* Disassemble system coprocessor (cp15) instruction. Return nonzero if
2511 instruction is not defined. */
2512 static int disas_cp15_insn(CPUState
*env
, DisasContext
*s
, uint32_t insn
)
2517 /* M profile cores use memory mapped registers instead of cp15. */
2518 if (arm_feature(env
, ARM_FEATURE_M
))
2521 if ((insn
& (1 << 25)) == 0) {
2522 if (insn
& (1 << 20)) {
2526 /* mcrr. Used for block cache operations, so implement as no-op. */
2529 if ((insn
& (1 << 4)) == 0) {
2533 if (IS_USER(s
) && !cp15_user_ok(insn
)) {
2537 /* Pre-v7 versions of the architecture implemented WFI via coprocessor
2538 * instructions rather than a separate instruction.
2540 if ((insn
& 0x0fff0fff) == 0x0e070f90) {
2541 /* 0,c7,c0,4: Standard v6 WFI (also used in some pre-v6 cores).
2542 * In v7, this must NOP.
2544 if (!arm_feature(env
, ARM_FEATURE_V7
)) {
2545 /* Wait for interrupt. */
2546 gen_set_pc_im(s
->pc
);
2547 s
->is_jmp
= DISAS_WFI
;
2552 if ((insn
& 0x0fff0fff) == 0x0e070f58) {
2553 /* 0,c7,c8,2: Not all pre-v6 cores implemented this WFI,
2554 * so this is slightly over-broad.
2556 if (!arm_feature(env
, ARM_FEATURE_V6
)) {
2557 /* Wait for interrupt. */
2558 gen_set_pc_im(s
->pc
);
2559 s
->is_jmp
= DISAS_WFI
;
2562 /* Otherwise fall through to handle via helper function.
2563 * In particular, on v7 and some v6 cores this is one of
2564 * the VA-PA registers.
2568 rd
= (insn
>> 12) & 0xf;
2570 if (cp15_tls_load_store(env
, s
, insn
, rd
))
2573 tmp2
= tcg_const_i32(insn
);
2574 if (insn
& ARM_CP_RW_BIT
) {
2575 tmp
= tcg_temp_new_i32();
2576 gen_helper_get_cp15(tmp
, cpu_env
, tmp2
);
2577 /* If the destination register is r15 then sets condition codes. */
2579 store_reg(s
, rd
, tmp
);
2581 tcg_temp_free_i32(tmp
);
2583 tmp
= load_reg(s
, rd
);
2584 gen_helper_set_cp15(cpu_env
, tmp2
, tmp
);
2585 tcg_temp_free_i32(tmp
);
2586 /* Normally we would always end the TB here, but Linux
2587 * arch/arm/mach-pxa/sleep.S expects two instructions following
2588 * an MMU enable to execute from cache. Imitate this behaviour. */
2589 if (!arm_feature(env
, ARM_FEATURE_XSCALE
) ||
2590 (insn
& 0x0fff0fff) != 0x0e010f10)
2593 tcg_temp_free_i32(tmp2
);
2597 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
2598 #define VFP_SREG(insn, bigbit, smallbit) \
2599 ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
2600 #define VFP_DREG(reg, insn, bigbit, smallbit) do { \
2601 if (arm_feature(env, ARM_FEATURE_VFP3)) { \
2602 reg = (((insn) >> (bigbit)) & 0x0f) \
2603 | (((insn) >> ((smallbit) - 4)) & 0x10); \
2605 if (insn & (1 << (smallbit))) \
2607 reg = ((insn) >> (bigbit)) & 0x0f; \
2610 #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
2611 #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
2612 #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7)
2613 #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7)
2614 #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5)
2615 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5)
2617 /* Move between integer and VFP cores. */
2618 static TCGv
gen_vfp_mrs(void)
2620 TCGv tmp
= tcg_temp_new_i32();
2621 tcg_gen_mov_i32(tmp
, cpu_F0s
);
2625 static void gen_vfp_msr(TCGv tmp
)
2627 tcg_gen_mov_i32(cpu_F0s
, tmp
);
2628 tcg_temp_free_i32(tmp
);
2631 static void gen_neon_dup_u8(TCGv var
, int shift
)
2633 TCGv tmp
= tcg_temp_new_i32();
2635 tcg_gen_shri_i32(var
, var
, shift
);
2636 tcg_gen_ext8u_i32(var
, var
);
2637 tcg_gen_shli_i32(tmp
, var
, 8);
2638 tcg_gen_or_i32(var
, var
, tmp
);
2639 tcg_gen_shli_i32(tmp
, var
, 16);
2640 tcg_gen_or_i32(var
, var
, tmp
);
2641 tcg_temp_free_i32(tmp
);
2644 static void gen_neon_dup_low16(TCGv var
)
2646 TCGv tmp
= tcg_temp_new_i32();
2647 tcg_gen_ext16u_i32(var
, var
);
2648 tcg_gen_shli_i32(tmp
, var
, 16);
2649 tcg_gen_or_i32(var
, var
, tmp
);
2650 tcg_temp_free_i32(tmp
);
2653 static void gen_neon_dup_high16(TCGv var
)
2655 TCGv tmp
= tcg_temp_new_i32();
2656 tcg_gen_andi_i32(var
, var
, 0xffff0000);
2657 tcg_gen_shri_i32(tmp
, var
, 16);
2658 tcg_gen_or_i32(var
, var
, tmp
);
2659 tcg_temp_free_i32(tmp
);
2662 static TCGv
gen_load_and_replicate(DisasContext
*s
, TCGv addr
, int size
)
2664 /* Load a single Neon element and replicate into a 32 bit TCG reg */
2668 tmp
= gen_ld8u(addr
, IS_USER(s
));
2669 gen_neon_dup_u8(tmp
, 0);
2672 tmp
= gen_ld16u(addr
, IS_USER(s
));
2673 gen_neon_dup_low16(tmp
);
2676 tmp
= gen_ld32(addr
, IS_USER(s
));
2678 default: /* Avoid compiler warnings. */
2684 /* Disassemble a VFP instruction. Returns nonzero if an error occured
2685 (ie. an undefined instruction). */
2686 static int disas_vfp_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
2688 uint32_t rd
, rn
, rm
, op
, i
, n
, offset
, delta_d
, delta_m
, bank_mask
;
2694 if (!arm_feature(env
, ARM_FEATURE_VFP
))
2697 if (!s
->vfp_enabled
) {
2698 /* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
2699 if ((insn
& 0x0fe00fff) != 0x0ee00a10)
2701 rn
= (insn
>> 16) & 0xf;
2702 if (rn
!= ARM_VFP_FPSID
&& rn
!= ARM_VFP_FPEXC
2703 && rn
!= ARM_VFP_MVFR1
&& rn
!= ARM_VFP_MVFR0
)
2706 dp
= ((insn
& 0xf00) == 0xb00);
2707 switch ((insn
>> 24) & 0xf) {
2709 if (insn
& (1 << 4)) {
2710 /* single register transfer */
2711 rd
= (insn
>> 12) & 0xf;
2716 VFP_DREG_N(rn
, insn
);
2719 if (insn
& 0x00c00060
2720 && !arm_feature(env
, ARM_FEATURE_NEON
))
2723 pass
= (insn
>> 21) & 1;
2724 if (insn
& (1 << 22)) {
2726 offset
= ((insn
>> 5) & 3) * 8;
2727 } else if (insn
& (1 << 5)) {
2729 offset
= (insn
& (1 << 6)) ? 16 : 0;
2734 if (insn
& ARM_CP_RW_BIT
) {
2736 tmp
= neon_load_reg(rn
, pass
);
2740 tcg_gen_shri_i32(tmp
, tmp
, offset
);
2741 if (insn
& (1 << 23))
2747 if (insn
& (1 << 23)) {
2749 tcg_gen_shri_i32(tmp
, tmp
, 16);
2755 tcg_gen_sari_i32(tmp
, tmp
, 16);
2764 store_reg(s
, rd
, tmp
);
2767 tmp
= load_reg(s
, rd
);
2768 if (insn
& (1 << 23)) {
2771 gen_neon_dup_u8(tmp
, 0);
2772 } else if (size
== 1) {
2773 gen_neon_dup_low16(tmp
);
2775 for (n
= 0; n
<= pass
* 2; n
++) {
2776 tmp2
= tcg_temp_new_i32();
2777 tcg_gen_mov_i32(tmp2
, tmp
);
2778 neon_store_reg(rn
, n
, tmp2
);
2780 neon_store_reg(rn
, n
, tmp
);
2785 tmp2
= neon_load_reg(rn
, pass
);
2786 gen_bfi(tmp
, tmp2
, tmp
, offset
, 0xff);
2787 tcg_temp_free_i32(tmp2
);
2790 tmp2
= neon_load_reg(rn
, pass
);
2791 gen_bfi(tmp
, tmp2
, tmp
, offset
, 0xffff);
2792 tcg_temp_free_i32(tmp2
);
2797 neon_store_reg(rn
, pass
, tmp
);
2801 if ((insn
& 0x6f) != 0x00)
2803 rn
= VFP_SREG_N(insn
);
2804 if (insn
& ARM_CP_RW_BIT
) {
2806 if (insn
& (1 << 21)) {
2807 /* system register */
2812 /* VFP2 allows access to FSID from userspace.
2813 VFP3 restricts all id registers to privileged
2816 && arm_feature(env
, ARM_FEATURE_VFP3
))
2818 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2823 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2825 case ARM_VFP_FPINST
:
2826 case ARM_VFP_FPINST2
:
2827 /* Not present in VFP3. */
2829 || arm_feature(env
, ARM_FEATURE_VFP3
))
2831 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2835 tmp
= load_cpu_field(vfp
.xregs
[ARM_VFP_FPSCR
]);
2836 tcg_gen_andi_i32(tmp
, tmp
, 0xf0000000);
2838 tmp
= tcg_temp_new_i32();
2839 gen_helper_vfp_get_fpscr(tmp
, cpu_env
);
2845 || !arm_feature(env
, ARM_FEATURE_VFP3
))
2847 tmp
= load_cpu_field(vfp
.xregs
[rn
]);
2853 gen_mov_F0_vreg(0, rn
);
2854 tmp
= gen_vfp_mrs();
2857 /* Set the 4 flag bits in the CPSR. */
2859 tcg_temp_free_i32(tmp
);
2861 store_reg(s
, rd
, tmp
);
2865 tmp
= load_reg(s
, rd
);
2866 if (insn
& (1 << 21)) {
2868 /* system register */
2873 /* Writes are ignored. */
2876 gen_helper_vfp_set_fpscr(cpu_env
, tmp
);
2877 tcg_temp_free_i32(tmp
);
2883 /* TODO: VFP subarchitecture support.
2884 * For now, keep the EN bit only */
2885 tcg_gen_andi_i32(tmp
, tmp
, 1 << 30);
2886 store_cpu_field(tmp
, vfp
.xregs
[rn
]);
2889 case ARM_VFP_FPINST
:
2890 case ARM_VFP_FPINST2
:
2891 store_cpu_field(tmp
, vfp
.xregs
[rn
]);
2898 gen_mov_vreg_F0(0, rn
);
2903 /* data processing */
2904 /* The opcode is in bits 23, 21, 20 and 6. */
2905 op
= ((insn
>> 20) & 8) | ((insn
>> 19) & 6) | ((insn
>> 6) & 1);
2909 rn
= ((insn
>> 15) & 0x1e) | ((insn
>> 7) & 1);
2911 /* rn is register number */
2912 VFP_DREG_N(rn
, insn
);
2915 if (op
== 15 && (rn
== 15 || ((rn
& 0x1c) == 0x18))) {
2916 /* Integer or single precision destination. */
2917 rd
= VFP_SREG_D(insn
);
2919 VFP_DREG_D(rd
, insn
);
2922 (((rn
& 0x1c) == 0x10) || ((rn
& 0x14) == 0x14))) {
2923 /* VCVT from int is always from S reg regardless of dp bit.
2924 * VCVT with immediate frac_bits has same format as SREG_M
2926 rm
= VFP_SREG_M(insn
);
2928 VFP_DREG_M(rm
, insn
);
2931 rn
= VFP_SREG_N(insn
);
2932 if (op
== 15 && rn
== 15) {
2933 /* Double precision destination. */
2934 VFP_DREG_D(rd
, insn
);
2936 rd
= VFP_SREG_D(insn
);
2938 /* NB that we implicitly rely on the encoding for the frac_bits
2939 * in VCVT of fixed to float being the same as that of an SREG_M
2941 rm
= VFP_SREG_M(insn
);
2944 veclen
= s
->vec_len
;
2945 if (op
== 15 && rn
> 3)
2948 /* Shut up compiler warnings. */
2959 /* Figure out what type of vector operation this is. */
2960 if ((rd
& bank_mask
) == 0) {
2965 delta_d
= (s
->vec_stride
>> 1) + 1;
2967 delta_d
= s
->vec_stride
+ 1;
2969 if ((rm
& bank_mask
) == 0) {
2970 /* mixed scalar/vector */
2979 /* Load the initial operands. */
2984 /* Integer source */
2985 gen_mov_F0_vreg(0, rm
);
2990 gen_mov_F0_vreg(dp
, rd
);
2991 gen_mov_F1_vreg(dp
, rm
);
2995 /* Compare with zero */
2996 gen_mov_F0_vreg(dp
, rd
);
3007 /* Source and destination the same. */
3008 gen_mov_F0_vreg(dp
, rd
);
3011 /* One source operand. */
3012 gen_mov_F0_vreg(dp
, rm
);
3016 /* Two source operands. */
3017 gen_mov_F0_vreg(dp
, rn
);
3018 gen_mov_F1_vreg(dp
, rm
);
3022 /* Perform the calculation. */
3024 case 0: /* mac: fd + (fn * fm) */
3026 gen_mov_F1_vreg(dp
, rd
);
3029 case 1: /* nmac: fd - (fn * fm) */
3032 gen_mov_F1_vreg(dp
, rd
);
3035 case 2: /* msc: -fd + (fn * fm) */
3037 gen_mov_F1_vreg(dp
, rd
);
3040 case 3: /* nmsc: -fd - (fn * fm) */
3043 gen_mov_F1_vreg(dp
, rd
);
3046 case 4: /* mul: fn * fm */
3049 case 5: /* nmul: -(fn * fm) */
3053 case 6: /* add: fn + fm */
3056 case 7: /* sub: fn - fm */
3059 case 8: /* div: fn / fm */
3062 case 14: /* fconst */
3063 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3066 n
= (insn
<< 12) & 0x80000000;
3067 i
= ((insn
>> 12) & 0x70) | (insn
& 0xf);
3074 tcg_gen_movi_i64(cpu_F0d
, ((uint64_t)n
) << 32);
3081 tcg_gen_movi_i32(cpu_F0s
, n
);
3084 case 15: /* extension space */
3098 case 4: /* vcvtb.f32.f16 */
3099 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3101 tmp
= gen_vfp_mrs();
3102 tcg_gen_ext16u_i32(tmp
, tmp
);
3103 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s
, tmp
, cpu_env
);
3104 tcg_temp_free_i32(tmp
);
3106 case 5: /* vcvtt.f32.f16 */
3107 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3109 tmp
= gen_vfp_mrs();
3110 tcg_gen_shri_i32(tmp
, tmp
, 16);
3111 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s
, tmp
, cpu_env
);
3112 tcg_temp_free_i32(tmp
);
3114 case 6: /* vcvtb.f16.f32 */
3115 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3117 tmp
= tcg_temp_new_i32();
3118 gen_helper_vfp_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
3119 gen_mov_F0_vreg(0, rd
);
3120 tmp2
= gen_vfp_mrs();
3121 tcg_gen_andi_i32(tmp2
, tmp2
, 0xffff0000);
3122 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
3123 tcg_temp_free_i32(tmp2
);
3126 case 7: /* vcvtt.f16.f32 */
3127 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
))
3129 tmp
= tcg_temp_new_i32();
3130 gen_helper_vfp_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
3131 tcg_gen_shli_i32(tmp
, tmp
, 16);
3132 gen_mov_F0_vreg(0, rd
);
3133 tmp2
= gen_vfp_mrs();
3134 tcg_gen_ext16u_i32(tmp2
, tmp2
);
3135 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
3136 tcg_temp_free_i32(tmp2
);
3148 case 11: /* cmpez */
3152 case 15: /* single<->double conversion */
3154 gen_helper_vfp_fcvtsd(cpu_F0s
, cpu_F0d
, cpu_env
);
3156 gen_helper_vfp_fcvtds(cpu_F0d
, cpu_F0s
, cpu_env
);
3158 case 16: /* fuito */
3161 case 17: /* fsito */
3164 case 20: /* fshto */
3165 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3167 gen_vfp_shto(dp
, 16 - rm
);
3169 case 21: /* fslto */
3170 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3172 gen_vfp_slto(dp
, 32 - rm
);
3174 case 22: /* fuhto */
3175 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3177 gen_vfp_uhto(dp
, 16 - rm
);
3179 case 23: /* fulto */
3180 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3182 gen_vfp_ulto(dp
, 32 - rm
);
3184 case 24: /* ftoui */
3187 case 25: /* ftouiz */
3190 case 26: /* ftosi */
3193 case 27: /* ftosiz */
3196 case 28: /* ftosh */
3197 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3199 gen_vfp_tosh(dp
, 16 - rm
);
3201 case 29: /* ftosl */
3202 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3204 gen_vfp_tosl(dp
, 32 - rm
);
3206 case 30: /* ftouh */
3207 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3209 gen_vfp_touh(dp
, 16 - rm
);
3211 case 31: /* ftoul */
3212 if (!arm_feature(env
, ARM_FEATURE_VFP3
))
3214 gen_vfp_toul(dp
, 32 - rm
);
3216 default: /* undefined */
3217 printf ("rn:%d\n", rn
);
3221 default: /* undefined */
3222 printf ("op:%d\n", op
);
3226 /* Write back the result. */
3227 if (op
== 15 && (rn
>= 8 && rn
<= 11))
3228 ; /* Comparison, do nothing. */
3229 else if (op
== 15 && dp
&& ((rn
& 0x1c) == 0x18))
3230 /* VCVT double to int: always integer result. */
3231 gen_mov_vreg_F0(0, rd
);
3232 else if (op
== 15 && rn
== 15)
3234 gen_mov_vreg_F0(!dp
, rd
);
3236 gen_mov_vreg_F0(dp
, rd
);
3238 /* break out of the loop if we have finished */
3242 if (op
== 15 && delta_m
== 0) {
3243 /* single source one-many */
3245 rd
= ((rd
+ delta_d
) & (bank_mask
- 1))
3247 gen_mov_vreg_F0(dp
, rd
);
3251 /* Setup the next operands. */
3253 rd
= ((rd
+ delta_d
) & (bank_mask
- 1))
3257 /* One source operand. */
3258 rm
= ((rm
+ delta_m
) & (bank_mask
- 1))
3260 gen_mov_F0_vreg(dp
, rm
);
3262 /* Two source operands. */
3263 rn
= ((rn
+ delta_d
) & (bank_mask
- 1))
3265 gen_mov_F0_vreg(dp
, rn
);
3267 rm
= ((rm
+ delta_m
) & (bank_mask
- 1))
3269 gen_mov_F1_vreg(dp
, rm
);
3277 if ((insn
& 0x03e00000) == 0x00400000) {
3278 /* two-register transfer */
3279 rn
= (insn
>> 16) & 0xf;
3280 rd
= (insn
>> 12) & 0xf;
3282 VFP_DREG_M(rm
, insn
);
3284 rm
= VFP_SREG_M(insn
);
3287 if (insn
& ARM_CP_RW_BIT
) {
3290 gen_mov_F0_vreg(0, rm
* 2);
3291 tmp
= gen_vfp_mrs();
3292 store_reg(s
, rd
, tmp
);
3293 gen_mov_F0_vreg(0, rm
* 2 + 1);
3294 tmp
= gen_vfp_mrs();
3295 store_reg(s
, rn
, tmp
);
3297 gen_mov_F0_vreg(0, rm
);
3298 tmp
= gen_vfp_mrs();
3299 store_reg(s
, rd
, tmp
);
3300 gen_mov_F0_vreg(0, rm
+ 1);
3301 tmp
= gen_vfp_mrs();
3302 store_reg(s
, rn
, tmp
);
3307 tmp
= load_reg(s
, rd
);
3309 gen_mov_vreg_F0(0, rm
* 2);
3310 tmp
= load_reg(s
, rn
);
3312 gen_mov_vreg_F0(0, rm
* 2 + 1);
3314 tmp
= load_reg(s
, rd
);
3316 gen_mov_vreg_F0(0, rm
);
3317 tmp
= load_reg(s
, rn
);
3319 gen_mov_vreg_F0(0, rm
+ 1);
3324 rn
= (insn
>> 16) & 0xf;
3326 VFP_DREG_D(rd
, insn
);
3328 rd
= VFP_SREG_D(insn
);
3329 if (s
->thumb
&& rn
== 15) {
3330 addr
= tcg_temp_new_i32();
3331 tcg_gen_movi_i32(addr
, s
->pc
& ~2);
3333 addr
= load_reg(s
, rn
);
3335 if ((insn
& 0x01200000) == 0x01000000) {
3336 /* Single load/store */
3337 offset
= (insn
& 0xff) << 2;
3338 if ((insn
& (1 << 23)) == 0)
3340 tcg_gen_addi_i32(addr
, addr
, offset
);
3341 if (insn
& (1 << 20)) {
3342 gen_vfp_ld(s
, dp
, addr
);
3343 gen_mov_vreg_F0(dp
, rd
);
3345 gen_mov_F0_vreg(dp
, rd
);
3346 gen_vfp_st(s
, dp
, addr
);
3348 tcg_temp_free_i32(addr
);
3350 /* load/store multiple */
3352 n
= (insn
>> 1) & 0x7f;
3356 if (insn
& (1 << 24)) /* pre-decrement */
3357 tcg_gen_addi_i32(addr
, addr
, -((insn
& 0xff) << 2));
3363 for (i
= 0; i
< n
; i
++) {
3364 if (insn
& ARM_CP_RW_BIT
) {
3366 gen_vfp_ld(s
, dp
, addr
);
3367 gen_mov_vreg_F0(dp
, rd
+ i
);
3370 gen_mov_F0_vreg(dp
, rd
+ i
);
3371 gen_vfp_st(s
, dp
, addr
);
3373 tcg_gen_addi_i32(addr
, addr
, offset
);
3375 if (insn
& (1 << 21)) {
3377 if (insn
& (1 << 24))
3378 offset
= -offset
* n
;
3379 else if (dp
&& (insn
& 1))
3385 tcg_gen_addi_i32(addr
, addr
, offset
);
3386 store_reg(s
, rn
, addr
);
3388 tcg_temp_free_i32(addr
);
3394 /* Should never happen. */
3400 static inline void gen_goto_tb(DisasContext
*s
, int n
, uint32_t dest
)
3402 TranslationBlock
*tb
;
3405 if ((tb
->pc
& TARGET_PAGE_MASK
) == (dest
& TARGET_PAGE_MASK
)) {
3407 gen_set_pc_im(dest
);
3408 tcg_gen_exit_tb((tcg_target_long
)tb
+ n
);
3410 gen_set_pc_im(dest
);
3415 static inline void gen_jmp (DisasContext
*s
, uint32_t dest
)
3417 if (unlikely(s
->singlestep_enabled
)) {
3418 /* An indirect jump so that we still trigger the debug exception. */
3423 gen_goto_tb(s
, 0, dest
);
3424 s
->is_jmp
= DISAS_TB_JUMP
;
3428 static inline void gen_mulxy(TCGv t0
, TCGv t1
, int x
, int y
)
3431 tcg_gen_sari_i32(t0
, t0
, 16);
3435 tcg_gen_sari_i32(t1
, t1
, 16);
3438 tcg_gen_mul_i32(t0
, t0
, t1
);
3441 /* Return the mask of PSR bits set by a MSR instruction. */
3442 static uint32_t msr_mask(CPUState
*env
, DisasContext
*s
, int flags
, int spsr
) {
3446 if (flags
& (1 << 0))
3448 if (flags
& (1 << 1))
3450 if (flags
& (1 << 2))
3452 if (flags
& (1 << 3))
3455 /* Mask out undefined bits. */
3456 mask
&= ~CPSR_RESERVED
;
3457 if (!arm_feature(env
, ARM_FEATURE_V4T
))
3459 if (!arm_feature(env
, ARM_FEATURE_V5
))
3460 mask
&= ~CPSR_Q
; /* V5TE in reality*/
3461 if (!arm_feature(env
, ARM_FEATURE_V6
))
3462 mask
&= ~(CPSR_E
| CPSR_GE
);
3463 if (!arm_feature(env
, ARM_FEATURE_THUMB2
))
3465 /* Mask out execution state bits. */
3468 /* Mask out privileged bits. */
3474 /* Returns nonzero if access to the PSR is not permitted. Marks t0 as dead. */
3475 static int gen_set_psr(DisasContext
*s
, uint32_t mask
, int spsr
, TCGv t0
)
3479 /* ??? This is also undefined in system mode. */
3483 tmp
= load_cpu_field(spsr
);
3484 tcg_gen_andi_i32(tmp
, tmp
, ~mask
);
3485 tcg_gen_andi_i32(t0
, t0
, mask
);
3486 tcg_gen_or_i32(tmp
, tmp
, t0
);
3487 store_cpu_field(tmp
, spsr
);
3489 gen_set_cpsr(t0
, mask
);
3491 tcg_temp_free_i32(t0
);
3496 /* Returns nonzero if access to the PSR is not permitted. */
3497 static int gen_set_psr_im(DisasContext
*s
, uint32_t mask
, int spsr
, uint32_t val
)
3500 tmp
= tcg_temp_new_i32();
3501 tcg_gen_movi_i32(tmp
, val
);
3502 return gen_set_psr(s
, mask
, spsr
, tmp
);
3505 /* Generate an old-style exception return. Marks pc as dead. */
3506 static void gen_exception_return(DisasContext
*s
, TCGv pc
)
3509 store_reg(s
, 15, pc
);
3510 tmp
= load_cpu_field(spsr
);
3511 gen_set_cpsr(tmp
, 0xffffffff);
3512 tcg_temp_free_i32(tmp
);
3513 s
->is_jmp
= DISAS_UPDATE
;
3516 /* Generate a v6 exception return. Marks both values as dead. */
3517 static void gen_rfe(DisasContext
*s
, TCGv pc
, TCGv cpsr
)
3519 gen_set_cpsr(cpsr
, 0xffffffff);
3520 tcg_temp_free_i32(cpsr
);
3521 store_reg(s
, 15, pc
);
3522 s
->is_jmp
= DISAS_UPDATE
;
3526 gen_set_condexec (DisasContext
*s
)
3528 if (s
->condexec_mask
) {
3529 uint32_t val
= (s
->condexec_cond
<< 4) | (s
->condexec_mask
>> 1);
3530 TCGv tmp
= tcg_temp_new_i32();
3531 tcg_gen_movi_i32(tmp
, val
);
3532 store_cpu_field(tmp
, condexec_bits
);
3536 static void gen_exception_insn(DisasContext
*s
, int offset
, int excp
)
3538 gen_set_condexec(s
