4 * Copyright (c) 2003 Fabrice Bellard
5 * Copyright (c) 2005-2007 CodeSourcery
6 * Copyright (c) 2007 OpenedHand, Ltd.
8 * This library is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2 of the License, or (at your option) any later version.
13 * This library is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
37 #define ENABLE_ARCH_5J 0
38 #define ENABLE_ARCH_6 arm_feature(env, ARM_FEATURE_V6)
39 #define ENABLE_ARCH_6K arm_feature(env, ARM_FEATURE_V6K)
40 #define ENABLE_ARCH_6T2 arm_feature(env, ARM_FEATURE_THUMB2)
41 #define ENABLE_ARCH_7 arm_feature(env, ARM_FEATURE_V7)
43 #define ARCH(x) do { if (!ENABLE_ARCH_##x) goto illegal_op; } while(0)
45 /* internal defines */
46 typedef struct DisasContext {
49 /* Nonzero if this instruction has been conditionally skipped. */
51 /* The label that will be jumped to when the instruction is skipped. */
53 /* Thumb-2 condtional execution bits. */
56 struct TranslationBlock *tb;
57 int singlestep_enabled;
59 #if !defined(CONFIG_USER_ONLY)
67 static uint32_t gen_opc_condexec_bits[OPC_BUF_SIZE];
69 #if defined(CONFIG_USER_ONLY)
72 #define IS_USER(s) (s->user)
75 /* These instructions trap after executing, so defer them until after the
76 conditional executions state has been updated. */
80 static TCGv_ptr cpu_env;
81 /* We reuse the same 64-bit temporaries for efficiency. */
82 static TCGv_i64 cpu_V0, cpu_V1, cpu_M0;
83 static TCGv_i32 cpu_R[16];
84 static TCGv_i32 cpu_exclusive_addr;
85 static TCGv_i32 cpu_exclusive_val;
86 static TCGv_i32 cpu_exclusive_high;
87 #ifdef CONFIG_USER_ONLY
88 static TCGv_i32 cpu_exclusive_test;
89 static TCGv_i32 cpu_exclusive_info;
92 /* FIXME: These should be removed. */
93 static TCGv cpu_F0s, cpu_F1s;
94 static TCGv_i64 cpu_F0d, cpu_F1d;
96 #include "gen-icount.h"
98 static const char *regnames[] =
99 { "r0", "r1", "r2", "r3", "r4", "r5", "r6", "r7",
100 "r8", "r9", "r10", "r11", "r12", "r13", "r14", "pc" };
102 /* initialize TCG globals. */
103 void arm_translate_init(void)
107 cpu_env = tcg_global_reg_new_ptr(TCG_AREG0, "env");
109 for (i = 0; i < 16; i++) {
110 cpu_R[i] = tcg_global_mem_new_i32(TCG_AREG0,
111 offsetof(CPUState, regs[i]),
114 cpu_exclusive_addr = tcg_global_mem_new_i32(TCG_AREG0,
115 offsetof(CPUState, exclusive_addr), "exclusive_addr");
116 cpu_exclusive_val = tcg_global_mem_new_i32(TCG_AREG0,
117 offsetof(CPUState, exclusive_val), "exclusive_val");
118 cpu_exclusive_high = tcg_global_mem_new_i32(TCG_AREG0,
119 offsetof(CPUState, exclusive_high), "exclusive_high");
120 #ifdef CONFIG_USER_ONLY
121 cpu_exclusive_test = tcg_global_mem_new_i32(TCG_AREG0,
122 offsetof(CPUState, exclusive_test), "exclusive_test");
123 cpu_exclusive_info = tcg_global_mem_new_i32(TCG_AREG0,
124 offsetof(CPUState, exclusive_info), "exclusive_info");
131 static int num_temps;
133 /* Allocate a temporary variable. */
134 static TCGv_i32 new_tmp(void)
137 return tcg_temp_new_i32();
140 /* Release a temporary variable. */
141 static void dead_tmp(TCGv tmp)
147 static inline TCGv load_cpu_offset(int offset)
149 TCGv tmp = new_tmp();
150 tcg_gen_ld_i32(tmp, cpu_env, offset);
154 #define load_cpu_field(name) load_cpu_offset(offsetof(CPUState, name))
156 static inline void store_cpu_offset(TCGv var, int offset)
158 tcg_gen_st_i32(var, cpu_env, offset);
162 #define store_cpu_field(var, name) \
163 store_cpu_offset(var, offsetof(CPUState, name))
165 /* Set a variable to the value of a CPU register. */
166 static void load_reg_var(DisasContext *s, TCGv var, int reg)
170 /* normaly, since we updated PC, we need only to add one insn */
172 addr = (long)s->pc + 2;
174 addr = (long)s->pc + 4;
175 tcg_gen_movi_i32(var, addr);
177 tcg_gen_mov_i32(var, cpu_R[reg]);
181 /* Create a new temporary and set it to the value of a CPU register. */
182 static inline TCGv load_reg(DisasContext *s, int reg)
184 TCGv tmp = new_tmp();
185 load_reg_var(s, tmp, reg);
189 /* Set a CPU register. The source must be a temporary and will be
191 static void store_reg(DisasContext *s, int reg, TCGv var)
194 tcg_gen_andi_i32(var, var, ~1);
195 s->is_jmp = DISAS_JUMP;
197 tcg_gen_mov_i32(cpu_R[reg], var);
201 /* Value extensions. */
202 #define gen_uxtb(var) tcg_gen_ext8u_i32(var, var)
203 #define gen_uxth(var) tcg_gen_ext16u_i32(var, var)
204 #define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
205 #define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
207 #define gen_sxtb16(var) gen_helper_sxtb16(var, var)
208 #define gen_uxtb16(var) gen_helper_uxtb16(var, var)
211 static inline void gen_set_cpsr(TCGv var, uint32_t mask)
213 TCGv tmp_mask = tcg_const_i32(mask);
214 gen_helper_cpsr_write(var, tmp_mask);
215 tcg_temp_free_i32(tmp_mask);
217 /* Set NZCV flags from the high 4 bits of var. */
218 #define gen_set_nzcv(var) gen_set_cpsr(var, CPSR_NZCV)
220 static void gen_exception(int excp)
222 TCGv tmp = new_tmp();
223 tcg_gen_movi_i32(tmp, excp);
224 gen_helper_exception(tmp);
228 static void gen_smul_dual(TCGv a, TCGv b)
230 TCGv tmp1 = new_tmp();
231 TCGv tmp2 = new_tmp();
232 tcg_gen_ext16s_i32(tmp1, a);
233 tcg_gen_ext16s_i32(tmp2, b);
234 tcg_gen_mul_i32(tmp1, tmp1, tmp2);
236 tcg_gen_sari_i32(a, a, 16);
237 tcg_gen_sari_i32(b, b, 16);
238 tcg_gen_mul_i32(b, b, a);
239 tcg_gen_mov_i32(a, tmp1);
243 /* Byteswap each halfword. */
244 static void gen_rev16(TCGv var)
246 TCGv tmp = new_tmp();
247 tcg_gen_shri_i32(tmp, var, 8);
248 tcg_gen_andi_i32(tmp, tmp, 0x00ff00ff);
249 tcg_gen_shli_i32(var, var, 8);
250 tcg_gen_andi_i32(var, var, 0xff00ff00);
251 tcg_gen_or_i32(var, var, tmp);
255 /* Byteswap low halfword and sign extend. */
256 static void gen_revsh(TCGv var)
258 tcg_gen_ext16u_i32(var, var);
259 tcg_gen_bswap16_i32(var, var);
260 tcg_gen_ext16s_i32(var, var);
263 /* Unsigned bitfield extract. */
264 static void gen_ubfx(TCGv var, int shift, uint32_t mask)
267 tcg_gen_shri_i32(var, var, shift);
268 tcg_gen_andi_i32(var, var, mask);
271 /* Signed bitfield extract. */
272 static void gen_sbfx(TCGv var, int shift, int width)
277 tcg_gen_sari_i32(var, var, shift);
278 if (shift + width < 32) {
279 signbit = 1u << (width - 1);
280 tcg_gen_andi_i32(var, var, (1u << width) - 1);
281 tcg_gen_xori_i32(var, var, signbit);
282 tcg_gen_subi_i32(var, var, signbit);
286 /* Bitfield insertion. Insert val into base. Clobbers base and val. */
287 static void gen_bfi(TCGv dest, TCGv base, TCGv val, int shift, uint32_t mask)
289 tcg_gen_andi_i32(val, val, mask);
290 tcg_gen_shli_i32(val, val, shift);
291 tcg_gen_andi_i32(base, base, ~(mask << shift));
292 tcg_gen_or_i32(dest, base, val);
295 /* Return (b << 32) + a. Mark inputs as dead */
296 static TCGv_i64 gen_addq_msw(TCGv_i64 a, TCGv b)
298 TCGv_i64 tmp64 = tcg_temp_new_i64();
300 tcg_gen_extu_i32_i64(tmp64, b);
302 tcg_gen_shli_i64(tmp64, tmp64, 32);
303 tcg_gen_add_i64(a, tmp64, a);
305 tcg_temp_free_i64(tmp64);
309 /* Return (b << 32) - a. Mark inputs as dead. */
310 static TCGv_i64 gen_subq_msw(TCGv_i64 a, TCGv b)
312 TCGv_i64 tmp64 = tcg_temp_new_i64();
314 tcg_gen_extu_i32_i64(tmp64, b);
316 tcg_gen_shli_i64(tmp64, tmp64, 32);
317 tcg_gen_sub_i64(a, tmp64, a);
319 tcg_temp_free_i64(tmp64);
323 /* FIXME: Most targets have native widening multiplication.
324 It would be good to use that instead of a full wide multiply. */
325 /* 32x32->64 multiply. Marks inputs as dead. */
326 static TCGv_i64 gen_mulu_i64_i32(TCGv a, TCGv b)
328 TCGv_i64 tmp1 = tcg_temp_new_i64();
329 TCGv_i64 tmp2 = tcg_temp_new_i64();
331 tcg_gen_extu_i32_i64(tmp1, a);
333 tcg_gen_extu_i32_i64(tmp2, b);
335 tcg_gen_mul_i64(tmp1, tmp1, tmp2);
336 tcg_temp_free_i64(tmp2);
340 static TCGv_i64 gen_muls_i64_i32(TCGv a, TCGv b)
342 TCGv_i64 tmp1 = tcg_temp_new_i64();
343 TCGv_i64 tmp2 = tcg_temp_new_i64();
345 tcg_gen_ext_i32_i64(tmp1, a);
347 tcg_gen_ext_i32_i64(tmp2, b);
349 tcg_gen_mul_i64(tmp1, tmp1, tmp2);
350 tcg_temp_free_i64(tmp2);
354 /* Swap low and high halfwords. */
355 static void gen_swap_half(TCGv var)
357 TCGv tmp = new_tmp();
358 tcg_gen_shri_i32(tmp, var, 16);
359 tcg_gen_shli_i32(var, var, 16);
360 tcg_gen_or_i32(var, var, tmp);
364 /* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
365 tmp = (t0 ^ t1) & 0x8000;
368 t0 = (t0 + t1) ^ tmp;
371 static void gen_add16(TCGv t0, TCGv t1)
373 TCGv tmp = new_tmp();
374 tcg_gen_xor_i32(tmp, t0, t1);
375 tcg_gen_andi_i32(tmp, tmp, 0x8000);
376 tcg_gen_andi_i32(t0, t0, ~0x8000);
377 tcg_gen_andi_i32(t1, t1, ~0x8000);
378 tcg_gen_add_i32(t0, t0, t1);
379 tcg_gen_xor_i32(t0, t0, tmp);
384 #define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, CF))
386 /* Set CF to the top bit of var. */
387 static void gen_set_CF_bit31(TCGv var)
389 TCGv tmp = new_tmp();
390 tcg_gen_shri_i32(tmp, var, 31);
395 /* Set N and Z flags from var. */
396 static inline void gen_logic_CC(TCGv var)
398 tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, NF));
399 tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, ZF));
403 static void gen_adc(TCGv t0, TCGv t1)
406 tcg_gen_add_i32(t0, t0, t1);
407 tmp = load_cpu_field(CF);
408 tcg_gen_add_i32(t0, t0, tmp);
412 /* dest = T0 + T1 + CF. */
413 static void gen_add_carry(TCGv dest, TCGv t0, TCGv t1)
416 tcg_gen_add_i32(dest, t0, t1);
417 tmp = load_cpu_field(CF);
418 tcg_gen_add_i32(dest, dest, tmp);
422 /* dest = T0 - T1 + CF - 1. */
423 static void gen_sub_carry(TCGv dest, TCGv t0, TCGv t1)
426 tcg_gen_sub_i32(dest, t0, t1);
427 tmp = load_cpu_field(CF);
428 tcg_gen_add_i32(dest, dest, tmp);
429 tcg_gen_subi_i32(dest, dest, 1);
433 /* FIXME: Implement this natively. */
434 #define tcg_gen_abs_i32(t0, t1) gen_helper_abs(t0, t1)
436 static void shifter_out_im(TCGv var, int shift)
438 TCGv tmp = new_tmp();
440 tcg_gen_andi_i32(tmp, var, 1);
442 tcg_gen_shri_i32(tmp, var, shift);
444 tcg_gen_andi_i32(tmp, tmp, 1);
450 /* Shift by immediate. Includes special handling for shift == 0. */
451 static inline void gen_arm_shift_im(TCGv var, int shiftop, int shift, int flags)
457 shifter_out_im(var, 32 - shift);
458 tcg_gen_shli_i32(var, var, shift);
464 tcg_gen_shri_i32(var, var, 31);
467 tcg_gen_movi_i32(var, 0);
470 shifter_out_im(var, shift - 1);
471 tcg_gen_shri_i32(var, var, shift);
478 shifter_out_im(var, shift - 1);
481 tcg_gen_sari_i32(var, var, shift);
483 case 3: /* ROR/RRX */
486 shifter_out_im(var, shift - 1);
487 tcg_gen_rotri_i32(var, var, shift); break;
489 TCGv tmp = load_cpu_field(CF);
491 shifter_out_im(var, 0);
492 tcg_gen_shri_i32(var, var, 1);
493 tcg_gen_shli_i32(tmp, tmp, 31);
494 tcg_gen_or_i32(var, var, tmp);
500 static inline void gen_arm_shift_reg(TCGv var, int shiftop,
501 TCGv shift, int flags)
505 case 0: gen_helper_shl_cc(var, var, shift); break;
506 case 1: gen_helper_shr_cc(var, var, shift); break;
507 case 2: gen_helper_sar_cc(var, var, shift); break;
508 case 3: gen_helper_ror_cc(var, var, shift); break;
512 case 0: gen_helper_shl(var, var, shift); break;
513 case 1: gen_helper_shr(var, var, shift); break;
514 case 2: gen_helper_sar(var, var, shift); break;
515 case 3: tcg_gen_andi_i32(shift, shift, 0x1f);
516 tcg_gen_rotr_i32(var, var, shift); break;
522 #define PAS_OP(pfx) \
524 case 0: gen_pas_helper(glue(pfx,add16)); break; \
525 case 1: gen_pas_helper(glue(pfx,addsubx)); break; \
526 case 2: gen_pas_helper(glue(pfx,subaddx)); break; \
527 case 3: gen_pas_helper(glue(pfx,sub16)); break; \
528 case 4: gen_pas_helper(glue(pfx,add8)); break; \
529 case 7: gen_pas_helper(glue(pfx,sub8)); break; \
531 static void gen_arm_parallel_addsub(int op1, int op2, TCGv a, TCGv b)
536 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
538 tmp = tcg_temp_new_ptr();
539 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
541 tcg_temp_free_ptr(tmp);
544 tmp = tcg_temp_new_ptr();
545 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
547 tcg_temp_free_ptr(tmp);
549 #undef gen_pas_helper
550 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
563 #undef gen_pas_helper
568 /* For unknown reasons Arm and Thumb-2 use arbitrarily different encodings. */
569 #define PAS_OP(pfx) \
571 case 0: gen_pas_helper(glue(pfx,add8)); break; \
572 case 1: gen_pas_helper(glue(pfx,add16)); break; \
573 case 2: gen_pas_helper(glue(pfx,addsubx)); break; \
574 case 4: gen_pas_helper(glue(pfx,sub8)); break; \
575 case 5: gen_pas_helper(glue(pfx,sub16)); break; \
576 case 6: gen_pas_helper(glue(pfx,subaddx)); break; \
578 static void gen_thumb2_parallel_addsub(int op1, int op2, TCGv a, TCGv b)
583 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b, tmp)
585 tmp = tcg_temp_new_ptr();
586 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
588 tcg_temp_free_ptr(tmp);
591 tmp = tcg_temp_new_ptr();
592 tcg_gen_addi_ptr(tmp, cpu_env, offsetof(CPUState, GE));
594 tcg_temp_free_ptr(tmp);
596 #undef gen_pas_helper
597 #define gen_pas_helper(name) glue(gen_helper_,name)(a, a, b)
610 #undef gen_pas_helper
615 static void gen_test_cc(int cc, int label)
623 tmp = load_cpu_field(ZF);
624 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
627 tmp = load_cpu_field(ZF);
628 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, label);
631 tmp = load_cpu_field(CF);
632 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, label);
635 tmp = load_cpu_field(CF);
636 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
639 tmp = load_cpu_field(NF);
640 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
643 tmp = load_cpu_field(NF);
644 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
647 tmp = load_cpu_field(VF);
648 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
651 tmp = load_cpu_field(VF);
652 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
654 case 8: /* hi: C && !Z */
655 inv = gen_new_label();
656 tmp = load_cpu_field(CF);
657 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, inv);
659 tmp = load_cpu_field(ZF);
660 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, label);
663 case 9: /* ls: !C || Z */
664 tmp = load_cpu_field(CF);
665 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
667 tmp = load_cpu_field(ZF);
668 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
670 case 10: /* ge: N == V -> N ^ V == 0 */
671 tmp = load_cpu_field(VF);
672 tmp2 = load_cpu_field(NF);
673 tcg_gen_xor_i32(tmp, tmp, tmp2);
675 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
677 case 11: /* lt: N != V -> N ^ V != 0 */
678 tmp = load_cpu_field(VF);
679 tmp2 = load_cpu_field(NF);
680 tcg_gen_xor_i32(tmp, tmp, tmp2);
682 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
684 case 12: /* gt: !Z && N == V */
685 inv = gen_new_label();
686 tmp = load_cpu_field(ZF);
687 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, inv);
689 tmp = load_cpu_field(VF);
690 tmp2 = load_cpu_field(NF);
691 tcg_gen_xor_i32(tmp, tmp, tmp2);
693 tcg_gen_brcondi_i32(TCG_COND_GE, tmp, 0, label);
696 case 13: /* le: Z || N != V */
697 tmp = load_cpu_field(ZF);
698 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, label);
700 tmp = load_cpu_field(VF);
701 tmp2 = load_cpu_field(NF);
702 tcg_gen_xor_i32(tmp, tmp, tmp2);
704 tcg_gen_brcondi_i32(TCG_COND_LT, tmp, 0, label);
707 fprintf(stderr, "Bad condition code 0x%x\n", cc);
713 static const uint8_t table_logic_cc[16] = {
732 /* Set PC and Thumb state from an immediate address. */
733 static inline void gen_bx_im(DisasContext *s, uint32_t addr)
737 s->is_jmp = DISAS_UPDATE;
738 if (s->thumb != (addr & 1)) {
740 tcg_gen_movi_i32(tmp, addr & 1);
741 tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, thumb));
744 tcg_gen_movi_i32(cpu_R[15], addr & ~1);
747 /* Set PC and Thumb state from var. var is marked as dead. */
748 static inline void gen_bx(DisasContext *s, TCGv var)
750 s->is_jmp = DISAS_UPDATE;
751 tcg_gen_andi_i32(cpu_R[15], var, ~1);
752 tcg_gen_andi_i32(var, var, 1);
753 store_cpu_field(var, thumb);
756 /* Variant of store_reg which uses branch&exchange logic when storing
757 to r15 in ARM architecture v7 and above. The source must be a temporary
758 and will be marked as dead. */
759 static inline void store_reg_bx(CPUState *env, DisasContext *s,
762 if (reg == 15 && ENABLE_ARCH_7) {
765 store_reg(s, reg, var);
769 static inline TCGv gen_ld8s(TCGv addr, int index)
771 TCGv tmp = new_tmp();
772 tcg_gen_qemu_ld8s(tmp, addr, index);
775 static inline TCGv gen_ld8u(TCGv addr, int index)
777 TCGv tmp = new_tmp();
778 tcg_gen_qemu_ld8u(tmp, addr, index);
781 static inline TCGv gen_ld16s(TCGv addr, int index)
783 TCGv tmp = new_tmp();
784 tcg_gen_qemu_ld16s(tmp, addr, index);
787 static inline TCGv gen_ld16u(TCGv addr, int index)
789 TCGv tmp = new_tmp();
790 tcg_gen_qemu_ld16u(tmp, addr, index);
793 static inline TCGv gen_ld32(TCGv addr, int index)
795 TCGv tmp = new_tmp();
796 tcg_gen_qemu_ld32u(tmp, addr, index);
799 static inline TCGv_i64 gen_ld64(TCGv addr, int index)
801 TCGv_i64 tmp = tcg_temp_new_i64();
802 tcg_gen_qemu_ld64(tmp, addr, index);
805 static inline void gen_st8(TCGv val, TCGv addr, int index)
807 tcg_gen_qemu_st8(val, addr, index);
810 static inline void gen_st16(TCGv val, TCGv addr, int index)
812 tcg_gen_qemu_st16(val, addr, index);
815 static inline void gen_st32(TCGv val, TCGv addr, int index)
817 tcg_gen_qemu_st32(val, addr, index);
820 static inline void gen_st64(TCGv_i64 val, TCGv addr, int index)
822 tcg_gen_qemu_st64(val, addr, index);
823 tcg_temp_free_i64(val);
826 static inline void gen_set_pc_im(uint32_t val)
828 tcg_gen_movi_i32(cpu_R[15], val);
831 /* Force a TB lookup after an instruction that changes the CPU state. */
832 static inline void gen_lookup_tb(DisasContext *s)
834 tcg_gen_movi_i32(cpu_R[15], s->pc & ~1);
835 s->is_jmp = DISAS_UPDATE;
838 static inline void gen_add_data_offset(DisasContext *s, unsigned int insn,
841 int val, rm, shift, shiftop;
844 if (!(insn & (1 << 25))) {
847 if (!(insn & (1 << 23)))
850 tcg_gen_addi_i32(var, var, val);
854 shift = (insn >> 7) & 0x1f;
855 shiftop = (insn >> 5) & 3;
856 offset = load_reg(s, rm);
857 gen_arm_shift_im(offset, shiftop, shift, 0);
858 if (!(insn & (1 << 23)))
859 tcg_gen_sub_i32(var, var, offset);
861 tcg_gen_add_i32(var, var, offset);
866 static inline void gen_add_datah_offset(DisasContext *s, unsigned int insn,
872 if (insn & (1 << 22)) {
874 val = (insn & 0xf) | ((insn >> 4) & 0xf0);
875 if (!(insn & (1 << 23)))
879 tcg_gen_addi_i32(var, var, val);
883 tcg_gen_addi_i32(var, var, extra);
885 offset = load_reg(s, rm);
886 if (!(insn & (1 << 23)))
887 tcg_gen_sub_i32(var, var, offset);
889 tcg_gen_add_i32(var, var, offset);
894 #define VFP_OP2(name) \
895 static inline void gen_vfp_##name(int dp) \
898 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, cpu_F1d, cpu_env); \
900 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, cpu_F1s, cpu_env); \
910 static inline void gen_vfp_abs(int dp)
913 gen_helper_vfp_absd(cpu_F0d, cpu_F0d);
915 gen_helper_vfp_abss(cpu_F0s, cpu_F0s);
918 static inline void gen_vfp_neg(int dp)
921 gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
923 gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
926 static inline void gen_vfp_sqrt(int dp)
929 gen_helper_vfp_sqrtd(cpu_F0d, cpu_F0d, cpu_env);
931 gen_helper_vfp_sqrts(cpu_F0s, cpu_F0s, cpu_env);
934 static inline void gen_vfp_cmp(int dp)
937 gen_helper_vfp_cmpd(cpu_F0d, cpu_F1d, cpu_env);
939 gen_helper_vfp_cmps(cpu_F0s, cpu_F1s, cpu_env);
942 static inline void gen_vfp_cmpe(int dp)
945 gen_helper_vfp_cmped(cpu_F0d, cpu_F1d, cpu_env);
947 gen_helper_vfp_cmpes(cpu_F0s, cpu_F1s, cpu_env);
950 static inline void gen_vfp_F1_ld0(int dp)
953 tcg_gen_movi_i64(cpu_F1d, 0);
955 tcg_gen_movi_i32(cpu_F1s, 0);
958 static inline void gen_vfp_uito(int dp)
961 gen_helper_vfp_uitod(cpu_F0d, cpu_F0s, cpu_env);
963 gen_helper_vfp_uitos(cpu_F0s, cpu_F0s, cpu_env);
966 static inline void gen_vfp_sito(int dp)
969 gen_helper_vfp_sitod(cpu_F0d, cpu_F0s, cpu_env);
971 gen_helper_vfp_sitos(cpu_F0s, cpu_F0s, cpu_env);
974 static inline void gen_vfp_toui(int dp)
977 gen_helper_vfp_touid(cpu_F0s, cpu_F0d, cpu_env);
979 gen_helper_vfp_touis(cpu_F0s, cpu_F0s, cpu_env);
982 static inline void gen_vfp_touiz(int dp)
985 gen_helper_vfp_touizd(cpu_F0s, cpu_F0d, cpu_env);
987 gen_helper_vfp_touizs(cpu_F0s, cpu_F0s, cpu_env);
990 static inline void gen_vfp_tosi(int dp)
993 gen_helper_vfp_tosid(cpu_F0s, cpu_F0d, cpu_env);
995 gen_helper_vfp_tosis(cpu_F0s, cpu_F0s, cpu_env);
998 static inline void gen_vfp_tosiz(int dp)
1001 gen_helper_vfp_tosizd(cpu_F0s, cpu_F0d, cpu_env);
1003 gen_helper_vfp_tosizs(cpu_F0s, cpu_F0s, cpu_env);
1006 #define VFP_GEN_FIX(name) \
1007 static inline void gen_vfp_##name(int dp, int shift) \
1009 TCGv tmp_shift = tcg_const_i32(shift); \
1011 gen_helper_vfp_##name##d(cpu_F0d, cpu_F0d, tmp_shift, cpu_env);\
1013 gen_helper_vfp_##name##s(cpu_F0s, cpu_F0s, tmp_shift, cpu_env);\
1014 tcg_temp_free_i32(tmp_shift); \
1026 static inline void gen_vfp_ld(DisasContext *s, int dp, TCGv addr)
1029 tcg_gen_qemu_ld64(cpu_F0d, addr, IS_USER(s));
1031 tcg_gen_qemu_ld32u(cpu_F0s, addr, IS_USER(s));
1034 static inline void gen_vfp_st(DisasContext *s, int dp, TCGv addr)
1037 tcg_gen_qemu_st64(cpu_F0d, addr, IS_USER(s));
1039 tcg_gen_qemu_st32(cpu_F0s, addr, IS_USER(s));
1043 vfp_reg_offset (int dp, int reg)
1046 return offsetof(CPUARMState, vfp.regs[reg]);
1048 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1049 + offsetof(CPU_DoubleU, l.upper);
1051 return offsetof(CPUARMState, vfp.regs[reg >> 1])
1052 + offsetof(CPU_DoubleU, l.lower);
1056 /* Return the offset of a 32-bit piece of a NEON register.
