root / target-arm / translate-a64.c @ 0ae39320
History | View | Annotate | Download (183.6 kB)
1 |
/*
|
---|---|
2 |
* AArch64 translation
|
3 |
*
|
4 |
* Copyright (c) 2013 Alexander Graf <agraf@suse.de>
|
5 |
*
|
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.
|
10 |
*
|
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.
|
15 |
*
|
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/>.
|
18 |
*/
|
19 |
#include <stdarg.h> |
20 |
#include <stdlib.h> |
21 |
#include <stdio.h> |
22 |
#include <string.h> |
23 |
#include <inttypes.h> |
24 |
|
25 |
#include "cpu.h" |
26 |
#include "tcg-op.h" |
27 |
#include "qemu/log.h" |
28 |
#include "translate.h" |
29 |
#include "qemu/host-utils.h" |
30 |
|
31 |
#include "exec/gen-icount.h" |
32 |
|
33 |
#include "helper.h" |
34 |
#define GEN_HELPER 1 |
35 |
#include "helper.h" |
36 |
|
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;
|
40 |
|
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;
|
48 |
#endif
|
49 |
|
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" |
55 |
}; |
56 |
|
57 |
enum a64_shift_type {
|
58 |
A64_SHIFT_TYPE_LSL = 0,
|
59 |
A64_SHIFT_TYPE_LSR = 1,
|
60 |
A64_SHIFT_TYPE_ASR = 2,
|
61 |
A64_SHIFT_TYPE_ROR = 3
|
62 |
}; |
63 |
|
64 |
/* Table based decoder typedefs - used when the relevant bits for decode
|
65 |
* are too awkwardly scattered across the instruction (eg SIMD).
|
66 |
*/
|
67 |
typedef void AArch64DecodeFn(DisasContext *s, uint32_t insn); |
68 |
|
69 |
typedef struct AArch64DecodeTable { |
70 |
uint32_t pattern; |
71 |
uint32_t mask; |
72 |
AArch64DecodeFn *disas_fn; |
73 |
} AArch64DecodeTable; |
74 |
|
75 |
/* initialize TCG globals. */
|
76 |
void a64_translate_init(void) |
77 |
{ |
78 |
int i;
|
79 |
|
80 |
cpu_pc = tcg_global_mem_new_i64(TCG_AREG0, |
81 |
offsetof(CPUARMState, pc), |
82 |
"pc");
|
83 |
for (i = 0; i < 32; i++) { |
84 |
cpu_X[i] = tcg_global_mem_new_i64(TCG_AREG0, |
85 |
offsetof(CPUARMState, xregs[i]), |
86 |
regnames[i]); |
87 |
} |
88 |
|
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");
|
93 |
|
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");
|
105 |
#endif
|
106 |
} |
107 |
|
108 |
void aarch64_cpu_dump_state(CPUState *cs, FILE *f,
|
109 |
fprintf_function cpu_fprintf, int flags)
|
110 |
{ |
111 |
ARMCPU *cpu = ARM_CPU(cs); |
112 |
CPUARMState *env = &cpu->env; |
113 |
uint32_t psr = pstate_read(env); |
114 |
int i;
|
115 |
|
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]);
|
120 |
if ((i % 4) == 3) { |
121 |
cpu_fprintf(f, "\n");
|
122 |
} else {
|
123 |
cpu_fprintf(f, " ");
|
124 |
} |
125 |
} |
126 |
cpu_fprintf(f, "PSTATE=%08x (flags %c%c%c%c)\n",
|
127 |
psr, |
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");
|
133 |
|
134 |
if (flags & CPU_DUMP_FPU) {
|
135 |
int numvfpregs = 32; |
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 " ", |
140 |
i, vhi, vlo); |
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", |
144 |
i + 1, vhi, vlo);
|
145 |
} |
146 |
cpu_fprintf(f, "FPCR: %08x FPSR: %08x\n",
|
147 |
vfp_get_fpcr(env), vfp_get_fpsr(env)); |
148 |
} |
149 |
} |
150 |
|
151 |
static int get_mem_index(DisasContext *s) |
152 |
{ |
153 |
#ifdef CONFIG_USER_ONLY
|
154 |
return 1; |
155 |
#else
|
156 |
return s->user;
|
157 |
#endif
|
158 |
} |
159 |
|
160 |
void gen_a64_set_pc_im(uint64_t val)
|
161 |
{ |
162 |
tcg_gen_movi_i64(cpu_pc, val); |
163 |
} |
164 |
|
165 |
static void gen_exception(int excp) |
166 |
{ |
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); |
171 |
} |
172 |
|
173 |
static void gen_exception_insn(DisasContext *s, int offset, int excp) |
174 |
{ |
175 |
gen_a64_set_pc_im(s->pc - offset); |
176 |
gen_exception(excp); |
177 |
s->is_jmp = DISAS_EXC; |
178 |
} |
179 |
|
180 |
static inline bool use_goto_tb(DisasContext *s, int n, uint64_t dest) |
181 |
{ |
182 |
/* No direct tb linking with singlestep or deterministic io */
|
183 |
if (s->singlestep_enabled || (s->tb->cflags & CF_LAST_IO)) {
|
184 |
return false; |
185 |
} |
186 |
|
187 |
/* Only link tbs from inside the same guest page */
|
188 |
if ((s->tb->pc & TARGET_PAGE_MASK) != (dest & TARGET_PAGE_MASK)) {
|
189 |
return false; |
190 |
} |
191 |
|
192 |
return true; |
193 |
} |
194 |
|
195 |
static inline void gen_goto_tb(DisasContext *s, int n, uint64_t dest) |
196 |
{ |
197 |
TranslationBlock *tb; |
198 |
|
199 |
tb = s->tb; |
200 |
if (use_goto_tb(s, n, dest)) {
|
201 |
tcg_gen_goto_tb(n); |
202 |
gen_a64_set_pc_im(dest); |
203 |
tcg_gen_exit_tb((tcg_target_long)tb + n); |
204 |
s->is_jmp = DISAS_TB_JUMP; |
205 |
} else {
|
206 |
gen_a64_set_pc_im(dest); |
207 |
if (s->singlestep_enabled) {
|
208 |
gen_exception(EXCP_DEBUG); |
209 |
} |
210 |
tcg_gen_exit_tb(0);
|
211 |
s->is_jmp = DISAS_JUMP; |
212 |
} |
213 |
} |
214 |
|
215 |
static void unallocated_encoding(DisasContext *s) |
216 |
{ |
217 |
gen_exception_insn(s, 4, EXCP_UDEF);
|
218 |
} |
219 |
|
220 |
#define unsupported_encoding(s, insn) \
|
221 |
do { \
|
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); \ |
227 |
} while (0); |
228 |
|
229 |
static void init_tmp_a64_array(DisasContext *s) |
230 |
{ |
231 |
#ifdef CONFIG_DEBUG_TCG
|
232 |
int i;
|
233 |
for (i = 0; i < ARRAY_SIZE(s->tmp_a64); i++) { |
234 |
TCGV_UNUSED_I64(s->tmp_a64[i]); |
235 |
} |
236 |
#endif
|
237 |
s->tmp_a64_count = 0;
|
238 |
} |
239 |
|
240 |
static void free_tmp_a64(DisasContext *s) |
241 |
{ |
242 |
int i;
|
243 |
for (i = 0; i < s->tmp_a64_count; i++) { |
244 |
tcg_temp_free_i64(s->tmp_a64[i]); |
245 |
} |
246 |
init_tmp_a64_array(s); |
247 |
} |
248 |
|
249 |
static TCGv_i64 new_tmp_a64(DisasContext *s)
|
250 |
{ |
251 |
assert(s->tmp_a64_count < TMP_A64_MAX); |
252 |
return s->tmp_a64[s->tmp_a64_count++] = tcg_temp_new_i64();
|
253 |
} |
254 |
|
255 |
static TCGv_i64 new_tmp_a64_zero(DisasContext *s)
|
256 |
{ |
257 |
TCGv_i64 t = new_tmp_a64(s); |
258 |
tcg_gen_movi_i64(t, 0);
|
259 |
return t;
|
260 |
} |
261 |
|
262 |
/*
|
263 |
* Register access functions
|
264 |
*
|
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.
|
269 |
*
|
270 |
* B1.2.1 Register mappings
|
271 |
*
|
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.
|
276 |
*/
|
277 |
static TCGv_i64 cpu_reg(DisasContext *s, int reg) |
278 |
{ |
279 |
if (reg == 31) { |
280 |
return new_tmp_a64_zero(s);
|
281 |
} else {
|
282 |
return cpu_X[reg];
|
283 |
} |
284 |
} |
285 |
|
286 |
/* register access for when 31 == SP */
|
287 |
static TCGv_i64 cpu_reg_sp(DisasContext *s, int reg) |
288 |
{ |
289 |
return cpu_X[reg];
|
290 |
} |
291 |
|
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.
|
295 |
*/
|
296 |
static TCGv_i64 read_cpu_reg(DisasContext *s, int reg, int sf) |
297 |
{ |
298 |
TCGv_i64 v = new_tmp_a64(s); |
299 |
if (reg != 31) { |
300 |
if (sf) {
|
301 |
tcg_gen_mov_i64(v, cpu_X[reg]); |
302 |
} else {
|
303 |
tcg_gen_ext32u_i64(v, cpu_X[reg]); |
304 |
} |
305 |
} else {
|
306 |
tcg_gen_movi_i64(v, 0);
|
307 |
} |
308 |
return v;
|
309 |
} |
310 |
|
311 |
static TCGv_i64 read_cpu_reg_sp(DisasContext *s, int reg, int sf) |
312 |
{ |
313 |
TCGv_i64 v = new_tmp_a64(s); |
314 |
if (sf) {
|
315 |
tcg_gen_mov_i64(v, cpu_X[reg]); |
316 |
} else {
|
317 |
tcg_gen_ext32u_i64(v, cpu_X[reg]); |
318 |
} |
319 |
return v;
|
320 |
} |
321 |
|
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.
|
325 |
*/
|
326 |
static inline int vec_reg_offset(int regno, int element, TCGMemOp size) |
327 |
{ |
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.
|
335 |
*/
|
336 |
offs += (16 - ((element + 1) * (1 << size))); |
337 |
offs ^= 8;
|
338 |
#else
|
339 |
offs += element * (1 << size);
|
340 |
#endif
|
341 |
return offs;
|
342 |
} |
343 |
|
344 |
/* Return the offset into CPUARMState of a slice (from
|
345 |
* the least significant end) of FP register Qn (ie
|
346 |
* Dn, Sn, Hn or Bn).
|
347 |
* (Note that this is not the same mapping as for A32; see cpu.h)
|
348 |
*/
|
349 |
static inline int fp_reg_offset(int regno, TCGMemOp size) |
350 |
{ |
351 |
int offs = offsetof(CPUARMState, vfp.regs[regno * 2]); |
352 |
#ifdef HOST_WORDS_BIGENDIAN
|
353 |
offs += (8 - (1 << size)); |
354 |
#endif
|
355 |
return offs;
|
356 |
} |
357 |
|
358 |
/* Offset of the high half of the 128 bit vector Qn */
|
359 |
static inline int fp_reg_hi_offset(int regno) |
360 |
{ |
361 |
return offsetof(CPUARMState, vfp.regs[regno * 2 + 1]); |
362 |
} |
363 |
|
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.
|
369 |
*/
|
370 |
static TCGv_i64 read_fp_dreg(DisasContext *s, int reg) |
371 |
{ |
372 |
TCGv_i64 v = tcg_temp_new_i64(); |
373 |
|
374 |
tcg_gen_ld_i64(v, cpu_env, fp_reg_offset(reg, MO_64)); |
375 |
return v;
|
376 |
} |
377 |
|
378 |
static TCGv_i32 read_fp_sreg(DisasContext *s, int reg) |
379 |
{ |
380 |
TCGv_i32 v = tcg_temp_new_i32(); |
381 |
|
382 |
tcg_gen_ld_i32(v, cpu_env, fp_reg_offset(reg, MO_32)); |
383 |
return v;
|
384 |
} |
385 |
|
386 |
static void write_fp_dreg(DisasContext *s, int reg, TCGv_i64 v) |
387 |
{ |
388 |
TCGv_i64 tcg_zero = tcg_const_i64(0);
|
389 |
|
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); |
393 |
} |
394 |
|
395 |
static void write_fp_sreg(DisasContext *s, int reg, TCGv_i32 v) |
396 |
{ |
397 |
TCGv_i64 tmp = tcg_temp_new_i64(); |
398 |
|
399 |
tcg_gen_extu_i32_i64(tmp, v); |
400 |
write_fp_dreg(s, reg, tmp); |
401 |
tcg_temp_free_i64(tmp); |
402 |
} |
403 |
|
404 |
static TCGv_ptr get_fpstatus_ptr(void) |
405 |
{ |
406 |
TCGv_ptr statusptr = tcg_temp_new_ptr(); |
407 |
int offset;
|
408 |
|
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.
|
412 |
*/
|
413 |
offset = offsetof(CPUARMState, vfp.fp_status); |
414 |
tcg_gen_addi_ptr(statusptr, cpu_env, offset); |
415 |
return statusptr;
|
416 |
} |
417 |
|
418 |
/* Set ZF and NF based on a 64 bit result. This is alas fiddlier
|
419 |
* than the 32 bit equivalent.
|
420 |
*/
|
421 |
static inline void gen_set_NZ64(TCGv_i64 result) |
422 |
{ |
423 |
TCGv_i64 flag = tcg_temp_new_i64(); |
424 |
|
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); |
430 |
} |
431 |
|
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) |
434 |
{ |
435 |
if (sf) {
|
436 |
gen_set_NZ64(result); |
437 |
} else {
|
438 |
tcg_gen_trunc_i64_i32(cpu_ZF, result); |
439 |
tcg_gen_trunc_i64_i32(cpu_NF, result); |
440 |
} |
441 |
tcg_gen_movi_i32(cpu_CF, 0);
|
442 |
tcg_gen_movi_i32(cpu_VF, 0);
|
443 |
} |
444 |
|
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) |
447 |
{ |
448 |
if (sf) {
|
449 |
TCGv_i64 result, flag, tmp; |
450 |
result = tcg_temp_new_i64(); |
451 |
flag = tcg_temp_new_i64(); |
452 |
tmp = tcg_temp_new_i64(); |
453 |
|
454 |
tcg_gen_movi_i64(tmp, 0);
|
455 |
tcg_gen_add2_i64(result, flag, t0, tmp, t1, tmp); |
456 |
|
457 |
tcg_gen_trunc_i64_i32(cpu_CF, flag); |
458 |
|
459 |
gen_set_NZ64(result); |
460 |
|
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); |
467 |
|
468 |
tcg_gen_mov_i64(dest, result); |
469 |
tcg_temp_free_i64(result); |
470 |
tcg_temp_free_i64(flag); |
471 |
} else {
|
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(); |
476 |
|
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); |
486 |
|
487 |
tcg_temp_free_i32(tmp); |
488 |
tcg_temp_free_i32(t0_32); |
489 |
tcg_temp_free_i32(t1_32); |
490 |
} |
491 |
} |
492 |
|
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) |
495 |
{ |
496 |
if (sf) {
|
497 |
/* 64 bit arithmetic */
|
498 |
TCGv_i64 result, flag, tmp; |
499 |
|
500 |
result = tcg_temp_new_i64(); |
501 |
flag = tcg_temp_new_i64(); |
502 |
tcg_gen_sub_i64(result, t0, t1); |
503 |
|
504 |
gen_set_NZ64(result); |
505 |
|
506 |
tcg_gen_setcond_i64(TCG_COND_GEU, flag, t0, t1); |
507 |
tcg_gen_trunc_i64_i32(cpu_CF, flag); |
508 |
|
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); |
519 |
} else {
|
520 |
/* 32 bit arithmetic */
|
521 |
TCGv_i32 t0_32 = tcg_temp_new_i32(); |
522 |
TCGv_i32 t1_32 = tcg_temp_new_i32(); |
523 |
TCGv_i32 tmp; |
524 |
|
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); |
538 |
} |
539 |
} |
540 |
|
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) |
543 |
{ |
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); |
549 |
|
550 |
if (!sf) {
|
551 |
tcg_gen_ext32u_i64(dest, dest); |
552 |
} |
553 |
} |
554 |
|
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) |
557 |
{ |
558 |
if (sf) {
|
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);
|
564 |
|
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); |
570 |
|
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); |
576 |
|
577 |
tcg_gen_mov_i64(dest, result); |
578 |
|
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); |
583 |
} else {
|
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);
|
588 |
|
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); |
593 |
|
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); |
599 |
|
600 |
tcg_temp_free_i32(tmp); |
601 |
tcg_temp_free_i32(t1_32); |
602 |
tcg_temp_free_i32(t0_32); |
603 |
} |
604 |
} |
605 |
|
606 |
/*
|
607 |
* Load/Store generators
|
608 |
*/
|
609 |
|
610 |
/*
|
611 |
* Store from GPR register to memory
|
612 |
*/
|
613 |
static void do_gpr_st(DisasContext *s, TCGv_i64 source, |
614 |
TCGv_i64 tcg_addr, int size)
|
615 |
{ |
616 |
g_assert(size <= 3);
|
617 |
tcg_gen_qemu_st_i64(source, tcg_addr, get_mem_index(s), MO_TE + size); |
618 |
} |
619 |
|
620 |
/*
|
621 |
* Load from memory to GPR register
|
622 |
*/
|
623 |
static void do_gpr_ld(DisasContext *s, TCGv_i64 dest, TCGv_i64 tcg_addr, |
624 |
int size, bool is_signed, bool extend) |
625 |
{ |
626 |
TCGMemOp memop = MO_TE + size; |
627 |
|
628 |
g_assert(size <= 3);
|
629 |
|
630 |
if (is_signed) {
|
631 |
memop += MO_SIGN; |
632 |
} |
633 |
|
634 |
tcg_gen_qemu_ld_i64(dest, tcg_addr, get_mem_index(s), memop); |
635 |
|
636 |
if (extend && is_signed) {
|
637 |
g_assert(size < 3);
|
638 |
tcg_gen_ext32u_i64(dest, dest); |
639 |
} |
640 |
} |
641 |
|
642 |
/*
|
643 |
* Store from FP register to memory
|
644 |
*/
|
645 |
static void do_fp_st(DisasContext *s, int srcidx, TCGv_i64 tcg_addr, int size) |
646 |
{ |
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)); |
650 |
if (size < 4) { |
651 |
tcg_gen_qemu_st_i64(tmp, tcg_addr, get_mem_index(s), MO_TE + size); |
652 |
} else {
|
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); |
660 |
} |
661 |
|
662 |
tcg_temp_free_i64(tmp); |
663 |
} |
664 |
|
665 |
/*
|
666 |
* Load from memory to FP register
|
667 |
*/
|
668 |
static void do_fp_ld(DisasContext *s, int destidx, TCGv_i64 tcg_addr, int size) |
669 |
{ |
670 |
/* This always zero-extends and writes to a full 128 bit wide vector */
|
671 |
TCGv_i64 tmplo = tcg_temp_new_i64(); |
672 |
TCGv_i64 tmphi; |
673 |
|
674 |
if (size < 4) { |
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); |
678 |
} else {
|
679 |
TCGv_i64 tcg_hiaddr; |
680 |
tmphi = tcg_temp_new_i64(); |
681 |
tcg_hiaddr = tcg_temp_new_i64(); |
682 |
|
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); |
687 |
} |
688 |
|
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)); |
691 |
|
692 |
tcg_temp_free_i64(tmplo); |
693 |
tcg_temp_free_i64(tmphi); |
694 |
} |
695 |
|
696 |
/*
|
697 |
* Vector load/store helpers.
|
698 |
*
|
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.
|
703 |
*
|
704 |
* The _i32 versions are useful when operating on 32 bit quantities
|
705 |
* (eg for floating point single or using Neon helper functions).
|
706 |
*/
|
707 |
|
708 |
/* Get value of an element within a vector register */
|
709 |
static void read_vec_element(DisasContext *s, TCGv_i64 tcg_dest, int srcidx, |
710 |
int element, TCGMemOp memop)
|
711 |
{ |
712 |
int vect_off = vec_reg_offset(srcidx, element, memop & MO_SIZE);
|
713 |
switch (memop) {
|
714 |
case MO_8:
|
715 |
tcg_gen_ld8u_i64(tcg_dest, cpu_env, vect_off); |
716 |
break;
|
717 |
case MO_16:
|
718 |
tcg_gen_ld16u_i64(tcg_dest, cpu_env, vect_off); |
719 |
break;
|
720 |
case MO_32:
|
721 |
tcg_gen_ld32u_i64(tcg_dest, cpu_env, vect_off); |
722 |
break;
|
723 |
case MO_8|MO_SIGN:
|
724 |
tcg_gen_ld8s_i64(tcg_dest, cpu_env, vect_off); |
725 |
break;
|
726 |
case MO_16|MO_SIGN:
|
727 |
tcg_gen_ld16s_i64(tcg_dest, cpu_env, vect_off); |
728 |
break;
|
729 |
case MO_32|MO_SIGN:
|
730 |
tcg_gen_ld32s_i64(tcg_dest, cpu_env, vect_off); |
731 |
break;
|
732 |
case MO_64:
|
733 |
case MO_64|MO_SIGN:
|
734 |
tcg_gen_ld_i64(tcg_dest, cpu_env, vect_off); |
735 |
break;
|
736 |
default:
|
737 |
g_assert_not_reached(); |
738 |
} |
739 |
} |
740 |
|
741 |
static void read_vec_element_i32(DisasContext *s, TCGv_i32 tcg_dest, int srcidx, |
742 |
int element, TCGMemOp memop)
|
743 |
{ |
744 |
int vect_off = vec_reg_offset(srcidx, element, memop & MO_SIZE);
|
745 |
switch (memop) {
|
746 |
case MO_8:
|
747 |
tcg_gen_ld8u_i32(tcg_dest, cpu_env, vect_off); |
748 |
break;
|
749 |
case MO_16:
|
750 |
tcg_gen_ld16u_i32(tcg_dest, cpu_env, vect_off); |
751 |
break;
|
752 |
case MO_8|MO_SIGN:
|
753 |
tcg_gen_ld8s_i32(tcg_dest, cpu_env, vect_off); |
754 |
break;
|
755 |
case MO_16|MO_SIGN:
|
756 |
tcg_gen_ld16s_i32(tcg_dest, cpu_env, vect_off); |
757 |
break;
|
758 |
case MO_32:
|
759 |
case MO_32|MO_SIGN:
|
760 |
tcg_gen_ld_i32(tcg_dest, cpu_env, vect_off); |
761 |
break;
|
762 |
default:
|
763 |
g_assert_not_reached(); |
764 |
} |
765 |
} |
766 |
|
767 |
/* Set value of an element within a vector register */
|
768 |
static void write_vec_element(DisasContext *s, TCGv_i64 tcg_src, int destidx, |
769 |
int element, TCGMemOp memop)
|
770 |
{ |
771 |
int vect_off = vec_reg_offset(destidx, element, memop & MO_SIZE);
|
772 |
switch (memop) {
|
773 |
case MO_8:
|
774 |
tcg_gen_st8_i64(tcg_src, cpu_env, vect_off); |
775 |
break;
|
776 |
case MO_16:
|
777 |
tcg_gen_st16_i64(tcg_src, cpu_env, vect_off); |
778 |
break;
|
779 |
case MO_32:
|
780 |
tcg_gen_st32_i64(tcg_src, cpu_env, vect_off); |
781 |
break;
|
782 |
case MO_64:
|
783 |
tcg_gen_st_i64(tcg_src, cpu_env, vect_off); |
784 |
break;
|
785 |
default:
|
786 |
g_assert_not_reached(); |
787 |
} |
788 |
} |
789 |
|
790 |
/* Clear the high 64 bits of a 128 bit vector (in general non-quad
|
791 |
* vector ops all need to do this).
|
792 |
*/
|
793 |
static void clear_vec_high(DisasContext *s, int rd) |
794 |
{ |
795 |
TCGv_i64 tcg_zero = tcg_const_i64(0);
|
796 |
|
797 |
write_vec_element(s, tcg_zero, rd, 1, MO_64);
|
798 |
tcg_temp_free_i64(tcg_zero); |
799 |
} |
800 |
|
801 |
/* Store from vector register to memory */
|
802 |
static void do_vec_st(DisasContext *s, int srcidx, int element, |
803 |
TCGv_i64 tcg_addr, int size)
|
804 |
{ |
805 |
TCGMemOp memop = MO_TE + size; |
806 |
TCGv_i64 tcg_tmp = tcg_temp_new_i64(); |
807 |
|
808 |
read_vec_element(s, tcg_tmp, srcidx, element, size); |
809 |
tcg_gen_qemu_st_i64(tcg_tmp, tcg_addr, get_mem_index(s), memop); |
810 |
|
811 |
tcg_temp_free_i64(tcg_tmp); |
812 |
} |
813 |
|
814 |
/* Load from memory to vector register */
|
815 |
static void do_vec_ld(DisasContext *s, int destidx, int element, |
816 |
TCGv_i64 tcg_addr, int size)
|
817 |
{ |
818 |
TCGMemOp memop = MO_TE + size; |
819 |
TCGv_i64 tcg_tmp = tcg_temp_new_i64(); |
820 |
|
821 |
tcg_gen_qemu_ld_i64(tcg_tmp, tcg_addr, get_mem_index(s), memop); |
822 |
write_vec_element(s, tcg_tmp, destidx, element, size); |
823 |
|
824 |
tcg_temp_free_i64(tcg_tmp); |
825 |
} |
826 |
|
827 |
/*
|
828 |
* This utility function is for doing register extension with an
|
829 |
* optional shift. You will likely want to pass a temporary for the
|
830 |
* destination register. See DecodeRegExtend() in the ARM ARM.
|
831 |
*/
|
832 |
static void ext_and_shift_reg(TCGv_i64 tcg_out, TCGv_i64 tcg_in, |
833 |
int option, unsigned int shift) |
834 |
{ |
835 |
int extsize = extract32(option, 0, 2); |
836 |
bool is_signed = extract32(option, 2, 1); |
837 |
|
838 |
if (is_signed) {
|
839 |
switch (extsize) {
|
840 |
case 0: |
841 |
tcg_gen_ext8s_i64(tcg_out, tcg_in); |
842 |
break;
|
843 |
case 1: |
844 |
tcg_gen_ext16s_i64(tcg_out, tcg_in); |
845 |
break;
|
846 |
case 2: |
847 |
tcg_gen_ext32s_i64(tcg_out, tcg_in); |
848 |
break;
|
849 |
case 3: |
850 |
tcg_gen_mov_i64(tcg_out, tcg_in); |
851 |
break;
|
852 |
} |
853 |
} else {
|
854 |
switch (extsize) {
|
855 |
case 0: |
856 |
tcg_gen_ext8u_i64(tcg_out, tcg_in); |
857 |
break;
|
858 |
case 1: |
859 |
tcg_gen_ext16u_i64(tcg_out, tcg_in); |
860 |
break;
|
861 |
case 2: |
862 |
tcg_gen_ext32u_i64(tcg_out, tcg_in); |
863 |
break;
|
864 |
case 3: |
865 |
tcg_gen_mov_i64(tcg_out, tcg_in); |
866 |
break;
|
867 |
} |
868 |
} |
869 |
|
870 |
if (shift) {
|
871 |
tcg_gen_shli_i64(tcg_out, tcg_out, shift); |
872 |
} |
873 |
} |
874 |
|
875 |
static inline void gen_check_sp_alignment(DisasContext *s) |
876 |
{ |
877 |
/* The AArch64 architecture mandates that (if enabled via PSTATE
|
878 |
* or SCTLR bits) there is a check that SP is 16-aligned on every
|
879 |
* SP-relative load or store (with an exception generated if it is not).
|
880 |
* In line with general QEMU practice regarding misaligned accesses,
|
881 |
* we omit these checks for the sake of guest program performance.
|
882 |
* This function is provided as a hook so we can more easily add these
|
883 |
* checks in future (possibly as a "favour catching guest program bugs
|
884 |
* over speed" user selectable option).
|
885 |
*/
|
886 |
} |
887 |
|
888 |
/*
|
889 |
* This provides a simple table based table lookup decoder. It is
|
890 |
* intended to be used when the relevant bits for decode are too
|
891 |
* awkwardly placed and switch/if based logic would be confusing and
|
892 |
* deeply nested. Since it's a linear search through the table, tables
|
893 |
* should be kept small.
|
894 |
*
|
895 |
* It returns the first handler where insn & mask == pattern, or
|
896 |
* NULL if there is no match.