);
3539 gen_set_pc_im(s
->pc
- offset
);
3540 gen_exception(excp
);
3541 s
->is_jmp
= DISAS_JUMP
;
3544 static void gen_nop_hint(DisasContext
*s
, int val
)
3548 gen_set_pc_im(s
->pc
);
3549 s
->is_jmp
= DISAS_WFI
;
3553 /* TODO: Implement SEV and WFE. May help SMP performance. */
3559 #define CPU_V001 cpu_V0, cpu_V0, cpu_V1
3561 static inline void gen_neon_add(int size
, TCGv t0
, TCGv t1
)
3564 case 0: gen_helper_neon_add_u8(t0
, t0
, t1
); break;
3565 case 1: gen_helper_neon_add_u16(t0
, t0
, t1
); break;
3566 case 2: tcg_gen_add_i32(t0
, t0
, t1
); break;
3571 static inline void gen_neon_rsb(int size
, TCGv t0
, TCGv t1
)
3574 case 0: gen_helper_neon_sub_u8(t0
, t1
, t0
); break;
3575 case 1: gen_helper_neon_sub_u16(t0
, t1
, t0
); break;
3576 case 2: tcg_gen_sub_i32(t0
, t1
, t0
); break;
3581 /* 32-bit pairwise ops end up the same as the elementwise versions. */
3582 #define gen_helper_neon_pmax_s32 gen_helper_neon_max_s32
3583 #define gen_helper_neon_pmax_u32 gen_helper_neon_max_u32
3584 #define gen_helper_neon_pmin_s32 gen_helper_neon_min_s32
3585 #define gen_helper_neon_pmin_u32 gen_helper_neon_min_u32
3587 #define GEN_NEON_INTEGER_OP_ENV(name) do { \
3588 switch ((size << 1) | u) { \
3590 gen_helper_neon_##name##_s8(tmp, cpu_env, tmp, tmp2); \
3593 gen_helper_neon_##name##_u8(tmp, cpu_env, tmp, tmp2); \
3596 gen_helper_neon_##name##_s16(tmp, cpu_env, tmp, tmp2); \
3599 gen_helper_neon_##name##_u16(tmp, cpu_env, tmp, tmp2); \
3602 gen_helper_neon_##name##_s32(tmp, cpu_env, tmp, tmp2); \
3605 gen_helper_neon_##name##_u32(tmp, cpu_env, tmp, tmp2); \
3607 default: return 1; \
3610 #define GEN_NEON_INTEGER_OP(name) do { \
3611 switch ((size << 1) | u) { \
3613 gen_helper_neon_##name##_s8(tmp, tmp, tmp2); \
3616 gen_helper_neon_##name##_u8(tmp, tmp, tmp2); \
3619 gen_helper_neon_##name##_s16(tmp, tmp, tmp2); \
3622 gen_helper_neon_##name##_u16(tmp, tmp, tmp2); \
3625 gen_helper_neon_##name##_s32(tmp, tmp, tmp2); \
3628 gen_helper_neon_##name##_u32(tmp, tmp, tmp2); \
3630 default: return 1; \
3633 static TCGv
neon_load_scratch(int scratch
)
3635 TCGv tmp
= tcg_temp_new_i32();
3636 tcg_gen_ld_i32(tmp
, cpu_env
, offsetof(CPUARMState
, vfp
.scratch
[scratch
]));
3640 static void neon_store_scratch(int scratch
, TCGv var
)
3642 tcg_gen_st_i32(var
, cpu_env
, offsetof(CPUARMState
, vfp
.scratch
[scratch
]));
3643 tcg_temp_free_i32(var
);
3646 static inline TCGv
neon_get_scalar(int size
, int reg
)
3650 tmp
= neon_load_reg(reg
& 7, reg
>> 4);
3652 gen_neon_dup_high16(tmp
);
3654 gen_neon_dup_low16(tmp
);
3657 tmp
= neon_load_reg(reg
& 15, reg
>> 4);
3662 static int gen_neon_unzip(int rd
, int rm
, int size
, int q
)
3665 if (!q
&& size
== 2) {
3668 tmp
= tcg_const_i32(rd
);
3669 tmp2
= tcg_const_i32(rm
);
3673 gen_helper_neon_qunzip8(tmp
, tmp2
);
3676 gen_helper_neon_qunzip16(tmp
, tmp2
);
3679 gen_helper_neon_qunzip32(tmp
, tmp2
);
3687 gen_helper_neon_unzip8(tmp
, tmp2
);
3690 gen_helper_neon_unzip16(tmp
, tmp2
);
3696 tcg_temp_free_i32(tmp
);
3697 tcg_temp_free_i32(tmp2
);
3701 static int gen_neon_zip(int rd
, int rm
, int size
, int q
)
3704 if (!q
&& size
== 2) {
3707 tmp
= tcg_const_i32(rd
);
3708 tmp2
= tcg_const_i32(rm
);
3712 gen_helper_neon_qzip8(tmp
, tmp2
);
3715 gen_helper_neon_qzip16(tmp
, tmp2
);
3718 gen_helper_neon_qzip32(tmp
, tmp2
);
3726 gen_helper_neon_zip8(tmp
, tmp2
);
3729 gen_helper_neon_zip16(tmp
, tmp2
);
3735 tcg_temp_free_i32(tmp
);
3736 tcg_temp_free_i32(tmp2
);
3740 static void gen_neon_trn_u8(TCGv t0
, TCGv t1
)
3744 rd
= tcg_temp_new_i32();
3745 tmp
= tcg_temp_new_i32();
3747 tcg_gen_shli_i32(rd
, t0
, 8);
3748 tcg_gen_andi_i32(rd
, rd
, 0xff00ff00);
3749 tcg_gen_andi_i32(tmp
, t1
, 0x00ff00ff);
3750 tcg_gen_or_i32(rd
, rd
, tmp
);
3752 tcg_gen_shri_i32(t1
, t1
, 8);
3753 tcg_gen_andi_i32(t1
, t1
, 0x00ff00ff);
3754 tcg_gen_andi_i32(tmp
, t0
, 0xff00ff00);
3755 tcg_gen_or_i32(t1
, t1
, tmp
);
3756 tcg_gen_mov_i32(t0
, rd
);
3758 tcg_temp_free_i32(tmp
);
3759 tcg_temp_free_i32(rd
);
3762 static void gen_neon_trn_u16(TCGv t0
, TCGv t1
)
3766 rd
= tcg_temp_new_i32();
3767 tmp
= tcg_temp_new_i32();
3769 tcg_gen_shli_i32(rd
, t0
, 16);
3770 tcg_gen_andi_i32(tmp
, t1
, 0xffff);
3771 tcg_gen_or_i32(rd
, rd
, tmp
);
3772 tcg_gen_shri_i32(t1
, t1
, 16);
3773 tcg_gen_andi_i32(tmp
, t0
, 0xffff0000);
3774 tcg_gen_or_i32(t1
, t1
, tmp
);
3775 tcg_gen_mov_i32(t0
, rd
);
3777 tcg_temp_free_i32(tmp
);
3778 tcg_temp_free_i32(rd
);
3786 } neon_ls_element_type
[11] = {
3800 /* Translate a NEON load/store element instruction. Return nonzero if the
3801 instruction is invalid. */
3802 static int disas_neon_ls_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
3821 if (!s
->vfp_enabled
)
3823 VFP_DREG_D(rd
, insn
);
3824 rn
= (insn
>> 16) & 0xf;
3826 load
= (insn
& (1 << 21)) != 0;
3827 if ((insn
& (1 << 23)) == 0) {
3828 /* Load store all elements. */
3829 op
= (insn
>> 8) & 0xf;
3830 size
= (insn
>> 6) & 3;
3833 nregs
= neon_ls_element_type
[op
].nregs
;
3834 interleave
= neon_ls_element_type
[op
].interleave
;
3835 spacing
= neon_ls_element_type
[op
].spacing
;
3836 if (size
== 3 && (interleave
| spacing
) != 1)
3838 addr
= tcg_temp_new_i32();
3839 load_reg_var(s
, addr
, rn
);
3840 stride
= (1 << size
) * interleave
;
3841 for (reg
= 0; reg
< nregs
; reg
++) {
3842 if (interleave
> 2 || (interleave
== 2 && nregs
== 2)) {
3843 load_reg_var(s
, addr
, rn
);
3844 tcg_gen_addi_i32(addr
, addr
, (1 << size
) * reg
);
3845 } else if (interleave
== 2 && nregs
== 4 && reg
== 2) {
3846 load_reg_var(s
, addr
, rn
);
3847 tcg_gen_addi_i32(addr
, addr
, 1 << size
);
3851 tmp64
= gen_ld64(addr
, IS_USER(s
));
3852 neon_store_reg64(tmp64
, rd
);
3853 tcg_temp_free_i64(tmp64
);
3855 tmp64
= tcg_temp_new_i64();
3856 neon_load_reg64(tmp64
, rd
);
3857 gen_st64(tmp64
, addr
, IS_USER(s
));
3859 tcg_gen_addi_i32(addr
, addr
, stride
);
3861 for (pass
= 0; pass
< 2; pass
++) {
3864 tmp
= gen_ld32(addr
, IS_USER(s
));
3865 neon_store_reg(rd
, pass
, tmp
);
3867 tmp
= neon_load_reg(rd
, pass
);
3868 gen_st32(tmp
, addr
, IS_USER(s
));
3870 tcg_gen_addi_i32(addr
, addr
, stride
);
3871 } else if (size
== 1) {
3873 tmp
= gen_ld16u(addr
, IS_USER(s
));
3874 tcg_gen_addi_i32(addr
, addr
, stride
);
3875 tmp2
= gen_ld16u(addr
, IS_USER(s
));
3876 tcg_gen_addi_i32(addr
, addr
, stride
);
3877 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
3878 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
3879 tcg_temp_free_i32(tmp2
);
3880 neon_store_reg(rd
, pass
, tmp
);
3882 tmp
= neon_load_reg(rd
, pass
);
3883 tmp2
= tcg_temp_new_i32();
3884 tcg_gen_shri_i32(tmp2
, tmp
, 16);
3885 gen_st16(tmp
, addr
, IS_USER(s
));
3886 tcg_gen_addi_i32(addr
, addr
, stride
);
3887 gen_st16(tmp2
, addr
, IS_USER(s
));
3888 tcg_gen_addi_i32(addr
, addr
, stride
);
3890 } else /* size == 0 */ {
3893 for (n
= 0; n
< 4; n
++) {
3894 tmp
= gen_ld8u(addr
, IS_USER(s
));
3895 tcg_gen_addi_i32(addr
, addr
, stride
);
3899 tcg_gen_shli_i32(tmp
, tmp
, n
* 8);
3900 tcg_gen_or_i32(tmp2
, tmp2
, tmp
);
3901 tcg_temp_free_i32(tmp
);
3904 neon_store_reg(rd
, pass
, tmp2
);
3906 tmp2
= neon_load_reg(rd
, pass
);
3907 for (n
= 0; n
< 4; n
++) {
3908 tmp
= tcg_temp_new_i32();
3910 tcg_gen_mov_i32(tmp
, tmp2
);
3912 tcg_gen_shri_i32(tmp
, tmp2
, n
* 8);
3914 gen_st8(tmp
, addr
, IS_USER(s
));
3915 tcg_gen_addi_i32(addr
, addr
, stride
);
3917 tcg_temp_free_i32(tmp2
);
3924 tcg_temp_free_i32(addr
);
3927 size
= (insn
>> 10) & 3;
3929 /* Load single element to all lanes. */
3930 int a
= (insn
>> 4) & 1;
3934 size
= (insn
>> 6) & 3;
3935 nregs
= ((insn
>> 8) & 3) + 1;
3938 if (nregs
!= 4 || a
== 0) {
3941 /* For VLD4 size==3 a == 1 means 32 bits at 16 byte alignment */
3944 if (nregs
== 1 && a
== 1 && size
== 0) {
3947 if (nregs
== 3 && a
== 1) {
3950 addr
= tcg_temp_new_i32();
3951 load_reg_var(s
, addr
, rn
);
3953 /* VLD1 to all lanes: bit 5 indicates how many Dregs to write */
3954 tmp
= gen_load_and_replicate(s
, addr
, size
);
3955 tcg_gen_st_i32(tmp
, cpu_env
, neon_reg_offset(rd
, 0));
3956 tcg_gen_st_i32(tmp
, cpu_env
, neon_reg_offset(rd
, 1));
3957 if (insn
& (1 << 5)) {
3958 tcg_gen_st_i32(tmp
, cpu_env
, neon_reg_offset(rd
+ 1, 0));
3959 tcg_gen_st_i32(tmp
, cpu_env
, neon_reg_offset(rd
+ 1, 1));
3961 tcg_temp_free_i32(tmp
);
3963 /* VLD2/3/4 to all lanes: bit 5 indicates register stride */
3964 stride
= (insn
& (1 << 5)) ? 2 : 1;
3965 for (reg
= 0; reg
< nregs
; reg
++) {
3966 tmp
= gen_load_and_replicate(s
, addr
, size
);
3967 tcg_gen_st_i32(tmp
, cpu_env
, neon_reg_offset(rd
, 0));
3968 tcg_gen_st_i32(tmp
, cpu_env
, neon_reg_offset(rd
, 1));
3969 tcg_temp_free_i32(tmp
);
3970 tcg_gen_addi_i32(addr
, addr
, 1 << size
);
3974 tcg_temp_free_i32(addr
);
3975 stride
= (1 << size
) * nregs
;
3977 /* Single element. */
3978 int idx
= (insn
>> 4) & 0xf;
3979 pass
= (insn
>> 7) & 1;
3982 shift
= ((insn
>> 5) & 3) * 8;
3986 shift
= ((insn
>> 6) & 1) * 16;
3987 stride
= (insn
& (1 << 5)) ? 2 : 1;
3991 stride
= (insn
& (1 << 6)) ? 2 : 1;
3996 nregs
= ((insn
>> 8) & 3) + 1;
3997 /* Catch the UNDEF cases. This is unavoidably a bit messy. */
4000 if (((idx
& (1 << size
)) != 0) ||
4001 (size
== 2 && ((idx
& 3) == 1 || (idx
& 3) == 2))) {
4006 if ((idx
& 1) != 0) {
4011 if (size
== 2 && (idx
& 2) != 0) {
4016 if ((size
== 2) && ((idx
& 3) == 3)) {
4023 if ((rd
+ stride
* (nregs
- 1)) > 31) {
4024 /* Attempts to write off the end of the register file
4025 * are UNPREDICTABLE; we choose to UNDEF because otherwise
4026 * the neon_load_reg() would write off the end of the array.
4030 addr
= tcg_temp_new_i32();
4031 load_reg_var(s
, addr
, rn
);
4032 for (reg
= 0; reg
< nregs
; reg
++) {
4036 tmp
= gen_ld8u(addr
, IS_USER(s
));
4039 tmp
= gen_ld16u(addr
, IS_USER(s
));
4042 tmp
= gen_ld32(addr
, IS_USER(s
));
4044 default: /* Avoid compiler warnings. */
4048 tmp2
= neon_load_reg(rd
, pass
);
4049 gen_bfi(tmp
, tmp2
, tmp
, shift
, size
? 0xffff : 0xff);
4050 tcg_temp_free_i32(tmp2
);
4052 neon_store_reg(rd
, pass
, tmp
);
4053 } else { /* Store */
4054 tmp
= neon_load_reg(rd
, pass
);
4056 tcg_gen_shri_i32(tmp
, tmp
, shift
);
4059 gen_st8(tmp
, addr
, IS_USER(s
));
4062 gen_st16(tmp
, addr
, IS_USER(s
));
4065 gen_st32(tmp
, addr
, IS_USER(s
));
4070 tcg_gen_addi_i32(addr
, addr
, 1 << size
);
4072 tcg_temp_free_i32(addr
);
4073 stride
= nregs
* (1 << size
);
4079 base
= load_reg(s
, rn
);
4081 tcg_gen_addi_i32(base
, base
, stride
);
4084 index
= load_reg(s
, rm
);
4085 tcg_gen_add_i32(base
, base
, index
);
4086 tcg_temp_free_i32(index
);
4088 store_reg(s
, rn
, base
);
4093 /* Bitwise select. dest = c ? t : f. Clobbers T and F. */
4094 static void gen_neon_bsl(TCGv dest
, TCGv t
, TCGv f
, TCGv c
)
4096 tcg_gen_and_i32(t
, t
, c
);
4097 tcg_gen_andc_i32(f
, f
, c
);
4098 tcg_gen_or_i32(dest
, t
, f
);
4101 static inline void gen_neon_narrow(int size
, TCGv dest
, TCGv_i64 src
)
4104 case 0: gen_helper_neon_narrow_u8(dest
, src
); break;
4105 case 1: gen_helper_neon_narrow_u16(dest
, src
); break;
4106 case 2: tcg_gen_trunc_i64_i32(dest
, src
); break;
4111 static inline void gen_neon_narrow_sats(int size
, TCGv dest
, TCGv_i64 src
)
4114 case 0: gen_helper_neon_narrow_sat_s8(dest
, src
); break;
4115 case 1: gen_helper_neon_narrow_sat_s16(dest
, src
); break;
4116 case 2: gen_helper_neon_narrow_sat_s32(dest
, src
); break;
4121 static inline void gen_neon_narrow_satu(int size
, TCGv dest
, TCGv_i64 src
)
4124 case 0: gen_helper_neon_narrow_sat_u8(dest
, src
); break;
4125 case 1: gen_helper_neon_narrow_sat_u16(dest
, src
); break;
4126 case 2: gen_helper_neon_narrow_sat_u32(dest
, src
); break;
4131 static inline void gen_neon_unarrow_sats(int size
, TCGv dest
, TCGv_i64 src
)
4134 case 0: gen_helper_neon_unarrow_sat8(dest
, src
); break;
4135 case 1: gen_helper_neon_unarrow_sat16(dest
, src
); break;
4136 case 2: gen_helper_neon_unarrow_sat32(dest
, src
); break;
4141 static inline void gen_neon_shift_narrow(int size
, TCGv var
, TCGv shift
,
4147 case 1: gen_helper_neon_rshl_u16(var
, var
, shift
); break;
4148 case 2: gen_helper_neon_rshl_u32(var
, var
, shift
); break;
4153 case 1: gen_helper_neon_rshl_s16(var
, var
, shift
); break;
4154 case 2: gen_helper_neon_rshl_s32(var
, var
, shift
); break;
4161 case 1: gen_helper_neon_shl_u16(var
, var
, shift
); break;
4162 case 2: gen_helper_neon_shl_u32(var
, var
, shift
); break;
4167 case 1: gen_helper_neon_shl_s16(var
, var
, shift
); break;
4168 case 2: gen_helper_neon_shl_s32(var
, var
, shift
); break;
4175 static inline void gen_neon_widen(TCGv_i64 dest
, TCGv src
, int size
, int u
)
4179 case 0: gen_helper_neon_widen_u8(dest
, src
); break;
4180 case 1: gen_helper_neon_widen_u16(dest
, src
); break;
4181 case 2: tcg_gen_extu_i32_i64(dest
, src
); break;
4186 case 0: gen_helper_neon_widen_s8(dest
, src
); break;
4187 case 1: gen_helper_neon_widen_s16(dest
, src
); break;
4188 case 2: tcg_gen_ext_i32_i64(dest
, src
); break;
4192 tcg_temp_free_i32(src
);
4195 static inline void gen_neon_addl(int size
)
4198 case 0: gen_helper_neon_addl_u16(CPU_V001
); break;
4199 case 1: gen_helper_neon_addl_u32(CPU_V001
); break;
4200 case 2: tcg_gen_add_i64(CPU_V001
); break;
4205 static inline void gen_neon_subl(int size
)
4208 case 0: gen_helper_neon_subl_u16(CPU_V001
); break;
4209 case 1: gen_helper_neon_subl_u32(CPU_V001
); break;
4210 case 2: tcg_gen_sub_i64(CPU_V001
); break;
4215 static inline void gen_neon_negl(TCGv_i64 var
, int size
)
4218 case 0: gen_helper_neon_negl_u16(var
, var
); break;
4219 case 1: gen_helper_neon_negl_u32(var
, var
); break;
4220 case 2: gen_helper_neon_negl_u64(var
, var
); break;
4225 static inline void gen_neon_addl_saturate(TCGv_i64 op0
, TCGv_i64 op1
, int size
)
4228 case 1: gen_helper_neon_addl_saturate_s32(op0
, op0
, op1
); break;
4229 case 2: gen_helper_neon_addl_saturate_s64(op0
, op0
, op1
); break;
4234 static inline void gen_neon_mull(TCGv_i64 dest
, TCGv a
, TCGv b
, int size
, int u
)
4238 switch ((size
<< 1) | u
) {
4239 case 0: gen_helper_neon_mull_s8(dest
, a
, b
); break;
4240 case 1: gen_helper_neon_mull_u8(dest
, a
, b
); break;
4241 case 2: gen_helper_neon_mull_s16(dest
, a
, b
); break;
4242 case 3: gen_helper_neon_mull_u16(dest
, a
, b
); break;
4244 tmp
= gen_muls_i64_i32(a
, b
);
4245 tcg_gen_mov_i64(dest
, tmp
);
4246 tcg_temp_free_i64(tmp
);
4249 tmp
= gen_mulu_i64_i32(a
, b
);
4250 tcg_gen_mov_i64(dest
, tmp
);
4251 tcg_temp_free_i64(tmp
);
4256 /* gen_helper_neon_mull_[su]{8|16} do not free their parameters.
4257 Don't forget to clean them now. */
4259 tcg_temp_free_i32(a
);
4260 tcg_temp_free_i32(b
);
4264 static void gen_neon_narrow_op(int op
, int u
, int size
, TCGv dest
, TCGv_i64 src
)
4268 gen_neon_unarrow_sats(size
, dest
, src
);
4270 gen_neon_narrow(size
, dest
, src
);
4274 gen_neon_narrow_satu(size
, dest
, src
);
4276 gen_neon_narrow_sats(size
, dest
, src
);
4281 /* Symbolic constants for op fields for Neon 3-register same-length.
4282 * The values correspond to bits [11:8,4]; see the ARM ARM DDI0406B
4285 #define NEON_3R_VHADD 0
4286 #define NEON_3R_VQADD 1
4287 #define NEON_3R_VRHADD 2
4288 #define NEON_3R_LOGIC 3 /* VAND,VBIC,VORR,VMOV,VORN,VEOR,VBIF,VBIT,VBSL */
4289 #define NEON_3R_VHSUB 4
4290 #define NEON_3R_VQSUB 5
4291 #define NEON_3R_VCGT 6
4292 #define NEON_3R_VCGE 7
4293 #define NEON_3R_VSHL 8
4294 #define NEON_3R_VQSHL 9
4295 #define NEON_3R_VRSHL 10
4296 #define NEON_3R_VQRSHL 11
4297 #define NEON_3R_VMAX 12
4298 #define NEON_3R_VMIN 13
4299 #define NEON_3R_VABD 14
4300 #define NEON_3R_VABA 15
4301 #define NEON_3R_VADD_VSUB 16
4302 #define NEON_3R_VTST_VCEQ 17
4303 #define NEON_3R_VML 18 /* VMLA, VMLAL, VMLS, VMLSL */
4304 #define NEON_3R_VMUL 19
4305 #define NEON_3R_VPMAX 20
4306 #define NEON_3R_VPMIN 21
4307 #define NEON_3R_VQDMULH_VQRDMULH 22
4308 #define NEON_3R_VPADD 23
4309 #define NEON_3R_FLOAT_ARITH 26 /* float VADD, VSUB, VPADD, VABD */
4310 #define NEON_3R_FLOAT_MULTIPLY 27 /* float VMLA, VMLS, VMUL */
4311 #define NEON_3R_FLOAT_CMP 28 /* float VCEQ, VCGE, VCGT */
4312 #define NEON_3R_FLOAT_ACMP 29 /* float VACGE, VACGT, VACLE, VACLT */
4313 #define NEON_3R_FLOAT_MINMAX 30 /* float VMIN, VMAX */
4314 #define NEON_3R_VRECPS_VRSQRTS 31 /* float VRECPS, VRSQRTS */
4316 static const uint8_t neon_3r_sizes
[] = {
4317 [NEON_3R_VHADD
] = 0x7,
4318 [NEON_3R_VQADD
] = 0xf,
4319 [NEON_3R_VRHADD
] = 0x7,
4320 [NEON_3R_LOGIC
] = 0xf, /* size field encodes op type */
4321 [NEON_3R_VHSUB
] = 0x7,
4322 [NEON_3R_VQSUB
] = 0xf,
4323 [NEON_3R_VCGT
] = 0x7,
4324 [NEON_3R_VCGE
] = 0x7,
4325 [NEON_3R_VSHL
] = 0xf,
4326 [NEON_3R_VQSHL
] = 0xf,
4327 [NEON_3R_VRSHL
] = 0xf,
4328 [NEON_3R_VQRSHL
] = 0xf,
4329 [NEON_3R_VMAX
] = 0x7,
4330 [NEON_3R_VMIN
] = 0x7,
4331 [NEON_3R_VABD
] = 0x7,
4332 [NEON_3R_VABA
] = 0x7,
4333 [NEON_3R_VADD_VSUB
] = 0xf,
4334 [NEON_3R_VTST_VCEQ
] = 0x7,
4335 [NEON_3R_VML
] = 0x7,
4336 [NEON_3R_VMUL
] = 0x7,
4337 [NEON_3R_VPMAX
] = 0x7,
4338 [NEON_3R_VPMIN
] = 0x7,
4339 [NEON_3R_VQDMULH_VQRDMULH
] = 0x6,
4340 [NEON_3R_VPADD
] = 0x7,
4341 [NEON_3R_FLOAT_ARITH
] = 0x5, /* size bit 1 encodes op */
4342 [NEON_3R_FLOAT_MULTIPLY
] = 0x5, /* size bit 1 encodes op */
4343 [NEON_3R_FLOAT_CMP
] = 0x5, /* size bit 1 encodes op */
4344 [NEON_3R_FLOAT_ACMP
] = 0x5, /* size bit 1 encodes op */
4345 [NEON_3R_FLOAT_MINMAX
] = 0x5, /* size bit 1 encodes op */
4346 [NEON_3R_VRECPS_VRSQRTS
] = 0x5, /* size bit 1 encodes op */
4349 /* Symbolic constants for op fields for Neon 2-register miscellaneous.
4350 * The values correspond to bits [17:16,10:7]; see the ARM ARM DDI0406B
4353 #define NEON_2RM_VREV64 0
4354 #define NEON_2RM_VREV32 1
4355 #define NEON_2RM_VREV16 2
4356 #define NEON_2RM_VPADDL 4
4357 #define NEON_2RM_VPADDL_U 5
4358 #define NEON_2RM_VCLS 8
4359 #define NEON_2RM_VCLZ 9
4360 #define NEON_2RM_VCNT 10
4361 #define NEON_2RM_VMVN 11
4362 #define NEON_2RM_VPADAL 12
4363 #define NEON_2RM_VPADAL_U 13
4364 #define NEON_2RM_VQABS 14
4365 #define NEON_2RM_VQNEG 15
4366 #define NEON_2RM_VCGT0 16
4367 #define NEON_2RM_VCGE0 17
4368 #define NEON_2RM_VCEQ0 18
4369 #define NEON_2RM_VCLE0 19
4370 #define NEON_2RM_VCLT0 20
4371 #define NEON_2RM_VABS 22
4372 #define NEON_2RM_VNEG 23
4373 #define NEON_2RM_VCGT0_F 24
4374 #define NEON_2RM_VCGE0_F 25
4375 #define NEON_2RM_VCEQ0_F 26
4376 #define NEON_2RM_VCLE0_F 27
4377 #define NEON_2RM_VCLT0_F 28
4378 #define NEON_2RM_VABS_F 30
4379 #define NEON_2RM_VNEG_F 31
4380 #define NEON_2RM_VSWP 32
4381 #define NEON_2RM_VTRN 33
4382 #define NEON_2RM_VUZP 34
4383 #define NEON_2RM_VZIP 35
4384 #define NEON_2RM_VMOVN 36 /* Includes VQMOVN, VQMOVUN */
4385 #define NEON_2RM_VQMOVN 37 /* Includes VQMOVUN */
4386 #define NEON_2RM_VSHLL 38
4387 #define NEON_2RM_VCVT_F16_F32 44
4388 #define NEON_2RM_VCVT_F32_F16 46
4389 #define NEON_2RM_VRECPE 56
4390 #define NEON_2RM_VRSQRTE 57
4391 #define NEON_2RM_VRECPE_F 58
4392 #define NEON_2RM_VRSQRTE_F 59
4393 #define NEON_2RM_VCVT_FS 60
4394 #define NEON_2RM_VCVT_FU 61
4395 #define NEON_2RM_VCVT_SF 62
4396 #define NEON_2RM_VCVT_UF 63
4398 static int neon_2rm_is_float_op(int op
)
4400 /* Return true if this neon 2reg-misc op is float-to-float */
4401 return (op
== NEON_2RM_VABS_F
|| op
== NEON_2RM_VNEG_F
||
4402 op
>= NEON_2RM_VRECPE_F
);
4405 /* Each entry in this array has bit n set if the insn allows
4406 * size value n (otherwise it will UNDEF). Since unallocated
4407 * op values will have no bits set they always UNDEF.