1057 zero is the least significant end of the register. */
1059 neon_reg_offset (int reg, int n)
1063 return vfp_reg_offset(0, sreg);
1066 static TCGv neon_load_reg(int reg, int pass)
1068 TCGv tmp = new_tmp();
1069 tcg_gen_ld_i32(tmp, cpu_env, neon_reg_offset(reg, pass));
1073 static void neon_store_reg(int reg, int pass, TCGv var)
1075 tcg_gen_st_i32(var, cpu_env, neon_reg_offset(reg, pass));
1079 static inline void neon_load_reg64(TCGv_i64 var, int reg)
1081 tcg_gen_ld_i64(var, cpu_env, vfp_reg_offset(1, reg));
1084 static inline void neon_store_reg64(TCGv_i64 var, int reg)
1086 tcg_gen_st_i64(var, cpu_env, vfp_reg_offset(1, reg));
1089 #define tcg_gen_ld_f32 tcg_gen_ld_i32
1090 #define tcg_gen_ld_f64 tcg_gen_ld_i64
1091 #define tcg_gen_st_f32 tcg_gen_st_i32
1092 #define tcg_gen_st_f64 tcg_gen_st_i64
1094 static inline void gen_mov_F0_vreg(int dp, int reg)
1097 tcg_gen_ld_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1099 tcg_gen_ld_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1102 static inline void gen_mov_F1_vreg(int dp, int reg)
1105 tcg_gen_ld_f64(cpu_F1d, cpu_env, vfp_reg_offset(dp, reg));
1107 tcg_gen_ld_f32(cpu_F1s, cpu_env, vfp_reg_offset(dp, reg));
1110 static inline void gen_mov_vreg_F0(int dp, int reg)
1113 tcg_gen_st_f64(cpu_F0d, cpu_env, vfp_reg_offset(dp, reg));
1115 tcg_gen_st_f32(cpu_F0s, cpu_env, vfp_reg_offset(dp, reg));
1118 #define ARM_CP_RW_BIT (1 << 20)
1120 static inline void iwmmxt_load_reg(TCGv_i64 var, int reg)
1122 tcg_gen_ld_i64(var, cpu_env, offsetof(CPUState, iwmmxt.regs[reg]));
1125 static inline void iwmmxt_store_reg(TCGv_i64 var, int reg)
1127 tcg_gen_st_i64(var, cpu_env, offsetof(CPUState, iwmmxt.regs[reg]));
1130 static inline TCGv iwmmxt_load_creg(int reg)
1132 TCGv var = new_tmp();
1133 tcg_gen_ld_i32(var, cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
1137 static inline void iwmmxt_store_creg(int reg, TCGv var)
1139 tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, iwmmxt.cregs[reg]));
1143 static inline void gen_op_iwmmxt_movq_wRn_M0(int rn)
1145 iwmmxt_store_reg(cpu_M0, rn);
1148 static inline void gen_op_iwmmxt_movq_M0_wRn(int rn)
1150 iwmmxt_load_reg(cpu_M0, rn);
1153 static inline void gen_op_iwmmxt_orq_M0_wRn(int rn)
1155 iwmmxt_load_reg(cpu_V1, rn);
1156 tcg_gen_or_i64(cpu_M0, cpu_M0, cpu_V1);
1159 static inline void gen_op_iwmmxt_andq_M0_wRn(int rn)
1161 iwmmxt_load_reg(cpu_V1, rn);
1162 tcg_gen_and_i64(cpu_M0, cpu_M0, cpu_V1);
1165 static inline void gen_op_iwmmxt_xorq_M0_wRn(int rn)
1167 iwmmxt_load_reg(cpu_V1, rn);
1168 tcg_gen_xor_i64(cpu_M0, cpu_M0, cpu_V1);
1171 #define IWMMXT_OP(name) \
1172 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1174 iwmmxt_load_reg(cpu_V1, rn); \
1175 gen_helper_iwmmxt_##name(cpu_M0, cpu_M0, cpu_V1); \
1178 #define IWMMXT_OP_ENV(name) \
1179 static inline void gen_op_iwmmxt_##name##_M0_wRn(int rn) \
1181 iwmmxt_load_reg(cpu_V1, rn); \
1182 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0, cpu_V1); \
1185 #define IWMMXT_OP_ENV_SIZE(name) \
1186 IWMMXT_OP_ENV(name##b) \
1187 IWMMXT_OP_ENV(name##w) \
1188 IWMMXT_OP_ENV(name##l)
1190 #define IWMMXT_OP_ENV1(name) \
1191 static inline void gen_op_iwmmxt_##name##_M0(void) \
1193 gen_helper_iwmmxt_##name(cpu_M0, cpu_env, cpu_M0); \
1207 IWMMXT_OP_ENV_SIZE(unpackl)
1208 IWMMXT_OP_ENV_SIZE(unpackh)
1210 IWMMXT_OP_ENV1(unpacklub)
1211 IWMMXT_OP_ENV1(unpackluw)
1212 IWMMXT_OP_ENV1(unpacklul)
1213 IWMMXT_OP_ENV1(unpackhub)
1214 IWMMXT_OP_ENV1(unpackhuw)
1215 IWMMXT_OP_ENV1(unpackhul)
1216 IWMMXT_OP_ENV1(unpacklsb)
1217 IWMMXT_OP_ENV1(unpacklsw)
1218 IWMMXT_OP_ENV1(unpacklsl)
1219 IWMMXT_OP_ENV1(unpackhsb)
1220 IWMMXT_OP_ENV1(unpackhsw)
1221 IWMMXT_OP_ENV1(unpackhsl)
1223 IWMMXT_OP_ENV_SIZE(cmpeq)
1224 IWMMXT_OP_ENV_SIZE(cmpgtu)
1225 IWMMXT_OP_ENV_SIZE(cmpgts)
1227 IWMMXT_OP_ENV_SIZE(mins)
1228 IWMMXT_OP_ENV_SIZE(minu)
1229 IWMMXT_OP_ENV_SIZE(maxs)
1230 IWMMXT_OP_ENV_SIZE(maxu)
1232 IWMMXT_OP_ENV_SIZE(subn)
1233 IWMMXT_OP_ENV_SIZE(addn)
1234 IWMMXT_OP_ENV_SIZE(subu)
1235 IWMMXT_OP_ENV_SIZE(addu)
1236 IWMMXT_OP_ENV_SIZE(subs)
1237 IWMMXT_OP_ENV_SIZE(adds)
1239 IWMMXT_OP_ENV(avgb0)
1240 IWMMXT_OP_ENV(avgb1)
1241 IWMMXT_OP_ENV(avgw0)
1242 IWMMXT_OP_ENV(avgw1)
1246 IWMMXT_OP_ENV(packuw)
1247 IWMMXT_OP_ENV(packul)
1248 IWMMXT_OP_ENV(packuq)
1249 IWMMXT_OP_ENV(packsw)
1250 IWMMXT_OP_ENV(packsl)
1251 IWMMXT_OP_ENV(packsq)
1253 static void gen_op_iwmmxt_set_mup(void)
1256 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1257 tcg_gen_ori_i32(tmp, tmp, 2);
1258 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1261 static void gen_op_iwmmxt_set_cup(void)
1264 tmp = load_cpu_field(iwmmxt.cregs[ARM_IWMMXT_wCon]);
1265 tcg_gen_ori_i32(tmp, tmp, 1);
1266 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCon]);
1269 static void gen_op_iwmmxt_setpsr_nz(void)
1271 TCGv tmp = new_tmp();
1272 gen_helper_iwmmxt_setpsr_nz(tmp, cpu_M0);
1273 store_cpu_field(tmp, iwmmxt.cregs[ARM_IWMMXT_wCASF]);
1276 static inline void gen_op_iwmmxt_addl_M0_wRn(int rn)
1278 iwmmxt_load_reg(cpu_V1, rn);
1279 tcg_gen_ext32u_i64(cpu_V1, cpu_V1);
1280 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1283 static inline int gen_iwmmxt_address(DisasContext *s, uint32_t insn, TCGv dest)
1289 rd = (insn >> 16) & 0xf;
1290 tmp = load_reg(s, rd);
1292 offset = (insn & 0xff) << ((insn >> 7) & 2);
1293 if (insn & (1 << 24)) {
1295 if (insn & (1 << 23))
1296 tcg_gen_addi_i32(tmp, tmp, offset);
1298 tcg_gen_addi_i32(tmp, tmp, -offset);
1299 tcg_gen_mov_i32(dest, tmp);
1300 if (insn & (1 << 21))
1301 store_reg(s, rd, tmp);
1304 } else if (insn & (1 << 21)) {
1306 tcg_gen_mov_i32(dest, tmp);
1307 if (insn & (1 << 23))
1308 tcg_gen_addi_i32(tmp, tmp, offset);
1310 tcg_gen_addi_i32(tmp, tmp, -offset);
1311 store_reg(s, rd, tmp);
1312 } else if (!(insn & (1 << 23)))
1317 static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask, TCGv dest)
1319 int rd = (insn >> 0) & 0xf;
1322 if (insn & (1 << 8)) {
1323 if (rd < ARM_IWMMXT_wCGR0 || rd > ARM_IWMMXT_wCGR3) {
1326 tmp = iwmmxt_load_creg(rd);
1330 iwmmxt_load_reg(cpu_V0, rd);
1331 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
1333 tcg_gen_andi_i32(tmp, tmp, mask);
1334 tcg_gen_mov_i32(dest, tmp);
1339 /* Disassemble an iwMMXt instruction. Returns nonzero if an error occured
1340 (ie. an undefined instruction). */
1341 static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn)
1344 int rdhi, rdlo, rd0, rd1, i;
1346 TCGv tmp, tmp2, tmp3;
1348 if ((insn & 0x0e000e00) == 0x0c000000) {
1349 if ((insn & 0x0fe00ff0) == 0x0c400000) {
1351 rdlo = (insn >> 12) & 0xf;
1352 rdhi = (insn >> 16) & 0xf;
1353 if (insn & ARM_CP_RW_BIT) { /* TMRRC */
1354 iwmmxt_load_reg(cpu_V0, wrd);
1355 tcg_gen_trunc_i64_i32(cpu_R[rdlo], cpu_V0);
1356 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
1357 tcg_gen_trunc_i64_i32(cpu_R[rdhi], cpu_V0);
1358 } else { /* TMCRR */
1359 tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]);
1360 iwmmxt_store_reg(cpu_V0, wrd);
1361 gen_op_iwmmxt_set_mup();
1366 wrd = (insn >> 12) & 0xf;
1368 if (gen_iwmmxt_address(s, insn, addr)) {
1372 if (insn & ARM_CP_RW_BIT) {
1373 if ((insn >> 28) == 0xf) { /* WLDRW wCx */
1375 tcg_gen_qemu_ld32u(tmp, addr, IS_USER(s));
1376 iwmmxt_store_creg(wrd, tmp);
1379 if (insn & (1 << 8)) {
1380 if (insn & (1 << 22)) { /* WLDRD */
1381 tcg_gen_qemu_ld64(cpu_M0, addr, IS_USER(s));
1383 } else { /* WLDRW wRd */
1384 tmp = gen_ld32(addr, IS_USER(s));
1387 if (insn & (1 << 22)) { /* WLDRH */
1388 tmp = gen_ld16u(addr, IS_USER(s));
1389 } else { /* WLDRB */
1390 tmp = gen_ld8u(addr, IS_USER(s));
1394 tcg_gen_extu_i32_i64(cpu_M0, tmp);
1397 gen_op_iwmmxt_movq_wRn_M0(wrd);
1400 if ((insn >> 28) == 0xf) { /* WSTRW wCx */
1401 tmp = iwmmxt_load_creg(wrd);
1402 gen_st32(tmp, addr, IS_USER(s));
1404 gen_op_iwmmxt_movq_M0_wRn(wrd);
1406 if (insn & (1 << 8)) {
1407 if (insn & (1 << 22)) { /* WSTRD */
1409 tcg_gen_qemu_st64(cpu_M0, addr, IS_USER(s));
1410 } else { /* WSTRW wRd */
1411 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1412 gen_st32(tmp, addr, IS_USER(s));
1415 if (insn & (1 << 22)) { /* WSTRH */
1416 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1417 gen_st16(tmp, addr, IS_USER(s));
1418 } else { /* WSTRB */
1419 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1420 gen_st8(tmp, addr, IS_USER(s));
1429 if ((insn & 0x0f000000) != 0x0e000000)
1432 switch (((insn >> 12) & 0xf00) | ((insn >> 4) & 0xff)) {
1433 case 0x000: /* WOR */
1434 wrd = (insn >> 12) & 0xf;
1435 rd0 = (insn >> 0) & 0xf;
1436 rd1 = (insn >> 16) & 0xf;
1437 gen_op_iwmmxt_movq_M0_wRn(rd0);
1438 gen_op_iwmmxt_orq_M0_wRn(rd1);
1439 gen_op_iwmmxt_setpsr_nz();
1440 gen_op_iwmmxt_movq_wRn_M0(wrd);
1441 gen_op_iwmmxt_set_mup();
1442 gen_op_iwmmxt_set_cup();
1444 case 0x011: /* TMCR */
1447 rd = (insn >> 12) & 0xf;
1448 wrd = (insn >> 16) & 0xf;
1450 case ARM_IWMMXT_wCID:
1451 case ARM_IWMMXT_wCASF:
1453 case ARM_IWMMXT_wCon:
1454 gen_op_iwmmxt_set_cup();
1456 case ARM_IWMMXT_wCSSF:
1457 tmp = iwmmxt_load_creg(wrd);
1458 tmp2 = load_reg(s, rd);
1459 tcg_gen_andc_i32(tmp, tmp, tmp2);
1461 iwmmxt_store_creg(wrd, tmp);
1463 case ARM_IWMMXT_wCGR0:
1464 case ARM_IWMMXT_wCGR1:
1465 case ARM_IWMMXT_wCGR2:
1466 case ARM_IWMMXT_wCGR3:
1467 gen_op_iwmmxt_set_cup();
1468 tmp = load_reg(s, rd);
1469 iwmmxt_store_creg(wrd, tmp);
1475 case 0x100: /* WXOR */
1476 wrd = (insn >> 12) & 0xf;
1477 rd0 = (insn >> 0) & 0xf;
1478 rd1 = (insn >> 16) & 0xf;
1479 gen_op_iwmmxt_movq_M0_wRn(rd0);
1480 gen_op_iwmmxt_xorq_M0_wRn(rd1);
1481 gen_op_iwmmxt_setpsr_nz();
1482 gen_op_iwmmxt_movq_wRn_M0(wrd);
1483 gen_op_iwmmxt_set_mup();
1484 gen_op_iwmmxt_set_cup();
1486 case 0x111: /* TMRC */
1489 rd = (insn >> 12) & 0xf;
1490 wrd = (insn >> 16) & 0xf;
1491 tmp = iwmmxt_load_creg(wrd);
1492 store_reg(s, rd, tmp);
1494 case 0x300: /* WANDN */
1495 wrd = (insn >> 12) & 0xf;
1496 rd0 = (insn >> 0) & 0xf;
1497 rd1 = (insn >> 16) & 0xf;
1498 gen_op_iwmmxt_movq_M0_wRn(rd0);
1499 tcg_gen_neg_i64(cpu_M0, cpu_M0);
1500 gen_op_iwmmxt_andq_M0_wRn(rd1);
1501 gen_op_iwmmxt_setpsr_nz();
1502 gen_op_iwmmxt_movq_wRn_M0(wrd);
1503 gen_op_iwmmxt_set_mup();
1504 gen_op_iwmmxt_set_cup();
1506 case 0x200: /* WAND */
1507 wrd = (insn >> 12) & 0xf;
1508 rd0 = (insn >> 0) & 0xf;
1509 rd1 = (insn >> 16) & 0xf;
1510 gen_op_iwmmxt_movq_M0_wRn(rd0);
1511 gen_op_iwmmxt_andq_M0_wRn(rd1);
1512 gen_op_iwmmxt_setpsr_nz();
1513 gen_op_iwmmxt_movq_wRn_M0(wrd);
1514 gen_op_iwmmxt_set_mup();
1515 gen_op_iwmmxt_set_cup();
1517 case 0x810: case 0xa10: /* WMADD */
1518 wrd = (insn >> 12) & 0xf;
1519 rd0 = (insn >> 0) & 0xf;
1520 rd1 = (insn >> 16) & 0xf;
1521 gen_op_iwmmxt_movq_M0_wRn(rd0);
1522 if (insn & (1 << 21))
1523 gen_op_iwmmxt_maddsq_M0_wRn(rd1);
1525 gen_op_iwmmxt_madduq_M0_wRn(rd1);
1526 gen_op_iwmmxt_movq_wRn_M0(wrd);
1527 gen_op_iwmmxt_set_mup();
1529 case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */
1530 wrd = (insn >> 12) & 0xf;
1531 rd0 = (insn >> 16) & 0xf;
1532 rd1 = (insn >> 0) & 0xf;
1533 gen_op_iwmmxt_movq_M0_wRn(rd0);
1534 switch ((insn >> 22) & 3) {
1536 gen_op_iwmmxt_unpacklb_M0_wRn(rd1);
1539 gen_op_iwmmxt_unpacklw_M0_wRn(rd1);
1542 gen_op_iwmmxt_unpackll_M0_wRn(rd1);
1547 gen_op_iwmmxt_movq_wRn_M0(wrd);
1548 gen_op_iwmmxt_set_mup();
1549 gen_op_iwmmxt_set_cup();
1551 case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */
1552 wrd = (insn >> 12) & 0xf;
1553 rd0 = (insn >> 16) & 0xf;
1554 rd1 = (insn >> 0) & 0xf;
1555 gen_op_iwmmxt_movq_M0_wRn(rd0);
1556 switch ((insn >> 22) & 3) {
1558 gen_op_iwmmxt_unpackhb_M0_wRn(rd1);
1561 gen_op_iwmmxt_unpackhw_M0_wRn(rd1);
1564 gen_op_iwmmxt_unpackhl_M0_wRn(rd1);
1569 gen_op_iwmmxt_movq_wRn_M0(wrd);
1570 gen_op_iwmmxt_set_mup();
1571 gen_op_iwmmxt_set_cup();
1573 case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */
1574 wrd = (insn >> 12) & 0xf;
1575 rd0 = (insn >> 16) & 0xf;
1576 rd1 = (insn >> 0) & 0xf;
1577 gen_op_iwmmxt_movq_M0_wRn(rd0);
1578 if (insn & (1 << 22))
1579 gen_op_iwmmxt_sadw_M0_wRn(rd1);
1581 gen_op_iwmmxt_sadb_M0_wRn(rd1);
1582 if (!(insn & (1 << 20)))
1583 gen_op_iwmmxt_addl_M0_wRn(wrd);
1584 gen_op_iwmmxt_movq_wRn_M0(wrd);
1585 gen_op_iwmmxt_set_mup();
1587 case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */
1588 wrd = (insn >> 12) & 0xf;
1589 rd0 = (insn >> 16) & 0xf;
1590 rd1 = (insn >> 0) & 0xf;
1591 gen_op_iwmmxt_movq_M0_wRn(rd0);
1592 if (insn & (1 << 21)) {
1593 if (insn & (1 << 20))
1594 gen_op_iwmmxt_mulshw_M0_wRn(rd1);
1596 gen_op_iwmmxt_mulslw_M0_wRn(rd1);
1598 if (insn & (1 << 20))
1599 gen_op_iwmmxt_muluhw_M0_wRn(rd1);
1601 gen_op_iwmmxt_mululw_M0_wRn(rd1);
1603 gen_op_iwmmxt_movq_wRn_M0(wrd);
1604 gen_op_iwmmxt_set_mup();
1606 case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */
1607 wrd = (insn >> 12) & 0xf;
1608 rd0 = (insn >> 16) & 0xf;
1609 rd1 = (insn >> 0) & 0xf;
1610 gen_op_iwmmxt_movq_M0_wRn(rd0);
1611 if (insn & (1 << 21))
1612 gen_op_iwmmxt_macsw_M0_wRn(rd1);
1614 gen_op_iwmmxt_macuw_M0_wRn(rd1);
1615 if (!(insn & (1 << 20))) {
1616 iwmmxt_load_reg(cpu_V1, wrd);
1617 tcg_gen_add_i64(cpu_M0, cpu_M0, cpu_V1);
1619 gen_op_iwmmxt_movq_wRn_M0(wrd);
1620 gen_op_iwmmxt_set_mup();
1622 case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */
1623 wrd = (insn >> 12) & 0xf;
1624 rd0 = (insn >> 16) & 0xf;
1625 rd1 = (insn >> 0) & 0xf;
1626 gen_op_iwmmxt_movq_M0_wRn(rd0);
1627 switch ((insn >> 22) & 3) {
1629 gen_op_iwmmxt_cmpeqb_M0_wRn(rd1);
1632 gen_op_iwmmxt_cmpeqw_M0_wRn(rd1);
1635 gen_op_iwmmxt_cmpeql_M0_wRn(rd1);
1640 gen_op_iwmmxt_movq_wRn_M0(wrd);
1641 gen_op_iwmmxt_set_mup();
1642 gen_op_iwmmxt_set_cup();
1644 case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */
1645 wrd = (insn >> 12) & 0xf;
1646 rd0 = (insn >> 16) & 0xf;
1647 rd1 = (insn >> 0) & 0xf;
1648 gen_op_iwmmxt_movq_M0_wRn(rd0);
1649 if (insn & (1 << 22)) {
1650 if (insn & (1 << 20))
1651 gen_op_iwmmxt_avgw1_M0_wRn(rd1);
1653 gen_op_iwmmxt_avgw0_M0_wRn(rd1);
1655 if (insn & (1 << 20))
1656 gen_op_iwmmxt_avgb1_M0_wRn(rd1);
1658 gen_op_iwmmxt_avgb0_M0_wRn(rd1);
1660 gen_op_iwmmxt_movq_wRn_M0(wrd);
1661 gen_op_iwmmxt_set_mup();
1662 gen_op_iwmmxt_set_cup();
1664 case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */
1665 wrd = (insn >> 12) & 0xf;
1666 rd0 = (insn >> 16) & 0xf;
1667 rd1 = (insn >> 0) & 0xf;
1668 gen_op_iwmmxt_movq_M0_wRn(rd0);
1669 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCGR0 + ((insn >> 20) & 3));
1670 tcg_gen_andi_i32(tmp, tmp, 7);
1671 iwmmxt_load_reg(cpu_V1, rd1);
1672 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp);
1674 gen_op_iwmmxt_movq_wRn_M0(wrd);
1675 gen_op_iwmmxt_set_mup();
1677 case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */
1678 if (((insn >> 6) & 3) == 3)
1680 rd = (insn >> 12) & 0xf;
1681 wrd = (insn >> 16) & 0xf;
1682 tmp = load_reg(s, rd);
1683 gen_op_iwmmxt_movq_M0_wRn(wrd);
1684 switch ((insn >> 6) & 3) {
1686 tmp2 = tcg_const_i32(0xff);
1687 tmp3 = tcg_const_i32((insn & 7) << 3);
1690 tmp2 = tcg_const_i32(0xffff);
1691 tmp3 = tcg_const_i32((insn & 3) << 4);
1694 tmp2 = tcg_const_i32(0xffffffff);
1695 tmp3 = tcg_const_i32((insn & 1) << 5);
1701 gen_helper_iwmmxt_insr(cpu_M0, cpu_M0, tmp, tmp2, tmp3);
1702 tcg_temp_free(tmp3);
1703 tcg_temp_free(tmp2);
1705 gen_op_iwmmxt_movq_wRn_M0(wrd);
1706 gen_op_iwmmxt_set_mup();
1708 case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */
1709 rd = (insn >> 12) & 0xf;
1710 wrd = (insn >> 16) & 0xf;
1711 if (rd == 15 || ((insn >> 22) & 3) == 3)
1713 gen_op_iwmmxt_movq_M0_wRn(wrd);
1715 switch ((insn >> 22) & 3) {
1717 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 7) << 3);
1718 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1720 tcg_gen_ext8s_i32(tmp, tmp);
1722 tcg_gen_andi_i32(tmp, tmp, 0xff);
1726 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 3) << 4);
1727 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1729 tcg_gen_ext16s_i32(tmp, tmp);
1731 tcg_gen_andi_i32(tmp, tmp, 0xffff);
1735 tcg_gen_shri_i64(cpu_M0, cpu_M0, (insn & 1) << 5);
1736 tcg_gen_trunc_i64_i32(tmp, cpu_M0);
1739 store_reg(s, rd, tmp);
1741 case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */
1742 if ((insn & 0x000ff008) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1744 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1745 switch ((insn >> 22) & 3) {
1747 tcg_gen_shri_i32(tmp, tmp, ((insn & 7) << 2) + 0);
1750 tcg_gen_shri_i32(tmp, tmp, ((insn & 3) << 3) + 4);
1753 tcg_gen_shri_i32(tmp, tmp, ((insn & 1) << 4) + 12);
1756 tcg_gen_shli_i32(tmp, tmp, 28);
1760 case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */
1761 if (((insn >> 6) & 3) == 3)
1763 rd = (insn >> 12) & 0xf;
1764 wrd = (insn >> 16) & 0xf;
1765 tmp = load_reg(s, rd);
1766 switch ((insn >> 6) & 3) {
1768 gen_helper_iwmmxt_bcstb(cpu_M0, tmp);
1771 gen_helper_iwmmxt_bcstw(cpu_M0, tmp);
1774 gen_helper_iwmmxt_bcstl(cpu_M0, tmp);
1778 gen_op_iwmmxt_movq_wRn_M0(wrd);
1779 gen_op_iwmmxt_set_mup();
1781 case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */
1782 if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1784 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1786 tcg_gen_mov_i32(tmp2, tmp);
1787 switch ((insn >> 22) & 3) {
1789 for (i = 0; i < 7; i ++) {
1790 tcg_gen_shli_i32(tmp2, tmp2, 4);
1791 tcg_gen_and_i32(tmp, tmp, tmp2);
1795 for (i = 0; i < 3; i ++) {
1796 tcg_gen_shli_i32(tmp2, tmp2, 8);
1797 tcg_gen_and_i32(tmp, tmp, tmp2);
1801 tcg_gen_shli_i32(tmp2, tmp2, 16);
1802 tcg_gen_and_i32(tmp, tmp, tmp2);
1809 case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */
1810 wrd = (insn >> 12) & 0xf;
1811 rd0 = (insn >> 16) & 0xf;
1812 gen_op_iwmmxt_movq_M0_wRn(rd0);
1813 switch ((insn >> 22) & 3) {
1815 gen_helper_iwmmxt_addcb(cpu_M0, cpu_M0);
1818 gen_helper_iwmmxt_addcw(cpu_M0, cpu_M0);
1821 gen_helper_iwmmxt_addcl(cpu_M0, cpu_M0);
1826 gen_op_iwmmxt_movq_wRn_M0(wrd);
1827 gen_op_iwmmxt_set_mup();
1829 case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */
1830 if ((insn & 0x000ff00f) != 0x0003f000 || ((insn >> 22) & 3) == 3)
1832 tmp = iwmmxt_load_creg(ARM_IWMMXT_wCASF);
1834 tcg_gen_mov_i32(tmp2, tmp);
1835 switch ((insn >> 22) & 3) {
1837 for (i = 0; i < 7; i ++) {
1838 tcg_gen_shli_i32(tmp2, tmp2, 4);
1839 tcg_gen_or_i32(tmp, tmp, tmp2);
1843 for (i = 0; i < 3; i ++) {
1844 tcg_gen_shli_i32(tmp2, tmp2, 8);
1845 tcg_gen_or_i32(tmp, tmp, tmp2);
1849 tcg_gen_shli_i32(tmp2, tmp2, 16);
1850 tcg_gen_or_i32(tmp, tmp, tmp2);
1857 case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */
1858 rd = (insn >> 12) & 0xf;
1859 rd0 = (insn >> 16) & 0xf;
1860 if ((insn & 0xf) != 0 || ((insn >> 22) & 3) == 3)
1862 gen_op_iwmmxt_movq_M0_wRn(rd0);
1864 switch ((insn >> 22) & 3) {
1866 gen_helper_iwmmxt_msbb(tmp, cpu_M0);
1869 gen_helper_iwmmxt_msbw(tmp, cpu_M0);
1872 gen_helper_iwmmxt_msbl(tmp, cpu_M0);
1875 store_reg(s, rd, tmp);
1877 case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */
1878 case 0x906: case 0xb06: case 0xd06: case 0xf06:
1879 wrd = (insn >> 12) & 0xf;
1880 rd0 = (insn >> 16) & 0xf;
1881 rd1 = (insn >> 0) & 0xf;
1882 gen_op_iwmmxt_movq_M0_wRn(rd0);
1883 switch ((insn >> 22) & 3) {
1885 if (insn & (1 << 21))
1886 gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1);
1888 gen_op_iwmmxt_cmpgtub_M0_wRn(rd1);
1891 if (insn & (1 << 21))
1892 gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1);
1894 gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1);
1897 if (insn & (1 << 21))
1898 gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1);
1900 gen_op_iwmmxt_cmpgtul_M0_wRn(rd1);
1905 gen_op_iwmmxt_movq_wRn_M0(wrd);
1906 gen_op_iwmmxt_set_mup();
1907 gen_op_iwmmxt_set_cup();
1909 case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */
1910 case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e:
1911 wrd = (insn >> 12) & 0xf;
1912 rd0 = (insn >> 16) & 0xf;
1913 gen_op_iwmmxt_movq_M0_wRn(rd0);
1914 switch ((insn >> 22) & 3) {
1916 if (insn & (1 << 21))
1917 gen_op_iwmmxt_unpacklsb_M0();
1919 gen_op_iwmmxt_unpacklub_M0();
1922 if (insn & (1 << 21))
1923 gen_op_iwmmxt_unpacklsw_M0();
1925 gen_op_iwmmxt_unpackluw_M0();
1928 if (insn & (1 << 21))
1929 gen_op_iwmmxt_unpacklsl_M0();
1931 gen_op_iwmmxt_unpacklul_M0();
1936 gen_op_iwmmxt_movq_wRn_M0(wrd);
1937 gen_op_iwmmxt_set_mup();
1938 gen_op_iwmmxt_set_cup();
1940 case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */
1941 case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c:
1942 wrd = (insn >> 12) & 0xf;
1943 rd0 = (insn >> 16) & 0xf;
1944 gen_op_iwmmxt_movq_M0_wRn(rd0);
1945 switch ((insn >> 22) & 3) {
1947 if (insn & (1 << 21))
1948 gen_op_iwmmxt_unpackhsb_M0();
1950 gen_op_iwmmxt_unpackhub_M0();
1953 if (insn & (1 << 21))
1954 gen_op_iwmmxt_unpackhsw_M0();
1956 gen_op_iwmmxt_unpackhuw_M0();
1959 if (insn & (1 << 21))
1960 gen_op_iwmmxt_unpackhsl_M0();
1962 gen_op_iwmmxt_unpackhul_M0();
1967 gen_op_iwmmxt_movq_wRn_M0(wrd);
1968 gen_op_iwmmxt_set_mup();
1969 gen_op_iwmmxt_set_cup();
1971 case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */
1972 case 0x214: case 0x614: case 0xa14: case 0xe14:
1973 if (((insn >> 22) & 3) == 0)
1975 wrd = (insn >> 12) & 0xf;
1976 rd0 = (insn >> 16) & 0xf;
1977 gen_op_iwmmxt_movq_M0_wRn(rd0);
1979 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
1983 switch ((insn >> 22) & 3) {
1985 gen_helper_iwmmxt_srlw(cpu_M0, cpu_env, cpu_M0, tmp);
1988 gen_helper_iwmmxt_srll(cpu_M0, cpu_env, cpu_M0, tmp);
1991 gen_helper_iwmmxt_srlq(cpu_M0, cpu_env, cpu_M0, tmp);
1995 gen_op_iwmmxt_movq_wRn_M0(wrd);
1996 gen_op_iwmmxt_set_mup();
1997 gen_op_iwmmxt_set_cup();
1999 case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */
2000 case 0x014: case 0x414: case 0x814: case 0xc14:
2001 if (((insn >> 22) & 3) == 0)
2003 wrd = (insn >> 12) & 0xf;
2004 rd0 = (insn >> 16) & 0xf;
2005 gen_op_iwmmxt_movq_M0_wRn(rd0);
2007 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2011 switch ((insn >> 22) & 3) {
2013 gen_helper_iwmmxt_sraw(cpu_M0, cpu_env, cpu_M0, tmp);
2016 gen_helper_iwmmxt_sral(cpu_M0, cpu_env, cpu_M0, tmp);
2019 gen_helper_iwmmxt_sraq(cpu_M0, cpu_env, cpu_M0, tmp);
2023 gen_op_iwmmxt_movq_wRn_M0(wrd);
2024 gen_op_iwmmxt_set_mup();
2025 gen_op_iwmmxt_set_cup();
2027 case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */
2028 case 0x114: case 0x514: case 0x914: case 0xd14:
2029 if (((insn >> 22) & 3) == 0)
2031 wrd = (insn >> 12) & 0xf;
2032 rd0 = (insn >> 16) & 0xf;
2033 gen_op_iwmmxt_movq_M0_wRn(rd0);
2035 if (gen_iwmmxt_shift(insn, 0xff, tmp)) {
2039 switch ((insn >> 22) & 3) {
2041 gen_helper_iwmmxt_sllw(cpu_M0, cpu_env, cpu_M0, tmp);
2044 gen_helper_iwmmxt_slll(cpu_M0, cpu_env, cpu_M0, tmp);
2047 gen_helper_iwmmxt_sllq(cpu_M0, cpu_env, cpu_M0, tmp);
2051 gen_op_iwmmxt_movq_wRn_M0(wrd);
2052 gen_op_iwmmxt_set_mup();
2053 gen_op_iwmmxt_set_cup();
2055 case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */
2056 case 0x314: case 0x714: case 0xb14: case 0xf14:
2057 if (((insn >> 22) & 3) == 0)
2059 wrd = (insn >> 12) & 0xf;
2060 rd0 = (insn >> 16) & 0xf;
2061 gen_op_iwmmxt_movq_M0_wRn(rd0);
2063 switch ((insn >> 22) & 3) {
2065 if (gen_iwmmxt_shift(insn, 0xf, tmp)) {
2069 gen_helper_iwmmxt_rorw(cpu_M0, cpu_env, cpu_M0, tmp);
2072 if (gen_iwmmxt_shift(insn, 0x1f, tmp)) {
2076 gen_helper_iwmmxt_rorl(cpu_M0, cpu_env, cpu_M0, tmp);
2079 if (gen_iwmmxt_shift(insn, 0x3f, tmp)) {
2083 gen_helper_iwmmxt_rorq(cpu_M0, cpu_env, cpu_M0, tmp);
2087 gen_op_iwmmxt_movq_wRn_M0(wrd);
2088 gen_op_iwmmxt_set_mup();
2089 gen_op_iwmmxt_set_cup();
2091 case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */
2092 case 0x916: case 0xb16: case 0xd16: case 0xf16:
2093 wrd = (insn >> 12) & 0xf;
2094 rd0 = (insn >> 16) & 0xf;
2095 rd1 = (insn >> 0) & 0xf;
2096 gen_op_iwmmxt_movq_M0_wRn(rd0);
2097 switch ((insn >> 22) & 3) {
2099 if (insn & (1 << 21))
2100 gen_op_iwmmxt_minsb_M0_wRn(rd1);
2102 gen_op_iwmmxt_minub_M0_wRn(rd1);
2105 if (insn & (1 << 21))
2106 gen_op_iwmmxt_minsw_M0_wRn(rd1);
2108 gen_op_iwmmxt_minuw_M0_wRn(rd1);
2111 if (insn & (1 << 21))
2112 gen_op_iwmmxt_minsl_M0_wRn(rd1);
2114 gen_op_iwmmxt_minul_M0_wRn(rd1);
2119 gen_op_iwmmxt_movq_wRn_M0(wrd);
2120 gen_op_iwmmxt_set_mup();
2122 case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */
2123 case 0x816: case 0xa16: case 0xc16: case 0xe16:
2124 wrd = (insn >> 12) & 0xf;
2125 rd0 = (insn >> 16) & 0xf;
2126 rd1 = (insn >> 0) & 0xf;
2127 gen_op_iwmmxt_movq_M0_wRn(rd0);
2128 switch ((insn >> 22) & 3) {
2130 if (insn & (1 << 21))
2131 gen_op_iwmmxt_maxsb_M0_wRn(rd1);
2133 gen_op_iwmmxt_maxub_M0_wRn(rd1);
2136 if (insn & (1 << 21))
2137 gen_op_iwmmxt_maxsw_M0_wRn(rd1);
2139 gen_op_iwmmxt_maxuw_M0_wRn(rd1);
2142 if (insn & (1 << 21))
2143 gen_op_iwmmxt_maxsl_M0_wRn(rd1);
2145 gen_op_iwmmxt_maxul_M0_wRn(rd1);
2150 gen_op_iwmmxt_movq_wRn_M0(wrd);
2151 gen_op_iwmmxt_set_mup();
2153 case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */
2154 case 0x402: case 0x502: case 0x602: case 0x702:
2155 wrd = (insn >> 12) & 0xf;
2156 rd0 = (insn >> 16) & 0xf;