|
897 |
* The table is terminated by an empty mask (i.e. 0)
|
898 |
*/
|
899 |
static inline AArch64DecodeFn *lookup_disas_fn(const AArch64DecodeTable *table, |
900 |
uint32_t insn) |
901 |
{ |
902 |
const AArch64DecodeTable *tptr = table;
|
903 |
|
904 |
while (tptr->mask) {
|
905 |
if ((insn & tptr->mask) == tptr->pattern) {
|
906 |
return tptr->disas_fn;
|
907 |
} |
908 |
tptr++; |
909 |
} |
910 |
return NULL; |
911 |
} |
912 |
|
913 |
/*
|
914 |
* the instruction disassembly implemented here matches
|
915 |
* the instruction encoding classifications in chapter 3 (C3)
|
916 |
* of the ARM Architecture Reference Manual (DDI0487A_a)
|
917 |
*/
|
918 |
|
919 |
/* C3.2.7 Unconditional branch (immediate)
|
920 |
* 31 30 26 25 0
|
921 |
* +----+-----------+-------------------------------------+
|
922 |
* | op | 0 0 1 0 1 | imm26 |
|
923 |
* +----+-----------+-------------------------------------+
|
924 |
*/
|
925 |
static void disas_uncond_b_imm(DisasContext *s, uint32_t insn) |
926 |
{ |
927 |
uint64_t addr = s->pc + sextract32(insn, 0, 26) * 4 - 4; |
928 |
|
929 |
if (insn & (1 << 31)) { |
930 |
/* C5.6.26 BL Branch with link */
|
931 |
tcg_gen_movi_i64(cpu_reg(s, 30), s->pc);
|
932 |
} |
933 |
|
934 |
/* C5.6.20 B Branch / C5.6.26 BL Branch with link */
|
935 |
gen_goto_tb(s, 0, addr);
|
936 |
} |
937 |
|
938 |
/* C3.2.1 Compare & branch (immediate)
|
939 |
* 31 30 25 24 23 5 4 0
|
940 |
* +----+-------------+----+---------------------+--------+
|
941 |
* | sf | 0 1 1 0 1 0 | op | imm19 | Rt |
|
942 |
* +----+-------------+----+---------------------+--------+
|
943 |
*/
|
944 |
static void disas_comp_b_imm(DisasContext *s, uint32_t insn) |
945 |
{ |
946 |
unsigned int sf, op, rt; |
947 |
uint64_t addr; |
948 |
int label_match;
|
949 |
TCGv_i64 tcg_cmp; |
950 |
|
951 |
sf = extract32(insn, 31, 1); |
952 |
op = extract32(insn, 24, 1); /* 0: CBZ; 1: CBNZ */ |
953 |
rt = extract32(insn, 0, 5); |
954 |
addr = s->pc + sextract32(insn, 5, 19) * 4 - 4; |
955 |
|
956 |
tcg_cmp = read_cpu_reg(s, rt, sf); |
957 |
label_match = gen_new_label(); |
958 |
|
959 |
tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, |
960 |
tcg_cmp, 0, label_match);
|
961 |
|
962 |
gen_goto_tb(s, 0, s->pc);
|
963 |
gen_set_label(label_match); |
964 |
gen_goto_tb(s, 1, addr);
|
965 |
} |
966 |
|
967 |
/* C3.2.5 Test & branch (immediate)
|
968 |
* 31 30 25 24 23 19 18 5 4 0
|
969 |
* +----+-------------+----+-------+-------------+------+
|
970 |
* | b5 | 0 1 1 0 1 1 | op | b40 | imm14 | Rt |
|
971 |
* +----+-------------+----+-------+-------------+------+
|
972 |
*/
|
973 |
static void disas_test_b_imm(DisasContext *s, uint32_t insn) |
974 |
{ |
975 |
unsigned int bit_pos, op, rt; |
976 |
uint64_t addr; |
977 |
int label_match;
|
978 |
TCGv_i64 tcg_cmp; |
979 |
|
980 |
bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5); |
981 |
op = extract32(insn, 24, 1); /* 0: TBZ; 1: TBNZ */ |
982 |
addr = s->pc + sextract32(insn, 5, 14) * 4 - 4; |
983 |
rt = extract32(insn, 0, 5); |
984 |
|
985 |
tcg_cmp = tcg_temp_new_i64(); |
986 |
tcg_gen_andi_i64(tcg_cmp, cpu_reg(s, rt), (1ULL << bit_pos));
|
987 |
label_match = gen_new_label(); |
988 |
tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ, |
989 |
tcg_cmp, 0, label_match);
|
990 |
tcg_temp_free_i64(tcg_cmp); |
991 |
gen_goto_tb(s, 0, s->pc);
|
992 |
gen_set_label(label_match); |
993 |
gen_goto_tb(s, 1, addr);
|
994 |
} |
995 |
|
996 |
/* C3.2.2 / C5.6.19 Conditional branch (immediate)
|
997 |
* 31 25 24 23 5 4 3 0
|
998 |
* +---------------+----+---------------------+----+------+
|
999 |
* | 0 1 0 1 0 1 0 | o1 | imm19 | o0 | cond |
|
1000 |
* +---------------+----+---------------------+----+------+
|
1001 |
*/
|
1002 |
static void disas_cond_b_imm(DisasContext *s, uint32_t insn) |
1003 |
{ |
1004 |
unsigned int cond; |
1005 |
uint64_t addr; |
1006 |
|
1007 |
if ((insn & (1 << 4)) || (insn & (1 << 24))) { |
1008 |
unallocated_encoding(s); |
1009 |
return;
|
1010 |
} |
1011 |
addr = s->pc + sextract32(insn, 5, 19) * 4 - 4; |
1012 |
cond = extract32(insn, 0, 4); |
1013 |
|
1014 |
if (cond < 0x0e) { |
1015 |
/* genuinely conditional branches */
|
1016 |
int label_match = gen_new_label();
|
1017 |
arm_gen_test_cc(cond, label_match); |
1018 |
gen_goto_tb(s, 0, s->pc);
|
1019 |
gen_set_label(label_match); |
1020 |
gen_goto_tb(s, 1, addr);
|
1021 |
} else {
|
1022 |
/* 0xe and 0xf are both "always" conditions */
|
1023 |
gen_goto_tb(s, 0, addr);
|
1024 |
} |
1025 |
} |
1026 |
|
1027 |
/* C5.6.68 HINT */
|
1028 |
static void handle_hint(DisasContext *s, uint32_t insn, |
1029 |
unsigned int op1, unsigned int op2, unsigned int crm) |
1030 |
{ |
1031 |
unsigned int selector = crm << 3 | op2; |
1032 |
|
1033 |
if (op1 != 3) { |
1034 |
unallocated_encoding(s); |
1035 |
return;
|
1036 |
} |
1037 |
|
1038 |
switch (selector) {
|
1039 |
case 0: /* NOP */ |
1040 |
return;
|
1041 |
case 1: /* YIELD */ |
1042 |
case 2: /* WFE */ |
1043 |
case 3: /* WFI */ |
1044 |
case 4: /* SEV */ |
1045 |
case 5: /* SEVL */ |
1046 |
/* we treat all as NOP at least for now */
|
1047 |
return;
|
1048 |
default:
|
1049 |
/* default specified as NOP equivalent */
|
1050 |
return;
|
1051 |
} |
1052 |
} |
1053 |
|
1054 |
static void gen_clrex(DisasContext *s, uint32_t insn) |
1055 |
{ |
1056 |
tcg_gen_movi_i64(cpu_exclusive_addr, -1);
|
1057 |
} |
1058 |
|
1059 |
/* CLREX, DSB, DMB, ISB */
|
1060 |
static void handle_sync(DisasContext *s, uint32_t insn, |
1061 |
unsigned int op1, unsigned int op2, unsigned int crm) |
1062 |
{ |
1063 |
if (op1 != 3) { |
1064 |
unallocated_encoding(s); |
1065 |
return;
|
1066 |
} |
1067 |
|
1068 |
switch (op2) {
|
1069 |
case 2: /* CLREX */ |
1070 |
gen_clrex(s, insn); |
1071 |
return;
|
1072 |
case 4: /* DSB */ |
1073 |
case 5: /* DMB */ |
1074 |
case 6: /* ISB */ |
1075 |
/* We don't emulate caches so barriers are no-ops */
|
1076 |
return;
|
1077 |
default:
|
1078 |
unallocated_encoding(s); |
1079 |
return;
|
1080 |
} |
1081 |
} |
1082 |
|
1083 |
/* C5.6.130 MSR (immediate) - move immediate to processor state field */
|
1084 |
static void handle_msr_i(DisasContext *s, uint32_t insn, |
1085 |
unsigned int op1, unsigned int op2, unsigned int crm) |
1086 |
{ |
1087 |
unsupported_encoding(s, insn); |
1088 |
} |
1089 |
|
1090 |
static void gen_get_nzcv(TCGv_i64 tcg_rt) |
1091 |
{ |
1092 |
TCGv_i32 tmp = tcg_temp_new_i32(); |
1093 |
TCGv_i32 nzcv = tcg_temp_new_i32(); |
1094 |
|
1095 |
/* build bit 31, N */
|
1096 |
tcg_gen_andi_i32(nzcv, cpu_NF, (1 << 31)); |
1097 |
/* build bit 30, Z */
|
1098 |
tcg_gen_setcondi_i32(TCG_COND_EQ, tmp, cpu_ZF, 0);
|
1099 |
tcg_gen_deposit_i32(nzcv, nzcv, tmp, 30, 1); |
1100 |
/* build bit 29, C */
|
1101 |
tcg_gen_deposit_i32(nzcv, nzcv, cpu_CF, 29, 1); |
1102 |
/* build bit 28, V */
|
1103 |
tcg_gen_shri_i32(tmp, cpu_VF, 31);
|
1104 |
tcg_gen_deposit_i32(nzcv, nzcv, tmp, 28, 1); |
1105 |
/* generate result */
|
1106 |
tcg_gen_extu_i32_i64(tcg_rt, nzcv); |
1107 |
|
1108 |
tcg_temp_free_i32(nzcv); |
1109 |
tcg_temp_free_i32(tmp); |
1110 |
} |
1111 |
|
1112 |
static void gen_set_nzcv(TCGv_i64 tcg_rt) |
1113 |
|
1114 |
{ |
1115 |
TCGv_i32 nzcv = tcg_temp_new_i32(); |
1116 |
|
1117 |
/* take NZCV from R[t] */
|
1118 |
tcg_gen_trunc_i64_i32(nzcv, tcg_rt); |
1119 |
|
1120 |
/* bit 31, N */
|
1121 |
tcg_gen_andi_i32(cpu_NF, nzcv, (1 << 31)); |
1122 |
/* bit 30, Z */
|
1123 |
tcg_gen_andi_i32(cpu_ZF, nzcv, (1 << 30)); |
1124 |
tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_ZF, cpu_ZF, 0);
|
1125 |
/* bit 29, C */
|
1126 |
tcg_gen_andi_i32(cpu_CF, nzcv, (1 << 29)); |
1127 |
tcg_gen_shri_i32(cpu_CF, cpu_CF, 29);
|
1128 |
/* bit 28, V */
|
1129 |
tcg_gen_andi_i32(cpu_VF, nzcv, (1 << 28)); |
1130 |
tcg_gen_shli_i32(cpu_VF, cpu_VF, 3);
|
1131 |
tcg_temp_free_i32(nzcv); |
1132 |
} |
1133 |
|
1134 |
/* C5.6.129 MRS - move from system register
|
1135 |
* C5.6.131 MSR (register) - move to system register
|
1136 |
* C5.6.204 SYS
|
1137 |
* C5.6.205 SYSL
|
1138 |
* These are all essentially the same insn in 'read' and 'write'
|
1139 |
* versions, with varying op0 fields.
|
1140 |
*/
|
1141 |
static void handle_sys(DisasContext *s, uint32_t insn, bool isread, |
1142 |
unsigned int op0, unsigned int op1, unsigned int op2, |
1143 |
unsigned int crn, unsigned int crm, unsigned int rt) |
1144 |
{ |
1145 |
const ARMCPRegInfo *ri;
|
1146 |
TCGv_i64 tcg_rt; |
1147 |
|
1148 |
ri = get_arm_cp_reginfo(s->cp_regs, |
1149 |
ENCODE_AA64_CP_REG(CP_REG_ARM64_SYSREG_CP, |
1150 |
crn, crm, op0, op1, op2)); |
1151 |
|
1152 |
if (!ri) {
|
1153 |
/* Unknown register */
|
1154 |
unallocated_encoding(s); |
1155 |
return;
|
1156 |
} |
1157 |
|
1158 |
/* Check access permissions */
|
1159 |
if (!cp_access_ok(s->current_pl, ri, isread)) {
|
1160 |
unallocated_encoding(s); |
1161 |
return;
|
1162 |
} |
1163 |
|
1164 |
/* Handle special cases first */
|
1165 |
switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) {
|
1166 |
case ARM_CP_NOP:
|
1167 |
return;
|
1168 |
case ARM_CP_NZCV:
|
1169 |
tcg_rt = cpu_reg(s, rt); |
1170 |
if (isread) {
|
1171 |
gen_get_nzcv(tcg_rt); |
1172 |
} else {
|
1173 |
gen_set_nzcv(tcg_rt); |
1174 |
} |
1175 |
return;
|
1176 |
default:
|
1177 |
break;
|
1178 |
} |
1179 |
|
1180 |
if (use_icount && (ri->type & ARM_CP_IO)) {
|
1181 |
gen_io_start(); |
1182 |
} |
1183 |
|
1184 |
tcg_rt = cpu_reg(s, rt); |
1185 |
|
1186 |
if (isread) {
|
1187 |
if (ri->type & ARM_CP_CONST) {
|
1188 |
tcg_gen_movi_i64(tcg_rt, ri->resetvalue); |
1189 |
} else if (ri->readfn) { |
1190 |
TCGv_ptr tmpptr; |
1191 |
gen_a64_set_pc_im(s->pc - 4);
|
1192 |
tmpptr = tcg_const_ptr(ri); |
1193 |
gen_helper_get_cp_reg64(tcg_rt, cpu_env, tmpptr); |
1194 |
tcg_temp_free_ptr(tmpptr); |
1195 |
} else {
|
1196 |
tcg_gen_ld_i64(tcg_rt, cpu_env, ri->fieldoffset); |
1197 |
} |
1198 |
} else {
|
1199 |
if (ri->type & ARM_CP_CONST) {
|
1200 |
/* If not forbidden by access permissions, treat as WI */
|
1201 |
return;
|
1202 |
} else if (ri->writefn) { |
1203 |
TCGv_ptr tmpptr; |
1204 |
gen_a64_set_pc_im(s->pc - 4);
|
1205 |
tmpptr = tcg_const_ptr(ri); |
1206 |
gen_helper_set_cp_reg64(cpu_env, tmpptr, tcg_rt); |
1207 |
tcg_temp_free_ptr(tmpptr); |
1208 |
} else {
|
1209 |
tcg_gen_st_i64(tcg_rt, cpu_env, ri->fieldoffset); |
1210 |
} |
1211 |
} |
1212 |
|
1213 |
if (use_icount && (ri->type & ARM_CP_IO)) {
|
1214 |
/* I/O operations must end the TB here (whether read or write) */
|
1215 |
gen_io_end(); |
1216 |
s->is_jmp = DISAS_UPDATE; |
1217 |
} else if (!isread && !(ri->type & ARM_CP_SUPPRESS_TB_END)) { |
1218 |
/* We default to ending the TB on a coprocessor register write,
|
1219 |
* but allow this to be suppressed by the register definition
|
1220 |
* (usually only necessary to work around guest bugs).
|
1221 |
*/
|
1222 |
s->is_jmp = DISAS_UPDATE; |
1223 |
} |
1224 |
} |
1225 |
|
1226 |
/* C3.2.4 System
|
1227 |
* 31 22 21 20 19 18 16 15 12 11 8 7 5 4 0
|
1228 |
* +---------------------+---+-----+-----+-------+-------+-----+------+
|
1229 |
* | 1 1 0 1 0 1 0 1 0 0 | L | op0 | op1 | CRn | CRm | op2 | Rt |
|
1230 |
* +---------------------+---+-----+-----+-------+-------+-----+------+
|
1231 |
*/
|
1232 |
static void disas_system(DisasContext *s, uint32_t insn) |
1233 |
{ |
1234 |
unsigned int l, op0, op1, crn, crm, op2, rt; |
1235 |
l = extract32(insn, 21, 1); |
1236 |
op0 = extract32(insn, 19, 2); |
1237 |
op1 = extract32(insn, 16, 3); |
1238 |
crn = extract32(insn, 12, 4); |
1239 |
crm = extract32(insn, 8, 4); |
1240 |
op2 = extract32(insn, 5, 3); |
1241 |
rt = extract32(insn, 0, 5); |
1242 |
|
1243 |
if (op0 == 0) { |
1244 |
if (l || rt != 31) { |
1245 |
unallocated_encoding(s); |
1246 |
return;
|
1247 |
} |
1248 |
switch (crn) {
|
1249 |
case 2: /* C5.6.68 HINT */ |
1250 |
handle_hint(s, insn, op1, op2, crm); |
1251 |
break;
|
1252 |
case 3: /* CLREX, DSB, DMB, ISB */ |
1253 |
handle_sync(s, insn, op1, op2, crm); |
1254 |
break;
|
1255 |
case 4: /* C5.6.130 MSR (immediate) */ |
1256 |
handle_msr_i(s, insn, op1, op2, crm); |
1257 |
break;
|
1258 |
default:
|
1259 |
unallocated_encoding(s); |
1260 |
break;
|
1261 |
} |
1262 |
return;
|
1263 |
} |
1264 |
handle_sys(s, insn, l, op0, op1, op2, crn, crm, rt); |
1265 |
} |
1266 |
|
1267 |
/* C3.2.3 Exception generation
|
1268 |
*
|
1269 |
* 31 24 23 21 20 5 4 2 1 0
|
1270 |
* +-----------------+-----+------------------------+-----+----+
|
1271 |
* | 1 1 0 1 0 1 0 0 | opc | imm16 | op2 | LL |
|
1272 |
* +-----------------------+------------------------+----------+
|
1273 |
*/
|
1274 |
static void disas_exc(DisasContext *s, uint32_t insn) |
1275 |
{ |
1276 |
int opc = extract32(insn, 21, 3); |
1277 |
int op2_ll = extract32(insn, 0, 5); |
1278 |
|
1279 |
switch (opc) {
|
1280 |
case 0: |
1281 |
/* SVC, HVC, SMC; since we don't support the Virtualization
|
1282 |
* or TrustZone extensions these all UNDEF except SVC.
|
1283 |
*/
|
1284 |
if (op2_ll != 1) { |
1285 |
unallocated_encoding(s); |
1286 |
break;
|
1287 |
} |
1288 |
gen_exception_insn(s, 0, EXCP_SWI);
|
1289 |
break;
|
1290 |
case 1: |
1291 |
if (op2_ll != 0) { |
1292 |
unallocated_encoding(s); |
1293 |
break;
|
1294 |
} |
1295 |
/* BRK */
|
1296 |
gen_exception_insn(s, 0, EXCP_BKPT);
|
1297 |
break;
|
1298 |
case 2: |
1299 |
if (op2_ll != 0) { |
1300 |
unallocated_encoding(s); |
1301 |
break;
|
1302 |
} |
1303 |
/* HLT */
|
1304 |
unsupported_encoding(s, insn); |
1305 |
break;
|
1306 |
case 5: |
1307 |
if (op2_ll < 1 || op2_ll > 3) { |
1308 |
unallocated_encoding(s); |
1309 |
break;
|
1310 |
} |
1311 |
/* DCPS1, DCPS2, DCPS3 */
|
1312 |
unsupported_encoding(s, insn); |
1313 |
break;
|
1314 |
default:
|
1315 |
unallocated_encoding(s); |
1316 |
break;
|
1317 |
} |
1318 |
} |
1319 |
|
1320 |
/* C3.2.7 Unconditional branch (register)
|
1321 |
* 31 25 24 21 20 16 15 10 9 5 4 0
|
1322 |
* +---------------+-------+-------+-------+------+-------+
|
1323 |
* | 1 1 0 1 0 1 1 | opc | op2 | op3 | Rn | op4 |
|
1324 |
* +---------------+-------+-------+-------+------+-------+
|
1325 |
*/
|
1326 |
static void disas_uncond_b_reg(DisasContext *s, uint32_t insn) |
1327 |
{ |
1328 |
unsigned int opc, op2, op3, rn, op4; |
1329 |
|
1330 |
opc = extract32(insn, 21, 4); |
1331 |
op2 = extract32(insn, 16, 5); |
1332 |
op3 = extract32(insn, 10, 6); |
1333 |
rn = extract32(insn, 5, 5); |
1334 |
op4 = extract32(insn, 0, 5); |
1335 |
|
1336 |
if (op4 != 0x0 || op3 != 0x0 || op2 != 0x1f) { |
1337 |
unallocated_encoding(s); |
1338 |
return;
|
1339 |
} |
1340 |
|
1341 |
switch (opc) {
|
1342 |
case 0: /* BR */ |
1343 |
case 2: /* RET */ |
1344 |
break;
|
1345 |
case 1: /* BLR */ |
1346 |
tcg_gen_movi_i64(cpu_reg(s, 30), s->pc);
|
1347 |
break;
|
1348 |
case 4: /* ERET */ |
1349 |
case 5: /* DRPS */ |
1350 |
if (rn != 0x1f) { |
1351 |
unallocated_encoding(s); |
1352 |
} else {
|
1353 |
unsupported_encoding(s, insn); |
1354 |
} |
1355 |
return;
|
1356 |
default:
|
1357 |
unallocated_encoding(s); |
1358 |
return;
|
1359 |
} |
1360 |
|
1361 |
tcg_gen_mov_i64(cpu_pc, cpu_reg(s, rn)); |
1362 |
s->is_jmp = DISAS_JUMP; |
1363 |
} |
1364 |
|
1365 |
/* C3.2 Branches, exception generating and system instructions */
|
1366 |
static void disas_b_exc_sys(DisasContext *s, uint32_t insn) |
1367 |
{ |
1368 |
switch (extract32(insn, 25, 7)) { |
1369 |
case 0x0a: case 0x0b: |
1370 |
case 0x4a: case 0x4b: /* Unconditional branch (immediate) */ |
1371 |
disas_uncond_b_imm(s, insn); |
1372 |
break;
|
1373 |
case 0x1a: case 0x5a: /* Compare & branch (immediate) */ |
1374 |
disas_comp_b_imm(s, insn); |
1375 |
break;
|
1376 |
case 0x1b: case 0x5b: /* Test & branch (immediate) */ |
1377 |
disas_test_b_imm(s, insn); |
1378 |
break;
|
1379 |
case 0x2a: /* Conditional branch (immediate) */ |
1380 |
disas_cond_b_imm(s, insn); |
1381 |
break;
|
1382 |
case 0x6a: /* Exception generation / System */ |
1383 |
if (insn & (1 << 24)) { |
1384 |
disas_system(s, insn); |
1385 |
} else {
|
1386 |
disas_exc(s, insn); |
1387 |
} |
1388 |
break;
|
1389 |
case 0x6b: /* Unconditional branch (register) */ |
1390 |
disas_uncond_b_reg(s, insn); |
1391 |
break;
|
1392 |
default:
|
1393 |
unallocated_encoding(s); |
1394 |
break;
|
1395 |
} |
1396 |
} |
1397 |
|
1398 |
/*
|
1399 |
* Load/Store exclusive instructions are implemented by remembering
|
1400 |
* the value/address loaded, and seeing if these are the same
|
1401 |
* when the store is performed. This is not actually the architecturally
|
1402 |
* mandated semantics, but it works for typical guest code sequences
|
1403 |
* and avoids having to monitor regular stores.
|
1404 |
*
|
1405 |
* In system emulation mode only one CPU will be running at once, so
|
1406 |
* this sequence is effectively atomic. In user emulation mode we
|
1407 |
* throw an exception and handle the atomic operation elsewhere.
|
1408 |
*/
|
1409 |
static void gen_load_exclusive(DisasContext *s, int rt, int rt2, |
1410 |
TCGv_i64 addr, int size, bool is_pair) |
1411 |
{ |
1412 |
TCGv_i64 tmp = tcg_temp_new_i64(); |
1413 |
TCGMemOp memop = MO_TE + size; |
1414 |
|
1415 |
g_assert(size <= 3);
|
1416 |
tcg_gen_qemu_ld_i64(tmp, addr, get_mem_index(s), memop); |
1417 |
|
1418 |
if (is_pair) {
|
1419 |
TCGv_i64 addr2 = tcg_temp_new_i64(); |
1420 |
TCGv_i64 hitmp = tcg_temp_new_i64(); |
1421 |
|
1422 |
g_assert(size >= 2);
|
1423 |
tcg_gen_addi_i64(addr2, addr, 1 << size);
|
1424 |
tcg_gen_qemu_ld_i64(hitmp, addr2, get_mem_index(s), memop); |
1425 |
tcg_temp_free_i64(addr2); |
1426 |
tcg_gen_mov_i64(cpu_exclusive_high, hitmp); |
1427 |
tcg_gen_mov_i64(cpu_reg(s, rt2), hitmp); |
1428 |
tcg_temp_free_i64(hitmp); |
1429 |
} |
1430 |
|
1431 |
tcg_gen_mov_i64(cpu_exclusive_val, tmp); |
1432 |
tcg_gen_mov_i64(cpu_reg(s, rt), tmp); |
1433 |
|
1434 |
tcg_temp_free_i64(tmp); |
1435 |
tcg_gen_mov_i64(cpu_exclusive_addr, addr); |
1436 |
} |
1437 |
|
1438 |
#ifdef CONFIG_USER_ONLY
|
1439 |
static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2, |
1440 |
TCGv_i64 addr, int size, int is_pair) |
1441 |
{ |
1442 |
tcg_gen_mov_i64(cpu_exclusive_test, addr); |
1443 |
tcg_gen_movi_i32(cpu_exclusive_info, |
1444 |
size | is_pair << 2 | (rd << 4) | (rt << 9) | (rt2 << 14)); |
1445 |
gen_exception_insn(s, 4, EXCP_STREX);
|
1446 |
} |
1447 |
#else
|
1448 |
static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2, |
1449 |
TCGv_i64 addr, int size, int is_pair) |
1450 |
{ |
1451 |
qemu_log_mask(LOG_UNIMP, |
1452 |
"%s:%d: system mode store_exclusive unsupported "
|
1453 |
"at pc=%016" PRIx64 "\n", |
1454 |
__FILE__, __LINE__, s->pc - 4);
|
1455 |
} |
1456 |
#endif
|
1457 |
|
1458 |
/* C3.3.6 Load/store exclusive
|
1459 |
*
|
1460 |
* 31 30 29 24 23 22 21 20 16 15 14 10 9 5 4 0
|
1461 |
* +-----+-------------+----+---+----+------+----+-------+------+------+
|
1462 |
* | sz | 0 0 1 0 0 0 | o2 | L | o1 | Rs | o0 | Rt2 | Rn | Rt |
|
1463 |
* +-----+-------------+----+---+----+------+----+-------+------+------+
|
1464 |
*
|
1465 |
* sz: 00 -> 8 bit, 01 -> 16 bit, 10 -> 32 bit, 11 -> 64 bit
|
1466 |
* L: 0 -> store, 1 -> load
|
1467 |
* o2: 0 -> exclusive, 1 -> not
|
1468 |
* o1: 0 -> single register, 1 -> register pair
|
1469 |
* o0: 1 -> load-acquire/store-release, 0 -> not
|
1470 |
*
|
1471 |
* o0 == 0 AND o2 == 1 is un-allocated
|
1472 |
* o1 == 1 is un-allocated except for 32 and 64 bit sizes
|
1473 |
*/
|
1474 |
static void disas_ldst_excl(DisasContext *s, uint32_t insn) |
1475 |
{ |
1476 |
int rt = extract32(insn, 0, 5); |
1477 |
int rn = extract32(insn, 5, 5); |
1478 |
int rt2 = extract32(insn, 10, 5); |
1479 |
int is_lasr = extract32(insn, 15, 1); |
1480 |
int rs = extract32(insn, 16, 5); |
1481 |
int is_pair = extract32(insn, 21, 1); |
1482 |
int is_store = !extract32(insn, 22, 1); |
1483 |
int is_excl = !extract32(insn, 23, 1); |
1484 |
int size = extract32(insn, 30, 2); |
1485 |
TCGv_i64 tcg_addr; |
1486 |
|
1487 |
if ((!is_excl && !is_lasr) ||
|
1488 |
(is_pair && size < 2)) {
|
1489 |
unallocated_encoding(s); |
1490 |
return;
|
1491 |
} |
1492 |
|
1493 |
if (rn == 31) { |
1494 |
gen_check_sp_alignment(s); |
1495 |
} |
1496 |
tcg_addr = read_cpu_reg_sp(s, rn, 1);
|
1497 |
|
1498 |
/* Note that since TCG is single threaded load-acquire/store-release
|
1499 |
* semantics require no extra if (is_lasr) { ... } handling.