4409 static const uint8_t neon_2rm_sizes
[] = {
4410 [NEON_2RM_VREV64
] = 0x7,
4411 [NEON_2RM_VREV32
] = 0x3,
4412 [NEON_2RM_VREV16
] = 0x1,
4413 [NEON_2RM_VPADDL
] = 0x7,
4414 [NEON_2RM_VPADDL_U
] = 0x7,
4415 [NEON_2RM_VCLS
] = 0x7,
4416 [NEON_2RM_VCLZ
] = 0x7,
4417 [NEON_2RM_VCNT
] = 0x1,
4418 [NEON_2RM_VMVN
] = 0x1,
4419 [NEON_2RM_VPADAL
] = 0x7,
4420 [NEON_2RM_VPADAL_U
] = 0x7,
4421 [NEON_2RM_VQABS
] = 0x7,
4422 [NEON_2RM_VQNEG
] = 0x7,
4423 [NEON_2RM_VCGT0
] = 0x7,
4424 [NEON_2RM_VCGE0
] = 0x7,
4425 [NEON_2RM_VCEQ0
] = 0x7,
4426 [NEON_2RM_VCLE0
] = 0x7,
4427 [NEON_2RM_VCLT0
] = 0x7,
4428 [NEON_2RM_VABS
] = 0x7,
4429 [NEON_2RM_VNEG
] = 0x7,
4430 [NEON_2RM_VCGT0_F
] = 0x4,
4431 [NEON_2RM_VCGE0_F
] = 0x4,
4432 [NEON_2RM_VCEQ0_F
] = 0x4,
4433 [NEON_2RM_VCLE0_F
] = 0x4,
4434 [NEON_2RM_VCLT0_F
] = 0x4,
4435 [NEON_2RM_VABS_F
] = 0x4,
4436 [NEON_2RM_VNEG_F
] = 0x4,
4437 [NEON_2RM_VSWP
] = 0x1,
4438 [NEON_2RM_VTRN
] = 0x7,
4439 [NEON_2RM_VUZP
] = 0x7,
4440 [NEON_2RM_VZIP
] = 0x7,
4441 [NEON_2RM_VMOVN
] = 0x7,
4442 [NEON_2RM_VQMOVN
] = 0x7,
4443 [NEON_2RM_VSHLL
] = 0x7,
4444 [NEON_2RM_VCVT_F16_F32
] = 0x2,
4445 [NEON_2RM_VCVT_F32_F16
] = 0x2,
4446 [NEON_2RM_VRECPE
] = 0x4,
4447 [NEON_2RM_VRSQRTE
] = 0x4,
4448 [NEON_2RM_VRECPE_F
] = 0x4,
4449 [NEON_2RM_VRSQRTE_F
] = 0x4,
4450 [NEON_2RM_VCVT_FS
] = 0x4,
4451 [NEON_2RM_VCVT_FU
] = 0x4,
4452 [NEON_2RM_VCVT_SF
] = 0x4,
4453 [NEON_2RM_VCVT_UF
] = 0x4,
4456 /* Translate a NEON data processing instruction. Return nonzero if the
4457 instruction is invalid.
4458 We process data in a mixture of 32-bit and 64-bit chunks.
4459 Mostly we use 32-bit chunks so we can use normal scalar instructions. */
4461 static int disas_neon_data_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
4473 TCGv tmp
, tmp2
, tmp3
, tmp4
, tmp5
;
4476 if (!s
->vfp_enabled
)
4478 q
= (insn
& (1 << 6)) != 0;
4479 u
= (insn
>> 24) & 1;
4480 VFP_DREG_D(rd
, insn
);
4481 VFP_DREG_N(rn
, insn
);
4482 VFP_DREG_M(rm
, insn
);
4483 size
= (insn
>> 20) & 3;
4484 if ((insn
& (1 << 23)) == 0) {
4485 /* Three register same length. */
4486 op
= ((insn
>> 7) & 0x1e) | ((insn
>> 4) & 1);
4487 /* Catch invalid op and bad size combinations: UNDEF */
4488 if ((neon_3r_sizes
[op
] & (1 << size
)) == 0) {
4491 /* All insns of this form UNDEF for either this condition or the
4492 * superset of cases "Q==1"; we catch the latter later.
4494 if (q
&& ((rd
| rn
| rm
) & 1)) {
4497 if (size
== 3 && op
!= NEON_3R_LOGIC
) {
4498 /* 64-bit element instructions. */
4499 for (pass
= 0; pass
< (q
? 2 : 1); pass
++) {
4500 neon_load_reg64(cpu_V0
, rn
+ pass
);
4501 neon_load_reg64(cpu_V1
, rm
+ pass
);
4505 gen_helper_neon_qadd_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4507 gen_helper_neon_qadd_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4512 gen_helper_neon_qsub_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4514 gen_helper_neon_qsub_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4519 gen_helper_neon_shl_u64(cpu_V0
, cpu_V1
, cpu_V0
);
4521 gen_helper_neon_shl_s64(cpu_V0
, cpu_V1
, cpu_V0
);
4526 gen_helper_neon_qshl_u64(cpu_V0
, cpu_V1
, cpu_V0
);
4528 gen_helper_neon_qshl_s64(cpu_V0
, cpu_V1
, cpu_V0
);
4533 gen_helper_neon_rshl_u64(cpu_V0
, cpu_V1
, cpu_V0
);
4535 gen_helper_neon_rshl_s64(cpu_V0
, cpu_V1
, cpu_V0
);
4538 case NEON_3R_VQRSHL
:
4540 gen_helper_neon_qrshl_u64(cpu_V0
, cpu_V1
, cpu_V0
);
4542 gen_helper_neon_qrshl_s64(cpu_V0
, cpu_V1
, cpu_V0
);
4545 case NEON_3R_VADD_VSUB
:
4547 tcg_gen_sub_i64(CPU_V001
);
4549 tcg_gen_add_i64(CPU_V001
);
4555 neon_store_reg64(cpu_V0
, rd
+ pass
);
4564 case NEON_3R_VQRSHL
:
4567 /* Shift instruction operands are reversed. */
4582 case NEON_3R_FLOAT_ARITH
:
4583 pairwise
= (u
&& size
< 2); /* if VPADD (float) */
4585 case NEON_3R_FLOAT_MINMAX
:
4586 pairwise
= u
; /* if VPMIN/VPMAX (float) */
4588 case NEON_3R_FLOAT_CMP
:
4590 /* no encoding for U=0 C=1x */
4594 case NEON_3R_FLOAT_ACMP
:
4599 case NEON_3R_VRECPS_VRSQRTS
:
4605 if (u
&& (size
!= 0)) {
4606 /* UNDEF on invalid size for polynomial subcase */
4614 if (pairwise
&& q
) {
4615 /* All the pairwise insns UNDEF if Q is set */
4619 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
4624 tmp
= neon_load_reg(rn
, 0);
4625 tmp2
= neon_load_reg(rn
, 1);
4627 tmp
= neon_load_reg(rm
, 0);
4628 tmp2
= neon_load_reg(rm
, 1);
4632 tmp
= neon_load_reg(rn
, pass
);
4633 tmp2
= neon_load_reg(rm
, pass
);
4637 GEN_NEON_INTEGER_OP(hadd
);
4640 GEN_NEON_INTEGER_OP(qadd
);
4642 case NEON_3R_VRHADD
:
4643 GEN_NEON_INTEGER_OP(rhadd
);
4645 case NEON_3R_LOGIC
: /* Logic ops. */
4646 switch ((u
<< 2) | size
) {
4648 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
4651 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
4654 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
4657 tcg_gen_orc_i32(tmp
, tmp
, tmp2
);
4660 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
4663 tmp3
= neon_load_reg(rd
, pass
);
4664 gen_neon_bsl(tmp
, tmp
, tmp2
, tmp3
);
4665 tcg_temp_free_i32(tmp3
);
4668 tmp3
= neon_load_reg(rd
, pass
);
4669 gen_neon_bsl(tmp
, tmp
, tmp3
, tmp2
);
4670 tcg_temp_free_i32(tmp3
);
4673 tmp3
= neon_load_reg(rd
, pass
);
4674 gen_neon_bsl(tmp
, tmp3
, tmp
, tmp2
);
4675 tcg_temp_free_i32(tmp3
);
4680 GEN_NEON_INTEGER_OP(hsub
);
4683 GEN_NEON_INTEGER_OP(qsub
);
4686 GEN_NEON_INTEGER_OP(cgt
);
4689 GEN_NEON_INTEGER_OP(cge
);
4692 GEN_NEON_INTEGER_OP(shl
);
4695 GEN_NEON_INTEGER_OP(qshl
);
4698 GEN_NEON_INTEGER_OP(rshl
);
4700 case NEON_3R_VQRSHL
:
4701 GEN_NEON_INTEGER_OP(qrshl
);
4704 GEN_NEON_INTEGER_OP(max
);
4707 GEN_NEON_INTEGER_OP(min
);
4710 GEN_NEON_INTEGER_OP(abd
);
4713 GEN_NEON_INTEGER_OP(abd
);
4714 tcg_temp_free_i32(tmp2
);
4715 tmp2
= neon_load_reg(rd
, pass
);
4716 gen_neon_add(size
, tmp
, tmp2
);
4718 case NEON_3R_VADD_VSUB
:
4719 if (!u
) { /* VADD */
4720 gen_neon_add(size
, tmp
, tmp2
);
4723 case 0: gen_helper_neon_sub_u8(tmp
, tmp
, tmp2
); break;
4724 case 1: gen_helper_neon_sub_u16(tmp
, tmp
, tmp2
); break;
4725 case 2: tcg_gen_sub_i32(tmp
, tmp
, tmp2
); break;
4730 case NEON_3R_VTST_VCEQ
:
4731 if (!u
) { /* VTST */
4733 case 0: gen_helper_neon_tst_u8(tmp
, tmp
, tmp2
); break;
4734 case 1: gen_helper_neon_tst_u16(tmp
, tmp
, tmp2
); break;
4735 case 2: gen_helper_neon_tst_u32(tmp
, tmp
, tmp2
); break;
4740 case 0: gen_helper_neon_ceq_u8(tmp
, tmp
, tmp2
); break;
4741 case 1: gen_helper_neon_ceq_u16(tmp
, tmp
, tmp2
); break;
4742 case 2: gen_helper_neon_ceq_u32(tmp
, tmp
, tmp2
); break;
4747 case NEON_3R_VML
: /* VMLA, VMLAL, VMLS,VMLSL */
4749 case 0: gen_helper_neon_mul_u8(tmp
, tmp
, tmp2
); break;
4750 case 1: gen_helper_neon_mul_u16(tmp
, tmp
, tmp2
); break;
4751 case 2: tcg_gen_mul_i32(tmp
, tmp
, tmp2
); break;
4754 tcg_temp_free_i32(tmp2
);
4755 tmp2
= neon_load_reg(rd
, pass
);
4757 gen_neon_rsb(size
, tmp
, tmp2
);
4759 gen_neon_add(size
, tmp
, tmp2
);
4763 if (u
) { /* polynomial */
4764 gen_helper_neon_mul_p8(tmp
, tmp
, tmp2
);
4765 } else { /* Integer */
4767 case 0: gen_helper_neon_mul_u8(tmp
, tmp
, tmp2
); break;
4768 case 1: gen_helper_neon_mul_u16(tmp
, tmp
, tmp2
); break;
4769 case 2: tcg_gen_mul_i32(tmp
, tmp
, tmp2
); break;
4775 GEN_NEON_INTEGER_OP(pmax
);
4778 GEN_NEON_INTEGER_OP(pmin
);
4780 case NEON_3R_VQDMULH_VQRDMULH
: /* Multiply high. */
4781 if (!u
) { /* VQDMULH */
4783 case 1: gen_helper_neon_qdmulh_s16(tmp
, tmp
, tmp2
); break;
4784 case 2: gen_helper_neon_qdmulh_s32(tmp
, tmp
, tmp2
); break;
4787 } else { /* VQRDMULH */
4789 case 1: gen_helper_neon_qrdmulh_s16(tmp
, tmp
, tmp2
); break;
4790 case 2: gen_helper_neon_qrdmulh_s32(tmp
, tmp
, tmp2
); break;
4797 case 0: gen_helper_neon_padd_u8(tmp
, tmp
, tmp2
); break;
4798 case 1: gen_helper_neon_padd_u16(tmp
, tmp
, tmp2
); break;
4799 case 2: tcg_gen_add_i32(tmp
, tmp
, tmp2
); break;
4803 case NEON_3R_FLOAT_ARITH
: /* Floating point arithmetic. */
4804 switch ((u
<< 2) | size
) {
4806 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
4809 gen_helper_neon_sub_f32(tmp
, tmp
, tmp2
);
4812 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
4815 gen_helper_neon_abd_f32(tmp
, tmp
, tmp2
);
4821 case NEON_3R_FLOAT_MULTIPLY
:
4822 gen_helper_neon_mul_f32(tmp
, tmp
, tmp2
);
4824 tcg_temp_free_i32(tmp2
);
4825 tmp2
= neon_load_reg(rd
, pass
);
4827 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
4829 gen_helper_neon_sub_f32(tmp
, tmp2
, tmp
);
4833 case NEON_3R_FLOAT_CMP
:
4835 gen_helper_neon_ceq_f32(tmp
, tmp
, tmp2
);
4838 gen_helper_neon_cge_f32(tmp
, tmp
, tmp2
);
4840 gen_helper_neon_cgt_f32(tmp
, tmp
, tmp2
);
4843 case NEON_3R_FLOAT_ACMP
:
4845 gen_helper_neon_acge_f32(tmp
, tmp
, tmp2
);
4847 gen_helper_neon_acgt_f32(tmp
, tmp
, tmp2
);
4849 case NEON_3R_FLOAT_MINMAX
:
4851 gen_helper_neon_max_f32(tmp
, tmp
, tmp2
);
4853 gen_helper_neon_min_f32(tmp
, tmp
, tmp2
);
4855 case NEON_3R_VRECPS_VRSQRTS
:
4857 gen_helper_recps_f32(tmp
, tmp
, tmp2
, cpu_env
);
4859 gen_helper_rsqrts_f32(tmp
, tmp
, tmp2
, cpu_env
);
4864 tcg_temp_free_i32(tmp2
);
4866 /* Save the result. For elementwise operations we can put it
4867 straight into the destination register. For pairwise operations
4868 we have to be careful to avoid clobbering the source operands. */
4869 if (pairwise
&& rd
== rm
) {
4870 neon_store_scratch(pass
, tmp
);
4872 neon_store_reg(rd
, pass
, tmp
);
4876 if (pairwise
&& rd
== rm
) {
4877 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
4878 tmp
= neon_load_scratch(pass
);
4879 neon_store_reg(rd
, pass
, tmp
);
4882 /* End of 3 register same size operations. */
4883 } else if (insn
& (1 << 4)) {
4884 if ((insn
& 0x00380080) != 0) {
4885 /* Two registers and shift. */
4886 op
= (insn
>> 8) & 0xf;
4887 if (insn
& (1 << 7)) {
4895 while ((insn
& (1 << (size
+ 19))) == 0)
4898 shift
= (insn
>> 16) & ((1 << (3 + size
)) - 1);
4899 /* To avoid excessive dumplication of ops we implement shift
4900 by immediate using the variable shift operations. */
4902 /* Shift by immediate:
4903 VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
4904 if (q
&& ((rd
| rm
) & 1)) {
4907 if (!u
&& (op
== 4 || op
== 6)) {
4910 /* Right shifts are encoded as N - shift, where N is the
4911 element size in bits. */
4913 shift
= shift
- (1 << (size
+ 3));
4921 imm
= (uint8_t) shift
;
4926 imm
= (uint16_t) shift
;
4937 for (pass
= 0; pass
< count
; pass
++) {
4939 neon_load_reg64(cpu_V0
, rm
+ pass
);
4940 tcg_gen_movi_i64(cpu_V1
, imm
);
4945 gen_helper_neon_shl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4947 gen_helper_neon_shl_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4952 gen_helper_neon_rshl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4954 gen_helper_neon_rshl_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4957 case 5: /* VSHL, VSLI */
4958 gen_helper_neon_shl_u64(cpu_V0
, cpu_V0
, cpu_V1
);
4960 case 6: /* VQSHLU */
4961 gen_helper_neon_qshlu_s64(cpu_V0
, cpu_V0
, cpu_V1
);
4965 gen_helper_neon_qshl_u64(cpu_V0
,
4968 gen_helper_neon_qshl_s64(cpu_V0
,
4973 if (op
== 1 || op
== 3) {
4975 neon_load_reg64(cpu_V1
, rd
+ pass
);
4976 tcg_gen_add_i64(cpu_V0
, cpu_V0
, cpu_V1
);
4977 } else if (op
== 4 || (op
== 5 && u
)) {
4979 neon_load_reg64(cpu_V1
, rd
+ pass
);
4981 if (shift
< -63 || shift
> 63) {
4985 mask
= 0xffffffffffffffffull
>> -shift
;
4987 mask
= 0xffffffffffffffffull
<< shift
;
4990 tcg_gen_andi_i64(cpu_V1
, cpu_V1
, ~mask
);
4991 tcg_gen_or_i64(cpu_V0
, cpu_V0
, cpu_V1
);
4993 neon_store_reg64(cpu_V0
, rd
+ pass
);
4994 } else { /* size < 3 */
4995 /* Operands in T0 and T1. */
4996 tmp
= neon_load_reg(rm
, pass
);
4997 tmp2
= tcg_temp_new_i32();
4998 tcg_gen_movi_i32(tmp2
, imm
);
5002 GEN_NEON_INTEGER_OP(shl
);
5006 GEN_NEON_INTEGER_OP(rshl
);
5009 case 5: /* VSHL, VSLI */
5011 case 0: gen_helper_neon_shl_u8(tmp
, tmp
, tmp2
); break;
5012 case 1: gen_helper_neon_shl_u16(tmp
, tmp
, tmp2
); break;
5013 case 2: gen_helper_neon_shl_u32(tmp
, tmp
, tmp2
); break;
5017 case 6: /* VQSHLU */
5020 gen_helper_neon_qshlu_s8(tmp
, tmp
, tmp2
);
5023 gen_helper_neon_qshlu_s16(tmp
, tmp
, tmp2
);
5026 gen_helper_neon_qshlu_s32(tmp
, tmp
, tmp2
);
5033 GEN_NEON_INTEGER_OP(qshl
);
5036 tcg_temp_free_i32(tmp2
);
5038 if (op
== 1 || op
== 3) {
5040 tmp2
= neon_load_reg(rd
, pass
);
5041 gen_neon_add(size
, tmp
, tmp2
);
5042 tcg_temp_free_i32(tmp2
);
5043 } else if (op
== 4 || (op
== 5 && u
)) {
5048 mask
= 0xff >> -shift
;
5050 mask
= (uint8_t)(0xff << shift
);
5056 mask
= 0xffff >> -shift
;
5058 mask
= (uint16_t)(0xffff << shift
);
5062 if (shift
< -31 || shift
> 31) {
5066 mask
= 0xffffffffu
>> -shift
;
5068 mask
= 0xffffffffu
<< shift
;
5074 tmp2
= neon_load_reg(rd
, pass
);
5075 tcg_gen_andi_i32(tmp
, tmp
, mask
);
5076 tcg_gen_andi_i32(tmp2
, tmp2
, ~mask
);
5077 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
5078 tcg_temp_free_i32(tmp2
);
5080 neon_store_reg(rd
, pass
, tmp
);
5083 } else if (op
< 10) {
5084 /* Shift by immediate and narrow:
5085 VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
5086 int input_unsigned
= (op
== 8) ? !u
: u
;
5090 shift
= shift
- (1 << (size
+ 3));
5093 tmp64
= tcg_const_i64(shift
);
5094 neon_load_reg64(cpu_V0
, rm
);
5095 neon_load_reg64(cpu_V1
, rm
+ 1);
5096 for (pass
= 0; pass
< 2; pass
++) {
5104 if (input_unsigned
) {
5105 gen_helper_neon_rshl_u64(cpu_V0
, in
, tmp64
);
5107 gen_helper_neon_rshl_s64(cpu_V0
, in
, tmp64
);
5110 if (input_unsigned
) {
5111 gen_helper_neon_shl_u64(cpu_V0
, in
, tmp64
);
5113 gen_helper_neon_shl_s64(cpu_V0
, in
, tmp64
);
5116 tmp
= tcg_temp_new_i32();
5117 gen_neon_narrow_op(op
== 8, u
, size
- 1, tmp
, cpu_V0
);
5118 neon_store_reg(rd
, pass
, tmp
);
5120 tcg_temp_free_i64(tmp64
);
5123 imm
= (uint16_t)shift
;
5127 imm
= (uint32_t)shift
;
5129 tmp2
= tcg_const_i32(imm
);
5130 tmp4
= neon_load_reg(rm
+ 1, 0);
5131 tmp5
= neon_load_reg(rm
+ 1, 1);
5132 for (pass
= 0; pass
< 2; pass
++) {
5134 tmp
= neon_load_reg(rm
, 0);
5138 gen_neon_shift_narrow(size
, tmp
, tmp2
, q
,
5141 tmp3
= neon_load_reg(rm
, 1);
5145 gen_neon_shift_narrow(size
, tmp3
, tmp2
, q
,
5147 tcg_gen_concat_i32_i64(cpu_V0
, tmp
, tmp3
);
5148 tcg_temp_free_i32(tmp
);
5149 tcg_temp_free_i32(tmp3
);
5150 tmp
= tcg_temp_new_i32();
5151 gen_neon_narrow_op(op
== 8, u
, size
- 1, tmp
, cpu_V0
);
5152 neon_store_reg(rd
, pass
, tmp
);
5154 tcg_temp_free_i32(tmp2
);
5156 } else if (op
== 10) {
5158 if (q
|| (rd
& 1)) {
5161 tmp
= neon_load_reg(rm
, 0);
5162 tmp2
= neon_load_reg(rm
, 1);
5163 for (pass
= 0; pass
< 2; pass
++) {
5167 gen_neon_widen(cpu_V0
, tmp
, size
, u
);
5170 /* The shift is less than the width of the source
5171 type, so we can just shift the whole register. */
5172 tcg_gen_shli_i64(cpu_V0
, cpu_V0
, shift
);
5173 /* Widen the result of shift: we need to clear
5174 * the potential overflow bits resulting from
5175 * left bits of the narrow input appearing as
5176 * right bits of left the neighbour narrow
5178 if (size
< 2 || !u
) {
5181 imm
= (0xffu
>> (8 - shift
));
5183 } else if (size
== 1) {
5184 imm
= 0xffff >> (16 - shift
);
5187 imm
= 0xffffffff >> (32 - shift
);
5190 imm64
= imm
| (((uint64_t)imm
) << 32);
5194 tcg_gen_andi_i64(cpu_V0
, cpu_V0
, ~imm64
);
5197 neon_store_reg64(cpu_V0
, rd
+ pass
);
5199 } else if (op
>= 14) {
5200 /* VCVT fixed-point. */
5201 if (!(insn
& (1 << 21)) || (q
&& ((rd
| rm
) & 1))) {
5204 /* We have already masked out the must-be-1 top bit of imm6,
5205 * hence this 32-shift where the ARM ARM has 64-imm6.
5208 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
5209 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, pass
));
5212 gen_vfp_ulto(0, shift
);
5214 gen_vfp_slto(0, shift
);
5217 gen_vfp_toul(0, shift
);
5219 gen_vfp_tosl(0, shift
);
5221 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, pass
));
5226 } else { /* (insn & 0x00380080) == 0 */
5228 if (q
&& (rd
& 1)) {
5232 op
= (insn
>> 8) & 0xf;
5233 /* One register and immediate. */
5234 imm
= (u
<< 7) | ((insn
>> 12) & 0x70) | (insn
& 0xf);
5235 invert
= (insn
& (1 << 5)) != 0;
5236 /* Note that op = 2,3,4,5,6,7,10,11,12,13 imm=0 is UNPREDICTABLE.
5237 * We choose to not special-case this and will behave as if a
5238 * valid constant encoding of 0 had been given.