2157 rd1 = (insn >> 0) & 0xf;
2158 gen_op_iwmmxt_movq_M0_wRn(rd0);
2159 tmp = tcg_const_i32((insn >> 20) & 3);
2160 iwmmxt_load_reg(cpu_V1, rd1);
2161 gen_helper_iwmmxt_align(cpu_M0, cpu_M0, cpu_V1, tmp);
2163 gen_op_iwmmxt_movq_wRn_M0(wrd);
2164 gen_op_iwmmxt_set_mup();
2166 case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */
2167 case 0x41a: case 0x51a: case 0x61a: case 0x71a:
2168 case 0x81a: case 0x91a: case 0xa1a: case 0xb1a:
2169 case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a:
2170 wrd = (insn >> 12) & 0xf;
2171 rd0 = (insn >> 16) & 0xf;
2172 rd1 = (insn >> 0) & 0xf;
2173 gen_op_iwmmxt_movq_M0_wRn(rd0);
2174 switch ((insn >> 20) & 0xf) {
2176 gen_op_iwmmxt_subnb_M0_wRn(rd1);
2179 gen_op_iwmmxt_subub_M0_wRn(rd1);
2182 gen_op_iwmmxt_subsb_M0_wRn(rd1);
2185 gen_op_iwmmxt_subnw_M0_wRn(rd1);
2188 gen_op_iwmmxt_subuw_M0_wRn(rd1);
2191 gen_op_iwmmxt_subsw_M0_wRn(rd1);
2194 gen_op_iwmmxt_subnl_M0_wRn(rd1);
2197 gen_op_iwmmxt_subul_M0_wRn(rd1);
2200 gen_op_iwmmxt_subsl_M0_wRn(rd1);
2205 gen_op_iwmmxt_movq_wRn_M0(wrd);
2206 gen_op_iwmmxt_set_mup();
2207 gen_op_iwmmxt_set_cup();
2209 case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */
2210 case 0x41e: case 0x51e: case 0x61e: case 0x71e:
2211 case 0x81e: case 0x91e: case 0xa1e: case 0xb1e:
2212 case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e:
2213 wrd = (insn >> 12) & 0xf;
2214 rd0 = (insn >> 16) & 0xf;
2215 gen_op_iwmmxt_movq_M0_wRn(rd0);
2216 tmp = tcg_const_i32(((insn >> 16) & 0xf0) | (insn & 0x0f));
2217 gen_helper_iwmmxt_shufh(cpu_M0, cpu_env, cpu_M0, tmp);
2219 gen_op_iwmmxt_movq_wRn_M0(wrd);
2220 gen_op_iwmmxt_set_mup();
2221 gen_op_iwmmxt_set_cup();
2223 case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */
2224 case 0x418: case 0x518: case 0x618: case 0x718:
2225 case 0x818: case 0x918: case 0xa18: case 0xb18:
2226 case 0xc18: case 0xd18: case 0xe18: case 0xf18:
2227 wrd = (insn >> 12) & 0xf;
2228 rd0 = (insn >> 16) & 0xf;
2229 rd1 = (insn >> 0) & 0xf;
2230 gen_op_iwmmxt_movq_M0_wRn(rd0);
2231 switch ((insn >> 20) & 0xf) {
2233 gen_op_iwmmxt_addnb_M0_wRn(rd1);
2236 gen_op_iwmmxt_addub_M0_wRn(rd1);
2239 gen_op_iwmmxt_addsb_M0_wRn(rd1);
2242 gen_op_iwmmxt_addnw_M0_wRn(rd1);
2245 gen_op_iwmmxt_adduw_M0_wRn(rd1);
2248 gen_op_iwmmxt_addsw_M0_wRn(rd1);
2251 gen_op_iwmmxt_addnl_M0_wRn(rd1);
2254 gen_op_iwmmxt_addul_M0_wRn(rd1);
2257 gen_op_iwmmxt_addsl_M0_wRn(rd1);
2262 gen_op_iwmmxt_movq_wRn_M0(wrd);
2263 gen_op_iwmmxt_set_mup();
2264 gen_op_iwmmxt_set_cup();
2266 case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */
2267 case 0x408: case 0x508: case 0x608: case 0x708:
2268 case 0x808: case 0x908: case 0xa08: case 0xb08:
2269 case 0xc08: case 0xd08: case 0xe08: case 0xf08:
2270 if (!(insn & (1 << 20)) || ((insn >> 22) & 3) == 0)
2272 wrd = (insn >> 12) & 0xf;
2273 rd0 = (insn >> 16) & 0xf;
2274 rd1 = (insn >> 0) & 0xf;
2275 gen_op_iwmmxt_movq_M0_wRn(rd0);
2276 switch ((insn >> 22) & 3) {
2278 if (insn & (1 << 21))
2279 gen_op_iwmmxt_packsw_M0_wRn(rd1);
2281 gen_op_iwmmxt_packuw_M0_wRn(rd1);
2284 if (insn & (1 << 21))
2285 gen_op_iwmmxt_packsl_M0_wRn(rd1);
2287 gen_op_iwmmxt_packul_M0_wRn(rd1);
2290 if (insn & (1 << 21))
2291 gen_op_iwmmxt_packsq_M0_wRn(rd1);
2293 gen_op_iwmmxt_packuq_M0_wRn(rd1);
2296 gen_op_iwmmxt_movq_wRn_M0(wrd);
2297 gen_op_iwmmxt_set_mup();
2298 gen_op_iwmmxt_set_cup();
2300 case 0x201: case 0x203: case 0x205: case 0x207:
2301 case 0x209: case 0x20b: case 0x20d: case 0x20f:
2302 case 0x211: case 0x213: case 0x215: case 0x217:
2303 case 0x219: case 0x21b: case 0x21d: case 0x21f:
2304 wrd = (insn >> 5) & 0xf;
2305 rd0 = (insn >> 12) & 0xf;
2306 rd1 = (insn >> 0) & 0xf;
2307 if (rd0 == 0xf || rd1 == 0xf)
2309 gen_op_iwmmxt_movq_M0_wRn(wrd);
2310 tmp = load_reg(s, rd0);
2311 tmp2 = load_reg(s, rd1);
2312 switch ((insn >> 16) & 0xf) {
2313 case 0x0: /* TMIA */
2314 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2);
2316 case 0x8: /* TMIAPH */
2317 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2);
2319 case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */
2320 if (insn & (1 << 16))
2321 tcg_gen_shri_i32(tmp, tmp, 16);
2322 if (insn & (1 << 17))
2323 tcg_gen_shri_i32(tmp2, tmp2, 16);
2324 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2);
2333 gen_op_iwmmxt_movq_wRn_M0(wrd);
2334 gen_op_iwmmxt_set_mup();
2343 /* Disassemble an XScale DSP instruction. Returns nonzero if an error occured
2344 (ie. an undefined instruction). */
2345 static int disas_dsp_insn(CPUState *env, DisasContext *s, uint32_t insn)
2347 int acc, rd0, rd1, rdhi, rdlo;
2350 if ((insn & 0x0ff00f10) == 0x0e200010) {
2351 /* Multiply with Internal Accumulate Format */
2352 rd0 = (insn >> 12) & 0xf;
2354 acc = (insn >> 5) & 7;
2359 tmp = load_reg(s, rd0);
2360 tmp2 = load_reg(s, rd1);
2361 switch ((insn >> 16) & 0xf) {
2363 gen_helper_iwmmxt_muladdsl(cpu_M0, cpu_M0, tmp, tmp2);
2365 case 0x8: /* MIAPH */
2366 gen_helper_iwmmxt_muladdsw(cpu_M0, cpu_M0, tmp, tmp2);
2368 case 0xc: /* MIABB */
2369 case 0xd: /* MIABT */
2370 case 0xe: /* MIATB */
2371 case 0xf: /* MIATT */
2372 if (insn & (1 << 16))
2373 tcg_gen_shri_i32(tmp, tmp, 16);
2374 if (insn & (1 << 17))
2375 tcg_gen_shri_i32(tmp2, tmp2, 16);
2376 gen_helper_iwmmxt_muladdswl(cpu_M0, cpu_M0, tmp, tmp2);
2384 gen_op_iwmmxt_movq_wRn_M0(acc);
2388 if ((insn & 0x0fe00ff8) == 0x0c400000) {
2389 /* Internal Accumulator Access Format */
2390 rdhi = (insn >> 16) & 0xf;
2391 rdlo = (insn >> 12) & 0xf;
2397 if (insn & ARM_CP_RW_BIT) { /* MRA */
2398 iwmmxt_load_reg(cpu_V0, acc);
2399 tcg_gen_trunc_i64_i32(cpu_R[rdlo], cpu_V0);
2400 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
2401 tcg_gen_trunc_i64_i32(cpu_R[rdhi], cpu_V0);
2402 tcg_gen_andi_i32(cpu_R[rdhi], cpu_R[rdhi], (1 << (40 - 32)) - 1);
2404 tcg_gen_concat_i32_i64(cpu_V0, cpu_R[rdlo], cpu_R[rdhi]);
2405 iwmmxt_store_reg(cpu_V0, acc);
2413 /* Disassemble system coprocessor instruction. Return nonzero if
2414 instruction is not defined. */
2415 static int disas_cp_insn(CPUState *env, DisasContext *s, uint32_t insn)
2418 uint32_t rd = (insn >> 12) & 0xf;
2419 uint32_t cp = (insn >> 8) & 0xf;
2424 if (insn & ARM_CP_RW_BIT) {
2425 if (!env->cp[cp].cp_read)
2427 gen_set_pc_im(s->pc);
2429 tmp2 = tcg_const_i32(insn);
2430 gen_helper_get_cp(tmp, cpu_env, tmp2);
2431 tcg_temp_free(tmp2);
2432 store_reg(s, rd, tmp);
2434 if (!env->cp[cp].cp_write)
2436 gen_set_pc_im(s->pc);
2437 tmp = load_reg(s, rd);
2438 tmp2 = tcg_const_i32(insn);
2439 gen_helper_set_cp(cpu_env, tmp2, tmp);
2440 tcg_temp_free(tmp2);
2446 static int cp15_user_ok(uint32_t insn)
2448 int cpn = (insn >> 16) & 0xf;
2449 int cpm = insn & 0xf;
2450 int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38);
2452 if (cpn == 13 && cpm == 0) {
2454 if (op == 2 || (op == 3 && (insn & ARM_CP_RW_BIT)))
2458 /* ISB, DSB, DMB. */
2459 if ((cpm == 5 && op == 4)
2460 || (cpm == 10 && (op == 4 || op == 5)))
2466 static int cp15_tls_load_store(CPUState *env, DisasContext *s, uint32_t insn, uint32_t rd)
2469 int cpn = (insn >> 16) & 0xf;
2470 int cpm = insn & 0xf;
2471 int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38);
2473 if (!arm_feature(env, ARM_FEATURE_V6K))
2476 if (!(cpn == 13 && cpm == 0))
2479 if (insn & ARM_CP_RW_BIT) {
2482 tmp = load_cpu_field(cp15.c13_tls1);
2485 tmp = load_cpu_field(cp15.c13_tls2);
2488 tmp = load_cpu_field(cp15.c13_tls3);
2493 store_reg(s, rd, tmp);
2496 tmp = load_reg(s, rd);
2499 store_cpu_field(tmp, cp15.c13_tls1);
2502 store_cpu_field(tmp, cp15.c13_tls2);
2505 store_cpu_field(tmp, cp15.c13_tls3);
2515 /* Disassemble system coprocessor (cp15) instruction. Return nonzero if
2516 instruction is not defined. */
2517 static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn)
2522 /* M profile cores use memory mapped registers instead of cp15. */
2523 if (arm_feature(env, ARM_FEATURE_M))
2526 if ((insn & (1 << 25)) == 0) {
2527 if (insn & (1 << 20)) {
2531 /* mcrr. Used for block cache operations, so implement as no-op. */
2534 if ((insn & (1 << 4)) == 0) {
2538 if (IS_USER(s) && !cp15_user_ok(insn)) {
2541 if ((insn & 0x0fff0fff) == 0x0e070f90
2542 || (insn & 0x0fff0fff) == 0x0e070f58) {
2543 /* Wait for interrupt. */
2544 gen_set_pc_im(s->pc);
2545 s->is_jmp = DISAS_WFI;
2548 rd = (insn >> 12) & 0xf;
2550 if (cp15_tls_load_store(env, s, insn, rd))
2553 tmp2 = tcg_const_i32(insn);
2554 if (insn & ARM_CP_RW_BIT) {
2556 gen_helper_get_cp15(tmp, cpu_env, tmp2);
2557 /* If the destination register is r15 then sets condition codes. */
2559 store_reg(s, rd, tmp);
2563 tmp = load_reg(s, rd);
2564 gen_helper_set_cp15(cpu_env, tmp2, tmp);
2566 /* Normally we would always end the TB here, but Linux
2567 * arch/arm/mach-pxa/sleep.S expects two instructions following
2568 * an MMU enable to execute from cache. Imitate this behaviour. */
2569 if (!arm_feature(env, ARM_FEATURE_XSCALE) ||
2570 (insn & 0x0fff0fff) != 0x0e010f10)
2573 tcg_temp_free_i32(tmp2);
2577 #define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n))
2578 #define VFP_SREG(insn, bigbit, smallbit) \
2579 ((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1))
2580 #define VFP_DREG(reg, insn, bigbit, smallbit) do { \
2581 if (arm_feature(env, ARM_FEATURE_VFP3)) { \
2582 reg = (((insn) >> (bigbit)) & 0x0f) \
2583 | (((insn) >> ((smallbit) - 4)) & 0x10); \
2585 if (insn & (1 << (smallbit))) \
2587 reg = ((insn) >> (bigbit)) & 0x0f; \
2590 #define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22)
2591 #define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22)
2592 #define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7)
2593 #define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7)
2594 #define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5)
2595 #define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5)
2597 /* Move between integer and VFP cores. */
2598 static TCGv gen_vfp_mrs(void)
2600 TCGv tmp = new_tmp();
2601 tcg_gen_mov_i32(tmp, cpu_F0s);
2605 static void gen_vfp_msr(TCGv tmp)
2607 tcg_gen_mov_i32(cpu_F0s, tmp);
2611 static void gen_neon_dup_u8(TCGv var, int shift)
2613 TCGv tmp = new_tmp();
2615 tcg_gen_shri_i32(var, var, shift);
2616 tcg_gen_ext8u_i32(var, var);
2617 tcg_gen_shli_i32(tmp, var, 8);
2618 tcg_gen_or_i32(var, var, tmp);
2619 tcg_gen_shli_i32(tmp, var, 16);
2620 tcg_gen_or_i32(var, var, tmp);
2624 static void gen_neon_dup_low16(TCGv var)
2626 TCGv tmp = new_tmp();
2627 tcg_gen_ext16u_i32(var, var);
2628 tcg_gen_shli_i32(tmp, var, 16);
2629 tcg_gen_or_i32(var, var, tmp);
2633 static void gen_neon_dup_high16(TCGv var)
2635 TCGv tmp = new_tmp();
2636 tcg_gen_andi_i32(var, var, 0xffff0000);
2637 tcg_gen_shri_i32(tmp, var, 16);
2638 tcg_gen_or_i32(var, var, tmp);
2642 /* Disassemble a VFP instruction. Returns nonzero if an error occured
2643 (ie. an undefined instruction). */
2644 static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
2646 uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask;
2652 if (!arm_feature(env, ARM_FEATURE_VFP))
2655 if (!s->vfp_enabled) {
2656 /* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
2657 if ((insn & 0x0fe00fff) != 0x0ee00a10)
2659 rn = (insn >> 16) & 0xf;
2660 if (rn != ARM_VFP_FPSID && rn != ARM_VFP_FPEXC
2661 && rn != ARM_VFP_MVFR1 && rn != ARM_VFP_MVFR0)
2664 dp = ((insn & 0xf00) == 0xb00);
2665 switch ((insn >> 24) & 0xf) {
2667 if (insn & (1 << 4)) {
2668 /* single register transfer */
2669 rd = (insn >> 12) & 0xf;
2674 VFP_DREG_N(rn, insn);
2677 if (insn & 0x00c00060
2678 && !arm_feature(env, ARM_FEATURE_NEON))
2681 pass = (insn >> 21) & 1;
2682 if (insn & (1 << 22)) {
2684 offset = ((insn >> 5) & 3) * 8;
2685 } else if (insn & (1 << 5)) {
2687 offset = (insn & (1 << 6)) ? 16 : 0;
2692 if (insn & ARM_CP_RW_BIT) {
2694 tmp = neon_load_reg(rn, pass);
2698 tcg_gen_shri_i32(tmp, tmp, offset);
2699 if (insn & (1 << 23))
2705 if (insn & (1 << 23)) {
2707 tcg_gen_shri_i32(tmp, tmp, 16);
2713 tcg_gen_sari_i32(tmp, tmp, 16);
2722 store_reg(s, rd, tmp);
2725 tmp = load_reg(s, rd);
2726 if (insn & (1 << 23)) {
2729 gen_neon_dup_u8(tmp, 0);
2730 } else if (size == 1) {
2731 gen_neon_dup_low16(tmp);
2733 for (n = 0; n <= pass * 2; n++) {
2735 tcg_gen_mov_i32(tmp2, tmp);
2736 neon_store_reg(rn, n, tmp2);
2738 neon_store_reg(rn, n, tmp);
2743 tmp2 = neon_load_reg(rn, pass);
2744 gen_bfi(tmp, tmp2, tmp, offset, 0xff);
2748 tmp2 = neon_load_reg(rn, pass);
2749 gen_bfi(tmp, tmp2, tmp, offset, 0xffff);
2755 neon_store_reg(rn, pass, tmp);
2759 if ((insn & 0x6f) != 0x00)
2761 rn = VFP_SREG_N(insn);
2762 if (insn & ARM_CP_RW_BIT) {
2764 if (insn & (1 << 21)) {
2765 /* system register */
2770 /* VFP2 allows access to FSID from userspace.
2771 VFP3 restricts all id registers to privileged
2774 && arm_feature(env, ARM_FEATURE_VFP3))
2776 tmp = load_cpu_field(vfp.xregs[rn]);
2781 tmp = load_cpu_field(vfp.xregs[rn]);
2783 case ARM_VFP_FPINST:
2784 case ARM_VFP_FPINST2:
2785 /* Not present in VFP3. */
2787 || arm_feature(env, ARM_FEATURE_VFP3))
2789 tmp = load_cpu_field(vfp.xregs[rn]);
2793 tmp = load_cpu_field(vfp.xregs[ARM_VFP_FPSCR]);
2794 tcg_gen_andi_i32(tmp, tmp, 0xf0000000);
2797 gen_helper_vfp_get_fpscr(tmp, cpu_env);
2803 || !arm_feature(env, ARM_FEATURE_VFP3))
2805 tmp = load_cpu_field(vfp.xregs[rn]);
2811 gen_mov_F0_vreg(0, rn);
2812 tmp = gen_vfp_mrs();
2815 /* Set the 4 flag bits in the CPSR. */
2819 store_reg(s, rd, tmp);
2823 tmp = load_reg(s, rd);
2824 if (insn & (1 << 21)) {
2826 /* system register */
2831 /* Writes are ignored. */
2834 gen_helper_vfp_set_fpscr(cpu_env, tmp);
2841 /* TODO: VFP subarchitecture support.
2842 * For now, keep the EN bit only */
2843 tcg_gen_andi_i32(tmp, tmp, 1 << 30);
2844 store_cpu_field(tmp, vfp.xregs[rn]);
2847 case ARM_VFP_FPINST:
2848 case ARM_VFP_FPINST2:
2849 store_cpu_field(tmp, vfp.xregs[rn]);
2856 gen_mov_vreg_F0(0, rn);
2861 /* data processing */
2862 /* The opcode is in bits 23, 21, 20 and 6. */
2863 op = ((insn >> 20) & 8) | ((insn >> 19) & 6) | ((insn >> 6) & 1);
2867 rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1);
2869 /* rn is register number */
2870 VFP_DREG_N(rn, insn);
2873 if (op == 15 && (rn == 15 || ((rn & 0x1c) == 0x18))) {
2874 /* Integer or single precision destination. */
2875 rd = VFP_SREG_D(insn);
2877 VFP_DREG_D(rd, insn);
2880 (((rn & 0x1c) == 0x10) || ((rn & 0x14) == 0x14))) {
2881 /* VCVT from int is always from S reg regardless of dp bit.
2882 * VCVT with immediate frac_bits has same format as SREG_M
2884 rm = VFP_SREG_M(insn);
2886 VFP_DREG_M(rm, insn);
2889 rn = VFP_SREG_N(insn);
2890 if (op == 15 && rn == 15) {
2891 /* Double precision destination. */
2892 VFP_DREG_D(rd, insn);
2894 rd = VFP_SREG_D(insn);
2896 /* NB that we implicitly rely on the encoding for the frac_bits
2897 * in VCVT of fixed to float being the same as that of an SREG_M
2899 rm = VFP_SREG_M(insn);
2902 veclen = s->vec_len;
2903 if (op == 15 && rn > 3)
2906 /* Shut up compiler warnings. */
2917 /* Figure out what type of vector operation this is. */
2918 if ((rd & bank_mask) == 0) {
2923 delta_d = (s->vec_stride >> 1) + 1;
2925 delta_d = s->vec_stride + 1;
2927 if ((rm & bank_mask) == 0) {
2928 /* mixed scalar/vector */
2937 /* Load the initial operands. */
2942 /* Integer source */
2943 gen_mov_F0_vreg(0, rm);
2948 gen_mov_F0_vreg(dp, rd);
2949 gen_mov_F1_vreg(dp, rm);
2953 /* Compare with zero */
2954 gen_mov_F0_vreg(dp, rd);
2965 /* Source and destination the same. */
2966 gen_mov_F0_vreg(dp, rd);
2969 /* One source operand. */
2970 gen_mov_F0_vreg(dp, rm);
2974 /* Two source operands. */
2975 gen_mov_F0_vreg(dp, rn);
2976 gen_mov_F1_vreg(dp, rm);
2980 /* Perform the calculation. */
2982 case 0: /* mac: fd + (fn * fm) */
2984 gen_mov_F1_vreg(dp, rd);
2987 case 1: /* nmac: fd - (fn * fm) */
2990 gen_mov_F1_vreg(dp, rd);
2993 case 2: /* msc: -fd + (fn * fm) */
2995 gen_mov_F1_vreg(dp, rd);
2998 case 3: /* nmsc: -fd - (fn * fm) */
3001 gen_mov_F1_vreg(dp, rd);
3004 case 4: /* mul: fn * fm */
3007 case 5: /* nmul: -(fn * fm) */
3011 case 6: /* add: fn + fm */
3014 case 7: /* sub: fn - fm */
3017 case 8: /* div: fn / fm */
3020 case 14: /* fconst */
3021 if (!arm_feature(env, ARM_FEATURE_VFP3))
3024 n = (insn << 12) & 0x80000000;
3025 i = ((insn >> 12) & 0x70) | (insn & 0xf);
3032 tcg_gen_movi_i64(cpu_F0d, ((uint64_t)n) << 32);
3039 tcg_gen_movi_i32(cpu_F0s, n);
3042 case 15: /* extension space */
3056 case 4: /* vcvtb.f32.f16 */
3057 if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
3059 tmp = gen_vfp_mrs();
3060 tcg_gen_ext16u_i32(tmp, tmp);
3061 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp, cpu_env);
3064 case 5: /* vcvtt.f32.f16 */
3065 if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
3067 tmp = gen_vfp_mrs();
3068 tcg_gen_shri_i32(tmp, tmp, 16);
3069 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp, cpu_env);
3072 case 6: /* vcvtb.f16.f32 */
3073 if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
3076 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
3077 gen_mov_F0_vreg(0, rd);
3078 tmp2 = gen_vfp_mrs();
3079 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
3080 tcg_gen_or_i32(tmp, tmp, tmp2);
3084 case 7: /* vcvtt.f16.f32 */
3085 if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
3088 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
3089 tcg_gen_shli_i32(tmp, tmp, 16);
3090 gen_mov_F0_vreg(0, rd);
3091 tmp2 = gen_vfp_mrs();
3092 tcg_gen_ext16u_i32(tmp2, tmp2);
3093 tcg_gen_or_i32(tmp, tmp, tmp2);
3106 case 11: /* cmpez */
3110 case 15: /* single<->double conversion */
3112 gen_helper_vfp_fcvtsd(cpu_F0s, cpu_F0d, cpu_env);
3114 gen_helper_vfp_fcvtds(cpu_F0d, cpu_F0s, cpu_env);
3116 case 16: /* fuito */
3119 case 17: /* fsito */
3122 case 20: /* fshto */
3123 if (!arm_feature(env, ARM_FEATURE_VFP3))
3125 gen_vfp_shto(dp, 16 - rm);
3127 case 21: /* fslto */
3128 if (!arm_feature(env, ARM_FEATURE_VFP3))
3130 gen_vfp_slto(dp, 32 - rm);
3132 case 22: /* fuhto */
3133 if (!arm_feature(env, ARM_FEATURE_VFP3))
3135 gen_vfp_uhto(dp, 16 - rm);
3137 case 23: /* fulto */
3138 if (!arm_feature(env, ARM_FEATURE_VFP3))
3140 gen_vfp_ulto(dp, 32 - rm);
3142 case 24: /* ftoui */
3145 case 25: /* ftouiz */
3148 case 26: /* ftosi */
3151 case 27: /* ftosiz */
3154 case 28: /* ftosh */
3155 if (!arm_feature(env, ARM_FEATURE_VFP3))
3157 gen_vfp_tosh(dp, 16 - rm);
3159 case 29: /* ftosl */
3160 if (!arm_feature(env, ARM_FEATURE_VFP3))
3162 gen_vfp_tosl(dp, 32 - rm);
3164 case 30: /* ftouh */
3165 if (!arm_feature(env, ARM_FEATURE_VFP3))
3167 gen_vfp_touh(dp, 16 - rm);
3169 case 31: /* ftoul */
3170 if (!arm_feature(env, ARM_FEATURE_VFP3))
3172 gen_vfp_toul(dp, 32 - rm);
3174 default: /* undefined */
3175 printf ("rn:%d\n", rn);
3179 default: /* undefined */
3180 printf ("op:%d\n", op);
3184 /* Write back the result. */
3185 if (op == 15 && (rn >= 8 && rn <= 11))
3186 ; /* Comparison, do nothing. */
3187 else if (op == 15 && dp && ((rn & 0x1c) == 0x18))
3188 /* VCVT double to int: always integer result. */
3189 gen_mov_vreg_F0(0, rd);
3190 else if (op == 15 && rn == 15)
3192 gen_mov_vreg_F0(!dp, rd);
3194 gen_mov_vreg_F0(dp, rd);
3196 /* break out of the loop if we have finished */
3200 if (op == 15 && delta_m == 0) {
3201 /* single source one-many */
3203 rd = ((rd + delta_d) & (bank_mask - 1))
3205 gen_mov_vreg_F0(dp, rd);
3209 /* Setup the next operands. */
3211 rd = ((rd + delta_d) & (bank_mask - 1))
3215 /* One source operand. */
3216 rm = ((rm + delta_m) & (bank_mask - 1))
3218 gen_mov_F0_vreg(dp, rm);
3220 /* Two source operands. */
3221 rn = ((rn + delta_d) & (bank_mask - 1))
3223 gen_mov_F0_vreg(dp, rn);
3225 rm = ((rm + delta_m) & (bank_mask - 1))
3227 gen_mov_F1_vreg(dp, rm);
3235 if (dp && (insn & 0x03e00000) == 0x00400000) {
3236 /* two-register transfer */
3237 rn = (insn >> 16) & 0xf;
3238 rd = (insn >> 12) & 0xf;
3240 VFP_DREG_M(rm, insn);
3242 rm = VFP_SREG_M(insn);
3245 if (insn & ARM_CP_RW_BIT) {
3248 gen_mov_F0_vreg(0, rm * 2);
3249 tmp = gen_vfp_mrs();
3250 store_reg(s, rd, tmp);
3251 gen_mov_F0_vreg(0, rm * 2 + 1);
3252 tmp = gen_vfp_mrs();
3253 store_reg(s, rn, tmp);
3255 gen_mov_F0_vreg(0, rm);
3256 tmp = gen_vfp_mrs();
3257 store_reg(s, rn, tmp);
3258 gen_mov_F0_vreg(0, rm + 1);
3259 tmp = gen_vfp_mrs();
3260 store_reg(s, rd, tmp);
3265 tmp = load_reg(s, rd);
3267 gen_mov_vreg_F0(0, rm * 2);
3268 tmp = load_reg(s, rn);
3270 gen_mov_vreg_F0(0, rm * 2 + 1);
3272 tmp = load_reg(s, rn);
3274 gen_mov_vreg_F0(0, rm);
3275 tmp = load_reg(s, rd);
3277 gen_mov_vreg_F0(0, rm + 1);
3282 rn = (insn >> 16) & 0xf;
3284 VFP_DREG_D(rd, insn);
3286 rd = VFP_SREG_D(insn);
3287 if (s->thumb && rn == 15) {
3289 tcg_gen_movi_i32(addr, s->pc & ~2);
3291 addr = load_reg(s, rn);
3293 if ((insn & 0x01200000) == 0x01000000) {
3294 /* Single load/store */
3295 offset = (insn & 0xff) << 2;
3296 if ((insn & (1 << 23)) == 0)
3298 tcg_gen_addi_i32(addr, addr, offset);
3299 if (insn & (1 << 20)) {
3300 gen_vfp_ld(s, dp, addr);
3301 gen_mov_vreg_F0(dp, rd);
3303 gen_mov_F0_vreg(dp, rd);
3304 gen_vfp_st(s, dp, addr);
3308 /* load/store multiple */
3310 n = (insn >> 1) & 0x7f;
3314 if (insn & (1 << 24)) /* pre-decrement */
3315 tcg_gen_addi_i32(addr, addr, -((insn & 0xff) << 2));
3321 for (i = 0; i < n; i++) {
3322 if (insn & ARM_CP_RW_BIT) {
3324 gen_vfp_ld(s, dp, addr);
3325 gen_mov_vreg_F0(dp, rd + i);
3328 gen_mov_F0_vreg(dp, rd + i);
3329 gen_vfp_st(s, dp, addr);
3331 tcg_gen_addi_i32(addr, addr, offset);
3333 if (insn & (1 << 21)) {
3335 if (insn & (1 << 24))
3336 offset = -offset * n;
3337 else if (dp && (insn & 1))
3343 tcg_gen_addi_i32(addr, addr, offset);
3344 store_reg(s, rn, addr);
3352 /* Should never happen. */
3358 static inline void gen_goto_tb(DisasContext *s, int n, uint32_t dest)
3360 TranslationBlock *tb;
3363 if ((tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK)) {
3365 gen_set_pc_im(dest);
3366 tcg_gen_exit_tb((long)tb + n);
3368 gen_set_pc_im(dest);
3373 static inline void gen_jmp (DisasContext *s, uint32_t dest)
3375 if (unlikely(s->singlestep_enabled)) {
3376 /* An indirect jump so that we still trigger the debug exception. */
3381 gen_goto_tb(s, 0, dest);
3382 s->is_jmp = DISAS_TB_JUMP;
3386 static inline void gen_mulxy(TCGv t0, TCGv t1, int x, int y)
3389 tcg_gen_sari_i32(t0, t0, 16);
3393 tcg_gen_sari_i32(t1, t1, 16);
3396 tcg_gen_mul_i32(t0, t0, t1);
3399 /* Return the mask of PSR bits set by a MSR instruction. */
3400 static uint32_t msr_mask(CPUState *env, DisasContext *s, int flags, int spsr) {
3404 if (flags & (1 << 0))
3406 if (flags & (1 << 1))
3408 if (flags & (1 << 2))
3410 if (flags & (1 << 3))
3413 /* Mask out undefined bits. */
3414 mask &= ~CPSR_RESERVED;
3415 if (!arm_feature(env, ARM_FEATURE_V6))
3416 mask &= ~(CPSR_E | CPSR_GE);
3417 if (!arm_feature(env, ARM_FEATURE_THUMB2))
3419 /* Mask out execution state bits. */
3422 /* Mask out privileged bits. */
3428 /* Returns nonzero if access to the PSR is not permitted. Marks t0 as dead. */
3429 static int gen_set_psr(DisasContext *s, uint32_t mask, int spsr, TCGv t0)
3433 /* ??? This is also undefined in system mode. */
3437 tmp = load_cpu_field(spsr);
3438 tcg_gen_andi_i32(tmp, tmp, ~mask);
3439 tcg_gen_andi_i32(t0, t0, mask);
3440 tcg_gen_or_i32(tmp, tmp, t0);
3441 store_cpu_field(tmp, spsr);
3443 gen_set_cpsr(t0, mask);
3450 /* Returns nonzero if access to the PSR is not permitted. */
3451 static int gen_set_psr_im(DisasContext *s, uint32_t mask, int spsr, uint32_t val)
3455 tcg_gen_movi_i32(tmp, val);
3456 return gen_set_psr(s, mask, spsr, tmp);
3459 /* Generate an old-style exception return. Marks pc as dead. */
3460 static void gen_exception_return(DisasContext *s, TCGv pc)
3463 store_reg(s, 15, pc);
3464 tmp = load_cpu_field(spsr);
3465 gen_set_cpsr(tmp, 0xffffffff);
3467 s->is_jmp = DISAS_UPDATE;
3470 /* Generate a v6 exception return. Marks both values as dead. */
3471 static void gen_rfe(DisasContext *s, TCGv pc, TCGv cpsr)
3473 gen_set_cpsr(cpsr, 0xffffffff);
3475 store_reg(s, 15, pc);
3476 s->is_jmp = DISAS_UPDATE;
3480 gen_set_condexec (DisasContext *s)
3482 if (s->condexec_mask) {
3483 uint32_t val = (s->condexec_cond << 4) | (s->condexec_mask >> 1);
3484 TCGv tmp = new_tmp();
3485 tcg_gen_movi_i32(tmp, val);
3486 store_cpu_field(tmp, condexec_bits);
3490 static void gen_exception_insn(DisasContext *s, int offset, int excp)
3492 gen_set_condexec(s);
3493 gen_set_pc_im(s->pc - offset);
3494 gen_exception(excp);
3495 s->is_jmp = DISAS_JUMP;
3498 static void gen_nop_hint(DisasContext *s, int val)
3502 gen_set_pc_im(s->pc);
3503 s->is_jmp = DISAS_WFI;
3507 /* TODO: Implement SEV and WFE. May help SMP performance. */
3513 #define CPU_V001 cpu_V0, cpu_V0, cpu_V1
3515 static inline int gen_neon_add(int size, TCGv t0, TCGv t1)
3518 case 0: gen_helper_neon_add_u8(t0, t0, t1); break;
3519 case 1: gen_helper_neon_add_u16(t0, t0, t1); break;
3520 case 2: tcg_gen_add_i32(t0, t0, t1); break;
3526 static inline void gen_neon_rsb(int size, TCGv t0, TCGv t1)
