6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
28 #include "translate.h"
29 #include "qemu/host-utils.h"
31 #include "exec/gen-icount.h"
37 static TCGv_i64 cpu_X[32];
38 static TCGv_i64 cpu_pc;
39 static TCGv_i32 cpu_NF, cpu_ZF, cpu_CF, cpu_VF;
41 /* Load/store exclusive handling */
42 static TCGv_i64 cpu_exclusive_addr;
43 static TCGv_i64 cpu_exclusive_val;
44 static TCGv_i64 cpu_exclusive_high;
45 #ifdef CONFIG_USER_ONLY
46 static TCGv_i64 cpu_exclusive_test;
47 static TCGv_i32 cpu_exclusive_info;
50 static const char *regnames[] = {
51 "x0", "x1", "x2", "x3", "x4", "x5", "x6", "x7",
52 "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15",
53 "x16", "x17", "x18", "x19", "x20", "x21", "x22", "x23",
54 "x24", "x25", "x26", "x27", "x28", "x29", "lr", "sp"
58 A64_SHIFT_TYPE_LSL = 0,
59 A64_SHIFT_TYPE_LSR = 1,
60 A64_SHIFT_TYPE_ASR = 2,
61 A64_SHIFT_TYPE_ROR = 3
64 /* Table based decoder typedefs - used when the relevant bits for decode
65 * are too awkwardly scattered across the instruction (eg SIMD).
67 typedef void AArch64DecodeFn(DisasContext *s, uint32_t insn);
69 typedef struct AArch64DecodeTable {
72 AArch64DecodeFn *disas_fn;
75 /* initialize TCG globals. */
76 void a64_translate_init(void)
80 cpu_pc = tcg_global_mem_new_i64(TCG_AREG0,
81 offsetof(CPUARMState, pc),
83 for (i = 0; i < 32; i++) {
84 cpu_X[i] = tcg_global_mem_new_i64(TCG_AREG0,
85 offsetof(CPUARMState, xregs[i]),
89 cpu_NF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, NF), "NF");
90 cpu_ZF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, ZF), "ZF");
91 cpu_CF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, CF), "CF");
92 cpu_VF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, VF), "VF");
94 cpu_exclusive_addr = tcg_global_mem_new_i64(TCG_AREG0,
95 offsetof(CPUARMState, exclusive_addr), "exclusive_addr");
96 cpu_exclusive_val = tcg_global_mem_new_i64(TCG_AREG0,
97 offsetof(CPUARMState, exclusive_val), "exclusive_val");
98 cpu_exclusive_high = tcg_global_mem_new_i64(TCG_AREG0,
99 offsetof(CPUARMState, exclusive_high), "exclusive_high");
100 #ifdef CONFIG_USER_ONLY
101 cpu_exclusive_test = tcg_global_mem_new_i64(TCG_AREG0,
102 offsetof(CPUARMState, exclusive_test), "exclusive_test");
103 cpu_exclusive_info = tcg_global_mem_new_i32(TCG_AREG0,
104 offsetof(CPUARMState, exclusive_info), "exclusive_info");
108 void aarch64_cpu_dump_state(CPUState *cs, FILE *f,
109 fprintf_function cpu_fprintf, int flags)
111 ARMCPU *cpu = ARM_CPU(cs);
112 CPUARMState *env = &cpu->env;
113 uint32_t psr = pstate_read(env);
116 cpu_fprintf(f, "PC=%016"PRIx64" SP=%016"PRIx64"\n",
117 env->pc, env->xregs[31]);
118 for (i = 0; i < 31; i++) {
119 cpu_fprintf(f, "X%02d=%016"PRIx64, i, env->xregs[i]);
121 cpu_fprintf(f, "\n");
126 cpu_fprintf(f, "PSTATE=%08x (flags %c%c%c%c)\n",
128 psr & PSTATE_N ? 'N' : '-',
129 psr & PSTATE_Z ? 'Z' : '-',
130 psr & PSTATE_C ? 'C' : '-',
131 psr & PSTATE_V ? 'V' : '-');
132 cpu_fprintf(f, "\n");
134 if (flags & CPU_DUMP_FPU) {
136 for (i = 0; i < numvfpregs; i += 2) {
137 uint64_t vlo = float64_val(env->vfp.regs[i * 2]);
138 uint64_t vhi = float64_val(env->vfp.regs[(i * 2) + 1]);
139 cpu_fprintf(f, "q%02d=%016" PRIx64 ":%016" PRIx64 " ",
141 vlo = float64_val(env->vfp.regs[(i + 1) * 2]);
142 vhi = float64_val(env->vfp.regs[((i + 1) * 2) + 1]);
143 cpu_fprintf(f, "q%02d=%016" PRIx64 ":%016" PRIx64 "\n",
146 cpu_fprintf(f, "FPCR: %08x FPSR: %08x\n",
147 vfp_get_fpcr(env), vfp_get_fpsr(env));
151 static int get_mem_index(DisasContext *s)
153 #ifdef CONFIG_USER_ONLY
160 void gen_a64_set_pc_im(uint64_t val)
162 tcg_gen_movi_i64(cpu_pc, val);
165 static void gen_exception(int excp)
167 TCGv_i32 tmp = tcg_temp_new_i32();
168 tcg_gen_movi_i32(tmp, excp);
169 gen_helper_exception(cpu_env, tmp);
170 tcg_temp_free_i32(tmp);
173 static void gen_exception_insn(DisasContext *s, int offset, int excp)
175 gen_a64_set_pc_im(s->pc - offset);
177 s->is_jmp = DISAS_EXC;
180 static inline bool use_goto_tb(DisasContext *s, int n, uint64_t dest)
182 /* No direct tb linking with singlestep or deterministic io */
183 if (s->singlestep_enabled || (s->tb->cflags & CF_LAST_IO)) {
187 /* Only link tbs from inside the same guest page */
188 if ((s->tb->pc & TARGET_PAGE_MASK) != (dest & TARGET_PAGE_MASK)) {
195 static inline void gen_goto_tb(DisasContext *s, int n, uint64_t dest)
197 TranslationBlock *tb;
200 if (use_goto_tb(s, n, dest)) {
202 gen_a64_set_pc_im(dest);
203 tcg_gen_exit_tb((tcg_target_long)tb + n);
204 s->is_jmp = DISAS_TB_JUMP;
206 gen_a64_set_pc_im(dest);
207 if (s->singlestep_enabled) {
208 gen_exception(EXCP_DEBUG);
211 s->is_jmp = DISAS_JUMP;
215 static void unallocated_encoding(DisasContext *s)
217 gen_exception_insn(s, 4, EXCP_UDEF);
220 #define unsupported_encoding(s, insn) \
222 qemu_log_mask(LOG_UNIMP, \
223 "%s:%d: unsupported instruction encoding 0x%08x " \
224 "at pc=%016" PRIx64 "\n", \
225 __FILE__, __LINE__, insn, s->pc - 4); \
226 unallocated_encoding(s); \
229 static void init_tmp_a64_array(DisasContext *s)
231 #ifdef CONFIG_DEBUG_TCG
233 for (i = 0; i < ARRAY_SIZE(s->tmp_a64); i++) {
234 TCGV_UNUSED_I64(s->tmp_a64[i]);
237 s->tmp_a64_count = 0;
240 static void free_tmp_a64(DisasContext *s)
243 for (i = 0; i < s->tmp_a64_count; i++) {
244 tcg_temp_free_i64(s->tmp_a64[i]);
246 init_tmp_a64_array(s);
249 static TCGv_i64 new_tmp_a64(DisasContext *s)
251 assert(s->tmp_a64_count < TMP_A64_MAX);
252 return s->tmp_a64[s->tmp_a64_count++] = tcg_temp_new_i64();
255 static TCGv_i64 new_tmp_a64_zero(DisasContext *s)
257 TCGv_i64 t = new_tmp_a64(s);
258 tcg_gen_movi_i64(t, 0);
263 * Register access functions
265 * These functions are used for directly accessing a register in where
266 * changes to the final register value are likely to be made. If you
267 * need to use a register for temporary calculation (e.g. index type
268 * operations) use the read_* form.
270 * B1.2.1 Register mappings
272 * In instruction register encoding 31 can refer to ZR (zero register) or
273 * the SP (stack pointer) depending on context. In QEMU's case we map SP
274 * to cpu_X[31] and ZR accesses to a temporary which can be discarded.
275 * This is the point of the _sp forms.
277 static TCGv_i64 cpu_reg(DisasContext *s, int reg)
280 return new_tmp_a64_zero(s);
286 /* register access for when 31 == SP */
287 static TCGv_i64 cpu_reg_sp(DisasContext *s, int reg)
292 /* read a cpu register in 32bit/64bit mode. Returns a TCGv_i64
293 * representing the register contents. This TCGv is an auto-freed
294 * temporary so it need not be explicitly freed, and may be modified.
296 static TCGv_i64 read_cpu_reg(DisasContext *s, int reg, int sf)
298 TCGv_i64 v = new_tmp_a64(s);
301 tcg_gen_mov_i64(v, cpu_X[reg]);
303 tcg_gen_ext32u_i64(v, cpu_X[reg]);
306 tcg_gen_movi_i64(v, 0);
311 static TCGv_i64 read_cpu_reg_sp(DisasContext *s, int reg, int sf)
313 TCGv_i64 v = new_tmp_a64(s);
315 tcg_gen_mov_i64(v, cpu_X[reg]);
317 tcg_gen_ext32u_i64(v, cpu_X[reg]);
322 /* Return the offset into CPUARMState of an element of specified
323 * size, 'element' places in from the least significant end of
324 * the FP/vector register Qn.
326 static inline int vec_reg_offset(int regno, int element, TCGMemOp size)
328 int offs = offsetof(CPUARMState, vfp.regs[regno * 2]);
329 #ifdef HOST_WORDS_BIGENDIAN
330 /* This is complicated slightly because vfp.regs[2n] is
331 * still the low half and vfp.regs[2n+1] the high half
332 * of the 128 bit vector, even on big endian systems.
333 * Calculate the offset assuming a fully bigendian 128 bits,
334 * then XOR to account for the order of the two 64 bit halves.
336 offs += (16 - ((element + 1) * (1 << size)));
339 offs += element * (1 << size);
344 /* Return the offset into CPUARMState of a slice (from
345 * the least significant end) of FP register Qn (ie
347 * (Note that this is not the same mapping as for A32; see cpu.h)
349 static inline int fp_reg_offset(int regno, TCGMemOp size)
351 int offs = offsetof(CPUARMState, vfp.regs[regno * 2]);
352 #ifdef HOST_WORDS_BIGENDIAN
353 offs += (8 - (1 << size));
358 /* Offset of the high half of the 128 bit vector Qn */
359 static inline int fp_reg_hi_offset(int regno)
361 return offsetof(CPUARMState, vfp.regs[regno * 2 + 1]);
364 /* Convenience accessors for reading and writing single and double
365 * FP registers. Writing clears the upper parts of the associated
366 * 128 bit vector register, as required by the architecture.
367 * Note that unlike the GP register accessors, the values returned
368 * by the read functions must be manually freed.
370 static TCGv_i64 read_fp_dreg(DisasContext *s, int reg)
372 TCGv_i64 v = tcg_temp_new_i64();
374 tcg_gen_ld_i64(v, cpu_env, fp_reg_offset(reg, MO_64));
378 static TCGv_i32 read_fp_sreg(DisasContext *s, int reg)
380 TCGv_i32 v = tcg_temp_new_i32();
382 tcg_gen_ld_i32(v, cpu_env, fp_reg_offset(reg, MO_32));
386 static void write_fp_dreg(DisasContext *s, int reg, TCGv_i64 v)
388 TCGv_i64 tcg_zero = tcg_const_i64(0);
390 tcg_gen_st_i64(v, cpu_env, fp_reg_offset(reg, MO_64));
391 tcg_gen_st_i64(tcg_zero, cpu_env, fp_reg_hi_offset(reg));
392 tcg_temp_free_i64(tcg_zero);
395 static void write_fp_sreg(DisasContext *s, int reg, TCGv_i32 v)
397 TCGv_i64 tmp = tcg_temp_new_i64();
399 tcg_gen_extu_i32_i64(tmp, v);
400 write_fp_dreg(s, reg, tmp);
401 tcg_temp_free_i64(tmp);
404 static TCGv_ptr get_fpstatus_ptr(void)
406 TCGv_ptr statusptr = tcg_temp_new_ptr();
409 /* In A64 all instructions (both FP and Neon) use the FPCR;
410 * there is no equivalent of the A32 Neon "standard FPSCR value"
411 * and all operations use vfp.fp_status.
413 offset = offsetof(CPUARMState, vfp.fp_status);
414 tcg_gen_addi_ptr(statusptr, cpu_env, offset);
418 /* Set ZF and NF based on a 64 bit result. This is alas fiddlier
419 * than the 32 bit equivalent.
421 static inline void gen_set_NZ64(TCGv_i64 result)
423 TCGv_i64 flag = tcg_temp_new_i64();
425 tcg_gen_setcondi_i64(TCG_COND_NE, flag, result, 0);
426 tcg_gen_trunc_i64_i32(cpu_ZF, flag);
427 tcg_gen_shri_i64(flag, result, 32);
428 tcg_gen_trunc_i64_i32(cpu_NF, flag);
429 tcg_temp_free_i64(flag);
432 /* Set NZCV as for a logical operation: NZ as per result, CV cleared. */
433 static inline void gen_logic_CC(int sf, TCGv_i64 result)
436 gen_set_NZ64(result);
438 tcg_gen_trunc_i64_i32(cpu_ZF, result);
439 tcg_gen_trunc_i64_i32(cpu_NF, result);
441 tcg_gen_movi_i32(cpu_CF, 0);
442 tcg_gen_movi_i32(cpu_VF, 0);
445 /* dest = T0 + T1; compute C, N, V and Z flags */
446 static void gen_add_CC(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
449 TCGv_i64 result, flag, tmp;
450 result = tcg_temp_new_i64();
451 flag = tcg_temp_new_i64();
452 tmp = tcg_temp_new_i64();
454 tcg_gen_movi_i64(tmp, 0);
455 tcg_gen_add2_i64(result, flag, t0, tmp, t1, tmp);
457 tcg_gen_trunc_i64_i32(cpu_CF, flag);
459 gen_set_NZ64(result);
461 tcg_gen_xor_i64(flag, result, t0);
462 tcg_gen_xor_i64(tmp, t0, t1);
463 tcg_gen_andc_i64(flag, flag, tmp);
464 tcg_temp_free_i64(tmp);
465 tcg_gen_shri_i64(flag, flag, 32);
466 tcg_gen_trunc_i64_i32(cpu_VF, flag);
468 tcg_gen_mov_i64(dest, result);
469 tcg_temp_free_i64(result);
470 tcg_temp_free_i64(flag);
472 /* 32 bit arithmetic */
473 TCGv_i32 t0_32 = tcg_temp_new_i32();
474 TCGv_i32 t1_32 = tcg_temp_new_i32();
475 TCGv_i32 tmp = tcg_temp_new_i32();
477 tcg_gen_movi_i32(tmp, 0);
478 tcg_gen_trunc_i64_i32(t0_32, t0);
479 tcg_gen_trunc_i64_i32(t1_32, t1);
480 tcg_gen_add2_i32(cpu_NF, cpu_CF, t0_32, tmp, t1_32, tmp);
481 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
482 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0_32);
483 tcg_gen_xor_i32(tmp, t0_32, t1_32);
484 tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
485 tcg_gen_extu_i32_i64(dest, cpu_NF);
487 tcg_temp_free_i32(tmp);
488 tcg_temp_free_i32(t0_32);
489 tcg_temp_free_i32(t1_32);
493 /* dest = T0 - T1; compute C, N, V and Z flags */
494 static void gen_sub_CC(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
497 /* 64 bit arithmetic */
498 TCGv_i64 result, flag, tmp;
500 result = tcg_temp_new_i64();
501 flag = tcg_temp_new_i64();
502 tcg_gen_sub_i64(result, t0, t1);
504 gen_set_NZ64(result);
506 tcg_gen_setcond_i64(TCG_COND_GEU, flag, t0, t1);
507 tcg_gen_trunc_i64_i32(cpu_CF, flag);
509 tcg_gen_xor_i64(flag, result, t0);
510 tmp = tcg_temp_new_i64();
511 tcg_gen_xor_i64(tmp, t0, t1);
512 tcg_gen_and_i64(flag, flag, tmp);
513 tcg_temp_free_i64(tmp);
514 tcg_gen_shri_i64(flag, flag, 32);
515 tcg_gen_trunc_i64_i32(cpu_VF, flag);
516 tcg_gen_mov_i64(dest, result);
517 tcg_temp_free_i64(flag);
518 tcg_temp_free_i64(result);
520 /* 32 bit arithmetic */
521 TCGv_i32 t0_32 = tcg_temp_new_i32();
522 TCGv_i32 t1_32 = tcg_temp_new_i32();
525 tcg_gen_trunc_i64_i32(t0_32, t0);
526 tcg_gen_trunc_i64_i32(t1_32, t1);
527 tcg_gen_sub_i32(cpu_NF, t0_32, t1_32);
528 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
529 tcg_gen_setcond_i32(TCG_COND_GEU, cpu_CF, t0_32, t1_32);
530 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0_32);
531 tmp = tcg_temp_new_i32();
532 tcg_gen_xor_i32(tmp, t0_32, t1_32);
533 tcg_temp_free_i32(t0_32);
534 tcg_temp_free_i32(t1_32);
535 tcg_gen_and_i32(cpu_VF, cpu_VF, tmp);
536 tcg_temp_free_i32(tmp);
537 tcg_gen_extu_i32_i64(dest, cpu_NF);
541 /* dest = T0 + T1 + CF; do not compute flags. */
542 static void gen_adc(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
544 TCGv_i64 flag = tcg_temp_new_i64();
545 tcg_gen_extu_i32_i64(flag, cpu_CF);
546 tcg_gen_add_i64(dest, t0, t1);
547 tcg_gen_add_i64(dest, dest, flag);
548 tcg_temp_free_i64(flag);
551 tcg_gen_ext32u_i64(dest, dest);
555 /* dest = T0 + T1 + CF; compute C, N, V and Z flags. */
556 static void gen_adc_CC(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
559 TCGv_i64 result, cf_64, vf_64, tmp;
560 result = tcg_temp_new_i64();
561 cf_64 = tcg_temp_new_i64();
562 vf_64 = tcg_temp_new_i64();
563 tmp = tcg_const_i64(0);
565 tcg_gen_extu_i32_i64(cf_64, cpu_CF);
566 tcg_gen_add2_i64(result, cf_64, t0, tmp, cf_64, tmp);
567 tcg_gen_add2_i64(result, cf_64, result, cf_64, t1, tmp);
568 tcg_gen_trunc_i64_i32(cpu_CF, cf_64);
569 gen_set_NZ64(result);
571 tcg_gen_xor_i64(vf_64, result, t0);
572 tcg_gen_xor_i64(tmp, t0, t1);
573 tcg_gen_andc_i64(vf_64, vf_64, tmp);
574 tcg_gen_shri_i64(vf_64, vf_64, 32);
575 tcg_gen_trunc_i64_i32(cpu_VF, vf_64);
577 tcg_gen_mov_i64(dest, result);
579 tcg_temp_free_i64(tmp);
580 tcg_temp_free_i64(vf_64);
581 tcg_temp_free_i64(cf_64);
582 tcg_temp_free_i64(result);
584 TCGv_i32 t0_32, t1_32, tmp;
585 t0_32 = tcg_temp_new_i32();
586 t1_32 = tcg_temp_new_i32();
587 tmp = tcg_const_i32(0);
589 tcg_gen_trunc_i64_i32(t0_32, t0);
590 tcg_gen_trunc_i64_i32(t1_32, t1);
591 tcg_gen_add2_i32(cpu_NF, cpu_CF, t0_32, tmp, cpu_CF, tmp);
592 tcg_gen_add2_i32(cpu_NF, cpu_CF, cpu_NF, cpu_CF, t1_32, tmp);
594 tcg_gen_mov_i32(cpu_ZF, cpu_NF);
595 tcg_gen_xor_i32(cpu_VF, cpu_NF, t0_32);
596 tcg_gen_xor_i32(tmp, t0_32, t1_32);
597 tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
598 tcg_gen_extu_i32_i64(dest, cpu_NF);
600 tcg_temp_free_i32(tmp);
601 tcg_temp_free_i32(t1_32);
602 tcg_temp_free_i32(t0_32);
607 * Load/Store generators
611 * Store from GPR register to memory
613 static void do_gpr_st(DisasContext *s, TCGv_i64 source,
614 TCGv_i64 tcg_addr, int size)
617 tcg_gen_qemu_st_i64(source, tcg_addr, get_mem_index(s), MO_TE + size);
621 * Load from memory to GPR register
623 static void do_gpr_ld(DisasContext *s, TCGv_i64 dest, TCGv_i64 tcg_addr,
624 int size, bool is_signed, bool extend)
626 TCGMemOp memop = MO_TE + size;
634 tcg_gen_qemu_ld_i64(dest, tcg_addr, get_mem_index(s), memop);
636 if (extend && is_signed) {
638 tcg_gen_ext32u_i64(dest, dest);
643 * Store from FP register to memory
645 static void do_fp_st(DisasContext *s, int srcidx, TCGv_i64 tcg_addr, int size)
647 /* This writes the bottom N bits of a 128 bit wide vector to memory */
648 TCGv_i64 tmp = tcg_temp_new_i64();
649 tcg_gen_ld_i64(tmp, cpu_env, fp_reg_offset(srcidx, MO_64));
651 tcg_gen_qemu_st_i64(tmp, tcg_addr, get_mem_index(s), MO_TE + size);
653 TCGv_i64 tcg_hiaddr = tcg_temp_new_i64();
654 tcg_gen_qemu_st_i64(tmp, tcg_addr, get_mem_index(s), MO_TEQ);
655 tcg_gen_qemu_st64(tmp, tcg_addr, get_mem_index(s));
656 tcg_gen_ld_i64(tmp, cpu_env, fp_reg_hi_offset(srcidx));
657 tcg_gen_addi_i64(tcg_hiaddr, tcg_addr, 8);
658 tcg_gen_qemu_st_i64(tmp, tcg_hiaddr, get_mem_index(s), MO_TEQ);
659 tcg_temp_free_i64(tcg_hiaddr);
662 tcg_temp_free_i64(tmp);
666 * Load from memory to FP register
668 static void do_fp_ld(DisasContext *s, int destidx, TCGv_i64 tcg_addr, int size)
670 /* This always zero-extends and writes to a full 128 bit wide vector */
671 TCGv_i64 tmplo = tcg_temp_new_i64();
675 TCGMemOp memop = MO_TE + size;
676 tmphi = tcg_const_i64(0);
677 tcg_gen_qemu_ld_i64(tmplo, tcg_addr, get_mem_index(s), memop);
680 tmphi = tcg_temp_new_i64();
681 tcg_hiaddr = tcg_temp_new_i64();
683 tcg_gen_qemu_ld_i64(tmplo, tcg_addr, get_mem_index(s), MO_TEQ);
684 tcg_gen_addi_i64(tcg_hiaddr, tcg_addr, 8);
685 tcg_gen_qemu_ld_i64(tmphi, tcg_hiaddr, get_mem_index(s), MO_TEQ);
686 tcg_temp_free_i64(tcg_hiaddr);
689 tcg_gen_st_i64(tmplo, cpu_env, fp_reg_offset(destidx, MO_64));
690 tcg_gen_st_i64(tmphi, cpu_env, fp_reg_hi_offset(destidx));
692 tcg_temp_free_i64(tmplo);
693 tcg_temp_free_i64(tmphi);
697 * Vector load/store helpers.