5257 imm
= (imm
<< 8) | (imm
<< 24);
5260 imm
= (imm
<< 8) | 0xff;
5263 imm
= (imm
<< 16) | 0xffff;
5266 imm
|= (imm
<< 8) | (imm
<< 16) | (imm
<< 24);
5274 imm
= ((imm
& 0x80) << 24) | ((imm
& 0x3f) << 19)
5275 | ((imm
& 0x40) ? (0x1f << 25) : (1 << 30));
5281 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
5282 if (op
& 1 && op
< 12) {
5283 tmp
= neon_load_reg(rd
, pass
);
5285 /* The immediate value has already been inverted, so
5287 tcg_gen_andi_i32(tmp
, tmp
, imm
);
5289 tcg_gen_ori_i32(tmp
, tmp
, imm
);
5293 tmp
= tcg_temp_new_i32();
5294 if (op
== 14 && invert
) {
5298 for (n
= 0; n
< 4; n
++) {
5299 if (imm
& (1 << (n
+ (pass
& 1) * 4)))
5300 val
|= 0xff << (n
* 8);
5302 tcg_gen_movi_i32(tmp
, val
);
5304 tcg_gen_movi_i32(tmp
, imm
);
5307 neon_store_reg(rd
, pass
, tmp
);
5310 } else { /* (insn & 0x00800010 == 0x00800000) */
5312 op
= (insn
>> 8) & 0xf;
5313 if ((insn
& (1 << 6)) == 0) {
5314 /* Three registers of different lengths. */
5318 /* undefreq: bit 0 : UNDEF if size != 0
5319 * bit 1 : UNDEF if size == 0
5320 * bit 2 : UNDEF if U == 1
5321 * Note that [1:0] set implies 'always UNDEF'
5324 /* prewiden, src1_wide, src2_wide, undefreq */
5325 static const int neon_3reg_wide
[16][4] = {
5326 {1, 0, 0, 0}, /* VADDL */
5327 {1, 1, 0, 0}, /* VADDW */
5328 {1, 0, 0, 0}, /* VSUBL */
5329 {1, 1, 0, 0}, /* VSUBW */
5330 {0, 1, 1, 0}, /* VADDHN */
5331 {0, 0, 0, 0}, /* VABAL */
5332 {0, 1, 1, 0}, /* VSUBHN */
5333 {0, 0, 0, 0}, /* VABDL */
5334 {0, 0, 0, 0}, /* VMLAL */
5335 {0, 0, 0, 6}, /* VQDMLAL */
5336 {0, 0, 0, 0}, /* VMLSL */
5337 {0, 0, 0, 6}, /* VQDMLSL */
5338 {0, 0, 0, 0}, /* Integer VMULL */
5339 {0, 0, 0, 2}, /* VQDMULL */
5340 {0, 0, 0, 5}, /* Polynomial VMULL */
5341 {0, 0, 0, 3}, /* Reserved: always UNDEF */
5344 prewiden
= neon_3reg_wide
[op
][0];
5345 src1_wide
= neon_3reg_wide
[op
][1];
5346 src2_wide
= neon_3reg_wide
[op
][2];
5347 undefreq
= neon_3reg_wide
[op
][3];
5349 if (((undefreq
& 1) && (size
!= 0)) ||
5350 ((undefreq
& 2) && (size
== 0)) ||
5351 ((undefreq
& 4) && u
)) {
5354 if ((src1_wide
&& (rn
& 1)) ||
5355 (src2_wide
&& (rm
& 1)) ||
5356 (!src2_wide
&& (rd
& 1))) {
5360 /* Avoid overlapping operands. Wide source operands are
5361 always aligned so will never overlap with wide
5362 destinations in problematic ways. */
5363 if (rd
== rm
&& !src2_wide
) {
5364 tmp
= neon_load_reg(rm
, 1);
5365 neon_store_scratch(2, tmp
);
5366 } else if (rd
== rn
&& !src1_wide
) {
5367 tmp
= neon_load_reg(rn
, 1);
5368 neon_store_scratch(2, tmp
);
5371 for (pass
= 0; pass
< 2; pass
++) {
5373 neon_load_reg64(cpu_V0
, rn
+ pass
);
5376 if (pass
== 1 && rd
== rn
) {
5377 tmp
= neon_load_scratch(2);
5379 tmp
= neon_load_reg(rn
, pass
);
5382 gen_neon_widen(cpu_V0
, tmp
, size
, u
);
5386 neon_load_reg64(cpu_V1
, rm
+ pass
);
5389 if (pass
== 1 && rd
== rm
) {
5390 tmp2
= neon_load_scratch(2);
5392 tmp2
= neon_load_reg(rm
, pass
);
5395 gen_neon_widen(cpu_V1
, tmp2
, size
, u
);
5399 case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
5400 gen_neon_addl(size
);
5402 case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */
5403 gen_neon_subl(size
);
5405 case 5: case 7: /* VABAL, VABDL */
5406 switch ((size
<< 1) | u
) {
5408 gen_helper_neon_abdl_s16(cpu_V0
, tmp
, tmp2
);
5411 gen_helper_neon_abdl_u16(cpu_V0
, tmp
, tmp2
);
5414 gen_helper_neon_abdl_s32(cpu_V0
, tmp
, tmp2
);
5417 gen_helper_neon_abdl_u32(cpu_V0
, tmp
, tmp2
);
5420 gen_helper_neon_abdl_s64(cpu_V0
, tmp
, tmp2
);
5423 gen_helper_neon_abdl_u64(cpu_V0
, tmp
, tmp2
);
5427 tcg_temp_free_i32(tmp2
);
5428 tcg_temp_free_i32(tmp
);
5430 case 8: case 9: case 10: case 11: case 12: case 13:
5431 /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
5432 gen_neon_mull(cpu_V0
, tmp
, tmp2
, size
, u
);
5434 case 14: /* Polynomial VMULL */
5435 gen_helper_neon_mull_p8(cpu_V0
, tmp
, tmp2
);
5436 tcg_temp_free_i32(tmp2
);
5437 tcg_temp_free_i32(tmp
);
5439 default: /* 15 is RESERVED: caught earlier */
5444 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5445 neon_store_reg64(cpu_V0
, rd
+ pass
);
5446 } else if (op
== 5 || (op
>= 8 && op
<= 11)) {
5448 neon_load_reg64(cpu_V1
, rd
+ pass
);
5450 case 10: /* VMLSL */
5451 gen_neon_negl(cpu_V0
, size
);
5453 case 5: case 8: /* VABAL, VMLAL */
5454 gen_neon_addl(size
);
5456 case 9: case 11: /* VQDMLAL, VQDMLSL */
5457 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5459 gen_neon_negl(cpu_V0
, size
);
5461 gen_neon_addl_saturate(cpu_V0
, cpu_V1
, size
);
5466 neon_store_reg64(cpu_V0
, rd
+ pass
);
5467 } else if (op
== 4 || op
== 6) {
5468 /* Narrowing operation. */
5469 tmp
= tcg_temp_new_i32();
5473 gen_helper_neon_narrow_high_u8(tmp
, cpu_V0
);
5476 gen_helper_neon_narrow_high_u16(tmp
, cpu_V0
);
5479 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
5480 tcg_gen_trunc_i64_i32(tmp
, cpu_V0
);
5487 gen_helper_neon_narrow_round_high_u8(tmp
, cpu_V0
);
5490 gen_helper_neon_narrow_round_high_u16(tmp
, cpu_V0
);
5493 tcg_gen_addi_i64(cpu_V0
, cpu_V0
, 1u << 31);
5494 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, 32);
5495 tcg_gen_trunc_i64_i32(tmp
, cpu_V0
);
5503 neon_store_reg(rd
, 0, tmp3
);
5504 neon_store_reg(rd
, 1, tmp
);
5507 /* Write back the result. */
5508 neon_store_reg64(cpu_V0
, rd
+ pass
);
5512 /* Two registers and a scalar. NB that for ops of this form
5513 * the ARM ARM labels bit 24 as Q, but it is in our variable
5520 case 1: /* Float VMLA scalar */
5521 case 5: /* Floating point VMLS scalar */
5522 case 9: /* Floating point VMUL scalar */
5527 case 0: /* Integer VMLA scalar */
5528 case 4: /* Integer VMLS scalar */
5529 case 8: /* Integer VMUL scalar */
5530 case 12: /* VQDMULH scalar */
5531 case 13: /* VQRDMULH scalar */
5532 if (u
&& ((rd
| rn
) & 1)) {
5535 tmp
= neon_get_scalar(size
, rm
);
5536 neon_store_scratch(0, tmp
);
5537 for (pass
= 0; pass
< (u
? 4 : 2); pass
++) {
5538 tmp
= neon_load_scratch(0);
5539 tmp2
= neon_load_reg(rn
, pass
);
5542 gen_helper_neon_qdmulh_s16(tmp
, tmp
, tmp2
);
5544 gen_helper_neon_qdmulh_s32(tmp
, tmp
, tmp2
);
5546 } else if (op
== 13) {
5548 gen_helper_neon_qrdmulh_s16(tmp
, tmp
, tmp2
);
5550 gen_helper_neon_qrdmulh_s32(tmp
, tmp
, tmp2
);
5552 } else if (op
& 1) {
5553 gen_helper_neon_mul_f32(tmp
, tmp
, tmp2
);
5556 case 0: gen_helper_neon_mul_u8(tmp
, tmp
, tmp2
); break;
5557 case 1: gen_helper_neon_mul_u16(tmp
, tmp
, tmp2
); break;
5558 case 2: tcg_gen_mul_i32(tmp
, tmp
, tmp2
); break;
5562 tcg_temp_free_i32(tmp2
);
5565 tmp2
= neon_load_reg(rd
, pass
);
5568 gen_neon_add(size
, tmp
, tmp2
);
5571 gen_helper_neon_add_f32(tmp
, tmp
, tmp2
);
5574 gen_neon_rsb(size
, tmp
, tmp2
);
5577 gen_helper_neon_sub_f32(tmp
, tmp2
, tmp
);
5582 tcg_temp_free_i32(tmp2
);
5584 neon_store_reg(rd
, pass
, tmp
);
5587 case 3: /* VQDMLAL scalar */
5588 case 7: /* VQDMLSL scalar */
5589 case 11: /* VQDMULL scalar */
5594 case 2: /* VMLAL sclar */
5595 case 6: /* VMLSL scalar */
5596 case 10: /* VMULL scalar */
5600 tmp2
= neon_get_scalar(size
, rm
);
5601 /* We need a copy of tmp2 because gen_neon_mull
5602 * deletes it during pass 0. */
5603 tmp4
= tcg_temp_new_i32();
5604 tcg_gen_mov_i32(tmp4
, tmp2
);
5605 tmp3
= neon_load_reg(rn
, 1);
5607 for (pass
= 0; pass
< 2; pass
++) {
5609 tmp
= neon_load_reg(rn
, 0);
5614 gen_neon_mull(cpu_V0
, tmp
, tmp2
, size
, u
);
5616 neon_load_reg64(cpu_V1
, rd
+ pass
);
5620 gen_neon_negl(cpu_V0
, size
);
5623 gen_neon_addl(size
);
5626 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5628 gen_neon_negl(cpu_V0
, size
);
5630 gen_neon_addl_saturate(cpu_V0
, cpu_V1
, size
);
5636 gen_neon_addl_saturate(cpu_V0
, cpu_V0
, size
);
5641 neon_store_reg64(cpu_V0
, rd
+ pass
);
5646 default: /* 14 and 15 are RESERVED */
5650 } else { /* size == 3 */
5653 imm
= (insn
>> 8) & 0xf;
5658 if (q
&& ((rd
| rn
| rm
) & 1)) {
5663 neon_load_reg64(cpu_V0
, rn
);
5665 neon_load_reg64(cpu_V1
, rn
+ 1);
5667 } else if (imm
== 8) {
5668 neon_load_reg64(cpu_V0
, rn
+ 1);
5670 neon_load_reg64(cpu_V1
, rm
);
5673 tmp64
= tcg_temp_new_i64();
5675 neon_load_reg64(cpu_V0
, rn
);
5676 neon_load_reg64(tmp64
, rn
+ 1);
5678 neon_load_reg64(cpu_V0
, rn
+ 1);
5679 neon_load_reg64(tmp64
, rm
);
5681 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, (imm
& 7) * 8);
5682 tcg_gen_shli_i64(cpu_V1
, tmp64
, 64 - ((imm
& 7) * 8));
5683 tcg_gen_or_i64(cpu_V0
, cpu_V0
, cpu_V1
);
5685 neon_load_reg64(cpu_V1
, rm
);
5687 neon_load_reg64(cpu_V1
, rm
+ 1);
5690 tcg_gen_shli_i64(cpu_V1
, cpu_V1
, 64 - (imm
* 8));
5691 tcg_gen_shri_i64(tmp64
, tmp64
, imm
* 8);
5692 tcg_gen_or_i64(cpu_V1
, cpu_V1
, tmp64
);
5693 tcg_temp_free_i64(tmp64
);
5696 neon_load_reg64(cpu_V0
, rn
);
5697 tcg_gen_shri_i64(cpu_V0
, cpu_V0
, imm
* 8);
5698 neon_load_reg64(cpu_V1
, rm
);
5699 tcg_gen_shli_i64(cpu_V1
, cpu_V1
, 64 - (imm
* 8));
5700 tcg_gen_or_i64(cpu_V0
, cpu_V0
, cpu_V1
);
5702 neon_store_reg64(cpu_V0
, rd
);
5704 neon_store_reg64(cpu_V1
, rd
+ 1);
5706 } else if ((insn
& (1 << 11)) == 0) {
5707 /* Two register misc. */
5708 op
= ((insn
>> 12) & 0x30) | ((insn
>> 7) & 0xf);
5709 size
= (insn
>> 18) & 3;
5710 /* UNDEF for unknown op values and bad op-size combinations */
5711 if ((neon_2rm_sizes
[op
] & (1 << size
)) == 0) {
5714 if ((op
!= NEON_2RM_VMOVN
&& op
!= NEON_2RM_VQMOVN
) &&
5715 q
&& ((rm
| rd
) & 1)) {
5719 case NEON_2RM_VREV64
:
5720 for (pass
= 0; pass
< (q
? 2 : 1); pass
++) {
5721 tmp
= neon_load_reg(rm
, pass
* 2);
5722 tmp2
= neon_load_reg(rm
, pass
* 2 + 1);
5724 case 0: tcg_gen_bswap32_i32(tmp
, tmp
); break;
5725 case 1: gen_swap_half(tmp
); break;
5726 case 2: /* no-op */ break;
5729 neon_store_reg(rd
, pass
* 2 + 1, tmp
);
5731 neon_store_reg(rd
, pass
* 2, tmp2
);
5734 case 0: tcg_gen_bswap32_i32(tmp2
, tmp2
); break;
5735 case 1: gen_swap_half(tmp2
); break;
5738 neon_store_reg(rd
, pass
* 2, tmp2
);
5742 case NEON_2RM_VPADDL
: case NEON_2RM_VPADDL_U
:
5743 case NEON_2RM_VPADAL
: case NEON_2RM_VPADAL_U
:
5744 for (pass
= 0; pass
< q
+ 1; pass
++) {
5745 tmp
= neon_load_reg(rm
, pass
* 2);
5746 gen_neon_widen(cpu_V0
, tmp
, size
, op
& 1);
5747 tmp
= neon_load_reg(rm
, pass
* 2 + 1);
5748 gen_neon_widen(cpu_V1
, tmp
, size
, op
& 1);
5750 case 0: gen_helper_neon_paddl_u16(CPU_V001
); break;
5751 case 1: gen_helper_neon_paddl_u32(CPU_V001
); break;
5752 case 2: tcg_gen_add_i64(CPU_V001
); break;
5755 if (op
>= NEON_2RM_VPADAL
) {
5757 neon_load_reg64(cpu_V1
, rd
+ pass
);
5758 gen_neon_addl(size
);
5760 neon_store_reg64(cpu_V0
, rd
+ pass
);
5766 for (n
= 0; n
< (q
? 4 : 2); n
+= 2) {
5767 tmp
= neon_load_reg(rm
, n
);
5768 tmp2
= neon_load_reg(rd
, n
+ 1);
5769 neon_store_reg(rm
, n
, tmp2
);
5770 neon_store_reg(rd
, n
+ 1, tmp
);
5777 if (gen_neon_unzip(rd
, rm
, size
, q
)) {
5782 if (gen_neon_zip(rd
, rm
, size
, q
)) {
5786 case NEON_2RM_VMOVN
: case NEON_2RM_VQMOVN
:
5787 /* also VQMOVUN; op field and mnemonics don't line up */
5792 for (pass
= 0; pass
< 2; pass
++) {
5793 neon_load_reg64(cpu_V0
, rm
+ pass
);
5794 tmp
= tcg_temp_new_i32();
5795 gen_neon_narrow_op(op
== NEON_2RM_VMOVN
, q
, size
,
5800 neon_store_reg(rd
, 0, tmp2
);
5801 neon_store_reg(rd
, 1, tmp
);
5805 case NEON_2RM_VSHLL
:
5806 if (q
|| (rd
& 1)) {
5809 tmp
= neon_load_reg(rm
, 0);
5810 tmp2
= neon_load_reg(rm
, 1);
5811 for (pass
= 0; pass
< 2; pass
++) {
5814 gen_neon_widen(cpu_V0
, tmp
, size
, 1);
5815 tcg_gen_shli_i64(cpu_V0
, cpu_V0
, 8 << size
);
5816 neon_store_reg64(cpu_V0
, rd
+ pass
);
5819 case NEON_2RM_VCVT_F16_F32
:
5820 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
) ||
5824 tmp
= tcg_temp_new_i32();
5825 tmp2
= tcg_temp_new_i32();
5826 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 0));
5827 gen_helper_neon_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
5828 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 1));
5829 gen_helper_neon_fcvt_f32_to_f16(tmp2
, cpu_F0s
, cpu_env
);
5830 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
5831 tcg_gen_or_i32(tmp2
, tmp2
, tmp
);
5832 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 2));
5833 gen_helper_neon_fcvt_f32_to_f16(tmp
, cpu_F0s
, cpu_env
);
5834 tcg_gen_ld_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rm
, 3));
5835 neon_store_reg(rd
, 0, tmp2
);
5836 tmp2
= tcg_temp_new_i32();
5837 gen_helper_neon_fcvt_f32_to_f16(tmp2
, cpu_F0s
, cpu_env
);
5838 tcg_gen_shli_i32(tmp2
, tmp2
, 16);
5839 tcg_gen_or_i32(tmp2
, tmp2
, tmp
);
5840 neon_store_reg(rd
, 1, tmp2
);
5841 tcg_temp_free_i32(tmp
);
5843 case NEON_2RM_VCVT_F32_F16
:
5844 if (!arm_feature(env
, ARM_FEATURE_VFP_FP16
) ||
5848 tmp3
= tcg_temp_new_i32();
5849 tmp
= neon_load_reg(rm
, 0);
5850 tmp2
= neon_load_reg(rm
, 1);
5851 tcg_gen_ext16u_i32(tmp3
, tmp
);
5852 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5853 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 0));
5854 tcg_gen_shri_i32(tmp3
, tmp
, 16);
5855 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5856 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 1));
5857 tcg_temp_free_i32(tmp
);
5858 tcg_gen_ext16u_i32(tmp3
, tmp2
);
5859 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5860 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 2));
5861 tcg_gen_shri_i32(tmp3
, tmp2
, 16);
5862 gen_helper_neon_fcvt_f16_to_f32(cpu_F0s
, tmp3
, cpu_env
);
5863 tcg_gen_st_f32(cpu_F0s
, cpu_env
, neon_reg_offset(rd
, 3));
5864 tcg_temp_free_i32(tmp2
);
5865 tcg_temp_free_i32(tmp3
);
5869 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
5870 if (neon_2rm_is_float_op(op
)) {
5871 tcg_gen_ld_f32(cpu_F0s
, cpu_env
,
5872 neon_reg_offset(rm
, pass
));
5875 tmp
= neon_load_reg(rm
, pass
);
5878 case NEON_2RM_VREV32
:
5880 case 0: tcg_gen_bswap32_i32(tmp
, tmp
); break;
5881 case 1: gen_swap_half(tmp
); break;
5885 case NEON_2RM_VREV16
:
5890 case 0: gen_helper_neon_cls_s8(tmp
, tmp
); break;
5891 case 1: gen_helper_neon_cls_s16(tmp
, tmp
); break;
5892 case 2: gen_helper_neon_cls_s32(tmp
, tmp
); break;
5898 case 0: gen_helper_neon_clz_u8(tmp
, tmp
); break;
5899 case 1: gen_helper_neon_clz_u16(tmp
, tmp
); break;
5900 case 2: gen_helper_clz(tmp
, tmp
); break;
5905 gen_helper_neon_cnt_u8(tmp
, tmp
);
5908 tcg_gen_not_i32(tmp
, tmp
);
5910 case NEON_2RM_VQABS
:
5912 case 0: gen_helper_neon_qabs_s8(tmp
, tmp
); break;
5913 case 1: gen_helper_neon_qabs_s16(tmp
, tmp
); break;
5914 case 2: gen_helper_neon_qabs_s32(tmp
, tmp
); break;
5918 case NEON_2RM_VQNEG
:
5920 case 0: gen_helper_neon_qneg_s8(tmp
, tmp
); break;
5921 case 1: gen_helper_neon_qneg_s16(tmp
, tmp
); break;
5922 case 2: gen_helper_neon_qneg_s32(tmp
, tmp
); break;
5926 case NEON_2RM_VCGT0
: case NEON_2RM_VCLE0
:
5927 tmp2
= tcg_const_i32(0);
5929 case 0: gen_helper_neon_cgt_s8(tmp
, tmp
, tmp2
); break;
5930 case 1: gen_helper_neon_cgt_s16(tmp
, tmp
, tmp2
); break;
5931 case 2: gen_helper_neon_cgt_s32(tmp
, tmp
, tmp2
); break;
5934 tcg_temp_free(tmp2
);
5935 if (op
== NEON_2RM_VCLE0
) {
5936 tcg_gen_not_i32(tmp
, tmp
);
5939 case NEON_2RM_VCGE0
: case NEON_2RM_VCLT0
:
5940 tmp2
= tcg_const_i32(0);
5942 case 0: gen_helper_neon_cge_s8(tmp
, tmp
, tmp2
); break;
5943 case 1: gen_helper_neon_cge_s16(tmp
, tmp
, tmp2
); break;
5944 case 2: gen_helper_neon_cge_s32(tmp
, tmp
, tmp2
); break;
5947 tcg_temp_free(tmp2
);
5948 if (op
== NEON_2RM_VCLT0
) {
5949 tcg_gen_not_i32(tmp
, tmp
);
5952 case NEON_2RM_VCEQ0
:
5953 tmp2
= tcg_const_i32(0);
5955 case 0: gen_helper_neon_ceq_u8(tmp
, tmp
, tmp2
); break;
5956 case 1: gen_helper_neon_ceq_u16(tmp
, tmp
, tmp2
); break;
5957 case 2: gen_helper_neon_ceq_u32(tmp
, tmp
, tmp2
); break;
5960 tcg_temp_free(tmp2
);
5964 case 0: gen_helper_neon_abs_s8(tmp
, tmp
); break;
5965 case 1: gen_helper_neon_abs_s16(tmp
, tmp
); break;
5966 case 2: tcg_gen_abs_i32(tmp
, tmp
); break;
5971 tmp2
= tcg_const_i32(0);
5972 gen_neon_rsb(size
, tmp
, tmp2
);
5973 tcg_temp_free(tmp2
);
5975 case NEON_2RM_VCGT0_F
:
5976 tmp2
= tcg_const_i32(0);
5977 gen_helper_neon_cgt_f32(tmp
, tmp
, tmp2
);
5978 tcg_temp_free(tmp2
);
5980 case NEON_2RM_VCGE0_F
:
5981 tmp2
= tcg_const_i32(0);
5982 gen_helper_neon_cge_f32(tmp
, tmp
, tmp2
);
5983 tcg_temp_free(tmp2
);
5985 case NEON_2RM_VCEQ0_F
:
5986 tmp2
= tcg_const_i32(0);
5987 gen_helper_neon_ceq_f32(tmp
, tmp
, tmp2
);
5988 tcg_temp_free(tmp2
);
5990 case NEON_2RM_VCLE0_F
:
5991 tmp2
= tcg_const_i32(0);
5992 gen_helper_neon_cge_f32(tmp
, tmp2
, tmp
);
5993 tcg_temp_free(tmp2
);
5995 case NEON_2RM_VCLT0_F
:
5996 tmp2
= tcg_const_i32(0);
5997 gen_helper_neon_cgt_f32(tmp
, tmp2
, tmp
);
5998 tcg_temp_free(tmp2
);
6000 case NEON_2RM_VABS_F
:
6003 case NEON_2RM_VNEG_F
:
6007 tmp2
= neon_load_reg(rd
, pass
);
6008 neon_store_reg(rm
, pass
, tmp2
);
6011 tmp2
= neon_load_reg(rd
, pass
);
6013 case 0: gen_neon_trn_u8(tmp
, tmp2
); break;
6014 case 1: gen_neon_trn_u16(tmp
, tmp2
); break;
6017 neon_store_reg(rm
, pass
, tmp2
);
6019 case NEON_2RM_VRECPE
:
6020 gen_helper_recpe_u32(tmp
, tmp
, cpu_env
);
6022 case NEON_2RM_VRSQRTE
:
6023 gen_helper_rsqrte_u32(tmp
, tmp
, cpu_env
);
6025 case NEON_2RM_VRECPE_F
:
6026 gen_helper_recpe_f32(cpu_F0s
, cpu_F0s
, cpu_env
);
6028 case NEON_2RM_VRSQRTE_F
:
6029 gen_helper_rsqrte_f32(cpu_F0s
, cpu_F0s
, cpu_env
);
6031 case NEON_2RM_VCVT_FS
: /* VCVT.F32.S32 */
6034 case NEON_2RM_VCVT_FU
: /* VCVT.F32.U32 */
6037 case NEON_2RM_VCVT_SF
: /* VCVT.S32.F32 */
6040 case NEON_2RM_VCVT_UF
: /* VCVT.U32.F32 */
6044 /* Reserved op values were caught by the
6045 * neon_2rm_sizes[] check earlier.
6049 if (neon_2rm_is_float_op(op
)) {
6050 tcg_gen_st_f32(cpu_F0s
, cpu_env
,
6051 neon_reg_offset(rd
, pass
));
6053 neon_store_reg(rd
, pass
, tmp
);
6058 } else if ((insn
& (1 << 10)) == 0) {
6060 int n
= ((insn
>> 8) & 3) + 1;
6061 if ((rn
+ n
) > 32) {
6062 /* This is UNPREDICTABLE; we choose to UNDEF to avoid the
6063 * helper function running off the end of the register file.
6068 if (insn
& (1 << 6)) {
6069 tmp
= neon_load_reg(rd
, 0);
6071 tmp
= tcg_temp_new_i32();
6072 tcg_gen_movi_i32(tmp
, 0);
6074 tmp2
= neon_load_reg(rm
, 0);
6075 tmp4
= tcg_const_i32(rn
);
6076 tmp5
= tcg_const_i32(n
);
6077 gen_helper_neon_tbl(tmp2
, tmp2
, tmp
, tmp4
, tmp5
);
6078 tcg_temp_free_i32(tmp
);
6079 if (insn
& (1 << 6)) {
6080 tmp
= neon_load_reg(rd
, 1);
6082 tmp
= tcg_temp_new_i32();
6083 tcg_gen_movi_i32(tmp
, 0);
6085 tmp3
= neon_load_reg(rm
, 1);
6086 gen_helper_neon_tbl(tmp3
, tmp3
, tmp
, tmp4
, tmp5
);
6087 tcg_temp_free_i32(tmp5
);
6088 tcg_temp_free_i32(tmp4
);
6089 neon_store_reg(rd
, 0, tmp2
);
6090 neon_store_reg(rd
, 1, tmp3
);
6091 tcg_temp_free_i32(tmp
);
6092 } else if ((insn
& 0x380) == 0) {
6094 if ((insn
& (7 << 16)) == 0 || (q
&& (rd
& 1))) {
6097 if (insn
& (1 << 19)) {
6098 tmp
= neon_load_reg(rm
, 1);
6100 tmp
= neon_load_reg(rm
, 0);
6102 if (insn
& (1 << 16)) {
6103 gen_neon_dup_u8(tmp
, ((insn
>> 17) & 3) * 8);
6104 } else if (insn
& (1 << 17)) {
6105 if ((insn
>> 18) & 1)
6106 gen_neon_dup_high16(tmp
);
6108 gen_neon_dup_low16(tmp
);
6110 for (pass
= 0; pass
< (q
? 4 : 2); pass
++) {
6111 tmp2
= tcg_temp_new_i32();
6112 tcg_gen_mov_i32(tmp2
, tmp
);
6113 neon_store_reg(rd
, pass
, tmp2
);
6115 tcg_temp_free_i32(tmp
);
6124 static int disas_cp14_read(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
6126 int crn
= (insn
>> 16) & 0xf;
6127 int crm
= insn
& 0xf;
6128 int op1
= (insn
>> 21) & 7;
6129 int op2
= (insn
>> 5) & 7;
6130 int rt
= (insn
>> 12) & 0xf;
6133 /* Minimal set of debug registers, since we don't support debug */
6134 if (op1
== 0 && crn
== 0 && op2
== 0) {
6137 /* DBGDIDR: just RAZ. In particular this means the
6138 * "debug architecture version" bits will read as
6139 * a reserved value, which should cause Linux to
6140 * not try to use the debug hardware.