|
1500 |
*/
|
1501 |
|
1502 |
if (is_excl) {
|
1503 |
if (!is_store) {
|
1504 |
gen_load_exclusive(s, rt, rt2, tcg_addr, size, is_pair); |
1505 |
} else {
|
1506 |
gen_store_exclusive(s, rs, rt, rt2, tcg_addr, size, is_pair); |
1507 |
} |
1508 |
} else {
|
1509 |
TCGv_i64 tcg_rt = cpu_reg(s, rt); |
1510 |
if (is_store) {
|
1511 |
do_gpr_st(s, tcg_rt, tcg_addr, size); |
1512 |
} else {
|
1513 |
do_gpr_ld(s, tcg_rt, tcg_addr, size, false, false); |
1514 |
} |
1515 |
if (is_pair) {
|
1516 |
TCGv_i64 tcg_rt2 = cpu_reg(s, rt); |
1517 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, 1 << size);
|
1518 |
if (is_store) {
|
1519 |
do_gpr_st(s, tcg_rt2, tcg_addr, size); |
1520 |
} else {
|
1521 |
do_gpr_ld(s, tcg_rt2, tcg_addr, size, false, false); |
1522 |
} |
1523 |
} |
1524 |
} |
1525 |
} |
1526 |
|
1527 |
/*
|
1528 |
* C3.3.5 Load register (literal)
|
1529 |
*
|
1530 |
* 31 30 29 27 26 25 24 23 5 4 0
|
1531 |
* +-----+-------+---+-----+-------------------+-------+
|
1532 |
* | opc | 0 1 1 | V | 0 0 | imm19 | Rt |
|
1533 |
* +-----+-------+---+-----+-------------------+-------+
|
1534 |
*
|
1535 |
* V: 1 -> vector (simd/fp)
|
1536 |
* opc (non-vector): 00 -> 32 bit, 01 -> 64 bit,
|
1537 |
* 10-> 32 bit signed, 11 -> prefetch
|
1538 |
* opc (vector): 00 -> 32 bit, 01 -> 64 bit, 10 -> 128 bit (11 unallocated)
|
1539 |
*/
|
1540 |
static void disas_ld_lit(DisasContext *s, uint32_t insn) |
1541 |
{ |
1542 |
int rt = extract32(insn, 0, 5); |
1543 |
int64_t imm = sextract32(insn, 5, 19) << 2; |
1544 |
bool is_vector = extract32(insn, 26, 1); |
1545 |
int opc = extract32(insn, 30, 2); |
1546 |
bool is_signed = false; |
1547 |
int size = 2; |
1548 |
TCGv_i64 tcg_rt, tcg_addr; |
1549 |
|
1550 |
if (is_vector) {
|
1551 |
if (opc == 3) { |
1552 |
unallocated_encoding(s); |
1553 |
return;
|
1554 |
} |
1555 |
size = 2 + opc;
|
1556 |
} else {
|
1557 |
if (opc == 3) { |
1558 |
/* PRFM (literal) : prefetch */
|
1559 |
return;
|
1560 |
} |
1561 |
size = 2 + extract32(opc, 0, 1); |
1562 |
is_signed = extract32(opc, 1, 1); |
1563 |
} |
1564 |
|
1565 |
tcg_rt = cpu_reg(s, rt); |
1566 |
|
1567 |
tcg_addr = tcg_const_i64((s->pc - 4) + imm);
|
1568 |
if (is_vector) {
|
1569 |
do_fp_ld(s, rt, tcg_addr, size); |
1570 |
} else {
|
1571 |
do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, false);
|
1572 |
} |
1573 |
tcg_temp_free_i64(tcg_addr); |
1574 |
} |
1575 |
|
1576 |
/*
|
1577 |
* C5.6.80 LDNP (Load Pair - non-temporal hint)
|
1578 |
* C5.6.81 LDP (Load Pair - non vector)
|
1579 |
* C5.6.82 LDPSW (Load Pair Signed Word - non vector)
|
1580 |
* C5.6.176 STNP (Store Pair - non-temporal hint)
|
1581 |
* C5.6.177 STP (Store Pair - non vector)
|
1582 |
* C6.3.165 LDNP (Load Pair of SIMD&FP - non-temporal hint)
|
1583 |
* C6.3.165 LDP (Load Pair of SIMD&FP)
|
1584 |
* C6.3.284 STNP (Store Pair of SIMD&FP - non-temporal hint)
|
1585 |
* C6.3.284 STP (Store Pair of SIMD&FP)
|
1586 |
*
|
1587 |
* 31 30 29 27 26 25 24 23 22 21 15 14 10 9 5 4 0
|
1588 |
* +-----+-------+---+---+-------+---+-----------------------------+
|
1589 |
* | opc | 1 0 1 | V | 0 | index | L | imm7 | Rt2 | Rn | Rt |
|
1590 |
* +-----+-------+---+---+-------+---+-------+-------+------+------+
|
1591 |
*
|
1592 |
* opc: LDP/STP/LDNP/STNP 00 -> 32 bit, 10 -> 64 bit
|
1593 |
* LDPSW 01
|
1594 |
* LDP/STP/LDNP/STNP (SIMD) 00 -> 32 bit, 01 -> 64 bit, 10 -> 128 bit
|
1595 |
* V: 0 -> GPR, 1 -> Vector
|
1596 |
* idx: 00 -> signed offset with non-temporal hint, 01 -> post-index,
|
1597 |
* 10 -> signed offset, 11 -> pre-index
|
1598 |
* L: 0 -> Store 1 -> Load
|
1599 |
*
|
1600 |
* Rt, Rt2 = GPR or SIMD registers to be stored
|
1601 |
* Rn = general purpose register containing address
|
1602 |
* imm7 = signed offset (multiple of 4 or 8 depending on size)
|
1603 |
*/
|
1604 |
static void disas_ldst_pair(DisasContext *s, uint32_t insn) |
1605 |
{ |
1606 |
int rt = extract32(insn, 0, 5); |
1607 |
int rn = extract32(insn, 5, 5); |
1608 |
int rt2 = extract32(insn, 10, 5); |
1609 |
int64_t offset = sextract32(insn, 15, 7); |
1610 |
int index = extract32(insn, 23, 2); |
1611 |
bool is_vector = extract32(insn, 26, 1); |
1612 |
bool is_load = extract32(insn, 22, 1); |
1613 |
int opc = extract32(insn, 30, 2); |
1614 |
|
1615 |
bool is_signed = false; |
1616 |
bool postindex = false; |
1617 |
bool wback = false; |
1618 |
|
1619 |
TCGv_i64 tcg_addr; /* calculated address */
|
1620 |
int size;
|
1621 |
|
1622 |
if (opc == 3) { |
1623 |
unallocated_encoding(s); |
1624 |
return;
|
1625 |
} |
1626 |
|
1627 |
if (is_vector) {
|
1628 |
size = 2 + opc;
|
1629 |
} else {
|
1630 |
size = 2 + extract32(opc, 1, 1); |
1631 |
is_signed = extract32(opc, 0, 1); |
1632 |
if (!is_load && is_signed) {
|
1633 |
unallocated_encoding(s); |
1634 |
return;
|
1635 |
} |
1636 |
} |
1637 |
|
1638 |
switch (index) {
|
1639 |
case 1: /* post-index */ |
1640 |
postindex = true;
|
1641 |
wback = true;
|
1642 |
break;
|
1643 |
case 0: |
1644 |
/* signed offset with "non-temporal" hint. Since we don't emulate
|
1645 |
* caches we don't care about hints to the cache system about
|
1646 |
* data access patterns, and handle this identically to plain
|
1647 |
* signed offset.
|
1648 |
*/
|
1649 |
if (is_signed) {
|
1650 |
/* There is no non-temporal-hint version of LDPSW */
|
1651 |
unallocated_encoding(s); |
1652 |
return;
|
1653 |
} |
1654 |
postindex = false;
|
1655 |
break;
|
1656 |
case 2: /* signed offset, rn not updated */ |
1657 |
postindex = false;
|
1658 |
break;
|
1659 |
case 3: /* pre-index */ |
1660 |
postindex = false;
|
1661 |
wback = true;
|
1662 |
break;
|
1663 |
} |
1664 |
|
1665 |
offset <<= size; |
1666 |
|
1667 |
if (rn == 31) { |
1668 |
gen_check_sp_alignment(s); |
1669 |
} |
1670 |
|
1671 |
tcg_addr = read_cpu_reg_sp(s, rn, 1);
|
1672 |
|
1673 |
if (!postindex) {
|
1674 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, offset); |
1675 |
} |
1676 |
|
1677 |
if (is_vector) {
|
1678 |
if (is_load) {
|
1679 |
do_fp_ld(s, rt, tcg_addr, size); |
1680 |
} else {
|
1681 |
do_fp_st(s, rt, tcg_addr, size); |
1682 |
} |
1683 |
} else {
|
1684 |
TCGv_i64 tcg_rt = cpu_reg(s, rt); |
1685 |
if (is_load) {
|
1686 |
do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, false);
|
1687 |
} else {
|
1688 |
do_gpr_st(s, tcg_rt, tcg_addr, size); |
1689 |
} |
1690 |
} |
1691 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, 1 << size);
|
1692 |
if (is_vector) {
|
1693 |
if (is_load) {
|
1694 |
do_fp_ld(s, rt2, tcg_addr, size); |
1695 |
} else {
|
1696 |
do_fp_st(s, rt2, tcg_addr, size); |
1697 |
} |
1698 |
} else {
|
1699 |
TCGv_i64 tcg_rt2 = cpu_reg(s, rt2); |
1700 |
if (is_load) {
|
1701 |
do_gpr_ld(s, tcg_rt2, tcg_addr, size, is_signed, false);
|
1702 |
} else {
|
1703 |
do_gpr_st(s, tcg_rt2, tcg_addr, size); |
1704 |
} |
1705 |
} |
1706 |
|
1707 |
if (wback) {
|
1708 |
if (postindex) {
|
1709 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, offset - (1 << size));
|
1710 |
} else {
|
1711 |
tcg_gen_subi_i64(tcg_addr, tcg_addr, 1 << size);
|
1712 |
} |
1713 |
tcg_gen_mov_i64(cpu_reg_sp(s, rn), tcg_addr); |
1714 |
} |
1715 |
} |
1716 |
|
1717 |
/*
|
1718 |
* C3.3.8 Load/store (immediate post-indexed)
|
1719 |
* C3.3.9 Load/store (immediate pre-indexed)
|
1720 |
* C3.3.12 Load/store (unscaled immediate)
|
1721 |
*
|
1722 |
* 31 30 29 27 26 25 24 23 22 21 20 12 11 10 9 5 4 0
|
1723 |
* +----+-------+---+-----+-----+---+--------+-----+------+------+
|
1724 |
* |size| 1 1 1 | V | 0 0 | opc | 0 | imm9 | idx | Rn | Rt |
|
1725 |
* +----+-------+---+-----+-----+---+--------+-----+------+------+
|
1726 |
*
|
1727 |
* idx = 01 -> post-indexed, 11 pre-indexed, 00 unscaled imm. (no writeback)
|
1728 |
* V = 0 -> non-vector
|
1729 |
* size: 00 -> 8 bit, 01 -> 16 bit, 10 -> 32 bit, 11 -> 64bit
|
1730 |
* opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
|
1731 |
*/
|
1732 |
static void disas_ldst_reg_imm9(DisasContext *s, uint32_t insn) |
1733 |
{ |
1734 |
int rt = extract32(insn, 0, 5); |
1735 |
int rn = extract32(insn, 5, 5); |
1736 |
int imm9 = sextract32(insn, 12, 9); |
1737 |
int opc = extract32(insn, 22, 2); |
1738 |
int size = extract32(insn, 30, 2); |
1739 |
int idx = extract32(insn, 10, 2); |
1740 |
bool is_signed = false; |
1741 |
bool is_store = false; |
1742 |
bool is_extended = false; |
1743 |
bool is_vector = extract32(insn, 26, 1); |
1744 |
bool post_index;
|
1745 |
bool writeback;
|
1746 |
|
1747 |
TCGv_i64 tcg_addr; |
1748 |
|
1749 |
if (is_vector) {
|
1750 |
size |= (opc & 2) << 1; |
1751 |
if (size > 4) { |
1752 |
unallocated_encoding(s); |
1753 |
return;
|
1754 |
} |
1755 |
is_store = ((opc & 1) == 0); |
1756 |
} else {
|
1757 |
if (size == 3 && opc == 2) { |
1758 |
/* PRFM - prefetch */
|
1759 |
return;
|
1760 |
} |
1761 |
if (opc == 3 && size > 1) { |
1762 |
unallocated_encoding(s); |
1763 |
return;
|
1764 |
} |
1765 |
is_store = (opc == 0);
|
1766 |
is_signed = opc & (1<<1); |
1767 |
is_extended = (size < 3) && (opc & 1); |
1768 |
} |
1769 |
|
1770 |
switch (idx) {
|
1771 |
case 0: |
1772 |
post_index = false;
|
1773 |
writeback = false;
|
1774 |
break;
|
1775 |
case 1: |
1776 |
post_index = true;
|
1777 |
writeback = true;
|
1778 |
break;
|
1779 |
case 3: |
1780 |
post_index = false;
|
1781 |
writeback = true;
|
1782 |
break;
|
1783 |
case 2: |
1784 |
g_assert(false);
|
1785 |
break;
|
1786 |
} |
1787 |
|
1788 |
if (rn == 31) { |
1789 |
gen_check_sp_alignment(s); |
1790 |
} |
1791 |
tcg_addr = read_cpu_reg_sp(s, rn, 1);
|
1792 |
|
1793 |
if (!post_index) {
|
1794 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, imm9); |
1795 |
} |
1796 |
|
1797 |
if (is_vector) {
|
1798 |
if (is_store) {
|
1799 |
do_fp_st(s, rt, tcg_addr, size); |
1800 |
} else {
|
1801 |
do_fp_ld(s, rt, tcg_addr, size); |
1802 |
} |
1803 |
} else {
|
1804 |
TCGv_i64 tcg_rt = cpu_reg(s, rt); |
1805 |
if (is_store) {
|
1806 |
do_gpr_st(s, tcg_rt, tcg_addr, size); |
1807 |
} else {
|
1808 |
do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended); |
1809 |
} |
1810 |
} |
1811 |
|
1812 |
if (writeback) {
|
1813 |
TCGv_i64 tcg_rn = cpu_reg_sp(s, rn); |
1814 |
if (post_index) {
|
1815 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, imm9); |
1816 |
} |
1817 |
tcg_gen_mov_i64(tcg_rn, tcg_addr); |
1818 |
} |
1819 |
} |
1820 |
|
1821 |
/*
|
1822 |
* C3.3.10 Load/store (register offset)
|
1823 |
*
|
1824 |
* 31 30 29 27 26 25 24 23 22 21 20 16 15 13 12 11 10 9 5 4 0
|
1825 |
* +----+-------+---+-----+-----+---+------+-----+--+-----+----+----+
|
1826 |
* |size| 1 1 1 | V | 0 0 | opc | 1 | Rm | opt | S| 1 0 | Rn | Rt |
|
1827 |
* +----+-------+---+-----+-----+---+------+-----+--+-----+----+----+
|
1828 |
*
|
1829 |
* For non-vector:
|
1830 |
* size: 00-> byte, 01 -> 16 bit, 10 -> 32bit, 11 -> 64bit
|
1831 |
* opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
|
1832 |
* For vector:
|
1833 |
* size is opc<1>:size<1:0> so 100 -> 128 bit; 110 and 111 unallocated
|
1834 |
* opc<0>: 0 -> store, 1 -> load
|
1835 |
* V: 1 -> vector/simd
|
1836 |
* opt: extend encoding (see DecodeRegExtend)
|
1837 |
* S: if S=1 then scale (essentially index by sizeof(size))
|
1838 |
* Rt: register to transfer into/out of
|
1839 |
* Rn: address register or SP for base
|
1840 |
* Rm: offset register or ZR for offset
|
1841 |
*/
|
1842 |
static void disas_ldst_reg_roffset(DisasContext *s, uint32_t insn) |
1843 |
{ |
1844 |
int rt = extract32(insn, 0, 5); |
1845 |
int rn = extract32(insn, 5, 5); |
1846 |
int shift = extract32(insn, 12, 1); |
1847 |
int rm = extract32(insn, 16, 5); |
1848 |
int opc = extract32(insn, 22, 2); |
1849 |
int opt = extract32(insn, 13, 3); |
1850 |
int size = extract32(insn, 30, 2); |
1851 |
bool is_signed = false; |
1852 |
bool is_store = false; |
1853 |
bool is_extended = false; |
1854 |
bool is_vector = extract32(insn, 26, 1); |
1855 |
|
1856 |
TCGv_i64 tcg_rm; |
1857 |
TCGv_i64 tcg_addr; |
1858 |
|
1859 |
if (extract32(opt, 1, 1) == 0) { |
1860 |
unallocated_encoding(s); |
1861 |
return;
|
1862 |
} |
1863 |
|
1864 |
if (is_vector) {
|
1865 |
size |= (opc & 2) << 1; |
1866 |
if (size > 4) { |
1867 |
unallocated_encoding(s); |
1868 |
return;
|
1869 |
} |
1870 |
is_store = !extract32(opc, 0, 1); |
1871 |
} else {
|
1872 |
if (size == 3 && opc == 2) { |
1873 |
/* PRFM - prefetch */
|
1874 |
return;
|
1875 |
} |
1876 |
if (opc == 3 && size > 1) { |
1877 |
unallocated_encoding(s); |
1878 |
return;
|
1879 |
} |
1880 |
is_store = (opc == 0);
|
1881 |
is_signed = extract32(opc, 1, 1); |
1882 |
is_extended = (size < 3) && extract32(opc, 0, 1); |
1883 |
} |
1884 |
|
1885 |
if (rn == 31) { |
1886 |
gen_check_sp_alignment(s); |
1887 |
} |
1888 |
tcg_addr = read_cpu_reg_sp(s, rn, 1);
|
1889 |
|
1890 |
tcg_rm = read_cpu_reg(s, rm, 1);
|
1891 |
ext_and_shift_reg(tcg_rm, tcg_rm, opt, shift ? size : 0);
|
1892 |
|
1893 |
tcg_gen_add_i64(tcg_addr, tcg_addr, tcg_rm); |
1894 |
|
1895 |
if (is_vector) {
|
1896 |
if (is_store) {
|
1897 |
do_fp_st(s, rt, tcg_addr, size); |
1898 |
} else {
|
1899 |
do_fp_ld(s, rt, tcg_addr, size); |
1900 |
} |
1901 |
} else {
|
1902 |
TCGv_i64 tcg_rt = cpu_reg(s, rt); |
1903 |
if (is_store) {
|
1904 |
do_gpr_st(s, tcg_rt, tcg_addr, size); |
1905 |
} else {
|
1906 |
do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended); |
1907 |
} |
1908 |
} |
1909 |
} |
1910 |
|
1911 |
/*
|
1912 |
* C3.3.13 Load/store (unsigned immediate)
|
1913 |
*
|
1914 |
* 31 30 29 27 26 25 24 23 22 21 10 9 5
|
1915 |
* +----+-------+---+-----+-----+------------+-------+------+
|
1916 |
* |size| 1 1 1 | V | 0 1 | opc | imm12 | Rn | Rt |
|
1917 |
* +----+-------+---+-----+-----+------------+-------+------+
|
1918 |
*
|
1919 |
* For non-vector:
|
1920 |
* size: 00-> byte, 01 -> 16 bit, 10 -> 32bit, 11 -> 64bit
|
1921 |
* opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
|
1922 |
* For vector:
|
1923 |
* size is opc<1>:size<1:0> so 100 -> 128 bit; 110 and 111 unallocated
|
1924 |
* opc<0>: 0 -> store, 1 -> load
|
1925 |
* Rn: base address register (inc SP)
|
1926 |
* Rt: target register
|
1927 |
*/
|
1928 |
static void disas_ldst_reg_unsigned_imm(DisasContext *s, uint32_t insn) |
1929 |
{ |
1930 |
int rt = extract32(insn, 0, 5); |
1931 |
int rn = extract32(insn, 5, 5); |
1932 |
unsigned int imm12 = extract32(insn, 10, 12); |
1933 |
bool is_vector = extract32(insn, 26, 1); |
1934 |
int size = extract32(insn, 30, 2); |
1935 |
int opc = extract32(insn, 22, 2); |
1936 |
unsigned int offset; |
1937 |
|
1938 |
TCGv_i64 tcg_addr; |
1939 |
|
1940 |
bool is_store;
|
1941 |
bool is_signed = false; |
1942 |
bool is_extended = false; |
1943 |
|
1944 |
if (is_vector) {
|
1945 |
size |= (opc & 2) << 1; |
1946 |
if (size > 4) { |
1947 |
unallocated_encoding(s); |
1948 |
return;
|
1949 |
} |
1950 |
is_store = !extract32(opc, 0, 1); |
1951 |
} else {
|
1952 |
if (size == 3 && opc == 2) { |
1953 |
/* PRFM - prefetch */
|
1954 |
return;
|
1955 |
} |
1956 |
if (opc == 3 && size > 1) { |
1957 |
unallocated_encoding(s); |
1958 |
return;
|
1959 |
} |
1960 |
is_store = (opc == 0);
|
1961 |
is_signed = extract32(opc, 1, 1); |
1962 |
is_extended = (size < 3) && extract32(opc, 0, 1); |
1963 |
} |
1964 |
|
1965 |
if (rn == 31) { |
1966 |
gen_check_sp_alignment(s); |
1967 |
} |
1968 |
tcg_addr = read_cpu_reg_sp(s, rn, 1);
|
1969 |
offset = imm12 << size; |
1970 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, offset); |
1971 |
|
1972 |
if (is_vector) {
|
1973 |
if (is_store) {
|
1974 |
do_fp_st(s, rt, tcg_addr, size); |
1975 |
} else {
|
1976 |
do_fp_ld(s, rt, tcg_addr, size); |
1977 |
} |
1978 |
} else {
|
1979 |
TCGv_i64 tcg_rt = cpu_reg(s, rt); |
1980 |
if (is_store) {
|
1981 |
do_gpr_st(s, tcg_rt, tcg_addr, size); |
1982 |
} else {
|
1983 |
do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended); |
1984 |
} |
1985 |
} |
1986 |
} |
1987 |
|
1988 |
/* Load/store register (immediate forms) */
|
1989 |
static void disas_ldst_reg_imm(DisasContext *s, uint32_t insn) |
1990 |
{ |
1991 |
switch (extract32(insn, 10, 2)) { |
1992 |
case 0: case 1: case 3: |
1993 |
/* Load/store register (unscaled immediate) */
|
1994 |
/* Load/store immediate pre/post-indexed */
|
1995 |
disas_ldst_reg_imm9(s, insn); |
1996 |
break;
|
1997 |
case 2: |
1998 |
/* Load/store register unprivileged */
|
1999 |
unsupported_encoding(s, insn); |
2000 |
break;
|
2001 |
default:
|
2002 |
unallocated_encoding(s); |
2003 |
break;
|
2004 |
} |
2005 |
} |
2006 |
|
2007 |
/* Load/store register (all forms) */
|
2008 |
static void disas_ldst_reg(DisasContext *s, uint32_t insn) |
2009 |
{ |
2010 |
switch (extract32(insn, 24, 2)) { |
2011 |
case 0: |
2012 |
if (extract32(insn, 21, 1) == 1 && extract32(insn, 10, 2) == 2) { |
2013 |
disas_ldst_reg_roffset(s, insn); |
2014 |
} else {
|
2015 |
disas_ldst_reg_imm(s, insn); |
2016 |
} |
2017 |
break;
|
2018 |
case 1: |
2019 |
disas_ldst_reg_unsigned_imm(s, insn); |
2020 |
break;
|
2021 |
default:
|
2022 |
unallocated_encoding(s); |
2023 |
break;
|
2024 |
} |
2025 |
} |
2026 |
|
2027 |
/* C3.3.1 AdvSIMD load/store multiple structures
|
2028 |
*
|
2029 |
* 31 30 29 23 22 21 16 15 12 11 10 9 5 4 0
|
2030 |
* +---+---+---------------+---+-------------+--------+------+------+------+
|
2031 |
* | 0 | Q | 0 0 1 1 0 0 0 | L | 0 0 0 0 0 0 | opcode | size | Rn | Rt |
|
2032 |
* +---+---+---------------+---+-------------+--------+------+------+------+
|
2033 |
*
|
2034 |
* C3.3.2 AdvSIMD load/store multiple structures (post-indexed)
|
2035 |
*
|
2036 |
* 31 30 29 23 22 21 20 16 15 12 11 10 9 5 4 0
|
2037 |
* +---+---+---------------+---+---+---------+--------+------+------+------+
|
2038 |
* | 0 | Q | 0 0 1 1 0 0 1 | L | 0 | Rm | opcode | size | Rn | Rt |
|
2039 |
* +---+---+---------------+---+---+---------+--------+------+------+------+
|
2040 |
*
|
2041 |
* Rt: first (or only) SIMD&FP register to be transferred
|
2042 |
* Rn: base address or SP
|
2043 |
* Rm (post-index only): post-index register (when !31) or size dependent #imm
|
2044 |
*/
|
2045 |
static void disas_ldst_multiple_struct(DisasContext *s, uint32_t insn) |
2046 |
{ |
2047 |
int rt = extract32(insn, 0, 5); |
2048 |
int rn = extract32(insn, 5, 5); |
2049 |
int size = extract32(insn, 10, 2); |
2050 |
int opcode = extract32(insn, 12, 4); |
2051 |
bool is_store = !extract32(insn, 22, 1); |
2052 |
bool is_postidx = extract32(insn, 23, 1); |
2053 |
bool is_q = extract32(insn, 30, 1); |
2054 |
TCGv_i64 tcg_addr, tcg_rn; |
2055 |
|
2056 |
int ebytes = 1 << size; |
2057 |
int elements = (is_q ? 128 : 64) / (8 << size); |
2058 |
int rpt; /* num iterations */ |
2059 |
int selem; /* structure elements */ |
2060 |
int r;
|
2061 |
|
2062 |
if (extract32(insn, 31, 1) || extract32(insn, 21, 1)) { |
2063 |
unallocated_encoding(s); |
2064 |
return;
|
2065 |
} |
2066 |
|
2067 |
/* From the shared decode logic */
|
2068 |
switch (opcode) {
|
2069 |
case 0x0: |
2070 |
rpt = 1;
|
2071 |
selem = 4;
|
2072 |
break;
|
2073 |
case 0x2: |
2074 |
rpt = 4;
|
2075 |
selem = 1;
|
2076 |
break;
|
2077 |
case 0x4: |
2078 |
rpt = 1;
|
2079 |
selem = 3;
|
2080 |
break;
|
2081 |
case 0x6: |
2082 |
rpt = 3;
|
2083 |
selem = 1;
|
2084 |
break;
|
2085 |
case 0x7: |
2086 |
rpt = 1;
|
2087 |
selem = 1;
|
2088 |
break;
|
2089 |
case 0x8: |
2090 |
rpt = 1;
|
2091 |
selem = 2;
|
2092 |
break;
|
2093 |
case 0xa: |
2094 |
rpt = 2;
|
2095 |
selem = 1;
|
2096 |
break;
|
2097 |
default:
|
2098 |
unallocated_encoding(s); |
2099 |
return;
|
2100 |
} |
2101 |
|
2102 |
if (size == 3 && !is_q && selem != 1) { |
2103 |
/* reserved */
|
2104 |
unallocated_encoding(s); |
2105 |
return;
|
2106 |
} |
2107 |
|
2108 |
if (rn == 31) { |
2109 |
gen_check_sp_alignment(s); |
2110 |
} |
2111 |
|
2112 |
tcg_rn = cpu_reg_sp(s, rn); |
2113 |
tcg_addr = tcg_temp_new_i64(); |
2114 |
tcg_gen_mov_i64(tcg_addr, tcg_rn); |
2115 |
|
2116 |
for (r = 0; r < rpt; r++) { |
2117 |
int e;
|
2118 |
for (e = 0; e < elements; e++) { |
2119 |
int tt = (rt + r) % 32; |
2120 |
int xs;
|
2121 |
for (xs = 0; xs < selem; xs++) { |
2122 |
if (is_store) {
|
2123 |
do_vec_st(s, tt, e, tcg_addr, size); |
2124 |
} else {
|
2125 |
do_vec_ld(s, tt, e, tcg_addr, size); |
2126 |
|
2127 |
/* For non-quad operations, setting a slice of the low
|
2128 |
* 64 bits of the register clears the high 64 bits (in
|
2129 |
* the ARM ARM pseudocode this is implicit in the fact
|
2130 |
* that 'rval' is a 64 bit wide variable). We optimize
|
2131 |
* by noticing that we only need to do this the first
|
2132 |
* time we touch a register.