699 * The principal difference between this and a FP load is that we don't
700 * zero extend as we are filling a partial chunk of the vector register.
701 * These functions don't support 128 bit loads/stores, which would be
702 * normal load/store operations.
705 /* Get value of an element within a vector register */
706 static void read_vec_element(DisasContext *s, TCGv_i64 tcg_dest, int srcidx,
707 int element, TCGMemOp memop)
709 int vect_off = vec_reg_offset(srcidx, element, memop & MO_SIZE);
712 tcg_gen_ld8u_i64(tcg_dest, cpu_env, vect_off);
715 tcg_gen_ld16u_i64(tcg_dest, cpu_env, vect_off);
718 tcg_gen_ld32u_i64(tcg_dest, cpu_env, vect_off);
721 tcg_gen_ld8s_i64(tcg_dest, cpu_env, vect_off);
724 tcg_gen_ld16s_i64(tcg_dest, cpu_env, vect_off);
727 tcg_gen_ld32s_i64(tcg_dest, cpu_env, vect_off);
731 tcg_gen_ld_i64(tcg_dest, cpu_env, vect_off);
734 g_assert_not_reached();
738 /* Set value of an element within a vector register */
739 static void write_vec_element(DisasContext *s, TCGv_i64 tcg_src, int destidx,
740 int element, TCGMemOp memop)
742 int vect_off = vec_reg_offset(destidx, element, memop & MO_SIZE);
745 tcg_gen_st8_i64(tcg_src, cpu_env, vect_off);
748 tcg_gen_st16_i64(tcg_src, cpu_env, vect_off);
751 tcg_gen_st32_i64(tcg_src, cpu_env, vect_off);
754 tcg_gen_st_i64(tcg_src, cpu_env, vect_off);
757 g_assert_not_reached();
761 /* Clear the high 64 bits of a 128 bit vector (in general non-quad
762 * vector ops all need to do this).
764 static void clear_vec_high(DisasContext *s, int rd)
766 TCGv_i64 tcg_zero = tcg_const_i64(0);
768 write_vec_element(s, tcg_zero, rd, 1, MO_64);
769 tcg_temp_free_i64(tcg_zero);
772 /* Store from vector register to memory */
773 static void do_vec_st(DisasContext *s, int srcidx, int element,
774 TCGv_i64 tcg_addr, int size)
776 TCGMemOp memop = MO_TE + size;
777 TCGv_i64 tcg_tmp = tcg_temp_new_i64();
779 read_vec_element(s, tcg_tmp, srcidx, element, size);
780 tcg_gen_qemu_st_i64(tcg_tmp, tcg_addr, get_mem_index(s), memop);
782 tcg_temp_free_i64(tcg_tmp);
785 /* Load from memory to vector register */
786 static void do_vec_ld(DisasContext *s, int destidx, int element,
787 TCGv_i64 tcg_addr, int size)
789 TCGMemOp memop = MO_TE + size;
790 TCGv_i64 tcg_tmp = tcg_temp_new_i64();
792 tcg_gen_qemu_ld_i64(tcg_tmp, tcg_addr, get_mem_index(s), memop);
793 write_vec_element(s, tcg_tmp, destidx, element, size);
795 tcg_temp_free_i64(tcg_tmp);
799 * This utility function is for doing register extension with an
800 * optional shift. You will likely want to pass a temporary for the
801 * destination register. See DecodeRegExtend() in the ARM ARM.
803 static void ext_and_shift_reg(TCGv_i64 tcg_out, TCGv_i64 tcg_in,
804 int option, unsigned int shift)
806 int extsize = extract32(option, 0, 2);
807 bool is_signed = extract32(option, 2, 1);
812 tcg_gen_ext8s_i64(tcg_out, tcg_in);
815 tcg_gen_ext16s_i64(tcg_out, tcg_in);
818 tcg_gen_ext32s_i64(tcg_out, tcg_in);
821 tcg_gen_mov_i64(tcg_out, tcg_in);
827 tcg_gen_ext8u_i64(tcg_out, tcg_in);
830 tcg_gen_ext16u_i64(tcg_out, tcg_in);
833 tcg_gen_ext32u_i64(tcg_out, tcg_in);
836 tcg_gen_mov_i64(tcg_out, tcg_in);
842 tcg_gen_shli_i64(tcg_out, tcg_out, shift);
846 static inline void gen_check_sp_alignment(DisasContext *s)
848 /* The AArch64 architecture mandates that (if enabled via PSTATE
849 * or SCTLR bits) there is a check that SP is 16-aligned on every
850 * SP-relative load or store (with an exception generated if it is not).
851 * In line with general QEMU practice regarding misaligned accesses,
852 * we omit these checks for the sake of guest program performance.
853 * This function is provided as a hook so we can more easily add these
854 * checks in future (possibly as a "favour catching guest program bugs
855 * over speed" user selectable option).
860 * This provides a simple table based table lookup decoder. It is
861 * intended to be used when the relevant bits for decode are too
862 * awkwardly placed and switch/if based logic would be confusing and
863 * deeply nested. Since it's a linear search through the table, tables
864 * should be kept small.
866 * It returns the first handler where insn & mask == pattern, or
867 * NULL if there is no match.
868 * The table is terminated by an empty mask (i.e. 0)
870 static inline AArch64DecodeFn *lookup_disas_fn(const AArch64DecodeTable *table,
873 const AArch64DecodeTable *tptr = table;
876 if ((insn & tptr->mask) == tptr->pattern) {
877 return tptr->disas_fn;
885 * the instruction disassembly implemented here matches
886 * the instruction encoding classifications in chapter 3 (C3)
887 * of the ARM Architecture Reference Manual (DDI0487A_a)
890 /* C3.2.7 Unconditional branch (immediate)
892 * +----+-----------+-------------------------------------+
893 * | op | 0 0 1 0 1 | imm26 |
894 * +----+-----------+-------------------------------------+
896 static void disas_uncond_b_imm(DisasContext *s, uint32_t insn)
898 uint64_t addr = s->pc + sextract32(insn, 0, 26) * 4 - 4;
900 if (insn & (1 << 31)) {
901 /* C5.6.26 BL Branch with link */
902 tcg_gen_movi_i64(cpu_reg(s, 30), s->pc);
905 /* C5.6.20 B Branch / C5.6.26 BL Branch with link */
906 gen_goto_tb(s, 0, addr);
909 /* C3.2.1 Compare & branch (immediate)
910 * 31 30 25 24 23 5 4 0
911 * +----+-------------+----+---------------------+--------+
912 * | sf | 0 1 1 0 1 0 | op | imm19 | Rt |
913 * +----+-------------+----+---------------------+--------+
915 static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
917 unsigned int sf, op, rt;
922 sf = extract32(insn, 31, 1);
923 op = extract32(insn, 24, 1); /* 0: CBZ; 1: CBNZ */
924 rt = extract32(insn, 0, 5);
925 addr = s->pc + sextract32(insn, 5, 19) * 4 - 4;
927 tcg_cmp = read_cpu_reg(s, rt, sf);
928 label_match = gen_new_label();
930 tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
931 tcg_cmp, 0, label_match);
933 gen_goto_tb(s, 0, s->pc);
934 gen_set_label(label_match);
935 gen_goto_tb(s, 1, addr);
938 /* C3.2.5 Test & branch (immediate)
939 * 31 30 25 24 23 19 18 5 4 0
940 * +----+-------------+----+-------+-------------+------+
941 * | b5 | 0 1 1 0 1 1 | op | b40 | imm14 | Rt |
942 * +----+-------------+----+-------+-------------+------+
944 static void disas_test_b_imm(DisasContext *s, uint32_t insn)
946 unsigned int bit_pos, op, rt;
951 bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5);
952 op = extract32(insn, 24, 1); /* 0: TBZ; 1: TBNZ */
953 addr = s->pc + sextract32(insn, 5, 14) * 4 - 4;
954 rt = extract32(insn, 0, 5);
956 tcg_cmp = tcg_temp_new_i64();
957 tcg_gen_andi_i64(tcg_cmp, cpu_reg(s, rt), (1ULL << bit_pos));
958 label_match = gen_new_label();
959 tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
960 tcg_cmp, 0, label_match);
961 tcg_temp_free_i64(tcg_cmp);
962 gen_goto_tb(s, 0, s->pc);
963 gen_set_label(label_match);
964 gen_goto_tb(s, 1, addr);
967 /* C3.2.2 / C5.6.19 Conditional branch (immediate)
968 * 31 25 24 23 5 4 3 0
969 * +---------------+----+---------------------+----+------+
970 * | 0 1 0 1 0 1 0 | o1 | imm19 | o0 | cond |
971 * +---------------+----+---------------------+----+------+
973 static void disas_cond_b_imm(DisasContext *s, uint32_t insn)
978 if ((insn & (1 << 4)) || (insn & (1 << 24))) {
979 unallocated_encoding(s);
982 addr = s->pc + sextract32(insn, 5, 19) * 4 - 4;
983 cond = extract32(insn, 0, 4);
986 /* genuinely conditional branches */
987 int label_match = gen_new_label();
988 arm_gen_test_cc(cond, label_match);
989 gen_goto_tb(s, 0, s->pc);
990 gen_set_label(label_match);
991 gen_goto_tb(s, 1, addr);
993 /* 0xe and 0xf are both "always" conditions */
994 gen_goto_tb(s, 0, addr);
999 static void handle_hint(DisasContext *s, uint32_t insn,
1000 unsigned int op1, unsigned int op2, unsigned int crm)
1002 unsigned int selector = crm << 3 | op2;
1005 unallocated_encoding(s);
1017 /* we treat all as NOP at least for now */
1020 /* default specified as NOP equivalent */
1025 static void gen_clrex(DisasContext *s, uint32_t insn)
1027 tcg_gen_movi_i64(cpu_exclusive_addr, -1);
1030 /* CLREX, DSB, DMB, ISB */
1031 static void handle_sync(DisasContext *s, uint32_t insn,
1032 unsigned int op1, unsigned int op2, unsigned int crm)
1035 unallocated_encoding(s);
1046 /* We don't emulate caches so barriers are no-ops */
1049 unallocated_encoding(s);
1054 /* C5.6.130 MSR (immediate) - move immediate to processor state field */
1055 static void handle_msr_i(DisasContext *s, uint32_t insn,
1056 unsigned int op1, unsigned int op2, unsigned int crm)
1058 unsupported_encoding(s, insn);
1061 static void gen_get_nzcv(TCGv_i64 tcg_rt)
1063 TCGv_i32 tmp = tcg_temp_new_i32();
1064 TCGv_i32 nzcv = tcg_temp_new_i32();
1066 /* build bit 31, N */
1067 tcg_gen_andi_i32(nzcv, cpu_NF, (1 << 31));
1068 /* build bit 30, Z */
1069 tcg_gen_setcondi_i32(TCG_COND_EQ, tmp, cpu_ZF, 0);
1070 tcg_gen_deposit_i32(nzcv, nzcv, tmp, 30, 1);
1071 /* build bit 29, C */
1072 tcg_gen_deposit_i32(nzcv, nzcv, cpu_CF, 29, 1);
1073 /* build bit 28, V */
1074 tcg_gen_shri_i32(tmp, cpu_VF, 31);
1075 tcg_gen_deposit_i32(nzcv, nzcv, tmp, 28, 1);
1076 /* generate result */
1077 tcg_gen_extu_i32_i64(tcg_rt, nzcv);
1079 tcg_temp_free_i32(nzcv);
1080 tcg_temp_free_i32(tmp);
1083 static void gen_set_nzcv(TCGv_i64 tcg_rt)
1086 TCGv_i32 nzcv = tcg_temp_new_i32();
1088 /* take NZCV from R[t] */
1089 tcg_gen_trunc_i64_i32(nzcv, tcg_rt);
1092 tcg_gen_andi_i32(cpu_NF, nzcv, (1 << 31));
1094 tcg_gen_andi_i32(cpu_ZF, nzcv, (1 << 30));
1095 tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_ZF, cpu_ZF, 0);
1097 tcg_gen_andi_i32(cpu_CF, nzcv, (1 << 29));
1098 tcg_gen_shri_i32(cpu_CF, cpu_CF, 29);
1100 tcg_gen_andi_i32(cpu_VF, nzcv, (1 << 28));
1101 tcg_gen_shli_i32(cpu_VF, cpu_VF, 3);
1102 tcg_temp_free_i32(nzcv);
1105 /* C5.6.129 MRS - move from system register
1106 * C5.6.131 MSR (register) - move to system register
1109 * These are all essentially the same insn in 'read' and 'write'
1110 * versions, with varying op0 fields.
1112 static void handle_sys(DisasContext *s, uint32_t insn, bool isread,
1113 unsigned int op0, unsigned int op1, unsigned int op2,
1114 unsigned int crn, unsigned int crm, unsigned int rt)
1116 const ARMCPRegInfo *ri;
1119 ri = get_arm_cp_reginfo(s->cp_regs,
1120 ENCODE_AA64_CP_REG(CP_REG_ARM64_SYSREG_CP,
1121 crn, crm, op0, op1, op2));
1124 /* Unknown register */
1125 unallocated_encoding(s);
1129 /* Check access permissions */
1130 if (!cp_access_ok(s->current_pl, ri, isread)) {
1131 unallocated_encoding(s);
1135 /* Handle special cases first */
1136 switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) {
1140 tcg_rt = cpu_reg(s, rt);
1142 gen_get_nzcv(tcg_rt);
1144 gen_set_nzcv(tcg_rt);
1151 if (use_icount && (ri->type & ARM_CP_IO)) {
1155 tcg_rt = cpu_reg(s, rt);
1158 if (ri->type & ARM_CP_CONST) {
1159 tcg_gen_movi_i64(tcg_rt, ri->resetvalue);
1160 } else if (ri->readfn) {
1162 gen_a64_set_pc_im(s->pc - 4);
1163 tmpptr = tcg_const_ptr(ri);
1164 gen_helper_get_cp_reg64(tcg_rt, cpu_env, tmpptr);
1165 tcg_temp_free_ptr(tmpptr);
1167 tcg_gen_ld_i64(tcg_rt, cpu_env, ri->fieldoffset);
1170 if (ri->type & ARM_CP_CONST) {
1171 /* If not forbidden by access permissions, treat as WI */
1173 } else if (ri->writefn) {
1175 gen_a64_set_pc_im(s->pc - 4);
1176 tmpptr = tcg_const_ptr(ri);
1177 gen_helper_set_cp_reg64(cpu_env, tmpptr, tcg_rt);
1178 tcg_temp_free_ptr(tmpptr);
1180 tcg_gen_st_i64(tcg_rt, cpu_env, ri->fieldoffset);
1184 if (use_icount && (ri->type & ARM_CP_IO)) {
1185 /* I/O operations must end the TB here (whether read or write) */
1187 s->is_jmp = DISAS_UPDATE;
1188 } else if (!isread && !(ri->type & ARM_CP_SUPPRESS_TB_END)) {
1189 /* We default to ending the TB on a coprocessor register write,
1190 * but allow this to be suppressed by the register definition
1191 * (usually only necessary to work around guest bugs).
1193 s->is_jmp = DISAS_UPDATE;
1198 * 31 22 21 20 19 18 16 15 12 11 8 7 5 4 0
1199 * +---------------------+---+-----+-----+-------+-------+-----+------+
1200 * | 1 1 0 1 0 1 0 1 0 0 | L | op0 | op1 | CRn | CRm | op2 | Rt |
1201 * +---------------------+---+-----+-----+-------+-------+-----+------+
1203 static void disas_system(DisasContext *s, uint32_t insn)
1205 unsigned int l, op0, op1, crn, crm, op2, rt;
1206 l = extract32(insn, 21, 1);
1207 op0 = extract32(insn, 19, 2);
1208 op1 = extract32(insn, 16, 3);
1209 crn = extract32(insn, 12, 4);
1210 crm = extract32(insn, 8, 4);
1211 op2 = extract32(insn, 5, 3);
1212 rt = extract32(insn, 0, 5);
1215 if (l || rt != 31) {
1216 unallocated_encoding(s);
1220 case 2: /* C5.6.68 HINT */
1221 handle_hint(s, insn, op1, op2, crm);
1223 case 3: /* CLREX, DSB, DMB, ISB */
1224 handle_sync(s, insn, op1, op2, crm);
1226 case 4: /* C5.6.130 MSR (immediate) */
1227 handle_msr_i(s, insn, op1, op2, crm);
1230 unallocated_encoding(s);
1235 handle_sys(s, insn, l, op0, op1, op2, crn, crm, rt);
1238 /* C3.2.3 Exception generation
1240 * 31 24 23 21 20 5 4 2 1 0
1241 * +-----------------+-----+------------------------+-----+----+
1242 * | 1 1 0 1 0 1 0 0 | opc | imm16 | op2 | LL |
1243 * +-----------------------+------------------------+----------+
1245 static void disas_exc(DisasContext *s, uint32_t insn)
1247 int opc = extract32(insn, 21, 3);
1248 int op2_ll = extract32(insn, 0, 5);
1252 /* SVC, HVC, SMC; since we don't support the Virtualization
1253 * or TrustZone extensions these all UNDEF except SVC.
1256 unallocated_encoding(s);
1259 gen_exception_insn(s, 0, EXCP_SWI);
1263 unallocated_encoding(s);
1267 gen_exception_insn(s, 0, EXCP_BKPT);
1271 unallocated_encoding(s);
1275 unsupported_encoding(s, insn);
1278 if (op2_ll < 1 || op2_ll > 3) {
1279 unallocated_encoding(s);
1282 /* DCPS1, DCPS2, DCPS3 */
1283 unsupported_encoding(s, insn);
1286 unallocated_encoding(s);
1291 /* C3.2.7 Unconditional branch (register)
1292 * 31 25 24 21 20 16 15 10 9 5 4 0
1293 * +---------------+-------+-------+-------+------+-------+
1294 * | 1 1 0 1 0 1 1 | opc | op2 | op3 | Rn | op4 |
1295 * +---------------+-------+-------+-------+------+-------+
1297 static void disas_uncond_b_reg(DisasContext *s, uint32_t insn)
1299 unsigned int opc, op2, op3, rn, op4;
1301 opc = extract32(insn, 21, 4);
1302 op2 = extract32(insn, 16, 5);
1303 op3 = extract32(insn, 10, 6);
1304 rn = extract32(insn, 5, 5);
1305 op4 = extract32(insn, 0, 5);
1307 if (op4 != 0x0 || op3 != 0x0 || op2 != 0x1f) {
1308 unallocated_encoding(s);
1317 tcg_gen_movi_i64(cpu_reg(s, 30), s->pc);
1322 unallocated_encoding(s);
1324 unsupported_encoding(s, insn);
1328 unallocated_encoding(s);
1332 tcg_gen_mov_i64(cpu_pc, cpu_reg(s, rn));
1333 s->is_jmp = DISAS_JUMP;
1336 /* C3.2 Branches, exception generating and system instructions */
1337 static void disas_b_exc_sys(DisasContext *s, uint32_t insn)
1339 switch (extract32(insn, 25, 7)) {
1340 case 0x0a: case 0x0b:
1341 case 0x4a: case 0x4b: /* Unconditional branch (immediate) */
1342 disas_uncond_b_imm(s, insn);
1344 case 0x1a: case 0x5a: /* Compare & branch (immediate) */
1345 disas_comp_b_imm(s, insn);
1347 case 0x1b: case 0x5b: /* Test & branch (immediate) */
1348 disas_test_b_imm(s, insn);
1350 case 0x2a: /* Conditional branch (immediate) */
1351 disas_cond_b_imm(s, insn);
1353 case 0x6a: /* Exception generation / System */
1354 if (insn & (1 << 24)) {
1355 disas_system(s, insn);
1360 case 0x6b: /* Unconditional branch (register) */
1361 disas_uncond_b_reg(s, insn);
1364 unallocated_encoding(s);
1370 * Load/Store exclusive instructions are implemented by remembering
1371 * the value/address loaded, and seeing if these are the same
1372 * when the store is performed. This is not actually the architecturally
1373 * mandated semantics, but it works for typical guest code sequences
1374 * and avoids having to monitor regular stores.