6142 tmp
= tcg_const_i32(0);
6143 store_reg(s
, rt
, tmp
);
6147 /* DBGDRAR and DBGDSAR: v7 only. Always RAZ since we
6148 * don't implement memory mapped debug components
6150 if (ENABLE_ARCH_7
) {
6151 tmp
= tcg_const_i32(0);
6152 store_reg(s
, rt
, tmp
);
6161 if (arm_feature(env
, ARM_FEATURE_THUMB2EE
)) {
6162 if (op1
== 6 && crn
== 0 && crm
== 0 && op2
== 0) {
6166 tmp
= load_cpu_field(teecr
);
6167 store_reg(s
, rt
, tmp
);
6170 if (op1
== 6 && crn
== 1 && crm
== 0 && op2
== 0) {
6172 if (IS_USER(s
) && (env
->teecr
& 1))
6174 tmp
= load_cpu_field(teehbr
);
6175 store_reg(s
, rt
, tmp
);
6179 fprintf(stderr
, "Unknown cp14 read op1:%d crn:%d crm:%d op2:%d\n",
6180 op1
, crn
, crm
, op2
);
6184 static int disas_cp14_write(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
6186 int crn
= (insn
>> 16) & 0xf;
6187 int crm
= insn
& 0xf;
6188 int op1
= (insn
>> 21) & 7;
6189 int op2
= (insn
>> 5) & 7;
6190 int rt
= (insn
>> 12) & 0xf;
6193 if (arm_feature(env
, ARM_FEATURE_THUMB2EE
)) {
6194 if (op1
== 6 && crn
== 0 && crm
== 0 && op2
== 0) {
6198 tmp
= load_reg(s
, rt
);
6199 gen_helper_set_teecr(cpu_env
, tmp
);
6200 tcg_temp_free_i32(tmp
);
6203 if (op1
== 6 && crn
== 1 && crm
== 0 && op2
== 0) {
6205 if (IS_USER(s
) && (env
->teecr
& 1))
6207 tmp
= load_reg(s
, rt
);
6208 store_cpu_field(tmp
, teehbr
);
6212 fprintf(stderr
, "Unknown cp14 write op1:%d crn:%d crm:%d op2:%d\n",
6213 op1
, crn
, crm
, op2
);
6217 static int disas_coproc_insn(CPUState
* env
, DisasContext
*s
, uint32_t insn
)
6221 cpnum
= (insn
>> 8) & 0xf;
6222 if (arm_feature(env
, ARM_FEATURE_XSCALE
)
6223 && ((env
->cp15
.c15_cpar
^ 0x3fff) & (1 << cpnum
)))
6229 if (arm_feature(env
, ARM_FEATURE_IWMMXT
)) {
6230 return disas_iwmmxt_insn(env
, s
, insn
);
6231 } else if (arm_feature(env
, ARM_FEATURE_XSCALE
)) {
6232 return disas_dsp_insn(env
, s
, insn
);
6237 return disas_vfp_insn (env
, s
, insn
);
6239 /* Coprocessors 7-15 are architecturally reserved by ARM.
6240 Unfortunately Intel decided to ignore this. */
6241 if (arm_feature(env
, ARM_FEATURE_XSCALE
))
6243 if (insn
& (1 << 20))
6244 return disas_cp14_read(env
, s
, insn
);
6246 return disas_cp14_write(env
, s
, insn
);
6248 return disas_cp15_insn (env
, s
, insn
);
6251 /* Unknown coprocessor. See if the board has hooked it. */
6252 return disas_cp_insn (env
, s
, insn
);
6257 /* Store a 64-bit value to a register pair. Clobbers val. */
6258 static void gen_storeq_reg(DisasContext
*s
, int rlow
, int rhigh
, TCGv_i64 val
)
6261 tmp
= tcg_temp_new_i32();
6262 tcg_gen_trunc_i64_i32(tmp
, val
);
6263 store_reg(s
, rlow
, tmp
);
6264 tmp
= tcg_temp_new_i32();
6265 tcg_gen_shri_i64(val
, val
, 32);
6266 tcg_gen_trunc_i64_i32(tmp
, val
);
6267 store_reg(s
, rhigh
, tmp
);
6270 /* load a 32-bit value from a register and perform a 64-bit accumulate. */
6271 static void gen_addq_lo(DisasContext
*s
, TCGv_i64 val
, int rlow
)
6276 /* Load value and extend to 64 bits. */
6277 tmp
= tcg_temp_new_i64();
6278 tmp2
= load_reg(s
, rlow
);
6279 tcg_gen_extu_i32_i64(tmp
, tmp2
);
6280 tcg_temp_free_i32(tmp2
);
6281 tcg_gen_add_i64(val
, val
, tmp
);
6282 tcg_temp_free_i64(tmp
);
6285 /* load and add a 64-bit value from a register pair. */
6286 static void gen_addq(DisasContext
*s
, TCGv_i64 val
, int rlow
, int rhigh
)
6292 /* Load 64-bit value rd:rn. */
6293 tmpl
= load_reg(s
, rlow
);
6294 tmph
= load_reg(s
, rhigh
);
6295 tmp
= tcg_temp_new_i64();
6296 tcg_gen_concat_i32_i64(tmp
, tmpl
, tmph
);
6297 tcg_temp_free_i32(tmpl
);
6298 tcg_temp_free_i32(tmph
);
6299 tcg_gen_add_i64(val
, val
, tmp
);
6300 tcg_temp_free_i64(tmp
);
6303 /* Set N and Z flags from a 64-bit value. */
6304 static void gen_logicq_cc(TCGv_i64 val
)
6306 TCGv tmp
= tcg_temp_new_i32();
6307 gen_helper_logicq_cc(tmp
, val
);
6309 tcg_temp_free_i32(tmp
);
6312 /* Load/Store exclusive instructions are implemented by remembering
6313 the value/address loaded, and seeing if these are the same
6314 when the store is performed. This should be is sufficient to implement
6315 the architecturally mandated semantics, and avoids having to monitor
6318 In system emulation mode only one CPU will be running at once, so
6319 this sequence is effectively atomic. In user emulation mode we
6320 throw an exception and handle the atomic operation elsewhere. */
6321 static void gen_load_exclusive(DisasContext
*s
, int rt
, int rt2
,
6322 TCGv addr
, int size
)
6328 tmp
= gen_ld8u(addr
, IS_USER(s
));
6331 tmp
= gen_ld16u(addr
, IS_USER(s
));
6335 tmp
= gen_ld32(addr
, IS_USER(s
));
6340 tcg_gen_mov_i32(cpu_exclusive_val
, tmp
);
6341 store_reg(s
, rt
, tmp
);
6343 TCGv tmp2
= tcg_temp_new_i32();
6344 tcg_gen_addi_i32(tmp2
, addr
, 4);
6345 tmp
= gen_ld32(tmp2
, IS_USER(s
));
6346 tcg_temp_free_i32(tmp2
);
6347 tcg_gen_mov_i32(cpu_exclusive_high
, tmp
);
6348 store_reg(s
, rt2
, tmp
);
6350 tcg_gen_mov_i32(cpu_exclusive_addr
, addr
);
6353 static void gen_clrex(DisasContext
*s
)
6355 tcg_gen_movi_i32(cpu_exclusive_addr
, -1);
6358 #ifdef CONFIG_USER_ONLY
6359 static void gen_store_exclusive(DisasContext
*s
, int rd
, int rt
, int rt2
,
6360 TCGv addr
, int size
)
6362 tcg_gen_mov_i32(cpu_exclusive_test
, addr
);
6363 tcg_gen_movi_i32(cpu_exclusive_info
,
6364 size
| (rd
<< 4) | (rt
<< 8) | (rt2
<< 12));
6365 gen_exception_insn(s
, 4, EXCP_STREX
);
6368 static void gen_store_exclusive(DisasContext
*s
, int rd
, int rt
, int rt2
,
6369 TCGv addr
, int size
)
6375 /* if (env->exclusive_addr == addr && env->exclusive_val == [addr]) {
6381 fail_label
= gen_new_label();
6382 done_label
= gen_new_label();
6383 tcg_gen_brcond_i32(TCG_COND_NE
, addr
, cpu_exclusive_addr
, fail_label
);
6386 tmp
= gen_ld8u(addr
, IS_USER(s
));
6389 tmp
= gen_ld16u(addr
, IS_USER(s
));
6393 tmp
= gen_ld32(addr
, IS_USER(s
));
6398 tcg_gen_brcond_i32(TCG_COND_NE
, tmp
, cpu_exclusive_val
, fail_label
);
6399 tcg_temp_free_i32(tmp
);
6401 TCGv tmp2
= tcg_temp_new_i32();
6402 tcg_gen_addi_i32(tmp2
, addr
, 4);
6403 tmp
= gen_ld32(tmp2
, IS_USER(s
));
6404 tcg_temp_free_i32(tmp2
);
6405 tcg_gen_brcond_i32(TCG_COND_NE
, tmp
, cpu_exclusive_high
, fail_label
);
6406 tcg_temp_free_i32(tmp
);
6408 tmp
= load_reg(s
, rt
);
6411 gen_st8(tmp
, addr
, IS_USER(s
));
6414 gen_st16(tmp
, addr
, IS_USER(s
));
6418 gen_st32(tmp
, addr
, IS_USER(s
));
6424 tcg_gen_addi_i32(addr
, addr
, 4);
6425 tmp
= load_reg(s
, rt2
);
6426 gen_st32(tmp
, addr
, IS_USER(s
));
6428 tcg_gen_movi_i32(cpu_R
[rd
], 0);
6429 tcg_gen_br(done_label
);
6430 gen_set_label(fail_label
);
6431 tcg_gen_movi_i32(cpu_R
[rd
], 1);
6432 gen_set_label(done_label
);
6433 tcg_gen_movi_i32(cpu_exclusive_addr
, -1);
6437 static void disas_arm_insn(CPUState
* env
, DisasContext
*s
)
6439 unsigned int cond
, insn
, val
, op1
, i
, shift
, rm
, rs
, rn
, rd
, sh
;
6446 insn
= ldl_code(s
->pc
);
6449 /* M variants do not implement ARM mode. */
6454 /* In ARMv3 and v4 the NV condition is UNPREDICTABLE; we
6455 * choose to UNDEF. In ARMv5 and above the space is used
6456 * for miscellaneous unconditional instructions.
6460 /* Unconditional instructions. */
6461 if (((insn
>> 25) & 7) == 1) {
6462 /* NEON Data processing. */
6463 if (!arm_feature(env
, ARM_FEATURE_NEON
))
6466 if (disas_neon_data_insn(env
, s
, insn
))
6470 if ((insn
& 0x0f100000) == 0x04000000) {
6471 /* NEON load/store. */
6472 if (!arm_feature(env
, ARM_FEATURE_NEON
))
6475 if (disas_neon_ls_insn(env
, s
, insn
))
6479 if (((insn
& 0x0f30f000) == 0x0510f000) ||
6480 ((insn
& 0x0f30f010) == 0x0710f000)) {
6481 if ((insn
& (1 << 22)) == 0) {
6483 if (!arm_feature(env
, ARM_FEATURE_V7MP
)) {
6487 /* Otherwise PLD; v5TE+ */
6491 if (((insn
& 0x0f70f000) == 0x0450f000) ||
6492 ((insn
& 0x0f70f010) == 0x0650f000)) {
6494 return; /* PLI; V7 */
6496 if (((insn
& 0x0f700000) == 0x04100000) ||
6497 ((insn
& 0x0f700010) == 0x06100000)) {
6498 if (!arm_feature(env
, ARM_FEATURE_V7MP
)) {
6501 return; /* v7MP: Unallocated memory hint: must NOP */
6504 if ((insn
& 0x0ffffdff) == 0x01010000) {
6507 if (insn
& (1 << 9)) {
6508 /* BE8 mode not implemented. */
6512 } else if ((insn
& 0x0fffff00) == 0x057ff000) {
6513 switch ((insn
>> 4) & 0xf) {
6522 /* We don't emulate caches so these are a no-op. */
6527 } else if ((insn
& 0x0e5fffe0) == 0x084d0500) {
6533 op1
= (insn
& 0x1f);
6534 addr
= tcg_temp_new_i32();
6535 tmp
= tcg_const_i32(op1
);
6536 gen_helper_get_r13_banked(addr
, cpu_env
, tmp
);
6537 tcg_temp_free_i32(tmp
);
6538 i
= (insn
>> 23) & 3;
6540 case 0: offset
= -4; break; /* DA */
6541 case 1: offset
= 0; break; /* IA */
6542 case 2: offset
= -8; break; /* DB */
6543 case 3: offset
= 4; break; /* IB */
6547 tcg_gen_addi_i32(addr
, addr
, offset
);
6548 tmp
= load_reg(s
, 14);
6549 gen_st32(tmp
, addr
, 0);
6550 tmp
= load_cpu_field(spsr
);
6551 tcg_gen_addi_i32(addr
, addr
, 4);
6552 gen_st32(tmp
, addr
, 0);
6553 if (insn
& (1 << 21)) {
6554 /* Base writeback. */
6556 case 0: offset
= -8; break;
6557 case 1: offset
= 4; break;
6558 case 2: offset
= -4; break;
6559 case 3: offset
= 0; break;
6563 tcg_gen_addi_i32(addr
, addr
, offset
);
6564 tmp
= tcg_const_i32(op1
);
6565 gen_helper_set_r13_banked(cpu_env
, tmp
, addr
);
6566 tcg_temp_free_i32(tmp
);
6567 tcg_temp_free_i32(addr
);
6569 tcg_temp_free_i32(addr
);
6572 } else if ((insn
& 0x0e50ffe0) == 0x08100a00) {
6578 rn
= (insn
>> 16) & 0xf;
6579 addr
= load_reg(s
, rn
);
6580 i
= (insn
>> 23) & 3;
6582 case 0: offset
= -4; break; /* DA */
6583 case 1: offset
= 0; break; /* IA */
6584 case 2: offset
= -8; break; /* DB */
6585 case 3: offset
= 4; break; /* IB */
6589 tcg_gen_addi_i32(addr
, addr
, offset
);
6590 /* Load PC into tmp and CPSR into tmp2. */
6591 tmp
= gen_ld32(addr
, 0);
6592 tcg_gen_addi_i32(addr
, addr
, 4);
6593 tmp2
= gen_ld32(addr
, 0);
6594 if (insn
& (1 << 21)) {
6595 /* Base writeback. */
6597 case 0: offset
= -8; break;
6598 case 1: offset
= 4; break;
6599 case 2: offset
= -4; break;
6600 case 3: offset
= 0; break;
6604 tcg_gen_addi_i32(addr
, addr
, offset
);
6605 store_reg(s
, rn
, addr
);
6607 tcg_temp_free_i32(addr
);
6609 gen_rfe(s
, tmp
, tmp2
);
6611 } else if ((insn
& 0x0e000000) == 0x0a000000) {
6612 /* branch link and change to thumb (blx <offset>) */
6615 val
= (uint32_t)s
->pc
;
6616 tmp
= tcg_temp_new_i32();
6617 tcg_gen_movi_i32(tmp
, val
);
6618 store_reg(s
, 14, tmp
);
6619 /* Sign-extend the 24-bit offset */
6620 offset
= (((int32_t)insn
) << 8) >> 8;
6621 /* offset * 4 + bit24 * 2 + (thumb bit) */
6622 val
+= (offset
<< 2) | ((insn
>> 23) & 2) | 1;
6623 /* pipeline offset */
6625 /* protected by ARCH(5); above, near the start of uncond block */
6628 } else if ((insn
& 0x0e000f00) == 0x0c000100) {
6629 if (arm_feature(env
, ARM_FEATURE_IWMMXT
)) {
6630 /* iWMMXt register transfer. */
6631 if (env
->cp15
.c15_cpar
& (1 << 1))
6632 if (!disas_iwmmxt_insn(env
, s
, insn
))
6635 } else if ((insn
& 0x0fe00000) == 0x0c400000) {
6636 /* Coprocessor double register transfer. */
6638 } else if ((insn
& 0x0f000010) == 0x0e000010) {
6639 /* Additional coprocessor register transfer. */
6640 } else if ((insn
& 0x0ff10020) == 0x01000000) {
6643 /* cps (privileged) */
6647 if (insn
& (1 << 19)) {
6648 if (insn
& (1 << 8))
6650 if (insn
& (1 << 7))
6652 if (insn
& (1 << 6))
6654 if (insn
& (1 << 18))
6657 if (insn
& (1 << 17)) {
6659 val
|= (insn
& 0x1f);
6662 gen_set_psr_im(s
, mask
, 0, val
);
6669 /* if not always execute, we generate a conditional jump to
6671 s
->condlabel
= gen_new_label();
6672 gen_test_cc(cond
^ 1, s
->condlabel
);
6675 if ((insn
& 0x0f900000) == 0x03000000) {
6676 if ((insn
& (1 << 21)) == 0) {
6678 rd
= (insn
>> 12) & 0xf;
6679 val
= ((insn
>> 4) & 0xf000) | (insn
& 0xfff);
6680 if ((insn
& (1 << 22)) == 0) {
6682 tmp
= tcg_temp_new_i32();
6683 tcg_gen_movi_i32(tmp
, val
);
6686 tmp
= load_reg(s
, rd
);
6687 tcg_gen_ext16u_i32(tmp
, tmp
);
6688 tcg_gen_ori_i32(tmp
, tmp
, val
<< 16);
6690 store_reg(s
, rd
, tmp
);
6692 if (((insn
>> 12) & 0xf) != 0xf)
6694 if (((insn
>> 16) & 0xf) == 0) {
6695 gen_nop_hint(s
, insn
& 0xff);
6697 /* CPSR = immediate */
6699 shift
= ((insn
>> 8) & 0xf) * 2;
6701 val
= (val
>> shift
) | (val
<< (32 - shift
));
6702 i
= ((insn
& (1 << 22)) != 0);
6703 if (gen_set_psr_im(s
, msr_mask(env
, s
, (insn
>> 16) & 0xf, i
), i
, val
))
6707 } else if ((insn
& 0x0f900000) == 0x01000000
6708 && (insn
& 0x00000090) != 0x00000090) {
6709 /* miscellaneous instructions */
6710 op1
= (insn
>> 21) & 3;
6711 sh
= (insn
>> 4) & 0xf;
6714 case 0x0: /* move program status register */
6717 tmp
= load_reg(s
, rm
);
6718 i
= ((op1
& 2) != 0);
6719 if (gen_set_psr(s
, msr_mask(env
, s
, (insn
>> 16) & 0xf, i
), i
, tmp
))
6723 rd
= (insn
>> 12) & 0xf;
6727 tmp
= load_cpu_field(spsr
);
6729 tmp
= tcg_temp_new_i32();
6730 gen_helper_cpsr_read(tmp
);
6732 store_reg(s
, rd
, tmp
);
6737 /* branch/exchange thumb (bx). */
6739 tmp
= load_reg(s
, rm
);
6741 } else if (op1
== 3) {
6744 rd
= (insn
>> 12) & 0xf;
6745 tmp
= load_reg(s
, rm
);
6746 gen_helper_clz(tmp
, tmp
);
6747 store_reg(s
, rd
, tmp
);
6755 /* Trivial implementation equivalent to bx. */
6756 tmp
= load_reg(s
, rm
);
6767 /* branch link/exchange thumb (blx) */
6768 tmp
= load_reg(s
, rm
);
6769 tmp2
= tcg_temp_new_i32();
6770 tcg_gen_movi_i32(tmp2
, s
->pc
);
6771 store_reg(s
, 14, tmp2
);
6774 case 0x5: /* saturating add/subtract */
6776 rd
= (insn
>> 12) & 0xf;
6777 rn
= (insn
>> 16) & 0xf;
6778 tmp
= load_reg(s
, rm
);
6779 tmp2
= load_reg(s
, rn
);
6781 gen_helper_double_saturate(tmp2
, tmp2
);
6783 gen_helper_sub_saturate(tmp
, tmp
, tmp2
);
6785 gen_helper_add_saturate(tmp
, tmp
, tmp2
);
6786 tcg_temp_free_i32(tmp2
);
6787 store_reg(s
, rd
, tmp
);
6790 /* SMC instruction (op1 == 3)
6791 and undefined instructions (op1 == 0 || op1 == 2)
6798 gen_exception_insn(s
, 4, EXCP_BKPT
);
6800 case 0x8: /* signed multiply */
6805 rs
= (insn
>> 8) & 0xf;
6806 rn
= (insn
>> 12) & 0xf;
6807 rd
= (insn
>> 16) & 0xf;
6809 /* (32 * 16) >> 16 */
6810 tmp
= load_reg(s
, rm
);
6811 tmp2
= load_reg(s
, rs
);
6813 tcg_gen_sari_i32(tmp2
, tmp2
, 16);
6816 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
6817 tcg_gen_shri_i64(tmp64
, tmp64
, 16);
6818 tmp
= tcg_temp_new_i32();
6819 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
6820 tcg_temp_free_i64(tmp64
);
6821 if ((sh
& 2) == 0) {
6822 tmp2
= load_reg(s
, rn
);
6823 gen_helper_add_setq(tmp
, tmp
, tmp2
);
6824 tcg_temp_free_i32(tmp2
);
6826 store_reg(s
, rd
, tmp
);
6829 tmp
= load_reg(s
, rm
);
6830 tmp2
= load_reg(s
, rs
);
6831 gen_mulxy(tmp
, tmp2
, sh
& 2, sh
& 4);
6832 tcg_temp_free_i32(tmp2
);
6834 tmp64
= tcg_temp_new_i64();
6835 tcg_gen_ext_i32_i64(tmp64
, tmp
);
6836 tcg_temp_free_i32(tmp
);
6837 gen_addq(s
, tmp64
, rn
, rd
);
6838 gen_storeq_reg(s
, rn
, rd
, tmp64
);
6839 tcg_temp_free_i64(tmp64
);
6842 tmp2
= load_reg(s
, rn
);
6843 gen_helper_add_setq(tmp
, tmp
, tmp2
);
6844 tcg_temp_free_i32(tmp2
);
6846 store_reg(s
, rd
, tmp
);
6853 } else if (((insn
& 0x0e000000) == 0 &&
6854 (insn
& 0x00000090) != 0x90) ||
6855 ((insn
& 0x0e000000) == (1 << 25))) {
6856 int set_cc
, logic_cc
, shiftop
;
6858 op1
= (insn
>> 21) & 0xf;
6859 set_cc
= (insn
>> 20) & 1;
6860 logic_cc
= table_logic_cc
[op1
] & set_cc
;
6862 /* data processing instruction */
6863 if (insn
& (1 << 25)) {
6864 /* immediate operand */
6866 shift
= ((insn
>> 8) & 0xf) * 2;
6868 val
= (val
>> shift
) | (val
<< (32 - shift
));
6870 tmp2
= tcg_temp_new_i32();
6871 tcg_gen_movi_i32(tmp2
, val
);
6872 if (logic_cc
&& shift
) {
6873 gen_set_CF_bit31(tmp2
);
6878 tmp2
= load_reg(s
, rm
);
6879 shiftop
= (insn
>> 5) & 3;
6880 if (!(insn
& (1 << 4))) {
6881 shift
= (insn
>> 7) & 0x1f;
6882 gen_arm_shift_im(tmp2
, shiftop
, shift
, logic_cc
);
6884 rs
= (insn
>> 8) & 0xf;
6885 tmp
= load_reg(s
, rs
);
6886 gen_arm_shift_reg(tmp2
, shiftop
, tmp
, logic_cc
);
6889 if (op1
!= 0x0f && op1
!= 0x0d) {
6890 rn
= (insn
>> 16) & 0xf;
6891 tmp
= load_reg(s
, rn
);
6895 rd
= (insn
>> 12) & 0xf;
6898 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
6902 store_reg_bx(env
, s
, rd
, tmp
);
6905 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
6909 store_reg_bx(env
, s
, rd
, tmp
);
6912 if (set_cc
&& rd
== 15) {
6913 /* SUBS r15, ... is used for exception return. */
6917 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
6918 gen_exception_return(s
, tmp
);
6921 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
6923 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
6925 store_reg_bx(env
, s
, rd
, tmp
);
6930 gen_helper_sub_cc(tmp
, tmp2
, tmp
);
6932 tcg_gen_sub_i32(tmp
, tmp2
, tmp
);
6934 store_reg_bx(env
, s
, rd
, tmp
);
6938 gen_helper_add_cc(tmp
, tmp
, tmp2
);
6940 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
6942 store_reg_bx(env
, s
, rd
, tmp
);
6946 gen_helper_adc_cc(tmp
, tmp
, tmp2
);
6948 gen_add_carry(tmp
, tmp
, tmp2
);
6950 store_reg_bx(env
, s
, rd
, tmp
);
6954 gen_helper_sbc_cc(tmp
, tmp
, tmp2
);
6956 gen_sub_carry(tmp
, tmp
, tmp2
);
6958 store_reg_bx(env
, s
, rd
, tmp
);
6962 gen_helper_sbc_cc(tmp
, tmp2
, tmp
);
6964 gen_sub_carry(tmp
, tmp2
, tmp
);
6966 store_reg_bx(env
, s
, rd
, tmp
);
6970 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
6973 tcg_temp_free_i32(tmp
);
6977 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
6980 tcg_temp_free_i32(tmp
);
6984 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
6986 tcg_temp_free_i32(tmp
);
6990 gen_helper_add_cc(tmp
, tmp
, tmp2
);
6992 tcg_temp_free_i32(tmp
);
6995 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
6999 store_reg_bx(env
, s
, rd
, tmp
);
7002 if (logic_cc
&& rd
== 15) {
7003 /* MOVS r15, ... is used for exception return. */
7007 gen_exception_return(s
, tmp2
);
7012 store_reg_bx(env
, s
, rd
, tmp2
);
7016 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
7020 store_reg_bx(env
, s
, rd
, tmp
);
7024 tcg_gen_not_i32(tmp2
, tmp2
);
7028 store_reg_bx(env
, s
, rd
, tmp2
);
7031 if (op1
!= 0x0f && op1
!= 0x0d) {
7032 tcg_temp_free_i32(tmp2
);
7035 /* other instructions */
7036 op1
= (insn
>> 24) & 0xf;
7040 /* multiplies, extra load/stores */
7041 sh
= (insn
>> 5) & 3;
7044 rd
= (insn
>> 16) & 0xf;
7045 rn
= (insn
>> 12) & 0xf;
7046 rs
= (insn
>> 8) & 0xf;
7048 op1
= (insn
>> 20) & 0xf;
7050 case 0: case 1: case 2: case 3: case 6:
7052 tmp
= load_reg(s
, rs
);
7053 tmp2
= load_reg(s
, rm
);
7054 tcg_gen_mul_i32(tmp
, tmp
, tmp2
);
7055 tcg_temp_free_i32(tmp2
);
7056 if (insn
& (1 << 22)) {
7057 /* Subtract (mls) */
7059 tmp2
= load_reg(s
, rn
);
7060 tcg_gen_sub_i32(tmp
, tmp2
, tmp
);
7061 tcg_temp_free_i32(tmp2
);
7062 } else if (insn
& (1 << 21)) {
7064 tmp2
= load_reg(s
, rn
);
7065 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7066 tcg_temp_free_i32(tmp2
);
7068 if (insn
& (1 << 20))
7070 store_reg(s
, rd
, tmp
);
7073 /* 64 bit mul double accumulate (UMAAL) */
7075 tmp
= load_reg(s
, rs
);
7076 tmp2
= load_reg(s
, rm
);
7077 tmp64
= gen_mulu_i64_i32(tmp
, tmp2
);
7078 gen_addq_lo(s
, tmp64
, rn
);
7079 gen_addq_lo(s
, tmp64
, rd
);
7080 gen_storeq_reg(s
, rn
, rd
, tmp64
);
7081 tcg_temp_free_i64(tmp64
);
7083 case 8: case 9: case 10: case 11:
7084 case 12: case 13: case 14: case 15:
7085 /* 64 bit mul: UMULL, UMLAL, SMULL, SMLAL. */
7086 tmp
= load_reg(s
, rs
);
7087 tmp2
= load_reg(s
, rm
);
7088 if (insn
& (1 << 22)) {
7089 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
7091 tmp64
= gen_mulu_i64_i32(tmp
, tmp2
);
7093 if (insn
& (1 << 21)) { /* mult accumulate */
7094 gen_addq(s
, tmp64
, rn
, rd
);
7096 if (insn
& (1 << 20)) {
7097 gen_logicq_cc(tmp64
);
7099 gen_storeq_reg(s
, rn
, rd
, tmp64
);
7100 tcg_temp_free_i64(tmp64
);
7106 rn
= (insn
>> 16) & 0xf;
7107 rd
= (insn
>> 12) & 0xf;
7108 if (insn
& (1 << 23)) {
7109 /* load/store exclusive */
7110 op1
= (insn
>> 21) & 0x3;
7115 addr
= tcg_temp_local_new_i32();
7116 load_reg_var(s
, addr
, rn
);
7117 if (insn
& (1 << 20)) {
7120 gen_load_exclusive(s
, rd
, 15, addr
, 2);
7122 case 1: /* ldrexd */
7123 gen_load_exclusive(s
, rd
, rd
+ 1, addr
, 3);
7125 case 2: /* ldrexb */
7126 gen_load_exclusive(s
, rd
, 15, addr
, 0);
7128 case 3: /* ldrexh */
7129 gen_load_exclusive(s
, rd
, 15, addr
, 1);
7138 gen_store_exclusive(s
, rd
, rm
, 15, addr
, 2);
7140 case 1: /* strexd */
7141 gen_store_exclusive(s
, rd
, rm
, rm
+ 1, addr
, 3);
7143 case 2: /* strexb */
7144 gen_store_exclusive(s
, rd
, rm
, 15, addr
, 0);
7146 case 3: /* strexh */
7147 gen_store_exclusive(s
, rd
, rm
, 15, addr
, 1);
7153 tcg_temp_free(addr
);
7155 /* SWP instruction */
7158 /* ??? This is not really atomic. However we know
7159 we never have multiple CPUs running in parallel,
7160 so it is good enough. */
7161 addr
= load_reg(s
, rn
);
7162 tmp
= load_reg(s
, rm
);
7163 if (insn
& (1 << 22)) {
7164 tmp2
= gen_ld8u(addr
, IS_USER(s
));
7165 gen_st8(tmp
, addr
, IS_USER(s
));
7167 tmp2
= gen_ld32(addr
, IS_USER(s
));
7168 gen_st32(tmp
, addr
, IS_USER(s
));
7170 tcg_temp_free_i32(addr
);
7171 store_reg(s
, rd
, tmp2
);
7177 /* Misc load/store */
7178 rn
= (insn
>> 16) & 0xf;
7179 rd
= (insn
>> 12) & 0xf;
7180 addr
= load_reg(s
, rn
);
7181 if (insn
& (1 << 24))
7182 gen_add_datah_offset(s
, insn
, 0, addr
);
7184 if (insn
& (1 << 20)) {
7188 tmp
= gen_ld16u(addr
, IS_USER(s
));
7191 tmp
= gen_ld8s(addr
, IS_USER(s
));
7195 tmp
= gen_ld16s(addr
, IS_USER(s
));
7199 } else if (sh
& 2) {
7204 tmp
= load_reg(s
, rd
);
7205 gen_st32(tmp
, addr
, IS_USER(s
));
7206 tcg_gen_addi_i32(addr
, addr
, 4);
7207 tmp
= load_reg(s
, rd
+ 1);
7208 gen_st32(tmp
, addr
, IS_USER(s
));
7212 tmp
= gen_ld32(addr
, IS_USER(s
));
7213 store_reg(s
, rd
, tmp
);
7214 tcg_gen_addi_i32(addr
, addr
, 4);
7215 tmp
= gen_ld32(addr
, IS_USER(s
));
7219 address_offset
= -4;
7222 tmp
= load_reg(s
, rd
);
7223 gen_st16(tmp
, addr
, IS_USER(s
));
7226 /* Perform base writeback before the loaded value to
7227 ensure correct behavior with overlapping index registers.