3529 case 0: gen_helper_neon_sub_u8(t0, t1, t0); break;
3530 case 1: gen_helper_neon_sub_u16(t0, t1, t0); break;
3531 case 2: tcg_gen_sub_i32(t0, t1, t0); break;
3536 /* 32-bit pairwise ops end up the same as the elementwise versions. */
3537 #define gen_helper_neon_pmax_s32 gen_helper_neon_max_s32
3538 #define gen_helper_neon_pmax_u32 gen_helper_neon_max_u32
3539 #define gen_helper_neon_pmin_s32 gen_helper_neon_min_s32
3540 #define gen_helper_neon_pmin_u32 gen_helper_neon_min_u32
3542 /* FIXME: This is wrong. They set the wrong overflow bit. */
3543 #define gen_helper_neon_qadd_s32(a, e, b, c) gen_helper_add_saturate(a, b, c)
3544 #define gen_helper_neon_qadd_u32(a, e, b, c) gen_helper_add_usaturate(a, b, c)
3545 #define gen_helper_neon_qsub_s32(a, e, b, c) gen_helper_sub_saturate(a, b, c)
3546 #define gen_helper_neon_qsub_u32(a, e, b, c) gen_helper_sub_usaturate(a, b, c)
3548 #define GEN_NEON_INTEGER_OP_ENV(name) do { \
3549 switch ((size << 1) | u) { \
3551 gen_helper_neon_##name##_s8(tmp, cpu_env, tmp, tmp2); \
3554 gen_helper_neon_##name##_u8(tmp, cpu_env, tmp, tmp2); \
3557 gen_helper_neon_##name##_s16(tmp, cpu_env, tmp, tmp2); \
3560 gen_helper_neon_##name##_u16(tmp, cpu_env, tmp, tmp2); \
3563 gen_helper_neon_##name##_s32(tmp, cpu_env, tmp, tmp2); \
3566 gen_helper_neon_##name##_u32(tmp, cpu_env, tmp, tmp2); \
3568 default: return 1; \
3571 #define GEN_NEON_INTEGER_OP(name) do { \
3572 switch ((size << 1) | u) { \
3574 gen_helper_neon_##name##_s8(tmp, tmp, tmp2); \
3577 gen_helper_neon_##name##_u8(tmp, tmp, tmp2); \
3580 gen_helper_neon_##name##_s16(tmp, tmp, tmp2); \
3583 gen_helper_neon_##name##_u16(tmp, tmp, tmp2); \
3586 gen_helper_neon_##name##_s32(tmp, tmp, tmp2); \
3589 gen_helper_neon_##name##_u32(tmp, tmp, tmp2); \
3591 default: return 1; \
3594 static TCGv neon_load_scratch(int scratch)
3596 TCGv tmp = new_tmp();
3597 tcg_gen_ld_i32(tmp, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch]));
3601 static void neon_store_scratch(int scratch, TCGv var)
3603 tcg_gen_st_i32(var, cpu_env, offsetof(CPUARMState, vfp.scratch[scratch]));
3607 static inline TCGv neon_get_scalar(int size, int reg)
3611 tmp = neon_load_reg(reg >> 1, reg & 1);
3613 tmp = neon_load_reg(reg >> 2, (reg >> 1) & 1);
3615 gen_neon_dup_low16(tmp);
3617 gen_neon_dup_high16(tmp);
3623 static void gen_neon_unzip_u8(TCGv t0, TCGv t1)
3631 tcg_gen_andi_i32(rd, t0, 0xff);
3632 tcg_gen_shri_i32(tmp, t0, 8);
3633 tcg_gen_andi_i32(tmp, tmp, 0xff00);
3634 tcg_gen_or_i32(rd, rd, tmp);
3635 tcg_gen_shli_i32(tmp, t1, 16);
3636 tcg_gen_andi_i32(tmp, tmp, 0xff0000);
3637 tcg_gen_or_i32(rd, rd, tmp);
3638 tcg_gen_shli_i32(tmp, t1, 8);
3639 tcg_gen_andi_i32(tmp, tmp, 0xff000000);
3640 tcg_gen_or_i32(rd, rd, tmp);
3642 tcg_gen_shri_i32(rm, t0, 8);
3643 tcg_gen_andi_i32(rm, rm, 0xff);
3644 tcg_gen_shri_i32(tmp, t0, 16);
3645 tcg_gen_andi_i32(tmp, tmp, 0xff00);
3646 tcg_gen_or_i32(rm, rm, tmp);
3647 tcg_gen_shli_i32(tmp, t1, 8);
3648 tcg_gen_andi_i32(tmp, tmp, 0xff0000);
3649 tcg_gen_or_i32(rm, rm, tmp);
3650 tcg_gen_andi_i32(tmp, t1, 0xff000000);
3651 tcg_gen_or_i32(t1, rm, tmp);
3652 tcg_gen_mov_i32(t0, rd);
3659 static void gen_neon_zip_u8(TCGv t0, TCGv t1)
3667 tcg_gen_andi_i32(rd, t0, 0xff);
3668 tcg_gen_shli_i32(tmp, t1, 8);
3669 tcg_gen_andi_i32(tmp, tmp, 0xff00);
3670 tcg_gen_or_i32(rd, rd, tmp);
3671 tcg_gen_shli_i32(tmp, t0, 16);
3672 tcg_gen_andi_i32(tmp, tmp, 0xff0000);
3673 tcg_gen_or_i32(rd, rd, tmp);
3674 tcg_gen_shli_i32(tmp, t1, 24);
3675 tcg_gen_andi_i32(tmp, tmp, 0xff000000);
3676 tcg_gen_or_i32(rd, rd, tmp);
3678 tcg_gen_andi_i32(rm, t1, 0xff000000);
3679 tcg_gen_shri_i32(tmp, t0, 8);
3680 tcg_gen_andi_i32(tmp, tmp, 0xff0000);
3681 tcg_gen_or_i32(rm, rm, tmp);
3682 tcg_gen_shri_i32(tmp, t1, 8);
3683 tcg_gen_andi_i32(tmp, tmp, 0xff00);
3684 tcg_gen_or_i32(rm, rm, tmp);
3685 tcg_gen_shri_i32(tmp, t0, 16);
3686 tcg_gen_andi_i32(tmp, tmp, 0xff);
3687 tcg_gen_or_i32(t1, rm, tmp);
3688 tcg_gen_mov_i32(t0, rd);
3695 static void gen_neon_zip_u16(TCGv t0, TCGv t1)
3702 tcg_gen_andi_i32(tmp, t0, 0xffff);
3703 tcg_gen_shli_i32(tmp2, t1, 16);
3704 tcg_gen_or_i32(tmp, tmp, tmp2);
3705 tcg_gen_andi_i32(t1, t1, 0xffff0000);
3706 tcg_gen_shri_i32(tmp2, t0, 16);
3707 tcg_gen_or_i32(t1, t1, tmp2);
3708 tcg_gen_mov_i32(t0, tmp);
3714 static void gen_neon_unzip(int reg, int q, int tmp, int size)
3719 for (n = 0; n < q + 1; n += 2) {
3720 t0 = neon_load_reg(reg, n);
3721 t1 = neon_load_reg(reg, n + 1);
3723 case 0: gen_neon_unzip_u8(t0, t1); break;
3724 case 1: gen_neon_zip_u16(t0, t1); break; /* zip and unzip are the same. */
3725 case 2: /* no-op */; break;
3728 neon_store_scratch(tmp + n, t0);
3729 neon_store_scratch(tmp + n + 1, t1);
3733 static void gen_neon_trn_u8(TCGv t0, TCGv t1)
3740 tcg_gen_shli_i32(rd, t0, 8);
3741 tcg_gen_andi_i32(rd, rd, 0xff00ff00);
3742 tcg_gen_andi_i32(tmp, t1, 0x00ff00ff);
3743 tcg_gen_or_i32(rd, rd, tmp);
3745 tcg_gen_shri_i32(t1, t1, 8);
3746 tcg_gen_andi_i32(t1, t1, 0x00ff00ff);
3747 tcg_gen_andi_i32(tmp, t0, 0xff00ff00);
3748 tcg_gen_or_i32(t1, t1, tmp);
3749 tcg_gen_mov_i32(t0, rd);
3755 static void gen_neon_trn_u16(TCGv t0, TCGv t1)
3762 tcg_gen_shli_i32(rd, t0, 16);
3763 tcg_gen_andi_i32(tmp, t1, 0xffff);
3764 tcg_gen_or_i32(rd, rd, tmp);
3765 tcg_gen_shri_i32(t1, t1, 16);
3766 tcg_gen_andi_i32(tmp, t0, 0xffff0000);
3767 tcg_gen_or_i32(t1, t1, tmp);
3768 tcg_gen_mov_i32(t0, rd);
3779 } neon_ls_element_type[11] = {
3793 /* Translate a NEON load/store element instruction. Return nonzero if the
3794 instruction is invalid. */
3795 static int disas_neon_ls_insn(CPUState * env, DisasContext *s, uint32_t insn)
3814 if (!s->vfp_enabled)
3816 VFP_DREG_D(rd, insn);
3817 rn = (insn >> 16) & 0xf;
3819 load = (insn & (1 << 21)) != 0;
3821 if ((insn & (1 << 23)) == 0) {
3822 /* Load store all elements. */
3823 op = (insn >> 8) & 0xf;
3824 size = (insn >> 6) & 3;
3827 nregs = neon_ls_element_type[op].nregs;
3828 interleave = neon_ls_element_type[op].interleave;
3829 spacing = neon_ls_element_type[op].spacing;
3830 if (size == 3 && (interleave | spacing) != 1)
3832 load_reg_var(s, addr, rn);
3833 stride = (1 << size) * interleave;
3834 for (reg = 0; reg < nregs; reg++) {
3835 if (interleave > 2 || (interleave == 2 && nregs == 2)) {
3836 load_reg_var(s, addr, rn);
3837 tcg_gen_addi_i32(addr, addr, (1 << size) * reg);
3838 } else if (interleave == 2 && nregs == 4 && reg == 2) {
3839 load_reg_var(s, addr, rn);
3840 tcg_gen_addi_i32(addr, addr, 1 << size);
3844 tmp64 = gen_ld64(addr, IS_USER(s));
3845 neon_store_reg64(tmp64, rd);
3846 tcg_temp_free_i64(tmp64);
3848 tmp64 = tcg_temp_new_i64();
3849 neon_load_reg64(tmp64, rd);
3850 gen_st64(tmp64, addr, IS_USER(s));
3852 tcg_gen_addi_i32(addr, addr, stride);
3854 for (pass = 0; pass < 2; pass++) {
3857 tmp = gen_ld32(addr, IS_USER(s));
3858 neon_store_reg(rd, pass, tmp);
3860 tmp = neon_load_reg(rd, pass);
3861 gen_st32(tmp, addr, IS_USER(s));
3863 tcg_gen_addi_i32(addr, addr, stride);
3864 } else if (size == 1) {
3866 tmp = gen_ld16u(addr, IS_USER(s));
3867 tcg_gen_addi_i32(addr, addr, stride);
3868 tmp2 = gen_ld16u(addr, IS_USER(s));
3869 tcg_gen_addi_i32(addr, addr, stride);
3870 tcg_gen_shli_i32(tmp2, tmp2, 16);
3871 tcg_gen_or_i32(tmp, tmp, tmp2);
3873 neon_store_reg(rd, pass, tmp);
3875 tmp = neon_load_reg(rd, pass);
3877 tcg_gen_shri_i32(tmp2, tmp, 16);
3878 gen_st16(tmp, addr, IS_USER(s));
3879 tcg_gen_addi_i32(addr, addr, stride);
3880 gen_st16(tmp2, addr, IS_USER(s));
3881 tcg_gen_addi_i32(addr, addr, stride);
3883 } else /* size == 0 */ {
3886 for (n = 0; n < 4; n++) {
3887 tmp = gen_ld8u(addr, IS_USER(s));
3888 tcg_gen_addi_i32(addr, addr, stride);
3892 tcg_gen_shli_i32(tmp, tmp, n * 8);
3893 tcg_gen_or_i32(tmp2, tmp2, tmp);
3897 neon_store_reg(rd, pass, tmp2);
3899 tmp2 = neon_load_reg(rd, pass);
3900 for (n = 0; n < 4; n++) {
3903 tcg_gen_mov_i32(tmp, tmp2);
3905 tcg_gen_shri_i32(tmp, tmp2, n * 8);
3907 gen_st8(tmp, addr, IS_USER(s));
3908 tcg_gen_addi_i32(addr, addr, stride);
3919 size = (insn >> 10) & 3;
3921 /* Load single element to all lanes. */
3924 size = (insn >> 6) & 3;
3925 nregs = ((insn >> 8) & 3) + 1;
3926 stride = (insn & (1 << 5)) ? 2 : 1;
3927 load_reg_var(s, addr, rn);
3928 for (reg = 0; reg < nregs; reg++) {
3931 tmp = gen_ld8u(addr, IS_USER(s));
3932 gen_neon_dup_u8(tmp, 0);
3935 tmp = gen_ld16u(addr, IS_USER(s));
3936 gen_neon_dup_low16(tmp);
3939 tmp = gen_ld32(addr, IS_USER(s));
3943 default: /* Avoid compiler warnings. */
3946 tcg_gen_addi_i32(addr, addr, 1 << size);
3948 tcg_gen_mov_i32(tmp2, tmp);
3949 neon_store_reg(rd, 0, tmp2);
3950 neon_store_reg(rd, 1, tmp);
3953 stride = (1 << size) * nregs;
3955 /* Single element. */
3956 pass = (insn >> 7) & 1;
3959 shift = ((insn >> 5) & 3) * 8;
3963 shift = ((insn >> 6) & 1) * 16;
3964 stride = (insn & (1 << 5)) ? 2 : 1;
3968 stride = (insn & (1 << 6)) ? 2 : 1;
3973 nregs = ((insn >> 8) & 3) + 1;
3974 load_reg_var(s, addr, rn);
3975 for (reg = 0; reg < nregs; reg++) {
3979 tmp = gen_ld8u(addr, IS_USER(s));
3982 tmp = gen_ld16u(addr, IS_USER(s));
3985 tmp = gen_ld32(addr, IS_USER(s));
3987 default: /* Avoid compiler warnings. */
3991 tmp2 = neon_load_reg(rd, pass);
3992 gen_bfi(tmp, tmp2, tmp, shift, size ? 0xffff : 0xff);
3995 neon_store_reg(rd, pass, tmp);
3996 } else { /* Store */
3997 tmp = neon_load_reg(rd, pass);
3999 tcg_gen_shri_i32(tmp, tmp, shift);
4002 gen_st8(tmp, addr, IS_USER(s));
4005 gen_st16(tmp, addr, IS_USER(s));
4008 gen_st32(tmp, addr, IS_USER(s));
4013 tcg_gen_addi_i32(addr, addr, 1 << size);
4015 stride = nregs * (1 << size);
4022 base = load_reg(s, rn);
4024 tcg_gen_addi_i32(base, base, stride);
4027 index = load_reg(s, rm);
4028 tcg_gen_add_i32(base, base, index);
4031 store_reg(s, rn, base);
4036 /* Bitwise select. dest = c ? t : f. Clobbers T and F. */
4037 static void gen_neon_bsl(TCGv dest, TCGv t, TCGv f, TCGv c)
4039 tcg_gen_and_i32(t, t, c);
4040 tcg_gen_andc_i32(f, f, c);
4041 tcg_gen_or_i32(dest, t, f);
4044 static inline void gen_neon_narrow(int size, TCGv dest, TCGv_i64 src)
4047 case 0: gen_helper_neon_narrow_u8(dest, src); break;
4048 case 1: gen_helper_neon_narrow_u16(dest, src); break;
4049 case 2: tcg_gen_trunc_i64_i32(dest, src); break;
4054 static inline void gen_neon_narrow_sats(int size, TCGv dest, TCGv_i64 src)
4057 case 0: gen_helper_neon_narrow_sat_s8(dest, cpu_env, src); break;
4058 case 1: gen_helper_neon_narrow_sat_s16(dest, cpu_env, src); break;
4059 case 2: gen_helper_neon_narrow_sat_s32(dest, cpu_env, src); break;
4064 static inline void gen_neon_narrow_satu(int size, TCGv dest, TCGv_i64 src)
4067 case 0: gen_helper_neon_narrow_sat_u8(dest, cpu_env, src); break;
4068 case 1: gen_helper_neon_narrow_sat_u16(dest, cpu_env, src); break;
4069 case 2: gen_helper_neon_narrow_sat_u32(dest, cpu_env, src); break;
4074 static inline void gen_neon_shift_narrow(int size, TCGv var, TCGv shift,
4080 case 1: gen_helper_neon_rshl_u16(var, var, shift); break;
4081 case 2: gen_helper_neon_rshl_u32(var, var, shift); break;
4086 case 1: gen_helper_neon_rshl_s16(var, var, shift); break;
4087 case 2: gen_helper_neon_rshl_s32(var, var, shift); break;
4094 case 1: gen_helper_neon_rshl_u16(var, var, shift); break;
4095 case 2: gen_helper_neon_rshl_u32(var, var, shift); break;
4100 case 1: gen_helper_neon_shl_s16(var, var, shift); break;
4101 case 2: gen_helper_neon_shl_s32(var, var, shift); break;
4108 static inline void gen_neon_widen(TCGv_i64 dest, TCGv src, int size, int u)
4112 case 0: gen_helper_neon_widen_u8(dest, src); break;
4113 case 1: gen_helper_neon_widen_u16(dest, src); break;
4114 case 2: tcg_gen_extu_i32_i64(dest, src); break;
4119 case 0: gen_helper_neon_widen_s8(dest, src); break;
4120 case 1: gen_helper_neon_widen_s16(dest, src); break;
4121 case 2: tcg_gen_ext_i32_i64(dest, src); break;
4128 static inline void gen_neon_addl(int size)
4131 case 0: gen_helper_neon_addl_u16(CPU_V001); break;
4132 case 1: gen_helper_neon_addl_u32(CPU_V001); break;
4133 case 2: tcg_gen_add_i64(CPU_V001); break;
4138 static inline void gen_neon_subl(int size)
4141 case 0: gen_helper_neon_subl_u16(CPU_V001); break;
4142 case 1: gen_helper_neon_subl_u32(CPU_V001); break;
4143 case 2: tcg_gen_sub_i64(CPU_V001); break;
4148 static inline void gen_neon_negl(TCGv_i64 var, int size)
4151 case 0: gen_helper_neon_negl_u16(var, var); break;
4152 case 1: gen_helper_neon_negl_u32(var, var); break;
4153 case 2: gen_helper_neon_negl_u64(var, var); break;
4158 static inline void gen_neon_addl_saturate(TCGv_i64 op0, TCGv_i64 op1, int size)
4161 case 1: gen_helper_neon_addl_saturate_s32(op0, cpu_env, op0, op1); break;
4162 case 2: gen_helper_neon_addl_saturate_s64(op0, cpu_env, op0, op1); break;
4167 static inline void gen_neon_mull(TCGv_i64 dest, TCGv a, TCGv b, int size, int u)
4171 switch ((size << 1) | u) {
4172 case 0: gen_helper_neon_mull_s8(dest, a, b); break;
4173 case 1: gen_helper_neon_mull_u8(dest, a, b); break;
4174 case 2: gen_helper_neon_mull_s16(dest, a, b); break;
4175 case 3: gen_helper_neon_mull_u16(dest, a, b); break;
4177 tmp = gen_muls_i64_i32(a, b);
4178 tcg_gen_mov_i64(dest, tmp);
4181 tmp = gen_mulu_i64_i32(a, b);
4182 tcg_gen_mov_i64(dest, tmp);
4188 /* Translate a NEON data processing instruction. Return nonzero if the
4189 instruction is invalid.
4190 We process data in a mixture of 32-bit and 64-bit chunks.
4191 Mostly we use 32-bit chunks so we can use normal scalar instructions. */
4193 static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
4206 TCGv tmp, tmp2, tmp3, tmp4, tmp5;
4209 if (!s->vfp_enabled)
4211 q = (insn & (1 << 6)) != 0;
4212 u = (insn >> 24) & 1;
4213 VFP_DREG_D(rd, insn);
4214 VFP_DREG_N(rn, insn);
4215 VFP_DREG_M(rm, insn);
4216 size = (insn >> 20) & 3;
4217 if ((insn & (1 << 23)) == 0) {
4218 /* Three register same length. */
4219 op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1);
4220 if (size == 3 && (op == 1 || op == 5 || op == 8 || op == 9
4221 || op == 10 || op == 11 || op == 16)) {
4222 /* 64-bit element instructions. */
4223 for (pass = 0; pass < (q ? 2 : 1); pass++) {
4224 neon_load_reg64(cpu_V0, rn + pass);
4225 neon_load_reg64(cpu_V1, rm + pass);
4229 gen_helper_neon_add_saturate_u64(CPU_V001);
4231 gen_helper_neon_add_saturate_s64(CPU_V001);
4236 gen_helper_neon_sub_saturate_u64(CPU_V001);
4238 gen_helper_neon_sub_saturate_s64(CPU_V001);
4243 gen_helper_neon_shl_u64(cpu_V0, cpu_V1, cpu_V0);
4245 gen_helper_neon_shl_s64(cpu_V0, cpu_V1, cpu_V0);
4250 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
4253 gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
4257 case 10: /* VRSHL */
4259 gen_helper_neon_rshl_u64(cpu_V0, cpu_V1, cpu_V0);
4261 gen_helper_neon_rshl_s64(cpu_V0, cpu_V1, cpu_V0);
4264 case 11: /* VQRSHL */
4266 gen_helper_neon_qrshl_u64(cpu_V0, cpu_env,
4269 gen_helper_neon_qrshl_s64(cpu_V0, cpu_env,
4275 tcg_gen_sub_i64(CPU_V001);
4277 tcg_gen_add_i64(CPU_V001);
4283 neon_store_reg64(cpu_V0, rd + pass);
4290 case 10: /* VRSHL */
4291 case 11: /* VQRSHL */
4294 /* Shift instruction operands are reversed. */
4301 case 20: /* VPMAX */
4302 case 21: /* VPMIN */
4303 case 23: /* VPADD */
4306 case 26: /* VPADD (float) */
4307 pairwise = (u && size < 2);
4309 case 30: /* VPMIN/VPMAX (float) */
4317 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4326 tmp = neon_load_reg(rn, n);
4327 tmp2 = neon_load_reg(rn, n + 1);
4329 tmp = neon_load_reg(rm, n);
4330 tmp2 = neon_load_reg(rm, n + 1);
4334 tmp = neon_load_reg(rn, pass);
4335 tmp2 = neon_load_reg(rm, pass);
4339 GEN_NEON_INTEGER_OP(hadd);
4342 GEN_NEON_INTEGER_OP_ENV(qadd);
4344 case 2: /* VRHADD */
4345 GEN_NEON_INTEGER_OP(rhadd);
4347 case 3: /* Logic ops. */
4348 switch ((u << 2) | size) {
4350 tcg_gen_and_i32(tmp, tmp, tmp2);
4353 tcg_gen_andc_i32(tmp, tmp, tmp2);
4356 tcg_gen_or_i32(tmp, tmp, tmp2);
4359 tcg_gen_orc_i32(tmp, tmp, tmp2);
4362 tcg_gen_xor_i32(tmp, tmp, tmp2);
4365 tmp3 = neon_load_reg(rd, pass);
4366 gen_neon_bsl(tmp, tmp, tmp2, tmp3);
4370 tmp3 = neon_load_reg(rd, pass);
4371 gen_neon_bsl(tmp, tmp, tmp3, tmp2);
4375 tmp3 = neon_load_reg(rd, pass);
4376 gen_neon_bsl(tmp, tmp3, tmp, tmp2);
4382 GEN_NEON_INTEGER_OP(hsub);
4385 GEN_NEON_INTEGER_OP_ENV(qsub);
4388 GEN_NEON_INTEGER_OP(cgt);
4391 GEN_NEON_INTEGER_OP(cge);
4394 GEN_NEON_INTEGER_OP(shl);
4397 GEN_NEON_INTEGER_OP_ENV(qshl);
4399 case 10: /* VRSHL */
4400 GEN_NEON_INTEGER_OP(rshl);
4402 case 11: /* VQRSHL */
4403 GEN_NEON_INTEGER_OP_ENV(qrshl);
4406 GEN_NEON_INTEGER_OP(max);
4409 GEN_NEON_INTEGER_OP(min);
4412 GEN_NEON_INTEGER_OP(abd);
4415 GEN_NEON_INTEGER_OP(abd);
4417 tmp2 = neon_load_reg(rd, pass);
4418 gen_neon_add(size, tmp, tmp2);
4421 if (!u) { /* VADD */
4422 if (gen_neon_add(size, tmp, tmp2))
4426 case 0: gen_helper_neon_sub_u8(tmp, tmp, tmp2); break;
4427 case 1: gen_helper_neon_sub_u16(tmp, tmp, tmp2); break;
4428 case 2: tcg_gen_sub_i32(tmp, tmp, tmp2); break;
4434 if (!u) { /* VTST */
4436 case 0: gen_helper_neon_tst_u8(tmp, tmp, tmp2); break;
4437 case 1: gen_helper_neon_tst_u16(tmp, tmp, tmp2); break;
4438 case 2: gen_helper_neon_tst_u32(tmp, tmp, tmp2); break;
4443 case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break;
4444 case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break;
4445 case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break;
4450 case 18: /* Multiply. */
4452 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
4453 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
4454 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
4458 tmp2 = neon_load_reg(rd, pass);
4460 gen_neon_rsb(size, tmp, tmp2);
4462 gen_neon_add(size, tmp, tmp2);
4466 if (u) { /* polynomial */
4467 gen_helper_neon_mul_p8(tmp, tmp, tmp2);
4468 } else { /* Integer */
4470 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
4471 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
4472 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
4477 case 20: /* VPMAX */
4478 GEN_NEON_INTEGER_OP(pmax);
4480 case 21: /* VPMIN */
4481 GEN_NEON_INTEGER_OP(pmin);
4483 case 22: /* Hultiply high. */
4484 if (!u) { /* VQDMULH */
4486 case 1: gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2); break;
4487 case 2: gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2); break;
4490 } else { /* VQRDHMUL */
4492 case 1: gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2); break;
4493 case 2: gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2); break;
4498 case 23: /* VPADD */
4502 case 0: gen_helper_neon_padd_u8(tmp, tmp, tmp2); break;
4503 case 1: gen_helper_neon_padd_u16(tmp, tmp, tmp2); break;
4504 case 2: tcg_gen_add_i32(tmp, tmp, tmp2); break;
4508 case 26: /* Floating point arithnetic. */
4509 switch ((u << 2) | size) {
4511 gen_helper_neon_add_f32(tmp, tmp, tmp2);
4514 gen_helper_neon_sub_f32(tmp, tmp, tmp2);
4517 gen_helper_neon_add_f32(tmp, tmp, tmp2);
4520 gen_helper_neon_abd_f32(tmp, tmp, tmp2);
4526 case 27: /* Float multiply. */
4527 gen_helper_neon_mul_f32(tmp, tmp, tmp2);
4530 tmp2 = neon_load_reg(rd, pass);
4532 gen_helper_neon_add_f32(tmp, tmp, tmp2);
4534 gen_helper_neon_sub_f32(tmp, tmp2, tmp);
4538 case 28: /* Float compare. */
4540 gen_helper_neon_ceq_f32(tmp, tmp, tmp2);
4543 gen_helper_neon_cge_f32(tmp, tmp, tmp2);
4545 gen_helper_neon_cgt_f32(tmp, tmp, tmp2);
4548 case 29: /* Float compare absolute. */
4552 gen_helper_neon_acge_f32(tmp, tmp, tmp2);
4554 gen_helper_neon_acgt_f32(tmp, tmp, tmp2);
4556 case 30: /* Float min/max. */
4558 gen_helper_neon_max_f32(tmp, tmp, tmp2);
4560 gen_helper_neon_min_f32(tmp, tmp, tmp2);
4564 gen_helper_recps_f32(tmp, tmp, tmp2, cpu_env);
4566 gen_helper_rsqrts_f32(tmp, tmp, tmp2, cpu_env);
4573 /* Save the result. For elementwise operations we can put it
4574 straight into the destination register. For pairwise operations
4575 we have to be careful to avoid clobbering the source operands. */
4576 if (pairwise && rd == rm) {
4577 neon_store_scratch(pass, tmp);
4579 neon_store_reg(rd, pass, tmp);
4583 if (pairwise && rd == rm) {
4584 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4585 tmp = neon_load_scratch(pass);
4586 neon_store_reg(rd, pass, tmp);
4589 /* End of 3 register same size operations. */
4590 } else if (insn & (1 << 4)) {
4591 if ((insn & 0x00380080) != 0) {
4592 /* Two registers and shift. */
4593 op = (insn >> 8) & 0xf;
4594 if (insn & (1 << 7)) {
4599 while ((insn & (1 << (size + 19))) == 0)
4602 shift = (insn >> 16) & ((1 << (3 + size)) - 1);
4603 /* To avoid excessive dumplication of ops we implement shift
4604 by immediate using the variable shift operations. */
4606 /* Shift by immediate:
4607 VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
4608 /* Right shifts are encoded as N - shift, where N is the
4609 element size in bits. */
4611 shift = shift - (1 << (size + 3));
4619 imm = (uint8_t) shift;
4624 imm = (uint16_t) shift;
4635 for (pass = 0; pass < count; pass++) {
4637 neon_load_reg64(cpu_V0, rm + pass);
4638 tcg_gen_movi_i64(cpu_V1, imm);
4643 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
4645 gen_helper_neon_shl_s64(cpu_V0, cpu_V0, cpu_V1);
4650 gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, cpu_V1);
4652 gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, cpu_V1);
4657 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
4659 case 5: /* VSHL, VSLI */
4660 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, cpu_V1);
4662 case 6: /* VQSHLU */
4664 gen_helper_neon_qshlu_s64(cpu_V0, cpu_env,
4672 gen_helper_neon_qshl_u64(cpu_V0, cpu_env,
4675 gen_helper_neon_qshl_s64(cpu_V0, cpu_env,
4680 if (op == 1 || op == 3) {
4682 neon_load_reg64(cpu_V0, rd + pass);
4683 tcg_gen_add_i64(cpu_V0, cpu_V0, cpu_V1);
4684 } else if (op == 4 || (op == 5 && u)) {
4686 cpu_abort(env, "VS[LR]I.64 not implemented");
4688 neon_store_reg64(cpu_V0, rd + pass);
4689 } else { /* size < 3 */
4690 /* Operands in T0 and T1. */
4691 tmp = neon_load_reg(rm, pass);
4693 tcg_gen_movi_i32(tmp2, imm);
4697 GEN_NEON_INTEGER_OP(shl);
4701 GEN_NEON_INTEGER_OP(rshl);
4706 GEN_NEON_INTEGER_OP(shl);
4708 case 5: /* VSHL, VSLI */
4710 case 0: gen_helper_neon_shl_u8(tmp, tmp, tmp2); break;
4711 case 1: gen_helper_neon_shl_u16(tmp, tmp, tmp2); break;
4712 case 2: gen_helper_neon_shl_u32(tmp, tmp, tmp2); break;
4716 case 6: /* VQSHLU */
4722 gen_helper_neon_qshlu_s8(tmp, cpu_env,
4726 gen_helper_neon_qshlu_s16(tmp, cpu_env,
4730 gen_helper_neon_qshlu_s32(tmp, cpu_env,
4738 GEN_NEON_INTEGER_OP_ENV(qshl);
4743 if (op == 1 || op == 3) {
4745 tmp2 = neon_load_reg(rd, pass);
4746 gen_neon_add(size, tmp2, tmp);
4748 } else if (op == 4 || (op == 5 && u)) {
4753 mask = 0xff >> -shift;
4755 mask = (uint8_t)(0xff << shift);
4761 mask = 0xffff >> -shift;
4763 mask = (uint16_t)(0xffff << shift);
4767 if (shift < -31 || shift > 31) {
4771 mask = 0xffffffffu >> -shift;
4773 mask = 0xffffffffu << shift;
4779 tmp2 = neon_load_reg(rd, pass);
4780 tcg_gen_andi_i32(tmp, tmp, mask);
4781 tcg_gen_andi_i32(tmp2, tmp2, ~mask);
4782 tcg_gen_or_i32(tmp, tmp, tmp2);
4785 neon_store_reg(rd, pass, tmp);
4788 } else if (op < 10) {
4789 /* Shift by immediate and narrow:
4790 VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
4791 shift = shift - (1 << (size + 3));
4795 imm = (uint16_t)shift;
4797 tmp2 = tcg_const_i32(imm);
4798 TCGV_UNUSED_I64(tmp64);
4801 imm = (uint32_t)shift;
4802 tmp2 = tcg_const_i32(imm);
4803 TCGV_UNUSED_I64(tmp64);
4806 tmp64 = tcg_const_i64(shift);
4813 for (pass = 0; pass < 2; pass++) {
4815 neon_load_reg64(cpu_V0, rm + pass);
4818 gen_helper_neon_rshl_u64(cpu_V0, cpu_V0, tmp64);
4820 gen_helper_neon_rshl_s64(cpu_V0, cpu_V0, tmp64);
4823 gen_helper_neon_shl_u64(cpu_V0, cpu_V0, tmp64);
4825 gen_helper_neon_shl_s64(cpu_V0, cpu_V0, tmp64);
4828 tmp = neon_load_reg(rm + pass, 0);
4829 gen_neon_shift_narrow(size, tmp, tmp2, q, u);
4830 tmp3 = neon_load_reg(rm + pass, 1);
4831 gen_neon_shift_narrow(size, tmp3, tmp2, q, u);
4832 tcg_gen_concat_i32_i64(cpu_V0, tmp, tmp3);
4837 if (op == 8 && !u) {
4838 gen_neon_narrow(size - 1, tmp, cpu_V0);
4841 gen_neon_narrow_sats(size - 1, tmp, cpu_V0);
4843 gen_neon_narrow_satu(size - 1, tmp, cpu_V0);
4845 neon_store_reg(rd, pass, tmp);
4848 tcg_temp_free_i64(tmp64);
4852 } else if (op == 10) {
4856 tmp = neon_load_reg(rm, 0);
4857 tmp2 = neon_load_reg(rm, 1);
4858 for (pass = 0; pass < 2; pass++) {
4862 gen_neon_widen(cpu_V0, tmp, size, u);
4865 /* The shift is less than the width of the source
4866 type, so we can just shift the whole register. */
4867 tcg_gen_shli_i64(cpu_V0, cpu_V0, shift);
4868 if (size < 2 || !u) {
4871 imm = (0xffu >> (8 - shift));
4874 imm = 0xffff >> (16 - shift);
4876 imm64 = imm | (((uint64_t)imm) << 32);
4877 tcg_gen_andi_i64(cpu_V0, cpu_V0, imm64);
4880 neon_store_reg64(cpu_V0, rd + pass);
4882 } else if (op >= 14) {
4883 /* VCVT fixed-point. */
4884 /* We have already masked out the must-be-1 top bit of imm6,
4885 * hence this 32-shift where the ARM ARM has 64-imm6.