1376 * In system emulation mode only one CPU will be running at once, so
1377 * this sequence is effectively atomic. In user emulation mode we
1378 * throw an exception and handle the atomic operation elsewhere.
1380 static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
1381 TCGv_i64 addr, int size, bool is_pair)
1383 TCGv_i64 tmp = tcg_temp_new_i64();
1384 TCGMemOp memop = MO_TE + size;
1386 g_assert(size <= 3);
1387 tcg_gen_qemu_ld_i64(tmp, addr, get_mem_index(s), memop);
1390 TCGv_i64 addr2 = tcg_temp_new_i64();
1391 TCGv_i64 hitmp = tcg_temp_new_i64();
1393 g_assert(size >= 2);
1394 tcg_gen_addi_i64(addr2, addr, 1 << size);
1395 tcg_gen_qemu_ld_i64(hitmp, addr2, get_mem_index(s), memop);
1396 tcg_temp_free_i64(addr2);
1397 tcg_gen_mov_i64(cpu_exclusive_high, hitmp);
1398 tcg_gen_mov_i64(cpu_reg(s, rt2), hitmp);
1399 tcg_temp_free_i64(hitmp);
1402 tcg_gen_mov_i64(cpu_exclusive_val, tmp);
1403 tcg_gen_mov_i64(cpu_reg(s, rt), tmp);
1405 tcg_temp_free_i64(tmp);
1406 tcg_gen_mov_i64(cpu_exclusive_addr, addr);
1409 #ifdef CONFIG_USER_ONLY
1410 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
1411 TCGv_i64 addr, int size, int is_pair)
1413 tcg_gen_mov_i64(cpu_exclusive_test, addr);
1414 tcg_gen_movi_i32(cpu_exclusive_info,
1415 size | is_pair << 2 | (rd << 4) | (rt << 9) | (rt2 << 14));
1416 gen_exception_insn(s, 4, EXCP_STREX);
1419 static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
1420 TCGv_i64 addr, int size, int is_pair)
1422 qemu_log_mask(LOG_UNIMP,
1423 "%s:%d: system mode store_exclusive unsupported "
1424 "at pc=%016" PRIx64 "\n",
1425 __FILE__, __LINE__, s->pc - 4);
1429 /* C3.3.6 Load/store exclusive
1431 * 31 30 29 24 23 22 21 20 16 15 14 10 9 5 4 0
1432 * +-----+-------------+----+---+----+------+----+-------+------+------+
1433 * | sz | 0 0 1 0 0 0 | o2 | L | o1 | Rs | o0 | Rt2 | Rn | Rt |
1434 * +-----+-------------+----+---+----+------+----+-------+------+------+
1436 * sz: 00 -> 8 bit, 01 -> 16 bit, 10 -> 32 bit, 11 -> 64 bit
1437 * L: 0 -> store, 1 -> load
1438 * o2: 0 -> exclusive, 1 -> not
1439 * o1: 0 -> single register, 1 -> register pair
1440 * o0: 1 -> load-acquire/store-release, 0 -> not
1442 * o0 == 0 AND o2 == 1 is un-allocated
1443 * o1 == 1 is un-allocated except for 32 and 64 bit sizes
1445 static void disas_ldst_excl(DisasContext *s, uint32_t insn)
1447 int rt = extract32(insn, 0, 5);
1448 int rn = extract32(insn, 5, 5);
1449 int rt2 = extract32(insn, 10, 5);
1450 int is_lasr = extract32(insn, 15, 1);
1451 int rs = extract32(insn, 16, 5);
1452 int is_pair = extract32(insn, 21, 1);
1453 int is_store = !extract32(insn, 22, 1);
1454 int is_excl = !extract32(insn, 23, 1);
1455 int size = extract32(insn, 30, 2);
1458 if ((!is_excl && !is_lasr) ||
1459 (is_pair && size < 2)) {
1460 unallocated_encoding(s);
1465 gen_check_sp_alignment(s);
1467 tcg_addr = read_cpu_reg_sp(s, rn, 1);
1469 /* Note that since TCG is single threaded load-acquire/store-release
1470 * semantics require no extra if (is_lasr) { ... } handling.
1475 gen_load_exclusive(s, rt, rt2, tcg_addr, size, is_pair);
1477 gen_store_exclusive(s, rs, rt, rt2, tcg_addr, size, is_pair);
1480 TCGv_i64 tcg_rt = cpu_reg(s, rt);
1482 do_gpr_st(s, tcg_rt, tcg_addr, size);
1484 do_gpr_ld(s, tcg_rt, tcg_addr, size, false, false);
1487 TCGv_i64 tcg_rt2 = cpu_reg(s, rt);
1488 tcg_gen_addi_i64(tcg_addr, tcg_addr, 1 << size);
1490 do_gpr_st(s, tcg_rt2, tcg_addr, size);
1492 do_gpr_ld(s, tcg_rt2, tcg_addr, size, false, false);
1499 * C3.3.5 Load register (literal)
1501 * 31 30 29 27 26 25 24 23 5 4 0
1502 * +-----+-------+---+-----+-------------------+-------+
1503 * | opc | 0 1 1 | V | 0 0 | imm19 | Rt |
1504 * +-----+-------+---+-----+-------------------+-------+
1506 * V: 1 -> vector (simd/fp)
1507 * opc (non-vector): 00 -> 32 bit, 01 -> 64 bit,
1508 * 10-> 32 bit signed, 11 -> prefetch
1509 * opc (vector): 00 -> 32 bit, 01 -> 64 bit, 10 -> 128 bit (11 unallocated)
1511 static void disas_ld_lit(DisasContext *s, uint32_t insn)
1513 int rt = extract32(insn, 0, 5);
1514 int64_t imm = sextract32(insn, 5, 19) << 2;
1515 bool is_vector = extract32(insn, 26, 1);
1516 int opc = extract32(insn, 30, 2);
1517 bool is_signed = false;
1519 TCGv_i64 tcg_rt, tcg_addr;
1523 unallocated_encoding(s);
1529 /* PRFM (literal) : prefetch */
1532 size = 2 + extract32(opc, 0, 1);
1533 is_signed = extract32(opc, 1, 1);
1536 tcg_rt = cpu_reg(s, rt);
1538 tcg_addr = tcg_const_i64((s->pc - 4) + imm);
1540 do_fp_ld(s, rt, tcg_addr, size);
1542 do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, false);
1544 tcg_temp_free_i64(tcg_addr);
1548 * C5.6.80 LDNP (Load Pair - non-temporal hint)
1549 * C5.6.81 LDP (Load Pair - non vector)
1550 * C5.6.82 LDPSW (Load Pair Signed Word - non vector)
1551 * C5.6.176 STNP (Store Pair - non-temporal hint)
1552 * C5.6.177 STP (Store Pair - non vector)
1553 * C6.3.165 LDNP (Load Pair of SIMD&FP - non-temporal hint)
1554 * C6.3.165 LDP (Load Pair of SIMD&FP)
1555 * C6.3.284 STNP (Store Pair of SIMD&FP - non-temporal hint)
1556 * C6.3.284 STP (Store Pair of SIMD&FP)
1558 * 31 30 29 27 26 25 24 23 22 21 15 14 10 9 5 4 0
1559 * +-----+-------+---+---+-------+---+-----------------------------+
1560 * | opc | 1 0 1 | V | 0 | index | L | imm7 | Rt2 | Rn | Rt |
1561 * +-----+-------+---+---+-------+---+-------+-------+------+------+
1563 * opc: LDP/STP/LDNP/STNP 00 -> 32 bit, 10 -> 64 bit
1565 * LDP/STP/LDNP/STNP (SIMD) 00 -> 32 bit, 01 -> 64 bit, 10 -> 128 bit
1566 * V: 0 -> GPR, 1 -> Vector
1567 * idx: 00 -> signed offset with non-temporal hint, 01 -> post-index,
1568 * 10 -> signed offset, 11 -> pre-index
1569 * L: 0 -> Store 1 -> Load
1571 * Rt, Rt2 = GPR or SIMD registers to be stored
1572 * Rn = general purpose register containing address
1573 * imm7 = signed offset (multiple of 4 or 8 depending on size)
1575 static void disas_ldst_pair(DisasContext *s, uint32_t insn)
1577 int rt = extract32(insn, 0, 5);
1578 int rn = extract32(insn, 5, 5);
1579 int rt2 = extract32(insn, 10, 5);
1580 int64_t offset = sextract32(insn, 15, 7);
1581 int index = extract32(insn, 23, 2);
1582 bool is_vector = extract32(insn, 26, 1);
1583 bool is_load = extract32(insn, 22, 1);
1584 int opc = extract32(insn, 30, 2);
1586 bool is_signed = false;
1587 bool postindex = false;
1590 TCGv_i64 tcg_addr; /* calculated address */
1594 unallocated_encoding(s);
1601 size = 2 + extract32(opc, 1, 1);
1602 is_signed = extract32(opc, 0, 1);
1603 if (!is_load && is_signed) {
1604 unallocated_encoding(s);
1610 case 1: /* post-index */
1615 /* signed offset with "non-temporal" hint. Since we don't emulate
1616 * caches we don't care about hints to the cache system about
1617 * data access patterns, and handle this identically to plain
1621 /* There is no non-temporal-hint version of LDPSW */
1622 unallocated_encoding(s);
1627 case 2: /* signed offset, rn not updated */
1630 case 3: /* pre-index */
1639 gen_check_sp_alignment(s);
1642 tcg_addr = read_cpu_reg_sp(s, rn, 1);
1645 tcg_gen_addi_i64(tcg_addr, tcg_addr, offset);
1650 do_fp_ld(s, rt, tcg_addr, size);
1652 do_fp_st(s, rt, tcg_addr, size);
1655 TCGv_i64 tcg_rt = cpu_reg(s, rt);
1657 do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, false);
1659 do_gpr_st(s, tcg_rt, tcg_addr, size);
1662 tcg_gen_addi_i64(tcg_addr, tcg_addr, 1 << size);
1665 do_fp_ld(s, rt2, tcg_addr, size);
1667 do_fp_st(s, rt2, tcg_addr, size);
1670 TCGv_i64 tcg_rt2 = cpu_reg(s, rt2);
1672 do_gpr_ld(s, tcg_rt2, tcg_addr, size, is_signed, false);
1674 do_gpr_st(s, tcg_rt2, tcg_addr, size);
1680 tcg_gen_addi_i64(tcg_addr, tcg_addr, offset - (1 << size));
1682 tcg_gen_subi_i64(tcg_addr, tcg_addr, 1 << size);
1684 tcg_gen_mov_i64(cpu_reg_sp(s, rn), tcg_addr);
1689 * C3.3.8 Load/store (immediate post-indexed)
1690 * C3.3.9 Load/store (immediate pre-indexed)
1691 * C3.3.12 Load/store (unscaled immediate)
1693 * 31 30 29 27 26 25 24 23 22 21 20 12 11 10 9 5 4 0
1694 * +----+-------+---+-----+-----+---+--------+-----+------+------+
1695 * |size| 1 1 1 | V | 0 0 | opc | 0 | imm9 | idx | Rn | Rt |
1696 * +----+-------+---+-----+-----+---+--------+-----+------+------+
1698 * idx = 01 -> post-indexed, 11 pre-indexed, 00 unscaled imm. (no writeback)
1699 * V = 0 -> non-vector
1700 * size: 00 -> 8 bit, 01 -> 16 bit, 10 -> 32 bit, 11 -> 64bit
1701 * opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
1703 static void disas_ldst_reg_imm9(DisasContext *s, uint32_t insn)
1705 int rt = extract32(insn, 0, 5);
1706 int rn = extract32(insn, 5, 5);
1707 int imm9 = sextract32(insn, 12, 9);
1708 int opc = extract32(insn, 22, 2);
1709 int size = extract32(insn, 30, 2);
1710 int idx = extract32(insn, 10, 2);
1711 bool is_signed = false;
1712 bool is_store = false;
1713 bool is_extended = false;
1714 bool is_vector = extract32(insn, 26, 1);
1721 size |= (opc & 2) << 1;
1723 unallocated_encoding(s);
1726 is_store = ((opc & 1) == 0);
1728 if (size == 3 && opc == 2) {
1729 /* PRFM - prefetch */
1732 if (opc == 3 && size > 1) {
1733 unallocated_encoding(s);
1736 is_store = (opc == 0);
1737 is_signed = opc & (1<<1);
1738 is_extended = (size < 3) && (opc & 1);
1760 gen_check_sp_alignment(s);
1762 tcg_addr = read_cpu_reg_sp(s, rn, 1);
1765 tcg_gen_addi_i64(tcg_addr, tcg_addr, imm9);
1770 do_fp_st(s, rt, tcg_addr, size);
1772 do_fp_ld(s, rt, tcg_addr, size);
1775 TCGv_i64 tcg_rt = cpu_reg(s, rt);
1777 do_gpr_st(s, tcg_rt, tcg_addr, size);
1779 do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended);
1784 TCGv_i64 tcg_rn = cpu_reg_sp(s, rn);
1786 tcg_gen_addi_i64(tcg_addr, tcg_addr, imm9);
1788 tcg_gen_mov_i64(tcg_rn, tcg_addr);
1793 * C3.3.10 Load/store (register offset)
1795 * 31 30 29 27 26 25 24 23 22 21 20 16 15 13 12 11 10 9 5 4 0
1796 * +----+-------+---+-----+-----+---+------+-----+--+-----+----+----+
1797 * |size| 1 1 1 | V | 0 0 | opc | 1 | Rm | opt | S| 1 0 | Rn | Rt |
1798 * +----+-------+---+-----+-----+---+------+-----+--+-----+----+----+
1801 * size: 00-> byte, 01 -> 16 bit, 10 -> 32bit, 11 -> 64bit
1802 * opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
1804 * size is opc<1>:size<1:0> so 100 -> 128 bit; 110 and 111 unallocated
1805 * opc<0>: 0 -> store, 1 -> load
1806 * V: 1 -> vector/simd
1807 * opt: extend encoding (see DecodeRegExtend)
1808 * S: if S=1 then scale (essentially index by sizeof(size))
1809 * Rt: register to transfer into/out of
1810 * Rn: address register or SP for base
1811 * Rm: offset register or ZR for offset
1813 static void disas_ldst_reg_roffset(DisasContext *s, uint32_t insn)
1815 int rt = extract32(insn, 0, 5);
1816 int rn = extract32(insn, 5, 5);
1817 int shift = extract32(insn, 12, 1);
1818 int rm = extract32(insn, 16, 5);
1819 int opc = extract32(insn, 22, 2);
1820 int opt = extract32(insn, 13, 3);
1821 int size = extract32(insn, 30, 2);
1822 bool is_signed = false;
1823 bool is_store = false;
1824 bool is_extended = false;
1825 bool is_vector = extract32(insn, 26, 1);
1830 if (extract32(opt, 1, 1) == 0) {
1831 unallocated_encoding(s);
1836 size |= (opc & 2) << 1;
1838 unallocated_encoding(s);
1841 is_store = !extract32(opc, 0, 1);
1843 if (size == 3 && opc == 2) {
1844 /* PRFM - prefetch */
1847 if (opc == 3 && size > 1) {
1848 unallocated_encoding(s);
1851 is_store = (opc == 0);
1852 is_signed = extract32(opc, 1, 1);
1853 is_extended = (size < 3) && extract32(opc, 0, 1);
1857 gen_check_sp_alignment(s);
1859 tcg_addr = read_cpu_reg_sp(s, rn, 1);
1861 tcg_rm = read_cpu_reg(s, rm, 1);
1862 ext_and_shift_reg(tcg_rm, tcg_rm, opt, shift ? size : 0);
1864 tcg_gen_add_i64(tcg_addr, tcg_addr, tcg_rm);
1868 do_fp_st(s, rt, tcg_addr, size);
1870 do_fp_ld(s, rt, tcg_addr, size);
1873 TCGv_i64 tcg_rt = cpu_reg(s, rt);
1875 do_gpr_st(s, tcg_rt, tcg_addr, size);
1877 do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended);
1883 * C3.3.13 Load/store (unsigned immediate)
1885 * 31 30 29 27 26 25 24 23 22 21 10 9 5
1886 * +----+-------+---+-----+-----+------------+-------+------+
1887 * |size| 1 1 1 | V | 0 1 | opc | imm12 | Rn | Rt |
1888 * +----+-------+---+-----+-----+------------+-------+------+
1891 * size: 00-> byte, 01 -> 16 bit, 10 -> 32bit, 11 -> 64bit
1892 * opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
1894 * size is opc<1>:size<1:0> so 100 -> 128 bit; 110 and 111 unallocated
1895 * opc<0>: 0 -> store, 1 -> load
1896 * Rn: base address register (inc SP)
1897 * Rt: target register
1899 static void disas_ldst_reg_unsigned_imm(DisasContext *s, uint32_t insn)
1901 int rt = extract32(insn, 0, 5);
1902 int rn = extract32(insn, 5, 5);
1903 unsigned int imm12 = extract32(insn, 10, 12);
1904 bool is_vector = extract32(insn, 26, 1);
1905 int size = extract32(insn, 30, 2);
1906 int opc = extract32(insn, 22, 2);
1907 unsigned int offset;
1912 bool is_signed = false;
1913 bool is_extended = false;
1916 size |= (opc & 2) << 1;
1918 unallocated_encoding(s);
1921 is_store = !extract32(opc, 0, 1);
1923 if (size == 3 && opc == 2) {
1924 /* PRFM - prefetch */
1927 if (opc == 3 && size > 1) {
1928 unallocated_encoding(s);
1931 is_store = (opc == 0);
1932 is_signed = extract32(opc, 1, 1);
1933 is_extended = (size < 3) && extract32(opc, 0, 1);
1937 gen_check_sp_alignment(s);
1939 tcg_addr = read_cpu_reg_sp(s, rn, 1);
1940 offset = imm12 << size;
1941 tcg_gen_addi_i64(tcg_addr, tcg_addr, offset);
1945 do_fp_st(s, rt, tcg_addr, size);
1947 do_fp_ld(s, rt, tcg_addr, size);
1950 TCGv_i64 tcg_rt = cpu_reg(s, rt);
1952 do_gpr_st(s, tcg_rt, tcg_addr, size);
1954 do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended);
1959 /* Load/store register (immediate forms) */
1960 static void disas_ldst_reg_imm(DisasContext *s, uint32_t insn)
1962 switch (extract32(insn, 10, 2)) {
1963 case 0: case 1: case 3:
1964 /* Load/store register (unscaled immediate) */
1965 /* Load/store immediate pre/post-indexed */
1966 disas_ldst_reg_imm9(s, insn);
1969 /* Load/store register unprivileged */
1970 unsupported_encoding(s, insn);
1973 unallocated_encoding(s);
1978 /* Load/store register (all forms) */
1979 static void disas_ldst_reg(DisasContext *s, uint32_t insn)
1981 switch (extract32(insn, 24, 2)) {
1983 if (extract32(insn, 21, 1) == 1 && extract32(insn, 10, 2) == 2) {
1984 disas_ldst_reg_roffset(s, insn);
1986 disas_ldst_reg_imm(s, insn);
1990 disas_ldst_reg_unsigned_imm(s, insn);
1993 unallocated_encoding(s);
1998 /* C3.3.1 AdvSIMD load/store multiple structures
2000 * 31 30 29 23 22 21 16 15 12 11 10 9 5 4 0
2001 * +---+---+---------------+---+-------------+--------+------+------+------+
2002 * | 0 | Q | 0 0 1 1 0 0 0 | L | 0 0 0 0 0 0 | opcode | size | Rn | Rt |
2003 * +---+---+---------------+---+-------------+--------+------+------+------+
2005 * C3.3.2 AdvSIMD load/store multiple structures (post-indexed)
2007 * 31 30 29 23 22 21 20 16 15 12 11 10 9 5 4 0
2008 * +---+---+---------------+---+---+---------+--------+------+------+------+
2009 * | 0 | Q | 0 0 1 1 0 0 1 | L | 0 | Rm | opcode | size | Rn | Rt |
2010 * +---+---+---------------+---+---+---------+--------+------+------+------+
2012 * Rt: first (or only) SIMD&FP register to be transferred
2013 * Rn: base address or SP
2014 * Rm (post-index only): post-index register (when !31) or size dependent #imm
2016 static void disas_ldst_multiple_struct(DisasContext *s, uint32_t insn)
2018 int rt = extract32(insn, 0, 5);
2019 int rn = extract32(insn, 5, 5);
2020 int size = extract32(insn, 10, 2);
2021 int opcode = extract32(insn, 12, 4);
2022 bool is_store = !extract32(insn, 22, 1);
2023 bool is_postidx = extract32(insn, 23, 1);
2024 bool is_q = extract32(insn, 30, 1);
2025 TCGv_i64 tcg_addr, tcg_rn;
2027 int ebytes = 1 << size;
2028 int elements = (is_q ? 128 : 64) / (8 << size);
2029 int rpt; /* num iterations */
2030 int selem; /* structure elements */
2033 if (extract32(insn, 31, 1) || extract32(insn, 21, 1)) {
2034 unallocated_encoding(s);
2038 /* From the shared decode logic */
2069 unallocated_encoding(s);
2073 if (size == 3 && !is_q && selem != 1) {
2075 unallocated_encoding(s);
2080 gen_check_sp_alignment(s);
2083 tcg_rn = cpu_reg_sp(s, rn);
2084 tcg_addr = tcg_temp_new_i64();
2085 tcg_gen_mov_i64(tcg_addr, tcg_rn);
2087 for (r = 0; r < rpt; r++) {
2089 for (e = 0; e < elements; e++) {
2090 int tt = (rt + r) % 32;
2092 for (xs = 0; xs < selem; xs++) {
2094 do_vec_st(s, tt, e, tcg_addr, size);
2096 do_vec_ld(s, tt, e, tcg_addr, size);
2098 /* For non-quad operations, setting a slice of the low
2099 * 64 bits of the register clears the high 64 bits (in
2100 * the ARM ARM pseudocode this is implicit in the fact
2101 * that 'rval' is a 64 bit wide variable). We optimize
2102 * by noticing that we only need to do this the first
2103 * time we touch a register.