|
2133 |
*/
|
2134 |
if (!is_q && e == 0 && (r == 0 || xs == selem - 1)) { |
2135 |
clear_vec_high(s, tt); |
2136 |
} |
2137 |
} |
2138 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, ebytes); |
2139 |
tt = (tt + 1) % 32; |
2140 |
} |
2141 |
} |
2142 |
} |
2143 |
|
2144 |
if (is_postidx) {
|
2145 |
int rm = extract32(insn, 16, 5); |
2146 |
if (rm == 31) { |
2147 |
tcg_gen_mov_i64(tcg_rn, tcg_addr); |
2148 |
} else {
|
2149 |
tcg_gen_add_i64(tcg_rn, tcg_rn, cpu_reg(s, rm)); |
2150 |
} |
2151 |
} |
2152 |
tcg_temp_free_i64(tcg_addr); |
2153 |
} |
2154 |
|
2155 |
/* C3.3.3 AdvSIMD load/store single structure
|
2156 |
*
|
2157 |
* 31 30 29 23 22 21 20 16 15 13 12 11 10 9 5 4 0
|
2158 |
* +---+---+---------------+-----+-----------+-----+---+------+------+------+
|
2159 |
* | 0 | Q | 0 0 1 1 0 1 0 | L R | 0 0 0 0 0 | opc | S | size | Rn | Rt |
|
2160 |
* +---+---+---------------+-----+-----------+-----+---+------+------+------+
|
2161 |
*
|
2162 |
* C3.3.4 AdvSIMD load/store single structure (post-indexed)
|
2163 |
*
|
2164 |
* 31 30 29 23 22 21 20 16 15 13 12 11 10 9 5 4 0
|
2165 |
* +---+---+---------------+-----+-----------+-----+---+------+------+------+
|
2166 |
* | 0 | Q | 0 0 1 1 0 1 1 | L R | Rm | opc | S | size | Rn | Rt |
|
2167 |
* +---+---+---------------+-----+-----------+-----+---+------+------+------+
|
2168 |
*
|
2169 |
* Rt: first (or only) SIMD&FP register to be transferred
|
2170 |
* Rn: base address or SP
|
2171 |
* Rm (post-index only): post-index register (when !31) or size dependent #imm
|
2172 |
* index = encoded in Q:S:size dependent on size
|
2173 |
*
|
2174 |
* lane_size = encoded in R, opc
|
2175 |
* transfer width = encoded in opc, S, size
|
2176 |
*/
|
2177 |
static void disas_ldst_single_struct(DisasContext *s, uint32_t insn) |
2178 |
{ |
2179 |
int rt = extract32(insn, 0, 5); |
2180 |
int rn = extract32(insn, 5, 5); |
2181 |
int size = extract32(insn, 10, 2); |
2182 |
int S = extract32(insn, 12, 1); |
2183 |
int opc = extract32(insn, 13, 3); |
2184 |
int R = extract32(insn, 21, 1); |
2185 |
int is_load = extract32(insn, 22, 1); |
2186 |
int is_postidx = extract32(insn, 23, 1); |
2187 |
int is_q = extract32(insn, 30, 1); |
2188 |
|
2189 |
int scale = extract32(opc, 1, 2); |
2190 |
int selem = (extract32(opc, 0, 1) << 1 | R) + 1; |
2191 |
bool replicate = false; |
2192 |
int index = is_q << 3 | S << 2 | size; |
2193 |
int ebytes, xs;
|
2194 |
TCGv_i64 tcg_addr, tcg_rn; |
2195 |
|
2196 |
switch (scale) {
|
2197 |
case 3: |
2198 |
if (!is_load || S) {
|
2199 |
unallocated_encoding(s); |
2200 |
return;
|
2201 |
} |
2202 |
scale = size; |
2203 |
replicate = true;
|
2204 |
break;
|
2205 |
case 0: |
2206 |
break;
|
2207 |
case 1: |
2208 |
if (extract32(size, 0, 1)) { |
2209 |
unallocated_encoding(s); |
2210 |
return;
|
2211 |
} |
2212 |
index >>= 1;
|
2213 |
break;
|
2214 |
case 2: |
2215 |
if (extract32(size, 1, 1)) { |
2216 |
unallocated_encoding(s); |
2217 |
return;
|
2218 |
} |
2219 |
if (!extract32(size, 0, 1)) { |
2220 |
index >>= 2;
|
2221 |
} else {
|
2222 |
if (S) {
|
2223 |
unallocated_encoding(s); |
2224 |
return;
|
2225 |
} |
2226 |
index >>= 3;
|
2227 |
scale = 3;
|
2228 |
} |
2229 |
break;
|
2230 |
default:
|
2231 |
g_assert_not_reached(); |
2232 |
} |
2233 |
|
2234 |
ebytes = 1 << scale;
|
2235 |
|
2236 |
if (rn == 31) { |
2237 |
gen_check_sp_alignment(s); |
2238 |
} |
2239 |
|
2240 |
tcg_rn = cpu_reg_sp(s, rn); |
2241 |
tcg_addr = tcg_temp_new_i64(); |
2242 |
tcg_gen_mov_i64(tcg_addr, tcg_rn); |
2243 |
|
2244 |
for (xs = 0; xs < selem; xs++) { |
2245 |
if (replicate) {
|
2246 |
/* Load and replicate to all elements */
|
2247 |
uint64_t mulconst; |
2248 |
TCGv_i64 tcg_tmp = tcg_temp_new_i64(); |
2249 |
|
2250 |
tcg_gen_qemu_ld_i64(tcg_tmp, tcg_addr, |
2251 |
get_mem_index(s), MO_TE + scale); |
2252 |
switch (scale) {
|
2253 |
case 0: |
2254 |
mulconst = 0x0101010101010101ULL;
|
2255 |
break;
|
2256 |
case 1: |
2257 |
mulconst = 0x0001000100010001ULL;
|
2258 |
break;
|
2259 |
case 2: |
2260 |
mulconst = 0x0000000100000001ULL;
|
2261 |
break;
|
2262 |
case 3: |
2263 |
mulconst = 0;
|
2264 |
break;
|
2265 |
default:
|
2266 |
g_assert_not_reached(); |
2267 |
} |
2268 |
if (mulconst) {
|
2269 |
tcg_gen_muli_i64(tcg_tmp, tcg_tmp, mulconst); |
2270 |
} |
2271 |
write_vec_element(s, tcg_tmp, rt, 0, MO_64);
|
2272 |
if (is_q) {
|
2273 |
write_vec_element(s, tcg_tmp, rt, 1, MO_64);
|
2274 |
} else {
|
2275 |
clear_vec_high(s, rt); |
2276 |
} |
2277 |
tcg_temp_free_i64(tcg_tmp); |
2278 |
} else {
|
2279 |
/* Load/store one element per register */
|
2280 |
if (is_load) {
|
2281 |
do_vec_ld(s, rt, index, tcg_addr, MO_TE + scale); |
2282 |
} else {
|
2283 |
do_vec_st(s, rt, index, tcg_addr, MO_TE + scale); |
2284 |
} |
2285 |
} |
2286 |
tcg_gen_addi_i64(tcg_addr, tcg_addr, ebytes); |
2287 |
rt = (rt + 1) % 32; |
2288 |
} |
2289 |
|
2290 |
if (is_postidx) {
|
2291 |
int rm = extract32(insn, 16, 5); |
2292 |
if (rm == 31) { |
2293 |
tcg_gen_mov_i64(tcg_rn, tcg_addr); |
2294 |
} else {
|
2295 |
tcg_gen_add_i64(tcg_rn, tcg_rn, cpu_reg(s, rm)); |
2296 |
} |
2297 |
} |
2298 |
tcg_temp_free_i64(tcg_addr); |
2299 |
} |
2300 |
|
2301 |
/* C3.3 Loads and stores */
|
2302 |
static void disas_ldst(DisasContext *s, uint32_t insn) |
2303 |
{ |
2304 |
switch (extract32(insn, 24, 6)) { |
2305 |
case 0x08: /* Load/store exclusive */ |
2306 |
disas_ldst_excl(s, insn); |
2307 |
break;
|
2308 |
case 0x18: case 0x1c: /* Load register (literal) */ |
2309 |
disas_ld_lit(s, insn); |
2310 |
break;
|
2311 |
case 0x28: case 0x29: |
2312 |
case 0x2c: case 0x2d: /* Load/store pair (all forms) */ |
2313 |
disas_ldst_pair(s, insn); |
2314 |
break;
|
2315 |
case 0x38: case 0x39: |
2316 |
case 0x3c: case 0x3d: /* Load/store register (all forms) */ |
2317 |
disas_ldst_reg(s, insn); |
2318 |
break;
|
2319 |
case 0x0c: /* AdvSIMD load/store multiple structures */ |
2320 |
disas_ldst_multiple_struct(s, insn); |
2321 |
break;
|
2322 |
case 0x0d: /* AdvSIMD load/store single structure */ |
2323 |
disas_ldst_single_struct(s, insn); |
2324 |
break;
|
2325 |
default:
|
2326 |
unallocated_encoding(s); |
2327 |
break;
|
2328 |
} |
2329 |
} |
2330 |
|
2331 |
/* C3.4.6 PC-rel. addressing
|
2332 |
* 31 30 29 28 24 23 5 4 0
|
2333 |
* +----+-------+-----------+-------------------+------+
|
2334 |
* | op | immlo | 1 0 0 0 0 | immhi | Rd |
|
2335 |
* +----+-------+-----------+-------------------+------+
|
2336 |
*/
|
2337 |
static void disas_pc_rel_adr(DisasContext *s, uint32_t insn) |
2338 |
{ |
2339 |
unsigned int page, rd; |
2340 |
uint64_t base; |
2341 |
int64_t offset; |
2342 |
|
2343 |
page = extract32(insn, 31, 1); |
2344 |
/* SignExtend(immhi:immlo) -> offset */
|
2345 |
offset = ((int64_t)sextract32(insn, 5, 19) << 2) | extract32(insn, 29, 2); |
2346 |
rd = extract32(insn, 0, 5); |
2347 |
base = s->pc - 4;
|
2348 |
|
2349 |
if (page) {
|
2350 |
/* ADRP (page based) */
|
2351 |
base &= ~0xfff;
|
2352 |
offset <<= 12;
|
2353 |
} |
2354 |
|
2355 |
tcg_gen_movi_i64(cpu_reg(s, rd), base + offset); |
2356 |
} |
2357 |
|
2358 |
/*
|
2359 |
* C3.4.1 Add/subtract (immediate)
|
2360 |
*
|
2361 |
* 31 30 29 28 24 23 22 21 10 9 5 4 0
|
2362 |
* +--+--+--+-----------+-----+-------------+-----+-----+
|
2363 |
* |sf|op| S| 1 0 0 0 1 |shift| imm12 | Rn | Rd |
|
2364 |
* +--+--+--+-----------+-----+-------------+-----+-----+
|
2365 |
*
|
2366 |
* sf: 0 -> 32bit, 1 -> 64bit
|
2367 |
* op: 0 -> add , 1 -> sub
|
2368 |
* S: 1 -> set flags
|
2369 |
* shift: 00 -> LSL imm by 0, 01 -> LSL imm by 12
|
2370 |
*/
|
2371 |
static void disas_add_sub_imm(DisasContext *s, uint32_t insn) |
2372 |
{ |
2373 |
int rd = extract32(insn, 0, 5); |
2374 |
int rn = extract32(insn, 5, 5); |
2375 |
uint64_t imm = extract32(insn, 10, 12); |
2376 |
int shift = extract32(insn, 22, 2); |
2377 |
bool setflags = extract32(insn, 29, 1); |
2378 |
bool sub_op = extract32(insn, 30, 1); |
2379 |
bool is_64bit = extract32(insn, 31, 1); |
2380 |
|
2381 |
TCGv_i64 tcg_rn = cpu_reg_sp(s, rn); |
2382 |
TCGv_i64 tcg_rd = setflags ? cpu_reg(s, rd) : cpu_reg_sp(s, rd); |
2383 |
TCGv_i64 tcg_result; |
2384 |
|
2385 |
switch (shift) {
|
2386 |
case 0x0: |
2387 |
break;
|
2388 |
case 0x1: |
2389 |
imm <<= 12;
|
2390 |
break;
|
2391 |
default:
|
2392 |
unallocated_encoding(s); |
2393 |
return;
|
2394 |
} |
2395 |
|
2396 |
tcg_result = tcg_temp_new_i64(); |
2397 |
if (!setflags) {
|
2398 |
if (sub_op) {
|
2399 |
tcg_gen_subi_i64(tcg_result, tcg_rn, imm); |
2400 |
} else {
|
2401 |
tcg_gen_addi_i64(tcg_result, tcg_rn, imm); |
2402 |
} |
2403 |
} else {
|
2404 |
TCGv_i64 tcg_imm = tcg_const_i64(imm); |
2405 |
if (sub_op) {
|
2406 |
gen_sub_CC(is_64bit, tcg_result, tcg_rn, tcg_imm); |
2407 |
} else {
|
2408 |
gen_add_CC(is_64bit, tcg_result, tcg_rn, tcg_imm); |
2409 |
} |
2410 |
tcg_temp_free_i64(tcg_imm); |
2411 |
} |
2412 |
|
2413 |
if (is_64bit) {
|
2414 |
tcg_gen_mov_i64(tcg_rd, tcg_result); |
2415 |
} else {
|
2416 |
tcg_gen_ext32u_i64(tcg_rd, tcg_result); |
2417 |
} |
2418 |
|
2419 |
tcg_temp_free_i64(tcg_result); |
2420 |
} |
2421 |
|
2422 |
/* The input should be a value in the bottom e bits (with higher
|
2423 |
* bits zero); returns that value replicated into every element
|
2424 |
* of size e in a 64 bit integer.
|
2425 |
*/
|
2426 |
static uint64_t bitfield_replicate(uint64_t mask, unsigned int e) |
2427 |
{ |
2428 |
assert(e != 0);
|
2429 |
while (e < 64) { |
2430 |
mask |= mask << e; |
2431 |
e *= 2;
|
2432 |
} |
2433 |
return mask;
|
2434 |
} |
2435 |
|
2436 |
/* Return a value with the bottom len bits set (where 0 < len <= 64) */
|
2437 |
static inline uint64_t bitmask64(unsigned int length) |
2438 |
{ |
2439 |
assert(length > 0 && length <= 64); |
2440 |
return ~0ULL >> (64 - length); |
2441 |
} |
2442 |
|
2443 |
/* Simplified variant of pseudocode DecodeBitMasks() for the case where we
|
2444 |
* only require the wmask. Returns false if the imms/immr/immn are a reserved
|
2445 |
* value (ie should cause a guest UNDEF exception), and true if they are
|
2446 |
* valid, in which case the decoded bit pattern is written to result.
|
2447 |
*/
|
2448 |
static bool logic_imm_decode_wmask(uint64_t *result, unsigned int immn, |
2449 |
unsigned int imms, unsigned int immr) |
2450 |
{ |
2451 |
uint64_t mask; |
2452 |
unsigned e, levels, s, r;
|
2453 |
int len;
|
2454 |
|
2455 |
assert(immn < 2 && imms < 64 && immr < 64); |
2456 |
|
2457 |
/* The bit patterns we create here are 64 bit patterns which
|
2458 |
* are vectors of identical elements of size e = 2, 4, 8, 16, 32 or
|
2459 |
* 64 bits each. Each element contains the same value: a run
|
2460 |
* of between 1 and e-1 non-zero bits, rotated within the
|
2461 |
* element by between 0 and e-1 bits.
|
2462 |
*
|
2463 |
* The element size and run length are encoded into immn (1 bit)
|
2464 |
* and imms (6 bits) as follows:
|
2465 |
* 64 bit elements: immn = 1, imms = <length of run - 1>
|
2466 |
* 32 bit elements: immn = 0, imms = 0 : <length of run - 1>
|
2467 |
* 16 bit elements: immn = 0, imms = 10 : <length of run - 1>
|
2468 |
* 8 bit elements: immn = 0, imms = 110 : <length of run - 1>
|
2469 |
* 4 bit elements: immn = 0, imms = 1110 : <length of run - 1>
|
2470 |
* 2 bit elements: immn = 0, imms = 11110 : <length of run - 1>
|
2471 |
* Notice that immn = 0, imms = 11111x is the only combination
|
2472 |
* not covered by one of the above options; this is reserved.
|
2473 |
* Further, <length of run - 1> all-ones is a reserved pattern.
|
2474 |
*
|
2475 |
* In all cases the rotation is by immr % e (and immr is 6 bits).
|
2476 |
*/
|
2477 |
|
2478 |
/* First determine the element size */
|
2479 |
len = 31 - clz32((immn << 6) | (~imms & 0x3f)); |
2480 |
if (len < 1) { |
2481 |
/* This is the immn == 0, imms == 0x11111x case */
|
2482 |
return false; |
2483 |
} |
2484 |
e = 1 << len;
|
2485 |
|
2486 |
levels = e - 1;
|
2487 |
s = imms & levels; |
2488 |
r = immr & levels; |
2489 |
|
2490 |
if (s == levels) {
|
2491 |
/* <length of run - 1> mustn't be all-ones. */
|
2492 |
return false; |
2493 |
} |
2494 |
|
2495 |
/* Create the value of one element: s+1 set bits rotated
|
2496 |
* by r within the element (which is e bits wide)...
|
2497 |
*/
|
2498 |
mask = bitmask64(s + 1);
|
2499 |
mask = (mask >> r) | (mask << (e - r)); |
2500 |
/* ...then replicate the element over the whole 64 bit value */
|
2501 |
mask = bitfield_replicate(mask, e); |
2502 |
*result = mask; |
2503 |
return true; |
2504 |
} |
2505 |
|
2506 |
/* C3.4.4 Logical (immediate)
|
2507 |
* 31 30 29 28 23 22 21 16 15 10 9 5 4 0
|
2508 |
* +----+-----+-------------+---+------+------+------+------+
|
2509 |
* | sf | opc | 1 0 0 1 0 0 | N | immr | imms | Rn | Rd |
|
2510 |
* +----+-----+-------------+---+------+------+------+------+
|
2511 |
*/
|
2512 |
static void disas_logic_imm(DisasContext *s, uint32_t insn) |
2513 |
{ |
2514 |
unsigned int sf, opc, is_n, immr, imms, rn, rd; |
2515 |
TCGv_i64 tcg_rd, tcg_rn; |
2516 |
uint64_t wmask; |
2517 |
bool is_and = false; |
2518 |
|
2519 |
sf = extract32(insn, 31, 1); |
2520 |
opc = extract32(insn, 29, 2); |
2521 |
is_n = extract32(insn, 22, 1); |
2522 |
immr = extract32(insn, 16, 6); |
2523 |
imms = extract32(insn, 10, 6); |
2524 |
rn = extract32(insn, 5, 5); |
2525 |
rd = extract32(insn, 0, 5); |
2526 |
|
2527 |
if (!sf && is_n) {
|
2528 |
unallocated_encoding(s); |
2529 |
return;
|
2530 |
} |
2531 |
|
2532 |
if (opc == 0x3) { /* ANDS */ |
2533 |
tcg_rd = cpu_reg(s, rd); |
2534 |
} else {
|
2535 |
tcg_rd = cpu_reg_sp(s, rd); |
2536 |
} |
2537 |
tcg_rn = cpu_reg(s, rn); |
2538 |
|
2539 |
if (!logic_imm_decode_wmask(&wmask, is_n, imms, immr)) {
|
2540 |
/* some immediate field values are reserved */
|
2541 |
unallocated_encoding(s); |
2542 |
return;
|
2543 |
} |
2544 |
|
2545 |
if (!sf) {
|
2546 |
wmask &= 0xffffffff;
|
2547 |
} |
2548 |
|
2549 |
switch (opc) {
|
2550 |
case 0x3: /* ANDS */ |
2551 |
case 0x0: /* AND */ |
2552 |
tcg_gen_andi_i64(tcg_rd, tcg_rn, wmask); |
2553 |
is_and = true;
|
2554 |
break;
|
2555 |
case 0x1: /* ORR */ |
2556 |
tcg_gen_ori_i64(tcg_rd, tcg_rn, wmask); |
2557 |
break;
|
2558 |
case 0x2: /* EOR */ |
2559 |
tcg_gen_xori_i64(tcg_rd, tcg_rn, wmask); |
2560 |
break;
|
2561 |
default:
|
2562 |
assert(FALSE); /* must handle all above */
|
2563 |
break;
|
2564 |
} |
2565 |
|
2566 |
if (!sf && !is_and) {
|
2567 |
/* zero extend final result; we know we can skip this for AND
|
2568 |
* since the immediate had the high 32 bits clear.
|
2569 |
*/
|
2570 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
2571 |
} |
2572 |
|
2573 |
if (opc == 3) { /* ANDS */ |
2574 |
gen_logic_CC(sf, tcg_rd); |
2575 |
} |
2576 |
} |
2577 |
|
2578 |
/*
|
2579 |
* C3.4.5 Move wide (immediate)
|
2580 |
*
|
2581 |
* 31 30 29 28 23 22 21 20 5 4 0
|
2582 |
* +--+-----+-------------+-----+----------------+------+
|
2583 |
* |sf| opc | 1 0 0 1 0 1 | hw | imm16 | Rd |
|
2584 |
* +--+-----+-------------+-----+----------------+------+
|
2585 |
*
|
2586 |
* sf: 0 -> 32 bit, 1 -> 64 bit
|
2587 |
* opc: 00 -> N, 10 -> Z, 11 -> K
|
2588 |
* hw: shift/16 (0,16, and sf only 32, 48)
|
2589 |
*/
|
2590 |
static void disas_movw_imm(DisasContext *s, uint32_t insn) |
2591 |
{ |
2592 |
int rd = extract32(insn, 0, 5); |
2593 |
uint64_t imm = extract32(insn, 5, 16); |
2594 |
int sf = extract32(insn, 31, 1); |
2595 |
int opc = extract32(insn, 29, 2); |
2596 |
int pos = extract32(insn, 21, 2) << 4; |
2597 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
2598 |
TCGv_i64 tcg_imm; |
2599 |
|
2600 |
if (!sf && (pos >= 32)) { |
2601 |
unallocated_encoding(s); |
2602 |
return;
|
2603 |
} |
2604 |
|
2605 |
switch (opc) {
|
2606 |
case 0: /* MOVN */ |
2607 |
case 2: /* MOVZ */ |
2608 |
imm <<= pos; |
2609 |
if (opc == 0) { |
2610 |
imm = ~imm; |
2611 |
} |
2612 |
if (!sf) {
|
2613 |
imm &= 0xffffffffu;
|
2614 |
} |
2615 |
tcg_gen_movi_i64(tcg_rd, imm); |
2616 |
break;
|
2617 |
case 3: /* MOVK */ |
2618 |
tcg_imm = tcg_const_i64(imm); |
2619 |
tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_imm, pos, 16);
|
2620 |
tcg_temp_free_i64(tcg_imm); |
2621 |
if (!sf) {
|
2622 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
2623 |
} |
2624 |
break;
|
2625 |
default:
|
2626 |
unallocated_encoding(s); |
2627 |
break;
|
2628 |
} |
2629 |
} |
2630 |
|
2631 |
/* C3.4.2 Bitfield
|
2632 |
* 31 30 29 28 23 22 21 16 15 10 9 5 4 0
|
2633 |
* +----+-----+-------------+---+------+------+------+------+
|
2634 |
* | sf | opc | 1 0 0 1 1 0 | N | immr | imms | Rn | Rd |
|
2635 |
* +----+-----+-------------+---+------+------+------+------+
|
2636 |
*/
|
2637 |
static void disas_bitfield(DisasContext *s, uint32_t insn) |
2638 |
{ |
2639 |
unsigned int sf, n, opc, ri, si, rn, rd, bitsize, pos, len; |
2640 |
TCGv_i64 tcg_rd, tcg_tmp; |
2641 |
|
2642 |
sf = extract32(insn, 31, 1); |
2643 |
opc = extract32(insn, 29, 2); |
2644 |
n = extract32(insn, 22, 1); |
2645 |
ri = extract32(insn, 16, 6); |
2646 |
si = extract32(insn, 10, 6); |
2647 |
rn = extract32(insn, 5, 5); |
2648 |
rd = extract32(insn, 0, 5); |
2649 |
bitsize = sf ? 64 : 32; |
2650 |
|
2651 |
if (sf != n || ri >= bitsize || si >= bitsize || opc > 2) { |
2652 |
unallocated_encoding(s); |
2653 |
return;
|
2654 |
} |
2655 |
|
2656 |
tcg_rd = cpu_reg(s, rd); |
2657 |
tcg_tmp = read_cpu_reg(s, rn, sf); |
2658 |
|
2659 |
/* OPTME: probably worth recognizing common cases of ext{8,16,32}{u,s} */
|
2660 |
|
2661 |
if (opc != 1) { /* SBFM or UBFM */ |
2662 |
tcg_gen_movi_i64(tcg_rd, 0);
|
2663 |
} |
2664 |
|
2665 |
/* do the bit move operation */
|
2666 |
if (si >= ri) {
|
2667 |
/* Wd<s-r:0> = Wn<s:r> */
|
2668 |
tcg_gen_shri_i64(tcg_tmp, tcg_tmp, ri); |
2669 |
pos = 0;
|
2670 |
len = (si - ri) + 1;
|
2671 |
} else {
|
2672 |
/* Wd<32+s-r,32-r> = Wn<s:0> */
|
2673 |
pos = bitsize - ri; |
2674 |
len = si + 1;
|
2675 |
} |
2676 |
|
2677 |
tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, pos, len); |
2678 |
|
2679 |
if (opc == 0) { /* SBFM - sign extend the destination field */ |
2680 |
tcg_gen_shli_i64(tcg_rd, tcg_rd, 64 - (pos + len));
|
2681 |
tcg_gen_sari_i64(tcg_rd, tcg_rd, 64 - (pos + len));
|
2682 |
} |
2683 |
|
2684 |
if (!sf) { /* zero extend final result */ |
2685 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
2686 |
} |
2687 |
} |
2688 |
|
2689 |
/* C3.4.3 Extract
|
2690 |
* 31 30 29 28 23 22 21 20 16 15 10 9 5 4 0
|
2691 |
* +----+------+-------------+---+----+------+--------+------+------+
|
2692 |
* | sf | op21 | 1 0 0 1 1 1 | N | o0 | Rm | imms | Rn | Rd |
|
2693 |
* +----+------+-------------+---+----+------+--------+------+------+
|
2694 |
*/
|
2695 |
static void disas_extract(DisasContext *s, uint32_t insn) |
2696 |
{ |
2697 |
unsigned int sf, n, rm, imm, rn, rd, bitsize, op21, op0; |
2698 |
|
2699 |
sf = extract32(insn, 31, 1); |
2700 |
n = extract32(insn, 22, 1); |
2701 |
rm = extract32(insn, 16, 5); |
2702 |
imm = extract32(insn, 10, 6); |
2703 |
rn = extract32(insn, 5, 5); |
2704 |
rd = extract32(insn, 0, 5); |
2705 |
op21 = extract32(insn, 29, 2); |
2706 |
op0 = extract32(insn, 21, 1); |
2707 |
bitsize = sf ? 64 : 32; |
2708 |
|
2709 |
if (sf != n || op21 || op0 || imm >= bitsize) {
|
2710 |
unallocated_encoding(s); |
2711 |
} else {
|
2712 |
TCGv_i64 tcg_rd, tcg_rm, tcg_rn; |
2713 |
|
2714 |
tcg_rd = cpu_reg(s, rd); |
2715 |
|
2716 |
if (imm) {
|
2717 |
/* OPTME: we can special case rm==rn as a rotate */
|
2718 |
tcg_rm = read_cpu_reg(s, rm, sf); |
2719 |
tcg_rn = read_cpu_reg(s, rn, sf); |
2720 |
tcg_gen_shri_i64(tcg_rm, tcg_rm, imm); |
2721 |
tcg_gen_shli_i64(tcg_rn, tcg_rn, bitsize - imm); |
2722 |
tcg_gen_or_i64(tcg_rd, tcg_rm, tcg_rn); |
2723 |
if (!sf) {
|
2724 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
2725 |
} |
2726 |
} else {
|
2727 |
/* tcg shl_i32/shl_i64 is undefined for 32/64 bit shifts,
|
2728 |
* so an extract from bit 0 is a special case.
|
2729 |
*/
|
2730 |
if (sf) {
|
2731 |
tcg_gen_mov_i64(tcg_rd, cpu_reg(s, rm)); |
2732 |
} else {
|
2733 |
tcg_gen_ext32u_i64(tcg_rd, cpu_reg(s, rm)); |
2734 |
} |
2735 |
} |
2736 |
|
2737 |
} |
2738 |
} |
2739 |
|
2740 |
/* C3.4 Data processing - immediate */
|
2741 |
static void disas_data_proc_imm(DisasContext *s, uint32_t insn) |
2742 |
{ |
2743 |
switch (extract32(insn, 23, 6)) { |
2744 |
case 0x20: case 0x21: /* PC-rel. addressing */ |
2745 |
disas_pc_rel_adr(s, insn); |
2746 |
break;
|
2747 |
case 0x22: case 0x23: /* Add/subtract (immediate) */ |
2748 |
disas_add_sub_imm(s, insn); |
2749 |
break;
|
2750 |
case 0x24: /* Logical (immediate) */ |
2751 |
disas_logic_imm(s, insn); |
2752 |
break;
|
2753 |
case 0x25: /* Move wide (immediate) */ |
2754 |
disas_movw_imm(s, insn); |
2755 |
break;
|
2756 |
case 0x26: /* Bitfield */ |
2757 |
disas_bitfield(s, insn); |
2758 |
break;
|
2759 |
case 0x27: /* Extract */ |
2760 |
disas_extract(s, insn); |
2761 |
break;
|
2762 |
default:
|
2763 |
unallocated_encoding(s); |
2764 |
break;
|
2765 |
} |
2766 |
} |
2767 |
|
2768 |
/* Shift a TCGv src by TCGv shift_amount, put result in dst.
|
2769 |
* Note that it is the caller's responsibility to ensure that the
|
2770 |
* shift amount is in range (ie 0..31 or 0..63) and provide the ARM
|
2771 |
* mandated semantics for out of range shifts.