4888 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4889 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, pass));
4892 gen_vfp_ulto(0, shift);
4894 gen_vfp_slto(0, shift);
4897 gen_vfp_toul(0, shift);
4899 gen_vfp_tosl(0, shift);
4901 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, pass));
4906 } else { /* (insn & 0x00380080) == 0 */
4909 op = (insn >> 8) & 0xf;
4910 /* One register and immediate. */
4911 imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf);
4912 invert = (insn & (1 << 5)) != 0;
4930 imm = (imm << 8) | (imm << 24);
4933 imm = (imm << 8) | 0xff;
4936 imm = (imm << 16) | 0xffff;
4939 imm |= (imm << 8) | (imm << 16) | (imm << 24);
4944 imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19)
4945 | ((imm & 0x40) ? (0x1f << 25) : (1 << 30));
4951 for (pass = 0; pass < (q ? 4 : 2); pass++) {
4952 if (op & 1 && op < 12) {
4953 tmp = neon_load_reg(rd, pass);
4955 /* The immediate value has already been inverted, so
4957 tcg_gen_andi_i32(tmp, tmp, imm);
4959 tcg_gen_ori_i32(tmp, tmp, imm);
4964 if (op == 14 && invert) {
4967 for (n = 0; n < 4; n++) {
4968 if (imm & (1 << (n + (pass & 1) * 4)))
4969 val |= 0xff << (n * 8);
4971 tcg_gen_movi_i32(tmp, val);
4973 tcg_gen_movi_i32(tmp, imm);
4976 neon_store_reg(rd, pass, tmp);
4979 } else { /* (insn & 0x00800010 == 0x00800000) */
4981 op = (insn >> 8) & 0xf;
4982 if ((insn & (1 << 6)) == 0) {
4983 /* Three registers of different lengths. */
4987 /* prewiden, src1_wide, src2_wide */
4988 static const int neon_3reg_wide[16][3] = {
4989 {1, 0, 0}, /* VADDL */
4990 {1, 1, 0}, /* VADDW */
4991 {1, 0, 0}, /* VSUBL */
4992 {1, 1, 0}, /* VSUBW */
4993 {0, 1, 1}, /* VADDHN */
4994 {0, 0, 0}, /* VABAL */
4995 {0, 1, 1}, /* VSUBHN */
4996 {0, 0, 0}, /* VABDL */
4997 {0, 0, 0}, /* VMLAL */
4998 {0, 0, 0}, /* VQDMLAL */
4999 {0, 0, 0}, /* VMLSL */
5000 {0, 0, 0}, /* VQDMLSL */
5001 {0, 0, 0}, /* Integer VMULL */
5002 {0, 0, 0}, /* VQDMULL */
5003 {0, 0, 0} /* Polynomial VMULL */
5006 prewiden = neon_3reg_wide[op][0];
5007 src1_wide = neon_3reg_wide[op][1];
5008 src2_wide = neon_3reg_wide[op][2];
5010 if (size == 0 && (op == 9 || op == 11 || op == 13))
5013 /* Avoid overlapping operands. Wide source operands are
5014 always aligned so will never overlap with wide
5015 destinations in problematic ways. */
5016 if (rd == rm && !src2_wide) {
5017 tmp = neon_load_reg(rm, 1);
5018 neon_store_scratch(2, tmp);
5019 } else if (rd == rn && !src1_wide) {
5020 tmp = neon_load_reg(rn, 1);
5021 neon_store_scratch(2, tmp);
5024 for (pass = 0; pass < 2; pass++) {
5026 neon_load_reg64(cpu_V0, rn + pass);
5029 if (pass == 1 && rd == rn) {
5030 tmp = neon_load_scratch(2);
5032 tmp = neon_load_reg(rn, pass);
5035 gen_neon_widen(cpu_V0, tmp, size, u);
5039 neon_load_reg64(cpu_V1, rm + pass);
5042 if (pass == 1 && rd == rm) {
5043 tmp2 = neon_load_scratch(2);
5045 tmp2 = neon_load_reg(rm, pass);
5048 gen_neon_widen(cpu_V1, tmp2, size, u);
5052 case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */
5053 gen_neon_addl(size);
5055 case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHN, VRSUBHN */
5056 gen_neon_subl(size);
5058 case 5: case 7: /* VABAL, VABDL */
5059 switch ((size << 1) | u) {
5061 gen_helper_neon_abdl_s16(cpu_V0, tmp, tmp2);
5064 gen_helper_neon_abdl_u16(cpu_V0, tmp, tmp2);
5067 gen_helper_neon_abdl_s32(cpu_V0, tmp, tmp2);
5070 gen_helper_neon_abdl_u32(cpu_V0, tmp, tmp2);
5073 gen_helper_neon_abdl_s64(cpu_V0, tmp, tmp2);
5076 gen_helper_neon_abdl_u64(cpu_V0, tmp, tmp2);
5083 case 8: case 9: case 10: case 11: case 12: case 13:
5084 /* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
5085 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
5089 case 14: /* Polynomial VMULL */
5090 cpu_abort(env, "Polynomial VMULL not implemented");
5092 default: /* 15 is RESERVED. */
5095 if (op == 5 || op == 13 || (op >= 8 && op <= 11)) {
5097 if (op == 10 || op == 11) {
5098 gen_neon_negl(cpu_V0, size);
5102 neon_load_reg64(cpu_V1, rd + pass);
5106 case 5: case 8: case 10: /* VABAL, VMLAL, VMLSL */
5107 gen_neon_addl(size);
5109 case 9: case 11: /* VQDMLAL, VQDMLSL */
5110 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5111 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
5114 case 13: /* VQDMULL */
5115 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5120 neon_store_reg64(cpu_V0, rd + pass);
5121 } else if (op == 4 || op == 6) {
5122 /* Narrowing operation. */
5127 gen_helper_neon_narrow_high_u8(tmp, cpu_V0);
5130 gen_helper_neon_narrow_high_u16(tmp, cpu_V0);
5133 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
5134 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
5141 gen_helper_neon_narrow_round_high_u8(tmp, cpu_V0);
5144 gen_helper_neon_narrow_round_high_u16(tmp, cpu_V0);
5147 tcg_gen_addi_i64(cpu_V0, cpu_V0, 1u << 31);
5148 tcg_gen_shri_i64(cpu_V0, cpu_V0, 32);
5149 tcg_gen_trunc_i64_i32(tmp, cpu_V0);
5157 neon_store_reg(rd, 0, tmp3);
5158 neon_store_reg(rd, 1, tmp);
5161 /* Write back the result. */
5162 neon_store_reg64(cpu_V0, rd + pass);
5166 /* Two registers and a scalar. */
5168 case 0: /* Integer VMLA scalar */
5169 case 1: /* Float VMLA scalar */
5170 case 4: /* Integer VMLS scalar */
5171 case 5: /* Floating point VMLS scalar */
5172 case 8: /* Integer VMUL scalar */
5173 case 9: /* Floating point VMUL scalar */
5174 case 12: /* VQDMULH scalar */
5175 case 13: /* VQRDMULH scalar */
5176 tmp = neon_get_scalar(size, rm);
5177 neon_store_scratch(0, tmp);
5178 for (pass = 0; pass < (u ? 4 : 2); pass++) {
5179 tmp = neon_load_scratch(0);
5180 tmp2 = neon_load_reg(rn, pass);
5183 gen_helper_neon_qdmulh_s16(tmp, cpu_env, tmp, tmp2);
5185 gen_helper_neon_qdmulh_s32(tmp, cpu_env, tmp, tmp2);
5187 } else if (op == 13) {
5189 gen_helper_neon_qrdmulh_s16(tmp, cpu_env, tmp, tmp2);
5191 gen_helper_neon_qrdmulh_s32(tmp, cpu_env, tmp, tmp2);
5193 } else if (op & 1) {
5194 gen_helper_neon_mul_f32(tmp, tmp, tmp2);
5197 case 0: gen_helper_neon_mul_u8(tmp, tmp, tmp2); break;
5198 case 1: gen_helper_neon_mul_u16(tmp, tmp, tmp2); break;
5199 case 2: tcg_gen_mul_i32(tmp, tmp, tmp2); break;
5206 tmp2 = neon_load_reg(rd, pass);
5209 gen_neon_add(size, tmp, tmp2);
5212 gen_helper_neon_add_f32(tmp, tmp, tmp2);
5215 gen_neon_rsb(size, tmp, tmp2);
5218 gen_helper_neon_sub_f32(tmp, tmp2, tmp);
5225 neon_store_reg(rd, pass, tmp);
5228 case 2: /* VMLAL sclar */
5229 case 3: /* VQDMLAL scalar */
5230 case 6: /* VMLSL scalar */
5231 case 7: /* VQDMLSL scalar */
5232 case 10: /* VMULL scalar */
5233 case 11: /* VQDMULL scalar */
5234 if (size == 0 && (op == 3 || op == 7 || op == 11))
5237 tmp2 = neon_get_scalar(size, rm);
5238 tmp3 = neon_load_reg(rn, 1);
5240 for (pass = 0; pass < 2; pass++) {
5242 tmp = neon_load_reg(rn, 0);
5246 gen_neon_mull(cpu_V0, tmp, tmp2, size, u);
5248 if (op == 6 || op == 7) {
5249 gen_neon_negl(cpu_V0, size);
5252 neon_load_reg64(cpu_V1, rd + pass);
5256 gen_neon_addl(size);
5259 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5260 gen_neon_addl_saturate(cpu_V0, cpu_V1, size);
5266 gen_neon_addl_saturate(cpu_V0, cpu_V0, size);
5271 neon_store_reg64(cpu_V0, rd + pass);
5277 default: /* 14 and 15 are RESERVED */
5281 } else { /* size == 3 */
5284 imm = (insn >> 8) & 0xf;
5290 neon_load_reg64(cpu_V0, rn);
5292 neon_load_reg64(cpu_V1, rn + 1);
5294 } else if (imm == 8) {
5295 neon_load_reg64(cpu_V0, rn + 1);
5297 neon_load_reg64(cpu_V1, rm);
5300 tmp64 = tcg_temp_new_i64();
5302 neon_load_reg64(cpu_V0, rn);
5303 neon_load_reg64(tmp64, rn + 1);
5305 neon_load_reg64(cpu_V0, rn + 1);
5306 neon_load_reg64(tmp64, rm);
5308 tcg_gen_shri_i64(cpu_V0, cpu_V0, (imm & 7) * 8);
5309 tcg_gen_shli_i64(cpu_V1, tmp64, 64 - ((imm & 7) * 8));
5310 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5312 neon_load_reg64(cpu_V1, rm);
5314 neon_load_reg64(cpu_V1, rm + 1);
5317 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
5318 tcg_gen_shri_i64(tmp64, tmp64, imm * 8);
5319 tcg_gen_or_i64(cpu_V1, cpu_V1, tmp64);
5320 tcg_temp_free_i64(tmp64);
5323 neon_load_reg64(cpu_V0, rn);
5324 tcg_gen_shri_i64(cpu_V0, cpu_V0, imm * 8);
5325 neon_load_reg64(cpu_V1, rm);
5326 tcg_gen_shli_i64(cpu_V1, cpu_V1, 64 - (imm * 8));
5327 tcg_gen_or_i64(cpu_V0, cpu_V0, cpu_V1);
5329 neon_store_reg64(cpu_V0, rd);
5331 neon_store_reg64(cpu_V1, rd + 1);
5333 } else if ((insn & (1 << 11)) == 0) {
5334 /* Two register misc. */
5335 op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf);
5336 size = (insn >> 18) & 3;
5338 case 0: /* VREV64 */
5341 for (pass = 0; pass < (q ? 2 : 1); pass++) {
5342 tmp = neon_load_reg(rm, pass * 2);
5343 tmp2 = neon_load_reg(rm, pass * 2 + 1);
5345 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
5346 case 1: gen_swap_half(tmp); break;
5347 case 2: /* no-op */ break;
5350 neon_store_reg(rd, pass * 2 + 1, tmp);
5352 neon_store_reg(rd, pass * 2, tmp2);
5355 case 0: tcg_gen_bswap32_i32(tmp2, tmp2); break;
5356 case 1: gen_swap_half(tmp2); break;
5359 neon_store_reg(rd, pass * 2, tmp2);
5363 case 4: case 5: /* VPADDL */
5364 case 12: case 13: /* VPADAL */
5367 for (pass = 0; pass < q + 1; pass++) {
5368 tmp = neon_load_reg(rm, pass * 2);
5369 gen_neon_widen(cpu_V0, tmp, size, op & 1);
5370 tmp = neon_load_reg(rm, pass * 2 + 1);
5371 gen_neon_widen(cpu_V1, tmp, size, op & 1);
5373 case 0: gen_helper_neon_paddl_u16(CPU_V001); break;
5374 case 1: gen_helper_neon_paddl_u32(CPU_V001); break;
5375 case 2: tcg_gen_add_i64(CPU_V001); break;
5380 neon_load_reg64(cpu_V1, rd + pass);
5381 gen_neon_addl(size);
5383 neon_store_reg64(cpu_V0, rd + pass);
5388 for (n = 0; n < (q ? 4 : 2); n += 2) {
5389 tmp = neon_load_reg(rm, n);
5390 tmp2 = neon_load_reg(rd, n + 1);
5391 neon_store_reg(rm, n, tmp2);
5392 neon_store_reg(rd, n + 1, tmp);
5400 Rd A3 A2 A1 A0 B2 B0 A2 A0
5401 Rm B3 B2 B1 B0 B3 B1 A3 A1
5405 gen_neon_unzip(rd, q, 0, size);
5406 gen_neon_unzip(rm, q, 4, size);
5408 static int unzip_order_q[8] =
5409 {0, 2, 4, 6, 1, 3, 5, 7};
5410 for (n = 0; n < 8; n++) {
5411 int reg = (n < 4) ? rd : rm;
5412 tmp = neon_load_scratch(unzip_order_q[n]);
5413 neon_store_reg(reg, n % 4, tmp);
5416 static int unzip_order[4] =
5418 for (n = 0; n < 4; n++) {
5419 int reg = (n < 2) ? rd : rm;
5420 tmp = neon_load_scratch(unzip_order[n]);
5421 neon_store_reg(reg, n % 2, tmp);
5427 Rd A3 A2 A1 A0 B1 A1 B0 A0
5428 Rm B3 B2 B1 B0 B3 A3 B2 A2
5432 count = (q ? 4 : 2);
5433 for (n = 0; n < count; n++) {
5434 tmp = neon_load_reg(rd, n);
5435 tmp2 = neon_load_reg(rd, n);
5437 case 0: gen_neon_zip_u8(tmp, tmp2); break;
5438 case 1: gen_neon_zip_u16(tmp, tmp2); break;
5439 case 2: /* no-op */; break;
5442 neon_store_scratch(n * 2, tmp);
5443 neon_store_scratch(n * 2 + 1, tmp2);
5445 for (n = 0; n < count * 2; n++) {
5446 int reg = (n < count) ? rd : rm;
5447 tmp = neon_load_scratch(n);
5448 neon_store_reg(reg, n % count, tmp);
5451 case 36: case 37: /* VMOVN, VQMOVUN, VQMOVN */
5455 for (pass = 0; pass < 2; pass++) {
5456 neon_load_reg64(cpu_V0, rm + pass);
5458 if (op == 36 && q == 0) {
5459 gen_neon_narrow(size, tmp, cpu_V0);
5461 gen_neon_narrow_satu(size, tmp, cpu_V0);
5463 gen_neon_narrow_sats(size, tmp, cpu_V0);
5468 neon_store_reg(rd, 0, tmp2);
5469 neon_store_reg(rd, 1, tmp);
5473 case 38: /* VSHLL */
5476 tmp = neon_load_reg(rm, 0);
5477 tmp2 = neon_load_reg(rm, 1);
5478 for (pass = 0; pass < 2; pass++) {
5481 gen_neon_widen(cpu_V0, tmp, size, 1);
5482 tcg_gen_shli_i64(cpu_V0, cpu_V0, 8 << size);
5483 neon_store_reg64(cpu_V0, rd + pass);
5486 case 44: /* VCVT.F16.F32 */
5487 if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
5491 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 0));
5492 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
5493 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 1));
5494 gen_helper_vfp_fcvt_f32_to_f16(tmp2, cpu_F0s, cpu_env);
5495 tcg_gen_shli_i32(tmp2, tmp2, 16);
5496 tcg_gen_or_i32(tmp2, tmp2, tmp);
5497 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 2));
5498 gen_helper_vfp_fcvt_f32_to_f16(tmp, cpu_F0s, cpu_env);
5499 tcg_gen_ld_f32(cpu_F0s, cpu_env, neon_reg_offset(rm, 3));
5500 neon_store_reg(rd, 0, tmp2);
5502 gen_helper_vfp_fcvt_f32_to_f16(tmp2, cpu_F0s, cpu_env);
5503 tcg_gen_shli_i32(tmp2, tmp2, 16);
5504 tcg_gen_or_i32(tmp2, tmp2, tmp);
5505 neon_store_reg(rd, 1, tmp2);
5508 case 46: /* VCVT.F32.F16 */
5509 if (!arm_feature(env, ARM_FEATURE_VFP_FP16))
5512 tmp = neon_load_reg(rm, 0);
5513 tmp2 = neon_load_reg(rm, 1);
5514 tcg_gen_ext16u_i32(tmp3, tmp);
5515 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5516 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 0));
5517 tcg_gen_shri_i32(tmp3, tmp, 16);
5518 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5519 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 1));
5521 tcg_gen_ext16u_i32(tmp3, tmp2);
5522 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5523 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 2));
5524 tcg_gen_shri_i32(tmp3, tmp2, 16);
5525 gen_helper_vfp_fcvt_f16_to_f32(cpu_F0s, tmp3, cpu_env);
5526 tcg_gen_st_f32(cpu_F0s, cpu_env, neon_reg_offset(rd, 3));
5532 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5533 if (op == 30 || op == 31 || op >= 58) {
5534 tcg_gen_ld_f32(cpu_F0s, cpu_env,
5535 neon_reg_offset(rm, pass));
5538 tmp = neon_load_reg(rm, pass);
5541 case 1: /* VREV32 */
5543 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
5544 case 1: gen_swap_half(tmp); break;
5548 case 2: /* VREV16 */
5555 case 0: gen_helper_neon_cls_s8(tmp, tmp); break;
5556 case 1: gen_helper_neon_cls_s16(tmp, tmp); break;
5557 case 2: gen_helper_neon_cls_s32(tmp, tmp); break;
5563 case 0: gen_helper_neon_clz_u8(tmp, tmp); break;
5564 case 1: gen_helper_neon_clz_u16(tmp, tmp); break;
5565 case 2: gen_helper_clz(tmp, tmp); break;
5572 gen_helper_neon_cnt_u8(tmp, tmp);
5577 tcg_gen_not_i32(tmp, tmp);
5579 case 14: /* VQABS */
5581 case 0: gen_helper_neon_qabs_s8(tmp, cpu_env, tmp); break;
5582 case 1: gen_helper_neon_qabs_s16(tmp, cpu_env, tmp); break;
5583 case 2: gen_helper_neon_qabs_s32(tmp, cpu_env, tmp); break;
5587 case 15: /* VQNEG */
5589 case 0: gen_helper_neon_qneg_s8(tmp, cpu_env, tmp); break;
5590 case 1: gen_helper_neon_qneg_s16(tmp, cpu_env, tmp); break;
5591 case 2: gen_helper_neon_qneg_s32(tmp, cpu_env, tmp); break;
5595 case 16: case 19: /* VCGT #0, VCLE #0 */
5596 tmp2 = tcg_const_i32(0);
5598 case 0: gen_helper_neon_cgt_s8(tmp, tmp, tmp2); break;
5599 case 1: gen_helper_neon_cgt_s16(tmp, tmp, tmp2); break;
5600 case 2: gen_helper_neon_cgt_s32(tmp, tmp, tmp2); break;
5603 tcg_temp_free(tmp2);
5605 tcg_gen_not_i32(tmp, tmp);
5607 case 17: case 20: /* VCGE #0, VCLT #0 */
5608 tmp2 = tcg_const_i32(0);
5610 case 0: gen_helper_neon_cge_s8(tmp, tmp, tmp2); break;
5611 case 1: gen_helper_neon_cge_s16(tmp, tmp, tmp2); break;
5612 case 2: gen_helper_neon_cge_s32(tmp, tmp, tmp2); break;
5615 tcg_temp_free(tmp2);
5617 tcg_gen_not_i32(tmp, tmp);
5619 case 18: /* VCEQ #0 */
5620 tmp2 = tcg_const_i32(0);
5622 case 0: gen_helper_neon_ceq_u8(tmp, tmp, tmp2); break;
5623 case 1: gen_helper_neon_ceq_u16(tmp, tmp, tmp2); break;
5624 case 2: gen_helper_neon_ceq_u32(tmp, tmp, tmp2); break;
5627 tcg_temp_free(tmp2);
5631 case 0: gen_helper_neon_abs_s8(tmp, tmp); break;
5632 case 1: gen_helper_neon_abs_s16(tmp, tmp); break;
5633 case 2: tcg_gen_abs_i32(tmp, tmp); break;
5640 tmp2 = tcg_const_i32(0);
5641 gen_neon_rsb(size, tmp, tmp2);
5642 tcg_temp_free(tmp2);
5644 case 24: case 27: /* Float VCGT #0, Float VCLE #0 */
5645 tmp2 = tcg_const_i32(0);
5646 gen_helper_neon_cgt_f32(tmp, tmp, tmp2);
5647 tcg_temp_free(tmp2);
5649 tcg_gen_not_i32(tmp, tmp);
5651 case 25: case 28: /* Float VCGE #0, Float VCLT #0 */
5652 tmp2 = tcg_const_i32(0);
5653 gen_helper_neon_cge_f32(tmp, tmp, tmp2);
5654 tcg_temp_free(tmp2);
5656 tcg_gen_not_i32(tmp, tmp);
5658 case 26: /* Float VCEQ #0 */
5659 tmp2 = tcg_const_i32(0);
5660 gen_helper_neon_ceq_f32(tmp, tmp, tmp2);
5661 tcg_temp_free(tmp2);
5663 case 30: /* Float VABS */
5666 case 31: /* Float VNEG */
5670 tmp2 = neon_load_reg(rd, pass);
5671 neon_store_reg(rm, pass, tmp2);
5674 tmp2 = neon_load_reg(rd, pass);
5676 case 0: gen_neon_trn_u8(tmp, tmp2); break;
5677 case 1: gen_neon_trn_u16(tmp, tmp2); break;
5681 neon_store_reg(rm, pass, tmp2);
5683 case 56: /* Integer VRECPE */
5684 gen_helper_recpe_u32(tmp, tmp, cpu_env);
5686 case 57: /* Integer VRSQRTE */
5687 gen_helper_rsqrte_u32(tmp, tmp, cpu_env);
5689 case 58: /* Float VRECPE */
5690 gen_helper_recpe_f32(cpu_F0s, cpu_F0s, cpu_env);
5692 case 59: /* Float VRSQRTE */
5693 gen_helper_rsqrte_f32(cpu_F0s, cpu_F0s, cpu_env);
5695 case 60: /* VCVT.F32.S32 */
5698 case 61: /* VCVT.F32.U32 */
5701 case 62: /* VCVT.S32.F32 */
5704 case 63: /* VCVT.U32.F32 */
5708 /* Reserved: 21, 29, 39-56 */
5711 if (op == 30 || op == 31 || op >= 58) {
5712 tcg_gen_st_f32(cpu_F0s, cpu_env,
5713 neon_reg_offset(rd, pass));
5715 neon_store_reg(rd, pass, tmp);
5720 } else if ((insn & (1 << 10)) == 0) {
5722 n = ((insn >> 5) & 0x18) + 8;
5723 if (insn & (1 << 6)) {
5724 tmp = neon_load_reg(rd, 0);
5727 tcg_gen_movi_i32(tmp, 0);
5729 tmp2 = neon_load_reg(rm, 0);
5730 tmp4 = tcg_const_i32(rn);
5731 tmp5 = tcg_const_i32(n);
5732 gen_helper_neon_tbl(tmp2, tmp2, tmp, tmp4, tmp5);
5734 if (insn & (1 << 6)) {
5735 tmp = neon_load_reg(rd, 1);
5738 tcg_gen_movi_i32(tmp, 0);
5740 tmp3 = neon_load_reg(rm, 1);
5741 gen_helper_neon_tbl(tmp3, tmp3, tmp, tmp4, tmp5);
5742 tcg_temp_free_i32(tmp5);
5743 tcg_temp_free_i32(tmp4);
5744 neon_store_reg(rd, 0, tmp2);
5745 neon_store_reg(rd, 1, tmp3);
5747 } else if ((insn & 0x380) == 0) {
5749 if (insn & (1 << 19)) {
5750 tmp = neon_load_reg(rm, 1);
5752 tmp = neon_load_reg(rm, 0);
5754 if (insn & (1 << 16)) {
5755 gen_neon_dup_u8(tmp, ((insn >> 17) & 3) * 8);
5756 } else if (insn & (1 << 17)) {
5757 if ((insn >> 18) & 1)
5758 gen_neon_dup_high16(tmp);
5760 gen_neon_dup_low16(tmp);
5762 for (pass = 0; pass < (q ? 4 : 2); pass++) {
5764 tcg_gen_mov_i32(tmp2, tmp);
5765 neon_store_reg(rd, pass, tmp2);
5776 static int disas_cp14_read(CPUState * env, DisasContext *s, uint32_t insn)
5778 int crn = (insn >> 16) & 0xf;
5779 int crm = insn & 0xf;
5780 int op1 = (insn >> 21) & 7;
5781 int op2 = (insn >> 5) & 7;
5782 int rt = (insn >> 12) & 0xf;
5785 if (arm_feature(env, ARM_FEATURE_THUMB2EE)) {
5786 if (op1 == 6 && crn == 0 && crm == 0 && op2 == 0) {
5790 tmp = load_cpu_field(teecr);
5791 store_reg(s, rt, tmp);
5794 if (op1 == 6 && crn == 1 && crm == 0 && op2 == 0) {
5796 if (IS_USER(s) && (env->teecr & 1))
5798 tmp = load_cpu_field(teehbr);
5799 store_reg(s, rt, tmp);
5803 fprintf(stderr, "Unknown cp14 read op1:%d crn:%d crm:%d op2:%d\n",
5804 op1, crn, crm, op2);
5808 static int disas_cp14_write(CPUState * env, DisasContext *s, uint32_t insn)
5810 int crn = (insn >> 16) & 0xf;
5811 int crm = insn & 0xf;
5812 int op1 = (insn >> 21) & 7;
5813 int op2 = (insn >> 5) & 7;
5814 int rt = (insn >> 12) & 0xf;
5817 if (arm_feature(env, ARM_FEATURE_THUMB2EE)) {
5818 if (op1 == 6 && crn == 0 && crm == 0 && op2 == 0) {
5822 tmp = load_reg(s, rt);
5823 gen_helper_set_teecr(cpu_env, tmp);
5827 if (op1 == 6 && crn == 1 && crm == 0 && op2 == 0) {
5829 if (IS_USER(s) && (env->teecr & 1))
5831 tmp = load_reg(s, rt);
5832 store_cpu_field(tmp, teehbr);
5836 fprintf(stderr, "Unknown cp14 write op1:%d crn:%d crm:%d op2:%d\n",
5837 op1, crn, crm, op2);
5841 static int disas_coproc_insn(CPUState * env, DisasContext *s, uint32_t insn)
5845 cpnum = (insn >> 8) & 0xf;
5846 if (arm_feature(env, ARM_FEATURE_XSCALE)
5847 && ((env->cp15.c15_cpar ^ 0x3fff) & (1 << cpnum)))
5853 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
5854 return disas_iwmmxt_insn(env, s, insn);
5855 } else if (arm_feature(env, ARM_FEATURE_XSCALE)) {
5856 return disas_dsp_insn(env, s, insn);
5861 return disas_vfp_insn (env, s, insn);
5863 /* Coprocessors 7-15 are architecturally reserved by ARM.