2105 if (!is_q && e == 0 && (r == 0 || xs == selem - 1)) {
2106 clear_vec_high(s, tt);
2109 tcg_gen_addi_i64(tcg_addr, tcg_addr, ebytes);
2116 int rm = extract32(insn, 16, 5);
2118 tcg_gen_mov_i64(tcg_rn, tcg_addr);
2120 tcg_gen_add_i64(tcg_rn, tcg_rn, cpu_reg(s, rm));
2123 tcg_temp_free_i64(tcg_addr);
2126 /* C3.3.3 AdvSIMD load/store single structure
2128 * 31 30 29 23 22 21 20 16 15 13 12 11 10 9 5 4 0
2129 * +---+---+---------------+-----+-----------+-----+---+------+------+------+
2130 * | 0 | Q | 0 0 1 1 0 1 0 | L R | 0 0 0 0 0 | opc | S | size | Rn | Rt |
2131 * +---+---+---------------+-----+-----------+-----+---+------+------+------+
2133 * C3.3.4 AdvSIMD load/store single structure (post-indexed)
2135 * 31 30 29 23 22 21 20 16 15 13 12 11 10 9 5 4 0
2136 * +---+---+---------------+-----+-----------+-----+---+------+------+------+
2137 * | 0 | Q | 0 0 1 1 0 1 1 | L R | Rm | opc | S | size | Rn | Rt |
2138 * +---+---+---------------+-----+-----------+-----+---+------+------+------+
2140 * Rt: first (or only) SIMD&FP register to be transferred
2141 * Rn: base address or SP
2142 * Rm (post-index only): post-index register (when !31) or size dependent #imm
2143 * index = encoded in Q:S:size dependent on size
2145 * lane_size = encoded in R, opc
2146 * transfer width = encoded in opc, S, size
2148 static void disas_ldst_single_struct(DisasContext *s, uint32_t insn)
2150 int rt = extract32(insn, 0, 5);
2151 int rn = extract32(insn, 5, 5);
2152 int size = extract32(insn, 10, 2);
2153 int S = extract32(insn, 12, 1);
2154 int opc = extract32(insn, 13, 3);
2155 int R = extract32(insn, 21, 1);
2156 int is_load = extract32(insn, 22, 1);
2157 int is_postidx = extract32(insn, 23, 1);
2158 int is_q = extract32(insn, 30, 1);
2160 int scale = extract32(opc, 1, 2);
2161 int selem = (extract32(opc, 0, 1) << 1 | R) + 1;
2162 bool replicate = false;
2163 int index = is_q << 3 | S << 2 | size;
2165 TCGv_i64 tcg_addr, tcg_rn;
2169 if (!is_load || S) {
2170 unallocated_encoding(s);
2179 if (extract32(size, 0, 1)) {
2180 unallocated_encoding(s);
2186 if (extract32(size, 1, 1)) {
2187 unallocated_encoding(s);
2190 if (!extract32(size, 0, 1)) {
2194 unallocated_encoding(s);
2202 g_assert_not_reached();
2205 ebytes = 1 << scale;
2208 gen_check_sp_alignment(s);
2211 tcg_rn = cpu_reg_sp(s, rn);
2212 tcg_addr = tcg_temp_new_i64();
2213 tcg_gen_mov_i64(tcg_addr, tcg_rn);
2215 for (xs = 0; xs < selem; xs++) {
2217 /* Load and replicate to all elements */
2219 TCGv_i64 tcg_tmp = tcg_temp_new_i64();
2221 tcg_gen_qemu_ld_i64(tcg_tmp, tcg_addr,
2222 get_mem_index(s), MO_TE + scale);
2225 mulconst = 0x0101010101010101ULL;
2228 mulconst = 0x0001000100010001ULL;
2231 mulconst = 0x0000000100000001ULL;
2237 g_assert_not_reached();
2240 tcg_gen_muli_i64(tcg_tmp, tcg_tmp, mulconst);
2242 write_vec_element(s, tcg_tmp, rt, 0, MO_64);
2244 write_vec_element(s, tcg_tmp, rt, 1, MO_64);
2246 clear_vec_high(s, rt);
2248 tcg_temp_free_i64(tcg_tmp);
2250 /* Load/store one element per register */
2252 do_vec_ld(s, rt, index, tcg_addr, MO_TE + scale);
2254 do_vec_st(s, rt, index, tcg_addr, MO_TE + scale);
2257 tcg_gen_addi_i64(tcg_addr, tcg_addr, ebytes);
2262 int rm = extract32(insn, 16, 5);
2264 tcg_gen_mov_i64(tcg_rn, tcg_addr);
2266 tcg_gen_add_i64(tcg_rn, tcg_rn, cpu_reg(s, rm));
2269 tcg_temp_free_i64(tcg_addr);
2272 /* C3.3 Loads and stores */
2273 static void disas_ldst(DisasContext *s, uint32_t insn)
2275 switch (extract32(insn, 24, 6)) {
2276 case 0x08: /* Load/store exclusive */
2277 disas_ldst_excl(s, insn);
2279 case 0x18: case 0x1c: /* Load register (literal) */
2280 disas_ld_lit(s, insn);
2282 case 0x28: case 0x29:
2283 case 0x2c: case 0x2d: /* Load/store pair (all forms) */
2284 disas_ldst_pair(s, insn);
2286 case 0x38: case 0x39:
2287 case 0x3c: case 0x3d: /* Load/store register (all forms) */
2288 disas_ldst_reg(s, insn);
2290 case 0x0c: /* AdvSIMD load/store multiple structures */
2291 disas_ldst_multiple_struct(s, insn);
2293 case 0x0d: /* AdvSIMD load/store single structure */
2294 disas_ldst_single_struct(s, insn);
2297 unallocated_encoding(s);
2302 /* C3.4.6 PC-rel. addressing
2303 * 31 30 29 28 24 23 5 4 0
2304 * +----+-------+-----------+-------------------+------+
2305 * | op | immlo | 1 0 0 0 0 | immhi | Rd |
2306 * +----+-------+-----------+-------------------+------+
2308 static void disas_pc_rel_adr(DisasContext *s, uint32_t insn)
2310 unsigned int page, rd;
2314 page = extract32(insn, 31, 1);
2315 /* SignExtend(immhi:immlo) -> offset */
2316 offset = ((int64_t)sextract32(insn, 5, 19) << 2) | extract32(insn, 29, 2);
2317 rd = extract32(insn, 0, 5);
2321 /* ADRP (page based) */
2326 tcg_gen_movi_i64(cpu_reg(s, rd), base + offset);
2330 * C3.4.1 Add/subtract (immediate)
2332 * 31 30 29 28 24 23 22 21 10 9 5 4 0
2333 * +--+--+--+-----------+-----+-------------+-----+-----+
2334 * |sf|op| S| 1 0 0 0 1 |shift| imm12 | Rn | Rd |
2335 * +--+--+--+-----------+-----+-------------+-----+-----+
2337 * sf: 0 -> 32bit, 1 -> 64bit
2338 * op: 0 -> add , 1 -> sub
2340 * shift: 00 -> LSL imm by 0, 01 -> LSL imm by 12
2342 static void disas_add_sub_imm(DisasContext *s, uint32_t insn)
2344 int rd = extract32(insn, 0, 5);
2345 int rn = extract32(insn, 5, 5);
2346 uint64_t imm = extract32(insn, 10, 12);
2347 int shift = extract32(insn, 22, 2);
2348 bool setflags = extract32(insn, 29, 1);
2349 bool sub_op = extract32(insn, 30, 1);
2350 bool is_64bit = extract32(insn, 31, 1);
2352 TCGv_i64 tcg_rn = cpu_reg_sp(s, rn);
2353 TCGv_i64 tcg_rd = setflags ? cpu_reg(s, rd) : cpu_reg_sp(s, rd);
2354 TCGv_i64 tcg_result;
2363 unallocated_encoding(s);
2367 tcg_result = tcg_temp_new_i64();
2370 tcg_gen_subi_i64(tcg_result, tcg_rn, imm);
2372 tcg_gen_addi_i64(tcg_result, tcg_rn, imm);
2375 TCGv_i64 tcg_imm = tcg_const_i64(imm);
2377 gen_sub_CC(is_64bit, tcg_result, tcg_rn, tcg_imm);
2379 gen_add_CC(is_64bit, tcg_result, tcg_rn, tcg_imm);
2381 tcg_temp_free_i64(tcg_imm);
2385 tcg_gen_mov_i64(tcg_rd, tcg_result);
2387 tcg_gen_ext32u_i64(tcg_rd, tcg_result);
2390 tcg_temp_free_i64(tcg_result);
2393 /* The input should be a value in the bottom e bits (with higher
2394 * bits zero); returns that value replicated into every element
2395 * of size e in a 64 bit integer.
2397 static uint64_t bitfield_replicate(uint64_t mask, unsigned int e)
2407 /* Return a value with the bottom len bits set (where 0 < len <= 64) */
2408 static inline uint64_t bitmask64(unsigned int length)
2410 assert(length > 0 && length <= 64);
2411 return ~0ULL >> (64 - length);
2414 /* Simplified variant of pseudocode DecodeBitMasks() for the case where we
2415 * only require the wmask. Returns false if the imms/immr/immn are a reserved
2416 * value (ie should cause a guest UNDEF exception), and true if they are
2417 * valid, in which case the decoded bit pattern is written to result.
2419 static bool logic_imm_decode_wmask(uint64_t *result, unsigned int immn,
2420 unsigned int imms, unsigned int immr)
2423 unsigned e, levels, s, r;
2426 assert(immn < 2 && imms < 64 && immr < 64);
2428 /* The bit patterns we create here are 64 bit patterns which
2429 * are vectors of identical elements of size e = 2, 4, 8, 16, 32 or
2430 * 64 bits each. Each element contains the same value: a run
2431 * of between 1 and e-1 non-zero bits, rotated within the
2432 * element by between 0 and e-1 bits.
2434 * The element size and run length are encoded into immn (1 bit)
2435 * and imms (6 bits) as follows:
2436 * 64 bit elements: immn = 1, imms = <length of run - 1>
2437 * 32 bit elements: immn = 0, imms = 0 : <length of run - 1>
2438 * 16 bit elements: immn = 0, imms = 10 : <length of run - 1>
2439 * 8 bit elements: immn = 0, imms = 110 : <length of run - 1>
2440 * 4 bit elements: immn = 0, imms = 1110 : <length of run - 1>
2441 * 2 bit elements: immn = 0, imms = 11110 : <length of run - 1>
2442 * Notice that immn = 0, imms = 11111x is the only combination
2443 * not covered by one of the above options; this is reserved.
2444 * Further, <length of run - 1> all-ones is a reserved pattern.
2446 * In all cases the rotation is by immr % e (and immr is 6 bits).
2449 /* First determine the element size */
2450 len = 31 - clz32((immn << 6) | (~imms & 0x3f));
2452 /* This is the immn == 0, imms == 0x11111x case */
2462 /* <length of run - 1> mustn't be all-ones. */
2466 /* Create the value of one element: s+1 set bits rotated
2467 * by r within the element (which is e bits wide)...
2469 mask = bitmask64(s + 1);
2470 mask = (mask >> r) | (mask << (e - r));
2471 /* ...then replicate the element over the whole 64 bit value */
2472 mask = bitfield_replicate(mask, e);
2477 /* C3.4.4 Logical (immediate)
2478 * 31 30 29 28 23 22 21 16 15 10 9 5 4 0
2479 * +----+-----+-------------+---+------+------+------+------+
2480 * | sf | opc | 1 0 0 1 0 0 | N | immr | imms | Rn | Rd |
2481 * +----+-----+-------------+---+------+------+------+------+
2483 static void disas_logic_imm(DisasContext *s, uint32_t insn)
2485 unsigned int sf, opc, is_n, immr, imms, rn, rd;
2486 TCGv_i64 tcg_rd, tcg_rn;
2488 bool is_and = false;
2490 sf = extract32(insn, 31, 1);
2491 opc = extract32(insn, 29, 2);
2492 is_n = extract32(insn, 22, 1);
2493 immr = extract32(insn, 16, 6);
2494 imms = extract32(insn, 10, 6);
2495 rn = extract32(insn, 5, 5);
2496 rd = extract32(insn, 0, 5);
2499 unallocated_encoding(s);
2503 if (opc == 0x3) { /* ANDS */
2504 tcg_rd = cpu_reg(s, rd);
2506 tcg_rd = cpu_reg_sp(s, rd);
2508 tcg_rn = cpu_reg(s, rn);
2510 if (!logic_imm_decode_wmask(&wmask, is_n, imms, immr)) {
2511 /* some immediate field values are reserved */
2512 unallocated_encoding(s);
2517 wmask &= 0xffffffff;
2521 case 0x3: /* ANDS */
2523 tcg_gen_andi_i64(tcg_rd, tcg_rn, wmask);
2527 tcg_gen_ori_i64(tcg_rd, tcg_rn, wmask);
2530 tcg_gen_xori_i64(tcg_rd, tcg_rn, wmask);
2533 assert(FALSE); /* must handle all above */
2537 if (!sf && !is_and) {
2538 /* zero extend final result; we know we can skip this for AND
2539 * since the immediate had the high 32 bits clear.
2541 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2544 if (opc == 3) { /* ANDS */
2545 gen_logic_CC(sf, tcg_rd);
2550 * C3.4.5 Move wide (immediate)
2552 * 31 30 29 28 23 22 21 20 5 4 0
2553 * +--+-----+-------------+-----+----------------+------+
2554 * |sf| opc | 1 0 0 1 0 1 | hw | imm16 | Rd |
2555 * +--+-----+-------------+-----+----------------+------+
2557 * sf: 0 -> 32 bit, 1 -> 64 bit
2558 * opc: 00 -> N, 10 -> Z, 11 -> K
2559 * hw: shift/16 (0,16, and sf only 32, 48)
2561 static void disas_movw_imm(DisasContext *s, uint32_t insn)
2563 int rd = extract32(insn, 0, 5);
2564 uint64_t imm = extract32(insn, 5, 16);
2565 int sf = extract32(insn, 31, 1);
2566 int opc = extract32(insn, 29, 2);
2567 int pos = extract32(insn, 21, 2) << 4;
2568 TCGv_i64 tcg_rd = cpu_reg(s, rd);
2571 if (!sf && (pos >= 32)) {
2572 unallocated_encoding(s);
2586 tcg_gen_movi_i64(tcg_rd, imm);
2589 tcg_imm = tcg_const_i64(imm);
2590 tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_imm, pos, 16);
2591 tcg_temp_free_i64(tcg_imm);
2593 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2597 unallocated_encoding(s);
2603 * 31 30 29 28 23 22 21 16 15 10 9 5 4 0
2604 * +----+-----+-------------+---+------+------+------+------+
2605 * | sf | opc | 1 0 0 1 1 0 | N | immr | imms | Rn | Rd |
2606 * +----+-----+-------------+---+------+------+------+------+
2608 static void disas_bitfield(DisasContext *s, uint32_t insn)
2610 unsigned int sf, n, opc, ri, si, rn, rd, bitsize, pos, len;
2611 TCGv_i64 tcg_rd, tcg_tmp;
2613 sf = extract32(insn, 31, 1);
2614 opc = extract32(insn, 29, 2);
2615 n = extract32(insn, 22, 1);
2616 ri = extract32(insn, 16, 6);
2617 si = extract32(insn, 10, 6);
2618 rn = extract32(insn, 5, 5);
2619 rd = extract32(insn, 0, 5);
2620 bitsize = sf ? 64 : 32;
2622 if (sf != n || ri >= bitsize || si >= bitsize || opc > 2) {
2623 unallocated_encoding(s);
2627 tcg_rd = cpu_reg(s, rd);
2628 tcg_tmp = read_cpu_reg(s, rn, sf);
2630 /* OPTME: probably worth recognizing common cases of ext{8,16,32}{u,s} */
2632 if (opc != 1) { /* SBFM or UBFM */
2633 tcg_gen_movi_i64(tcg_rd, 0);
2636 /* do the bit move operation */
2638 /* Wd<s-r:0> = Wn<s:r> */
2639 tcg_gen_shri_i64(tcg_tmp, tcg_tmp, ri);
2641 len = (si - ri) + 1;
2643 /* Wd<32+s-r,32-r> = Wn<s:0> */
2648 tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, pos, len);
2650 if (opc == 0) { /* SBFM - sign extend the destination field */
2651 tcg_gen_shli_i64(tcg_rd, tcg_rd, 64 - (pos + len));
2652 tcg_gen_sari_i64(tcg_rd, tcg_rd, 64 - (pos + len));
2655 if (!sf) { /* zero extend final result */
2656 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2661 * 31 30 29 28 23 22 21 20 16 15 10 9 5 4 0
2662 * +----+------+-------------+---+----+------+--------+------+------+
2663 * | sf | op21 | 1 0 0 1 1 1 | N | o0 | Rm | imms | Rn | Rd |
2664 * +----+------+-------------+---+----+------+--------+------+------+
2666 static void disas_extract(DisasContext *s, uint32_t insn)
2668 unsigned int sf, n, rm, imm, rn, rd, bitsize, op21, op0;
2670 sf = extract32(insn, 31, 1);
2671 n = extract32(insn, 22, 1);
2672 rm = extract32(insn, 16, 5);
2673 imm = extract32(insn, 10, 6);
2674 rn = extract32(insn, 5, 5);
2675 rd = extract32(insn, 0, 5);
2676 op21 = extract32(insn, 29, 2);
2677 op0 = extract32(insn, 21, 1);
2678 bitsize = sf ? 64 : 32;
2680 if (sf != n || op21 || op0 || imm >= bitsize) {
2681 unallocated_encoding(s);
2683 TCGv_i64 tcg_rd, tcg_rm, tcg_rn;
2685 tcg_rd = cpu_reg(s, rd);
2688 /* OPTME: we can special case rm==rn as a rotate */
2689 tcg_rm = read_cpu_reg(s, rm, sf);
2690 tcg_rn = read_cpu_reg(s, rn, sf);
2691 tcg_gen_shri_i64(tcg_rm, tcg_rm, imm);
2692 tcg_gen_shli_i64(tcg_rn, tcg_rn, bitsize - imm);
2693 tcg_gen_or_i64(tcg_rd, tcg_rm, tcg_rn);
2695 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2698 /* tcg shl_i32/shl_i64 is undefined for 32/64 bit shifts,
2699 * so an extract from bit 0 is a special case.
2702 tcg_gen_mov_i64(tcg_rd, cpu_reg(s, rm));
2704 tcg_gen_ext32u_i64(tcg_rd, cpu_reg(s, rm));
2711 /* C3.4 Data processing - immediate */
2712 static void disas_data_proc_imm(DisasContext *s, uint32_t insn)
2714 switch (extract32(insn, 23, 6)) {
2715 case 0x20: case 0x21: /* PC-rel. addressing */
2716 disas_pc_rel_adr(s, insn);
2718 case 0x22: case 0x23: /* Add/subtract (immediate) */
2719 disas_add_sub_imm(s, insn);
2721 case 0x24: /* Logical (immediate) */
2722 disas_logic_imm(s, insn);
2724 case 0x25: /* Move wide (immediate) */
2725 disas_movw_imm(s, insn);
2727 case 0x26: /* Bitfield */
2728 disas_bitfield(s, insn);
2730 case 0x27: /* Extract */
2731 disas_extract(s, insn);
2734 unallocated_encoding(s);
2739 /* Shift a TCGv src by TCGv shift_amount, put result in dst.
2740 * Note that it is the caller's responsibility to ensure that the
2741 * shift amount is in range (ie 0..31 or 0..63) and provide the ARM
2742 * mandated semantics for out of range shifts.
2744 static void shift_reg(TCGv_i64 dst, TCGv_i64 src, int sf,
2745 enum a64_shift_type shift_type, TCGv_i64 shift_amount)
2747 switch (shift_type) {
2748 case A64_SHIFT_TYPE_LSL:
2749 tcg_gen_shl_i64(dst, src, shift_amount);
2751 case A64_SHIFT_TYPE_LSR:
2752 tcg_gen_shr_i64(dst, src, shift_amount);
2754 case A64_SHIFT_TYPE_ASR:
2756 tcg_gen_ext32s_i64(dst, src);
2758 tcg_gen_sar_i64(dst, sf ? src : dst, shift_amount);
2760 case A64_SHIFT_TYPE_ROR:
2762 tcg_gen_rotr_i64(dst, src, shift_amount);
2765 t0 = tcg_temp_new_i32();
2766 t1 = tcg_temp_new_i32();
2767 tcg_gen_trunc_i64_i32(t0, src);
2768 tcg_gen_trunc_i64_i32(t1, shift_amount);
2769 tcg_gen_rotr_i32(t0, t0, t1);
2770 tcg_gen_extu_i32_i64(dst, t0);
2771 tcg_temp_free_i32(t0);
2772 tcg_temp_free_i32(t1);
2776 assert(FALSE); /* all shift types should be handled */
2780 if (!sf) { /* zero extend final result */
2781 tcg_gen_ext32u_i64(dst, dst);
2785 /* Shift a TCGv src by immediate, put result in dst.
2786 * The shift amount must be in range (this should always be true as the
2787 * relevant instructions will UNDEF on bad shift immediates).