|
2772 |
*/
|
2773 |
static void shift_reg(TCGv_i64 dst, TCGv_i64 src, int sf, |
2774 |
enum a64_shift_type shift_type, TCGv_i64 shift_amount)
|
2775 |
{ |
2776 |
switch (shift_type) {
|
2777 |
case A64_SHIFT_TYPE_LSL:
|
2778 |
tcg_gen_shl_i64(dst, src, shift_amount); |
2779 |
break;
|
2780 |
case A64_SHIFT_TYPE_LSR:
|
2781 |
tcg_gen_shr_i64(dst, src, shift_amount); |
2782 |
break;
|
2783 |
case A64_SHIFT_TYPE_ASR:
|
2784 |
if (!sf) {
|
2785 |
tcg_gen_ext32s_i64(dst, src); |
2786 |
} |
2787 |
tcg_gen_sar_i64(dst, sf ? src : dst, shift_amount); |
2788 |
break;
|
2789 |
case A64_SHIFT_TYPE_ROR:
|
2790 |
if (sf) {
|
2791 |
tcg_gen_rotr_i64(dst, src, shift_amount); |
2792 |
} else {
|
2793 |
TCGv_i32 t0, t1; |
2794 |
t0 = tcg_temp_new_i32(); |
2795 |
t1 = tcg_temp_new_i32(); |
2796 |
tcg_gen_trunc_i64_i32(t0, src); |
2797 |
tcg_gen_trunc_i64_i32(t1, shift_amount); |
2798 |
tcg_gen_rotr_i32(t0, t0, t1); |
2799 |
tcg_gen_extu_i32_i64(dst, t0); |
2800 |
tcg_temp_free_i32(t0); |
2801 |
tcg_temp_free_i32(t1); |
2802 |
} |
2803 |
break;
|
2804 |
default:
|
2805 |
assert(FALSE); /* all shift types should be handled */
|
2806 |
break;
|
2807 |
} |
2808 |
|
2809 |
if (!sf) { /* zero extend final result */ |
2810 |
tcg_gen_ext32u_i64(dst, dst); |
2811 |
} |
2812 |
} |
2813 |
|
2814 |
/* Shift a TCGv src by immediate, put result in dst.
|
2815 |
* The shift amount must be in range (this should always be true as the
|
2816 |
* relevant instructions will UNDEF on bad shift immediates).
|
2817 |
*/
|
2818 |
static void shift_reg_imm(TCGv_i64 dst, TCGv_i64 src, int sf, |
2819 |
enum a64_shift_type shift_type, unsigned int shift_i) |
2820 |
{ |
2821 |
assert(shift_i < (sf ? 64 : 32)); |
2822 |
|
2823 |
if (shift_i == 0) { |
2824 |
tcg_gen_mov_i64(dst, src); |
2825 |
} else {
|
2826 |
TCGv_i64 shift_const; |
2827 |
|
2828 |
shift_const = tcg_const_i64(shift_i); |
2829 |
shift_reg(dst, src, sf, shift_type, shift_const); |
2830 |
tcg_temp_free_i64(shift_const); |
2831 |
} |
2832 |
} |
2833 |
|
2834 |
/* C3.5.10 Logical (shifted register)
|
2835 |
* 31 30 29 28 24 23 22 21 20 16 15 10 9 5 4 0
|
2836 |
* +----+-----+-----------+-------+---+------+--------+------+------+
|
2837 |
* | sf | opc | 0 1 0 1 0 | shift | N | Rm | imm6 | Rn | Rd |
|
2838 |
* +----+-----+-----------+-------+---+------+--------+------+------+
|
2839 |
*/
|
2840 |
static void disas_logic_reg(DisasContext *s, uint32_t insn) |
2841 |
{ |
2842 |
TCGv_i64 tcg_rd, tcg_rn, tcg_rm; |
2843 |
unsigned int sf, opc, shift_type, invert, rm, shift_amount, rn, rd; |
2844 |
|
2845 |
sf = extract32(insn, 31, 1); |
2846 |
opc = extract32(insn, 29, 2); |
2847 |
shift_type = extract32(insn, 22, 2); |
2848 |
invert = extract32(insn, 21, 1); |
2849 |
rm = extract32(insn, 16, 5); |
2850 |
shift_amount = extract32(insn, 10, 6); |
2851 |
rn = extract32(insn, 5, 5); |
2852 |
rd = extract32(insn, 0, 5); |
2853 |
|
2854 |
if (!sf && (shift_amount & (1 << 5))) { |
2855 |
unallocated_encoding(s); |
2856 |
return;
|
2857 |
} |
2858 |
|
2859 |
tcg_rd = cpu_reg(s, rd); |
2860 |
|
2861 |
if (opc == 1 && shift_amount == 0 && shift_type == 0 && rn == 31) { |
2862 |
/* Unshifted ORR and ORN with WZR/XZR is the standard encoding for
|
2863 |
* register-register MOV and MVN, so it is worth special casing.
|
2864 |
*/
|
2865 |
tcg_rm = cpu_reg(s, rm); |
2866 |
if (invert) {
|
2867 |
tcg_gen_not_i64(tcg_rd, tcg_rm); |
2868 |
if (!sf) {
|
2869 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
2870 |
} |
2871 |
} else {
|
2872 |
if (sf) {
|
2873 |
tcg_gen_mov_i64(tcg_rd, tcg_rm); |
2874 |
} else {
|
2875 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rm); |
2876 |
} |
2877 |
} |
2878 |
return;
|
2879 |
} |
2880 |
|
2881 |
tcg_rm = read_cpu_reg(s, rm, sf); |
2882 |
|
2883 |
if (shift_amount) {
|
2884 |
shift_reg_imm(tcg_rm, tcg_rm, sf, shift_type, shift_amount); |
2885 |
} |
2886 |
|
2887 |
tcg_rn = cpu_reg(s, rn); |
2888 |
|
2889 |
switch (opc | (invert << 2)) { |
2890 |
case 0: /* AND */ |
2891 |
case 3: /* ANDS */ |
2892 |
tcg_gen_and_i64(tcg_rd, tcg_rn, tcg_rm); |
2893 |
break;
|
2894 |
case 1: /* ORR */ |
2895 |
tcg_gen_or_i64(tcg_rd, tcg_rn, tcg_rm); |
2896 |
break;
|
2897 |
case 2: /* EOR */ |
2898 |
tcg_gen_xor_i64(tcg_rd, tcg_rn, tcg_rm); |
2899 |
break;
|
2900 |
case 4: /* BIC */ |
2901 |
case 7: /* BICS */ |
2902 |
tcg_gen_andc_i64(tcg_rd, tcg_rn, tcg_rm); |
2903 |
break;
|
2904 |
case 5: /* ORN */ |
2905 |
tcg_gen_orc_i64(tcg_rd, tcg_rn, tcg_rm); |
2906 |
break;
|
2907 |
case 6: /* EON */ |
2908 |
tcg_gen_eqv_i64(tcg_rd, tcg_rn, tcg_rm); |
2909 |
break;
|
2910 |
default:
|
2911 |
assert(FALSE); |
2912 |
break;
|
2913 |
} |
2914 |
|
2915 |
if (!sf) {
|
2916 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
2917 |
} |
2918 |
|
2919 |
if (opc == 3) { |
2920 |
gen_logic_CC(sf, tcg_rd); |
2921 |
} |
2922 |
} |
2923 |
|
2924 |
/*
|
2925 |
* C3.5.1 Add/subtract (extended register)
|
2926 |
*
|
2927 |
* 31|30|29|28 24|23 22|21|20 16|15 13|12 10|9 5|4 0|
|
2928 |
* +--+--+--+-----------+-----+--+-------+------+------+----+----+
|
2929 |
* |sf|op| S| 0 1 0 1 1 | opt | 1| Rm |option| imm3 | Rn | Rd |
|
2930 |
* +--+--+--+-----------+-----+--+-------+------+------+----+----+
|
2931 |
*
|
2932 |
* sf: 0 -> 32bit, 1 -> 64bit
|
2933 |
* op: 0 -> add , 1 -> sub
|
2934 |
* S: 1 -> set flags
|
2935 |
* opt: 00
|
2936 |
* option: extension type (see DecodeRegExtend)
|
2937 |
* imm3: optional shift to Rm
|
2938 |
*
|
2939 |
* Rd = Rn + LSL(extend(Rm), amount)
|
2940 |
*/
|
2941 |
static void disas_add_sub_ext_reg(DisasContext *s, uint32_t insn) |
2942 |
{ |
2943 |
int rd = extract32(insn, 0, 5); |
2944 |
int rn = extract32(insn, 5, 5); |
2945 |
int imm3 = extract32(insn, 10, 3); |
2946 |
int option = extract32(insn, 13, 3); |
2947 |
int rm = extract32(insn, 16, 5); |
2948 |
bool setflags = extract32(insn, 29, 1); |
2949 |
bool sub_op = extract32(insn, 30, 1); |
2950 |
bool sf = extract32(insn, 31, 1); |
2951 |
|
2952 |
TCGv_i64 tcg_rm, tcg_rn; /* temps */
|
2953 |
TCGv_i64 tcg_rd; |
2954 |
TCGv_i64 tcg_result; |
2955 |
|
2956 |
if (imm3 > 4) { |
2957 |
unallocated_encoding(s); |
2958 |
return;
|
2959 |
} |
2960 |
|
2961 |
/* non-flag setting ops may use SP */
|
2962 |
if (!setflags) {
|
2963 |
tcg_rn = read_cpu_reg_sp(s, rn, sf); |
2964 |
tcg_rd = cpu_reg_sp(s, rd); |
2965 |
} else {
|
2966 |
tcg_rn = read_cpu_reg(s, rn, sf); |
2967 |
tcg_rd = cpu_reg(s, rd); |
2968 |
} |
2969 |
|
2970 |
tcg_rm = read_cpu_reg(s, rm, sf); |
2971 |
ext_and_shift_reg(tcg_rm, tcg_rm, option, imm3); |
2972 |
|
2973 |
tcg_result = tcg_temp_new_i64(); |
2974 |
|
2975 |
if (!setflags) {
|
2976 |
if (sub_op) {
|
2977 |
tcg_gen_sub_i64(tcg_result, tcg_rn, tcg_rm); |
2978 |
} else {
|
2979 |
tcg_gen_add_i64(tcg_result, tcg_rn, tcg_rm); |
2980 |
} |
2981 |
} else {
|
2982 |
if (sub_op) {
|
2983 |
gen_sub_CC(sf, tcg_result, tcg_rn, tcg_rm); |
2984 |
} else {
|
2985 |
gen_add_CC(sf, tcg_result, tcg_rn, tcg_rm); |
2986 |
} |
2987 |
} |
2988 |
|
2989 |
if (sf) {
|
2990 |
tcg_gen_mov_i64(tcg_rd, tcg_result); |
2991 |
} else {
|
2992 |
tcg_gen_ext32u_i64(tcg_rd, tcg_result); |
2993 |
} |
2994 |
|
2995 |
tcg_temp_free_i64(tcg_result); |
2996 |
} |
2997 |
|
2998 |
/*
|
2999 |
* C3.5.2 Add/subtract (shifted register)
|
3000 |
*
|
3001 |
* 31 30 29 28 24 23 22 21 20 16 15 10 9 5 4 0
|
3002 |
* +--+--+--+-----------+-----+--+-------+---------+------+------+
|
3003 |
* |sf|op| S| 0 1 0 1 1 |shift| 0| Rm | imm6 | Rn | Rd |
|
3004 |
* +--+--+--+-----------+-----+--+-------+---------+------+------+
|
3005 |
*
|
3006 |
* sf: 0 -> 32bit, 1 -> 64bit
|
3007 |
* op: 0 -> add , 1 -> sub
|
3008 |
* S: 1 -> set flags
|
3009 |
* shift: 00 -> LSL, 01 -> LSR, 10 -> ASR, 11 -> RESERVED
|
3010 |
* imm6: Shift amount to apply to Rm before the add/sub
|
3011 |
*/
|
3012 |
static void disas_add_sub_reg(DisasContext *s, uint32_t insn) |
3013 |
{ |
3014 |
int rd = extract32(insn, 0, 5); |
3015 |
int rn = extract32(insn, 5, 5); |
3016 |
int imm6 = extract32(insn, 10, 6); |
3017 |
int rm = extract32(insn, 16, 5); |
3018 |
int shift_type = extract32(insn, 22, 2); |
3019 |
bool setflags = extract32(insn, 29, 1); |
3020 |
bool sub_op = extract32(insn, 30, 1); |
3021 |
bool sf = extract32(insn, 31, 1); |
3022 |
|
3023 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
3024 |
TCGv_i64 tcg_rn, tcg_rm; |
3025 |
TCGv_i64 tcg_result; |
3026 |
|
3027 |
if ((shift_type == 3) || (!sf && (imm6 > 31))) { |
3028 |
unallocated_encoding(s); |
3029 |
return;
|
3030 |
} |
3031 |
|
3032 |
tcg_rn = read_cpu_reg(s, rn, sf); |
3033 |
tcg_rm = read_cpu_reg(s, rm, sf); |
3034 |
|
3035 |
shift_reg_imm(tcg_rm, tcg_rm, sf, shift_type, imm6); |
3036 |
|
3037 |
tcg_result = tcg_temp_new_i64(); |
3038 |
|
3039 |
if (!setflags) {
|
3040 |
if (sub_op) {
|
3041 |
tcg_gen_sub_i64(tcg_result, tcg_rn, tcg_rm); |
3042 |
} else {
|
3043 |
tcg_gen_add_i64(tcg_result, tcg_rn, tcg_rm); |
3044 |
} |
3045 |
} else {
|
3046 |
if (sub_op) {
|
3047 |
gen_sub_CC(sf, tcg_result, tcg_rn, tcg_rm); |
3048 |
} else {
|
3049 |
gen_add_CC(sf, tcg_result, tcg_rn, tcg_rm); |
3050 |
} |
3051 |
} |
3052 |
|
3053 |
if (sf) {
|
3054 |
tcg_gen_mov_i64(tcg_rd, tcg_result); |
3055 |
} else {
|
3056 |
tcg_gen_ext32u_i64(tcg_rd, tcg_result); |
3057 |
} |
3058 |
|
3059 |
tcg_temp_free_i64(tcg_result); |
3060 |
} |
3061 |
|
3062 |
/* C3.5.9 Data-processing (3 source)
|
3063 |
|
3064 |
31 30 29 28 24 23 21 20 16 15 14 10 9 5 4 0
|
3065 |
+--+------+-----------+------+------+----+------+------+------+
|
3066 |
|sf| op54 | 1 1 0 1 1 | op31 | Rm | o0 | Ra | Rn | Rd |
|
3067 |
+--+------+-----------+------+------+----+------+------+------+
|
3068 |
|
3069 |
*/
|
3070 |
static void disas_data_proc_3src(DisasContext *s, uint32_t insn) |
3071 |
{ |
3072 |
int rd = extract32(insn, 0, 5); |
3073 |
int rn = extract32(insn, 5, 5); |
3074 |
int ra = extract32(insn, 10, 5); |
3075 |
int rm = extract32(insn, 16, 5); |
3076 |
int op_id = (extract32(insn, 29, 3) << 4) | |
3077 |
(extract32(insn, 21, 3) << 1) | |
3078 |
extract32(insn, 15, 1); |
3079 |
bool sf = extract32(insn, 31, 1); |
3080 |
bool is_sub = extract32(op_id, 0, 1); |
3081 |
bool is_high = extract32(op_id, 2, 1); |
3082 |
bool is_signed = false; |
3083 |
TCGv_i64 tcg_op1; |
3084 |
TCGv_i64 tcg_op2; |
3085 |
TCGv_i64 tcg_tmp; |
3086 |
|
3087 |
/* Note that op_id is sf:op54:op31:o0 so it includes the 32/64 size flag */
|
3088 |
switch (op_id) {
|
3089 |
case 0x42: /* SMADDL */ |
3090 |
case 0x43: /* SMSUBL */ |
3091 |
case 0x44: /* SMULH */ |
3092 |
is_signed = true;
|
3093 |
break;
|
3094 |
case 0x0: /* MADD (32bit) */ |
3095 |
case 0x1: /* MSUB (32bit) */ |
3096 |
case 0x40: /* MADD (64bit) */ |
3097 |
case 0x41: /* MSUB (64bit) */ |
3098 |
case 0x4a: /* UMADDL */ |
3099 |
case 0x4b: /* UMSUBL */ |
3100 |
case 0x4c: /* UMULH */ |
3101 |
break;
|
3102 |
default:
|
3103 |
unallocated_encoding(s); |
3104 |
return;
|
3105 |
} |
3106 |
|
3107 |
if (is_high) {
|
3108 |
TCGv_i64 low_bits = tcg_temp_new_i64(); /* low bits discarded */
|
3109 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
3110 |
TCGv_i64 tcg_rn = cpu_reg(s, rn); |
3111 |
TCGv_i64 tcg_rm = cpu_reg(s, rm); |
3112 |
|
3113 |
if (is_signed) {
|
3114 |
tcg_gen_muls2_i64(low_bits, tcg_rd, tcg_rn, tcg_rm); |
3115 |
} else {
|
3116 |
tcg_gen_mulu2_i64(low_bits, tcg_rd, tcg_rn, tcg_rm); |
3117 |
} |
3118 |
|
3119 |
tcg_temp_free_i64(low_bits); |
3120 |
return;
|
3121 |
} |
3122 |
|
3123 |
tcg_op1 = tcg_temp_new_i64(); |
3124 |
tcg_op2 = tcg_temp_new_i64(); |
3125 |
tcg_tmp = tcg_temp_new_i64(); |
3126 |
|
3127 |
if (op_id < 0x42) { |
3128 |
tcg_gen_mov_i64(tcg_op1, cpu_reg(s, rn)); |
3129 |
tcg_gen_mov_i64(tcg_op2, cpu_reg(s, rm)); |
3130 |
} else {
|
3131 |
if (is_signed) {
|
3132 |
tcg_gen_ext32s_i64(tcg_op1, cpu_reg(s, rn)); |
3133 |
tcg_gen_ext32s_i64(tcg_op2, cpu_reg(s, rm)); |
3134 |
} else {
|
3135 |
tcg_gen_ext32u_i64(tcg_op1, cpu_reg(s, rn)); |
3136 |
tcg_gen_ext32u_i64(tcg_op2, cpu_reg(s, rm)); |
3137 |
} |
3138 |
} |
3139 |
|
3140 |
if (ra == 31 && !is_sub) { |
3141 |
/* Special-case MADD with rA == XZR; it is the standard MUL alias */
|
3142 |
tcg_gen_mul_i64(cpu_reg(s, rd), tcg_op1, tcg_op2); |
3143 |
} else {
|
3144 |
tcg_gen_mul_i64(tcg_tmp, tcg_op1, tcg_op2); |
3145 |
if (is_sub) {
|
3146 |
tcg_gen_sub_i64(cpu_reg(s, rd), cpu_reg(s, ra), tcg_tmp); |
3147 |
} else {
|
3148 |
tcg_gen_add_i64(cpu_reg(s, rd), cpu_reg(s, ra), tcg_tmp); |
3149 |
} |
3150 |
} |
3151 |
|
3152 |
if (!sf) {
|
3153 |
tcg_gen_ext32u_i64(cpu_reg(s, rd), cpu_reg(s, rd)); |
3154 |
} |
3155 |
|
3156 |
tcg_temp_free_i64(tcg_op1); |
3157 |
tcg_temp_free_i64(tcg_op2); |
3158 |
tcg_temp_free_i64(tcg_tmp); |
3159 |
} |
3160 |
|
3161 |
/* C3.5.3 - Add/subtract (with carry)
|
3162 |
* 31 30 29 28 27 26 25 24 23 22 21 20 16 15 10 9 5 4 0
|
3163 |
* +--+--+--+------------------------+------+---------+------+-----+
|
3164 |
* |sf|op| S| 1 1 0 1 0 0 0 0 | rm | opcode2 | Rn | Rd |
|
3165 |
* +--+--+--+------------------------+------+---------+------+-----+
|
3166 |
* [000000]
|
3167 |
*/
|
3168 |
|
3169 |
static void disas_adc_sbc(DisasContext *s, uint32_t insn) |
3170 |
{ |
3171 |
unsigned int sf, op, setflags, rm, rn, rd; |
3172 |
TCGv_i64 tcg_y, tcg_rn, tcg_rd; |
3173 |
|
3174 |
if (extract32(insn, 10, 6) != 0) { |
3175 |
unallocated_encoding(s); |
3176 |
return;
|
3177 |
} |
3178 |
|
3179 |
sf = extract32(insn, 31, 1); |
3180 |
op = extract32(insn, 30, 1); |
3181 |
setflags = extract32(insn, 29, 1); |
3182 |
rm = extract32(insn, 16, 5); |
3183 |
rn = extract32(insn, 5, 5); |
3184 |
rd = extract32(insn, 0, 5); |
3185 |
|
3186 |
tcg_rd = cpu_reg(s, rd); |
3187 |
tcg_rn = cpu_reg(s, rn); |
3188 |
|
3189 |
if (op) {
|
3190 |
tcg_y = new_tmp_a64(s); |
3191 |
tcg_gen_not_i64(tcg_y, cpu_reg(s, rm)); |
3192 |
} else {
|
3193 |
tcg_y = cpu_reg(s, rm); |
3194 |
} |
3195 |
|
3196 |
if (setflags) {
|
3197 |
gen_adc_CC(sf, tcg_rd, tcg_rn, tcg_y); |
3198 |
} else {
|
3199 |
gen_adc(sf, tcg_rd, tcg_rn, tcg_y); |
3200 |
} |
3201 |
} |
3202 |
|
3203 |
/* C3.5.4 - C3.5.5 Conditional compare (immediate / register)
|
3204 |
* 31 30 29 28 27 26 25 24 23 22 21 20 16 15 12 11 10 9 5 4 3 0
|
3205 |
* +--+--+--+------------------------+--------+------+----+--+------+--+-----+
|
3206 |
* |sf|op| S| 1 1 0 1 0 0 1 0 |imm5/rm | cond |i/r |o2| Rn |o3|nzcv |
|
3207 |
* +--+--+--+------------------------+--------+------+----+--+------+--+-----+
|
3208 |
* [1] y [0] [0]
|
3209 |
*/
|
3210 |
static void disas_cc(DisasContext *s, uint32_t insn) |
3211 |
{ |
3212 |
unsigned int sf, op, y, cond, rn, nzcv, is_imm; |
3213 |
int label_continue = -1; |
3214 |
TCGv_i64 tcg_tmp, tcg_y, tcg_rn; |
3215 |
|
3216 |
if (!extract32(insn, 29, 1)) { |
3217 |
unallocated_encoding(s); |
3218 |
return;
|
3219 |
} |
3220 |
if (insn & (1 << 10 | 1 << 4)) { |
3221 |
unallocated_encoding(s); |
3222 |
return;
|
3223 |
} |
3224 |
sf = extract32(insn, 31, 1); |
3225 |
op = extract32(insn, 30, 1); |
3226 |
is_imm = extract32(insn, 11, 1); |
3227 |
y = extract32(insn, 16, 5); /* y = rm (reg) or imm5 (imm) */ |
3228 |
cond = extract32(insn, 12, 4); |
3229 |
rn = extract32(insn, 5, 5); |
3230 |
nzcv = extract32(insn, 0, 4); |
3231 |
|
3232 |
if (cond < 0x0e) { /* not always */ |
3233 |
int label_match = gen_new_label();
|
3234 |
label_continue = gen_new_label(); |
3235 |
arm_gen_test_cc(cond, label_match); |
3236 |
/* nomatch: */
|
3237 |
tcg_tmp = tcg_temp_new_i64(); |
3238 |
tcg_gen_movi_i64(tcg_tmp, nzcv << 28);
|
3239 |
gen_set_nzcv(tcg_tmp); |
3240 |
tcg_temp_free_i64(tcg_tmp); |
3241 |
tcg_gen_br(label_continue); |
3242 |
gen_set_label(label_match); |
3243 |
} |
3244 |
/* match, or condition is always */
|
3245 |
if (is_imm) {
|
3246 |
tcg_y = new_tmp_a64(s); |
3247 |
tcg_gen_movi_i64(tcg_y, y); |
3248 |
} else {
|
3249 |
tcg_y = cpu_reg(s, y); |
3250 |
} |
3251 |
tcg_rn = cpu_reg(s, rn); |
3252 |
|
3253 |
tcg_tmp = tcg_temp_new_i64(); |
3254 |
if (op) {
|
3255 |
gen_sub_CC(sf, tcg_tmp, tcg_rn, tcg_y); |
3256 |
} else {
|
3257 |
gen_add_CC(sf, tcg_tmp, tcg_rn, tcg_y); |
3258 |
} |
3259 |
tcg_temp_free_i64(tcg_tmp); |
3260 |
|
3261 |
if (cond < 0x0e) { /* continue */ |
3262 |
gen_set_label(label_continue); |
3263 |
} |
3264 |
} |
3265 |
|
3266 |
/* C3.5.6 Conditional select
|
3267 |
* 31 30 29 28 21 20 16 15 12 11 10 9 5 4 0
|
3268 |
* +----+----+---+-----------------+------+------+-----+------+------+
|
3269 |
* | sf | op | S | 1 1 0 1 0 1 0 0 | Rm | cond | op2 | Rn | Rd |
|
3270 |
* +----+----+---+-----------------+------+------+-----+------+------+
|
3271 |
*/
|
3272 |
static void disas_cond_select(DisasContext *s, uint32_t insn) |
3273 |
{ |
3274 |
unsigned int sf, else_inv, rm, cond, else_inc, rn, rd; |
3275 |
TCGv_i64 tcg_rd, tcg_src; |
3276 |
|
3277 |
if (extract32(insn, 29, 1) || extract32(insn, 11, 1)) { |
3278 |
/* S == 1 or op2<1> == 1 */
|
3279 |
unallocated_encoding(s); |
3280 |
return;
|
3281 |
} |
3282 |
sf = extract32(insn, 31, 1); |
3283 |
else_inv = extract32(insn, 30, 1); |
3284 |
rm = extract32(insn, 16, 5); |
3285 |
cond = extract32(insn, 12, 4); |
3286 |
else_inc = extract32(insn, 10, 1); |
3287 |
rn = extract32(insn, 5, 5); |
3288 |
rd = extract32(insn, 0, 5); |
3289 |
|
3290 |
if (rd == 31) { |
3291 |
/* silly no-op write; until we use movcond we must special-case
|
3292 |
* this to avoid a dead temporary across basic blocks.
|
3293 |
*/
|
3294 |
return;
|
3295 |
} |
3296 |
|
3297 |
tcg_rd = cpu_reg(s, rd); |
3298 |
|
3299 |
if (cond >= 0x0e) { /* condition "always" */ |
3300 |
tcg_src = read_cpu_reg(s, rn, sf); |
3301 |
tcg_gen_mov_i64(tcg_rd, tcg_src); |
3302 |
} else {
|
3303 |
/* OPTME: we could use movcond here, at the cost of duplicating
|
3304 |
* a lot of the arm_gen_test_cc() logic.