5864 Unfortunately Intel decided to ignore this. */
5865 if (arm_feature(env, ARM_FEATURE_XSCALE))
5867 if (insn & (1 << 20))
5868 return disas_cp14_read(env, s, insn);
5870 return disas_cp14_write(env, s, insn);
5872 return disas_cp15_insn (env, s, insn);
5875 /* Unknown coprocessor. See if the board has hooked it. */
5876 return disas_cp_insn (env, s, insn);
5881 /* Store a 64-bit value to a register pair. Clobbers val. */
5882 static void gen_storeq_reg(DisasContext *s, int rlow, int rhigh, TCGv_i64 val)
5886 tcg_gen_trunc_i64_i32(tmp, val);
5887 store_reg(s, rlow, tmp);
5889 tcg_gen_shri_i64(val, val, 32);
5890 tcg_gen_trunc_i64_i32(tmp, val);
5891 store_reg(s, rhigh, tmp);
5894 /* load a 32-bit value from a register and perform a 64-bit accumulate. */
5895 static void gen_addq_lo(DisasContext *s, TCGv_i64 val, int rlow)
5900 /* Load value and extend to 64 bits. */
5901 tmp = tcg_temp_new_i64();
5902 tmp2 = load_reg(s, rlow);
5903 tcg_gen_extu_i32_i64(tmp, tmp2);
5905 tcg_gen_add_i64(val, val, tmp);
5906 tcg_temp_free_i64(tmp);
5909 /* load and add a 64-bit value from a register pair. */
5910 static void gen_addq(DisasContext *s, TCGv_i64 val, int rlow, int rhigh)
5916 /* Load 64-bit value rd:rn. */
5917 tmpl = load_reg(s, rlow);
5918 tmph = load_reg(s, rhigh);
5919 tmp = tcg_temp_new_i64();
5920 tcg_gen_concat_i32_i64(tmp, tmpl, tmph);
5923 tcg_gen_add_i64(val, val, tmp);
5924 tcg_temp_free_i64(tmp);
5927 /* Set N and Z flags from a 64-bit value. */
5928 static void gen_logicq_cc(TCGv_i64 val)
5930 TCGv tmp = new_tmp();
5931 gen_helper_logicq_cc(tmp, val);
5936 /* Load/Store exclusive instructions are implemented by remembering
5937 the value/address loaded, and seeing if these are the same
5938 when the store is performed. This should be is sufficient to implement
5939 the architecturally mandated semantics, and avoids having to monitor
5942 In system emulation mode only one CPU will be running at once, so
5943 this sequence is effectively atomic. In user emulation mode we
5944 throw an exception and handle the atomic operation elsewhere. */
5945 static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
5946 TCGv addr, int size)
5952 tmp = gen_ld8u(addr, IS_USER(s));
5955 tmp = gen_ld16u(addr, IS_USER(s));
5959 tmp = gen_ld32(addr, IS_USER(s));
5964 tcg_gen_mov_i32(cpu_exclusive_val, tmp);
5965 store_reg(s, rt, tmp);
5967 TCGv tmp2 = new_tmp();
5968 tcg_gen_addi_i32(tmp2, addr, 4);
5969 tmp = gen_ld32(tmp2, IS_USER(s));
5971 tcg_gen_mov_i32(cpu_exclusive_high, tmp);
5972 store_reg(s, rt2, tmp);
5974 tcg_gen_mov_i32(cpu_exclusive_addr, addr);
5977 static void gen_clrex(DisasContext *s)
5979 tcg_gen_movi_i32(cpu_exclusive_addr, -1);
5982 #ifdef CONFIG_USER_ONLY
5983 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
5984 TCGv addr, int size)
5986 tcg_gen_mov_i32(cpu_exclusive_test, addr);
5987 tcg_gen_movi_i32(cpu_exclusive_info,
5988 size | (rd << 4) | (rt << 8) | (rt2 << 12));
5989 gen_exception_insn(s, 4, EXCP_STREX);
5992 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
5993 TCGv addr, int size)
5999 /* if (env->exclusive_addr == addr && env->exclusive_val == [addr]) {
6005 fail_label = gen_new_label();
6006 done_label = gen_new_label();
6007 tcg_gen_brcond_i32(TCG_COND_NE, addr, cpu_exclusive_addr, fail_label);
6010 tmp = gen_ld8u(addr, IS_USER(s));
6013 tmp = gen_ld16u(addr, IS_USER(s));
6017 tmp = gen_ld32(addr, IS_USER(s));
6022 tcg_gen_brcond_i32(TCG_COND_NE, tmp, cpu_exclusive_val, fail_label);
6025 TCGv tmp2 = new_tmp();
6026 tcg_gen_addi_i32(tmp2, addr, 4);
6027 tmp = gen_ld32(tmp2, IS_USER(s));
6029 tcg_gen_brcond_i32(TCG_COND_NE, tmp, cpu_exclusive_high, fail_label);
6032 tmp = load_reg(s, rt);
6035 gen_st8(tmp, addr, IS_USER(s));
6038 gen_st16(tmp, addr, IS_USER(s));
6042 gen_st32(tmp, addr, IS_USER(s));
6048 tcg_gen_addi_i32(addr, addr, 4);
6049 tmp = load_reg(s, rt2);
6050 gen_st32(tmp, addr, IS_USER(s));
6052 tcg_gen_movi_i32(cpu_R[rd], 0);
6053 tcg_gen_br(done_label);
6054 gen_set_label(fail_label);
6055 tcg_gen_movi_i32(cpu_R[rd], 1);
6056 gen_set_label(done_label);
6057 tcg_gen_movi_i32(cpu_exclusive_addr, -1);
6061 static void disas_arm_insn(CPUState * env, DisasContext *s)
6063 unsigned int cond, insn, val, op1, i, shift, rm, rs, rn, rd, sh;
6070 insn = ldl_code(s->pc);
6073 /* M variants do not implement ARM mode. */
6078 /* Unconditional instructions. */
6079 if (((insn >> 25) & 7) == 1) {
6080 /* NEON Data processing. */
6081 if (!arm_feature(env, ARM_FEATURE_NEON))
6084 if (disas_neon_data_insn(env, s, insn))
6088 if ((insn & 0x0f100000) == 0x04000000) {
6089 /* NEON load/store. */
6090 if (!arm_feature(env, ARM_FEATURE_NEON))
6093 if (disas_neon_ls_insn(env, s, insn))
6097 if ((insn & 0x0d70f000) == 0x0550f000)
6099 else if ((insn & 0x0ffffdff) == 0x01010000) {
6102 if (insn & (1 << 9)) {
6103 /* BE8 mode not implemented. */
6107 } else if ((insn & 0x0fffff00) == 0x057ff000) {
6108 switch ((insn >> 4) & 0xf) {
6117 /* We don't emulate caches so these are a no-op. */
6122 } else if ((insn & 0x0e5fffe0) == 0x084d0500) {
6128 op1 = (insn & 0x1f);
6130 tmp = tcg_const_i32(op1);
6131 gen_helper_get_r13_banked(addr, cpu_env, tmp);
6132 tcg_temp_free_i32(tmp);
6133 i = (insn >> 23) & 3;
6135 case 0: offset = -4; break; /* DA */
6136 case 1: offset = 0; break; /* IA */
6137 case 2: offset = -8; break; /* DB */
6138 case 3: offset = 4; break; /* IB */
6142 tcg_gen_addi_i32(addr, addr, offset);
6143 tmp = load_reg(s, 14);
6144 gen_st32(tmp, addr, 0);
6145 tmp = load_cpu_field(spsr);
6146 tcg_gen_addi_i32(addr, addr, 4);
6147 gen_st32(tmp, addr, 0);
6148 if (insn & (1 << 21)) {
6149 /* Base writeback. */
6151 case 0: offset = -8; break;
6152 case 1: offset = 4; break;
6153 case 2: offset = -4; break;
6154 case 3: offset = 0; break;
6158 tcg_gen_addi_i32(addr, addr, offset);
6159 tmp = tcg_const_i32(op1);
6160 gen_helper_set_r13_banked(cpu_env, tmp, addr);
6161 tcg_temp_free_i32(tmp);
6167 } else if ((insn & 0x0e50ffe0) == 0x08100a00) {
6173 rn = (insn >> 16) & 0xf;
6174 addr = load_reg(s, rn);
6175 i = (insn >> 23) & 3;
6177 case 0: offset = -4; break; /* DA */
6178 case 1: offset = 0; break; /* IA */
6179 case 2: offset = -8; break; /* DB */
6180 case 3: offset = 4; break; /* IB */
6184 tcg_gen_addi_i32(addr, addr, offset);
6185 /* Load PC into tmp and CPSR into tmp2. */
6186 tmp = gen_ld32(addr, 0);
6187 tcg_gen_addi_i32(addr, addr, 4);
6188 tmp2 = gen_ld32(addr, 0);
6189 if (insn & (1 << 21)) {
6190 /* Base writeback. */
6192 case 0: offset = -8; break;
6193 case 1: offset = 4; break;
6194 case 2: offset = -4; break;
6195 case 3: offset = 0; break;
6199 tcg_gen_addi_i32(addr, addr, offset);
6200 store_reg(s, rn, addr);
6204 gen_rfe(s, tmp, tmp2);
6206 } else if ((insn & 0x0e000000) == 0x0a000000) {
6207 /* branch link and change to thumb (blx <offset>) */
6210 val = (uint32_t)s->pc;
6212 tcg_gen_movi_i32(tmp, val);
6213 store_reg(s, 14, tmp);
6214 /* Sign-extend the 24-bit offset */
6215 offset = (((int32_t)insn) << 8) >> 8;
6216 /* offset * 4 + bit24 * 2 + (thumb bit) */
6217 val += (offset << 2) | ((insn >> 23) & 2) | 1;
6218 /* pipeline offset */
6222 } else if ((insn & 0x0e000f00) == 0x0c000100) {
6223 if (arm_feature(env, ARM_FEATURE_IWMMXT)) {
6224 /* iWMMXt register transfer. */
6225 if (env->cp15.c15_cpar & (1 << 1))
6226 if (!disas_iwmmxt_insn(env, s, insn))
6229 } else if ((insn & 0x0fe00000) == 0x0c400000) {
6230 /* Coprocessor double register transfer. */
6231 } else if ((insn & 0x0f000010) == 0x0e000010) {
6232 /* Additional coprocessor register transfer. */
6233 } else if ((insn & 0x0ff10020) == 0x01000000) {
6236 /* cps (privileged) */
6240 if (insn & (1 << 19)) {
6241 if (insn & (1 << 8))
6243 if (insn & (1 << 7))
6245 if (insn & (1 << 6))
6247 if (insn & (1 << 18))
6250 if (insn & (1 << 17)) {
6252 val |= (insn & 0x1f);
6255 gen_set_psr_im(s, mask, 0, val);
6262 /* if not always execute, we generate a conditional jump to
6264 s->condlabel = gen_new_label();
6265 gen_test_cc(cond ^ 1, s->condlabel);
6268 if ((insn & 0x0f900000) == 0x03000000) {
6269 if ((insn & (1 << 21)) == 0) {
6271 rd = (insn >> 12) & 0xf;
6272 val = ((insn >> 4) & 0xf000) | (insn & 0xfff);
6273 if ((insn & (1 << 22)) == 0) {
6276 tcg_gen_movi_i32(tmp, val);
6279 tmp = load_reg(s, rd);
6280 tcg_gen_ext16u_i32(tmp, tmp);
6281 tcg_gen_ori_i32(tmp, tmp, val << 16);
6283 store_reg(s, rd, tmp);
6285 if (((insn >> 12) & 0xf) != 0xf)
6287 if (((insn >> 16) & 0xf) == 0) {
6288 gen_nop_hint(s, insn & 0xff);
6290 /* CPSR = immediate */
6292 shift = ((insn >> 8) & 0xf) * 2;
6294 val = (val >> shift) | (val << (32 - shift));
6295 i = ((insn & (1 << 22)) != 0);
6296 if (gen_set_psr_im(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i, val))
6300 } else if ((insn & 0x0f900000) == 0x01000000
6301 && (insn & 0x00000090) != 0x00000090) {
6302 /* miscellaneous instructions */
6303 op1 = (insn >> 21) & 3;
6304 sh = (insn >> 4) & 0xf;
6307 case 0x0: /* move program status register */
6310 tmp = load_reg(s, rm);
6311 i = ((op1 & 2) != 0);
6312 if (gen_set_psr(s, msr_mask(env, s, (insn >> 16) & 0xf, i), i, tmp))
6316 rd = (insn >> 12) & 0xf;
6320 tmp = load_cpu_field(spsr);
6323 gen_helper_cpsr_read(tmp);
6325 store_reg(s, rd, tmp);
6330 /* branch/exchange thumb (bx). */
6331 tmp = load_reg(s, rm);
6333 } else if (op1 == 3) {
6335 rd = (insn >> 12) & 0xf;
6336 tmp = load_reg(s, rm);
6337 gen_helper_clz(tmp, tmp);
6338 store_reg(s, rd, tmp);
6346 /* Trivial implementation equivalent to bx. */
6347 tmp = load_reg(s, rm);
6357 /* branch link/exchange thumb (blx) */
6358 tmp = load_reg(s, rm);
6360 tcg_gen_movi_i32(tmp2, s->pc);
6361 store_reg(s, 14, tmp2);
6364 case 0x5: /* saturating add/subtract */
6365 rd = (insn >> 12) & 0xf;
6366 rn = (insn >> 16) & 0xf;
6367 tmp = load_reg(s, rm);
6368 tmp2 = load_reg(s, rn);
6370 gen_helper_double_saturate(tmp2, tmp2);
6372 gen_helper_sub_saturate(tmp, tmp, tmp2);
6374 gen_helper_add_saturate(tmp, tmp, tmp2);
6376 store_reg(s, rd, tmp);
6379 /* SMC instruction (op1 == 3)
6380 and undefined instructions (op1 == 0 || op1 == 2)
6386 gen_exception_insn(s, 4, EXCP_BKPT);
6388 case 0x8: /* signed multiply */
6392 rs = (insn >> 8) & 0xf;
6393 rn = (insn >> 12) & 0xf;
6394 rd = (insn >> 16) & 0xf;
6396 /* (32 * 16) >> 16 */
6397 tmp = load_reg(s, rm);
6398 tmp2 = load_reg(s, rs);
6400 tcg_gen_sari_i32(tmp2, tmp2, 16);
6403 tmp64 = gen_muls_i64_i32(tmp, tmp2);
6404 tcg_gen_shri_i64(tmp64, tmp64, 16);
6406 tcg_gen_trunc_i64_i32(tmp, tmp64);
6407 tcg_temp_free_i64(tmp64);
6408 if ((sh & 2) == 0) {
6409 tmp2 = load_reg(s, rn);
6410 gen_helper_add_setq(tmp, tmp, tmp2);
6413 store_reg(s, rd, tmp);
6416 tmp = load_reg(s, rm);
6417 tmp2 = load_reg(s, rs);
6418 gen_mulxy(tmp, tmp2, sh & 2, sh & 4);
6421 tmp64 = tcg_temp_new_i64();
6422 tcg_gen_ext_i32_i64(tmp64, tmp);
6424 gen_addq(s, tmp64, rn, rd);
6425 gen_storeq_reg(s, rn, rd, tmp64);
6426 tcg_temp_free_i64(tmp64);
6429 tmp2 = load_reg(s, rn);
6430 gen_helper_add_setq(tmp, tmp, tmp2);
6433 store_reg(s, rd, tmp);
6440 } else if (((insn & 0x0e000000) == 0 &&
6441 (insn & 0x00000090) != 0x90) ||
6442 ((insn & 0x0e000000) == (1 << 25))) {
6443 int set_cc, logic_cc, shiftop;
6445 op1 = (insn >> 21) & 0xf;
6446 set_cc = (insn >> 20) & 1;
6447 logic_cc = table_logic_cc[op1] & set_cc;
6449 /* data processing instruction */
6450 if (insn & (1 << 25)) {
6451 /* immediate operand */
6453 shift = ((insn >> 8) & 0xf) * 2;
6455 val = (val >> shift) | (val << (32 - shift));
6458 tcg_gen_movi_i32(tmp2, val);
6459 if (logic_cc && shift) {
6460 gen_set_CF_bit31(tmp2);
6465 tmp2 = load_reg(s, rm);
6466 shiftop = (insn >> 5) & 3;
6467 if (!(insn & (1 << 4))) {
6468 shift = (insn >> 7) & 0x1f;
6469 gen_arm_shift_im(tmp2, shiftop, shift, logic_cc);
6471 rs = (insn >> 8) & 0xf;
6472 tmp = load_reg(s, rs);
6473 gen_arm_shift_reg(tmp2, shiftop, tmp, logic_cc);
6476 if (op1 != 0x0f && op1 != 0x0d) {
6477 rn = (insn >> 16) & 0xf;
6478 tmp = load_reg(s, rn);
6482 rd = (insn >> 12) & 0xf;
6485 tcg_gen_and_i32(tmp, tmp, tmp2);
6489 store_reg_bx(env, s, rd, tmp);
6492 tcg_gen_xor_i32(tmp, tmp, tmp2);
6496 store_reg_bx(env, s, rd, tmp);
6499 if (set_cc && rd == 15) {
6500 /* SUBS r15, ... is used for exception return. */
6504 gen_helper_sub_cc(tmp, tmp, tmp2);
6505 gen_exception_return(s, tmp);
6508 gen_helper_sub_cc(tmp, tmp, tmp2);
6510 tcg_gen_sub_i32(tmp, tmp, tmp2);
6512 store_reg_bx(env, s, rd, tmp);
6517 gen_helper_sub_cc(tmp, tmp2, tmp);
6519 tcg_gen_sub_i32(tmp, tmp2, tmp);
6521 store_reg_bx(env, s, rd, tmp);
6525 gen_helper_add_cc(tmp, tmp, tmp2);
6527 tcg_gen_add_i32(tmp, tmp, tmp2);
6529 store_reg_bx(env, s, rd, tmp);
6533 gen_helper_adc_cc(tmp, tmp, tmp2);
6535 gen_add_carry(tmp, tmp, tmp2);
6537 store_reg_bx(env, s, rd, tmp);
6541 gen_helper_sbc_cc(tmp, tmp, tmp2);
6543 gen_sub_carry(tmp, tmp, tmp2);
6545 store_reg_bx(env, s, rd, tmp);
6549 gen_helper_sbc_cc(tmp, tmp2, tmp);
6551 gen_sub_carry(tmp, tmp2, tmp);
6553 store_reg_bx(env, s, rd, tmp);
6557 tcg_gen_and_i32(tmp, tmp, tmp2);
6564 tcg_gen_xor_i32(tmp, tmp, tmp2);
6571 gen_helper_sub_cc(tmp, tmp, tmp2);
6577 gen_helper_add_cc(tmp, tmp, tmp2);
6582 tcg_gen_or_i32(tmp, tmp, tmp2);
6586 store_reg_bx(env, s, rd, tmp);
6589 if (logic_cc && rd == 15) {
6590 /* MOVS r15, ... is used for exception return. */
6594 gen_exception_return(s, tmp2);
6599 store_reg_bx(env, s, rd, tmp2);
6603 tcg_gen_andc_i32(tmp, tmp, tmp2);
6607 store_reg_bx(env, s, rd, tmp);
6611 tcg_gen_not_i32(tmp2, tmp2);
6615 store_reg_bx(env, s, rd, tmp2);
6618 if (op1 != 0x0f && op1 != 0x0d) {
6622 /* other instructions */
6623 op1 = (insn >> 24) & 0xf;
6627 /* multiplies, extra load/stores */
6628 sh = (insn >> 5) & 3;
6631 rd = (insn >> 16) & 0xf;
6632 rn = (insn >> 12) & 0xf;
6633 rs = (insn >> 8) & 0xf;
6635 op1 = (insn >> 20) & 0xf;
6637 case 0: case 1: case 2: case 3: case 6:
6639 tmp = load_reg(s, rs);
6640 tmp2 = load_reg(s, rm);
6641 tcg_gen_mul_i32(tmp, tmp, tmp2);
6643 if (insn & (1 << 22)) {
6644 /* Subtract (mls) */
6646 tmp2 = load_reg(s, rn);
6647 tcg_gen_sub_i32(tmp, tmp2, tmp);
6649 } else if (insn & (1 << 21)) {
6651 tmp2 = load_reg(s, rn);
6652 tcg_gen_add_i32(tmp, tmp, tmp2);
6655 if (insn & (1 << 20))
6657 store_reg(s, rd, tmp);
6660 /* 64 bit mul double accumulate (UMAAL) */
6662 tmp = load_reg(s, rs);
6663 tmp2 = load_reg(s, rm);
6664 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
6665 gen_addq_lo(s, tmp64, rn);
6666 gen_addq_lo(s, tmp64, rd);
6667 gen_storeq_reg(s, rn, rd, tmp64);
6668 tcg_temp_free_i64(tmp64);
6670 case 8: case 9: case 10: case 11:
6671 case 12: case 13: case 14: case 15:
6672 /* 64 bit mul: UMULL, UMLAL, SMULL, SMLAL. */
6673 tmp = load_reg(s, rs);
6674 tmp2 = load_reg(s, rm);
6675 if (insn & (1 << 22)) {
6676 tmp64 = gen_muls_i64_i32(tmp, tmp2);
6678 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
6680 if (insn & (1 << 21)) { /* mult accumulate */
6681 gen_addq(s, tmp64, rn, rd);
6683 if (insn & (1 << 20)) {
6684 gen_logicq_cc(tmp64);
6686 gen_storeq_reg(s, rn, rd, tmp64);
6687 tcg_temp_free_i64(tmp64);
6693 rn = (insn >> 16) & 0xf;
6694 rd = (insn >> 12) & 0xf;
6695 if (insn & (1 << 23)) {
6696 /* load/store exclusive */
6697 op1 = (insn >> 21) & 0x3;
6702 addr = tcg_temp_local_new_i32();
6703 load_reg_var(s, addr, rn);
6704 if (insn & (1 << 20)) {
6707 gen_load_exclusive(s, rd, 15, addr, 2);
6709 case 1: /* ldrexd */
6710 gen_load_exclusive(s, rd, rd + 1, addr, 3);
6712 case 2: /* ldrexb */
6713 gen_load_exclusive(s, rd, 15, addr, 0);
6715 case 3: /* ldrexh */
6716 gen_load_exclusive(s, rd, 15, addr, 1);
6725 gen_store_exclusive(s, rd, rm, 15, addr, 2);
6727 case 1: /* strexd */
6728 gen_store_exclusive(s, rd, rm, rm + 1, addr, 3);
6730 case 2: /* strexb */
6731 gen_store_exclusive(s, rd, rm, 15, addr, 0);
6733 case 3: /* strexh */
6734 gen_store_exclusive(s, rd, rm, 15, addr, 1);
6740 tcg_temp_free(addr);
6742 /* SWP instruction */
6745 /* ??? This is not really atomic. However we know
6746 we never have multiple CPUs running in parallel,
6747 so it is good enough. */
6748 addr = load_reg(s, rn);
6749 tmp = load_reg(s, rm);
6750 if (insn & (1 << 22)) {
6751 tmp2 = gen_ld8u(addr, IS_USER(s));
6752 gen_st8(tmp, addr, IS_USER(s));
6754 tmp2 = gen_ld32(addr, IS_USER(s));
6755 gen_st32(tmp, addr, IS_USER(s));
6758 store_reg(s, rd, tmp2);
6764 /* Misc load/store */
6765 rn = (insn >> 16) & 0xf;
6766 rd = (insn >> 12) & 0xf;
6767 addr = load_reg(s, rn);
6768 if (insn & (1 << 24))
6769 gen_add_datah_offset(s, insn, 0, addr);
6771 if (insn & (1 << 20)) {
6775 tmp = gen_ld16u(addr, IS_USER(s));
6778 tmp = gen_ld8s(addr, IS_USER(s));
6782 tmp = gen_ld16s(addr, IS_USER(s));
6786 } else if (sh & 2) {
6790 tmp = load_reg(s, rd);
6791 gen_st32(tmp, addr, IS_USER(s));
6792 tcg_gen_addi_i32(addr, addr, 4);
6793 tmp = load_reg(s, rd + 1);
6794 gen_st32(tmp, addr, IS_USER(s));
6798 tmp = gen_ld32(addr, IS_USER(s));
6799 store_reg(s, rd, tmp);
6800 tcg_gen_addi_i32(addr, addr, 4);
6801 tmp = gen_ld32(addr, IS_USER(s));
6805 address_offset = -4;
6808 tmp = load_reg(s, rd);
6809 gen_st16(tmp, addr, IS_USER(s));
6812 /* Perform base writeback before the loaded value to
6813 ensure correct behavior with overlapping index registers.
6814 ldrd with base writeback is is undefined if the
6815 destination and index registers overlap. */
6816 if (!(insn & (1 << 24))) {
6817 gen_add_datah_offset(s, insn, address_offset, addr);
6818 store_reg(s, rn, addr);
6819 } else if (insn & (1 << 21)) {
6821 tcg_gen_addi_i32(addr, addr, address_offset);
6822 store_reg(s, rn, addr);
6827 /* Complete the load. */
6828 store_reg(s, rd, tmp);
6837 if (insn & (1 << 4)) {
6839 /* Armv6 Media instructions. */
6841 rn = (insn >> 16) & 0xf;
6842 rd = (insn >> 12) & 0xf;
6843 rs = (insn >> 8) & 0xf;
6844 switch ((insn >> 23) & 3) {
6845 case 0: /* Parallel add/subtract. */
6846 op1 = (insn >> 20) & 7;
6847 tmp = load_reg(s, rn);
6848 tmp2 = load_reg(s, rm);
6849 sh = (insn >> 5) & 7;
6850 if ((op1 & 3) == 0 || sh == 5 || sh == 6)
6852 gen_arm_parallel_addsub(op1, sh, tmp, tmp2);
6854 store_reg(s, rd, tmp);
6857 if ((insn & 0x00700020) == 0) {
6858 /* Halfword pack. */
6859 tmp = load_reg(s, rn);
6860 tmp2 = load_reg(s, rm);
6861 shift = (insn >> 7) & 0x1f;
6862 if (insn & (1 << 6)) {
6866 tcg_gen_sari_i32(tmp2, tmp2, shift);
6867 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
6868 tcg_gen_ext16u_i32(tmp2, tmp2);
6872 tcg_gen_shli_i32(tmp2, tmp2, shift);
6873 tcg_gen_ext16u_i32(tmp, tmp);
6874 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
6876 tcg_gen_or_i32(tmp, tmp, tmp2);
6878 store_reg(s, rd, tmp);
6879 } else if ((insn & 0x00200020) == 0x00200000) {
6881 tmp = load_reg(s, rm);
6882 shift = (insn >> 7) & 0x1f;
6883 if (insn & (1 << 6)) {
6886 tcg_gen_sari_i32(tmp, tmp, shift);
6888 tcg_gen_shli_i32(tmp, tmp, shift);
6890 sh = (insn >> 16) & 0x1f;
6892 tmp2 = tcg_const_i32(sh);
6893 if (insn & (1 << 22))
6894 gen_helper_usat(tmp, tmp, tmp2);
6896 gen_helper_ssat(tmp, tmp, tmp2);
6897 tcg_temp_free_i32(tmp2);
6899 store_reg(s, rd, tmp);
6900 } else if ((insn & 0x00300fe0) == 0x00200f20) {
6902 tmp = load_reg(s, rm);
6903 sh = (insn >> 16) & 0x1f;
6905 tmp2 = tcg_const_i32(sh);
6906 if (insn & (1 << 22))
6907 gen_helper_usat16(tmp, tmp, tmp2);
6909 gen_helper_ssat16(tmp, tmp, tmp2);
6910 tcg_temp_free_i32(tmp2);
6912 store_reg(s, rd, tmp);
6913 } else if ((insn & 0x00700fe0) == 0x00000fa0) {
6915 tmp = load_reg(s, rn);
6916 tmp2 = load_reg(s, rm);
6918 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUState, GE));
6919 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
6922 store_reg(s, rd, tmp);
6923 } else if ((insn & 0x000003e0) == 0x00000060) {
6924 tmp = load_reg(s, rm);
6925 shift = (insn >> 10) & 3;
6926 /* ??? In many cases it's not neccessary to do a
6927 rotate, a shift is sufficient. */
6929 tcg_gen_rotri_i32(tmp, tmp, shift * 8);
6930 op1 = (insn >> 20) & 7;
6932 case 0: gen_sxtb16(tmp); break;
6933 case 2: gen_sxtb(tmp); break;
6934 case 3: gen_sxth(tmp); break;
6935 case 4: gen_uxtb16(tmp); break;
6936 case 6: gen_uxtb(tmp); break;
6937 case 7: gen_uxth(tmp); break;
6938 default: goto illegal_op;
6941 tmp2 = load_reg(s, rn);
6942 if ((op1 & 3) == 0) {
6943 gen_add16(tmp, tmp2);
6945 tcg_gen_add_i32(tmp, tmp, tmp2);
6949 store_reg(s, rd, tmp);
6950 } else if ((insn & 0x003f0f60) == 0x003f0f20) {
6952 tmp = load_reg(s, rm);
6953 if (insn & (1 << 22)) {
6954 if (insn & (1 << 7)) {
6958 gen_helper_rbit(tmp, tmp);
6961 if (insn & (1 << 7))
6964 tcg_gen_bswap32_i32(tmp, tmp);
6966 store_reg(s, rd, tmp);
6971 case 2: /* Multiplies (Type 3). */
6972 tmp = load_reg(s, rm);
6973 tmp2 = load_reg(s, rs);
6974 if (insn & (1 << 20)) {
6975 /* Signed multiply most significant [accumulate].