2789 static void shift_reg_imm(TCGv_i64 dst, TCGv_i64 src, int sf,
2790 enum a64_shift_type shift_type, unsigned int shift_i)
2792 assert(shift_i < (sf ? 64 : 32));
2795 tcg_gen_mov_i64(dst, src);
2797 TCGv_i64 shift_const;
2799 shift_const = tcg_const_i64(shift_i);
2800 shift_reg(dst, src, sf, shift_type, shift_const);
2801 tcg_temp_free_i64(shift_const);
2805 /* C3.5.10 Logical (shifted register)
2806 * 31 30 29 28 24 23 22 21 20 16 15 10 9 5 4 0
2807 * +----+-----+-----------+-------+---+------+--------+------+------+
2808 * | sf | opc | 0 1 0 1 0 | shift | N | Rm | imm6 | Rn | Rd |
2809 * +----+-----+-----------+-------+---+------+--------+------+------+
2811 static void disas_logic_reg(DisasContext *s, uint32_t insn)
2813 TCGv_i64 tcg_rd, tcg_rn, tcg_rm;
2814 unsigned int sf, opc, shift_type, invert, rm, shift_amount, rn, rd;
2816 sf = extract32(insn, 31, 1);
2817 opc = extract32(insn, 29, 2);
2818 shift_type = extract32(insn, 22, 2);
2819 invert = extract32(insn, 21, 1);
2820 rm = extract32(insn, 16, 5);
2821 shift_amount = extract32(insn, 10, 6);
2822 rn = extract32(insn, 5, 5);
2823 rd = extract32(insn, 0, 5);
2825 if (!sf && (shift_amount & (1 << 5))) {
2826 unallocated_encoding(s);
2830 tcg_rd = cpu_reg(s, rd);
2832 if (opc == 1 && shift_amount == 0 && shift_type == 0 && rn == 31) {
2833 /* Unshifted ORR and ORN with WZR/XZR is the standard encoding for
2834 * register-register MOV and MVN, so it is worth special casing.
2836 tcg_rm = cpu_reg(s, rm);
2838 tcg_gen_not_i64(tcg_rd, tcg_rm);
2840 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2844 tcg_gen_mov_i64(tcg_rd, tcg_rm);
2846 tcg_gen_ext32u_i64(tcg_rd, tcg_rm);
2852 tcg_rm = read_cpu_reg(s, rm, sf);
2855 shift_reg_imm(tcg_rm, tcg_rm, sf, shift_type, shift_amount);
2858 tcg_rn = cpu_reg(s, rn);
2860 switch (opc | (invert << 2)) {
2863 tcg_gen_and_i64(tcg_rd, tcg_rn, tcg_rm);
2866 tcg_gen_or_i64(tcg_rd, tcg_rn, tcg_rm);
2869 tcg_gen_xor_i64(tcg_rd, tcg_rn, tcg_rm);
2873 tcg_gen_andc_i64(tcg_rd, tcg_rn, tcg_rm);
2876 tcg_gen_orc_i64(tcg_rd, tcg_rn, tcg_rm);
2879 tcg_gen_eqv_i64(tcg_rd, tcg_rn, tcg_rm);
2887 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2891 gen_logic_CC(sf, tcg_rd);
2896 * C3.5.1 Add/subtract (extended register)
2898 * 31|30|29|28 24|23 22|21|20 16|15 13|12 10|9 5|4 0|
2899 * +--+--+--+-----------+-----+--+-------+------+------+----+----+
2900 * |sf|op| S| 0 1 0 1 1 | opt | 1| Rm |option| imm3 | Rn | Rd |
2901 * +--+--+--+-----------+-----+--+-------+------+------+----+----+
2903 * sf: 0 -> 32bit, 1 -> 64bit
2904 * op: 0 -> add , 1 -> sub
2907 * option: extension type (see DecodeRegExtend)
2908 * imm3: optional shift to Rm
2910 * Rd = Rn + LSL(extend(Rm), amount)
2912 static void disas_add_sub_ext_reg(DisasContext *s, uint32_t insn)
2914 int rd = extract32(insn, 0, 5);
2915 int rn = extract32(insn, 5, 5);
2916 int imm3 = extract32(insn, 10, 3);
2917 int option = extract32(insn, 13, 3);
2918 int rm = extract32(insn, 16, 5);
2919 bool setflags = extract32(insn, 29, 1);
2920 bool sub_op = extract32(insn, 30, 1);
2921 bool sf = extract32(insn, 31, 1);
2923 TCGv_i64 tcg_rm, tcg_rn; /* temps */
2925 TCGv_i64 tcg_result;
2928 unallocated_encoding(s);
2932 /* non-flag setting ops may use SP */
2934 tcg_rn = read_cpu_reg_sp(s, rn, sf);
2935 tcg_rd = cpu_reg_sp(s, rd);
2937 tcg_rn = read_cpu_reg(s, rn, sf);
2938 tcg_rd = cpu_reg(s, rd);
2941 tcg_rm = read_cpu_reg(s, rm, sf);
2942 ext_and_shift_reg(tcg_rm, tcg_rm, option, imm3);
2944 tcg_result = tcg_temp_new_i64();
2948 tcg_gen_sub_i64(tcg_result, tcg_rn, tcg_rm);
2950 tcg_gen_add_i64(tcg_result, tcg_rn, tcg_rm);
2954 gen_sub_CC(sf, tcg_result, tcg_rn, tcg_rm);
2956 gen_add_CC(sf, tcg_result, tcg_rn, tcg_rm);
2961 tcg_gen_mov_i64(tcg_rd, tcg_result);
2963 tcg_gen_ext32u_i64(tcg_rd, tcg_result);
2966 tcg_temp_free_i64(tcg_result);
2970 * C3.5.2 Add/subtract (shifted register)
2972 * 31 30 29 28 24 23 22 21 20 16 15 10 9 5 4 0
2973 * +--+--+--+-----------+-----+--+-------+---------+------+------+
2974 * |sf|op| S| 0 1 0 1 1 |shift| 0| Rm | imm6 | Rn | Rd |
2975 * +--+--+--+-----------+-----+--+-------+---------+------+------+
2977 * sf: 0 -> 32bit, 1 -> 64bit
2978 * op: 0 -> add , 1 -> sub
2980 * shift: 00 -> LSL, 01 -> LSR, 10 -> ASR, 11 -> RESERVED
2981 * imm6: Shift amount to apply to Rm before the add/sub
2983 static void disas_add_sub_reg(DisasContext *s, uint32_t insn)
2985 int rd = extract32(insn, 0, 5);
2986 int rn = extract32(insn, 5, 5);
2987 int imm6 = extract32(insn, 10, 6);
2988 int rm = extract32(insn, 16, 5);
2989 int shift_type = extract32(insn, 22, 2);
2990 bool setflags = extract32(insn, 29, 1);
2991 bool sub_op = extract32(insn, 30, 1);
2992 bool sf = extract32(insn, 31, 1);
2994 TCGv_i64 tcg_rd = cpu_reg(s, rd);
2995 TCGv_i64 tcg_rn, tcg_rm;
2996 TCGv_i64 tcg_result;
2998 if ((shift_type == 3) || (!sf && (imm6 > 31))) {
2999 unallocated_encoding(s);
3003 tcg_rn = read_cpu_reg(s, rn, sf);
3004 tcg_rm = read_cpu_reg(s, rm, sf);
3006 shift_reg_imm(tcg_rm, tcg_rm, sf, shift_type, imm6);
3008 tcg_result = tcg_temp_new_i64();
3012 tcg_gen_sub_i64(tcg_result, tcg_rn, tcg_rm);
3014 tcg_gen_add_i64(tcg_result, tcg_rn, tcg_rm);
3018 gen_sub_CC(sf, tcg_result, tcg_rn, tcg_rm);
3020 gen_add_CC(sf, tcg_result, tcg_rn, tcg_rm);
3025 tcg_gen_mov_i64(tcg_rd, tcg_result);
3027 tcg_gen_ext32u_i64(tcg_rd, tcg_result);
3030 tcg_temp_free_i64(tcg_result);
3033 /* C3.5.9 Data-processing (3 source)
3035 31 30 29 28 24 23 21 20 16 15 14 10 9 5 4 0
3036 +--+------+-----------+------+------+----+------+------+------+
3037 |sf| op54 | 1 1 0 1 1 | op31 | Rm | o0 | Ra | Rn | Rd |
3038 +--+------+-----------+------+------+----+------+------+------+
3041 static void disas_data_proc_3src(DisasContext *s, uint32_t insn)
3043 int rd = extract32(insn, 0, 5);
3044 int rn = extract32(insn, 5, 5);
3045 int ra = extract32(insn, 10, 5);
3046 int rm = extract32(insn, 16, 5);
3047 int op_id = (extract32(insn, 29, 3) << 4) |
3048 (extract32(insn, 21, 3) << 1) |
3049 extract32(insn, 15, 1);
3050 bool sf = extract32(insn, 31, 1);
3051 bool is_sub = extract32(op_id, 0, 1);
3052 bool is_high = extract32(op_id, 2, 1);
3053 bool is_signed = false;
3058 /* Note that op_id is sf:op54:op31:o0 so it includes the 32/64 size flag */
3060 case 0x42: /* SMADDL */
3061 case 0x43: /* SMSUBL */
3062 case 0x44: /* SMULH */
3065 case 0x0: /* MADD (32bit) */
3066 case 0x1: /* MSUB (32bit) */
3067 case 0x40: /* MADD (64bit) */
3068 case 0x41: /* MSUB (64bit) */
3069 case 0x4a: /* UMADDL */
3070 case 0x4b: /* UMSUBL */
3071 case 0x4c: /* UMULH */
3074 unallocated_encoding(s);
3079 TCGv_i64 low_bits = tcg_temp_new_i64(); /* low bits discarded */
3080 TCGv_i64 tcg_rd = cpu_reg(s, rd);
3081 TCGv_i64 tcg_rn = cpu_reg(s, rn);
3082 TCGv_i64 tcg_rm = cpu_reg(s, rm);
3085 tcg_gen_muls2_i64(low_bits, tcg_rd, tcg_rn, tcg_rm);
3087 tcg_gen_mulu2_i64(low_bits, tcg_rd, tcg_rn, tcg_rm);
3090 tcg_temp_free_i64(low_bits);
3094 tcg_op1 = tcg_temp_new_i64();
3095 tcg_op2 = tcg_temp_new_i64();
3096 tcg_tmp = tcg_temp_new_i64();
3099 tcg_gen_mov_i64(tcg_op1, cpu_reg(s, rn));
3100 tcg_gen_mov_i64(tcg_op2, cpu_reg(s, rm));
3103 tcg_gen_ext32s_i64(tcg_op1, cpu_reg(s, rn));
3104 tcg_gen_ext32s_i64(tcg_op2, cpu_reg(s, rm));
3106 tcg_gen_ext32u_i64(tcg_op1, cpu_reg(s, rn));
3107 tcg_gen_ext32u_i64(tcg_op2, cpu_reg(s, rm));
3111 if (ra == 31 && !is_sub) {
3112 /* Special-case MADD with rA == XZR; it is the standard MUL alias */
3113 tcg_gen_mul_i64(cpu_reg(s, rd), tcg_op1, tcg_op2);
3115 tcg_gen_mul_i64(tcg_tmp, tcg_op1, tcg_op2);
3117 tcg_gen_sub_i64(cpu_reg(s, rd), cpu_reg(s, ra), tcg_tmp);
3119 tcg_gen_add_i64(cpu_reg(s, rd), cpu_reg(s, ra), tcg_tmp);
3124 tcg_gen_ext32u_i64(cpu_reg(s, rd), cpu_reg(s, rd));
3127 tcg_temp_free_i64(tcg_op1);
3128 tcg_temp_free_i64(tcg_op2);
3129 tcg_temp_free_i64(tcg_tmp);
3132 /* C3.5.3 - Add/subtract (with carry)
3133 * 31 30 29 28 27 26 25 24 23 22 21 20 16 15 10 9 5 4 0
3134 * +--+--+--+------------------------+------+---------+------+-----+
3135 * |sf|op| S| 1 1 0 1 0 0 0 0 | rm | opcode2 | Rn | Rd |
3136 * +--+--+--+------------------------+------+---------+------+-----+
3140 static void disas_adc_sbc(DisasContext *s, uint32_t insn)
3142 unsigned int sf, op, setflags, rm, rn, rd;
3143 TCGv_i64 tcg_y, tcg_rn, tcg_rd;
3145 if (extract32(insn, 10, 6) != 0) {
3146 unallocated_encoding(s);
3150 sf = extract32(insn, 31, 1);
3151 op = extract32(insn, 30, 1);
3152 setflags = extract32(insn, 29, 1);
3153 rm = extract32(insn, 16, 5);
3154 rn = extract32(insn, 5, 5);
3155 rd = extract32(insn, 0, 5);
3157 tcg_rd = cpu_reg(s, rd);
3158 tcg_rn = cpu_reg(s, rn);
3161 tcg_y = new_tmp_a64(s);
3162 tcg_gen_not_i64(tcg_y, cpu_reg(s, rm));
3164 tcg_y = cpu_reg(s, rm);
3168 gen_adc_CC(sf, tcg_rd, tcg_rn, tcg_y);
3170 gen_adc(sf, tcg_rd, tcg_rn, tcg_y);
3174 /* C3.5.4 - C3.5.5 Conditional compare (immediate / register)
3175 * 31 30 29 28 27 26 25 24 23 22 21 20 16 15 12 11 10 9 5 4 3 0
3176 * +--+--+--+------------------------+--------+------+----+--+------+--+-----+
3177 * |sf|op| S| 1 1 0 1 0 0 1 0 |imm5/rm | cond |i/r |o2| Rn |o3|nzcv |
3178 * +--+--+--+------------------------+--------+------+----+--+------+--+-----+
3181 static void disas_cc(DisasContext *s, uint32_t insn)
3183 unsigned int sf, op, y, cond, rn, nzcv, is_imm;
3184 int label_continue = -1;
3185 TCGv_i64 tcg_tmp, tcg_y, tcg_rn;
3187 if (!extract32(insn, 29, 1)) {
3188 unallocated_encoding(s);
3191 if (insn & (1 << 10 | 1 << 4)) {
3192 unallocated_encoding(s);
3195 sf = extract32(insn, 31, 1);
3196 op = extract32(insn, 30, 1);
3197 is_imm = extract32(insn, 11, 1);
3198 y = extract32(insn, 16, 5); /* y = rm (reg) or imm5 (imm) */
3199 cond = extract32(insn, 12, 4);
3200 rn = extract32(insn, 5, 5);
3201 nzcv = extract32(insn, 0, 4);
3203 if (cond < 0x0e) { /* not always */
3204 int label_match = gen_new_label();
3205 label_continue = gen_new_label();
3206 arm_gen_test_cc(cond, label_match);
3208 tcg_tmp = tcg_temp_new_i64();
3209 tcg_gen_movi_i64(tcg_tmp, nzcv << 28);
3210 gen_set_nzcv(tcg_tmp);
3211 tcg_temp_free_i64(tcg_tmp);
3212 tcg_gen_br(label_continue);
3213 gen_set_label(label_match);
3215 /* match, or condition is always */
3217 tcg_y = new_tmp_a64(s);
3218 tcg_gen_movi_i64(tcg_y, y);
3220 tcg_y = cpu_reg(s, y);
3222 tcg_rn = cpu_reg(s, rn);
3224 tcg_tmp = tcg_temp_new_i64();
3226 gen_sub_CC(sf, tcg_tmp, tcg_rn, tcg_y);
3228 gen_add_CC(sf, tcg_tmp, tcg_rn, tcg_y);
3230 tcg_temp_free_i64(tcg_tmp);
3232 if (cond < 0x0e) { /* continue */
3233 gen_set_label(label_continue);
3237 /* C3.5.6 Conditional select
3238 * 31 30 29 28 21 20 16 15 12 11 10 9 5 4 0
3239 * +----+----+---+-----------------+------+------+-----+------+------+
3240 * | sf | op | S | 1 1 0 1 0 1 0 0 | Rm | cond | op2 | Rn | Rd |
3241 * +----+----+---+-----------------+------+------+-----+------+------+
3243 static void disas_cond_select(DisasContext *s, uint32_t insn)
3245 unsigned int sf, else_inv, rm, cond, else_inc, rn, rd;
3246 TCGv_i64 tcg_rd, tcg_src;
3248 if (extract32(insn, 29, 1) || extract32(insn, 11, 1)) {
3249 /* S == 1 or op2<1> == 1 */
3250 unallocated_encoding(s);
3253 sf = extract32(insn, 31, 1);
3254 else_inv = extract32(insn, 30, 1);
3255 rm = extract32(insn, 16, 5);
3256 cond = extract32(insn, 12, 4);
3257 else_inc = extract32(insn, 10, 1);
3258 rn = extract32(insn, 5, 5);
3259 rd = extract32(insn, 0, 5);
3262 /* silly no-op write; until we use movcond we must special-case
3263 * this to avoid a dead temporary across basic blocks.
3268 tcg_rd = cpu_reg(s, rd);
3270 if (cond >= 0x0e) { /* condition "always" */
3271 tcg_src = read_cpu_reg(s, rn, sf);
3272 tcg_gen_mov_i64(tcg_rd, tcg_src);
3274 /* OPTME: we could use movcond here, at the cost of duplicating
3275 * a lot of the arm_gen_test_cc() logic.