7228 ldrd with base writeback is is undefined if the
7229 destination and index registers overlap. */
7230 if (!(insn
& (1 << 24))) {
7231 gen_add_datah_offset(s
, insn
, address_offset
, addr
);
7232 store_reg(s
, rn
, addr
);
7233 } else if (insn
& (1 << 21)) {
7235 tcg_gen_addi_i32(addr
, addr
, address_offset
);
7236 store_reg(s
, rn
, addr
);
7238 tcg_temp_free_i32(addr
);
7241 /* Complete the load. */
7242 store_reg(s
, rd
, tmp
);
7251 if (insn
& (1 << 4)) {
7253 /* Armv6 Media instructions. */
7255 rn
= (insn
>> 16) & 0xf;
7256 rd
= (insn
>> 12) & 0xf;
7257 rs
= (insn
>> 8) & 0xf;
7258 switch ((insn
>> 23) & 3) {
7259 case 0: /* Parallel add/subtract. */
7260 op1
= (insn
>> 20) & 7;
7261 tmp
= load_reg(s
, rn
);
7262 tmp2
= load_reg(s
, rm
);
7263 sh
= (insn
>> 5) & 7;
7264 if ((op1
& 3) == 0 || sh
== 5 || sh
== 6)
7266 gen_arm_parallel_addsub(op1
, sh
, tmp
, tmp2
);
7267 tcg_temp_free_i32(tmp2
);
7268 store_reg(s
, rd
, tmp
);
7271 if ((insn
& 0x00700020) == 0) {
7272 /* Halfword pack. */
7273 tmp
= load_reg(s
, rn
);
7274 tmp2
= load_reg(s
, rm
);
7275 shift
= (insn
>> 7) & 0x1f;
7276 if (insn
& (1 << 6)) {
7280 tcg_gen_sari_i32(tmp2
, tmp2
, shift
);
7281 tcg_gen_andi_i32(tmp
, tmp
, 0xffff0000);
7282 tcg_gen_ext16u_i32(tmp2
, tmp2
);
7286 tcg_gen_shli_i32(tmp2
, tmp2
, shift
);
7287 tcg_gen_ext16u_i32(tmp
, tmp
);
7288 tcg_gen_andi_i32(tmp2
, tmp2
, 0xffff0000);
7290 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
7291 tcg_temp_free_i32(tmp2
);
7292 store_reg(s
, rd
, tmp
);
7293 } else if ((insn
& 0x00200020) == 0x00200000) {
7295 tmp
= load_reg(s
, rm
);
7296 shift
= (insn
>> 7) & 0x1f;
7297 if (insn
& (1 << 6)) {
7300 tcg_gen_sari_i32(tmp
, tmp
, shift
);
7302 tcg_gen_shli_i32(tmp
, tmp
, shift
);
7304 sh
= (insn
>> 16) & 0x1f;
7305 tmp2
= tcg_const_i32(sh
);
7306 if (insn
& (1 << 22))
7307 gen_helper_usat(tmp
, tmp
, tmp2
);
7309 gen_helper_ssat(tmp
, tmp
, tmp2
);
7310 tcg_temp_free_i32(tmp2
);
7311 store_reg(s
, rd
, tmp
);
7312 } else if ((insn
& 0x00300fe0) == 0x00200f20) {
7314 tmp
= load_reg(s
, rm
);
7315 sh
= (insn
>> 16) & 0x1f;
7316 tmp2
= tcg_const_i32(sh
);
7317 if (insn
& (1 << 22))
7318 gen_helper_usat16(tmp
, tmp
, tmp2
);
7320 gen_helper_ssat16(tmp
, tmp
, tmp2
);
7321 tcg_temp_free_i32(tmp2
);
7322 store_reg(s
, rd
, tmp
);
7323 } else if ((insn
& 0x00700fe0) == 0x00000fa0) {
7325 tmp
= load_reg(s
, rn
);
7326 tmp2
= load_reg(s
, rm
);
7327 tmp3
= tcg_temp_new_i32();
7328 tcg_gen_ld_i32(tmp3
, cpu_env
, offsetof(CPUState
, GE
));
7329 gen_helper_sel_flags(tmp
, tmp3
, tmp
, tmp2
);
7330 tcg_temp_free_i32(tmp3
);
7331 tcg_temp_free_i32(tmp2
);
7332 store_reg(s
, rd
, tmp
);
7333 } else if ((insn
& 0x000003e0) == 0x00000060) {
7334 tmp
= load_reg(s
, rm
);
7335 shift
= (insn
>> 10) & 3;
7336 /* ??? In many cases it's not neccessary to do a
7337 rotate, a shift is sufficient. */
7339 tcg_gen_rotri_i32(tmp
, tmp
, shift
* 8);
7340 op1
= (insn
>> 20) & 7;
7342 case 0: gen_sxtb16(tmp
); break;
7343 case 2: gen_sxtb(tmp
); break;
7344 case 3: gen_sxth(tmp
); break;
7345 case 4: gen_uxtb16(tmp
); break;
7346 case 6: gen_uxtb(tmp
); break;
7347 case 7: gen_uxth(tmp
); break;
7348 default: goto illegal_op
;
7351 tmp2
= load_reg(s
, rn
);
7352 if ((op1
& 3) == 0) {
7353 gen_add16(tmp
, tmp2
);
7355 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7356 tcg_temp_free_i32(tmp2
);
7359 store_reg(s
, rd
, tmp
);
7360 } else if ((insn
& 0x003f0f60) == 0x003f0f20) {
7362 tmp
= load_reg(s
, rm
);
7363 if (insn
& (1 << 22)) {
7364 if (insn
& (1 << 7)) {
7368 gen_helper_rbit(tmp
, tmp
);
7371 if (insn
& (1 << 7))
7374 tcg_gen_bswap32_i32(tmp
, tmp
);
7376 store_reg(s
, rd
, tmp
);
7381 case 2: /* Multiplies (Type 3). */
7382 tmp
= load_reg(s
, rm
);
7383 tmp2
= load_reg(s
, rs
);
7384 if (insn
& (1 << 20)) {
7385 /* Signed multiply most significant [accumulate].
7386 (SMMUL, SMMLA, SMMLS) */
7387 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
7390 tmp
= load_reg(s
, rd
);
7391 if (insn
& (1 << 6)) {
7392 tmp64
= gen_subq_msw(tmp64
, tmp
);
7394 tmp64
= gen_addq_msw(tmp64
, tmp
);
7397 if (insn
& (1 << 5)) {
7398 tcg_gen_addi_i64(tmp64
, tmp64
, 0x80000000u
);
7400 tcg_gen_shri_i64(tmp64
, tmp64
, 32);
7401 tmp
= tcg_temp_new_i32();
7402 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
7403 tcg_temp_free_i64(tmp64
);
7404 store_reg(s
, rn
, tmp
);
7406 if (insn
& (1 << 5))
7407 gen_swap_half(tmp2
);
7408 gen_smul_dual(tmp
, tmp2
);
7409 if (insn
& (1 << 6)) {
7410 /* This subtraction cannot overflow. */
7411 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
7413 /* This addition cannot overflow 32 bits;
7414 * however it may overflow considered as a signed
7415 * operation, in which case we must set the Q flag.
7417 gen_helper_add_setq(tmp
, tmp
, tmp2
);
7419 tcg_temp_free_i32(tmp2
);
7420 if (insn
& (1 << 22)) {
7421 /* smlald, smlsld */
7422 tmp64
= tcg_temp_new_i64();
7423 tcg_gen_ext_i32_i64(tmp64
, tmp
);
7424 tcg_temp_free_i32(tmp
);
7425 gen_addq(s
, tmp64
, rd
, rn
);
7426 gen_storeq_reg(s
, rd
, rn
, tmp64
);
7427 tcg_temp_free_i64(tmp64
);
7429 /* smuad, smusd, smlad, smlsd */
7432 tmp2
= load_reg(s
, rd
);
7433 gen_helper_add_setq(tmp
, tmp
, tmp2
);
7434 tcg_temp_free_i32(tmp2
);
7436 store_reg(s
, rn
, tmp
);
7441 op1
= ((insn
>> 17) & 0x38) | ((insn
>> 5) & 7);
7443 case 0: /* Unsigned sum of absolute differences. */
7445 tmp
= load_reg(s
, rm
);
7446 tmp2
= load_reg(s
, rs
);
7447 gen_helper_usad8(tmp
, tmp
, tmp2
);
7448 tcg_temp_free_i32(tmp2
);
7450 tmp2
= load_reg(s
, rd
);
7451 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
7452 tcg_temp_free_i32(tmp2
);
7454 store_reg(s
, rn
, tmp
);
7456 case 0x20: case 0x24: case 0x28: case 0x2c:
7457 /* Bitfield insert/clear. */
7459 shift
= (insn
>> 7) & 0x1f;
7460 i
= (insn
>> 16) & 0x1f;
7463 tmp
= tcg_temp_new_i32();
7464 tcg_gen_movi_i32(tmp
, 0);
7466 tmp
= load_reg(s
, rm
);
7469 tmp2
= load_reg(s
, rd
);
7470 gen_bfi(tmp
, tmp2
, tmp
, shift
, (1u << i
) - 1);
7471 tcg_temp_free_i32(tmp2
);
7473 store_reg(s
, rd
, tmp
);
7475 case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
7476 case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
7478 tmp
= load_reg(s
, rm
);
7479 shift
= (insn
>> 7) & 0x1f;
7480 i
= ((insn
>> 16) & 0x1f) + 1;
7485 gen_ubfx(tmp
, shift
, (1u << i
) - 1);
7487 gen_sbfx(tmp
, shift
, i
);
7490 store_reg(s
, rd
, tmp
);
7500 /* Check for undefined extension instructions
7501 * per the ARM Bible IE:
7502 * xxxx 0111 1111 xxxx xxxx xxxx 1111 xxxx
7504 sh
= (0xf << 20) | (0xf << 4);
7505 if (op1
== 0x7 && ((insn
& sh
) == sh
))
7509 /* load/store byte/word */
7510 rn
= (insn
>> 16) & 0xf;
7511 rd
= (insn
>> 12) & 0xf;
7512 tmp2
= load_reg(s
, rn
);
7513 i
= (IS_USER(s
) || (insn
& 0x01200000) == 0x00200000);
7514 if (insn
& (1 << 24))
7515 gen_add_data_offset(s
, insn
, tmp2
);
7516 if (insn
& (1 << 20)) {
7518 if (insn
& (1 << 22)) {
7519 tmp
= gen_ld8u(tmp2
, i
);
7521 tmp
= gen_ld32(tmp2
, i
);
7525 tmp
= load_reg(s
, rd
);
7526 if (insn
& (1 << 22))
7527 gen_st8(tmp
, tmp2
, i
);
7529 gen_st32(tmp
, tmp2
, i
);
7531 if (!(insn
& (1 << 24))) {
7532 gen_add_data_offset(s
, insn
, tmp2
);
7533 store_reg(s
, rn
, tmp2
);
7534 } else if (insn
& (1 << 21)) {
7535 store_reg(s
, rn
, tmp2
);
7537 tcg_temp_free_i32(tmp2
);
7539 if (insn
& (1 << 20)) {
7540 /* Complete the load. */
7541 store_reg_from_load(env
, s
, rd
, tmp
);
7547 int j
, n
, user
, loaded_base
;
7549 /* load/store multiple words */
7550 /* XXX: store correct base if write back */
7552 if (insn
& (1 << 22)) {
7554 goto illegal_op
; /* only usable in supervisor mode */
7556 if ((insn
& (1 << 15)) == 0)
7559 rn
= (insn
>> 16) & 0xf;
7560 addr
= load_reg(s
, rn
);
7562 /* compute total size */
7564 TCGV_UNUSED(loaded_var
);
7567 if (insn
& (1 << i
))
7570 /* XXX: test invalid n == 0 case ? */
7571 if (insn
& (1 << 23)) {
7572 if (insn
& (1 << 24)) {
7574 tcg_gen_addi_i32(addr
, addr
, 4);
7576 /* post increment */
7579 if (insn
& (1 << 24)) {
7581 tcg_gen_addi_i32(addr
, addr
, -(n
* 4));
7583 /* post decrement */
7585 tcg_gen_addi_i32(addr
, addr
, -((n
- 1) * 4));
7590 if (insn
& (1 << i
)) {
7591 if (insn
& (1 << 20)) {
7593 tmp
= gen_ld32(addr
, IS_USER(s
));
7595 tmp2
= tcg_const_i32(i
);
7596 gen_helper_set_user_reg(tmp2
, tmp
);
7597 tcg_temp_free_i32(tmp2
);
7598 tcg_temp_free_i32(tmp
);
7599 } else if (i
== rn
) {
7603 store_reg_from_load(env
, s
, i
, tmp
);
7608 /* special case: r15 = PC + 8 */
7609 val
= (long)s
->pc
+ 4;
7610 tmp
= tcg_temp_new_i32();
7611 tcg_gen_movi_i32(tmp
, val
);
7613 tmp
= tcg_temp_new_i32();
7614 tmp2
= tcg_const_i32(i
);
7615 gen_helper_get_user_reg(tmp
, tmp2
);
7616 tcg_temp_free_i32(tmp2
);
7618 tmp
= load_reg(s
, i
);
7620 gen_st32(tmp
, addr
, IS_USER(s
));
7623 /* no need to add after the last transfer */
7625 tcg_gen_addi_i32(addr
, addr
, 4);
7628 if (insn
& (1 << 21)) {
7630 if (insn
& (1 << 23)) {
7631 if (insn
& (1 << 24)) {
7634 /* post increment */
7635 tcg_gen_addi_i32(addr
, addr
, 4);
7638 if (insn
& (1 << 24)) {
7641 tcg_gen_addi_i32(addr
, addr
, -((n
- 1) * 4));
7643 /* post decrement */
7644 tcg_gen_addi_i32(addr
, addr
, -(n
* 4));
7647 store_reg(s
, rn
, addr
);
7649 tcg_temp_free_i32(addr
);
7652 store_reg(s
, rn
, loaded_var
);
7654 if ((insn
& (1 << 22)) && !user
) {
7655 /* Restore CPSR from SPSR. */
7656 tmp
= load_cpu_field(spsr
);
7657 gen_set_cpsr(tmp
, 0xffffffff);
7658 tcg_temp_free_i32(tmp
);
7659 s
->is_jmp
= DISAS_UPDATE
;
7668 /* branch (and link) */
7669 val
= (int32_t)s
->pc
;
7670 if (insn
& (1 << 24)) {
7671 tmp
= tcg_temp_new_i32();
7672 tcg_gen_movi_i32(tmp
, val
);
7673 store_reg(s
, 14, tmp
);
7675 offset
= (((int32_t)insn
<< 8) >> 8);
7676 val
+= (offset
<< 2) + 4;
7684 if (disas_coproc_insn(env
, s
, insn
))
7689 gen_set_pc_im(s
->pc
);
7690 s
->is_jmp
= DISAS_SWI
;
7694 gen_exception_insn(s
, 4, EXCP_UDEF
);
7700 /* Return true if this is a Thumb-2 logical op. */
7702 thumb2_logic_op(int op
)
7707 /* Generate code for a Thumb-2 data processing operation. If CONDS is nonzero
7708 then set condition code flags based on the result of the operation.
7709 If SHIFTER_OUT is nonzero then set the carry flag for logical operations
7710 to the high bit of T1.
7711 Returns zero if the opcode is valid. */
7714 gen_thumb2_data_op(DisasContext
*s
, int op
, int conds
, uint32_t shifter_out
, TCGv t0
, TCGv t1
)
7721 tcg_gen_and_i32(t0
, t0
, t1
);
7725 tcg_gen_andc_i32(t0
, t0
, t1
);
7729 tcg_gen_or_i32(t0
, t0
, t1
);
7733 tcg_gen_orc_i32(t0
, t0
, t1
);
7737 tcg_gen_xor_i32(t0
, t0
, t1
);
7742 gen_helper_add_cc(t0
, t0
, t1
);
7744 tcg_gen_add_i32(t0
, t0
, t1
);
7748 gen_helper_adc_cc(t0
, t0
, t1
);
7754 gen_helper_sbc_cc(t0
, t0
, t1
);
7756 gen_sub_carry(t0
, t0
, t1
);
7760 gen_helper_sub_cc(t0
, t0
, t1
);
7762 tcg_gen_sub_i32(t0
, t0
, t1
);
7766 gen_helper_sub_cc(t0
, t1
, t0
);
7768 tcg_gen_sub_i32(t0
, t1
, t0
);
7770 default: /* 5, 6, 7, 9, 12, 15. */
7776 gen_set_CF_bit31(t1
);
7781 /* Translate a 32-bit thumb instruction. Returns nonzero if the instruction
7783 static int disas_thumb2_insn(CPUState
*env
, DisasContext
*s
, uint16_t insn_hw1
)
7785 uint32_t insn
, imm
, shift
, offset
;
7786 uint32_t rd
, rn
, rm
, rs
;
7797 if (!(arm_feature(env
, ARM_FEATURE_THUMB2
)
7798 || arm_feature (env
, ARM_FEATURE_M
))) {
7799 /* Thumb-1 cores may need to treat bl and blx as a pair of
7800 16-bit instructions to get correct prefetch abort behavior. */
7802 if ((insn
& (1 << 12)) == 0) {
7804 /* Second half of blx. */
7805 offset
= ((insn
& 0x7ff) << 1);
7806 tmp
= load_reg(s
, 14);
7807 tcg_gen_addi_i32(tmp
, tmp
, offset
);
7808 tcg_gen_andi_i32(tmp
, tmp
, 0xfffffffc);
7810 tmp2
= tcg_temp_new_i32();
7811 tcg_gen_movi_i32(tmp2
, s
->pc
| 1);
7812 store_reg(s
, 14, tmp2
);
7816 if (insn
& (1 << 11)) {
7817 /* Second half of bl. */
7818 offset
= ((insn
& 0x7ff) << 1) | 1;
7819 tmp
= load_reg(s
, 14);
7820 tcg_gen_addi_i32(tmp
, tmp
, offset
);
7822 tmp2
= tcg_temp_new_i32();
7823 tcg_gen_movi_i32(tmp2
, s
->pc
| 1);
7824 store_reg(s
, 14, tmp2
);
7828 if ((s
->pc
& ~TARGET_PAGE_MASK
) == 0) {
7829 /* Instruction spans a page boundary. Implement it as two
7830 16-bit instructions in case the second half causes an
7832 offset
= ((int32_t)insn
<< 21) >> 9;
7833 tcg_gen_movi_i32(cpu_R
[14], s
->pc
+ 2 + offset
);
7836 /* Fall through to 32-bit decode. */
7839 insn
= lduw_code(s
->pc
);
7841 insn
|= (uint32_t)insn_hw1
<< 16;
7843 if ((insn
& 0xf800e800) != 0xf000e800) {
7847 rn
= (insn
>> 16) & 0xf;
7848 rs
= (insn
>> 12) & 0xf;
7849 rd
= (insn
>> 8) & 0xf;
7851 switch ((insn
>> 25) & 0xf) {
7852 case 0: case 1: case 2: case 3:
7853 /* 16-bit instructions. Should never happen. */
7856 if (insn
& (1 << 22)) {
7857 /* Other load/store, table branch. */
7858 if (insn
& 0x01200000) {
7859 /* Load/store doubleword. */
7861 addr
= tcg_temp_new_i32();
7862 tcg_gen_movi_i32(addr
, s
->pc
& ~3);
7864 addr
= load_reg(s
, rn
);
7866 offset
= (insn
& 0xff) * 4;
7867 if ((insn
& (1 << 23)) == 0)
7869 if (insn
& (1 << 24)) {
7870 tcg_gen_addi_i32(addr
, addr
, offset
);
7873 if (insn
& (1 << 20)) {
7875 tmp
= gen_ld32(addr
, IS_USER(s
));
7876 store_reg(s
, rs
, tmp
);
7877 tcg_gen_addi_i32(addr
, addr
, 4);
7878 tmp
= gen_ld32(addr
, IS_USER(s
));
7879 store_reg(s
, rd
, tmp
);
7882 tmp
= load_reg(s
, rs
);
7883 gen_st32(tmp
, addr
, IS_USER(s
));
7884 tcg_gen_addi_i32(addr
, addr
, 4);
7885 tmp
= load_reg(s
, rd
);
7886 gen_st32(tmp
, addr
, IS_USER(s
));
7888 if (insn
& (1 << 21)) {
7889 /* Base writeback. */
7892 tcg_gen_addi_i32(addr
, addr
, offset
- 4);
7893 store_reg(s
, rn
, addr
);
7895 tcg_temp_free_i32(addr
);
7897 } else if ((insn
& (1 << 23)) == 0) {
7898 /* Load/store exclusive word. */
7899 addr
= tcg_temp_local_new();
7900 load_reg_var(s
, addr
, rn
);
7901 tcg_gen_addi_i32(addr
, addr
, (insn
& 0xff) << 2);
7902 if (insn
& (1 << 20)) {
7903 gen_load_exclusive(s
, rs
, 15, addr
, 2);
7905 gen_store_exclusive(s
, rd
, rs
, 15, addr
, 2);
7907 tcg_temp_free(addr
);
7908 } else if ((insn
& (1 << 6)) == 0) {
7911 addr
= tcg_temp_new_i32();
7912 tcg_gen_movi_i32(addr
, s
->pc
);
7914 addr
= load_reg(s
, rn
);
7916 tmp
= load_reg(s
, rm
);
7917 tcg_gen_add_i32(addr
, addr
, tmp
);
7918 if (insn
& (1 << 4)) {
7920 tcg_gen_add_i32(addr
, addr
, tmp
);
7921 tcg_temp_free_i32(tmp
);
7922 tmp
= gen_ld16u(addr
, IS_USER(s
));
7924 tcg_temp_free_i32(tmp
);
7925 tmp
= gen_ld8u(addr
, IS_USER(s
));
7927 tcg_temp_free_i32(addr
);
7928 tcg_gen_shli_i32(tmp
, tmp
, 1);
7929 tcg_gen_addi_i32(tmp
, tmp
, s
->pc
);
7930 store_reg(s
, 15, tmp
);
7932 /* Load/store exclusive byte/halfword/doubleword. */
7934 op
= (insn
>> 4) & 0x3;
7938 addr
= tcg_temp_local_new();
7939 load_reg_var(s
, addr
, rn
);
7940 if (insn
& (1 << 20)) {
7941 gen_load_exclusive(s
, rs
, rd
, addr
, op
);
7943 gen_store_exclusive(s
, rm
, rs
, rd
, addr
, op
);
7945 tcg_temp_free(addr
);
7948 /* Load/store multiple, RFE, SRS. */
7949 if (((insn
>> 23) & 1) == ((insn
>> 24) & 1)) {
7950 /* Not available in user mode. */
7953 if (insn
& (1 << 20)) {
7955 addr
= load_reg(s
, rn
);
7956 if ((insn
& (1 << 24)) == 0)
7957 tcg_gen_addi_i32(addr
, addr
, -8);
7958 /* Load PC into tmp and CPSR into tmp2. */
7959 tmp
= gen_ld32(addr
, 0);
7960 tcg_gen_addi_i32(addr
, addr
, 4);
7961 tmp2
= gen_ld32(addr
, 0);
7962 if (insn
& (1 << 21)) {
7963 /* Base writeback. */
7964 if (insn
& (1 << 24)) {
7965 tcg_gen_addi_i32(addr
, addr
, 4);
7967 tcg_gen_addi_i32(addr
, addr
, -4);
7969 store_reg(s
, rn
, addr
);
7971 tcg_temp_free_i32(addr
);
7973 gen_rfe(s
, tmp
, tmp2
);
7977 addr
= tcg_temp_new_i32();
7978 tmp
= tcg_const_i32(op
);
7979 gen_helper_get_r13_banked(addr
, cpu_env
, tmp
);
7980 tcg_temp_free_i32(tmp
);
7981 if ((insn
& (1 << 24)) == 0) {
7982 tcg_gen_addi_i32(addr
, addr
, -8);
7984 tmp
= load_reg(s
, 14);
7985 gen_st32(tmp
, addr
, 0);
7986 tcg_gen_addi_i32(addr
, addr
, 4);
7987 tmp
= tcg_temp_new_i32();
7988 gen_helper_cpsr_read(tmp
);
7989 gen_st32(tmp
, addr
, 0);
7990 if (insn
& (1 << 21)) {
7991 if ((insn
& (1 << 24)) == 0) {
7992 tcg_gen_addi_i32(addr
, addr
, -4);
7994 tcg_gen_addi_i32(addr
, addr
, 4);
7996 tmp
= tcg_const_i32(op
);
7997 gen_helper_set_r13_banked(cpu_env
, tmp
, addr
);
7998 tcg_temp_free_i32(tmp
);
8000 tcg_temp_free_i32(addr
);
8005 /* Load/store multiple. */
8006 addr
= load_reg(s
, rn
);
8008 for (i
= 0; i
< 16; i
++) {
8009 if (insn
& (1 << i
))
8012 if (insn
& (1 << 24)) {
8013 tcg_gen_addi_i32(addr
, addr
, -offset
);
8016 for (i
= 0; i
< 16; i
++) {
8017 if ((insn
& (1 << i
)) == 0)
8019 if (insn
& (1 << 20)) {
8021 tmp
= gen_ld32(addr
, IS_USER(s
));
8025 store_reg(s
, i
, tmp
);
8029 tmp
= load_reg(s
, i
);
8030 gen_st32(tmp
, addr
, IS_USER(s
));
8032 tcg_gen_addi_i32(addr
, addr
, 4);
8034 if (insn
& (1 << 21)) {
8035 /* Base register writeback. */
8036 if (insn
& (1 << 24)) {
8037 tcg_gen_addi_i32(addr
, addr
, -offset
);
8039 /* Fault if writeback register is in register list. */
8040 if (insn
& (1 << rn
))
8042 store_reg(s
, rn
, addr
);
8044 tcg_temp_free_i32(addr
);
8051 op
= (insn
>> 21) & 0xf;
8053 /* Halfword pack. */
8054 tmp
= load_reg(s
, rn
);
8055 tmp2
= load_reg(s
, rm
);
8056 shift
= ((insn
>> 10) & 0x1c) | ((insn
>> 6) & 0x3);
8057 if (insn
& (1 << 5)) {
8061 tcg_gen_sari_i32(tmp2
, tmp2
, shift
);
8062 tcg_gen_andi_i32(tmp
, tmp
, 0xffff0000);
8063 tcg_gen_ext16u_i32(tmp2
, tmp2
);
8067 tcg_gen_shli_i32(tmp2
, tmp2
, shift
);
8068 tcg_gen_ext16u_i32(tmp
, tmp
);
8069 tcg_gen_andi_i32(tmp2
, tmp2
, 0xffff0000);
8071 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
8072 tcg_temp_free_i32(tmp2
);
8073 store_reg(s
, rd
, tmp
);
8075 /* Data processing register constant shift. */
8077 tmp
= tcg_temp_new_i32();
8078 tcg_gen_movi_i32(tmp
, 0);
8080 tmp
= load_reg(s
, rn
);
8082 tmp2
= load_reg(s
, rm
);
8084 shiftop
= (insn
>> 4) & 3;
8085 shift
= ((insn
>> 6) & 3) | ((insn
>> 10) & 0x1c);
8086 conds
= (insn
& (1 << 20)) != 0;
8087 logic_cc
= (conds
&& thumb2_logic_op(op
));
8088 gen_arm_shift_im(tmp2
, shiftop
, shift
, logic_cc
);
8089 if (gen_thumb2_data_op(s
, op
, conds
, 0, tmp
, tmp2
))
8091 tcg_temp_free_i32(tmp2
);
8093 store_reg(s
, rd
, tmp
);
8095 tcg_temp_free_i32(tmp
);
8099 case 13: /* Misc data processing. */
8100 op
= ((insn
>> 22) & 6) | ((insn
>> 7) & 1);
8101 if (op
< 4 && (insn
& 0xf000) != 0xf000)
8104 case 0: /* Register controlled shift. */
8105 tmp
= load_reg(s
, rn
);
8106 tmp2
= load_reg(s
, rm
);
8107 if ((insn
& 0x70) != 0)
8109 op
= (insn
>> 21) & 3;
8110 logic_cc
= (insn
& (1 << 20)) != 0;
8111 gen_arm_shift_reg(tmp
, op
, tmp2
, logic_cc
);
8114 store_reg_bx(env
, s
, rd
, tmp
);
8116 case 1: /* Sign/zero extend. */
8117 tmp
= load_reg(s
, rm
);
8118 shift
= (insn
>> 4) & 3;
8119 /* ??? In many cases it's not neccessary to do a
8120 rotate, a shift is sufficient. */
8122 tcg_gen_rotri_i32(tmp
, tmp
, shift
* 8);
8123 op
= (insn
>> 20) & 7;
8125 case 0: gen_sxth(tmp
); break;
8126 case 1: gen_uxth(tmp
); break;
8127 case 2: gen_sxtb16(tmp
); break;
8128 case 3: gen_uxtb16(tmp
); break;
8129 case 4: gen_sxtb(tmp
); break;
8130 case 5: gen_uxtb(tmp
); break;
8131 default: goto illegal_op
;
8134 tmp2
= load_reg(s
, rn
);
8135 if ((op
>> 1) == 1) {
8136 gen_add16(tmp
, tmp2
);
8138 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8139 tcg_temp_free_i32(tmp2
);
8142 store_reg(s
, rd
, tmp
);
8144 case 2: /* SIMD add/subtract. */
8145 op
= (insn
>> 20) & 7;
8146 shift
= (insn
>> 4) & 7;
8147 if ((op
& 3) == 3 || (shift
& 3) == 3)
8149 tmp
= load_reg(s
, rn
);
8150 tmp2
= load_reg(s
, rm
);
8151 gen_thumb2_parallel_addsub(op
, shift
, tmp
, tmp2
);
8152 tcg_temp_free_i32(tmp2
);
8153 store_reg(s
, rd
, tmp
);
8155 case 3: /* Other data processing. */
8156 op
= ((insn
>> 17) & 0x38) | ((insn
>> 4) & 7);
8158 /* Saturating add/subtract. */
8159 tmp
= load_reg(s
, rn
);
8160 tmp2
= load_reg(s
, rm
);
8162 gen_helper_double_saturate(tmp
, tmp
);
8164 gen_helper_sub_saturate(tmp
, tmp2
, tmp
);
8166 gen_helper_add_saturate(tmp
, tmp
, tmp2
);
8167 tcg_temp_free_i32(tmp2
);
8169 tmp
= load_reg(s
, rn
);
8171 case 0x0a: /* rbit */
8172 gen_helper_rbit(tmp
, tmp
);
8174 case 0x08: /* rev */
8175 tcg_gen_bswap32_i32(tmp
, tmp
);
8177 case 0x09: /* rev16 */
8180 case 0x0b: /* revsh */
8183 case 0x10: /* sel */
8184 tmp2
= load_reg(s
, rm
);
8185 tmp3
= tcg_temp_new_i32();
8186 tcg_gen_ld_i32(tmp3
, cpu_env
, offsetof(CPUState
, GE
));
8187 gen_helper_sel_flags(tmp
, tmp3
, tmp
, tmp2
);
8188 tcg_temp_free_i32(tmp3
);
8189 tcg_temp_free_i32(tmp2
);
8191 case 0x18: /* clz */
8192 gen_helper_clz(tmp
, tmp
);
8198 store_reg(s
, rd
, tmp
);
8200 case 4: case 5: /* 32-bit multiply. Sum of absolute differences. */
8201 op
= (insn
>> 4) & 0xf;
8202 tmp
= load_reg(s
, rn
);
8203 tmp2
= load_reg(s
, rm
);
8204 switch ((insn
>> 20) & 7) {
8205 case 0: /* 32 x 32 -> 32 */
8206 tcg_gen_mul_i32(tmp
, tmp
, tmp2
);
8207 tcg_temp_free_i32(tmp2
);
8209 tmp2
= load_reg(s
, rs
);
8211 tcg_gen_sub_i32(tmp
, tmp2
, tmp
);
8213 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8214 tcg_temp_free_i32(tmp2
);
8217 case 1: /* 16 x 16 -> 32 */
8218 gen_mulxy(tmp
, tmp2
, op
& 2, op
& 1);
8219 tcg_temp_free_i32(tmp2
);
8221 tmp2
= load_reg(s
, rs
);
8222 gen_helper_add_setq(tmp
, tmp
, tmp2
);
8223 tcg_temp_free_i32(tmp2
);
8226 case 2: /* Dual multiply add. */
8227 case 4: /* Dual multiply subtract. */
8229 gen_swap_half(tmp2
);
8230 gen_smul_dual(tmp
, tmp2
);
8231 if (insn
& (1 << 22)) {
8232 /* This subtraction cannot overflow. */
8233 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
8235 /* This addition cannot overflow 32 bits;
8236 * however it may overflow considered as a signed
8237 * operation, in which case we must set the Q flag.