|
3305 |
*/
|
3306 |
int label_match = gen_new_label();
|
3307 |
int label_continue = gen_new_label();
|
3308 |
|
3309 |
arm_gen_test_cc(cond, label_match); |
3310 |
/* nomatch: */
|
3311 |
tcg_src = cpu_reg(s, rm); |
3312 |
|
3313 |
if (else_inv && else_inc) {
|
3314 |
tcg_gen_neg_i64(tcg_rd, tcg_src); |
3315 |
} else if (else_inv) { |
3316 |
tcg_gen_not_i64(tcg_rd, tcg_src); |
3317 |
} else if (else_inc) { |
3318 |
tcg_gen_addi_i64(tcg_rd, tcg_src, 1);
|
3319 |
} else {
|
3320 |
tcg_gen_mov_i64(tcg_rd, tcg_src); |
3321 |
} |
3322 |
if (!sf) {
|
3323 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
3324 |
} |
3325 |
tcg_gen_br(label_continue); |
3326 |
/* match: */
|
3327 |
gen_set_label(label_match); |
3328 |
tcg_src = read_cpu_reg(s, rn, sf); |
3329 |
tcg_gen_mov_i64(tcg_rd, tcg_src); |
3330 |
/* continue: */
|
3331 |
gen_set_label(label_continue); |
3332 |
} |
3333 |
} |
3334 |
|
3335 |
static void handle_clz(DisasContext *s, unsigned int sf, |
3336 |
unsigned int rn, unsigned int rd) |
3337 |
{ |
3338 |
TCGv_i64 tcg_rd, tcg_rn; |
3339 |
tcg_rd = cpu_reg(s, rd); |
3340 |
tcg_rn = cpu_reg(s, rn); |
3341 |
|
3342 |
if (sf) {
|
3343 |
gen_helper_clz64(tcg_rd, tcg_rn); |
3344 |
} else {
|
3345 |
TCGv_i32 tcg_tmp32 = tcg_temp_new_i32(); |
3346 |
tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn); |
3347 |
gen_helper_clz(tcg_tmp32, tcg_tmp32); |
3348 |
tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32); |
3349 |
tcg_temp_free_i32(tcg_tmp32); |
3350 |
} |
3351 |
} |
3352 |
|
3353 |
static void handle_cls(DisasContext *s, unsigned int sf, |
3354 |
unsigned int rn, unsigned int rd) |
3355 |
{ |
3356 |
TCGv_i64 tcg_rd, tcg_rn; |
3357 |
tcg_rd = cpu_reg(s, rd); |
3358 |
tcg_rn = cpu_reg(s, rn); |
3359 |
|
3360 |
if (sf) {
|
3361 |
gen_helper_cls64(tcg_rd, tcg_rn); |
3362 |
} else {
|
3363 |
TCGv_i32 tcg_tmp32 = tcg_temp_new_i32(); |
3364 |
tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn); |
3365 |
gen_helper_cls32(tcg_tmp32, tcg_tmp32); |
3366 |
tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32); |
3367 |
tcg_temp_free_i32(tcg_tmp32); |
3368 |
} |
3369 |
} |
3370 |
|
3371 |
static void handle_rbit(DisasContext *s, unsigned int sf, |
3372 |
unsigned int rn, unsigned int rd) |
3373 |
{ |
3374 |
TCGv_i64 tcg_rd, tcg_rn; |
3375 |
tcg_rd = cpu_reg(s, rd); |
3376 |
tcg_rn = cpu_reg(s, rn); |
3377 |
|
3378 |
if (sf) {
|
3379 |
gen_helper_rbit64(tcg_rd, tcg_rn); |
3380 |
} else {
|
3381 |
TCGv_i32 tcg_tmp32 = tcg_temp_new_i32(); |
3382 |
tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn); |
3383 |
gen_helper_rbit(tcg_tmp32, tcg_tmp32); |
3384 |
tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32); |
3385 |
tcg_temp_free_i32(tcg_tmp32); |
3386 |
} |
3387 |
} |
3388 |
|
3389 |
/* C5.6.149 REV with sf==1, opcode==3 ("REV64") */
|
3390 |
static void handle_rev64(DisasContext *s, unsigned int sf, |
3391 |
unsigned int rn, unsigned int rd) |
3392 |
{ |
3393 |
if (!sf) {
|
3394 |
unallocated_encoding(s); |
3395 |
return;
|
3396 |
} |
3397 |
tcg_gen_bswap64_i64(cpu_reg(s, rd), cpu_reg(s, rn)); |
3398 |
} |
3399 |
|
3400 |
/* C5.6.149 REV with sf==0, opcode==2
|
3401 |
* C5.6.151 REV32 (sf==1, opcode==2)
|
3402 |
*/
|
3403 |
static void handle_rev32(DisasContext *s, unsigned int sf, |
3404 |
unsigned int rn, unsigned int rd) |
3405 |
{ |
3406 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
3407 |
|
3408 |
if (sf) {
|
3409 |
TCGv_i64 tcg_tmp = tcg_temp_new_i64(); |
3410 |
TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf); |
3411 |
|
3412 |
/* bswap32_i64 requires zero high word */
|
3413 |
tcg_gen_ext32u_i64(tcg_tmp, tcg_rn); |
3414 |
tcg_gen_bswap32_i64(tcg_rd, tcg_tmp); |
3415 |
tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
|
3416 |
tcg_gen_bswap32_i64(tcg_tmp, tcg_tmp); |
3417 |
tcg_gen_concat32_i64(tcg_rd, tcg_rd, tcg_tmp); |
3418 |
|
3419 |
tcg_temp_free_i64(tcg_tmp); |
3420 |
} else {
|
3421 |
tcg_gen_ext32u_i64(tcg_rd, cpu_reg(s, rn)); |
3422 |
tcg_gen_bswap32_i64(tcg_rd, tcg_rd); |
3423 |
} |
3424 |
} |
3425 |
|
3426 |
/* C5.6.150 REV16 (opcode==1) */
|
3427 |
static void handle_rev16(DisasContext *s, unsigned int sf, |
3428 |
unsigned int rn, unsigned int rd) |
3429 |
{ |
3430 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
3431 |
TCGv_i64 tcg_tmp = tcg_temp_new_i64(); |
3432 |
TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf); |
3433 |
|
3434 |
tcg_gen_andi_i64(tcg_tmp, tcg_rn, 0xffff);
|
3435 |
tcg_gen_bswap16_i64(tcg_rd, tcg_tmp); |
3436 |
|
3437 |
tcg_gen_shri_i64(tcg_tmp, tcg_rn, 16);
|
3438 |
tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
|
3439 |
tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp); |
3440 |
tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 16, 16); |
3441 |
|
3442 |
if (sf) {
|
3443 |
tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
|
3444 |
tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
|
3445 |
tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp); |
3446 |
tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 32, 16); |
3447 |
|
3448 |
tcg_gen_shri_i64(tcg_tmp, tcg_rn, 48);
|
3449 |
tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp); |
3450 |
tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 48, 16); |
3451 |
} |
3452 |
|
3453 |
tcg_temp_free_i64(tcg_tmp); |
3454 |
} |
3455 |
|
3456 |
/* C3.5.7 Data-processing (1 source)
|
3457 |
* 31 30 29 28 21 20 16 15 10 9 5 4 0
|
3458 |
* +----+---+---+-----------------+---------+--------+------+------+
|
3459 |
* | sf | 1 | S | 1 1 0 1 0 1 1 0 | opcode2 | opcode | Rn | Rd |
|
3460 |
* +----+---+---+-----------------+---------+--------+------+------+
|
3461 |
*/
|
3462 |
static void disas_data_proc_1src(DisasContext *s, uint32_t insn) |
3463 |
{ |
3464 |
unsigned int sf, opcode, rn, rd; |
3465 |
|
3466 |
if (extract32(insn, 29, 1) || extract32(insn, 16, 5)) { |
3467 |
unallocated_encoding(s); |
3468 |
return;
|
3469 |
} |
3470 |
|
3471 |
sf = extract32(insn, 31, 1); |
3472 |
opcode = extract32(insn, 10, 6); |
3473 |
rn = extract32(insn, 5, 5); |
3474 |
rd = extract32(insn, 0, 5); |
3475 |
|
3476 |
switch (opcode) {
|
3477 |
case 0: /* RBIT */ |
3478 |
handle_rbit(s, sf, rn, rd); |
3479 |
break;
|
3480 |
case 1: /* REV16 */ |
3481 |
handle_rev16(s, sf, rn, rd); |
3482 |
break;
|
3483 |
case 2: /* REV32 */ |
3484 |
handle_rev32(s, sf, rn, rd); |
3485 |
break;
|
3486 |
case 3: /* REV64 */ |
3487 |
handle_rev64(s, sf, rn, rd); |
3488 |
break;
|
3489 |
case 4: /* CLZ */ |
3490 |
handle_clz(s, sf, rn, rd); |
3491 |
break;
|
3492 |
case 5: /* CLS */ |
3493 |
handle_cls(s, sf, rn, rd); |
3494 |
break;
|
3495 |
} |
3496 |
} |
3497 |
|
3498 |
static void handle_div(DisasContext *s, bool is_signed, unsigned int sf, |
3499 |
unsigned int rm, unsigned int rn, unsigned int rd) |
3500 |
{ |
3501 |
TCGv_i64 tcg_n, tcg_m, tcg_rd; |
3502 |
tcg_rd = cpu_reg(s, rd); |
3503 |
|
3504 |
if (!sf && is_signed) {
|
3505 |
tcg_n = new_tmp_a64(s); |
3506 |
tcg_m = new_tmp_a64(s); |
3507 |
tcg_gen_ext32s_i64(tcg_n, cpu_reg(s, rn)); |
3508 |
tcg_gen_ext32s_i64(tcg_m, cpu_reg(s, rm)); |
3509 |
} else {
|
3510 |
tcg_n = read_cpu_reg(s, rn, sf); |
3511 |
tcg_m = read_cpu_reg(s, rm, sf); |
3512 |
} |
3513 |
|
3514 |
if (is_signed) {
|
3515 |
gen_helper_sdiv64(tcg_rd, tcg_n, tcg_m); |
3516 |
} else {
|
3517 |
gen_helper_udiv64(tcg_rd, tcg_n, tcg_m); |
3518 |
} |
3519 |
|
3520 |
if (!sf) { /* zero extend final result */ |
3521 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
3522 |
} |
3523 |
} |
3524 |
|
3525 |
/* C5.6.115 LSLV, C5.6.118 LSRV, C5.6.17 ASRV, C5.6.154 RORV */
|
3526 |
static void handle_shift_reg(DisasContext *s, |
3527 |
enum a64_shift_type shift_type, unsigned int sf, |
3528 |
unsigned int rm, unsigned int rn, unsigned int rd) |
3529 |
{ |
3530 |
TCGv_i64 tcg_shift = tcg_temp_new_i64(); |
3531 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
3532 |
TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf); |
3533 |
|
3534 |
tcg_gen_andi_i64(tcg_shift, cpu_reg(s, rm), sf ? 63 : 31); |
3535 |
shift_reg(tcg_rd, tcg_rn, sf, shift_type, tcg_shift); |
3536 |
tcg_temp_free_i64(tcg_shift); |
3537 |
} |
3538 |
|
3539 |
/* C3.5.8 Data-processing (2 source)
|
3540 |
* 31 30 29 28 21 20 16 15 10 9 5 4 0
|
3541 |
* +----+---+---+-----------------+------+--------+------+------+
|
3542 |
* | sf | 0 | S | 1 1 0 1 0 1 1 0 | Rm | opcode | Rn | Rd |
|
3543 |
* +----+---+---+-----------------+------+--------+------+------+
|
3544 |
*/
|
3545 |
static void disas_data_proc_2src(DisasContext *s, uint32_t insn) |
3546 |
{ |
3547 |
unsigned int sf, rm, opcode, rn, rd; |
3548 |
sf = extract32(insn, 31, 1); |
3549 |
rm = extract32(insn, 16, 5); |
3550 |
opcode = extract32(insn, 10, 6); |
3551 |
rn = extract32(insn, 5, 5); |
3552 |
rd = extract32(insn, 0, 5); |
3553 |
|
3554 |
if (extract32(insn, 29, 1)) { |
3555 |
unallocated_encoding(s); |
3556 |
return;
|
3557 |
} |
3558 |
|
3559 |
switch (opcode) {
|
3560 |
case 2: /* UDIV */ |
3561 |
handle_div(s, false, sf, rm, rn, rd);
|
3562 |
break;
|
3563 |
case 3: /* SDIV */ |
3564 |
handle_div(s, true, sf, rm, rn, rd);
|
3565 |
break;
|
3566 |
case 8: /* LSLV */ |
3567 |
handle_shift_reg(s, A64_SHIFT_TYPE_LSL, sf, rm, rn, rd); |
3568 |
break;
|
3569 |
case 9: /* LSRV */ |
3570 |
handle_shift_reg(s, A64_SHIFT_TYPE_LSR, sf, rm, rn, rd); |
3571 |
break;
|
3572 |
case 10: /* ASRV */ |
3573 |
handle_shift_reg(s, A64_SHIFT_TYPE_ASR, sf, rm, rn, rd); |
3574 |
break;
|
3575 |
case 11: /* RORV */ |
3576 |
handle_shift_reg(s, A64_SHIFT_TYPE_ROR, sf, rm, rn, rd); |
3577 |
break;
|
3578 |
case 16: |
3579 |
case 17: |
3580 |
case 18: |
3581 |
case 19: |
3582 |
case 20: |
3583 |
case 21: |
3584 |
case 22: |
3585 |
case 23: /* CRC32 */ |
3586 |
unsupported_encoding(s, insn); |
3587 |
break;
|
3588 |
default:
|
3589 |
unallocated_encoding(s); |
3590 |
break;
|
3591 |
} |
3592 |
} |
3593 |
|
3594 |
/* C3.5 Data processing - register */
|
3595 |
static void disas_data_proc_reg(DisasContext *s, uint32_t insn) |
3596 |
{ |
3597 |
switch (extract32(insn, 24, 5)) { |
3598 |
case 0x0a: /* Logical (shifted register) */ |
3599 |
disas_logic_reg(s, insn); |
3600 |
break;
|
3601 |
case 0x0b: /* Add/subtract */ |
3602 |
if (insn & (1 << 21)) { /* (extended register) */ |
3603 |
disas_add_sub_ext_reg(s, insn); |
3604 |
} else {
|
3605 |
disas_add_sub_reg(s, insn); |
3606 |
} |
3607 |
break;
|
3608 |
case 0x1b: /* Data-processing (3 source) */ |
3609 |
disas_data_proc_3src(s, insn); |
3610 |
break;
|
3611 |
case 0x1a: |
3612 |
switch (extract32(insn, 21, 3)) { |
3613 |
case 0x0: /* Add/subtract (with carry) */ |
3614 |
disas_adc_sbc(s, insn); |
3615 |
break;
|
3616 |
case 0x2: /* Conditional compare */ |
3617 |
disas_cc(s, insn); /* both imm and reg forms */
|
3618 |
break;
|
3619 |
case 0x4: /* Conditional select */ |
3620 |
disas_cond_select(s, insn); |
3621 |
break;
|
3622 |
case 0x6: /* Data-processing */ |
3623 |
if (insn & (1 << 30)) { /* (1 source) */ |
3624 |
disas_data_proc_1src(s, insn); |
3625 |
} else { /* (2 source) */ |
3626 |
disas_data_proc_2src(s, insn); |
3627 |
} |
3628 |
break;
|
3629 |
default:
|
3630 |
unallocated_encoding(s); |
3631 |
break;
|
3632 |
} |
3633 |
break;
|
3634 |
default:
|
3635 |
unallocated_encoding(s); |
3636 |
break;
|
3637 |
} |
3638 |
} |
3639 |
|
3640 |
static void handle_fp_compare(DisasContext *s, bool is_double, |
3641 |
unsigned int rn, unsigned int rm, |
3642 |
bool cmp_with_zero, bool signal_all_nans) |
3643 |
{ |
3644 |
TCGv_i64 tcg_flags = tcg_temp_new_i64(); |
3645 |
TCGv_ptr fpst = get_fpstatus_ptr(); |
3646 |
|
3647 |
if (is_double) {
|
3648 |
TCGv_i64 tcg_vn, tcg_vm; |
3649 |
|
3650 |
tcg_vn = read_fp_dreg(s, rn); |
3651 |
if (cmp_with_zero) {
|
3652 |
tcg_vm = tcg_const_i64(0);
|
3653 |
} else {
|
3654 |
tcg_vm = read_fp_dreg(s, rm); |
3655 |
} |
3656 |
if (signal_all_nans) {
|
3657 |
gen_helper_vfp_cmped_a64(tcg_flags, tcg_vn, tcg_vm, fpst); |
3658 |
} else {
|
3659 |
gen_helper_vfp_cmpd_a64(tcg_flags, tcg_vn, tcg_vm, fpst); |
3660 |
} |
3661 |
tcg_temp_free_i64(tcg_vn); |
3662 |
tcg_temp_free_i64(tcg_vm); |
3663 |
} else {
|
3664 |
TCGv_i32 tcg_vn, tcg_vm; |
3665 |
|
3666 |
tcg_vn = read_fp_sreg(s, rn); |
3667 |
if (cmp_with_zero) {
|
3668 |
tcg_vm = tcg_const_i32(0);
|
3669 |
} else {
|
3670 |
tcg_vm = read_fp_sreg(s, rm); |
3671 |
} |
3672 |
if (signal_all_nans) {
|
3673 |
gen_helper_vfp_cmpes_a64(tcg_flags, tcg_vn, tcg_vm, fpst); |
3674 |
} else {
|
3675 |
gen_helper_vfp_cmps_a64(tcg_flags, tcg_vn, tcg_vm, fpst); |
3676 |
} |
3677 |
tcg_temp_free_i32(tcg_vn); |
3678 |
tcg_temp_free_i32(tcg_vm); |
3679 |
} |
3680 |
|
3681 |
tcg_temp_free_ptr(fpst); |
3682 |
|
3683 |
gen_set_nzcv(tcg_flags); |
3684 |
|
3685 |
tcg_temp_free_i64(tcg_flags); |
3686 |
} |
3687 |
|
3688 |
/* C3.6.22 Floating point compare
|
3689 |
* 31 30 29 28 24 23 22 21 20 16 15 14 13 10 9 5 4 0
|
3690 |
* +---+---+---+-----------+------+---+------+-----+---------+------+-------+
|
3691 |
* | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | op | 1 0 0 0 | Rn | op2 |
|
3692 |
* +---+---+---+-----------+------+---+------+-----+---------+------+-------+
|
3693 |
*/
|
3694 |
static void disas_fp_compare(DisasContext *s, uint32_t insn) |
3695 |
{ |
3696 |
unsigned int mos, type, rm, op, rn, opc, op2r; |
3697 |
|
3698 |
mos = extract32(insn, 29, 3); |
3699 |
type = extract32(insn, 22, 2); /* 0 = single, 1 = double */ |
3700 |
rm = extract32(insn, 16, 5); |
3701 |
op = extract32(insn, 14, 2); |
3702 |
rn = extract32(insn, 5, 5); |
3703 |
opc = extract32(insn, 3, 2); |
3704 |
op2r = extract32(insn, 0, 3); |
3705 |
|
3706 |
if (mos || op || op2r || type > 1) { |
3707 |
unallocated_encoding(s); |
3708 |
return;
|
3709 |
} |
3710 |
|
3711 |
handle_fp_compare(s, type, rn, rm, opc & 1, opc & 2); |
3712 |
} |
3713 |
|
3714 |
/* C3.6.23 Floating point conditional compare
|
3715 |
* 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 3 0
|
3716 |
* +---+---+---+-----------+------+---+------+------+-----+------+----+------+
|
3717 |
* | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | cond | 0 1 | Rn | op | nzcv |
|
3718 |
* +---+---+---+-----------+------+---+------+------+-----+------+----+------+
|
3719 |
*/
|
3720 |
static void disas_fp_ccomp(DisasContext *s, uint32_t insn) |
3721 |
{ |
3722 |
unsigned int mos, type, rm, cond, rn, op, nzcv; |
3723 |
TCGv_i64 tcg_flags; |
3724 |
int label_continue = -1; |
3725 |
|
3726 |
mos = extract32(insn, 29, 3); |
3727 |
type = extract32(insn, 22, 2); /* 0 = single, 1 = double */ |
3728 |
rm = extract32(insn, 16, 5); |
3729 |
cond = extract32(insn, 12, 4); |
3730 |
rn = extract32(insn, 5, 5); |
3731 |
op = extract32(insn, 4, 1); |
3732 |
nzcv = extract32(insn, 0, 4); |
3733 |
|
3734 |
if (mos || type > 1) { |
3735 |
unallocated_encoding(s); |
3736 |
return;
|
3737 |
} |
3738 |
|
3739 |
if (cond < 0x0e) { /* not always */ |
3740 |
int label_match = gen_new_label();
|
3741 |
label_continue = gen_new_label(); |
3742 |
arm_gen_test_cc(cond, label_match); |
3743 |
/* nomatch: */
|
3744 |
tcg_flags = tcg_const_i64(nzcv << 28);
|
3745 |
gen_set_nzcv(tcg_flags); |
3746 |
tcg_temp_free_i64(tcg_flags); |
3747 |
tcg_gen_br(label_continue); |
3748 |
gen_set_label(label_match); |
3749 |
} |
3750 |
|
3751 |
handle_fp_compare(s, type, rn, rm, false, op);
|
3752 |
|
3753 |
if (cond < 0x0e) { |
3754 |
gen_set_label(label_continue); |
3755 |
} |
3756 |
} |
3757 |
|
3758 |
/* copy src FP register to dst FP register; type specifies single or double */
|
3759 |
static void gen_mov_fp2fp(DisasContext *s, int type, int dst, int src) |
3760 |
{ |
3761 |
if (type) {
|
3762 |
TCGv_i64 v = read_fp_dreg(s, src); |
3763 |
write_fp_dreg(s, dst, v); |
3764 |
tcg_temp_free_i64(v); |
3765 |
} else {
|
3766 |
TCGv_i32 v = read_fp_sreg(s, src); |
3767 |
write_fp_sreg(s, dst, v); |
3768 |
tcg_temp_free_i32(v); |
3769 |
} |
3770 |
} |
3771 |
|
3772 |
/* C3.6.24 Floating point conditional select
|
3773 |
* 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
|
3774 |
* +---+---+---+-----------+------+---+------+------+-----+------+------+
|
3775 |
* | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | cond | 1 1 | Rn | Rd |
|
3776 |
* +---+---+---+-----------+------+---+------+------+-----+------+------+
|
3777 |
*/
|
3778 |
static void disas_fp_csel(DisasContext *s, uint32_t insn) |
3779 |
{ |
3780 |
unsigned int mos, type, rm, cond, rn, rd; |
3781 |
int label_continue = -1; |
3782 |
|
3783 |
mos = extract32(insn, 29, 3); |
3784 |
type = extract32(insn, 22, 2); /* 0 = single, 1 = double */ |
3785 |
rm = extract32(insn, 16, 5); |
3786 |
cond = extract32(insn, 12, 4); |
3787 |
rn = extract32(insn, 5, 5); |
3788 |
rd = extract32(insn, 0, 5); |
3789 |
|
3790 |
if (mos || type > 1) { |
3791 |
unallocated_encoding(s); |
3792 |
return;
|
3793 |
} |
3794 |
|
3795 |
if (cond < 0x0e) { /* not always */ |
3796 |
int label_match = gen_new_label();
|
3797 |
label_continue = gen_new_label(); |
3798 |
arm_gen_test_cc(cond, label_match); |
3799 |
/* nomatch: */
|
3800 |
gen_mov_fp2fp(s, type, rd, rm); |
3801 |
tcg_gen_br(label_continue); |
3802 |
gen_set_label(label_match); |
3803 |
} |
3804 |
|
3805 |
gen_mov_fp2fp(s, type, rd, rn); |
3806 |
|
3807 |
if (cond < 0x0e) { /* continue */ |
3808 |
gen_set_label(label_continue); |
3809 |
} |
3810 |
} |
3811 |
|
3812 |
/* C3.6.25 Floating-point data-processing (1 source) - single precision */
|
3813 |
static void handle_fp_1src_single(DisasContext *s, int opcode, int rd, int rn) |
3814 |
{ |
3815 |
TCGv_ptr fpst; |
3816 |
TCGv_i32 tcg_op; |
3817 |
TCGv_i32 tcg_res; |
3818 |
|
3819 |
fpst = get_fpstatus_ptr(); |
3820 |
tcg_op = read_fp_sreg(s, rn); |
3821 |
tcg_res = tcg_temp_new_i32(); |
3822 |
|
3823 |
switch (opcode) {
|
3824 |
case 0x0: /* FMOV */ |
3825 |
tcg_gen_mov_i32(tcg_res, tcg_op); |
3826 |
break;
|
3827 |
case 0x1: /* FABS */ |
3828 |
gen_helper_vfp_abss(tcg_res, tcg_op); |
3829 |
break;
|
3830 |
case 0x2: /* FNEG */ |
3831 |
gen_helper_vfp_negs(tcg_res, tcg_op); |
3832 |
break;
|
3833 |
case 0x3: /* FSQRT */ |
3834 |
gen_helper_vfp_sqrts(tcg_res, tcg_op, cpu_env); |
3835 |
break;
|
3836 |
case 0x8: /* FRINTN */ |
3837 |
case 0x9: /* FRINTP */ |
3838 |
case 0xa: /* FRINTM */ |
3839 |
case 0xb: /* FRINTZ */ |
3840 |
case 0xc: /* FRINTA */ |
3841 |
{ |
3842 |
TCGv_i32 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(opcode & 7));
|
3843 |
|
3844 |
gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env); |
3845 |
gen_helper_rints(tcg_res, tcg_op, fpst); |
3846 |
|
3847 |
gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env); |
3848 |
tcg_temp_free_i32(tcg_rmode); |
3849 |
break;
|
3850 |
} |
3851 |
case 0xe: /* FRINTX */ |
3852 |
gen_helper_rints_exact(tcg_res, tcg_op, fpst); |
3853 |
break;
|
3854 |
case 0xf: /* FRINTI */ |
3855 |
gen_helper_rints(tcg_res, tcg_op, fpst); |
3856 |
break;
|
3857 |
default:
|
3858 |
abort(); |
3859 |
} |
3860 |
|
3861 |
write_fp_sreg(s, rd, tcg_res); |
3862 |
|
3863 |
tcg_temp_free_ptr(fpst); |
3864 |
tcg_temp_free_i32(tcg_op); |
3865 |
tcg_temp_free_i32(tcg_res); |
3866 |
} |
3867 |
|
3868 |
/* C3.6.25 Floating-point data-processing (1 source) - double precision */
|
3869 |
static void handle_fp_1src_double(DisasContext *s, int opcode, int rd, int rn) |
3870 |
{ |
3871 |
TCGv_ptr fpst; |
3872 |
TCGv_i64 tcg_op; |
3873 |
TCGv_i64 tcg_res; |
3874 |
|
3875 |
fpst = get_fpstatus_ptr(); |
3876 |
tcg_op = read_fp_dreg(s, rn); |
3877 |
tcg_res = tcg_temp_new_i64(); |
3878 |
|
3879 |
switch (opcode) {
|
3880 |
case 0x0: /* FMOV */ |
3881 |
tcg_gen_mov_i64(tcg_res, tcg_op); |
3882 |
break;
|
3883 |
case 0x1: /* FABS */ |
3884 |
gen_helper_vfp_absd(tcg_res, tcg_op); |
3885 |
break;
|
3886 |
case 0x2: /* FNEG */ |
3887 |
gen_helper_vfp_negd(tcg_res, tcg_op); |
3888 |
break;
|
3889 |
case 0x3: /* FSQRT */ |
3890 |
gen_helper_vfp_sqrtd(tcg_res, tcg_op, cpu_env); |
3891 |
break;
|
3892 |
case 0x8: /* FRINTN */ |
3893 |
case 0x9: /* FRINTP */ |
3894 |
case 0xa: /* FRINTM */ |
3895 |
case 0xb: /* FRINTZ */ |
3896 |
case 0xc: /* FRINTA */ |
3897 |
{ |
3898 |
TCGv_i32 tcg_rmode = tcg_const_i32(arm_rmode_to_sf(opcode & 7));