3277 int label_match = gen_new_label();
3278 int label_continue = gen_new_label();
3280 arm_gen_test_cc(cond, label_match);
3282 tcg_src = cpu_reg(s, rm);
3284 if (else_inv && else_inc) {
3285 tcg_gen_neg_i64(tcg_rd, tcg_src);
3286 } else if (else_inv) {
3287 tcg_gen_not_i64(tcg_rd, tcg_src);
3288 } else if (else_inc) {
3289 tcg_gen_addi_i64(tcg_rd, tcg_src, 1);
3291 tcg_gen_mov_i64(tcg_rd, tcg_src);
3294 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
3296 tcg_gen_br(label_continue);
3298 gen_set_label(label_match);
3299 tcg_src = read_cpu_reg(s, rn, sf);
3300 tcg_gen_mov_i64(tcg_rd, tcg_src);
3302 gen_set_label(label_continue);
3306 static void handle_clz(DisasContext *s, unsigned int sf,
3307 unsigned int rn, unsigned int rd)
3309 TCGv_i64 tcg_rd, tcg_rn;
3310 tcg_rd = cpu_reg(s, rd);
3311 tcg_rn = cpu_reg(s, rn);
3314 gen_helper_clz64(tcg_rd, tcg_rn);
3316 TCGv_i32 tcg_tmp32 = tcg_temp_new_i32();
3317 tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn);
3318 gen_helper_clz(tcg_tmp32, tcg_tmp32);
3319 tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32);
3320 tcg_temp_free_i32(tcg_tmp32);
3324 static void handle_cls(DisasContext *s, unsigned int sf,
3325 unsigned int rn, unsigned int rd)
3327 TCGv_i64 tcg_rd, tcg_rn;
3328 tcg_rd = cpu_reg(s, rd);
3329 tcg_rn = cpu_reg(s, rn);
3332 gen_helper_cls64(tcg_rd, tcg_rn);
3334 TCGv_i32 tcg_tmp32 = tcg_temp_new_i32();
3335 tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn);
3336 gen_helper_cls32(tcg_tmp32, tcg_tmp32);
3337 tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32);
3338 tcg_temp_free_i32(tcg_tmp32);
3342 static void handle_rbit(DisasContext *s, unsigned int sf,
3343 unsigned int rn, unsigned int rd)
3345 TCGv_i64 tcg_rd, tcg_rn;
3346 tcg_rd = cpu_reg(s, rd);
3347 tcg_rn = cpu_reg(s, rn);
3350 gen_helper_rbit64(tcg_rd, tcg_rn);
3352 TCGv_i32 tcg_tmp32 = tcg_temp_new_i32();
3353 tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn);
3354 gen_helper_rbit(tcg_tmp32, tcg_tmp32);
3355 tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32);
3356 tcg_temp_free_i32(tcg_tmp32);
3360 /* C5.6.149 REV with sf==1, opcode==3 ("REV64") */
3361 static void handle_rev64(DisasContext *s, unsigned int sf,
3362 unsigned int rn, unsigned int rd)
3365 unallocated_encoding(s);
3368 tcg_gen_bswap64_i64(cpu_reg(s, rd), cpu_reg(s, rn));
3371 /* C5.6.149 REV with sf==0, opcode==2
3372 * C5.6.151 REV32 (sf==1, opcode==2)
3374 static void handle_rev32(DisasContext *s, unsigned int sf,
3375 unsigned int rn, unsigned int rd)
3377 TCGv_i64 tcg_rd = cpu_reg(s, rd);
3380 TCGv_i64 tcg_tmp = tcg_temp_new_i64();
3381 TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
3383 /* bswap32_i64 requires zero high word */
3384 tcg_gen_ext32u_i64(tcg_tmp, tcg_rn);
3385 tcg_gen_bswap32_i64(tcg_rd, tcg_tmp);
3386 tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
3387 tcg_gen_bswap32_i64(tcg_tmp, tcg_tmp);
3388 tcg_gen_concat32_i64(tcg_rd, tcg_rd, tcg_tmp);
3390 tcg_temp_free_i64(tcg_tmp);
3392 tcg_gen_ext32u_i64(tcg_rd, cpu_reg(s, rn));
3393 tcg_gen_bswap32_i64(tcg_rd, tcg_rd);
3397 /* C5.6.150 REV16 (opcode==1) */
3398 static void handle_rev16(DisasContext *s, unsigned int sf,
3399 unsigned int rn, unsigned int rd)
3401 TCGv_i64 tcg_rd = cpu_reg(s, rd);
3402 TCGv_i64 tcg_tmp = tcg_temp_new_i64();
3403 TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
3405 tcg_gen_andi_i64(tcg_tmp, tcg_rn, 0xffff);
3406 tcg_gen_bswap16_i64(tcg_rd, tcg_tmp);
3408 tcg_gen_shri_i64(tcg_tmp, tcg_rn, 16);
3409 tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
3410 tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
3411 tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 16, 16);
3414 tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
3415 tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
3416 tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
3417 tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 32, 16);
3419 tcg_gen_shri_i64(tcg_tmp, tcg_rn, 48);
3420 tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
3421 tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 48, 16);
3424 tcg_temp_free_i64(tcg_tmp);
3427 /* C3.5.7 Data-processing (1 source)
3428 * 31 30 29 28 21 20 16 15 10 9 5 4 0
3429 * +----+---+---+-----------------+---------+--------+------+------+
3430 * | sf | 1 | S | 1 1 0 1 0 1 1 0 | opcode2 | opcode | Rn | Rd |
3431 * +----+---+---+-----------------+---------+--------+------+------+
3433 static void disas_data_proc_1src(DisasContext *s, uint32_t insn)
3435 unsigned int sf, opcode, rn, rd;
3437 if (extract32(insn, 29, 1) || extract32(insn, 16, 5)) {
3438 unallocated_encoding(s);
3442 sf = extract32(insn, 31, 1);
3443 opcode = extract32(insn, 10, 6);
3444 rn = extract32(insn, 5, 5);
3445 rd = extract32(insn, 0, 5);
3449 handle_rbit(s, sf, rn, rd);
3452 handle_rev16(s, sf, rn, rd);
3455 handle_rev32(s, sf, rn, rd);
3458 handle_rev64(s, sf, rn, rd);
3461 handle_clz(s, sf, rn, rd);
3464 handle_cls(s, sf, rn, rd);
3469 static void handle_div(DisasContext *s, bool is_signed, unsigned int sf,
3470 unsigned int rm, unsigned int rn, unsigned int rd)
3472 TCGv_i64 tcg_n, tcg_m, tcg_rd;
3473 tcg_rd = cpu_reg(s, rd);
3475 if (!sf && is_signed) {
3476 tcg_n = new_tmp_a64(s);
3477 tcg_m = new_tmp_a64(s);
3478 tcg_gen_ext32s_i64(tcg_n, cpu_reg(s, rn));
3479 tcg_gen_ext32s_i64(tcg_m, cpu_reg(s, rm));
3481 tcg_n = read_cpu_reg(s, rn, sf);
3482 tcg_m = read_cpu_reg(s, rm, sf);
3486 gen_helper_sdiv64(tcg_rd, tcg_n, tcg_m);
3488 gen_helper_udiv64(tcg_rd, tcg_n, tcg_m);
3491 if (!sf) { /* zero extend final result */
3492 tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
3496 /* C5.6.115 LSLV, C5.6.118 LSRV, C5.6.17 ASRV, C5.6.154 RORV */
3497 static void handle_shift_reg(DisasContext *s,
3498 enum a64_shift_type shift_type, unsigned int sf,
3499 unsigned int rm, unsigned int rn, unsigned int rd)
3501 TCGv_i64 tcg_shift = tcg_temp_new_i64();
3502 TCGv_i64 tcg_rd = cpu_reg(s, rd);
3503 TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
3505 tcg_gen_andi_i64(tcg_shift, cpu_reg(s, rm), sf ? 63 : 31);
3506 shift_reg(tcg_rd, tcg_rn, sf, shift_type, tcg_shift);
3507 tcg_temp_free_i64(tcg_shift);
3510 /* C3.5.8 Data-processing (2 source)
3511 * 31 30 29 28 21 20 16 15 10 9 5 4 0
3512 * +----+---+---+-----------------+------+--------+------+------+
3513 * | sf | 0 | S | 1 1 0 1 0 1 1 0 | Rm | opcode | Rn | Rd |
3514 * +----+---+---+-----------------+------+--------+------+------+
3516 static void disas_data_proc_2src(DisasContext *s, uint32_t insn)
3518 unsigned int sf, rm, opcode, rn, rd;
3519 sf = extract32(insn, 31, 1);
3520 rm = extract32(insn, 16, 5);
3521 opcode = extract32(insn, 10, 6);
3522 rn = extract32(insn, 5, 5);
3523 rd = extract32(insn, 0, 5);
3525 if (extract32(insn, 29, 1)) {
3526 unallocated_encoding(s);
3532 handle_div(s, false, sf, rm, rn, rd);
3535 handle_div(s, true, sf, rm, rn, rd);
3538 handle_shift_reg(s, A64_SHIFT_TYPE_LSL, sf, rm, rn, rd);
3541 handle_shift_reg(s, A64_SHIFT_TYPE_LSR, sf, rm, rn, rd);
3544 handle_shift_reg(s, A64_SHIFT_TYPE_ASR, sf, rm, rn, rd);
3547 handle_shift_reg(s, A64_SHIFT_TYPE_ROR, sf, rm, rn, rd);
3556 case 23: /* CRC32 */
3557 unsupported_encoding(s, insn);
3560 unallocated_encoding(s);
3565 /* C3.5 Data processing - register */
3566 static void disas_data_proc_reg(DisasContext *s, uint32_t insn)
3568 switch (extract32(insn, 24, 5)) {
3569 case 0x0a: /* Logical (shifted register) */
3570 disas_logic_reg(s, insn);
3572 case 0x0b: /* Add/subtract */
3573 if (insn & (1 << 21)) { /* (extended register) */
3574 disas_add_sub_ext_reg(s, insn);
3576 disas_add_sub_reg(s, insn);
3579 case 0x1b: /* Data-processing (3 source) */
3580 disas_data_proc_3src(s, insn);
3583 switch (extract32(insn, 21, 3)) {
3584 case 0x0: /* Add/subtract (with carry) */
3585 disas_adc_sbc(s, insn);
3587 case 0x2: /* Conditional compare */
3588 disas_cc(s, insn); /* both imm and reg forms */
3590 case 0x4: /* Conditional select */
3591 disas_cond_select(s, insn);
3593 case 0x6: /* Data-processing */
3594 if (insn & (1 << 30)) { /* (1 source) */
3595 disas_data_proc_1src(s, insn);
3596 } else { /* (2 source) */
3597 disas_data_proc_2src(s, insn);
3601 unallocated_encoding(s);
3606 unallocated_encoding(s);
3611 /* Convert ARM rounding mode to softfloat */
3612 static inline int arm_rmode_to_sf(int rmode)
3615 case FPROUNDING_TIEAWAY:
3616 rmode = float_round_ties_away;
3618 case FPROUNDING_ODD:
3619 /* FIXME: add support for TIEAWAY and ODD */
3620 qemu_log_mask(LOG_UNIMP, "arm: unimplemented rounding mode: %d\n",
3622 case FPROUNDING_TIEEVEN:
3624 rmode = float_round_nearest_even;
3626 case FPROUNDING_POSINF:
3627 rmode = float_round_up;
3629 case FPROUNDING_NEGINF:
3630 rmode = float_round_down;
3632 case FPROUNDING_ZERO:
3633 rmode = float_round_to_zero;
3639 static void handle_fp_compare(DisasContext *s, bool is_double,
3640 unsigned int rn, unsigned int rm,
3641 bool cmp_with_zero, bool signal_all_nans)
3643 TCGv_i64 tcg_flags = tcg_temp_new_i64();
3644 TCGv_ptr fpst = get_fpstatus_ptr();
3647 TCGv_i64 tcg_vn, tcg_vm;
3649 tcg_vn = read_fp_dreg(s, rn);
3650 if (cmp_with_zero) {
3651 tcg_vm = tcg_const_i64(0);
3653 tcg_vm = read_fp_dreg(s, rm);
3655 if (signal_all_nans) {
3656 gen_helper_vfp_cmped_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3658 gen_helper_vfp_cmpd_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3660 tcg_temp_free_i64(tcg_vn);
3661 tcg_temp_free_i64(tcg_vm);
3663 TCGv_i32 tcg_vn, tcg_vm;
3665 tcg_vn = read_fp_sreg(s, rn);
3666 if (cmp_with_zero) {
3667 tcg_vm = tcg_const_i32(0);
3669 tcg_vm = read_fp_sreg(s, rm);
3671 if (signal_all_nans) {
3672 gen_helper_vfp_cmpes_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3674 gen_helper_vfp_cmps_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3676 tcg_temp_free_i32(tcg_vn);
3677 tcg_temp_free_i32(tcg_vm);
3680 tcg_temp_free_ptr(fpst);
3682 gen_set_nzcv(tcg_flags);
3684 tcg_temp_free_i64(tcg_flags);
3687 /* C3.6.22 Floating point compare
3688 * 31 30 29 28 24 23 22 21 20 16 15 14 13 10 9 5 4 0
3689 * +---+---+---+-----------+------+---+------+-----+---------+------+-------+
3690 * | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | op | 1 0 0 0 | Rn | op2 |
3691 * +---+---+---+-----------+------+---+------+-----+---------+------+-------+
3693 static void disas_fp_compare(DisasContext *s, uint32_t insn)
3695 unsigned int mos, type, rm, op, rn, opc, op2r;
3697 mos = extract32(insn, 29, 3);
3698 type = extract32(insn, 22, 2); /* 0 = single, 1 = double */
3699 rm = extract32(insn, 16, 5);
3700 op = extract32(insn, 14, 2);
3701 rn = extract32(insn, 5, 5);
3702 opc = extract32(insn, 3, 2);
3703 op2r = extract32(insn, 0, 3);
3705 if (mos || op || op2r || type > 1) {
3706 unallocated_encoding(s);
3710 handle_fp_compare(s, type, rn, rm, opc & 1, opc & 2);
3713 /* C3.6.23 Floating point conditional compare
3714 * 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 3 0
3715 * +---+---+---+-----------+------+---+------+------+-----+------+----+------+
3716 * | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | cond | 0 1 | Rn | op | nzcv |
3717 * +---+---+---+-----------+------+---+------+------+-----+------+----+------+
3719 static void disas_fp_ccomp(DisasContext *s, uint32_t insn)
3721 unsigned int mos, type, rm, cond, rn, op, nzcv;
3723 int label_continue = -1;
3725 mos = extract32(insn, 29, 3);
3726 type = extract32(insn, 22, 2); /* 0 = single, 1 = double */
3727 rm = extract32(insn, 16, 5);
3728 cond = extract32(insn, 12, 4);
3729 rn = extract32(insn, 5, 5);
3730 op = extract32(insn, 4, 1);
3731 nzcv = extract32(insn, 0, 4);
3733 if (mos || type > 1) {
3734 unallocated_encoding(s);
3738 if (cond < 0x0e) { /* not always */
3739 int label_match = gen_new_label();
3740 label_continue = gen_new_label();
3741 arm_gen_test_cc(cond, label_match);
3743 tcg_flags = tcg_const_i64(nzcv << 28);
3744 gen_set_nzcv(tcg_flags);
3745 tcg_temp_free_i64(tcg_flags);
3746 tcg_gen_br(label_continue);
3747 gen_set_label(label_match);
3750 handle_fp_compare(s, type, rn, rm, false, op);
3753 gen_set_label(label_continue);
3757 /* copy src FP register to dst FP register; type specifies single or double */
3758 static void gen_mov_fp2fp(DisasContext *s, int type, int dst, int src)
3761 TCGv_i64 v = read_fp_dreg(s, src);
3762 write_fp_dreg(s, dst, v);
3763 tcg_temp_free_i64(v);
3765 TCGv_i32 v = read_fp_sreg(s, src);
3766 write_fp_sreg(s, dst, v);
3767 tcg_temp_free_i32(v);
3771 /* C3.6.24 Floating point conditional select
3772 * 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
3773 * +---+---+---+-----------+------+---+------+------+-----+------+------+
3774 * | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | cond | 1 1 | Rn | Rd |
3775 * +---+---+---+-----------+------+---+------+------+-----+------+------+
3777 static void disas_fp_csel(DisasContext *s, uint32_t insn)
3779 unsigned int mos, type, rm, cond, rn, rd;
3780 int label_continue = -1;
3782 mos = extract32(insn, 29, 3);
3783 type = extract32(insn, 22, 2); /* 0 = single, 1 = double */
3784 rm = extract32(insn, 16, 5);
3785 cond = extract32(insn, 12, 4);
3786 rn = extract32(insn, 5, 5);
3787 rd = extract32(insn, 0, 5);
3789 if (mos || type > 1) {
3790 unallocated_encoding(s);
3794 if (cond < 0x0e) { /* not always */
3795 int label_match = gen_new_label();
3796 label_continue = gen_new_label();
3797 arm_gen_test_cc(cond, label_match);
3799 gen_mov_fp2fp(s, type, rd, rm);
3800 tcg_gen_br(label_continue);
3801 gen_set_label(label_match);
3804 gen_mov_fp2fp(s, type, rd, rn);
3806 if (cond < 0x0e) { /* continue */
3807 gen_set_label(label_continue);
3811 /* C3.6.25 Floating-point data-processing (1 source) - single precision */
3812 static void handle_fp_1src_single(DisasContext *s, int opcode, int rd, int rn)
3818 fpst = get_fpstatus_ptr();
3819 tcg_op = read_fp_sreg(s, rn);
3820 tcg_res = tcg_temp_new_i32();
3823 case 0x0: /* FMOV */
3824 tcg_gen_mov_i32(tcg_res, tcg_op);
3826 case 0x1: /* FABS */
3827 gen_helper_vfp_abss(tcg_res, tcg_op);
3829 case 0x2: /* FNEG */
3830 gen_helper_vfp_negs(tcg_res, tcg_op);
3832 case 0x3: /* FSQRT */
3833 gen_helper_vfp_sqrts(tcg_res, tcg_op, cpu_env);
3835 case 0x8: /* FRINTN */
3836 case 0x9: /* FRINTP */
3837 case 0xa: /* FRINTM */
3838 case 0xb: /* FRINTZ */
3839 case 0xc: /* FRINTA */
3841 TCGv_i32 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(opcode & 7));
3843 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3844 gen_helper_rints(tcg_res, tcg_op, fpst);
3846 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3847 tcg_temp_free_i32(tcg_rmode);
3850 case 0xe: /* FRINTX */
3851 gen_helper_rints_exact(tcg_res, tcg_op, fpst);
3853 case 0xf: /* FRINTI */
3854 gen_helper_rints(tcg_res, tcg_op, fpst);
3860 write_fp_sreg(s, rd, tcg_res);
3862 tcg_temp_free_ptr(fpst);
3863 tcg_temp_free_i32(tcg_op);
3864 tcg_temp_free_i32(tcg_res);
3867 /* C3.6.25 Floating-point data-processing (1 source) - double precision */
3868 static void handle_fp_1src_double(DisasContext *s, int opcode, int rd, int rn)
3874 fpst = get_fpstatus_ptr();
3875 tcg_op = read_fp_dreg(s, rn);
3876 tcg_res = tcg_temp_new_i64();
3879 case 0x0: /* FMOV */
3880 tcg_gen_mov_i64(tcg_res, tcg_op);
3882 case 0x1: /* FABS */
3883 gen_helper_vfp_absd(tcg_res, tcg_op);
3885 case 0x2: /* FNEG */
3886 gen_helper_vfp_negd(tcg_res, tcg_op);
3888 case 0x3: /* FSQRT */
3889 gen_helper_vfp_sqrtd(tcg_res, tcg_op, cpu_env);
3891 case 0x8: /* FRINTN */
3892 case 0x9: /* FRINTP */
3893 case 0xa: /* FRINTM */
3894 case 0xb: /* FRINTZ */
3895 case 0xc: /* FRINTA */
3897 TCGv_i32 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(opcode & 7));
3899 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3900 gen_helper_rintd(tcg_res, tcg_op, fpst);
3902 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3903 tcg_temp_free_i32(tcg_rmode);
3906 case 0xe: /* FRINTX */
3907 gen_helper_rintd_exact(tcg_res, tcg_op, fpst);
3909 case 0xf: /* FRINTI */
3910 gen_helper_rintd(tcg_res, tcg_op, fpst);
3916 write_fp_dreg(s, rd, tcg_res);
3918 tcg_temp_free_ptr(fpst);
3919 tcg_temp_free_i64(tcg_op);
3920 tcg_temp_free_i64(tcg_res);
3923 static void handle_fp_fcvt(DisasContext *s, int opcode,
3924 int rd, int rn, int dtype, int ntype)
3929 TCGv_i32 tcg_rn = read_fp_sreg(s, rn);
3931 /* Single to double */
3932 TCGv_i64 tcg_rd = tcg_temp_new_i64();
3933 gen_helper_vfp_fcvtds(tcg_rd, tcg_rn, cpu_env);
3934 write_fp_dreg(s, rd, tcg_rd);
3935 tcg_temp_free_i64(tcg_rd);
3937 /* Single to half */
3938 TCGv_i32 tcg_rd = tcg_temp_new_i32();
3939 gen_helper_vfp_fcvt_f32_to_f16(tcg_rd, tcg_rn, cpu_env);
3940 /* write_fp_sreg is OK here because top half of tcg_rd is zero */
3941 write_fp_sreg(s, rd, tcg_rd);
3942 tcg_temp_free_i32(tcg_rd);
3944 tcg_temp_free_i32(tcg_rn);
3949 TCGv_i64 tcg_rn = read_fp_dreg(s, rn);
3950 TCGv_i32 tcg_rd = tcg_temp_new_i32();
3952 /* Double to single */
3953 gen_helper_vfp_fcvtsd(tcg_rd, tcg_rn, cpu_env);
3955 /* Double to half */
3956 gen_helper_vfp_fcvt_f64_to_f16(tcg_rd, tcg_rn, cpu_env);
3957 /* write_fp_sreg is OK here because top half of tcg_rd is zero */
3959 write_fp_sreg(s, rd, tcg_rd);
3960 tcg_temp_free_i32(tcg_rd);
3961 tcg_temp_free_i64(tcg_rn);
3966 TCGv_i32 tcg_rn = read_fp_sreg(s, rn);
3967 tcg_gen_ext16u_i32(tcg_rn, tcg_rn);
3969 /* Half to single */
3970 TCGv_i32 tcg_rd = tcg_temp_new_i32();
3971 gen_helper_vfp_fcvt_f16_to_f32(tcg_rd, tcg_rn, cpu_env);
3972 write_fp_sreg(s, rd, tcg_rd);
3973 tcg_temp_free_i32(tcg_rd);
3975 /* Half to double */
3976 TCGv_i64 tcg_rd = tcg_temp_new_i64();
3977 gen_helper_vfp_fcvt_f16_to_f64(tcg_rd, tcg_rn, cpu_env);
3978 write_fp_dreg(s, rd, tcg_rd);
3979 tcg_temp_free_i64(tcg_rd);
3981 tcg_temp_free_i32(tcg_rn);
3989 /* C3.6.25 Floating point data-processing (1 source)
3990 * 31 30 29 28 24 23 22 21 20 15 14 10 9 5 4 0
3991 * +---+---+---+-----------+------+---+--------+-----------+------+------+
3992 * | M | 0 | S | 1 1 1 1 0 | type | 1 | opcode | 1 0 0 0 0 | Rn | Rd |
3993 * +---+---+---+-----------+------+---+--------+-----------+------+------+
3995 static void disas_fp_1src(DisasContext *s, uint32_t insn)
3997 int type = extract32(insn, 22, 2);
3998 int opcode = extract32(insn, 15, 6);
3999 int rn = extract32(insn, 5, 5);
4000 int rd = extract32(insn, 0, 5);
4003 case 0x4: case 0x5: case 0x7:
4005 /* FCVT between half, single and double precision */
4006 int dtype = extract32(opcode, 0, 2);
4007 if (type == 2 || dtype == type) {
4008 unallocated_encoding(s);
4011 handle_fp_fcvt(s, opcode, rd, rn, dtype, type);
4017 /* 32-to-32 and 64-to-64 ops */
4020 handle_fp_1src_single(s, opcode, rd, rn);
4023 handle_fp_1src_double(s, opcode, rd, rn);
4026 unallocated_encoding(s);
4030 unallocated_encoding(s);
4035 /* C3.6.26 Floating-point data-processing (2 source) - single precision */
4036 static void handle_fp_2src_single(DisasContext *s, int opcode,
4037 int rd, int rn, int rm)
4044 tcg_res = tcg_temp_new_i32();
4045 fpst = get_fpstatus_ptr();
4046 tcg_op1 = read_fp_sreg(s, rn);
4047 tcg_op2 = read_fp_sreg(s, rm);
4050 case 0x0: /* FMUL */
4051 gen_helper_vfp_muls(tcg_res, tcg_op1, tcg_op2, fpst);
4053 case 0x1: /* FDIV */
4054 gen_helper_vfp_divs(tcg_res, tcg_op1, tcg_op2, fpst);
4056 case 0x2: /* FADD */
4057 gen_helper_vfp_adds(tcg_res, tcg_op1, tcg_op2, fpst);
4059 case 0x3: /* FSUB */
4060 gen_helper_vfp_subs(tcg_res, tcg_op1, tcg_op2, fpst);
4062 case 0x4: /* FMAX */
4063 gen_helper_vfp_maxs(tcg_res, tcg_op1, tcg_op2, fpst);
4065 case 0x5: /* FMIN */
4066 gen_helper_vfp_mins(tcg_res, tcg_op1, tcg_op2, fpst);
4068 case 0x6: /* FMAXNM */
4069 gen_helper_vfp_maxnums(tcg_res, tcg_op1, tcg_op2, fpst);
4071 case 0x7: /* FMINNM */
4072 gen_helper_vfp_minnums(tcg_res, tcg_op1, tcg_op2, fpst);
4074 case 0x8: /* FNMUL */
4075 gen_helper_vfp_muls(tcg_res, tcg_op1, tcg_op2, fpst);
4076 gen_helper_vfp_negs(tcg_res, tcg_res);
4080 write_fp_sreg(s, rd, tcg_res);
4082 tcg_temp_free_ptr(fpst);
4083 tcg_temp_free_i32(tcg_op1);
4084 tcg_temp_free_i32(tcg_op2);
4085 tcg_temp_free_i32(tcg_res);
4088 /* C3.6.26 Floating-point data-processing (2 source) - double precision */
4089 static void handle_fp_2src_double(DisasContext *s, int opcode,
4090 int rd, int rn, int rm)
4097 tcg_res = tcg_temp_new_i64();
4098 fpst = get_fpstatus_ptr();
4099 tcg_op1 = read_fp_dreg(s, rn);
4100 tcg_op2 = read_fp_dreg(s, rm);
4103 case 0x0: /* FMUL */
4104 gen_helper_vfp_muld(tcg_res, tcg_op1, tcg_op2, fpst);
4106 case 0x1: /* FDIV */
4107 gen_helper_vfp_divd(tcg_res, tcg_op1, tcg_op2, fpst);
4109 case 0x2: /* FADD */
4110 gen_helper_vfp_addd(tcg_res, tcg_op1, tcg_op2, fpst);
4112 case 0x3: /* FSUB */
4113 gen_helper_vfp_subd(tcg_res, tcg_op1, tcg_op2, fpst);
4115 case 0x4: /* FMAX */
4116 gen_helper_vfp_maxd(tcg_res, tcg_op1, tcg_op2, fpst);
4118 case 0x5: /* FMIN */
4119 gen_helper_vfp_mind(tcg_res, tcg_op1, tcg_op2, fpst);
4121 case 0x6: /* FMAXNM */
4122 gen_helper_vfp_maxnumd(tcg_res, tcg_op1, tcg_op2, fpst);
4124 case 0x7: /* FMINNM */
4125 gen_helper_vfp_minnumd(tcg_res, tcg_op1, tcg_op2, fpst);
4127 case 0x8: /* FNMUL */
4128 gen_helper_vfp_muld(tcg_res, tcg_op1, tcg_op2, fpst);
4129 gen_helper_vfp_negd(tcg_res, tcg_res);
4133 write_fp_dreg(s, rd, tcg_res);
4135 tcg_temp_free_ptr(fpst);
4136 tcg_temp_free_i64(tcg_op1);
4137 tcg_temp_free_i64(tcg_op2);
4138 tcg_temp_free_i64(tcg_res);
4141 /* C3.6.26 Floating point data-processing (2 source)
4142 * 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
4143 * +---+---+---+-----------+------+---+------+--------+-----+------+------+
4144 * | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | opcode | 1 0 | Rn | Rd |
4145 * +---+---+---+-----------+------+---+------+--------+-----+------+------+
4147 static void disas_fp_2src(DisasContext *s, uint32_t insn)
4149 int type = extract32(insn, 22, 2);
4150 int rd = extract32(insn, 0, 5);
4151 int rn = extract32(insn, 5, 5);
4152 int rm = extract32(insn, 16, 5);
4153 int opcode = extract32(insn, 12, 4);
4156 unallocated_encoding(s);
4162 handle_fp_2src_single(s, opcode, rd, rn, rm);
4165 handle_fp_2src_double(s, opcode, rd, rn, rm);
4168 unallocated_encoding(s);
4172 /* C3.6.27 Floating-point data-processing (3 source) - single precision */
4173 static void handle_fp_3src_single(DisasContext *s, bool o0, bool o1,
4174 int rd, int rn, int rm, int ra)
4176 TCGv_i32 tcg_op1, tcg_op2, tcg_op3;
4177 TCGv_i32 tcg_res = tcg_temp_new_i32();
4178 TCGv_ptr fpst = get_fpstatus_ptr();
4180 tcg_op1 = read_fp_sreg(s, rn);
4181 tcg_op2 = read_fp_sreg(s, rm);
4182 tcg_op3 = read_fp_sreg(s, ra);
4184 /* These are fused multiply-add, and must be done as one
4185 * floating point operation with no rounding between the
4186 * multiplication and addition steps.