6976 (SMMUL, SMMLA, SMMLS) */
6977 tmp64 = gen_muls_i64_i32(tmp, tmp2);
6980 tmp = load_reg(s, rd);
6981 if (insn & (1 << 6)) {
6982 tmp64 = gen_subq_msw(tmp64, tmp);
6984 tmp64 = gen_addq_msw(tmp64, tmp);
6987 if (insn & (1 << 5)) {
6988 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
6990 tcg_gen_shri_i64(tmp64, tmp64, 32);
6992 tcg_gen_trunc_i64_i32(tmp, tmp64);
6993 tcg_temp_free_i64(tmp64);
6994 store_reg(s, rn, tmp);
6996 if (insn & (1 << 5))
6997 gen_swap_half(tmp2);
6998 gen_smul_dual(tmp, tmp2);
6999 /* This addition cannot overflow. */
7000 if (insn & (1 << 6)) {
7001 tcg_gen_sub_i32(tmp, tmp, tmp2);
7003 tcg_gen_add_i32(tmp, tmp, tmp2);
7006 if (insn & (1 << 22)) {
7007 /* smlald, smlsld */
7008 tmp64 = tcg_temp_new_i64();
7009 tcg_gen_ext_i32_i64(tmp64, tmp);
7011 gen_addq(s, tmp64, rd, rn);
7012 gen_storeq_reg(s, rd, rn, tmp64);
7013 tcg_temp_free_i64(tmp64);
7015 /* smuad, smusd, smlad, smlsd */
7018 tmp2 = load_reg(s, rd);
7019 gen_helper_add_setq(tmp, tmp, tmp2);
7022 store_reg(s, rn, tmp);
7027 op1 = ((insn >> 17) & 0x38) | ((insn >> 5) & 7);
7029 case 0: /* Unsigned sum of absolute differences. */
7031 tmp = load_reg(s, rm);
7032 tmp2 = load_reg(s, rs);
7033 gen_helper_usad8(tmp, tmp, tmp2);
7036 tmp2 = load_reg(s, rd);
7037 tcg_gen_add_i32(tmp, tmp, tmp2);
7040 store_reg(s, rn, tmp);
7042 case 0x20: case 0x24: case 0x28: case 0x2c:
7043 /* Bitfield insert/clear. */
7045 shift = (insn >> 7) & 0x1f;
7046 i = (insn >> 16) & 0x1f;
7050 tcg_gen_movi_i32(tmp, 0);
7052 tmp = load_reg(s, rm);
7055 tmp2 = load_reg(s, rd);
7056 gen_bfi(tmp, tmp2, tmp, shift, (1u << i) - 1);
7059 store_reg(s, rd, tmp);
7061 case 0x12: case 0x16: case 0x1a: case 0x1e: /* sbfx */
7062 case 0x32: case 0x36: case 0x3a: case 0x3e: /* ubfx */
7064 tmp = load_reg(s, rm);
7065 shift = (insn >> 7) & 0x1f;
7066 i = ((insn >> 16) & 0x1f) + 1;
7071 gen_ubfx(tmp, shift, (1u << i) - 1);
7073 gen_sbfx(tmp, shift, i);
7076 store_reg(s, rd, tmp);
7086 /* Check for undefined extension instructions
7087 * per the ARM Bible IE:
7088 * xxxx 0111 1111 xxxx xxxx xxxx 1111 xxxx
7090 sh = (0xf << 20) | (0xf << 4);
7091 if (op1 == 0x7 && ((insn & sh) == sh))
7095 /* load/store byte/word */
7096 rn = (insn >> 16) & 0xf;
7097 rd = (insn >> 12) & 0xf;
7098 tmp2 = load_reg(s, rn);
7099 i = (IS_USER(s) || (insn & 0x01200000) == 0x00200000);
7100 if (insn & (1 << 24))
7101 gen_add_data_offset(s, insn, tmp2);
7102 if (insn & (1 << 20)) {
7104 if (insn & (1 << 22)) {
7105 tmp = gen_ld8u(tmp2, i);
7107 tmp = gen_ld32(tmp2, i);
7111 tmp = load_reg(s, rd);
7112 if (insn & (1 << 22))
7113 gen_st8(tmp, tmp2, i);
7115 gen_st32(tmp, tmp2, i);
7117 if (!(insn & (1 << 24))) {
7118 gen_add_data_offset(s, insn, tmp2);
7119 store_reg(s, rn, tmp2);
7120 } else if (insn & (1 << 21)) {
7121 store_reg(s, rn, tmp2);
7125 if (insn & (1 << 20)) {
7126 /* Complete the load. */
7130 store_reg(s, rd, tmp);
7136 int j, n, user, loaded_base;
7138 /* load/store multiple words */
7139 /* XXX: store correct base if write back */
7141 if (insn & (1 << 22)) {
7143 goto illegal_op; /* only usable in supervisor mode */
7145 if ((insn & (1 << 15)) == 0)
7148 rn = (insn >> 16) & 0xf;
7149 addr = load_reg(s, rn);
7151 /* compute total size */
7153 TCGV_UNUSED(loaded_var);
7156 if (insn & (1 << i))
7159 /* XXX: test invalid n == 0 case ? */
7160 if (insn & (1 << 23)) {
7161 if (insn & (1 << 24)) {
7163 tcg_gen_addi_i32(addr, addr, 4);
7165 /* post increment */
7168 if (insn & (1 << 24)) {
7170 tcg_gen_addi_i32(addr, addr, -(n * 4));
7172 /* post decrement */
7174 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
7179 if (insn & (1 << i)) {
7180 if (insn & (1 << 20)) {
7182 tmp = gen_ld32(addr, IS_USER(s));
7186 tmp2 = tcg_const_i32(i);
7187 gen_helper_set_user_reg(tmp2, tmp);
7188 tcg_temp_free_i32(tmp2);
7190 } else if (i == rn) {
7194 store_reg(s, i, tmp);
7199 /* special case: r15 = PC + 8 */
7200 val = (long)s->pc + 4;
7202 tcg_gen_movi_i32(tmp, val);
7205 tmp2 = tcg_const_i32(i);
7206 gen_helper_get_user_reg(tmp, tmp2);
7207 tcg_temp_free_i32(tmp2);
7209 tmp = load_reg(s, i);
7211 gen_st32(tmp, addr, IS_USER(s));
7214 /* no need to add after the last transfer */
7216 tcg_gen_addi_i32(addr, addr, 4);
7219 if (insn & (1 << 21)) {
7221 if (insn & (1 << 23)) {
7222 if (insn & (1 << 24)) {
7225 /* post increment */
7226 tcg_gen_addi_i32(addr, addr, 4);
7229 if (insn & (1 << 24)) {
7232 tcg_gen_addi_i32(addr, addr, -((n - 1) * 4));
7234 /* post decrement */
7235 tcg_gen_addi_i32(addr, addr, -(n * 4));
7238 store_reg(s, rn, addr);
7243 store_reg(s, rn, loaded_var);
7245 if ((insn & (1 << 22)) && !user) {
7246 /* Restore CPSR from SPSR. */
7247 tmp = load_cpu_field(spsr);
7248 gen_set_cpsr(tmp, 0xffffffff);
7250 s->is_jmp = DISAS_UPDATE;
7259 /* branch (and link) */
7260 val = (int32_t)s->pc;
7261 if (insn & (1 << 24)) {
7263 tcg_gen_movi_i32(tmp, val);
7264 store_reg(s, 14, tmp);
7266 offset = (((int32_t)insn << 8) >> 8);
7267 val += (offset << 2) + 4;
7275 if (disas_coproc_insn(env, s, insn))
7280 gen_set_pc_im(s->pc);
7281 s->is_jmp = DISAS_SWI;
7285 gen_exception_insn(s, 4, EXCP_UDEF);
7291 /* Return true if this is a Thumb-2 logical op. */
7293 thumb2_logic_op(int op)
7298 /* Generate code for a Thumb-2 data processing operation. If CONDS is nonzero
7299 then set condition code flags based on the result of the operation.
7300 If SHIFTER_OUT is nonzero then set the carry flag for logical operations
7301 to the high bit of T1.
7302 Returns zero if the opcode is valid. */
7305 gen_thumb2_data_op(DisasContext *s, int op, int conds, uint32_t shifter_out, TCGv t0, TCGv t1)
7312 tcg_gen_and_i32(t0, t0, t1);
7316 tcg_gen_andc_i32(t0, t0, t1);
7320 tcg_gen_or_i32(t0, t0, t1);
7324 tcg_gen_not_i32(t1, t1);
7325 tcg_gen_or_i32(t0, t0, t1);
7329 tcg_gen_xor_i32(t0, t0, t1);
7334 gen_helper_add_cc(t0, t0, t1);
7336 tcg_gen_add_i32(t0, t0, t1);
7340 gen_helper_adc_cc(t0, t0, t1);
7346 gen_helper_sbc_cc(t0, t0, t1);
7348 gen_sub_carry(t0, t0, t1);
7352 gen_helper_sub_cc(t0, t0, t1);
7354 tcg_gen_sub_i32(t0, t0, t1);
7358 gen_helper_sub_cc(t0, t1, t0);
7360 tcg_gen_sub_i32(t0, t1, t0);
7362 default: /* 5, 6, 7, 9, 12, 15. */
7368 gen_set_CF_bit31(t1);
7373 /* Translate a 32-bit thumb instruction. Returns nonzero if the instruction
7375 static int disas_thumb2_insn(CPUState *env, DisasContext *s, uint16_t insn_hw1)
7377 uint32_t insn, imm, shift, offset;
7378 uint32_t rd, rn, rm, rs;
7389 if (!(arm_feature(env, ARM_FEATURE_THUMB2)
7390 || arm_feature (env, ARM_FEATURE_M))) {
7391 /* Thumb-1 cores may need to treat bl and blx as a pair of
7392 16-bit instructions to get correct prefetch abort behavior. */
7394 if ((insn & (1 << 12)) == 0) {
7395 /* Second half of blx. */
7396 offset = ((insn & 0x7ff) << 1);
7397 tmp = load_reg(s, 14);
7398 tcg_gen_addi_i32(tmp, tmp, offset);
7399 tcg_gen_andi_i32(tmp, tmp, 0xfffffffc);
7402 tcg_gen_movi_i32(tmp2, s->pc | 1);
7403 store_reg(s, 14, tmp2);
7407 if (insn & (1 << 11)) {
7408 /* Second half of bl. */
7409 offset = ((insn & 0x7ff) << 1) | 1;
7410 tmp = load_reg(s, 14);
7411 tcg_gen_addi_i32(tmp, tmp, offset);
7414 tcg_gen_movi_i32(tmp2, s->pc | 1);
7415 store_reg(s, 14, tmp2);
7419 if ((s->pc & ~TARGET_PAGE_MASK) == 0) {
7420 /* Instruction spans a page boundary. Implement it as two
7421 16-bit instructions in case the second half causes an
7423 offset = ((int32_t)insn << 21) >> 9;
7424 tcg_gen_movi_i32(cpu_R[14], s->pc + 2 + offset);
7427 /* Fall through to 32-bit decode. */
7430 insn = lduw_code(s->pc);
7432 insn |= (uint32_t)insn_hw1 << 16;
7434 if ((insn & 0xf800e800) != 0xf000e800) {
7438 rn = (insn >> 16) & 0xf;
7439 rs = (insn >> 12) & 0xf;
7440 rd = (insn >> 8) & 0xf;
7442 switch ((insn >> 25) & 0xf) {
7443 case 0: case 1: case 2: case 3:
7444 /* 16-bit instructions. Should never happen. */
7447 if (insn & (1 << 22)) {
7448 /* Other load/store, table branch. */
7449 if (insn & 0x01200000) {
7450 /* Load/store doubleword. */
7453 tcg_gen_movi_i32(addr, s->pc & ~3);
7455 addr = load_reg(s, rn);
7457 offset = (insn & 0xff) * 4;
7458 if ((insn & (1 << 23)) == 0)
7460 if (insn & (1 << 24)) {
7461 tcg_gen_addi_i32(addr, addr, offset);
7464 if (insn & (1 << 20)) {
7466 tmp = gen_ld32(addr, IS_USER(s));
7467 store_reg(s, rs, tmp);
7468 tcg_gen_addi_i32(addr, addr, 4);
7469 tmp = gen_ld32(addr, IS_USER(s));
7470 store_reg(s, rd, tmp);
7473 tmp = load_reg(s, rs);
7474 gen_st32(tmp, addr, IS_USER(s));
7475 tcg_gen_addi_i32(addr, addr, 4);
7476 tmp = load_reg(s, rd);
7477 gen_st32(tmp, addr, IS_USER(s));
7479 if (insn & (1 << 21)) {
7480 /* Base writeback. */
7483 tcg_gen_addi_i32(addr, addr, offset - 4);
7484 store_reg(s, rn, addr);
7488 } else if ((insn & (1 << 23)) == 0) {
7489 /* Load/store exclusive word. */
7490 addr = tcg_temp_local_new();
7491 load_reg_var(s, addr, rn);
7492 tcg_gen_addi_i32(addr, addr, (insn & 0xff) << 2);
7493 if (insn & (1 << 20)) {
7494 gen_load_exclusive(s, rs, 15, addr, 2);
7496 gen_store_exclusive(s, rd, rs, 15, addr, 2);
7498 tcg_temp_free(addr);
7499 } else if ((insn & (1 << 6)) == 0) {
7503 tcg_gen_movi_i32(addr, s->pc);
7505 addr = load_reg(s, rn);
7507 tmp = load_reg(s, rm);
7508 tcg_gen_add_i32(addr, addr, tmp);
7509 if (insn & (1 << 4)) {
7511 tcg_gen_add_i32(addr, addr, tmp);
7513 tmp = gen_ld16u(addr, IS_USER(s));
7516 tmp = gen_ld8u(addr, IS_USER(s));
7519 tcg_gen_shli_i32(tmp, tmp, 1);
7520 tcg_gen_addi_i32(tmp, tmp, s->pc);
7521 store_reg(s, 15, tmp);
7523 /* Load/store exclusive byte/halfword/doubleword. */
7525 op = (insn >> 4) & 0x3;
7529 addr = tcg_temp_local_new();
7530 load_reg_var(s, addr, rn);
7531 if (insn & (1 << 20)) {
7532 gen_load_exclusive(s, rs, rd, addr, op);
7534 gen_store_exclusive(s, rm, rs, rd, addr, op);
7536 tcg_temp_free(addr);
7539 /* Load/store multiple, RFE, SRS. */
7540 if (((insn >> 23) & 1) == ((insn >> 24) & 1)) {
7541 /* Not available in user mode. */
7544 if (insn & (1 << 20)) {
7546 addr = load_reg(s, rn);
7547 if ((insn & (1 << 24)) == 0)
7548 tcg_gen_addi_i32(addr, addr, -8);
7549 /* Load PC into tmp and CPSR into tmp2. */
7550 tmp = gen_ld32(addr, 0);
7551 tcg_gen_addi_i32(addr, addr, 4);
7552 tmp2 = gen_ld32(addr, 0);
7553 if (insn & (1 << 21)) {
7554 /* Base writeback. */
7555 if (insn & (1 << 24)) {
7556 tcg_gen_addi_i32(addr, addr, 4);
7558 tcg_gen_addi_i32(addr, addr, -4);
7560 store_reg(s, rn, addr);
7564 gen_rfe(s, tmp, tmp2);
7569 tmp = tcg_const_i32(op);
7570 gen_helper_get_r13_banked(addr, cpu_env, tmp);
7571 tcg_temp_free_i32(tmp);
7572 if ((insn & (1 << 24)) == 0) {
7573 tcg_gen_addi_i32(addr, addr, -8);
7575 tmp = load_reg(s, 14);
7576 gen_st32(tmp, addr, 0);
7577 tcg_gen_addi_i32(addr, addr, 4);
7579 gen_helper_cpsr_read(tmp);
7580 gen_st32(tmp, addr, 0);
7581 if (insn & (1 << 21)) {
7582 if ((insn & (1 << 24)) == 0) {
7583 tcg_gen_addi_i32(addr, addr, -4);
7585 tcg_gen_addi_i32(addr, addr, 4);
7587 tmp = tcg_const_i32(op);
7588 gen_helper_set_r13_banked(cpu_env, tmp, addr);
7589 tcg_temp_free_i32(tmp);
7596 /* Load/store multiple. */
7597 addr = load_reg(s, rn);
7599 for (i = 0; i < 16; i++) {
7600 if (insn & (1 << i))
7603 if (insn & (1 << 24)) {
7604 tcg_gen_addi_i32(addr, addr, -offset);
7607 for (i = 0; i < 16; i++) {
7608 if ((insn & (1 << i)) == 0)
7610 if (insn & (1 << 20)) {
7612 tmp = gen_ld32(addr, IS_USER(s));
7616 store_reg(s, i, tmp);
7620 tmp = load_reg(s, i);
7621 gen_st32(tmp, addr, IS_USER(s));
7623 tcg_gen_addi_i32(addr, addr, 4);
7625 if (insn & (1 << 21)) {
7626 /* Base register writeback. */
7627 if (insn & (1 << 24)) {
7628 tcg_gen_addi_i32(addr, addr, -offset);
7630 /* Fault if writeback register is in register list. */
7631 if (insn & (1 << rn))
7633 store_reg(s, rn, addr);
7642 op = (insn >> 21) & 0xf;
7644 /* Halfword pack. */
7645 tmp = load_reg(s, rn);
7646 tmp2 = load_reg(s, rm);
7647 shift = ((insn >> 10) & 0x1c) | ((insn >> 6) & 0x3);
7648 if (insn & (1 << 5)) {
7652 tcg_gen_sari_i32(tmp2, tmp2, shift);
7653 tcg_gen_andi_i32(tmp, tmp, 0xffff0000);
7654 tcg_gen_ext16u_i32(tmp2, tmp2);
7658 tcg_gen_shli_i32(tmp2, tmp2, shift);
7659 tcg_gen_ext16u_i32(tmp, tmp);
7660 tcg_gen_andi_i32(tmp2, tmp2, 0xffff0000);
7662 tcg_gen_or_i32(tmp, tmp, tmp2);
7664 store_reg(s, rd, tmp);
7666 /* Data processing register constant shift. */
7669 tcg_gen_movi_i32(tmp, 0);
7671 tmp = load_reg(s, rn);
7673 tmp2 = load_reg(s, rm);
7675 shiftop = (insn >> 4) & 3;
7676 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
7677 conds = (insn & (1 << 20)) != 0;
7678 logic_cc = (conds && thumb2_logic_op(op));
7679 gen_arm_shift_im(tmp2, shiftop, shift, logic_cc);
7680 if (gen_thumb2_data_op(s, op, conds, 0, tmp, tmp2))
7684 store_reg(s, rd, tmp);
7690 case 13: /* Misc data processing. */
7691 op = ((insn >> 22) & 6) | ((insn >> 7) & 1);
7692 if (op < 4 && (insn & 0xf000) != 0xf000)
7695 case 0: /* Register controlled shift. */
7696 tmp = load_reg(s, rn);
7697 tmp2 = load_reg(s, rm);
7698 if ((insn & 0x70) != 0)
7700 op = (insn >> 21) & 3;
7701 logic_cc = (insn & (1 << 20)) != 0;
7702 gen_arm_shift_reg(tmp, op, tmp2, logic_cc);
7705 store_reg_bx(env, s, rd, tmp);
7707 case 1: /* Sign/zero extend. */
7708 tmp = load_reg(s, rm);
7709 shift = (insn >> 4) & 3;
7710 /* ??? In many cases it's not neccessary to do a
7711 rotate, a shift is sufficient. */
7713 tcg_gen_rotri_i32(tmp, tmp, shift * 8);
7714 op = (insn >> 20) & 7;
7716 case 0: gen_sxth(tmp); break;
7717 case 1: gen_uxth(tmp); break;
7718 case 2: gen_sxtb16(tmp); break;
7719 case 3: gen_uxtb16(tmp); break;
7720 case 4: gen_sxtb(tmp); break;
7721 case 5: gen_uxtb(tmp); break;
7722 default: goto illegal_op;
7725 tmp2 = load_reg(s, rn);
7726 if ((op >> 1) == 1) {
7727 gen_add16(tmp, tmp2);
7729 tcg_gen_add_i32(tmp, tmp, tmp2);
7733 store_reg(s, rd, tmp);
7735 case 2: /* SIMD add/subtract. */
7736 op = (insn >> 20) & 7;
7737 shift = (insn >> 4) & 7;
7738 if ((op & 3) == 3 || (shift & 3) == 3)
7740 tmp = load_reg(s, rn);
7741 tmp2 = load_reg(s, rm);
7742 gen_thumb2_parallel_addsub(op, shift, tmp, tmp2);
7744 store_reg(s, rd, tmp);
7746 case 3: /* Other data processing. */
7747 op = ((insn >> 17) & 0x38) | ((insn >> 4) & 7);
7749 /* Saturating add/subtract. */
7750 tmp = load_reg(s, rn);
7751 tmp2 = load_reg(s, rm);
7753 gen_helper_double_saturate(tmp, tmp);
7755 gen_helper_sub_saturate(tmp, tmp2, tmp);
7757 gen_helper_add_saturate(tmp, tmp, tmp2);
7760 tmp = load_reg(s, rn);
7762 case 0x0a: /* rbit */
7763 gen_helper_rbit(tmp, tmp);
7765 case 0x08: /* rev */
7766 tcg_gen_bswap32_i32(tmp, tmp);
7768 case 0x09: /* rev16 */
7771 case 0x0b: /* revsh */
7774 case 0x10: /* sel */
7775 tmp2 = load_reg(s, rm);
7777 tcg_gen_ld_i32(tmp3, cpu_env, offsetof(CPUState, GE));
7778 gen_helper_sel_flags(tmp, tmp3, tmp, tmp2);
7782 case 0x18: /* clz */
7783 gen_helper_clz(tmp, tmp);
7789 store_reg(s, rd, tmp);
7791 case 4: case 5: /* 32-bit multiply. Sum of absolute differences. */
7792 op = (insn >> 4) & 0xf;
7793 tmp = load_reg(s, rn);
7794 tmp2 = load_reg(s, rm);
7795 switch ((insn >> 20) & 7) {
7796 case 0: /* 32 x 32 -> 32 */
7797 tcg_gen_mul_i32(tmp, tmp, tmp2);
7800 tmp2 = load_reg(s, rs);
7802 tcg_gen_sub_i32(tmp, tmp2, tmp);
7804 tcg_gen_add_i32(tmp, tmp, tmp2);
7808 case 1: /* 16 x 16 -> 32 */
7809 gen_mulxy(tmp, tmp2, op & 2, op & 1);
7812 tmp2 = load_reg(s, rs);
7813 gen_helper_add_setq(tmp, tmp, tmp2);
7817 case 2: /* Dual multiply add. */
7818 case 4: /* Dual multiply subtract. */
7820 gen_swap_half(tmp2);
7821 gen_smul_dual(tmp, tmp2);
7822 /* This addition cannot overflow. */
7823 if (insn & (1 << 22)) {
7824 tcg_gen_sub_i32(tmp, tmp, tmp2);
7826 tcg_gen_add_i32(tmp, tmp, tmp2);
7831 tmp2 = load_reg(s, rs);
7832 gen_helper_add_setq(tmp, tmp, tmp2);
7836 case 3: /* 32 * 16 -> 32msb */
7838 tcg_gen_sari_i32(tmp2, tmp2, 16);
7841 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7842 tcg_gen_shri_i64(tmp64, tmp64, 16);
7844 tcg_gen_trunc_i64_i32(tmp, tmp64);
7845 tcg_temp_free_i64(tmp64);
7848 tmp2 = load_reg(s, rs);
7849 gen_helper_add_setq(tmp, tmp, tmp2);
7853 case 5: case 6: /* 32 * 32 -> 32msb (SMMUL, SMMLA, SMMLS) */
7854 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7856 tmp = load_reg(s, rs);
7857 if (insn & (1 << 20)) {
7858 tmp64 = gen_addq_msw(tmp64, tmp);
7860 tmp64 = gen_subq_msw(tmp64, tmp);
7863 if (insn & (1 << 4)) {
7864 tcg_gen_addi_i64(tmp64, tmp64, 0x80000000u);
7866 tcg_gen_shri_i64(tmp64, tmp64, 32);
7868 tcg_gen_trunc_i64_i32(tmp, tmp64);
7869 tcg_temp_free_i64(tmp64);
7871 case 7: /* Unsigned sum of absolute differences. */
7872 gen_helper_usad8(tmp, tmp, tmp2);
7875 tmp2 = load_reg(s, rs);
7876 tcg_gen_add_i32(tmp, tmp, tmp2);
7881 store_reg(s, rd, tmp);
7883 case 6: case 7: /* 64-bit multiply, Divide. */
7884 op = ((insn >> 4) & 0xf) | ((insn >> 16) & 0x70);
7885 tmp = load_reg(s, rn);
7886 tmp2 = load_reg(s, rm);
7887 if ((op & 0x50) == 0x10) {
7889 if (!arm_feature(env, ARM_FEATURE_DIV))
7892 gen_helper_udiv(tmp, tmp, tmp2);
7894 gen_helper_sdiv(tmp, tmp, tmp2);
7896 store_reg(s, rd, tmp);
7897 } else if ((op & 0xe) == 0xc) {
7898 /* Dual multiply accumulate long. */
7900 gen_swap_half(tmp2);
7901 gen_smul_dual(tmp, tmp2);
7903 tcg_gen_sub_i32(tmp, tmp, tmp2);
7905 tcg_gen_add_i32(tmp, tmp, tmp2);
7909 tmp64 = tcg_temp_new_i64();
7910 tcg_gen_ext_i32_i64(tmp64, tmp);
7912 gen_addq(s, tmp64, rs, rd);
7913 gen_storeq_reg(s, rs, rd, tmp64);
7914 tcg_temp_free_i64(tmp64);
7917 /* Unsigned 64-bit multiply */
7918 tmp64 = gen_mulu_i64_i32(tmp, tmp2);
7922 gen_mulxy(tmp, tmp2, op & 2, op & 1);
7924 tmp64 = tcg_temp_new_i64();
7925 tcg_gen_ext_i32_i64(tmp64, tmp);
7928 /* Signed 64-bit multiply */
7929 tmp64 = gen_muls_i64_i32(tmp, tmp2);
7934 gen_addq_lo(s, tmp64, rs);
7935 gen_addq_lo(s, tmp64, rd);
7936 } else if (op & 0x40) {
7937 /* 64-bit accumulate. */
7938 gen_addq(s, tmp64, rs, rd);
7940 gen_storeq_reg(s, rs, rd, tmp64);
7941 tcg_temp_free_i64(tmp64);
7946 case 6: case 7: case 14: case 15:
7948 if (((insn >> 24) & 3) == 3) {
7949 /* Translate into the equivalent ARM encoding. */
7950 insn = (insn & 0xe2ffffff) | ((insn & (1 << 28)) >> 4);
7951 if (disas_neon_data_insn(env, s, insn))
7954 if (insn & (1 << 28))
7956 if (disas_coproc_insn (env, s, insn))
7960 case 8: case 9: case 10: case 11:
7961 if (insn & (1 << 15)) {
7962 /* Branches, misc control. */
7963 if (insn & 0x5000) {
7964 /* Unconditional branch. */
7965 /* signextend(hw1[10:0]) -> offset[:12]. */
7966 offset = ((int32_t)insn << 5) >> 9 & ~(int32_t)0xfff;
7967 /* hw1[10:0] -> offset[11:1]. */
7968 offset |= (insn & 0x7ff) << 1;
7969 /* (~hw2[13, 11] ^ offset[24]) -> offset[23,22]
7970 offset[24:22] already have the same value because of the
7971 sign extension above. */
7972 offset ^= ((~insn) & (1 << 13)) << 10;
7973 offset ^= ((~insn) & (1 << 11)) << 11;
7975 if (insn & (1 << 14)) {
7976 /* Branch and link. */
7977 tcg_gen_movi_i32(cpu_R[14], s->pc | 1);
7981 if (insn & (1 << 12)) {
7986 offset &= ~(uint32_t)2;
7987 gen_bx_im(s, offset);
7989 } else if (((insn >> 23) & 7) == 7) {
7991 if (insn & (1 << 13))
7994 if (insn & (1 << 26)) {
7995 /* Secure monitor call (v6Z) */
7996 goto illegal_op; /* not implemented. */
7998 op = (insn >> 20) & 7;
8000 case 0: /* msr cpsr. */
8002 tmp = load_reg(s, rn);
8003 addr = tcg_const_i32(insn & 0xff);
8004 gen_helper_v7m_msr(cpu_env, addr, tmp);
8005 tcg_temp_free_i32(addr);
8011 case 1: /* msr spsr. */
8014 tmp = load_reg(s, rn);
8016 msr_mask(env, s, (insn >> 8) & 0xf, op == 1),
8020 case 2: /* cps, nop-hint. */
8021 if (((insn >> 8) & 7) == 0) {
8022 gen_nop_hint(s, insn & 0xff);
8024 /* Implemented as NOP in user mode. */
8029 if (insn & (1 << 10)) {
8030 if (insn & (1 << 7))
8032 if (insn & (1 << 6))
8034 if (insn & (1 << 5))
8036 if (insn & (1 << 9))
8037 imm = CPSR_A | CPSR_I | CPSR_F;
8039 if (insn & (1 << 8)) {
8041 imm |= (insn & 0x1f);
8044 gen_set_psr_im(s, offset, 0, imm);
8047 case 3: /* Special control operations. */
8049 op = (insn >> 4) & 0xf;
8057 /* These execute as NOPs. */
8064 /* Trivial implementation equivalent to bx. */
8065 tmp = load_reg(s, rn);
8068 case 5: /* Exception return. */
8072 if (rn != 14 || rd != 15) {
8075 tmp = load_reg(s, rn);
8076 tcg_gen_subi_i32(tmp, tmp, insn & 0xff);
8077 gen_exception_return(s, tmp);
8079 case 6: /* mrs cpsr. */
8082 addr = tcg_const_i32(insn & 0xff);
8083 gen_helper_v7m_mrs(tmp, cpu_env, addr);
8084 tcg_temp_free_i32(addr);
8086 gen_helper_cpsr_read(tmp);
8088 store_reg(s, rd, tmp);
8090 case 7: /* mrs spsr. */
8091 /* Not accessible in user mode. */
8092 if (IS_USER(s) || IS_M(env))
8094 tmp = load_cpu_field(spsr);
8095 store_reg(s, rd, tmp);
8100 /* Conditional branch. */
8101 op = (insn >> 22) & 0xf;
8102 /* Generate a conditional jump to next instruction. */
8103 s->condlabel = gen_new_label();
8104 gen_test_cc(op ^ 1, s->condlabel);
8107 /* offset[11:1] = insn[10:0] */
8108 offset = (insn & 0x7ff) << 1;
8109 /* offset[17:12] = insn[21:16]. */
8110 offset |= (insn & 0x003f0000) >> 4;
8111 /* offset[31:20] = insn[26]. */
8112 offset |= ((int32_t)((insn << 5) & 0x80000000)) >> 11;
8113 /* offset[18] = insn[13]. */
8114 offset |= (insn & (1 << 13)) << 5;
8115 /* offset[19] = insn[11]. */
8116 offset |= (insn & (1 << 11)) << 8;
8118 /* jump to the offset */
8119 gen_jmp(s, s->pc + offset);
8122 /* Data processing immediate. */
8123 if (insn & (1 << 25)) {
8124 if (insn & (1 << 24)) {
8125 if (insn & (1 << 20))
8127 /* Bitfield/Saturate. */
8128 op = (insn >> 21) & 7;
8130 shift = ((insn >> 6) & 3) | ((insn >> 10) & 0x1c);
8133 tcg_gen_movi_i32(tmp, 0);
8135 tmp = load_reg(s, rn);
8138 case 2: /* Signed bitfield extract. */
8140 if (shift + imm > 32)
8143 gen_sbfx(tmp, shift, imm);
8145 case 6: /* Unsigned bitfield extract. */
8147 if (shift + imm > 32)
8150 gen_ubfx(tmp, shift, (1u << imm) - 1);
8152 case 3: /* Bitfield insert/clear. */
8155 imm = imm + 1 - shift;
8157 tmp2 = load_reg(s, rd);
8158 gen_bfi(tmp, tmp2, tmp, shift, (1u << imm) - 1);
8164 default: /* Saturate. */
8167 tcg_gen_sari_i32(tmp, tmp, shift);
8169 tcg_gen_shli_i32(tmp, tmp, shift);
8171 tmp2 = tcg_const_i32(imm);
8174 if ((op & 1) && shift == 0)
8175 gen_helper_usat16(tmp, tmp, tmp2);
8177 gen_helper_usat(tmp, tmp, tmp2);
8180 if ((op & 1) && shift == 0)
8181 gen_helper_ssat16(tmp, tmp, tmp2);
8183 gen_helper_ssat(tmp, tmp, tmp2);
8185 tcg_temp_free_i32(tmp2);
8188 store_reg(s, rd, tmp);
8190 imm = ((insn & 0x04000000) >> 15)
8191 | ((insn & 0x7000) >> 4) | (insn & 0xff);
8192 if (insn & (1 << 22)) {
8193 /* 16-bit immediate. */
8194 imm |= (insn >> 4) & 0xf000;
8195 if (insn & (1 << 23)) {
8197 tmp = load_reg(s, rd);