8239 gen_helper_add_setq(tmp
, tmp
, tmp2
);
8241 tcg_temp_free_i32(tmp2
);
8244 tmp2
= load_reg(s
, rs
);
8245 gen_helper_add_setq(tmp
, tmp
, tmp2
);
8246 tcg_temp_free_i32(tmp2
);
8249 case 3: /* 32 * 16 -> 32msb */
8251 tcg_gen_sari_i32(tmp2
, tmp2
, 16);
8254 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
8255 tcg_gen_shri_i64(tmp64
, tmp64
, 16);
8256 tmp
= tcg_temp_new_i32();
8257 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
8258 tcg_temp_free_i64(tmp64
);
8261 tmp2
= load_reg(s
, rs
);
8262 gen_helper_add_setq(tmp
, tmp
, tmp2
);
8263 tcg_temp_free_i32(tmp2
);
8266 case 5: case 6: /* 32 * 32 -> 32msb (SMMUL, SMMLA, SMMLS) */
8267 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
8269 tmp
= load_reg(s
, rs
);
8270 if (insn
& (1 << 20)) {
8271 tmp64
= gen_addq_msw(tmp64
, tmp
);
8273 tmp64
= gen_subq_msw(tmp64
, tmp
);
8276 if (insn
& (1 << 4)) {
8277 tcg_gen_addi_i64(tmp64
, tmp64
, 0x80000000u
);
8279 tcg_gen_shri_i64(tmp64
, tmp64
, 32);
8280 tmp
= tcg_temp_new_i32();
8281 tcg_gen_trunc_i64_i32(tmp
, tmp64
);
8282 tcg_temp_free_i64(tmp64
);
8284 case 7: /* Unsigned sum of absolute differences. */
8285 gen_helper_usad8(tmp
, tmp
, tmp2
);
8286 tcg_temp_free_i32(tmp2
);
8288 tmp2
= load_reg(s
, rs
);
8289 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8290 tcg_temp_free_i32(tmp2
);
8294 store_reg(s
, rd
, tmp
);
8296 case 6: case 7: /* 64-bit multiply, Divide. */
8297 op
= ((insn
>> 4) & 0xf) | ((insn
>> 16) & 0x70);
8298 tmp
= load_reg(s
, rn
);
8299 tmp2
= load_reg(s
, rm
);
8300 if ((op
& 0x50) == 0x10) {
8302 if (!arm_feature(env
, ARM_FEATURE_DIV
))
8305 gen_helper_udiv(tmp
, tmp
, tmp2
);
8307 gen_helper_sdiv(tmp
, tmp
, tmp2
);
8308 tcg_temp_free_i32(tmp2
);
8309 store_reg(s
, rd
, tmp
);
8310 } else if ((op
& 0xe) == 0xc) {
8311 /* Dual multiply accumulate long. */
8313 gen_swap_half(tmp2
);
8314 gen_smul_dual(tmp
, tmp2
);
8316 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
8318 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8320 tcg_temp_free_i32(tmp2
);
8322 tmp64
= tcg_temp_new_i64();
8323 tcg_gen_ext_i32_i64(tmp64
, tmp
);
8324 tcg_temp_free_i32(tmp
);
8325 gen_addq(s
, tmp64
, rs
, rd
);
8326 gen_storeq_reg(s
, rs
, rd
, tmp64
);
8327 tcg_temp_free_i64(tmp64
);
8330 /* Unsigned 64-bit multiply */
8331 tmp64
= gen_mulu_i64_i32(tmp
, tmp2
);
8335 gen_mulxy(tmp
, tmp2
, op
& 2, op
& 1);
8336 tcg_temp_free_i32(tmp2
);
8337 tmp64
= tcg_temp_new_i64();
8338 tcg_gen_ext_i32_i64(tmp64
, tmp
);
8339 tcg_temp_free_i32(tmp
);
8341 /* Signed 64-bit multiply */
8342 tmp64
= gen_muls_i64_i32(tmp
, tmp2
);
8347 gen_addq_lo(s
, tmp64
, rs
);
8348 gen_addq_lo(s
, tmp64
, rd
);
8349 } else if (op
& 0x40) {
8350 /* 64-bit accumulate. */
8351 gen_addq(s
, tmp64
, rs
, rd
);
8353 gen_storeq_reg(s
, rs
, rd
, tmp64
);
8354 tcg_temp_free_i64(tmp64
);
8359 case 6: case 7: case 14: case 15:
8361 if (((insn
>> 24) & 3) == 3) {
8362 /* Translate into the equivalent ARM encoding. */
8363 insn
= (insn
& 0xe2ffffff) | ((insn
& (1 << 28)) >> 4) | (1 << 28);
8364 if (disas_neon_data_insn(env
, s
, insn
))
8367 if (insn
& (1 << 28))
8369 if (disas_coproc_insn (env
, s
, insn
))
8373 case 8: case 9: case 10: case 11:
8374 if (insn
& (1 << 15)) {
8375 /* Branches, misc control. */
8376 if (insn
& 0x5000) {
8377 /* Unconditional branch. */
8378 /* signextend(hw1[10:0]) -> offset[:12]. */
8379 offset
= ((int32_t)insn
<< 5) >> 9 & ~(int32_t)0xfff;
8380 /* hw1[10:0] -> offset[11:1]. */
8381 offset
|= (insn
& 0x7ff) << 1;
8382 /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
8383 offset[24:22] already have the same value because of the
8384 sign extension above. */
8385 offset
^= ((~insn
) & (1 << 13)) << 10;
8386 offset
^= ((~insn
) & (1 << 11)) << 11;
8388 if (insn
& (1 << 14)) {
8389 /* Branch and link. */
8390 tcg_gen_movi_i32(cpu_R
[14], s
->pc
| 1);
8394 if (insn
& (1 << 12)) {
8399 offset
&= ~(uint32_t)2;
8400 /* thumb2 bx, no need to check */
8401 gen_bx_im(s
, offset
);
8403 } else if (((insn
>> 23) & 7) == 7) {
8405 if (insn
& (1 << 13))
8408 if (insn
& (1 << 26)) {
8409 /* Secure monitor call (v6Z) */
8410 goto illegal_op
; /* not implemented. */
8412 op
= (insn
>> 20) & 7;
8414 case 0: /* msr cpsr. */
8416 tmp
= load_reg(s
, rn
);
8417 addr
= tcg_const_i32(insn
& 0xff);
8418 gen_helper_v7m_msr(cpu_env
, addr
, tmp
);
8419 tcg_temp_free_i32(addr
);
8420 tcg_temp_free_i32(tmp
);
8425 case 1: /* msr spsr. */
8428 tmp
= load_reg(s
, rn
);
8430 msr_mask(env
, s
, (insn
>> 8) & 0xf, op
== 1),
8434 case 2: /* cps, nop-hint. */
8435 if (((insn
>> 8) & 7) == 0) {
8436 gen_nop_hint(s
, insn
& 0xff);
8438 /* Implemented as NOP in user mode. */
8443 if (insn
& (1 << 10)) {
8444 if (insn
& (1 << 7))
8446 if (insn
& (1 << 6))
8448 if (insn
& (1 << 5))
8450 if (insn
& (1 << 9))
8451 imm
= CPSR_A
| CPSR_I
| CPSR_F
;
8453 if (insn
& (1 << 8)) {
8455 imm
|= (insn
& 0x1f);
8458 gen_set_psr_im(s
, offset
, 0, imm
);
8461 case 3: /* Special control operations. */
8463 op
= (insn
>> 4) & 0xf;
8471 /* These execute as NOPs. */
8478 /* Trivial implementation equivalent to bx. */
8479 tmp
= load_reg(s
, rn
);
8482 case 5: /* Exception return. */
8486 if (rn
!= 14 || rd
!= 15) {
8489 tmp
= load_reg(s
, rn
);
8490 tcg_gen_subi_i32(tmp
, tmp
, insn
& 0xff);
8491 gen_exception_return(s
, tmp
);
8493 case 6: /* mrs cpsr. */
8494 tmp
= tcg_temp_new_i32();
8496 addr
= tcg_const_i32(insn
& 0xff);
8497 gen_helper_v7m_mrs(tmp
, cpu_env
, addr
);
8498 tcg_temp_free_i32(addr
);
8500 gen_helper_cpsr_read(tmp
);
8502 store_reg(s
, rd
, tmp
);
8504 case 7: /* mrs spsr. */
8505 /* Not accessible in user mode. */
8506 if (IS_USER(s
) || IS_M(env
))
8508 tmp
= load_cpu_field(spsr
);
8509 store_reg(s
, rd
, tmp
);
8514 /* Conditional branch. */
8515 op
= (insn
>> 22) & 0xf;
8516 /* Generate a conditional jump to next instruction. */
8517 s
->condlabel
= gen_new_label();
8518 gen_test_cc(op
^ 1, s
->condlabel
);
8521 /* offset[11:1] = insn[10:0] */
8522 offset
= (insn
& 0x7ff) << 1;
8523 /* offset[17:12] = insn[21:16]. */
8524 offset
|= (insn
& 0x003f0000) >> 4;
8525 /* offset[31:20] = insn[26]. */
8526 offset
|= ((int32_t)((insn
<< 5) & 0x80000000)) >> 11;
8527 /* offset[18] = insn[13]. */
8528 offset
|= (insn
& (1 << 13)) << 5;
8529 /* offset[19] = insn[11]. */
8530 offset
|= (insn
& (1 << 11)) << 8;
8532 /* jump to the offset */
8533 gen_jmp(s
, s
->pc
+ offset
);
8536 /* Data processing immediate. */
8537 if (insn
& (1 << 25)) {
8538 if (insn
& (1 << 24)) {
8539 if (insn
& (1 << 20))
8541 /* Bitfield/Saturate. */
8542 op
= (insn
>> 21) & 7;
8544 shift
= ((insn
>> 6) & 3) | ((insn
>> 10) & 0x1c);
8546 tmp
= tcg_temp_new_i32();
8547 tcg_gen_movi_i32(tmp
, 0);
8549 tmp
= load_reg(s
, rn
);
8552 case 2: /* Signed bitfield extract. */
8554 if (shift
+ imm
> 32)
8557 gen_sbfx(tmp
, shift
, imm
);
8559 case 6: /* Unsigned bitfield extract. */
8561 if (shift
+ imm
> 32)
8564 gen_ubfx(tmp
, shift
, (1u << imm
) - 1);
8566 case 3: /* Bitfield insert/clear. */
8569 imm
= imm
+ 1 - shift
;
8571 tmp2
= load_reg(s
, rd
);
8572 gen_bfi(tmp
, tmp2
, tmp
, shift
, (1u << imm
) - 1);
8573 tcg_temp_free_i32(tmp2
);
8578 default: /* Saturate. */
8581 tcg_gen_sari_i32(tmp
, tmp
, shift
);
8583 tcg_gen_shli_i32(tmp
, tmp
, shift
);
8585 tmp2
= tcg_const_i32(imm
);
8588 if ((op
& 1) && shift
== 0)
8589 gen_helper_usat16(tmp
, tmp
, tmp2
);
8591 gen_helper_usat(tmp
, tmp
, tmp2
);
8594 if ((op
& 1) && shift
== 0)
8595 gen_helper_ssat16(tmp
, tmp
, tmp2
);
8597 gen_helper_ssat(tmp
, tmp
, tmp2
);
8599 tcg_temp_free_i32(tmp2
);
8602 store_reg(s
, rd
, tmp
);
8604 imm
= ((insn
& 0x04000000) >> 15)
8605 | ((insn
& 0x7000) >> 4) | (insn
& 0xff);
8606 if (insn
& (1 << 22)) {
8607 /* 16-bit immediate. */
8608 imm
|= (insn
>> 4) & 0xf000;
8609 if (insn
& (1 << 23)) {
8611 tmp
= load_reg(s
, rd
);
8612 tcg_gen_ext16u_i32(tmp
, tmp
);
8613 tcg_gen_ori_i32(tmp
, tmp
, imm
<< 16);
8616 tmp
= tcg_temp_new_i32();
8617 tcg_gen_movi_i32(tmp
, imm
);
8620 /* Add/sub 12-bit immediate. */
8622 offset
= s
->pc
& ~(uint32_t)3;
8623 if (insn
& (1 << 23))
8627 tmp
= tcg_temp_new_i32();
8628 tcg_gen_movi_i32(tmp
, offset
);
8630 tmp
= load_reg(s
, rn
);
8631 if (insn
& (1 << 23))
8632 tcg_gen_subi_i32(tmp
, tmp
, imm
);
8634 tcg_gen_addi_i32(tmp
, tmp
, imm
);
8637 store_reg(s
, rd
, tmp
);
8640 int shifter_out
= 0;
8641 /* modified 12-bit immediate. */
8642 shift
= ((insn
& 0x04000000) >> 23) | ((insn
& 0x7000) >> 12);
8643 imm
= (insn
& 0xff);
8646 /* Nothing to do. */
8648 case 1: /* 00XY00XY */
8651 case 2: /* XY00XY00 */
8655 case 3: /* XYXYXYXY */
8659 default: /* Rotated constant. */
8660 shift
= (shift
<< 1) | (imm
>> 7);
8662 imm
= imm
<< (32 - shift
);
8666 tmp2
= tcg_temp_new_i32();
8667 tcg_gen_movi_i32(tmp2
, imm
);
8668 rn
= (insn
>> 16) & 0xf;
8670 tmp
= tcg_temp_new_i32();
8671 tcg_gen_movi_i32(tmp
, 0);
8673 tmp
= load_reg(s
, rn
);
8675 op
= (insn
>> 21) & 0xf;
8676 if (gen_thumb2_data_op(s
, op
, (insn
& (1 << 20)) != 0,
8677 shifter_out
, tmp
, tmp2
))
8679 tcg_temp_free_i32(tmp2
);
8680 rd
= (insn
>> 8) & 0xf;
8682 store_reg(s
, rd
, tmp
);
8684 tcg_temp_free_i32(tmp
);
8689 case 12: /* Load/store single data item. */
8694 if ((insn
& 0x01100000) == 0x01000000) {
8695 if (disas_neon_ls_insn(env
, s
, insn
))
8699 op
= ((insn
>> 21) & 3) | ((insn
>> 22) & 4);
8701 if (!(insn
& (1 << 20))) {
8705 /* Byte or halfword load space with dest == r15 : memory hints.
8706 * Catch them early so we don't emit pointless addressing code.