4187 * NB that doing the negations here as separate steps is
4188 * correct : an input NaN should come out with its sign bit
4189 * flipped if it is a negated-input.
4192 gen_helper_vfp_negs(tcg_op3, tcg_op3);
4196 gen_helper_vfp_negs(tcg_op1, tcg_op1);
4199 gen_helper_vfp_muladds(tcg_res, tcg_op1, tcg_op2, tcg_op3, fpst);
4201 write_fp_sreg(s, rd, tcg_res);
4203 tcg_temp_free_ptr(fpst);
4204 tcg_temp_free_i32(tcg_op1);
4205 tcg_temp_free_i32(tcg_op2);
4206 tcg_temp_free_i32(tcg_op3);
4207 tcg_temp_free_i32(tcg_res);
4210 /* C3.6.27 Floating-point data-processing (3 source) - double precision */
4211 static void handle_fp_3src_double(DisasContext *s, bool o0, bool o1,
4212 int rd, int rn, int rm, int ra)
4214 TCGv_i64 tcg_op1, tcg_op2, tcg_op3;
4215 TCGv_i64 tcg_res = tcg_temp_new_i64();
4216 TCGv_ptr fpst = get_fpstatus_ptr();
4218 tcg_op1 = read_fp_dreg(s, rn);
4219 tcg_op2 = read_fp_dreg(s, rm);
4220 tcg_op3 = read_fp_dreg(s, ra);
4222 /* These are fused multiply-add, and must be done as one
4223 * floating point operation with no rounding between the
4224 * multiplication and addition steps.
4225 * NB that doing the negations here as separate steps is
4226 * correct : an input NaN should come out with its sign bit
4227 * flipped if it is a negated-input.
4230 gen_helper_vfp_negd(tcg_op3, tcg_op3);
4234 gen_helper_vfp_negd(tcg_op1, tcg_op1);
4237 gen_helper_vfp_muladdd(tcg_res, tcg_op1, tcg_op2, tcg_op3, fpst);
4239 write_fp_dreg(s, rd, tcg_res);
4241 tcg_temp_free_ptr(fpst);
4242 tcg_temp_free_i64(tcg_op1);
4243 tcg_temp_free_i64(tcg_op2);
4244 tcg_temp_free_i64(tcg_op3);
4245 tcg_temp_free_i64(tcg_res);
4248 /* C3.6.27 Floating point data-processing (3 source)
4249 * 31 30 29 28 24 23 22 21 20 16 15 14 10 9 5 4 0
4250 * +---+---+---+-----------+------+----+------+----+------+------+------+
4251 * | M | 0 | S | 1 1 1 1 1 | type | o1 | Rm | o0 | Ra | Rn | Rd |
4252 * +---+---+---+-----------+------+----+------+----+------+------+------+
4254 static void disas_fp_3src(DisasContext *s, uint32_t insn)
4256 int type = extract32(insn, 22, 2);
4257 int rd = extract32(insn, 0, 5);
4258 int rn = extract32(insn, 5, 5);
4259 int ra = extract32(insn, 10, 5);
4260 int rm = extract32(insn, 16, 5);
4261 bool o0 = extract32(insn, 15, 1);
4262 bool o1 = extract32(insn, 21, 1);
4266 handle_fp_3src_single(s, o0, o1, rd, rn, rm, ra);
4269 handle_fp_3src_double(s, o0, o1, rd, rn, rm, ra);
4272 unallocated_encoding(s);
4276 /* C3.6.28 Floating point immediate
4277 * 31 30 29 28 24 23 22 21 20 13 12 10 9 5 4 0
4278 * +---+---+---+-----------+------+---+------------+-------+------+------+
4279 * | M | 0 | S | 1 1 1 1 0 | type | 1 | imm8 | 1 0 0 | imm5 | Rd |
4280 * +---+---+---+-----------+------+---+------------+-------+------+------+
4282 static void disas_fp_imm(DisasContext *s, uint32_t insn)
4284 int rd = extract32(insn, 0, 5);
4285 int imm8 = extract32(insn, 13, 8);
4286 int is_double = extract32(insn, 22, 2);
4290 if (is_double > 1) {
4291 unallocated_encoding(s);
4295 /* The imm8 encodes the sign bit, enough bits to represent
4296 * an exponent in the range 01....1xx to 10....0xx,
4297 * and the most significant 4 bits of the mantissa; see
4298 * VFPExpandImm() in the v8 ARM ARM.
4301 imm = (extract32(imm8, 7, 1) ? 0x8000 : 0) |
4302 (extract32(imm8, 6, 1) ? 0x3fc0 : 0x4000) |
4303 extract32(imm8, 0, 6);
4306 imm = (extract32(imm8, 7, 1) ? 0x8000 : 0) |
4307 (extract32(imm8, 6, 1) ? 0x3e00 : 0x4000) |
4308 (extract32(imm8, 0, 6) << 3);
4312 tcg_res = tcg_const_i64(imm);
4313 write_fp_dreg(s, rd, tcg_res);
4314 tcg_temp_free_i64(tcg_res);
4317 /* Handle floating point <=> fixed point conversions. Note that we can
4318 * also deal with fp <=> integer conversions as a special case (scale == 64)
4319 * OPTME: consider handling that special case specially or at least skipping
4320 * the call to scalbn in the helpers for zero shifts.
4322 static void handle_fpfpcvt(DisasContext *s, int rd, int rn, int opcode,
4323 bool itof, int rmode, int scale, int sf, int type)
4325 bool is_signed = !(opcode & 1);
4326 bool is_double = type;
4327 TCGv_ptr tcg_fpstatus;
4330 tcg_fpstatus = get_fpstatus_ptr();
4332 tcg_shift = tcg_const_i32(64 - scale);
4335 TCGv_i64 tcg_int = cpu_reg(s, rn);
4337 TCGv_i64 tcg_extend = new_tmp_a64(s);
4340 tcg_gen_ext32s_i64(tcg_extend, tcg_int);
4342 tcg_gen_ext32u_i64(tcg_extend, tcg_int);
4345 tcg_int = tcg_extend;
4349 TCGv_i64 tcg_double = tcg_temp_new_i64();
4351 gen_helper_vfp_sqtod(tcg_double, tcg_int,
4352 tcg_shift, tcg_fpstatus);
4354 gen_helper_vfp_uqtod(tcg_double, tcg_int,
4355 tcg_shift, tcg_fpstatus);
4357 write_fp_dreg(s, rd, tcg_double);
4358 tcg_temp_free_i64(tcg_double);
4360 TCGv_i32 tcg_single = tcg_temp_new_i32();
4362 gen_helper_vfp_sqtos(tcg_single, tcg_int,
4363 tcg_shift, tcg_fpstatus);
4365 gen_helper_vfp_uqtos(tcg_single, tcg_int,
4366 tcg_shift, tcg_fpstatus);
4368 write_fp_sreg(s, rd, tcg_single);
4369 tcg_temp_free_i32(tcg_single);
4372 TCGv_i64 tcg_int = cpu_reg(s, rd);
4375 if (extract32(opcode, 2, 1)) {
4376 /* There are too many rounding modes to all fit into rmode,
4377 * so FCVTA[US] is a special case.
4379 rmode = FPROUNDING_TIEAWAY;
4382 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode));
4384 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
4387 TCGv_i64 tcg_double = read_fp_dreg(s, rn);
4390 gen_helper_vfp_tosld(tcg_int, tcg_double,
4391 tcg_shift, tcg_fpstatus);
4393 gen_helper_vfp_tosqd(tcg_int, tcg_double,
4394 tcg_shift, tcg_fpstatus);
4398 gen_helper_vfp_tould(tcg_int, tcg_double,
4399 tcg_shift, tcg_fpstatus);
4401 gen_helper_vfp_touqd(tcg_int, tcg_double,
4402 tcg_shift, tcg_fpstatus);
4405 tcg_temp_free_i64(tcg_double);
4407 TCGv_i32 tcg_single = read_fp_sreg(s, rn);
4410 gen_helper_vfp_tosqs(tcg_int, tcg_single,
4411 tcg_shift, tcg_fpstatus);
4413 gen_helper_vfp_touqs(tcg_int, tcg_single,
4414 tcg_shift, tcg_fpstatus);
4417 TCGv_i32 tcg_dest = tcg_temp_new_i32();
4419 gen_helper_vfp_tosls(tcg_dest, tcg_single,
4420 tcg_shift, tcg_fpstatus);
4422 gen_helper_vfp_touls(tcg_dest, tcg_single,
4423 tcg_shift, tcg_fpstatus);
4425 tcg_gen_extu_i32_i64(tcg_int, tcg_dest);
4426 tcg_temp_free_i32(tcg_dest);
4428 tcg_temp_free_i32(tcg_single);
4431 gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
4432 tcg_temp_free_i32(tcg_rmode);
4435 tcg_gen_ext32u_i64(tcg_int, tcg_int);
4439 tcg_temp_free_ptr(tcg_fpstatus);
4440 tcg_temp_free_i32(tcg_shift);
4443 /* C3.6.29 Floating point <-> fixed point conversions
4444 * 31 30 29 28 24 23 22 21 20 19 18 16 15 10 9 5 4 0
4445 * +----+---+---+-----------+------+---+-------+--------+-------+------+------+
4446 * | sf | 0 | S | 1 1 1 1 0 | type | 0 | rmode | opcode | scale | Rn | Rd |
4447 * +----+---+---+-----------+------+---+-------+--------+-------+------+------+
4449 static void disas_fp_fixed_conv(DisasContext *s, uint32_t insn)
4451 int rd = extract32(insn, 0, 5);
4452 int rn = extract32(insn, 5, 5);
4453 int scale = extract32(insn, 10, 6);
4454 int opcode = extract32(insn, 16, 3);
4455 int rmode = extract32(insn, 19, 2);
4456 int type = extract32(insn, 22, 2);
4457 bool sbit = extract32(insn, 29, 1);
4458 bool sf = extract32(insn, 31, 1);
4461 if (sbit || (type > 1)
4462 || (!sf && scale < 32)) {
4463 unallocated_encoding(s);
4467 switch ((rmode << 3) | opcode) {
4468 case 0x2: /* SCVTF */
4469 case 0x3: /* UCVTF */
4472 case 0x18: /* FCVTZS */
4473 case 0x19: /* FCVTZU */
4477 unallocated_encoding(s);
4481 handle_fpfpcvt(s, rd, rn, opcode, itof, FPROUNDING_ZERO, scale, sf, type);
4484 static void handle_fmov(DisasContext *s, int rd, int rn, int type, bool itof)
4486 /* FMOV: gpr to or from float, double, or top half of quad fp reg,
4487 * without conversion.
4491 TCGv_i64 tcg_rn = cpu_reg(s, rn);
4497 TCGv_i64 tmp = tcg_temp_new_i64();
4498 tcg_gen_ext32u_i64(tmp, tcg_rn);
4499 tcg_gen_st_i64(tmp, cpu_env, fp_reg_offset(rd, MO_64));
4500 tcg_gen_movi_i64(tmp, 0);
4501 tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(rd));
4502 tcg_temp_free_i64(tmp);
4508 TCGv_i64 tmp = tcg_const_i64(0);
4509 tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_offset(rd, MO_64));
4510 tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(rd));
4511 tcg_temp_free_i64(tmp);
4515 /* 64 bit to top half. */
4516 tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_hi_offset(rd));
4520 TCGv_i64 tcg_rd = cpu_reg(s, rd);
4525 tcg_gen_ld32u_i64(tcg_rd, cpu_env, fp_reg_offset(rn, MO_32));
4529 tcg_gen_ld_i64(tcg_rd, cpu_env, fp_reg_offset(rn, MO_64));
4532 /* 64 bits from top half */
4533 tcg_gen_ld_i64(tcg_rd, cpu_env, fp_reg_hi_offset(rn));
4539 /* C3.6.30 Floating point <-> integer conversions
4540 * 31 30 29 28 24 23 22 21 20 19 18 16 15 10 9 5 4 0
4541 * +----+---+---+-----------+------+---+-------+-----+-------------+----+----+
4542 * | sf | 0 | S | 1 1 1 1 0 | type | 1 | rmode | opc | 0 0 0 0 0 0 | Rn | Rd |
4543 * +----+---+---+-----------+------+---+-------+-----+-------------+----+----+
4545 static void disas_fp_int_conv(DisasContext *s, uint32_t insn)
4547 int rd = extract32(insn, 0, 5);
4548 int rn = extract32(insn, 5, 5);
4549 int opcode = extract32(insn, 16, 3);
4550 int rmode = extract32(insn, 19, 2);
4551 int type = extract32(insn, 22, 2);
4552 bool sbit = extract32(insn, 29, 1);
4553 bool sf = extract32(insn, 31, 1);
4556 unallocated_encoding(s);
4562 bool itof = opcode & 1;
4565 unallocated_encoding(s);
4569 switch (sf << 3 | type << 1 | rmode) {
4570 case 0x0: /* 32 bit */
4571 case 0xa: /* 64 bit */
4572 case 0xd: /* 64 bit to top half of quad */
4575 /* all other sf/type/rmode combinations are invalid */
4576 unallocated_encoding(s);
4580 handle_fmov(s, rd, rn, type, itof);
4582 /* actual FP conversions */
4583 bool itof = extract32(opcode, 1, 1);
4585 if (type > 1 || (rmode != 0 && opcode > 1)) {
4586 unallocated_encoding(s);
4590 handle_fpfpcvt(s, rd, rn, opcode, itof, rmode, 64, sf, type);
4594 /* FP-specific subcases of table C3-6 (SIMD and FP data processing)
4595 * 31 30 29 28 25 24 0
4596 * +---+---+---+---------+-----------------------------+
4597 * | | 0 | | 1 1 1 1 | |
4598 * +---+---+---+---------+-----------------------------+
4600 static void disas_data_proc_fp(DisasContext *s, uint32_t insn)
4602 if (extract32(insn, 24, 1)) {
4603 /* Floating point data-processing (3 source) */
4604 disas_fp_3src(s, insn);
4605 } else if (extract32(insn, 21, 1) == 0) {
4606 /* Floating point to fixed point conversions */
4607 disas_fp_fixed_conv(s, insn);
4609 switch (extract32(insn, 10, 2)) {
4611 /* Floating point conditional compare */
4612 disas_fp_ccomp(s, insn);
4615 /* Floating point data-processing (2 source) */
4616 disas_fp_2src(s, insn);
4619 /* Floating point conditional select */
4620 disas_fp_csel(s, insn);
4623 switch (ctz32(extract32(insn, 12, 4))) {
4624 case 0: /* [15:12] == xxx1 */
4625 /* Floating point immediate */
4626 disas_fp_imm(s, insn);
4628 case 1: /* [15:12] == xx10 */
4629 /* Floating point compare */
4630 disas_fp_compare(s, insn);
4632 case 2: /* [15:12] == x100 */
4633 /* Floating point data-processing (1 source) */
4634 disas_fp_1src(s, insn);
4636 case 3: /* [15:12] == 1000 */
4637 unallocated_encoding(s);
4639 default: /* [15:12] == 0000 */
4640 /* Floating point <-> integer conversions */
4641 disas_fp_int_conv(s, insn);
4649 static void do_ext64(DisasContext *s, TCGv_i64 tcg_left, TCGv_i64 tcg_right,
4652 /* Extract 64 bits from the middle of two concatenated 64 bit
4653 * vector register slices left:right. The extracted bits start
4654 * at 'pos' bits into the right (least significant) side.
4655 * We return the result in tcg_right, and guarantee not to
4658 TCGv_i64 tcg_tmp = tcg_temp_new_i64();
4659 assert(pos > 0 && pos < 64);
4661 tcg_gen_shri_i64(tcg_right, tcg_right, pos);
4662 tcg_gen_shli_i64(tcg_tmp, tcg_left, 64 - pos);
4663 tcg_gen_or_i64(tcg_right, tcg_right, tcg_tmp);
4665 tcg_temp_free_i64(tcg_tmp);
4669 * 31 30 29 24 23 22 21 20 16 15 14 11 10 9 5 4 0
4670 * +---+---+-------------+-----+---+------+---+------+---+------+------+
4671 * | 0 | Q | 1 0 1 1 1 0 | op2 | 0 | Rm | 0 | imm4 | 0 | Rn | Rd |
4672 * +---+---+-------------+-----+---+------+---+------+---+------+------+
4674 static void disas_simd_ext(DisasContext *s, uint32_t insn)
4676 int is_q = extract32(insn, 30, 1);
4677 int op2 = extract32(insn, 22, 2);
4678 int imm4 = extract32(insn, 11, 4);
4679 int rm = extract32(insn, 16, 5);
4680 int rn = extract32(insn, 5, 5);
4681 int rd = extract32(insn, 0, 5);
4682 int pos = imm4 << 3;
4683 TCGv_i64 tcg_resl, tcg_resh;
4685 if (op2 != 0 || (!is_q && extract32(imm4, 3, 1))) {
4686 unallocated_encoding(s);
4690 tcg_resh = tcg_temp_new_i64();
4691 tcg_resl = tcg_temp_new_i64();
4693 /* Vd gets bits starting at pos bits into Vm:Vn. This is
4694 * either extracting 128 bits from a 128:128 concatenation, or
4695 * extracting 64 bits from a 64:64 concatenation.