|
3899 |
|
3900 |
gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env); |
3901 |
gen_helper_rintd(tcg_res, tcg_op, fpst); |
3902 |
|
3903 |
gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env); |
3904 |
tcg_temp_free_i32(tcg_rmode); |
3905 |
break;
|
3906 |
} |
3907 |
case 0xe: /* FRINTX */ |
3908 |
gen_helper_rintd_exact(tcg_res, tcg_op, fpst); |
3909 |
break;
|
3910 |
case 0xf: /* FRINTI */ |
3911 |
gen_helper_rintd(tcg_res, tcg_op, fpst); |
3912 |
break;
|
3913 |
default:
|
3914 |
abort(); |
3915 |
} |
3916 |
|
3917 |
write_fp_dreg(s, rd, tcg_res); |
3918 |
|
3919 |
tcg_temp_free_ptr(fpst); |
3920 |
tcg_temp_free_i64(tcg_op); |
3921 |
tcg_temp_free_i64(tcg_res); |
3922 |
} |
3923 |
|
3924 |
static void handle_fp_fcvt(DisasContext *s, int opcode, |
3925 |
int rd, int rn, int dtype, int ntype) |
3926 |
{ |
3927 |
switch (ntype) {
|
3928 |
case 0x0: |
3929 |
{ |
3930 |
TCGv_i32 tcg_rn = read_fp_sreg(s, rn); |
3931 |
if (dtype == 1) { |
3932 |
/* Single to double */
|
3933 |
TCGv_i64 tcg_rd = tcg_temp_new_i64(); |
3934 |
gen_helper_vfp_fcvtds(tcg_rd, tcg_rn, cpu_env); |
3935 |
write_fp_dreg(s, rd, tcg_rd); |
3936 |
tcg_temp_free_i64(tcg_rd); |
3937 |
} else {
|
3938 |
/* Single to half */
|
3939 |
TCGv_i32 tcg_rd = tcg_temp_new_i32(); |
3940 |
gen_helper_vfp_fcvt_f32_to_f16(tcg_rd, tcg_rn, cpu_env); |
3941 |
/* write_fp_sreg is OK here because top half of tcg_rd is zero */
|
3942 |
write_fp_sreg(s, rd, tcg_rd); |
3943 |
tcg_temp_free_i32(tcg_rd); |
3944 |
} |
3945 |
tcg_temp_free_i32(tcg_rn); |
3946 |
break;
|
3947 |
} |
3948 |
case 0x1: |
3949 |
{ |
3950 |
TCGv_i64 tcg_rn = read_fp_dreg(s, rn); |
3951 |
TCGv_i32 tcg_rd = tcg_temp_new_i32(); |
3952 |
if (dtype == 0) { |
3953 |
/* Double to single */
|
3954 |
gen_helper_vfp_fcvtsd(tcg_rd, tcg_rn, cpu_env); |
3955 |
} else {
|
3956 |
/* Double to half */
|
3957 |
gen_helper_vfp_fcvt_f64_to_f16(tcg_rd, tcg_rn, cpu_env); |
3958 |
/* write_fp_sreg is OK here because top half of tcg_rd is zero */
|
3959 |
} |
3960 |
write_fp_sreg(s, rd, tcg_rd); |
3961 |
tcg_temp_free_i32(tcg_rd); |
3962 |
tcg_temp_free_i64(tcg_rn); |
3963 |
break;
|
3964 |
} |
3965 |
case 0x3: |
3966 |
{ |
3967 |
TCGv_i32 tcg_rn = read_fp_sreg(s, rn); |
3968 |
tcg_gen_ext16u_i32(tcg_rn, tcg_rn); |
3969 |
if (dtype == 0) { |
3970 |
/* Half to single */
|
3971 |
TCGv_i32 tcg_rd = tcg_temp_new_i32(); |
3972 |
gen_helper_vfp_fcvt_f16_to_f32(tcg_rd, tcg_rn, cpu_env); |
3973 |
write_fp_sreg(s, rd, tcg_rd); |
3974 |
tcg_temp_free_i32(tcg_rd); |
3975 |
} else {
|
3976 |
/* Half to double */
|
3977 |
TCGv_i64 tcg_rd = tcg_temp_new_i64(); |
3978 |
gen_helper_vfp_fcvt_f16_to_f64(tcg_rd, tcg_rn, cpu_env); |
3979 |
write_fp_dreg(s, rd, tcg_rd); |
3980 |
tcg_temp_free_i64(tcg_rd); |
3981 |
} |
3982 |
tcg_temp_free_i32(tcg_rn); |
3983 |
break;
|
3984 |
} |
3985 |
default:
|
3986 |
abort(); |
3987 |
} |
3988 |
} |
3989 |
|
3990 |
/* C3.6.25 Floating point data-processing (1 source)
|
3991 |
* 31 30 29 28 24 23 22 21 20 15 14 10 9 5 4 0
|
3992 |
* +---+---+---+-----------+------+---+--------+-----------+------+------+
|
3993 |
* | M | 0 | S | 1 1 1 1 0 | type | 1 | opcode | 1 0 0 0 0 | Rn | Rd |
|
3994 |
* +---+---+---+-----------+------+---+--------+-----------+------+------+
|
3995 |
*/
|
3996 |
static void disas_fp_1src(DisasContext *s, uint32_t insn) |
3997 |
{ |
3998 |
int type = extract32(insn, 22, 2); |
3999 |
int opcode = extract32(insn, 15, 6); |
4000 |
int rn = extract32(insn, 5, 5); |
4001 |
int rd = extract32(insn, 0, 5); |
4002 |
|
4003 |
switch (opcode) {
|
4004 |
case 0x4: case 0x5: case 0x7: |
4005 |
{ |
4006 |
/* FCVT between half, single and double precision */
|
4007 |
int dtype = extract32(opcode, 0, 2); |
4008 |
if (type == 2 || dtype == type) { |
4009 |
unallocated_encoding(s); |
4010 |
return;
|
4011 |
} |
4012 |
handle_fp_fcvt(s, opcode, rd, rn, dtype, type); |
4013 |
break;
|
4014 |
} |
4015 |
case 0x0 ... 0x3: |
4016 |
case 0x8 ... 0xc: |
4017 |
case 0xe ... 0xf: |
4018 |
/* 32-to-32 and 64-to-64 ops */
|
4019 |
switch (type) {
|
4020 |
case 0: |
4021 |
handle_fp_1src_single(s, opcode, rd, rn); |
4022 |
break;
|
4023 |
case 1: |
4024 |
handle_fp_1src_double(s, opcode, rd, rn); |
4025 |
break;
|
4026 |
default:
|
4027 |
unallocated_encoding(s); |
4028 |
} |
4029 |
break;
|
4030 |
default:
|
4031 |
unallocated_encoding(s); |
4032 |
break;
|
4033 |
} |
4034 |
} |
4035 |
|
4036 |
/* C3.6.26 Floating-point data-processing (2 source) - single precision */
|
4037 |
static void handle_fp_2src_single(DisasContext *s, int opcode, |
4038 |
int rd, int rn, int rm) |
4039 |
{ |
4040 |
TCGv_i32 tcg_op1; |
4041 |
TCGv_i32 tcg_op2; |
4042 |
TCGv_i32 tcg_res; |
4043 |
TCGv_ptr fpst; |
4044 |
|
4045 |
tcg_res = tcg_temp_new_i32(); |
4046 |
fpst = get_fpstatus_ptr(); |
4047 |
tcg_op1 = read_fp_sreg(s, rn); |
4048 |
tcg_op2 = read_fp_sreg(s, rm); |
4049 |
|
4050 |
switch (opcode) {
|
4051 |
case 0x0: /* FMUL */ |
4052 |
gen_helper_vfp_muls(tcg_res, tcg_op1, tcg_op2, fpst); |
4053 |
break;
|
4054 |
case 0x1: /* FDIV */ |
4055 |
gen_helper_vfp_divs(tcg_res, tcg_op1, tcg_op2, fpst); |
4056 |
break;
|
4057 |
case 0x2: /* FADD */ |
4058 |
gen_helper_vfp_adds(tcg_res, tcg_op1, tcg_op2, fpst); |
4059 |
break;
|
4060 |
case 0x3: /* FSUB */ |
4061 |
gen_helper_vfp_subs(tcg_res, tcg_op1, tcg_op2, fpst); |
4062 |
break;
|
4063 |
case 0x4: /* FMAX */ |
4064 |
gen_helper_vfp_maxs(tcg_res, tcg_op1, tcg_op2, fpst); |
4065 |
break;
|
4066 |
case 0x5: /* FMIN */ |
4067 |
gen_helper_vfp_mins(tcg_res, tcg_op1, tcg_op2, fpst); |
4068 |
break;
|
4069 |
case 0x6: /* FMAXNM */ |
4070 |
gen_helper_vfp_maxnums(tcg_res, tcg_op1, tcg_op2, fpst); |
4071 |
break;
|
4072 |
case 0x7: /* FMINNM */ |
4073 |
gen_helper_vfp_minnums(tcg_res, tcg_op1, tcg_op2, fpst); |
4074 |
break;
|
4075 |
case 0x8: /* FNMUL */ |
4076 |
gen_helper_vfp_muls(tcg_res, tcg_op1, tcg_op2, fpst); |
4077 |
gen_helper_vfp_negs(tcg_res, tcg_res); |
4078 |
break;
|
4079 |
} |
4080 |
|
4081 |
write_fp_sreg(s, rd, tcg_res); |
4082 |
|
4083 |
tcg_temp_free_ptr(fpst); |
4084 |
tcg_temp_free_i32(tcg_op1); |
4085 |
tcg_temp_free_i32(tcg_op2); |
4086 |
tcg_temp_free_i32(tcg_res); |
4087 |
} |
4088 |
|
4089 |
/* C3.6.26 Floating-point data-processing (2 source) - double precision */
|
4090 |
static void handle_fp_2src_double(DisasContext *s, int opcode, |
4091 |
int rd, int rn, int rm) |
4092 |
{ |
4093 |
TCGv_i64 tcg_op1; |
4094 |
TCGv_i64 tcg_op2; |
4095 |
TCGv_i64 tcg_res; |
4096 |
TCGv_ptr fpst; |
4097 |
|
4098 |
tcg_res = tcg_temp_new_i64(); |
4099 |
fpst = get_fpstatus_ptr(); |
4100 |
tcg_op1 = read_fp_dreg(s, rn); |
4101 |
tcg_op2 = read_fp_dreg(s, rm); |
4102 |
|
4103 |
switch (opcode) {
|
4104 |
case 0x0: /* FMUL */ |
4105 |
gen_helper_vfp_muld(tcg_res, tcg_op1, tcg_op2, fpst); |
4106 |
break;
|
4107 |
case 0x1: /* FDIV */ |
4108 |
gen_helper_vfp_divd(tcg_res, tcg_op1, tcg_op2, fpst); |
4109 |
break;
|
4110 |
case 0x2: /* FADD */ |
4111 |
gen_helper_vfp_addd(tcg_res, tcg_op1, tcg_op2, fpst); |
4112 |
break;
|
4113 |
case 0x3: /* FSUB */ |
4114 |
gen_helper_vfp_subd(tcg_res, tcg_op1, tcg_op2, fpst); |
4115 |
break;
|
4116 |
case 0x4: /* FMAX */ |
4117 |
gen_helper_vfp_maxd(tcg_res, tcg_op1, tcg_op2, fpst); |
4118 |
break;
|
4119 |
case 0x5: /* FMIN */ |
4120 |
gen_helper_vfp_mind(tcg_res, tcg_op1, tcg_op2, fpst); |
4121 |
break;
|
4122 |
case 0x6: /* FMAXNM */ |
4123 |
gen_helper_vfp_maxnumd(tcg_res, tcg_op1, tcg_op2, fpst); |
4124 |
break;
|
4125 |
case 0x7: /* FMINNM */ |
4126 |
gen_helper_vfp_minnumd(tcg_res, tcg_op1, tcg_op2, fpst); |
4127 |
break;
|
4128 |
case 0x8: /* FNMUL */ |
4129 |
gen_helper_vfp_muld(tcg_res, tcg_op1, tcg_op2, fpst); |
4130 |
gen_helper_vfp_negd(tcg_res, tcg_res); |
4131 |
break;
|
4132 |
} |
4133 |
|
4134 |
write_fp_dreg(s, rd, tcg_res); |
4135 |
|
4136 |
tcg_temp_free_ptr(fpst); |
4137 |
tcg_temp_free_i64(tcg_op1); |
4138 |
tcg_temp_free_i64(tcg_op2); |
4139 |
tcg_temp_free_i64(tcg_res); |
4140 |
} |
4141 |
|
4142 |
/* C3.6.26 Floating point data-processing (2 source)
|
4143 |
* 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
|
4144 |
* +---+---+---+-----------+------+---+------+--------+-----+------+------+
|
4145 |
* | M | 0 | S | 1 1 1 1 0 | type | 1 | Rm | opcode | 1 0 | Rn | Rd |
|
4146 |
* +---+---+---+-----------+------+---+------+--------+-----+------+------+
|
4147 |
*/
|
4148 |
static void disas_fp_2src(DisasContext *s, uint32_t insn) |
4149 |
{ |
4150 |
int type = extract32(insn, 22, 2); |
4151 |
int rd = extract32(insn, 0, 5); |
4152 |
int rn = extract32(insn, 5, 5); |
4153 |
int rm = extract32(insn, 16, 5); |
4154 |
int opcode = extract32(insn, 12, 4); |
4155 |
|
4156 |
if (opcode > 8) { |
4157 |
unallocated_encoding(s); |
4158 |
return;
|
4159 |
} |
4160 |
|
4161 |
switch (type) {
|
4162 |
case 0: |
4163 |
handle_fp_2src_single(s, opcode, rd, rn, rm); |
4164 |
break;
|
4165 |
case 1: |
4166 |
handle_fp_2src_double(s, opcode, rd, rn, rm); |
4167 |
break;
|
4168 |
default:
|
4169 |
unallocated_encoding(s); |
4170 |
} |
4171 |
} |
4172 |
|
4173 |
/* C3.6.27 Floating-point data-processing (3 source) - single precision */
|
4174 |
static void handle_fp_3src_single(DisasContext *s, bool o0, bool o1, |
4175 |
int rd, int rn, int rm, int ra) |
4176 |
{ |
4177 |
TCGv_i32 tcg_op1, tcg_op2, tcg_op3; |
4178 |
TCGv_i32 tcg_res = tcg_temp_new_i32(); |
4179 |
TCGv_ptr fpst = get_fpstatus_ptr(); |
4180 |
|
4181 |
tcg_op1 = read_fp_sreg(s, rn); |
4182 |
tcg_op2 = read_fp_sreg(s, rm); |
4183 |
tcg_op3 = read_fp_sreg(s, ra); |
4184 |
|
4185 |
/* These are fused multiply-add, and must be done as one
|
4186 |
* floating point operation with no rounding between the
|
4187 |
* multiplication and addition steps.
|
4188 |
* NB that doing the negations here as separate steps is
|
4189 |
* correct : an input NaN should come out with its sign bit
|
4190 |
* flipped if it is a negated-input.
|
4191 |
*/
|
4192 |
if (o1 == true) { |
4193 |
gen_helper_vfp_negs(tcg_op3, tcg_op3); |
4194 |
} |
4195 |
|
4196 |
if (o0 != o1) {
|
4197 |
gen_helper_vfp_negs(tcg_op1, tcg_op1); |
4198 |
} |
4199 |
|
4200 |
gen_helper_vfp_muladds(tcg_res, tcg_op1, tcg_op2, tcg_op3, fpst); |
4201 |
|
4202 |
write_fp_sreg(s, rd, tcg_res); |
4203 |
|
4204 |
tcg_temp_free_ptr(fpst); |
4205 |
tcg_temp_free_i32(tcg_op1); |
4206 |
tcg_temp_free_i32(tcg_op2); |
4207 |
tcg_temp_free_i32(tcg_op3); |
4208 |
tcg_temp_free_i32(tcg_res); |
4209 |
} |
4210 |
|
4211 |
/* C3.6.27 Floating-point data-processing (3 source) - double precision */
|
4212 |
static void handle_fp_3src_double(DisasContext *s, bool o0, bool o1, |
4213 |
int rd, int rn, int rm, int ra) |
4214 |
{ |
4215 |
TCGv_i64 tcg_op1, tcg_op2, tcg_op3; |
4216 |
TCGv_i64 tcg_res = tcg_temp_new_i64(); |
4217 |
TCGv_ptr fpst = get_fpstatus_ptr(); |
4218 |
|
4219 |
tcg_op1 = read_fp_dreg(s, rn); |
4220 |
tcg_op2 = read_fp_dreg(s, rm); |
4221 |
tcg_op3 = read_fp_dreg(s, ra); |
4222 |
|
4223 |
/* These are fused multiply-add, and must be done as one
|
4224 |
* floating point operation with no rounding between the
|
4225 |
* multiplication and addition steps.
|
4226 |
* NB that doing the negations here as separate steps is
|
4227 |
* correct : an input NaN should come out with its sign bit
|
4228 |
* flipped if it is a negated-input.
|
4229 |
*/
|
4230 |
if (o1 == true) { |
4231 |
gen_helper_vfp_negd(tcg_op3, tcg_op3); |
4232 |
} |
4233 |
|
4234 |
if (o0 != o1) {
|
4235 |
gen_helper_vfp_negd(tcg_op1, tcg_op1); |
4236 |
} |
4237 |
|
4238 |
gen_helper_vfp_muladdd(tcg_res, tcg_op1, tcg_op2, tcg_op3, fpst); |
4239 |
|
4240 |
write_fp_dreg(s, rd, tcg_res); |
4241 |
|
4242 |
tcg_temp_free_ptr(fpst); |
4243 |
tcg_temp_free_i64(tcg_op1); |
4244 |
tcg_temp_free_i64(tcg_op2); |
4245 |
tcg_temp_free_i64(tcg_op3); |
4246 |
tcg_temp_free_i64(tcg_res); |
4247 |
} |
4248 |
|
4249 |
/* C3.6.27 Floating point data-processing (3 source)
|
4250 |
* 31 30 29 28 24 23 22 21 20 16 15 14 10 9 5 4 0
|
4251 |
* +---+---+---+-----------+------+----+------+----+------+------+------+
|
4252 |
* | M | 0 | S | 1 1 1 1 1 | type | o1 | Rm | o0 | Ra | Rn | Rd |
|
4253 |
* +---+---+---+-----------+------+----+------+----+------+------+------+
|
4254 |
*/
|
4255 |
static void disas_fp_3src(DisasContext *s, uint32_t insn) |
4256 |
{ |
4257 |
int type = extract32(insn, 22, 2); |
4258 |
int rd = extract32(insn, 0, 5); |
4259 |
int rn = extract32(insn, 5, 5); |
4260 |
int ra = extract32(insn, 10, 5); |
4261 |
int rm = extract32(insn, 16, 5); |
4262 |
bool o0 = extract32(insn, 15, 1); |
4263 |
bool o1 = extract32(insn, 21, 1); |
4264 |
|
4265 |
switch (type) {
|
4266 |
case 0: |
4267 |
handle_fp_3src_single(s, o0, o1, rd, rn, rm, ra); |
4268 |
break;
|
4269 |
case 1: |
4270 |
handle_fp_3src_double(s, o0, o1, rd, rn, rm, ra); |
4271 |
break;
|
4272 |
default:
|
4273 |
unallocated_encoding(s); |
4274 |
} |
4275 |
} |
4276 |
|
4277 |
/* C3.6.28 Floating point immediate
|
4278 |
* 31 30 29 28 24 23 22 21 20 13 12 10 9 5 4 0
|
4279 |
* +---+---+---+-----------+------+---+------------+-------+------+------+
|
4280 |
* | M | 0 | S | 1 1 1 1 0 | type | 1 | imm8 | 1 0 0 | imm5 | Rd |
|
4281 |
* +---+---+---+-----------+------+---+------------+-------+------+------+
|
4282 |
*/
|
4283 |
static void disas_fp_imm(DisasContext *s, uint32_t insn) |
4284 |
{ |
4285 |
int rd = extract32(insn, 0, 5); |
4286 |
int imm8 = extract32(insn, 13, 8); |
4287 |
int is_double = extract32(insn, 22, 2); |
4288 |
uint64_t imm; |
4289 |
TCGv_i64 tcg_res; |
4290 |
|
4291 |
if (is_double > 1) { |
4292 |
unallocated_encoding(s); |
4293 |
return;
|
4294 |
} |
4295 |
|
4296 |
/* The imm8 encodes the sign bit, enough bits to represent
|
4297 |
* an exponent in the range 01....1xx to 10....0xx,
|
4298 |
* and the most significant 4 bits of the mantissa; see
|
4299 |
* VFPExpandImm() in the v8 ARM ARM.
|
4300 |
*/
|
4301 |
if (is_double) {
|
4302 |
imm = (extract32(imm8, 7, 1) ? 0x8000 : 0) | |
4303 |
(extract32(imm8, 6, 1) ? 0x3fc0 : 0x4000) | |
4304 |
extract32(imm8, 0, 6); |
4305 |
imm <<= 48;
|
4306 |
} else {
|
4307 |
imm = (extract32(imm8, 7, 1) ? 0x8000 : 0) | |
4308 |
(extract32(imm8, 6, 1) ? 0x3e00 : 0x4000) | |
4309 |
(extract32(imm8, 0, 6) << 3); |
4310 |
imm <<= 16;
|
4311 |
} |
4312 |
|
4313 |
tcg_res = tcg_const_i64(imm); |
4314 |
write_fp_dreg(s, rd, tcg_res); |
4315 |
tcg_temp_free_i64(tcg_res); |
4316 |
} |
4317 |
|
4318 |
/* Handle floating point <=> fixed point conversions. Note that we can
|
4319 |
* also deal with fp <=> integer conversions as a special case (scale == 64)
|
4320 |
* OPTME: consider handling that special case specially or at least skipping
|
4321 |
* the call to scalbn in the helpers for zero shifts.
|
4322 |
*/
|
4323 |
static void handle_fpfpcvt(DisasContext *s, int rd, int rn, int opcode, |
4324 |
bool itof, int rmode, int scale, int sf, int type) |
4325 |
{ |
4326 |
bool is_signed = !(opcode & 1); |
4327 |
bool is_double = type;
|
4328 |
TCGv_ptr tcg_fpstatus; |
4329 |
TCGv_i32 tcg_shift; |
4330 |
|
4331 |
tcg_fpstatus = get_fpstatus_ptr(); |
4332 |
|
4333 |
tcg_shift = tcg_const_i32(64 - scale);
|
4334 |
|
4335 |
if (itof) {
|
4336 |
TCGv_i64 tcg_int = cpu_reg(s, rn); |
4337 |
if (!sf) {
|
4338 |
TCGv_i64 tcg_extend = new_tmp_a64(s); |
4339 |
|
4340 |
if (is_signed) {
|
4341 |
tcg_gen_ext32s_i64(tcg_extend, tcg_int); |
4342 |
} else {
|
4343 |
tcg_gen_ext32u_i64(tcg_extend, tcg_int); |
4344 |
} |
4345 |
|
4346 |
tcg_int = tcg_extend; |
4347 |
} |
4348 |
|
4349 |
if (is_double) {
|
4350 |
TCGv_i64 tcg_double = tcg_temp_new_i64(); |
4351 |
if (is_signed) {
|
4352 |
gen_helper_vfp_sqtod(tcg_double, tcg_int, |
4353 |
tcg_shift, tcg_fpstatus); |
4354 |
} else {
|
4355 |
gen_helper_vfp_uqtod(tcg_double, tcg_int, |
4356 |
tcg_shift, tcg_fpstatus); |
4357 |
} |
4358 |
write_fp_dreg(s, rd, tcg_double); |
4359 |
tcg_temp_free_i64(tcg_double); |
4360 |
} else {
|
4361 |
TCGv_i32 tcg_single = tcg_temp_new_i32(); |
4362 |
if (is_signed) {
|
4363 |
gen_helper_vfp_sqtos(tcg_single, tcg_int, |
4364 |
tcg_shift, tcg_fpstatus); |
4365 |
} else {
|
4366 |
gen_helper_vfp_uqtos(tcg_single, tcg_int, |
4367 |
tcg_shift, tcg_fpstatus); |
4368 |
} |
4369 |
write_fp_sreg(s, rd, tcg_single); |
4370 |
tcg_temp_free_i32(tcg_single); |
4371 |
} |
4372 |
} else {
|
4373 |
TCGv_i64 tcg_int = cpu_reg(s, rd); |
4374 |
TCGv_i32 tcg_rmode; |
4375 |
|
4376 |
if (extract32(opcode, 2, 1)) { |
4377 |
/* There are too many rounding modes to all fit into rmode,
|
4378 |
* so FCVTA[US] is a special case.
|
4379 |
*/
|
4380 |
rmode = FPROUNDING_TIEAWAY; |
4381 |
} |
4382 |
|
4383 |
tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode)); |
4384 |
|
4385 |
gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env); |
4386 |
|
4387 |
if (is_double) {
|
4388 |
TCGv_i64 tcg_double = read_fp_dreg(s, rn); |
4389 |
if (is_signed) {
|
4390 |
if (!sf) {
|
4391 |
gen_helper_vfp_tosld(tcg_int, tcg_double, |
4392 |
tcg_shift, tcg_fpstatus); |
4393 |
} else {
|
4394 |
gen_helper_vfp_tosqd(tcg_int, tcg_double, |
4395 |
tcg_shift, tcg_fpstatus); |
4396 |
} |
4397 |
} else {
|
4398 |
if (!sf) {
|
4399 |
gen_helper_vfp_tould(tcg_int, tcg_double, |
4400 |
tcg_shift, tcg_fpstatus); |
4401 |
} else {
|
4402 |
gen_helper_vfp_touqd(tcg_int, tcg_double, |
4403 |
tcg_shift, tcg_fpstatus); |
4404 |
} |
4405 |
} |
4406 |
tcg_temp_free_i64(tcg_double); |
4407 |
} else {
|
4408 |
TCGv_i32 tcg_single = read_fp_sreg(s, rn); |
4409 |
if (sf) {
|
4410 |
if (is_signed) {
|
4411 |
gen_helper_vfp_tosqs(tcg_int, tcg_single, |
4412 |
tcg_shift, tcg_fpstatus); |
4413 |
} else {
|
4414 |
gen_helper_vfp_touqs(tcg_int, tcg_single, |
4415 |
tcg_shift, tcg_fpstatus); |
4416 |
} |
4417 |
} else {
|
4418 |
TCGv_i32 tcg_dest = tcg_temp_new_i32(); |
4419 |
if (is_signed) {
|
4420 |
gen_helper_vfp_tosls(tcg_dest, tcg_single, |
4421 |
tcg_shift, tcg_fpstatus); |
4422 |
} else {
|
4423 |
gen_helper_vfp_touls(tcg_dest, tcg_single, |
4424 |
tcg_shift, tcg_fpstatus); |
4425 |
} |
4426 |
tcg_gen_extu_i32_i64(tcg_int, tcg_dest); |
4427 |
tcg_temp_free_i32(tcg_dest); |
4428 |
} |
4429 |
tcg_temp_free_i32(tcg_single); |
4430 |
} |
4431 |
|
4432 |
gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env); |
4433 |
tcg_temp_free_i32(tcg_rmode); |
4434 |
|
4435 |
if (!sf) {
|
4436 |
tcg_gen_ext32u_i64(tcg_int, tcg_int); |
4437 |
} |
4438 |
} |
4439 |
|
4440 |
tcg_temp_free_ptr(tcg_fpstatus); |
4441 |
tcg_temp_free_i32(tcg_shift); |
4442 |
} |
4443 |
|
4444 |
/* C3.6.29 Floating point <-> fixed point conversions
|
4445 |
* 31 30 29 28 24 23 22 21 20 19 18 16 15 10 9 5 4 0
|
4446 |
* +----+---+---+-----------+------+---+-------+--------+-------+------+------+
|
4447 |
* | sf | 0 | S | 1 1 1 1 0 | type | 0 | rmode | opcode | scale | Rn | Rd |
|
4448 |
* +----+---+---+-----------+------+---+-------+--------+-------+------+------+
|
4449 |
*/
|
4450 |
static void disas_fp_fixed_conv(DisasContext *s, uint32_t insn) |
4451 |
{ |
4452 |
int rd = extract32(insn, 0, 5); |
4453 |
int rn = extract32(insn, 5, 5); |
4454 |
int scale = extract32(insn, 10, 6); |
4455 |
int opcode = extract32(insn, 16, 3); |
4456 |
int rmode = extract32(insn, 19, 2); |
4457 |
int type = extract32(insn, 22, 2); |
4458 |
bool sbit = extract32(insn, 29, 1); |
4459 |
bool sf = extract32(insn, 31, 1); |
4460 |
bool itof;
|
4461 |
|
4462 |
if (sbit || (type > 1) |
4463 |
|| (!sf && scale < 32)) {
|
4464 |
unallocated_encoding(s); |
4465 |
return;
|
4466 |
} |
4467 |
|
4468 |
switch ((rmode << 3) | opcode) { |
4469 |
case 0x2: /* SCVTF */ |
4470 |
case 0x3: /* UCVTF */ |
4471 |
itof = true;
|
4472 |
break;
|
4473 |
case 0x18: /* FCVTZS */ |
4474 |
case 0x19: /* FCVTZU */ |
4475 |
itof = false;
|
4476 |
break;
|
4477 |
default:
|
4478 |
unallocated_encoding(s); |
4479 |
return;
|
4480 |
} |
4481 |
|
4482 |
handle_fpfpcvt(s, rd, rn, opcode, itof, FPROUNDING_ZERO, scale, sf, type); |
4483 |
} |
4484 |
|
4485 |
static void handle_fmov(DisasContext *s, int rd, int rn, int type, bool itof) |
4486 |
{ |
4487 |
/* FMOV: gpr to or from float, double, or top half of quad fp reg,
|
4488 |
* without conversion.