8198 tcg_gen_ext16u_i32(tmp, tmp);
8199 tcg_gen_ori_i32(tmp, tmp, imm << 16);
8203 tcg_gen_movi_i32(tmp, imm);
8206 /* Add/sub 12-bit immediate. */
8208 offset = s->pc & ~(uint32_t)3;
8209 if (insn & (1 << 23))
8214 tcg_gen_movi_i32(tmp, offset);
8216 tmp = load_reg(s, rn);
8217 if (insn & (1 << 23))
8218 tcg_gen_subi_i32(tmp, tmp, imm);
8220 tcg_gen_addi_i32(tmp, tmp, imm);
8223 store_reg(s, rd, tmp);
8226 int shifter_out = 0;
8227 /* modified 12-bit immediate. */
8228 shift = ((insn & 0x04000000) >> 23) | ((insn & 0x7000) >> 12);
8229 imm = (insn & 0xff);
8232 /* Nothing to do. */
8234 case 1: /* 00XY00XY */
8237 case 2: /* XY00XY00 */
8241 case 3: /* XYXYXYXY */
8245 default: /* Rotated constant. */
8246 shift = (shift << 1) | (imm >> 7);
8248 imm = imm << (32 - shift);
8253 tcg_gen_movi_i32(tmp2, imm);
8254 rn = (insn >> 16) & 0xf;
8257 tcg_gen_movi_i32(tmp, 0);
8259 tmp = load_reg(s, rn);
8261 op = (insn >> 21) & 0xf;
8262 if (gen_thumb2_data_op(s, op, (insn & (1 << 20)) != 0,
8263 shifter_out, tmp, tmp2))
8266 rd = (insn >> 8) & 0xf;
8268 store_reg(s, rd, tmp);
8275 case 12: /* Load/store single data item. */
8280 if ((insn & 0x01100000) == 0x01000000) {
8281 if (disas_neon_ls_insn(env, s, insn))
8289 /* s->pc has already been incremented by 4. */
8290 imm = s->pc & 0xfffffffc;
8291 if (insn & (1 << 23))
8292 imm += insn & 0xfff;
8294 imm -= insn & 0xfff;
8295 tcg_gen_movi_i32(addr, imm);
8297 addr = load_reg(s, rn);
8298 if (insn & (1 << 23)) {
8299 /* Positive offset. */
8301 tcg_gen_addi_i32(addr, addr, imm);
8303 op = (insn >> 8) & 7;
8306 case 0: case 8: /* Shifted Register. */
8307 shift = (insn >> 4) & 0xf;
8310 tmp = load_reg(s, rm);
8312 tcg_gen_shli_i32(tmp, tmp, shift);
8313 tcg_gen_add_i32(addr, addr, tmp);
8316 case 4: /* Negative offset. */
8317 tcg_gen_addi_i32(addr, addr, -imm);
8319 case 6: /* User privilege. */
8320 tcg_gen_addi_i32(addr, addr, imm);
8323 case 1: /* Post-decrement. */
8326 case 3: /* Post-increment. */
8330 case 5: /* Pre-decrement. */
8333 case 7: /* Pre-increment. */
8334 tcg_gen_addi_i32(addr, addr, imm);
8342 op = ((insn >> 21) & 3) | ((insn >> 22) & 4);
8343 if (insn & (1 << 20)) {
8345 if (rs == 15 && op != 2) {
8348 /* Memory hint. Implemented as NOP. */
8351 case 0: tmp = gen_ld8u(addr, user); break;
8352 case 4: tmp = gen_ld8s(addr, user); break;
8353 case 1: tmp = gen_ld16u(addr, user); break;
8354 case 5: tmp = gen_ld16s(addr, user); break;
8355 case 2: tmp = gen_ld32(addr, user); break;
8356 default: goto illegal_op;
8361 store_reg(s, rs, tmp);
8368 tmp = load_reg(s, rs);
8370 case 0: gen_st8(tmp, addr, user); break;
8371 case 1: gen_st16(tmp, addr, user); break;
8372 case 2: gen_st32(tmp, addr, user); break;
8373 default: goto illegal_op;
8377 tcg_gen_addi_i32(addr, addr, imm);
8379 store_reg(s, rn, addr);
8393 static void disas_thumb_insn(CPUState *env, DisasContext *s)
8395 uint32_t val, insn, op, rm, rn, rd, shift, cond;
8402 if (s->condexec_mask) {
8403 cond = s->condexec_cond;
8404 if (cond != 0x0e) { /* Skip conditional when condition is AL. */
8405 s->condlabel = gen_new_label();
8406 gen_test_cc(cond ^ 1, s->condlabel);
8411 insn = lduw_code(s->pc);
8414 switch (insn >> 12) {
8418 op = (insn >> 11) & 3;
8421 rn = (insn >> 3) & 7;
8422 tmp = load_reg(s, rn);
8423 if (insn & (1 << 10)) {
8426 tcg_gen_movi_i32(tmp2, (insn >> 6) & 7);
8429 rm = (insn >> 6) & 7;
8430 tmp2 = load_reg(s, rm);
8432 if (insn & (1 << 9)) {
8433 if (s->condexec_mask)
8434 tcg_gen_sub_i32(tmp, tmp, tmp2);
8436 gen_helper_sub_cc(tmp, tmp, tmp2);
8438 if (s->condexec_mask)
8439 tcg_gen_add_i32(tmp, tmp, tmp2);
8441 gen_helper_add_cc(tmp, tmp, tmp2);
8444 store_reg(s, rd, tmp);
8446 /* shift immediate */
8447 rm = (insn >> 3) & 7;
8448 shift = (insn >> 6) & 0x1f;
8449 tmp = load_reg(s, rm);
8450 gen_arm_shift_im(tmp, op, shift, s->condexec_mask == 0);
8451 if (!s->condexec_mask)
8453 store_reg(s, rd, tmp);
8457 /* arithmetic large immediate */
8458 op = (insn >> 11) & 3;
8459 rd = (insn >> 8) & 0x7;
8460 if (op == 0) { /* mov */
8462 tcg_gen_movi_i32(tmp, insn & 0xff);
8463 if (!s->condexec_mask)
8465 store_reg(s, rd, tmp);
8467 tmp = load_reg(s, rd);
8469 tcg_gen_movi_i32(tmp2, insn & 0xff);
8472 gen_helper_sub_cc(tmp, tmp, tmp2);
8477 if (s->condexec_mask)
8478 tcg_gen_add_i32(tmp, tmp, tmp2);
8480 gen_helper_add_cc(tmp, tmp, tmp2);
8482 store_reg(s, rd, tmp);
8485 if (s->condexec_mask)
8486 tcg_gen_sub_i32(tmp, tmp, tmp2);
8488 gen_helper_sub_cc(tmp, tmp, tmp2);
8490 store_reg(s, rd, tmp);
8496 if (insn & (1 << 11)) {
8497 rd = (insn >> 8) & 7;
8498 /* load pc-relative. Bit 1 of PC is ignored. */
8499 val = s->pc + 2 + ((insn & 0xff) * 4);
8500 val &= ~(uint32_t)2;
8502 tcg_gen_movi_i32(addr, val);
8503 tmp = gen_ld32(addr, IS_USER(s));
8505 store_reg(s, rd, tmp);
8508 if (insn & (1 << 10)) {
8509 /* data processing extended or blx */
8510 rd = (insn & 7) | ((insn >> 4) & 8);
8511 rm = (insn >> 3) & 0xf;
8512 op = (insn >> 8) & 3;
8515 tmp = load_reg(s, rd);
8516 tmp2 = load_reg(s, rm);
8517 tcg_gen_add_i32(tmp, tmp, tmp2);
8519 store_reg(s, rd, tmp);
8522 tmp = load_reg(s, rd);
8523 tmp2 = load_reg(s, rm);
8524 gen_helper_sub_cc(tmp, tmp, tmp2);
8528 case 2: /* mov/cpy */
8529 tmp = load_reg(s, rm);
8530 store_reg(s, rd, tmp);
8532 case 3:/* branch [and link] exchange thumb register */
8533 tmp = load_reg(s, rm);
8534 if (insn & (1 << 7)) {
8535 val = (uint32_t)s->pc | 1;
8537 tcg_gen_movi_i32(tmp2, val);
8538 store_reg(s, 14, tmp2);
8546 /* data processing register */
8548 rm = (insn >> 3) & 7;
8549 op = (insn >> 6) & 0xf;
8550 if (op == 2 || op == 3 || op == 4 || op == 7) {
8551 /* the shift/rotate ops want the operands backwards */
8560 if (op == 9) { /* neg */
8562 tcg_gen_movi_i32(tmp, 0);
8563 } else if (op != 0xf) { /* mvn doesn't read its first operand */
8564 tmp = load_reg(s, rd);
8569 tmp2 = load_reg(s, rm);
8572 tcg_gen_and_i32(tmp, tmp, tmp2);
8573 if (!s->condexec_mask)
8577 tcg_gen_xor_i32(tmp, tmp, tmp2);
8578 if (!s->condexec_mask)
8582 if (s->condexec_mask) {
8583 gen_helper_shl(tmp2, tmp2, tmp);
8585 gen_helper_shl_cc(tmp2, tmp2, tmp);
8590 if (s->condexec_mask) {
8591 gen_helper_shr(tmp2, tmp2, tmp);
8593 gen_helper_shr_cc(tmp2, tmp2, tmp);
8598 if (s->condexec_mask) {
8599 gen_helper_sar(tmp2, tmp2, tmp);
8601 gen_helper_sar_cc(tmp2, tmp2, tmp);
8606 if (s->condexec_mask)
8609 gen_helper_adc_cc(tmp, tmp, tmp2);
8612 if (s->condexec_mask)
8613 gen_sub_carry(tmp, tmp, tmp2);
8615 gen_helper_sbc_cc(tmp, tmp, tmp2);
8618 if (s->condexec_mask) {
8619 tcg_gen_andi_i32(tmp, tmp, 0x1f);
8620 tcg_gen_rotr_i32(tmp2, tmp2, tmp);
8622 gen_helper_ror_cc(tmp2, tmp2, tmp);
8627 tcg_gen_and_i32(tmp, tmp, tmp2);
8632 if (s->condexec_mask)
8633 tcg_gen_neg_i32(tmp, tmp2);
8635 gen_helper_sub_cc(tmp, tmp, tmp2);
8638 gen_helper_sub_cc(tmp, tmp, tmp2);
8642 gen_helper_add_cc(tmp, tmp, tmp2);
8646 tcg_gen_or_i32(tmp, tmp, tmp2);
8647 if (!s->condexec_mask)
8651 tcg_gen_mul_i32(tmp, tmp, tmp2);
8652 if (!s->condexec_mask)
8656 tcg_gen_andc_i32(tmp, tmp, tmp2);
8657 if (!s->condexec_mask)
8661 tcg_gen_not_i32(tmp2, tmp2);
8662 if (!s->condexec_mask)
8670 store_reg(s, rm, tmp2);
8674 store_reg(s, rd, tmp);
8684 /* load/store register offset. */
8686 rn = (insn >> 3) & 7;
8687 rm = (insn >> 6) & 7;
8688 op = (insn >> 9) & 7;
8689 addr = load_reg(s, rn);
8690 tmp = load_reg(s, rm);
8691 tcg_gen_add_i32(addr, addr, tmp);
8694 if (op < 3) /* store */
8695 tmp = load_reg(s, rd);
8699 gen_st32(tmp, addr, IS_USER(s));
8702 gen_st16(tmp, addr, IS_USER(s));
8705 gen_st8(tmp, addr, IS_USER(s));
8708 tmp = gen_ld8s(addr, IS_USER(s));
8711 tmp = gen_ld32(addr, IS_USER(s));
8714 tmp = gen_ld16u(addr, IS_USER(s));
8717 tmp = gen_ld8u(addr, IS_USER(s));
8720 tmp = gen_ld16s(addr, IS_USER(s));
8723 if (op >= 3) /* load */
8724 store_reg(s, rd, tmp);
8729 /* load/store word immediate offset */
8731 rn = (insn >> 3) & 7;
8732 addr = load_reg(s, rn);
8733 val = (insn >> 4) & 0x7c;
8734 tcg_gen_addi_i32(addr, addr, val);
8736 if (insn & (1 << 11)) {
8738 tmp = gen_ld32(addr, IS_USER(s));
8739 store_reg(s, rd, tmp);
8742 tmp = load_reg(s, rd);
8743 gen_st32(tmp, addr, IS_USER(s));
8749 /* load/store byte immediate offset */
8751 rn = (insn >> 3) & 7;
8752 addr = load_reg(s, rn);
8753 val = (insn >> 6) & 0x1f;
8754 tcg_gen_addi_i32(addr, addr, val);
8756 if (insn & (1 << 11)) {
8758 tmp = gen_ld8u(addr, IS_USER(s));
8759 store_reg(s, rd, tmp);
8762 tmp = load_reg(s, rd);
8763 gen_st8(tmp, addr, IS_USER(s));
8769 /* load/store halfword immediate offset */
8771 rn = (insn >> 3) & 7;
8772 addr = load_reg(s, rn);
8773 val = (insn >> 5) & 0x3e;
8774 tcg_gen_addi_i32(addr, addr, val);
8776 if (insn & (1 << 11)) {
8778 tmp = gen_ld16u(addr, IS_USER(s));
8779 store_reg(s, rd, tmp);
8782 tmp = load_reg(s, rd);
8783 gen_st16(tmp, addr, IS_USER(s));
8789 /* load/store from stack */
8790 rd = (insn >> 8) & 7;
8791 addr = load_reg(s, 13);
8792 val = (insn & 0xff) * 4;
8793 tcg_gen_addi_i32(addr, addr, val);
8795 if (insn & (1 << 11)) {
8797 tmp = gen_ld32(addr, IS_USER(s));
8798 store_reg(s, rd, tmp);
8801 tmp = load_reg(s, rd);
8802 gen_st32(tmp, addr, IS_USER(s));
8808 /* add to high reg */
8809 rd = (insn >> 8) & 7;
8810 if (insn & (1 << 11)) {
8812 tmp = load_reg(s, 13);
8814 /* PC. bit 1 is ignored. */
8816 tcg_gen_movi_i32(tmp, (s->pc + 2) & ~(uint32_t)2);
8818 val = (insn & 0xff) * 4;
8819 tcg_gen_addi_i32(tmp, tmp, val);
8820 store_reg(s, rd, tmp);
8825 op = (insn >> 8) & 0xf;
8828 /* adjust stack pointer */
8829 tmp = load_reg(s, 13);
8830 val = (insn & 0x7f) * 4;
8831 if (insn & (1 << 7))
8832 val = -(int32_t)val;
8833 tcg_gen_addi_i32(tmp, tmp, val);
8834 store_reg(s, 13, tmp);
8837 case 2: /* sign/zero extend. */
8840 rm = (insn >> 3) & 7;
8841 tmp = load_reg(s, rm);
8842 switch ((insn >> 6) & 3) {
8843 case 0: gen_sxth(tmp); break;
8844 case 1: gen_sxtb(tmp); break;
8845 case 2: gen_uxth(tmp); break;
8846 case 3: gen_uxtb(tmp); break;
8848 store_reg(s, rd, tmp);
8850 case 4: case 5: case 0xc: case 0xd:
8852 addr = load_reg(s, 13);
8853 if (insn & (1 << 8))
8857 for (i = 0; i < 8; i++) {
8858 if (insn & (1 << i))
8861 if ((insn & (1 << 11)) == 0) {
8862 tcg_gen_addi_i32(addr, addr, -offset);
8864 for (i = 0; i < 8; i++) {
8865 if (insn & (1 << i)) {
8866 if (insn & (1 << 11)) {
8868 tmp = gen_ld32(addr, IS_USER(s));
8869 store_reg(s, i, tmp);
8872 tmp = load_reg(s, i);
8873 gen_st32(tmp, addr, IS_USER(s));
8875 /* advance to the next address. */
8876 tcg_gen_addi_i32(addr, addr, 4);
8880 if (insn & (1 << 8)) {
8881 if (insn & (1 << 11)) {
8883 tmp = gen_ld32(addr, IS_USER(s));
8884 /* don't set the pc until the rest of the instruction
8888 tmp = load_reg(s, 14);
8889 gen_st32(tmp, addr, IS_USER(s));
8891 tcg_gen_addi_i32(addr, addr, 4);
8893 if ((insn & (1 << 11)) == 0) {
8894 tcg_gen_addi_i32(addr, addr, -offset);
8896 /* write back the new stack pointer */
8897 store_reg(s, 13, addr);
8898 /* set the new PC value */
8899 if ((insn & 0x0900) == 0x0900)
8903 case 1: case 3: case 9: case 11: /* czb */
8905 tmp = load_reg(s, rm);
8906 s->condlabel = gen_new_label();
8908 if (insn & (1 << 11))
8909 tcg_gen_brcondi_i32(TCG_COND_EQ, tmp, 0, s->condlabel);
8911 tcg_gen_brcondi_i32(TCG_COND_NE, tmp, 0, s->condlabel);
8913 offset = ((insn & 0xf8) >> 2) | (insn & 0x200) >> 3;
8914 val = (uint32_t)s->pc + 2;
8919 case 15: /* IT, nop-hint. */
8920 if ((insn & 0xf) == 0) {
8921 gen_nop_hint(s, (insn >> 4) & 0xf);
8925 s->condexec_cond = (insn >> 4) & 0xe;
8926 s->condexec_mask = insn & 0x1f;
8927 /* No actual code generated for this insn, just setup state. */
8930 case 0xe: /* bkpt */
8931 gen_exception_insn(s, 2, EXCP_BKPT);
8936 rn = (insn >> 3) & 0x7;
8938 tmp = load_reg(s, rn);
8939 switch ((insn >> 6) & 3) {
8940 case 0: tcg_gen_bswap32_i32(tmp, tmp); break;
8941 case 1: gen_rev16(tmp); break;
8942 case 3: gen_revsh(tmp); break;
8943 default: goto illegal_op;
8945 store_reg(s, rd, tmp);
8953 tmp = tcg_const_i32((insn & (1 << 4)) != 0);
8956 addr = tcg_const_i32(16);
8957 gen_helper_v7m_msr(cpu_env, addr, tmp);
8958 tcg_temp_free_i32(addr);
8962 addr = tcg_const_i32(17);
8963 gen_helper_v7m_msr(cpu_env, addr, tmp);
8964 tcg_temp_free_i32(addr);
8966 tcg_temp_free_i32(tmp);
8969 if (insn & (1 << 4))
8970 shift = CPSR_A | CPSR_I | CPSR_F;
8973 gen_set_psr_im(s, ((insn & 7) << 6), 0, shift);
8983 /* load/store multiple */
8984 rn = (insn >> 8) & 0x7;
8985 addr = load_reg(s, rn);
8986 for (i = 0; i < 8; i++) {
8987 if (insn & (1 << i)) {
8988 if (insn & (1 << 11)) {
8990 tmp = gen_ld32(addr, IS_USER(s));
8991 store_reg(s, i, tmp);
8994 tmp = load_reg(s, i);
8995 gen_st32(tmp, addr, IS_USER(s));
8997 /* advance to the next address */
8998 tcg_gen_addi_i32(addr, addr, 4);
9001 /* Base register writeback. */
9002 if ((insn & (1 << rn)) == 0) {
9003 store_reg(s, rn, addr);
9010 /* conditional branch or swi */
9011 cond = (insn >> 8) & 0xf;
9017 gen_set_pc_im(s->pc);
9018 s->is_jmp = DISAS_SWI;
9021 /* generate a conditional jump to next instruction */
9022 s->condlabel = gen_new_label();
9023 gen_test_cc(cond ^ 1, s->condlabel);
9026 /* jump to the offset */
9027 val = (uint32_t)s->pc + 2;
9028 offset = ((int32_t)insn << 24) >> 24;
9034 if (insn & (1 << 11)) {
9035 if (disas_thumb2_insn(env, s, insn))
9039 /* unconditional branch */
9040 val = (uint32_t)s->pc;
9041 offset = ((int32_t)insn << 21) >> 21;
9042 val += (offset << 1) + 2;
9047 if (disas_thumb2_insn(env, s, insn))
9053 gen_exception_insn(s, 4, EXCP_UDEF);
9057 gen_exception_insn(s, 2, EXCP_UDEF);
9060 /* generate intermediate code in gen_opc_buf and gen_opparam_buf for
9061 basic block 'tb'. If search_pc is TRUE, also generate PC
9062 information for each intermediate instruction. */
9063 static inline void gen_intermediate_code_internal(CPUState *env,
9064 TranslationBlock *tb,
9067 DisasContext dc1, *dc = &dc1;
9069 uint16_t *gen_opc_end;
9071 target_ulong pc_start;
9072 uint32_t next_page_start;
9076 /* generate intermediate code */
9083 gen_opc_end = gen_opc_buf + OPC_MAX_SIZE;
9085 dc->is_jmp = DISAS_NEXT;
9087 dc->singlestep_enabled = env->singlestep_enabled;
9089 dc->thumb = ARM_TBFLAG_THUMB(tb->flags);
9090 dc->condexec_mask = (ARM_TBFLAG_CONDEXEC(tb->flags) & 0xf) << 1;
9091 dc->condexec_cond = ARM_TBFLAG_CONDEXEC(tb->flags) >> 4;
9092 #if !defined(CONFIG_USER_ONLY)
9093 dc->user = (ARM_TBFLAG_PRIV(tb->flags) == 0);
9095 dc->vfp_enabled = ARM_TBFLAG_VFPEN(tb->flags);
9096 dc->vec_len = ARM_TBFLAG_VECLEN(tb->flags);
9097 dc->vec_stride = ARM_TBFLAG_VECSTRIDE(tb->flags);
9098 cpu_F0s = tcg_temp_new_i32();
9099 cpu_F1s = tcg_temp_new_i32();
9100 cpu_F0d = tcg_temp_new_i64();
9101 cpu_F1d = tcg_temp_new_i64();
9104 /* FIXME: cpu_M0 can probably be the same as cpu_V0. */
9105 cpu_M0 = tcg_temp_new_i64();
9106 next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
9109 max_insns = tb->cflags & CF_COUNT_MASK;
9111 max_insns = CF_COUNT_MASK;
9115 /* A note on handling of the condexec (IT) bits:
9117 * We want to avoid the overhead of having to write the updated condexec
9118 * bits back to the CPUState for every instruction in an IT block. So:
9119 * (1) if the condexec bits are not already zero then we write
9120 * zero back into the CPUState now. This avoids complications trying
9121 * to do it at the end of the block. (For example if we don't do this
9122 * it's hard to identify whether we can safely skip writing condexec
9123 * at the end of the TB, which we definitely want to do for the case
9124 * where a TB doesn't do anything with the IT state at all.)
9125 * (2) if we are going to leave the TB then we call gen_set_condexec()
9126 * which will write the correct value into CPUState if zero is wrong.
9127 * This is done both for leaving the TB at the end, and for leaving
9128 * it because of an exception we know will happen, which is done in
9129 * gen_exception_insn(). The latter is necessary because we need to
9130 * leave the TB with the PC/IT state just prior to execution of the
9131 * instruction which caused the exception.
9132 * (3) if we leave the TB unexpectedly (eg a data abort on a load)
9133 * then the CPUState will be wrong and we need to reset it.
9134 * This is handled in the same way as restoration of the
9135 * PC in these situations: we will be called again with search_pc=1
9136 * and generate a mapping of the condexec bits for each PC in
9137 * gen_opc_condexec_bits[]. gen_pc_load[] then uses this to restore
9138 * the condexec bits.
9140 * Note that there are no instructions which can read the condexec
9141 * bits, and none which can write non-static values to them, so
9142 * we don't need to care about whether CPUState is correct in the
9146 /* Reset the conditional execution bits immediately. This avoids
9147 complications trying to do it at the end of the block. */
9148 if (dc->condexec_mask || dc->condexec_cond)
9150 TCGv tmp = new_tmp();
9151 tcg_gen_movi_i32(tmp, 0);
9152 store_cpu_field(tmp, condexec_bits);
9155 #ifdef CONFIG_USER_ONLY
9156 /* Intercept jump to the magic kernel page. */
9157 if (dc->pc >= 0xffff0000) {
9158 /* We always get here via a jump, so know we are not in a
9159 conditional execution block. */
9160 gen_exception(EXCP_KERNEL_TRAP);
9161 dc->is_jmp = DISAS_UPDATE;
9165 if (dc->pc >= 0xfffffff0 && IS_M(env)) {
9166 /* We always get here via a jump, so know we are not in a
9167 conditional execution block. */
9168 gen_exception(EXCP_EXCEPTION_EXIT);
9169 dc->is_jmp = DISAS_UPDATE;
9174 if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
9175 QTAILQ_FOREACH(bp, &env->breakpoints, entry) {
9176 if (bp->pc == dc->pc) {
9177 gen_exception_insn(dc, 0, EXCP_DEBUG);
9178 /* Advance PC so that clearing the breakpoint will
9179 invalidate this TB. */
9181 goto done_generating;
9187 j = gen_opc_ptr - gen_opc_buf;
9191 gen_opc_instr_start[lj++] = 0;
9193 gen_opc_pc[lj] = dc->pc;
9194 gen_opc_condexec_bits[lj] = (dc->condexec_cond << 4) | (dc->condexec_mask >> 1);
9195 gen_opc_instr_start[lj] = 1;
9196 gen_opc_icount[lj] = num_insns;
9199 if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO))
9202 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP))) {
9203 tcg_gen_debug_insn_start(dc->pc);
9207 disas_thumb_insn(env, dc);
9208 if (dc->condexec_mask) {
9209 dc->condexec_cond = (dc->condexec_cond & 0xe)
9210 | ((dc->condexec_mask >> 4) & 1);
9211 dc->condexec_mask = (dc->condexec_mask << 1) & 0x1f;
9212 if (dc->condexec_mask == 0) {
9213 dc->condexec_cond = 0;
9217 disas_arm_insn(env, dc);
9220 fprintf(stderr, "Internal resource leak before %08x\n", dc->pc);
9224 if (dc->condjmp && !dc->is_jmp) {
9225 gen_set_label(dc->condlabel);
9228 /* Translation stops when a conditional branch is encountered.
9229 * Otherwise the subsequent code could get translated several times.
9230 * Also stop translation when a page boundary is reached. This
9231 * ensures prefetch aborts occur at the right place. */
9233 } while (!dc->is_jmp && gen_opc_ptr < gen_opc_end &&
9234 !env->singlestep_enabled &&
9236 dc->pc < next_page_start &&
9237 num_insns < max_insns);
9239 if (tb->cflags & CF_LAST_IO) {
9241 /* FIXME: This can theoretically happen with self-modifying
9243 cpu_abort(env, "IO on conditional branch instruction");
9248 /* At this stage dc->condjmp will only be set when the skipped
9249 instruction was a conditional branch or trap, and the PC has
9250 already been written. */
9251 if (unlikely(env->singlestep_enabled)) {
9252 /* Make sure the pc is updated, and raise a debug exception. */
9254 gen_set_condexec(dc);
9255 if (dc->is_jmp == DISAS_SWI) {
9256 gen_exception(EXCP_SWI);
9258 gen_exception(EXCP_DEBUG);
9260 gen_set_label(dc->condlabel);
9262 if (dc->condjmp || !dc->is_jmp) {
9263 gen_set_pc_im(dc->pc);
9266 gen_set_condexec(dc);
9267 if (dc->is_jmp == DISAS_SWI && !dc->condjmp) {
9268 gen_exception(EXCP_SWI);
9270 /* FIXME: Single stepping a WFI insn will not halt
9272 gen_exception(EXCP_DEBUG);
9275 /* While branches must always occur at the end of an IT block,
9276 there are a few other things that can cause us to terminate
9277 the TB in the middel of an IT block:
9278 - Exception generating instructions (bkpt, swi, undefined).
9280 - Hardware watchpoints.
9281 Hardware breakpoints have already been handled and skip this code.
9283 gen_set_condexec(dc);
9284 switch(dc->is_jmp) {
9286 gen_goto_tb(dc, 1, dc->pc);
9291 /* indicate that the hash table must be used to find the next TB */
9295 /* nothing more to generate */
9301 gen_exception(EXCP_SWI);
9305 gen_set_label(dc->condlabel);
9306 gen_set_condexec(dc);
9307 gen_goto_tb(dc, 1, dc->pc);
9313 gen_icount_end(tb, num_insns);
9314 *gen_opc_ptr = INDEX_op_end;
9317 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
9318 qemu_log("----------------\n");
9319 qemu_log("IN: %s\n", lookup_symbol(pc_start));
9320 log_target_disas(pc_start, dc->pc - pc_start, dc->thumb);
9325 j = gen_opc_ptr - gen_opc_buf;
9328 gen_opc_instr_start[lj++] = 0;
9330 tb->size = dc->pc - pc_start;
9331 tb->icount = num_insns;
9335 void gen_intermediate_code(CPUState *env, TranslationBlock *tb)
9337 gen_intermediate_code_internal(env, tb, 0);
9340 void gen_intermediate_code_pc(CPUState *env, TranslationBlock *tb)
9342 gen_intermediate_code_internal(env, tb, 1);
9345 static const char *cpu_mode_names[16] = {
9346 "usr", "fiq", "irq", "svc", "???", "???", "???", "abt",
9347 "???", "???", "???", "und", "???", "???", "???", "sys"
9350 void cpu_dump_state(CPUState *env, FILE *f, fprintf_function cpu_fprintf,
9360 /* ??? This assumes float64 and double have the same layout.
9361 Oh well, it's only debug dumps. */
9370 cpu_fprintf(f, "R%02d=%08x", i, env->regs[i]);
9372 cpu_fprintf(f, "\n");
9374 cpu_fprintf(f, " ");
9376 psr = cpsr_read(env);
9377 cpu_fprintf(f, "PSR=%08x %c%c%c%c %c %s%d\n",
9379 psr & (1 << 31) ? 'N' : '-',
9380 psr & (1 << 30) ? 'Z' : '-',
9381 psr & (1 << 29) ? 'C' : '-',
9382 psr & (1 << 28) ? 'V' : '-',
9383 psr & CPSR_T ? 'T' : 'A',
9384 cpu_mode_names[psr & 0xf], (psr & 0x10) ? 32 : 26);
9387 for (i = 0; i < 16; i++) {
9388 d.d = env->vfp.regs[i];
9392 cpu_fprintf(f, "s%02d=%08x(%8g) s%02d=%08x(%8g) d%02d=%08x%08x(%8g)\n",
9393 i * 2, (int)s0.i, s0.s,
9394 i * 2 + 1, (int)s1.i, s1.s,
9395 i, (int)(uint32_t)d.l.upper, (int)(uint32_t)d.l.lower,
9398 cpu_fprintf(f, "FPSCR: %08x\n", (int)env->vfp.xregs[ARM_VFP_FPSCR]);
9402 void gen_pc_load(CPUState *env, TranslationBlock *tb,
9403 unsigned long searched_pc, int pc_pos, void *puc)
9405 env->regs[15] = gen_opc_pc[pc_pos];
9406 env->condexec_bits = gen_opc_condexec_bits[pc_pos];