4698 read_vec_element(s, tcg_resl, rn, 0, MO_64);
4700 read_vec_element(s, tcg_resh, rm, 0, MO_64);
4701 do_ext64(s, tcg_resh, tcg_resl, pos);
4703 tcg_gen_movi_i64(tcg_resh, 0);
4710 EltPosns eltposns[] = { {rn, 0}, {rn, 1}, {rm, 0}, {rm, 1} };
4711 EltPosns *elt = eltposns;
4718 read_vec_element(s, tcg_resl, elt->reg, elt->elt, MO_64);
4720 read_vec_element(s, tcg_resh, elt->reg, elt->elt, MO_64);
4723 do_ext64(s, tcg_resh, tcg_resl, pos);
4724 tcg_hh = tcg_temp_new_i64();
4725 read_vec_element(s, tcg_hh, elt->reg, elt->elt, MO_64);
4726 do_ext64(s, tcg_hh, tcg_resh, pos);
4727 tcg_temp_free_i64(tcg_hh);
4731 write_vec_element(s, tcg_resl, rd, 0, MO_64);
4732 tcg_temp_free_i64(tcg_resl);
4733 write_vec_element(s, tcg_resh, rd, 1, MO_64);
4734 tcg_temp_free_i64(tcg_resh);
4738 * 31 30 29 24 23 22 21 20 16 15 14 13 12 11 10 9 5 4 0
4739 * +---+---+-------------+-----+---+------+---+-----+----+-----+------+------+
4740 * | 0 | Q | 0 0 1 1 1 0 | op2 | 0 | Rm | 0 | len | op | 0 0 | Rn | Rd |
4741 * +---+---+-------------+-----+---+------+---+-----+----+-----+------+------+
4743 static void disas_simd_tb(DisasContext *s, uint32_t insn)
4745 int op2 = extract32(insn, 22, 2);
4746 int is_q = extract32(insn, 30, 1);
4747 int rm = extract32(insn, 16, 5);
4748 int rn = extract32(insn, 5, 5);
4749 int rd = extract32(insn, 0, 5);
4750 int is_tblx = extract32(insn, 12, 1);
4751 int len = extract32(insn, 13, 2);
4752 TCGv_i64 tcg_resl, tcg_resh, tcg_idx;
4753 TCGv_i32 tcg_regno, tcg_numregs;
4756 unallocated_encoding(s);
4760 /* This does a table lookup: for every byte element in the input
4761 * we index into a table formed from up to four vector registers,
4762 * and then the output is the result of the lookups. Our helper
4763 * function does the lookup operation for a single 64 bit part of
4766 tcg_resl = tcg_temp_new_i64();
4767 tcg_resh = tcg_temp_new_i64();
4770 read_vec_element(s, tcg_resl, rd, 0, MO_64);
4772 tcg_gen_movi_i64(tcg_resl, 0);
4774 if (is_tblx && is_q) {
4775 read_vec_element(s, tcg_resh, rd, 1, MO_64);
4777 tcg_gen_movi_i64(tcg_resh, 0);
4780 tcg_idx = tcg_temp_new_i64();
4781 tcg_regno = tcg_const_i32(rn);
4782 tcg_numregs = tcg_const_i32(len + 1);
4783 read_vec_element(s, tcg_idx, rm, 0, MO_64);
4784 gen_helper_simd_tbl(tcg_resl, cpu_env, tcg_resl, tcg_idx,
4785 tcg_regno, tcg_numregs);
4787 read_vec_element(s, tcg_idx, rm, 1, MO_64);
4788 gen_helper_simd_tbl(tcg_resh, cpu_env, tcg_resh, tcg_idx,
4789 tcg_regno, tcg_numregs);
4791 tcg_temp_free_i64(tcg_idx);
4792 tcg_temp_free_i32(tcg_regno);
4793 tcg_temp_free_i32(tcg_numregs);
4795 write_vec_element(s, tcg_resl, rd, 0, MO_64);
4796 tcg_temp_free_i64(tcg_resl);
4797 write_vec_element(s, tcg_resh, rd, 1, MO_64);
4798 tcg_temp_free_i64(tcg_resh);
4801 /* C3.6.3 ZIP/UZP/TRN
4802 * 31 30 29 24 23 22 21 20 16 15 14 12 11 10 9 5 4 0
4803 * +---+---+-------------+------+---+------+---+------------------+------+
4804 * | 0 | Q | 0 0 1 1 1 0 | size | 0 | Rm | 0 | opc | 1 0 | Rn | Rd |
4805 * +---+---+-------------+------+---+------+---+------------------+------+
4807 static void disas_simd_zip_trn(DisasContext *s, uint32_t insn)
4809 unsupported_encoding(s, insn);
4812 /* C3.6.4 AdvSIMD across lanes
4813 * 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
4814 * +---+---+---+-----------+------+-----------+--------+-----+------+------+
4815 * | 0 | Q | U | 0 1 1 1 0 | size | 1 1 0 0 0 | opcode | 1 0 | Rn | Rd |
4816 * +---+---+---+-----------+------+-----------+--------+-----+------+------+
4818 static void disas_simd_across_lanes(DisasContext *s, uint32_t insn)
4820 unsupported_encoding(s, insn);
4823 /* C3.6.5 AdvSIMD copy
4824 * 31 30 29 28 21 20 16 15 14 11 10 9 5 4 0
4825 * +---+---+----+-----------------+------+---+------+---+------+------+
4826 * | 0 | Q | op | 0 1 1 1 0 0 0 0 | imm5 | 0 | imm4 | 1 | Rn | Rd |
4827 * +---+---+----+-----------------+------+---+------+---+------+------+
4829 static void disas_simd_copy(DisasContext *s, uint32_t insn)
4831 unsupported_encoding(s, insn);
4834 /* C3.6.6 AdvSIMD modified immediate
4835 * 31 30 29 28 19 18 16 15 12 11 10 9 5 4 0
4836 * +---+---+----+---------------------+-----+-------+----+---+-------+------+
4837 * | 0 | Q | op | 0 1 1 1 1 0 0 0 0 0 | abc | cmode | o2 | 1 | defgh | Rd |
4838 * +---+---+----+---------------------+-----+-------+----+---+-------+------+
4840 static void disas_simd_mod_imm(DisasContext *s, uint32_t insn)
4842 unsupported_encoding(s, insn);
4845 /* C3.6.7 AdvSIMD scalar copy
4846 * 31 30 29 28 21 20 16 15 14 11 10 9 5 4 0
4847 * +-----+----+-----------------+------+---+------+---+------+------+
4848 * | 0 1 | op | 1 1 1 1 0 0 0 0 | imm5 | 0 | imm4 | 1 | Rn | Rd |
4849 * +-----+----+-----------------+------+---+------+---+------+------+
4851 static void disas_simd_scalar_copy(DisasContext *s, uint32_t insn)
4853 unsupported_encoding(s, insn);
4856 /* C3.6.8 AdvSIMD scalar pairwise
4857 * 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
4858 * +-----+---+-----------+------+-----------+--------+-----+------+------+
4859 * | 0 1 | U | 1 1 1 1 0 | size | 1 1 0 0 0 | opcode | 1 0 | Rn | Rd |
4860 * +-----+---+-----------+------+-----------+--------+-----+------+------+
4862 static void disas_simd_scalar_pairwise(DisasContext *s, uint32_t insn)
4864 unsupported_encoding(s, insn);
4867 /* C3.6.9 AdvSIMD scalar shift by immediate
4868 * 31 30 29 28 23 22 19 18 16 15 11 10 9 5 4 0
4869 * +-----+---+-------------+------+------+--------+---+------+------+
4870 * | 0 1 | U | 1 1 1 1 1 0 | immh | immb | opcode | 1 | Rn | Rd |
4871 * +-----+---+-------------+------+------+--------+---+------+------+
4873 static void disas_simd_scalar_shift_imm(DisasContext *s, uint32_t insn)
4875 unsupported_encoding(s, insn);
4878 /* C3.6.10 AdvSIMD scalar three different
4879 * 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
4880 * +-----+---+-----------+------+---+------+--------+-----+------+------+
4881 * | 0 1 | U | 1 1 1 1 0 | size | 1 | Rm | opcode | 0 0 | Rn | Rd |
4882 * +-----+---+-----------+------+---+------+--------+-----+------+------+
4884 static void disas_simd_scalar_three_reg_diff(DisasContext *s, uint32_t insn)
4886 unsupported_encoding(s, insn);
4889 /* C3.6.11 AdvSIMD scalar three same
4890 * 31 30 29 28 24 23 22 21 20 16 15 11 10 9 5 4 0
4891 * +-----+---+-----------+------+---+------+--------+---+------+------+
4892 * | 0 1 | U | 1 1 1 1 0 | size | 1 | Rm | opcode | 1 | Rn | Rd |
4893 * +-----+---+-----------+------+---+------+--------+---+------+------+
4895 static void disas_simd_scalar_three_reg_same(DisasContext *s, uint32_t insn)
4897 unsupported_encoding(s, insn);
4900 /* C3.6.12 AdvSIMD scalar two reg misc
4901 * 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
4902 * +-----+---+-----------+------+-----------+--------+-----+------+------+
4903 * | 0 1 | U | 1 1 1 1 0 | size | 1 0 0 0 0 | opcode | 1 0 | Rn | Rd |
4904 * +-----+---+-----------+------+-----------+--------+-----+------+------+
4906 static void disas_simd_scalar_two_reg_misc(DisasContext *s, uint32_t insn)
4908 unsupported_encoding(s, insn);
4911 /* C3.6.13 AdvSIMD scalar x indexed element
4912 * 31 30 29 28 24 23 22 21 20 19 16 15 12 11 10 9 5 4 0
4913 * +-----+---+-----------+------+---+---+------+-----+---+---+------+------+
4914 * | 0 1 | U | 1 1 1 1 1 | size | L | M | Rm | opc | H | 0 | Rn | Rd |
4915 * +-----+---+-----------+------+---+---+------+-----+---+---+------+------+
4917 static void disas_simd_scalar_indexed(DisasContext *s, uint32_t insn)
4919 unsupported_encoding(s, insn);
4922 /* C3.6.14 AdvSIMD shift by immediate
4923 * 31 30 29 28 23 22 19 18 16 15 11 10 9 5 4 0
4924 * +---+---+---+-------------+------+------+--------+---+------+------+
4925 * | 0 | Q | U | 0 1 1 1 1 0 | immh | immb | opcode | 1 | Rn | Rd |
4926 * +---+---+---+-------------+------+------+--------+---+------+------+
4928 static void disas_simd_shift_imm(DisasContext *s, uint32_t insn)
4930 unsupported_encoding(s, insn);
4933 /* C3.6.15 AdvSIMD three different
4934 * 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
4935 * +---+---+---+-----------+------+---+------+--------+-----+------+------+
4936 * | 0 | Q | U | 0 1 1 1 0 | size | 1 | Rm | opcode | 0 0 | Rn | Rd |
4937 * +---+---+---+-----------+------+---+------+--------+-----+------+------+
4939 static void disas_simd_three_reg_diff(DisasContext *s, uint32_t insn)
4941 unsupported_encoding(s, insn);
4944 /* C3.6.16 AdvSIMD three same
4945 * 31 30 29 28 24 23 22 21 20 16 15 11 10 9 5 4 0
4946 * +---+---+---+-----------+------+---+------+--------+---+------+------+
4947 * | 0 | Q | U | 0 1 1 1 0 | size | 1 | Rm | opcode | 1 | Rn | Rd |
4948 * +---+---+---+-----------+------+---+------+--------+---+------+------+
4950 static void disas_simd_three_reg_same(DisasContext *s, uint32_t insn)
4952 unsupported_encoding(s, insn);
4955 /* C3.6.17 AdvSIMD two reg misc
4956 * 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
4957 * +---+---+---+-----------+------+-----------+--------+-----+------+------+
4958 * | 0 | Q | U | 0 1 1 1 0 | size | 1 0 0 0 0 | opcode | 1 0 | Rn | Rd |
4959 * +---+---+---+-----------+------+-----------+--------+-----+------+------+
4961 static void disas_simd_two_reg_misc(DisasContext *s, uint32_t insn)
4963 unsupported_encoding(s, insn);
4966 /* C3.6.18 AdvSIMD vector x indexed element
4967 * 31 30 29 28 24 23 22 21 20 19 16 15 12 11 10 9 5 4 0
4968 * +---+---+---+-----------+------+---+---+------+-----+---+---+------+------+
4969 * | 0 | Q | U | 0 1 1 1 1 | size | L | M | Rm | opc | H | 0 | Rn | Rd |
4970 * +---+---+---+-----------+------+---+---+------+-----+---+---+------+------+
4972 static void disas_simd_indexed_vector(DisasContext *s, uint32_t insn)
4974 unsupported_encoding(s, insn);
4977 /* C3.6.19 Crypto AES
4978 * 31 24 23 22 21 17 16 12 11 10 9 5 4 0
4979 * +-----------------+------+-----------+--------+-----+------+------+
4980 * | 0 1 0 0 1 1 1 0 | size | 1 0 1 0 0 | opcode | 1 0 | Rn | Rd |
4981 * +-----------------+------+-----------+--------+-----+------+------+
4983 static void disas_crypto_aes(DisasContext *s, uint32_t insn)
4985 unsupported_encoding(s, insn);
4988 /* C3.6.20 Crypto three-reg SHA
4989 * 31 24 23 22 21 20 16 15 14 12 11 10 9 5 4 0
4990 * +-----------------+------+---+------+---+--------+-----+------+------+
4991 * | 0 1 0 1 1 1 1 0 | size | 0 | Rm | 0 | opcode | 0 0 | Rn | Rd |
4992 * +-----------------+------+---+------+---+--------+-----+------+------+
4994 static void disas_crypto_three_reg_sha(DisasContext *s, uint32_t insn)
4996 unsupported_encoding(s, insn);
4999 /* C3.6.21 Crypto two-reg SHA
5000 * 31 24 23 22 21 17 16 12 11 10 9 5 4 0
5001 * +-----------------+------+-----------+--------+-----+------+------+
5002 * | 0 1 0 1 1 1 1 0 | size | 1 0 1 0 0 | opcode | 1 0 | Rn | Rd |
5003 * +-----------------+------+-----------+--------+-----+------+------+
5005 static void disas_crypto_two_reg_sha(DisasContext *s, uint32_t insn)
5007 unsupported_encoding(s, insn);
5010 /* C3.6 Data processing - SIMD, inc Crypto
5012 * As the decode gets a little complex we are using a table based
5013 * approach for this part of the decode.
5015 static const AArch64DecodeTable data_proc_simd[] = {
5016 /* pattern , mask , fn */
5017 { 0x0e200400, 0x9f200400, disas_simd_three_reg_same },
5018 { 0x0e200000, 0x9f200c00, disas_simd_three_reg_diff },
5019 { 0x0e200800, 0x9f3e0c00, disas_simd_two_reg_misc },
5020 { 0x0e300800, 0x9f3e0c00, disas_simd_across_lanes },
5021 { 0x0e000400, 0x9fe08400, disas_simd_copy },
5022 { 0x0f000000, 0x9f000400, disas_simd_indexed_vector },
5023 /* simd_mod_imm decode is a subset of simd_shift_imm, so must precede it */
5024 { 0x0f000400, 0x9ff80400, disas_simd_mod_imm },
5025 { 0x0f000400, 0x9f800400, disas_simd_shift_imm },
5026 { 0x0e000000, 0xbf208c00, disas_simd_tb },
5027 { 0x0e000800, 0xbf208c00, disas_simd_zip_trn },
5028 { 0x2e000000, 0xbf208400, disas_simd_ext },
5029 { 0x5e200400, 0xdf200400, disas_simd_scalar_three_reg_same },
5030 { 0x5e200000, 0xdf200c00, disas_simd_scalar_three_reg_diff },
5031 { 0x5e200800, 0xdf3e0c00, disas_simd_scalar_two_reg_misc },
5032 { 0x5e300800, 0xdf3e0c00, disas_simd_scalar_pairwise },
5033 { 0x5e000400, 0xdfe08400, disas_simd_scalar_copy },
5034 { 0x5f000000, 0xdf000400, disas_simd_scalar_indexed },
5035 { 0x5f000400, 0xdf800400, disas_simd_scalar_shift_imm },
5036 { 0x4e280800, 0xff3e0c00, disas_crypto_aes },
5037 { 0x5e000000, 0xff208c00, disas_crypto_three_reg_sha },
5038 { 0x5e280800, 0xff3e0c00, disas_crypto_two_reg_sha },
5039 { 0x00000000, 0x00000000, NULL }
5042 static void disas_data_proc_simd(DisasContext *s, uint32_t insn)
5044 /* Note that this is called with all non-FP cases from
5045 * table C3-6 so it must UNDEF for entries not specifically
5046 * allocated to instructions in that table.
5048 AArch64DecodeFn *fn = lookup_disas_fn(&data_proc_simd[0], insn);
5052 unallocated_encoding(s);
5056 /* C3.6 Data processing - SIMD and floating point */
5057 static void disas_data_proc_simd_fp(DisasContext *s, uint32_t insn)
5059 if (extract32(insn, 28, 1) == 1 && extract32(insn, 30, 1) == 0) {
5060 disas_data_proc_fp(s, insn);
5062 /* SIMD, including crypto */
5063 disas_data_proc_simd(s, insn);
5067 /* C3.1 A64 instruction index by encoding */
5068 static void disas_a64_insn(CPUARMState *env, DisasContext *s)
5072 insn = arm_ldl_code(env, s->pc, s->bswap_code);
5076 switch (extract32(insn, 25, 4)) {
5077 case 0x0: case 0x1: case 0x2: case 0x3: /* UNALLOCATED */
5078 unallocated_encoding(s);
5080 case 0x8: case 0x9: /* Data processing - immediate */
5081 disas_data_proc_imm(s, insn);
5083 case 0xa: case 0xb: /* Branch, exception generation and system insns */
5084 disas_b_exc_sys(s, insn);
5089 case 0xe: /* Loads and stores */
5090 disas_ldst(s, insn);
5093 case 0xd: /* Data processing - register */
5094 disas_data_proc_reg(s, insn);
5097 case 0xf: /* Data processing - SIMD and floating point */
5098 disas_data_proc_simd_fp(s, insn);
5101 assert(FALSE); /* all 15 cases should be handled above */
5105 /* if we allocated any temporaries, free them here */
5109 void gen_intermediate_code_internal_a64(ARMCPU *cpu,
5110 TranslationBlock *tb,
5113 CPUState *cs = CPU(cpu);
5114 CPUARMState *env = &cpu->env;
5115 DisasContext dc1, *dc = &dc1;
5117 uint16_t *gen_opc_end;
5119 target_ulong pc_start;
5120 target_ulong next_page_start;
5128 gen_opc_end = tcg_ctx.gen_opc_buf + OPC_MAX_SIZE;
5130 dc->is_jmp = DISAS_NEXT;
5132 dc->singlestep_enabled = cs->singlestep_enabled;
5138 dc->condexec_mask = 0;
5139 dc->condexec_cond = 0;
5140 #if !defined(CONFIG_USER_ONLY)
5143 dc->vfp_enabled = 0;
5146 dc->cp_regs = cpu->cp_regs;
5147 dc->current_pl = arm_current_pl(env);
5149 init_tmp_a64_array(dc);
5151 next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
5154 max_insns = tb->cflags & CF_COUNT_MASK;
5155 if (max_insns == 0) {
5156 max_insns = CF_COUNT_MASK;
5161 tcg_clear_temp_count();
5164 if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
5165 QTAILQ_FOREACH(bp, &env->breakpoints, entry) {
5166 if (bp->pc == dc->pc) {
5167 gen_exception_insn(dc, 0, EXCP_DEBUG);
5168 /* Advance PC so that clearing the breakpoint will
5169 invalidate this TB. */
5171 goto done_generating;
5177 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
5181 tcg_ctx.gen_opc_instr_start[lj++] = 0;
5184 tcg_ctx.gen_opc_pc[lj] = dc->pc;
5185 tcg_ctx.gen_opc_instr_start[lj] = 1;
5186 tcg_ctx.gen_opc_icount[lj] = num_insns;
5189 if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO)) {
5193 if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP | CPU_LOG_TB_OP_OPT))) {
5194 tcg_gen_debug_insn_start(dc->pc);
5197 disas_a64_insn(env, dc);
5199 if (tcg_check_temp_count()) {
5200 fprintf(stderr, "TCG temporary leak before "TARGET_FMT_lx"\n",
5204 /* Translation stops when a conditional branch is encountered.
5205 * Otherwise the subsequent code could get translated several times.
5206 * Also stop translation when a page boundary is reached. This
5207 * ensures prefetch aborts occur at the right place.
5210 } while (!dc->is_jmp && tcg_ctx.gen_opc_ptr < gen_opc_end &&
5211 !cs->singlestep_enabled &&
5213 dc->pc < next_page_start &&
5214 num_insns < max_insns);
5216 if (tb->cflags & CF_LAST_IO) {
5220 if (unlikely(cs->singlestep_enabled) && dc->is_jmp != DISAS_EXC) {
5221 /* Note that this means single stepping WFI doesn't halt the CPU.
5222 * For conditional branch insns this is harmless unreachable code as
5223 * gen_goto_tb() has already handled emitting the debug exception
5224 * (and thus a tb-jump is not possible when singlestepping).
5226 assert(dc->is_jmp != DISAS_TB_JUMP);
5227 if (dc->is_jmp != DISAS_JUMP) {
5228 gen_a64_set_pc_im(dc->pc);
5230 gen_exception(EXCP_DEBUG);
5232 switch (dc->is_jmp) {
5234 gen_goto_tb(dc, 1, dc->pc);
5238 gen_a64_set_pc_im(dc->pc);
5241 /* indicate that the hash table must be used to find the next TB */
5249 /* This is a special case because we don't want to just halt the CPU
5250 * if trying to debug across a WFI.
5252 gen_helper_wfi(cpu_env);
5258 gen_tb_end(tb, num_insns);
5259 *tcg_ctx.gen_opc_ptr = INDEX_op_end;
5262 if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
5263 qemu_log("----------------\n");
5264 qemu_log("IN: %s\n", lookup_symbol(pc_start));
5265 log_target_disas(env, pc_start, dc->pc - pc_start,
5266 dc->thumb | (dc->bswap_code << 1));
5271 j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
5274 tcg_ctx.gen_opc_instr_start[lj++] = 0;
5277 tb->size = dc->pc - pc_start;
5278 tb->icount = num_insns;