8707 * This space is a mix of:
8708 * PLD/PLDW/PLI, which we implement as NOPs (note that unlike
8709 * the ARM encodings, PLDW space doesn't UNDEF for non-v7MP
8711 * unallocated hints, which must be treated as NOPs
8712 * UNPREDICTABLE space, which we NOP or UNDEF depending on
8713 * which is easiest for the decoding logic
8714 * Some space which must UNDEF
8716 int op1
= (insn
>> 23) & 3;
8717 int op2
= (insn
>> 6) & 0x3f;
8722 /* UNPREDICTABLE or unallocated hint */
8726 return 0; /* PLD* or unallocated hint */
8728 if ((op2
== 0) || ((op2
& 0x3c) == 0x30)) {
8729 return 0; /* PLD* or unallocated hint */
8731 /* UNDEF space, or an UNPREDICTABLE */
8737 addr
= tcg_temp_new_i32();
8739 /* s->pc has already been incremented by 4. */
8740 imm
= s
->pc
& 0xfffffffc;
8741 if (insn
& (1 << 23))
8742 imm
+= insn
& 0xfff;
8744 imm
-= insn
& 0xfff;
8745 tcg_gen_movi_i32(addr
, imm
);
8747 addr
= load_reg(s
, rn
);
8748 if (insn
& (1 << 23)) {
8749 /* Positive offset. */
8751 tcg_gen_addi_i32(addr
, addr
, imm
);
8754 switch ((insn
>> 8) & 0xf) {
8755 case 0x0: /* Shifted Register. */
8756 shift
= (insn
>> 4) & 0xf;
8758 tcg_temp_free_i32(addr
);
8761 tmp
= load_reg(s
, rm
);
8763 tcg_gen_shli_i32(tmp
, tmp
, shift
);
8764 tcg_gen_add_i32(addr
, addr
, tmp
);
8765 tcg_temp_free_i32(tmp
);
8767 case 0xc: /* Negative offset. */
8768 tcg_gen_addi_i32(addr
, addr
, -imm
);
8770 case 0xe: /* User privilege. */
8771 tcg_gen_addi_i32(addr
, addr
, imm
);
8774 case 0x9: /* Post-decrement. */
8777 case 0xb: /* Post-increment. */
8781 case 0xd: /* Pre-decrement. */
8784 case 0xf: /* Pre-increment. */
8785 tcg_gen_addi_i32(addr
, addr
, imm
);
8789 tcg_temp_free_i32(addr
);
8794 if (insn
& (1 << 20)) {
8797 case 0: tmp
= gen_ld8u(addr
, user
); break;
8798 case 4: tmp
= gen_ld8s(addr
, user
); break;
8799 case 1: tmp
= gen_ld16u(addr
, user
); break;
8800 case 5: tmp
= gen_ld16s(addr
, user
); break;
8801 case 2: tmp
= gen_ld32(addr
, user
); break;
8803 tcg_temp_free_i32(addr
);
8809 store_reg(s
, rs
, tmp
);
8813 tmp
= load_reg(s
, rs
);
8815 case 0: gen_st8(tmp
, addr
, user
); break;
8816 case 1: gen_st16(tmp
, addr
, user
); break;
8817 case 2: gen_st32(tmp
, addr
, user
); break;
8819 tcg_temp_free_i32(addr
);
8824 tcg_gen_addi_i32(addr
, addr
, imm
);
8826 store_reg(s
, rn
, addr
);
8828 tcg_temp_free_i32(addr
);
8840 static void disas_thumb_insn(CPUState
*env
, DisasContext
*s
)
8842 uint32_t val
, insn
, op
, rm
, rn
, rd
, shift
, cond
;
8849 if (s
->condexec_mask
) {
8850 cond
= s
->condexec_cond
;
8851 if (cond
!= 0x0e) { /* Skip conditional when condition is AL. */
8852 s
->condlabel
= gen_new_label();
8853 gen_test_cc(cond
^ 1, s
->condlabel
);
8858 insn
= lduw_code(s
->pc
);
8861 switch (insn
>> 12) {
8865 op
= (insn
>> 11) & 3;
8868 rn
= (insn
>> 3) & 7;
8869 tmp
= load_reg(s
, rn
);
8870 if (insn
& (1 << 10)) {
8872 tmp2
= tcg_temp_new_i32();
8873 tcg_gen_movi_i32(tmp2
, (insn
>> 6) & 7);
8876 rm
= (insn
>> 6) & 7;
8877 tmp2
= load_reg(s
, rm
);
8879 if (insn
& (1 << 9)) {
8880 if (s
->condexec_mask
)
8881 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
8883 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8885 if (s
->condexec_mask
)
8886 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8888 gen_helper_add_cc(tmp
, tmp
, tmp2
);
8890 tcg_temp_free_i32(tmp2
);
8891 store_reg(s
, rd
, tmp
);
8893 /* shift immediate */
8894 rm
= (insn
>> 3) & 7;
8895 shift
= (insn
>> 6) & 0x1f;
8896 tmp
= load_reg(s
, rm
);
8897 gen_arm_shift_im(tmp
, op
, shift
, s
->condexec_mask
== 0);
8898 if (!s
->condexec_mask
)
8900 store_reg(s
, rd
, tmp
);
8904 /* arithmetic large immediate */
8905 op
= (insn
>> 11) & 3;
8906 rd
= (insn
>> 8) & 0x7;
8907 if (op
== 0) { /* mov */
8908 tmp
= tcg_temp_new_i32();
8909 tcg_gen_movi_i32(tmp
, insn
& 0xff);
8910 if (!s
->condexec_mask
)
8912 store_reg(s
, rd
, tmp
);
8914 tmp
= load_reg(s
, rd
);
8915 tmp2
= tcg_temp_new_i32();
8916 tcg_gen_movi_i32(tmp2
, insn
& 0xff);
8919 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8920 tcg_temp_free_i32(tmp
);
8921 tcg_temp_free_i32(tmp2
);
8924 if (s
->condexec_mask
)
8925 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8927 gen_helper_add_cc(tmp
, tmp
, tmp2
);
8928 tcg_temp_free_i32(tmp2
);
8929 store_reg(s
, rd
, tmp
);
8932 if (s
->condexec_mask
)
8933 tcg_gen_sub_i32(tmp
, tmp
, tmp2
);
8935 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8936 tcg_temp_free_i32(tmp2
);
8937 store_reg(s
, rd
, tmp
);
8943 if (insn
& (1 << 11)) {
8944 rd
= (insn
>> 8) & 7;
8945 /* load pc-relative. Bit 1 of PC is ignored. */
8946 val
= s
->pc
+ 2 + ((insn
& 0xff) * 4);
8947 val
&= ~(uint32_t)2;
8948 addr
= tcg_temp_new_i32();
8949 tcg_gen_movi_i32(addr
, val
);
8950 tmp
= gen_ld32(addr
, IS_USER(s
));
8951 tcg_temp_free_i32(addr
);
8952 store_reg(s
, rd
, tmp
);
8955 if (insn
& (1 << 10)) {
8956 /* data processing extended or blx */
8957 rd
= (insn
& 7) | ((insn
>> 4) & 8);
8958 rm
= (insn
>> 3) & 0xf;
8959 op
= (insn
>> 8) & 3;
8962 tmp
= load_reg(s
, rd
);
8963 tmp2
= load_reg(s
, rm
);
8964 tcg_gen_add_i32(tmp
, tmp
, tmp2
);
8965 tcg_temp_free_i32(tmp2
);
8966 store_reg(s
, rd
, tmp
);
8969 tmp
= load_reg(s
, rd
);
8970 tmp2
= load_reg(s
, rm
);
8971 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
8972 tcg_temp_free_i32(tmp2
);
8973 tcg_temp_free_i32(tmp
);
8975 case 2: /* mov/cpy */
8976 tmp
= load_reg(s
, rm
);
8977 store_reg(s
, rd
, tmp
);
8979 case 3:/* branch [and link] exchange thumb register */
8980 tmp
= load_reg(s
, rm
);
8981 if (insn
& (1 << 7)) {
8983 val
= (uint32_t)s
->pc
| 1;
8984 tmp2
= tcg_temp_new_i32();
8985 tcg_gen_movi_i32(tmp2
, val
);
8986 store_reg(s
, 14, tmp2
);
8988 /* already thumb, no need to check */
8995 /* data processing register */
8997 rm
= (insn
>> 3) & 7;
8998 op
= (insn
>> 6) & 0xf;
8999 if (op
== 2 || op
== 3 || op
== 4 || op
== 7) {
9000 /* the shift/rotate ops want the operands backwards */
9009 if (op
== 9) { /* neg */
9010 tmp
= tcg_temp_new_i32();
9011 tcg_gen_movi_i32(tmp
, 0);
9012 } else if (op
!= 0xf) { /* mvn doesn't read its first operand */
9013 tmp
= load_reg(s
, rd
);
9018 tmp2
= load_reg(s
, rm
);
9021 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
9022 if (!s
->condexec_mask
)
9026 tcg_gen_xor_i32(tmp
, tmp
, tmp2
);
9027 if (!s
->condexec_mask
)
9031 if (s
->condexec_mask
) {
9032 gen_helper_shl(tmp2
, tmp2
, tmp
);
9034 gen_helper_shl_cc(tmp2
, tmp2
, tmp
);
9039 if (s
->condexec_mask
) {
9040 gen_helper_shr(tmp2
, tmp2
, tmp
);
9042 gen_helper_shr_cc(tmp2
, tmp2
, tmp
);
9047 if (s
->condexec_mask
) {
9048 gen_helper_sar(tmp2
, tmp2
, tmp
);
9050 gen_helper_sar_cc(tmp2
, tmp2
, tmp
);
9055 if (s
->condexec_mask
)
9058 gen_helper_adc_cc(tmp
, tmp
, tmp2
);
9061 if (s
->condexec_mask
)
9062 gen_sub_carry(tmp
, tmp
, tmp2
);
9064 gen_helper_sbc_cc(tmp
, tmp
, tmp2
);
9067 if (s
->condexec_mask
) {
9068 tcg_gen_andi_i32(tmp
, tmp
, 0x1f);
9069 tcg_gen_rotr_i32(tmp2
, tmp2
, tmp
);
9071 gen_helper_ror_cc(tmp2
, tmp2
, tmp
);
9076 tcg_gen_and_i32(tmp
, tmp
, tmp2
);
9081 if (s
->condexec_mask
)
9082 tcg_gen_neg_i32(tmp
, tmp2
);
9084 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
9087 gen_helper_sub_cc(tmp
, tmp
, tmp2
);
9091 gen_helper_add_cc(tmp
, tmp
, tmp2
);
9095 tcg_gen_or_i32(tmp
, tmp
, tmp2
);
9096 if (!s
->condexec_mask
)
9100 tcg_gen_mul_i32(tmp
, tmp
, tmp2
);
9101 if (!s
->condexec_mask
)
9105 tcg_gen_andc_i32(tmp
, tmp
, tmp2
);
9106 if (!s
->condexec_mask
)
9110 tcg_gen_not_i32(tmp2
, tmp2
);
9111 if (!s
->condexec_mask
)
9119 store_reg(s
, rm
, tmp2
);
9121 tcg_temp_free_i32(tmp
);
9123 store_reg(s
, rd
, tmp
);
9124 tcg_temp_free_i32(tmp2
);
9127 tcg_temp_free_i32(tmp
);
9128 tcg_temp_free_i32(tmp2
);
9133 /* load/store register offset. */
9135 rn
= (insn
>> 3) & 7;
9136 rm
= (insn
>> 6) & 7;
9137 op
= (insn
>> 9) & 7;
9138 addr
= load_reg(s
, rn
);
9139 tmp
= load_reg(s
, rm
);
9140 tcg_gen_add_i32(addr
, addr
, tmp
);
9141 tcg_temp_free_i32(tmp
);
9143 if (op
< 3) /* store */
9144 tmp
= load_reg(s
, rd
);
9148 gen_st32(tmp
, addr
, IS_USER(s
));
9151 gen_st16(tmp
, addr
, IS_USER(s
));
9154 gen_st8(tmp
, addr
, IS_USER(s
));
9157 tmp
= gen_ld8s(addr
, IS_USER(s
));
9160 tmp
= gen_ld32(addr
, IS_USER(s
));
9163 tmp
= gen_ld16u(addr
, IS_USER(s
));
9166 tmp
= gen_ld8u(addr
, IS_USER(s
));
9169 tmp
= gen_ld16s(addr
, IS_USER(s
));
9172 if (op
>= 3) /* load */
9173 store_reg(s
, rd
, tmp
);
9174 tcg_temp_free_i32(addr
);
9178 /* load/store word immediate offset */
9180 rn
= (insn
>> 3) & 7;
9181 addr
= load_reg(s
, rn
);
9182 val
= (insn
>> 4) & 0x7c;
9183 tcg_gen_addi_i32(addr
, addr
, val
);
9185 if (insn
& (1 << 11)) {
9187 tmp
= gen_ld32(addr
, IS_USER(s
));
9188 store_reg(s
, rd
, tmp
);
9191 tmp
= load_reg(s
, rd
);
9192 gen_st32(tmp
, addr
, IS_USER(s
));
9194 tcg_temp_free_i32(addr
);
9198 /* load/store byte immediate offset */
9200 rn
= (insn
>> 3) & 7;
9201 addr
= load_reg(s
, rn
);
9202 val
= (insn
>> 6) & 0x1f;
9203 tcg_gen_addi_i32(addr
, addr
, val
);
9205 if (insn
& (1 << 11)) {
9207 tmp
= gen_ld8u(addr
, IS_USER(s
));
9208 store_reg(s
, rd
, tmp
);
9211 tmp
= load_reg(s
, rd
);
9212 gen_st8(tmp
, addr
, IS_USER(s
));
9214 tcg_temp_free_i32(addr
);
9218 /* load/store halfword immediate offset */
9220 rn
= (insn
>> 3) & 7;
9221 addr
= load_reg(s
, rn
);
9222 val
= (insn
>> 5) & 0x3e;
9223 tcg_gen_addi_i32(addr
, addr
, val
);
9225 if (insn
& (1 << 11)) {
9227 tmp
= gen_ld16u(addr
, IS_USER(s
));
9228 store_reg(s
, rd
, tmp
);
9231 tmp
= load_reg(s
, rd
);
9232 gen_st16(tmp
, addr
, IS_USER(s
));
9234 tcg_temp_free_i32(addr
);
9238 /* load/store from stack */
9239 rd
= (insn
>> 8) & 7;
9240 addr
= load_reg(s
, 13);
9241 val
= (insn
& 0xff) * 4;
9242 tcg_gen_addi_i32(addr
, addr
, val
);
9244 if (insn
& (1 << 11)) {
9246 tmp
= gen_ld32(addr
, IS_USER(s
));
9247 store_reg(s
, rd
, tmp
);
9250 tmp
= load_reg(s
, rd
);
9251 gen_st32(tmp
, addr
, IS_USER(s
));
9253 tcg_temp_free_i32(addr
);
9257 /* add to high reg */
9258 rd
= (insn
>> 8) & 7;
9259 if (insn
& (1 << 11)) {
9261 tmp
= load_reg(s
, 13);
9263 /* PC. bit 1 is ignored. */
9264 tmp
= tcg_temp_new_i32();
9265 tcg_gen_movi_i32(tmp
, (s
->pc
+ 2) & ~(uint32_t)2);
9267 val
= (insn
& 0xff) * 4;
9268 tcg_gen_addi_i32(tmp
, tmp
, val
);
9269 store_reg(s
, rd
, tmp
);
9274 op
= (insn
>> 8) & 0xf;
9277 /* adjust stack pointer */
9278 tmp
= load_reg(s
, 13);
9279 val
= (insn
& 0x7f) * 4;
9280 if (insn
& (1 << 7))
9281 val
= -(int32_t)val
;
9282 tcg_gen_addi_i32(tmp
, tmp
, val
);
9283 store_reg(s
, 13, tmp
);
9286 case 2: /* sign/zero extend. */
9289 rm
= (insn
>> 3) & 7;
9290 tmp
= load_reg(s
, rm
);
9291 switch ((insn
>> 6) & 3) {
9292 case 0: gen_sxth(tmp
); break;
9293 case 1: gen_sxtb(tmp
); break;
9294 case 2: gen_uxth(tmp
); break;
9295 case 3: gen_uxtb(tmp
); break;
9297 store_reg(s
, rd
, tmp
);
9299 case 4: case 5: case 0xc: case 0xd:
9301 addr
= load_reg(s
, 13);
9302 if (insn
& (1 << 8))
9306 for (i
= 0; i
< 8; i
++) {
9307 if (insn
& (1 << i
))
9310 if ((insn
& (1 << 11)) == 0) {
9311 tcg_gen_addi_i32(addr
, addr
, -offset
);
9313 for (i
= 0; i
< 8; i
++) {
9314 if (insn
& (1 << i
)) {
9315 if (insn
& (1 << 11)) {
9317 tmp
= gen_ld32(addr
, IS_USER(s
));
9318 store_reg(s
, i
, tmp
);
9321 tmp
= load_reg(s
, i
);
9322 gen_st32(tmp
, addr
, IS_USER(s
));
9324 /* advance to the next address. */
9325 tcg_gen_addi_i32(addr
, addr
, 4);
9329 if (insn
& (1 << 8)) {
9330 if (insn
& (1 << 11)) {
9332 tmp
= gen_ld32(addr
, IS_USER(s
));
9333 /* don't set the pc until the rest of the instruction
9337 tmp
= load_reg(s
, 14);
9338 gen_st32(tmp
, addr
, IS_USER(s
));
9340 tcg_gen_addi_i32(addr
, addr
, 4);
9342 if ((insn
& (1 << 11)) == 0) {
9343 tcg_gen_addi_i32(addr
, addr
, -offset
);
9345 /* write back the new stack pointer */
9346 store_reg(s
, 13, addr
);
9347 /* set the new PC value */
9348 if ((insn
& 0x0900) == 0x0900) {
9349 store_reg_from_load(env
, s
, 15, tmp
);
9353 case 1: case 3: case 9: case 11: /* czb */
9355 tmp
= load_reg(s
, rm
);
9356 s
->condlabel
= gen_new_label();
9358 if (insn
& (1 << 11))
9359 tcg_gen_brcondi_i32(TCG_COND_EQ
, tmp
, 0, s
->condlabel
);
9361 tcg_gen_brcondi_i32(TCG_COND_NE
, tmp
, 0, s
->condlabel
);
9362 tcg_temp_free_i32(tmp
);
9363 offset
= ((insn
& 0xf8) >> 2) | (insn
& 0x200) >> 3;
9364 val
= (uint32_t)s
->pc
+ 2;
9369 case 15: /* IT, nop-hint. */
9370 if ((insn
& 0xf) == 0) {
9371 gen_nop_hint(s
, (insn
>> 4) & 0xf);
9375 s
->condexec_cond
= (insn
>> 4) & 0xe;
9376 s
->condexec_mask
= insn
& 0x1f;
9377 /* No actual code generated for this insn, just setup state. */
9380 case 0xe: /* bkpt */
9382 gen_exception_insn(s
, 2, EXCP_BKPT
);
9387 rn
= (insn
>> 3) & 0x7;
9389 tmp
= load_reg(s
, rn
);
9390 switch ((insn
>> 6) & 3) {
9391 case 0: tcg_gen_bswap32_i32(tmp
, tmp
); break;
9392 case 1: gen_rev16(tmp
); break;
9393 case 3: gen_revsh(tmp
); break;
9394 default: goto illegal_op
;
9396 store_reg(s
, rd
, tmp
);
9404 tmp
= tcg_const_i32((insn
& (1 << 4)) != 0);
9407 addr
= tcg_const_i32(16);
9408 gen_helper_v7m_msr(cpu_env
, addr
, tmp
);
9409 tcg_temp_free_i32(addr
);
9413 addr
= tcg_const_i32(17);
9414 gen_helper_v7m_msr(cpu_env
, addr
, tmp
);
9415 tcg_temp_free_i32(addr
);
9417 tcg_temp_free_i32(tmp
);
9420 if (insn
& (1 << 4))
9421 shift
= CPSR_A
| CPSR_I
| CPSR_F
;
9424 gen_set_psr_im(s
, ((insn
& 7) << 6), 0, shift
);
9434 /* load/store multiple */
9435 rn
= (insn
>> 8) & 0x7;
9436 addr
= load_reg(s
, rn
);
9437 for (i
= 0; i
< 8; i
++) {
9438 if (insn
& (1 << i
)) {
9439 if (insn
& (1 << 11)) {
9441 tmp
= gen_ld32(addr
, IS_USER(s
));
9442 store_reg(s
, i
, tmp
);
9445 tmp
= load_reg(s
, i
);
9446 gen_st32(tmp
, addr
, IS_USER(s
));
9448 /* advance to the next address */
9449 tcg_gen_addi_i32(addr
, addr
, 4);
9452 /* Base register writeback. */
9453 if ((insn
& (1 << rn
)) == 0) {
9454 store_reg(s
, rn
, addr
);
9456 tcg_temp_free_i32(addr
);
9461 /* conditional branch or swi */
9462 cond
= (insn
>> 8) & 0xf;
9468 gen_set_pc_im(s
->pc
);
9469 s
->is_jmp
= DISAS_SWI
;
9472 /* generate a conditional jump to next instruction */
9473 s
->condlabel
= gen_new_label();
9474 gen_test_cc(cond
^ 1, s
->condlabel
);
9477 /* jump to the offset */
9478 val
= (uint32_t)s
->pc
+ 2;
9479 offset
= ((int32_t)insn
<< 24) >> 24;
9485 if (insn
& (1 << 11)) {
9486 if (disas_thumb2_insn(env
, s
, insn
))
9490 /* unconditional branch */
9491 val
= (uint32_t)s
->pc
;
9492 offset
= ((int32_t)insn
<< 21) >> 21;
9493 val
+= (offset
<< 1) + 2;
9498 if (disas_thumb2_insn(env
, s
, insn
))
9504 gen_exception_insn(s
, 4, EXCP_UDEF
);
9508 gen_exception_insn(s
, 2, EXCP_UDEF
);
9511 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
9512 basic block 'tb'. If search_pc is TRUE, also generate PC
9513 information for each intermediate instruction. */
9514 static inline void gen_intermediate_code_internal(CPUState
*env
,
9515 TranslationBlock
*tb
,
9518 DisasContext dc1
, *dc
= &dc1
;
9520 uint16_t *gen_opc_end
;
9522 target_ulong pc_start
;
9523 uint32_t next_page_start
;
9527 /* generate intermediate code */
9532 gen_opc_end
= gen_opc_buf
+ OPC_MAX_SIZE
;
9534 dc
->is_jmp
= DISAS_NEXT
;
9536 dc
->singlestep_enabled
= env
->singlestep_enabled
;
9538 dc
->thumb
= ARM_TBFLAG_THUMB(tb
->flags
);
9539 dc
->condexec_mask
= (ARM_TBFLAG_CONDEXEC(tb
->flags
) & 0xf) << 1;
9540 dc
->condexec_cond
= ARM_TBFLAG_CONDEXEC(tb
->flags
) >> 4;
9541 #if !defined(CONFIG_USER_ONLY)
9542 dc
->user
= (ARM_TBFLAG_PRIV(tb
->flags
) == 0);
9544 dc
->vfp_enabled
= ARM_TBFLAG_VFPEN(tb
->flags
);
9545 dc
->vec_len
= ARM_TBFLAG_VECLEN(tb
->flags
);
9546 dc
->vec_stride
= ARM_TBFLAG_VECSTRIDE(tb
->flags
);
9547 cpu_F0s
= tcg_temp_new_i32();
9548 cpu_F1s
= tcg_temp_new_i32();
9549 cpu_F0d
= tcg_temp_new_i64();
9550 cpu_F1d
= tcg_temp_new_i64();
9553 /* FIXME: cpu_M0 can probably be the same as cpu_V0. */
9554 cpu_M0
= tcg_temp_new_i64();
9555 next_page_start
= (pc_start
& TARGET_PAGE_MASK
) + TARGET_PAGE_SIZE
;
9558 max_insns
= tb
->cflags
& CF_COUNT_MASK
;
9560 max_insns
= CF_COUNT_MASK
;
9564 tcg_clear_temp_count();
9566 /* A note on handling of the condexec (IT) bits:
9568 * We want to avoid the overhead of having to write the updated condexec
9569 * bits back to the CPUState for every instruction in an IT block. So:
9570 * (1) if the condexec bits are not already zero then we write
9571 * zero back into the CPUState now. This avoids complications trying
9572 * to do it at the end of the block. (For example if we don't do this
9573 * it's hard to identify whether we can safely skip writing condexec
9574 * at the end of the TB, which we definitely want to do for the case
9575 * where a TB doesn't do anything with the IT state at all.)
9576 * (2) if we are going to leave the TB then we call gen_set_condexec()
9577 * which will write the correct value into CPUState if zero is wrong.
9578 * This is done both for leaving the TB at the end, and for leaving
9579 * it because of an exception we know will happen, which is done in
9580 * gen_exception_insn(). The latter is necessary because we need to
9581 * leave the TB with the PC/IT state just prior to execution of the
9582 * instruction which caused the exception.
9583 * (3) if we leave the TB unexpectedly (eg a data abort on a load)
9584 * then the CPUState will be wrong and we need to reset it.
9585 * This is handled in the same way as restoration of the
9586 * PC in these situations: we will be called again with search_pc=1
9587 * and generate a mapping of the condexec bits for each PC in
9588 * gen_opc_condexec_bits[]. restore_state_to_opc() then uses
9589 * this to restore the condexec bits.
9591 * Note that there are no instructions which can read the condexec
9592 * bits, and none which can write non-static values to them, so
9593 * we don't need to care about whether CPUState is correct in the
9597 /* Reset the conditional execution bits immediately. This avoids
9598 complications trying to do it at the end of the block. */
9599 if (dc
->condexec_mask
|| dc
->condexec_cond
)
9601 TCGv tmp
= tcg_temp_new_i32();
9602 tcg_gen_movi_i32(tmp
, 0);
9603 store_cpu_field(tmp
, condexec_bits
);
9606 #ifdef CONFIG_USER_ONLY
9607 /* Intercept jump to the magic kernel page. */
9608 if (dc
->pc
>= 0xffff0000) {
9609 /* We always get here via a jump, so know we are not in a
9610 conditional execution block. */
9611 gen_exception(EXCP_KERNEL_TRAP
);
9612 dc
->is_jmp
= DISAS_UPDATE
;
9616 if (dc
->pc
>= 0xfffffff0 && IS_M(env
)) {
9617 /* We always get here via a jump, so know we are not in a
9618 conditional execution block. */
9619 gen_exception(EXCP_EXCEPTION_EXIT
);
9620 dc
->is_jmp
= DISAS_UPDATE
;
9625 if (unlikely(!QTAILQ_EMPTY(&env
->breakpoints
))) {
9626 QTAILQ_FOREACH(bp
, &env
->breakpoints
, entry
) {
9627 if (bp
->pc
== dc
->pc
) {
9628 gen_exception_insn(dc
, 0, EXCP_DEBUG
);
9629 /* Advance PC so that clearing the breakpoint will
9630 invalidate this TB. */
9632 goto done_generating
;
9638 j
= gen_opc_ptr
- gen_opc_buf
;
9642 gen_opc_instr_start
[lj
++] = 0;
9644 gen_opc_pc
[lj
] = dc
->pc
;
9645 gen_opc_condexec_bits
[lj
] = (dc
->condexec_cond
<< 4) | (dc
->condexec_mask
>> 1);
9646 gen_opc_instr_start
[lj
] = 1;
9647 gen_opc_icount
[lj
] = num_insns
;
9650 if (num_insns
+ 1 == max_insns
&& (tb
->cflags
& CF_LAST_IO
))
9653 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP
))) {
9654 tcg_gen_debug_insn_start(dc
->pc
);
9658 disas_thumb_insn(env
, dc
);
9659 if (dc
->condexec_mask
) {
9660 dc
->condexec_cond
= (dc
->condexec_cond
& 0xe)
9661 | ((dc
->condexec_mask
>> 4) & 1);
9662 dc
->condexec_mask
= (dc
->condexec_mask
<< 1) & 0x1f;
9663 if (dc
->condexec_mask
== 0) {
9664 dc
->condexec_cond
= 0;
9668 disas_arm_insn(env
, dc
);
9671 if (dc
->condjmp
&& !dc
->is_jmp
) {
9672 gen_set_label(dc
->condlabel
);
9676 if (tcg_check_temp_count()) {
9677 fprintf(stderr
, "TCG temporary leak before %08x\n", dc
->pc
);
9680 /* Translation stops when a conditional branch is encountered.
9681 * Otherwise the subsequent code could get translated several times.
9682 * Also stop translation when a page boundary is reached. This
9683 * ensures prefetch aborts occur at the right place. */
9685 } while (!dc
->is_jmp
&& gen_opc_ptr
< gen_opc_end
&&
9686 !env
->singlestep_enabled
&&
9688 dc
->pc
< next_page_start
&&
9689 num_insns
< max_insns
);
9691 if (tb
->cflags
& CF_LAST_IO
) {
9693 /* FIXME: This can theoretically happen with self-modifying
9695 cpu_abort(env
, "IO on conditional branch instruction");
9700 /* At this stage dc->condjmp will only be set when the skipped
9701 instruction was a conditional branch or trap, and the PC has
9702 already been written. */
9703 if (unlikely(env
->singlestep_enabled
)) {
9704 /* Make sure the pc is updated, and raise a debug exception. */
9706 gen_set_condexec(dc
);
9707 if (dc
->is_jmp
== DISAS_SWI
) {
9708 gen_exception(EXCP_SWI
);
9710 gen_exception(EXCP_DEBUG
);
9712 gen_set_label(dc
->condlabel
);
9714 if (dc
->condjmp
|| !dc
->is_jmp
) {
9715 gen_set_pc_im(dc
->pc
);
9718 gen_set_condexec(dc
);
9719 if (dc
->is_jmp
== DISAS_SWI
&& !dc
->condjmp
) {
9720 gen_exception(EXCP_SWI
);
9722 /* FIXME: Single stepping a WFI insn will not halt
9724 gen_exception(EXCP_DEBUG
);
9727 /* While branches must always occur at the end of an IT block,
9728 there are a few other things that can cause us to terminate
9729 the TB in the middel of an IT block:
9730 - Exception generating instructions (bkpt, swi, undefined).
9732 - Hardware watchpoints.
9733 Hardware breakpoints have already been handled and skip this code.
9735 gen_set_condexec(dc
);
9736 switch(dc
->is_jmp
) {
9738 gen_goto_tb(dc
, 1, dc
->pc
);
9743 /* indicate that the hash table must be used to find the next TB */
9747 /* nothing more to generate */
9753 gen_exception(EXCP_SWI
);
9757 gen_set_label(dc
->condlabel
);
9758 gen_set_condexec(dc
);
9759 gen_goto_tb(dc
, 1, dc
->pc
);
9765 gen_icount_end(tb
, num_insns
);
9766 *gen_opc_ptr
= INDEX_op_end
;
9769 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM
)) {
9770 qemu_log("----------------\n");
9771 qemu_log("IN: %s\n", lookup_symbol(pc_start
));
9772 log_target_disas(pc_start
, dc
->pc
- pc_start
, dc
->thumb
);
9777 j
= gen_opc_ptr
- gen_opc_buf
;
9780 gen_opc_instr_start
[lj
++] = 0;
9782 tb
->size
= dc
->pc
- pc_start
;
9783 tb
->icount
= num_insns
;
9787 void gen_intermediate_code(CPUState
*env
, TranslationBlock
*tb
)
9789 gen_intermediate_code_internal(env
, tb
, 0);
9792 void gen_intermediate_code_pc(CPUState
*env
, TranslationBlock
*tb
)
9794 gen_intermediate_code_internal(env
, tb
, 1);
9797 static const char *cpu_mode_names
[16] = {
9798 "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
9799 "???", "???", "???", "und", "???", "???", "???", "sys"
9802 void cpu_dump_state(CPUState
*env
, FILE *f
, fprintf_function cpu_fprintf
,
9812 /* ??? This assumes float64 and double have the same layout.
9813 Oh well, it's only debug dumps. */
9822 cpu_fprintf(f
, "R%02d=%08x", i
, env
->regs
[i
]);
9824 cpu_fprintf(f
, "\n");
9826 cpu_fprintf(f
, " ");
9828 psr
= cpsr_read(env
);
9829 cpu_fprintf(f
, "PSR=%08x %c%c%c%c %c %s%d\n",
9831 psr
& (1 << 31) ? 'N' : '-',
9832 psr
& (1 << 30) ? 'Z' : '-',
9833 psr
& (1 << 29) ? 'C' : '-',
9834 psr
& (1 << 28) ? 'V' : '-',
9835 psr
& CPSR_T
? 'T' : 'A',
9836 cpu_mode_names
[psr
& 0xf], (psr
& 0x10) ? 32 : 26);
9839 for (i
= 0; i
< 16; i
++) {
9840 d
.d
= env
->vfp
.regs
[i
];
9844 cpu_fprintf(f
, "s%02d=%08x(%8g) s%02d=%08x(%8g) d%02d=%08x%08x(%8g)\n",
9845 i
* 2, (int)s0
.i
, s0
.s
,
9846 i
* 2 + 1, (int)s1
.i
, s1
.s
,
9847 i
, (int)(uint32_t)d
.l
.upper
, (int)(uint32_t)d
.l
.lower
,
9850 cpu_fprintf(f
, "FPSCR: %08x\n", (int)env
->vfp
.xregs
[ARM_VFP_FPSCR
]);
9854 void restore_state_to_opc(CPUState
*env
, TranslationBlock
*tb
, int pc_pos
)
9856 env
->regs
[15] = gen_opc_pc
[pc_pos
];
9857 env
->condexec_bits
= gen_opc_condexec_bits
[pc_pos
];