|
4489 |
*/
|
4490 |
|
4491 |
if (itof) {
|
4492 |
TCGv_i64 tcg_rn = cpu_reg(s, rn); |
4493 |
|
4494 |
switch (type) {
|
4495 |
case 0: |
4496 |
{ |
4497 |
/* 32 bit */
|
4498 |
TCGv_i64 tmp = tcg_temp_new_i64(); |
4499 |
tcg_gen_ext32u_i64(tmp, tcg_rn); |
4500 |
tcg_gen_st_i64(tmp, cpu_env, fp_reg_offset(rd, MO_64)); |
4501 |
tcg_gen_movi_i64(tmp, 0);
|
4502 |
tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(rd)); |
4503 |
tcg_temp_free_i64(tmp); |
4504 |
break;
|
4505 |
} |
4506 |
case 1: |
4507 |
{ |
4508 |
/* 64 bit */
|
4509 |
TCGv_i64 tmp = tcg_const_i64(0);
|
4510 |
tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_offset(rd, MO_64)); |
4511 |
tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(rd)); |
4512 |
tcg_temp_free_i64(tmp); |
4513 |
break;
|
4514 |
} |
4515 |
case 2: |
4516 |
/* 64 bit to top half. */
|
4517 |
tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_hi_offset(rd)); |
4518 |
break;
|
4519 |
} |
4520 |
} else {
|
4521 |
TCGv_i64 tcg_rd = cpu_reg(s, rd); |
4522 |
|
4523 |
switch (type) {
|
4524 |
case 0: |
4525 |
/* 32 bit */
|
4526 |
tcg_gen_ld32u_i64(tcg_rd, cpu_env, fp_reg_offset(rn, MO_32)); |
4527 |
break;
|
4528 |
case 1: |
4529 |
/* 64 bit */
|
4530 |
tcg_gen_ld_i64(tcg_rd, cpu_env, fp_reg_offset(rn, MO_64)); |
4531 |
break;
|
4532 |
case 2: |
4533 |
/* 64 bits from top half */
|
4534 |
tcg_gen_ld_i64(tcg_rd, cpu_env, fp_reg_hi_offset(rn)); |
4535 |
break;
|
4536 |
} |
4537 |
} |
4538 |
} |
4539 |
|
4540 |
/* C3.6.30 Floating point <-> integer conversions
|
4541 |
* 31 30 29 28 24 23 22 21 20 19 18 16 15 10 9 5 4 0
|
4542 |
* +----+---+---+-----------+------+---+-------+-----+-------------+----+----+
|
4543 |
* | sf | 0 | S | 1 1 1 1 0 | type | 1 | rmode | opc | 0 0 0 0 0 0 | Rn | Rd |
|
4544 |
* +----+---+---+-----------+------+---+-------+-----+-------------+----+----+
|
4545 |
*/
|
4546 |
static void disas_fp_int_conv(DisasContext *s, uint32_t insn) |
4547 |
{ |
4548 |
int rd = extract32(insn, 0, 5); |
4549 |
int rn = extract32(insn, 5, 5); |
4550 |
int opcode = extract32(insn, 16, 3); |
4551 |
int rmode = extract32(insn, 19, 2); |
4552 |
int type = extract32(insn, 22, 2); |
4553 |
bool sbit = extract32(insn, 29, 1); |
4554 |
bool sf = extract32(insn, 31, 1); |
4555 |
|
4556 |
if (sbit) {
|
4557 |
unallocated_encoding(s); |
4558 |
return;
|
4559 |
} |
4560 |
|
4561 |
if (opcode > 5) { |
4562 |
/* FMOV */
|
4563 |
bool itof = opcode & 1; |
4564 |
|
4565 |
if (rmode >= 2) { |
4566 |
unallocated_encoding(s); |
4567 |
return;
|
4568 |
} |
4569 |
|
4570 |
switch (sf << 3 | type << 1 | rmode) { |
4571 |
case 0x0: /* 32 bit */ |
4572 |
case 0xa: /* 64 bit */ |
4573 |
case 0xd: /* 64 bit to top half of quad */ |
4574 |
break;
|
4575 |
default:
|
4576 |
/* all other sf/type/rmode combinations are invalid */
|
4577 |
unallocated_encoding(s); |
4578 |
break;
|
4579 |
} |
4580 |
|
4581 |
handle_fmov(s, rd, rn, type, itof); |
4582 |
} else {
|
4583 |
/* actual FP conversions */
|
4584 |
bool itof = extract32(opcode, 1, 1); |
4585 |
|
4586 |
if (type > 1 || (rmode != 0 && opcode > 1)) { |
4587 |
unallocated_encoding(s); |
4588 |
return;
|
4589 |
} |
4590 |
|
4591 |
handle_fpfpcvt(s, rd, rn, opcode, itof, rmode, 64, sf, type);
|
4592 |
} |
4593 |
} |
4594 |
|
4595 |
/* FP-specific subcases of table C3-6 (SIMD and FP data processing)
|
4596 |
* 31 30 29 28 25 24 0
|
4597 |
* +---+---+---+---------+-----------------------------+
|
4598 |
* | | 0 | | 1 1 1 1 | |
|
4599 |
* +---+---+---+---------+-----------------------------+
|
4600 |
*/
|
4601 |
static void disas_data_proc_fp(DisasContext *s, uint32_t insn) |
4602 |
{ |
4603 |
if (extract32(insn, 24, 1)) { |
4604 |
/* Floating point data-processing (3 source) */
|
4605 |
disas_fp_3src(s, insn); |
4606 |
} else if (extract32(insn, 21, 1) == 0) { |
4607 |
/* Floating point to fixed point conversions */
|
4608 |
disas_fp_fixed_conv(s, insn); |
4609 |
} else {
|
4610 |
switch (extract32(insn, 10, 2)) { |
4611 |
case 1: |
4612 |
/* Floating point conditional compare */
|
4613 |
disas_fp_ccomp(s, insn); |
4614 |
break;
|
4615 |
case 2: |
4616 |
/* Floating point data-processing (2 source) */
|
4617 |
disas_fp_2src(s, insn); |
4618 |
break;
|
4619 |
case 3: |
4620 |
/* Floating point conditional select */
|
4621 |
disas_fp_csel(s, insn); |
4622 |
break;
|
4623 |
case 0: |
4624 |
switch (ctz32(extract32(insn, 12, 4))) { |
4625 |
case 0: /* [15:12] == xxx1 */ |
4626 |
/* Floating point immediate */
|
4627 |
disas_fp_imm(s, insn); |
4628 |
break;
|
4629 |
case 1: /* [15:12] == xx10 */ |
4630 |
/* Floating point compare */
|
4631 |
disas_fp_compare(s, insn); |
4632 |
break;
|
4633 |
case 2: /* [15:12] == x100 */ |
4634 |
/* Floating point data-processing (1 source) */
|
4635 |
disas_fp_1src(s, insn); |
4636 |
break;
|
4637 |
case 3: /* [15:12] == 1000 */ |
4638 |
unallocated_encoding(s); |
4639 |
break;
|
4640 |
default: /* [15:12] == 0000 */ |
4641 |
/* Floating point <-> integer conversions */
|
4642 |
disas_fp_int_conv(s, insn); |
4643 |
break;
|
4644 |
} |
4645 |
break;
|
4646 |
} |
4647 |
} |
4648 |
} |
4649 |
|
4650 |
static void do_ext64(DisasContext *s, TCGv_i64 tcg_left, TCGv_i64 tcg_right, |
4651 |
int pos)
|
4652 |
{ |
4653 |
/* Extract 64 bits from the middle of two concatenated 64 bit
|
4654 |
* vector register slices left:right. The extracted bits start
|
4655 |
* at 'pos' bits into the right (least significant) side.
|
4656 |
* We return the result in tcg_right, and guarantee not to
|
4657 |
* trash tcg_left.
|
4658 |
*/
|
4659 |
TCGv_i64 tcg_tmp = tcg_temp_new_i64(); |
4660 |
assert(pos > 0 && pos < 64); |
4661 |
|
4662 |
tcg_gen_shri_i64(tcg_right, tcg_right, pos); |
4663 |
tcg_gen_shli_i64(tcg_tmp, tcg_left, 64 - pos);
|
4664 |
tcg_gen_or_i64(tcg_right, tcg_right, tcg_tmp); |
4665 |
|
4666 |
tcg_temp_free_i64(tcg_tmp); |
4667 |
} |
4668 |
|
4669 |
/* C3.6.1 EXT
|
4670 |
* 31 30 29 24 23 22 21 20 16 15 14 11 10 9 5 4 0
|
4671 |
* +---+---+-------------+-----+---+------+---+------+---+------+------+
|
4672 |
* | 0 | Q | 1 0 1 1 1 0 | op2 | 0 | Rm | 0 | imm4 | 0 | Rn | Rd |
|
4673 |
* +---+---+-------------+-----+---+------+---+------+---+------+------+
|
4674 |
*/
|
4675 |
static void disas_simd_ext(DisasContext *s, uint32_t insn) |
4676 |
{ |
4677 |
int is_q = extract32(insn, 30, 1); |
4678 |
int op2 = extract32(insn, 22, 2); |
4679 |
int imm4 = extract32(insn, 11, 4); |
4680 |
int rm = extract32(insn, 16, 5); |
4681 |
int rn = extract32(insn, 5, 5); |
4682 |
int rd = extract32(insn, 0, 5); |
4683 |
int pos = imm4 << 3; |
4684 |
TCGv_i64 tcg_resl, tcg_resh; |
4685 |
|
4686 |
if (op2 != 0 || (!is_q && extract32(imm4, 3, 1))) { |
4687 |
unallocated_encoding(s); |
4688 |
return;
|
4689 |
} |
4690 |
|
4691 |
tcg_resh = tcg_temp_new_i64(); |
4692 |
tcg_resl = tcg_temp_new_i64(); |
4693 |
|
4694 |
/* Vd gets bits starting at pos bits into Vm:Vn. This is
|
4695 |
* either extracting 128 bits from a 128:128 concatenation, or
|
4696 |
* extracting 64 bits from a 64:64 concatenation.
|
4697 |
*/
|
4698 |
if (!is_q) {
|
4699 |
read_vec_element(s, tcg_resl, rn, 0, MO_64);
|
4700 |
if (pos != 0) { |
4701 |
read_vec_element(s, tcg_resh, rm, 0, MO_64);
|
4702 |
do_ext64(s, tcg_resh, tcg_resl, pos); |
4703 |
} |
4704 |
tcg_gen_movi_i64(tcg_resh, 0);
|
4705 |
} else {
|
4706 |
TCGv_i64 tcg_hh; |
4707 |
typedef struct { |
4708 |
int reg;
|
4709 |
int elt;
|
4710 |
} EltPosns; |
4711 |
EltPosns eltposns[] = { {rn, 0}, {rn, 1}, {rm, 0}, {rm, 1} }; |
4712 |
EltPosns *elt = eltposns; |
4713 |
|
4714 |
if (pos >= 64) { |
4715 |
elt++; |
4716 |
pos -= 64;
|
4717 |
} |
4718 |
|
4719 |
read_vec_element(s, tcg_resl, elt->reg, elt->elt, MO_64); |
4720 |
elt++; |
4721 |
read_vec_element(s, tcg_resh, elt->reg, elt->elt, MO_64); |
4722 |
elt++; |
4723 |
if (pos != 0) { |
4724 |
do_ext64(s, tcg_resh, tcg_resl, pos); |
4725 |
tcg_hh = tcg_temp_new_i64(); |
4726 |
read_vec_element(s, tcg_hh, elt->reg, elt->elt, MO_64); |
4727 |
do_ext64(s, tcg_hh, tcg_resh, pos); |
4728 |
tcg_temp_free_i64(tcg_hh); |
4729 |
} |
4730 |
} |
4731 |
|
4732 |
write_vec_element(s, tcg_resl, rd, 0, MO_64);
|
4733 |
tcg_temp_free_i64(tcg_resl); |
4734 |
write_vec_element(s, tcg_resh, rd, 1, MO_64);
|
4735 |
tcg_temp_free_i64(tcg_resh); |
4736 |
} |
4737 |
|
4738 |
/* C3.6.2 TBL/TBX
|
4739 |
* 31 30 29 24 23 22 21 20 16 15 14 13 12 11 10 9 5 4 0
|
4740 |
* +---+---+-------------+-----+---+------+---+-----+----+-----+------+------+
|
4741 |
* | 0 | Q | 0 0 1 1 1 0 | op2 | 0 | Rm | 0 | len | op | 0 0 | Rn | Rd |
|
4742 |
* +---+---+-------------+-----+---+------+---+-----+----+-----+------+------+
|
4743 |
*/
|
4744 |
static void disas_simd_tb(DisasContext *s, uint32_t insn) |
4745 |
{ |
4746 |
int op2 = extract32(insn, 22, 2); |
4747 |
int is_q = extract32(insn, 30, 1); |
4748 |
int rm = extract32(insn, 16, 5); |
4749 |
int rn = extract32(insn, 5, 5); |
4750 |
int rd = extract32(insn, 0, 5); |
4751 |
int is_tblx = extract32(insn, 12, 1); |
4752 |
int len = extract32(insn, 13, 2); |
4753 |
TCGv_i64 tcg_resl, tcg_resh, tcg_idx; |
4754 |
TCGv_i32 tcg_regno, tcg_numregs; |
4755 |
|
4756 |
if (op2 != 0) { |
4757 |
unallocated_encoding(s); |
4758 |
return;
|
4759 |
} |
4760 |
|
4761 |
/* This does a table lookup: for every byte element in the input
|
4762 |
* we index into a table formed from up to four vector registers,
|
4763 |
* and then the output is the result of the lookups. Our helper
|
4764 |
* function does the lookup operation for a single 64 bit part of
|
4765 |
* the input.
|
4766 |
*/
|
4767 |
tcg_resl = tcg_temp_new_i64(); |
4768 |
tcg_resh = tcg_temp_new_i64(); |
4769 |
|
4770 |
if (is_tblx) {
|
4771 |
read_vec_element(s, tcg_resl, rd, 0, MO_64);
|
4772 |
} else {
|
4773 |
tcg_gen_movi_i64(tcg_resl, 0);
|
4774 |
} |
4775 |
if (is_tblx && is_q) {
|
4776 |
read_vec_element(s, tcg_resh, rd, 1, MO_64);
|
4777 |
} else {
|
4778 |
tcg_gen_movi_i64(tcg_resh, 0);
|
4779 |
} |
4780 |
|
4781 |
tcg_idx = tcg_temp_new_i64(); |
4782 |
tcg_regno = tcg_const_i32(rn); |
4783 |
tcg_numregs = tcg_const_i32(len + 1);
|
4784 |
read_vec_element(s, tcg_idx, rm, 0, MO_64);
|
4785 |
gen_helper_simd_tbl(tcg_resl, cpu_env, tcg_resl, tcg_idx, |
4786 |
tcg_regno, tcg_numregs); |
4787 |
if (is_q) {
|
4788 |
read_vec_element(s, tcg_idx, rm, 1, MO_64);
|
4789 |
gen_helper_simd_tbl(tcg_resh, cpu_env, tcg_resh, tcg_idx, |
4790 |
tcg_regno, tcg_numregs); |
4791 |
} |
4792 |
tcg_temp_free_i64(tcg_idx); |
4793 |
tcg_temp_free_i32(tcg_regno); |
4794 |
tcg_temp_free_i32(tcg_numregs); |
4795 |
|
4796 |
write_vec_element(s, tcg_resl, rd, 0, MO_64);
|
4797 |
tcg_temp_free_i64(tcg_resl); |
4798 |
write_vec_element(s, tcg_resh, rd, 1, MO_64);
|
4799 |
tcg_temp_free_i64(tcg_resh); |
4800 |
} |
4801 |
|
4802 |
/* C3.6.3 ZIP/UZP/TRN
|
4803 |
* 31 30 29 24 23 22 21 20 16 15 14 12 11 10 9 5 4 0
|
4804 |
* +---+---+-------------+------+---+------+---+------------------+------+
|
4805 |
* | 0 | Q | 0 0 1 1 1 0 | size | 0 | Rm | 0 | opc | 1 0 | Rn | Rd |
|
4806 |
* +---+---+-------------+------+---+------+---+------------------+------+
|
4807 |
*/
|
4808 |
static void disas_simd_zip_trn(DisasContext *s, uint32_t insn) |
4809 |
{ |
4810 |
int rd = extract32(insn, 0, 5); |
4811 |
int rn = extract32(insn, 5, 5); |
4812 |
int rm = extract32(insn, 16, 5); |
4813 |
int size = extract32(insn, 22, 2); |
4814 |
/* opc field bits [1:0] indicate ZIP/UZP/TRN;
|
4815 |
* bit 2 indicates 1 vs 2 variant of the insn.
|
4816 |
*/
|
4817 |
int opcode = extract32(insn, 12, 2); |
4818 |
bool part = extract32(insn, 14, 1); |
4819 |
bool is_q = extract32(insn, 30, 1); |
4820 |
int esize = 8 << size; |
4821 |
int i, ofs;
|
4822 |
int datasize = is_q ? 128 : 64; |
4823 |
int elements = datasize / esize;
|
4824 |
TCGv_i64 tcg_res, tcg_resl, tcg_resh; |
4825 |
|
4826 |
if (opcode == 0 || (size == 3 && !is_q)) { |
4827 |
unallocated_encoding(s); |
4828 |
return;
|
4829 |
} |
4830 |
|
4831 |
tcg_resl = tcg_const_i64(0);
|
4832 |
tcg_resh = tcg_const_i64(0);
|
4833 |
tcg_res = tcg_temp_new_i64(); |
4834 |
|
4835 |
for (i = 0; i < elements; i++) { |
4836 |
switch (opcode) {
|
4837 |
case 1: /* UZP1/2 */ |
4838 |
{ |
4839 |
int midpoint = elements / 2; |
4840 |
if (i < midpoint) {
|
4841 |
read_vec_element(s, tcg_res, rn, 2 * i + part, size);
|
4842 |
} else {
|
4843 |
read_vec_element(s, tcg_res, rm, |
4844 |
2 * (i - midpoint) + part, size);
|
4845 |
} |
4846 |
break;
|
4847 |
} |
4848 |
case 2: /* TRN1/2 */ |
4849 |
if (i & 1) { |
4850 |
read_vec_element(s, tcg_res, rm, (i & ~1) + part, size);
|
4851 |
} else {
|
4852 |
read_vec_element(s, tcg_res, rn, (i & ~1) + part, size);
|
4853 |
} |
4854 |
break;
|
4855 |
case 3: /* ZIP1/2 */ |
4856 |
{ |
4857 |
int base = part * elements / 2; |
4858 |
if (i & 1) { |
4859 |
read_vec_element(s, tcg_res, rm, base + (i >> 1), size);
|
4860 |
} else {
|
4861 |
read_vec_element(s, tcg_res, rn, base + (i >> 1), size);
|
4862 |
} |
4863 |
break;
|
4864 |
} |
4865 |
default:
|
4866 |
g_assert_not_reached(); |
4867 |
} |
4868 |
|
4869 |
ofs = i * esize; |
4870 |
if (ofs < 64) { |
4871 |
tcg_gen_shli_i64(tcg_res, tcg_res, ofs); |
4872 |
tcg_gen_or_i64(tcg_resl, tcg_resl, tcg_res); |
4873 |
} else {
|
4874 |
tcg_gen_shli_i64(tcg_res, tcg_res, ofs - 64);
|
4875 |
tcg_gen_or_i64(tcg_resh, tcg_resh, tcg_res); |
4876 |
} |
4877 |
} |
4878 |
|
4879 |
tcg_temp_free_i64(tcg_res); |
4880 |
|
4881 |
write_vec_element(s, tcg_resl, rd, 0, MO_64);
|
4882 |
tcg_temp_free_i64(tcg_resl); |
4883 |
write_vec_element(s, tcg_resh, rd, 1, MO_64);
|
4884 |
tcg_temp_free_i64(tcg_resh); |
4885 |
} |
4886 |
|
4887 |
static void do_minmaxop(DisasContext *s, TCGv_i32 tcg_elt1, TCGv_i32 tcg_elt2, |
4888 |
int opc, bool is_min, TCGv_ptr fpst) |
4889 |
{ |
4890 |
/* Helper function for disas_simd_across_lanes: do a single precision
|
4891 |
* min/max operation on the specified two inputs,
|
4892 |
* and return the result in tcg_elt1.
|
4893 |
*/
|
4894 |
if (opc == 0xc) { |
4895 |
if (is_min) {
|
4896 |
gen_helper_vfp_minnums(tcg_elt1, tcg_elt1, tcg_elt2, fpst); |
4897 |
} else {
|
4898 |
gen_helper_vfp_maxnums(tcg_elt1, tcg_elt1, tcg_elt2, fpst); |
4899 |
} |
4900 |
} else {
|
4901 |
assert(opc == 0xf);
|
4902 |
if (is_min) {
|
4903 |
gen_helper_vfp_mins(tcg_elt1, tcg_elt1, tcg_elt2, fpst); |
4904 |
} else {
|
4905 |
gen_helper_vfp_maxs(tcg_elt1, tcg_elt1, tcg_elt2, fpst); |
4906 |
} |
4907 |
} |
4908 |
} |
4909 |
|
4910 |
/* C3.6.4 AdvSIMD across lanes
|
4911 |
* 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
|
4912 |
* +---+---+---+-----------+------+-----------+--------+-----+------+------+
|
4913 |
* | 0 | Q | U | 0 1 1 1 0 | size | 1 1 0 0 0 | opcode | 1 0 | Rn | Rd |
|
4914 |
* +---+---+---+-----------+------+-----------+--------+-----+------+------+
|
4915 |
*/
|
4916 |
static void disas_simd_across_lanes(DisasContext *s, uint32_t insn) |
4917 |
{ |
4918 |
int rd = extract32(insn, 0, 5); |
4919 |
int rn = extract32(insn, 5, 5); |
4920 |
int size = extract32(insn, 22, 2); |
4921 |
int opcode = extract32(insn, 12, 5); |
4922 |
bool is_q = extract32(insn, 30, 1); |
4923 |
bool is_u = extract32(insn, 29, 1); |
4924 |
bool is_fp = false; |
4925 |
bool is_min = false; |
4926 |
int esize;
|
4927 |
int elements;
|
4928 |
int i;
|
4929 |
TCGv_i64 tcg_res, tcg_elt; |
4930 |
|
4931 |
switch (opcode) {
|
4932 |
case 0x1b: /* ADDV */ |
4933 |
if (is_u) {
|
4934 |
unallocated_encoding(s); |
4935 |
return;
|
4936 |
} |
4937 |
/* fall through */
|
4938 |
case 0x3: /* SADDLV, UADDLV */ |
4939 |
case 0xa: /* SMAXV, UMAXV */ |
4940 |
case 0x1a: /* SMINV, UMINV */ |
4941 |
if (size == 3 || (size == 2 && !is_q)) { |
4942 |
unallocated_encoding(s); |
4943 |
return;
|
4944 |
} |
4945 |
break;
|
4946 |
case 0xc: /* FMAXNMV, FMINNMV */ |
4947 |
case 0xf: /* FMAXV, FMINV */ |
4948 |
if (!is_u || !is_q || extract32(size, 0, 1)) { |
4949 |
unallocated_encoding(s); |
4950 |
return;
|
4951 |
} |
4952 |
/* Bit 1 of size field encodes min vs max, and actual size is always
|
4953 |
* 32 bits: adjust the size variable so following code can rely on it
|
4954 |
*/
|
4955 |
is_min = extract32(size, 1, 1); |
4956 |
is_fp = true;
|
4957 |
size = 2;
|
4958 |
break;
|
4959 |
default:
|
4960 |
unallocated_encoding(s); |
4961 |
return;
|
4962 |
} |
4963 |
|
4964 |
esize = 8 << size;
|
4965 |
elements = (is_q ? 128 : 64) / esize; |
4966 |
|
4967 |
tcg_res = tcg_temp_new_i64(); |
4968 |
tcg_elt = tcg_temp_new_i64(); |
4969 |
|
4970 |
/* These instructions operate across all lanes of a vector
|
4971 |
* to produce a single result. We can guarantee that a 64
|
4972 |
* bit intermediate is sufficient:
|
4973 |
* + for [US]ADDLV the maximum element size is 32 bits, and
|
4974 |
* the result type is 64 bits
|
4975 |
* + for FMAX*V, FMIN*V, ADDV the intermediate type is the
|
4976 |
* same as the element size, which is 32 bits at most
|
4977 |
* For the integer operations we can choose to work at 64
|
4978 |
* or 32 bits and truncate at the end; for simplicity
|
4979 |
* we use 64 bits always. The floating point
|
4980 |
* ops do require 32 bit intermediates, though.
|
4981 |
*/
|
4982 |
if (!is_fp) {
|
4983 |
read_vec_element(s, tcg_res, rn, 0, size | (is_u ? 0 : MO_SIGN)); |
4984 |
|
4985 |
for (i = 1; i < elements; i++) { |
4986 |
read_vec_element(s, tcg_elt, rn, i, size | (is_u ? 0 : MO_SIGN));
|
4987 |
|
4988 |
switch (opcode) {
|
4989 |
case 0x03: /* SADDLV / UADDLV */ |
4990 |
case 0x1b: /* ADDV */ |
4991 |
tcg_gen_add_i64(tcg_res, tcg_res, tcg_elt); |
4992 |
break;
|
4993 |
case 0x0a: /* SMAXV / UMAXV */ |
4994 |
tcg_gen_movcond_i64(is_u ? TCG_COND_GEU : TCG_COND_GE, |
4995 |
tcg_res, |
4996 |
tcg_res, tcg_elt, tcg_res, tcg_elt); |
4997 |
break;
|
4998 |
case 0x1a: /* SMINV / UMINV */ |
4999 |
tcg_gen_movcond_i64(is_u ? TCG_COND_LEU : TCG_COND_LE, |
5000 |
tcg_res, |
5001 |
tcg_res, tcg_elt, tcg_res, tcg_elt); |
5002 |
break;
|
5003 |
break;
|
5004 |
default:
|
5005 |
g_assert_not_reached(); |
5006 |
} |
5007 |
|
5008 |
} |
5009 |
} else {
|
5010 |
/* Floating point ops which work on 32 bit (single) intermediates.
|
5011 |
* Note that correct NaN propagation requires that we do these
|
5012 |
* operations in exactly the order specified by the pseudocode.
|
5013 |
*/
|
5014 |
TCGv_i32 tcg_elt1 = tcg_temp_new_i32(); |
5015 |
TCGv_i32 tcg_elt2 = tcg_temp_new_i32(); |
5016 |
TCGv_i32 tcg_elt3 = tcg_temp_new_i32(); |
5017 |
TCGv_ptr fpst = get_fpstatus_ptr(); |
5018 |
|
5019 |
assert(esize == 32);
|
5020 |
assert(elements == 4);
|
5021 |
|
5022 |
read_vec_element(s, tcg_elt, rn, 0, MO_32);
|
5023 |
tcg_gen_trunc_i64_i32(tcg_elt1, tcg_elt); |
5024 |
read_vec_element(s, tcg_elt, rn, 1, MO_32);
|
5025 |
tcg_gen_trunc_i64_i32(tcg_elt2, tcg_elt); |
5026 |
|
5027 |
do_minmaxop(s, tcg_elt1, tcg_elt2, opcode, is_min, fpst); |
5028 |
|
5029 |
read_vec_element(s, tcg_elt, rn, 2, MO_32);
|
5030 |
tcg_gen_trunc_i64_i32(tcg_elt2, tcg_elt); |
5031 |
read_vec_element(s, tcg_elt, rn, 3, MO_32);
|
5032 |
tcg_gen_trunc_i64_i32(tcg_elt3, tcg_elt); |
5033 |
|
5034 |
do_minmaxop(s, tcg_elt2, tcg_elt3, opcode, is_min, fpst); |
5035 |
|
5036 |
do_minmaxop(s, tcg_elt1, tcg_elt2, opcode, is_min, fpst); |
5037 |
|
5038 |
tcg_gen_extu_i32_i64(tcg_res, tcg_elt1); |
5039 |
tcg_temp_free_i32(tcg_elt1); |
5040 |
tcg_temp_free_i32(tcg_elt2); |
5041 |
tcg_temp_free_i32(tcg_elt3); |
5042 |
tcg_temp_free_ptr(fpst); |
5043 |
} |
5044 |
|
5045 |
tcg_temp_free_i64(tcg_elt); |
5046 |
|
5047 |
/* Now truncate the result to the width required for the final output */
|
5048 |
if (opcode == 0x03) { |
5049 |
/* SADDLV, UADDLV: result is 2*esize */
|
5050 |
size++; |
5051 |
} |
5052 |
|
5053 |
switch (size) {
|
5054 |
case 0: |
5055 |
tcg_gen_ext8u_i64(tcg_res, tcg_res); |
5056 |
break;
|
5057 |
case 1: |
5058 |
tcg_gen_ext16u_i64(tcg_res, tcg_res); |
5059 |
break;
|
5060 |
case 2: |
5061 |
tcg_gen_ext32u_i64(tcg_res, tcg_res); |
5062 |
break;
|
5063 |
case 3: |
5064 |
break;
|
5065 |
default:
|
5066 |
g_assert_not_reached(); |
5067 |
} |
5068 |
|
5069 |
write_fp_dreg(s, rd, tcg_res); |
5070 |
tcg_temp_free_i64(tcg_res); |
5071 |
} |
5072 |
|
5073 |
/* C6.3.31 DUP (Element, Vector)
|
5074 |
*
|
5075 |
* 31 30 29 21 20 16 15 10 9 5 4 0
|
5076 |
* +---+---+-------------------+--------+-------------+------+------+
|
5077 |
* | 0 | Q | 0 0 1 1 1 0 0 0 0 | imm5 | 0 0 0 0 0 1 | Rn | Rd |
|
5078 |
* +---+---+-------------------+--------+-------------+------+------+
|
5079 |
*
|
5080 |
* size: encoded in imm5 (see ARM ARM LowestSetBit())
|
5081 |
*/
|
5082 |
static void handle_simd_dupe(DisasContext *s, int is_q, int rd, int rn, |
5083 |
int imm5)
|
5084 |
{ |
5085 |
int size = ctz32(imm5);
|
5086 |
int esize = 8 << size; |
5087 |
int elements = (is_q ? 128 : 64) / esize; |
5088 |
int index, i;
|
5089 |
TCGv_i64 tmp; |
5090 |
|
5091 |
if (size > 3 || (size == 3 && !is_q)) { |
5092 |
unallocated_encoding(s); |
5093 |
return;
|
5094 |
} |
5095 |
|
5096 |
index = imm5 >> (size + 1);
|
5097 |
|
5098 |
tmp = tcg_temp_new_i64(); |
5099 |
read_vec_element(s, tmp, rn, index, size); |
5100 |
|
5101 |
for (i = 0; i < elements; i++) { |
5102 |
write_vec_element(s, tmp, rd, i, size); |
5103 |
} |
5104 |
|
5105 |
if (!is_q) {
|
5106 |
clear_vec_high(s, rd); |
5107 |
} |
5108 |
|
5109 |
tcg_temp_free_i64(tmp); |
5110 |
} |
5111 |
|
5112 |
/* C6.3.31 DUP (element, scalar)
|
5113 |
* 31 21 20 16 15 10 9 5 4 0
|
5114 |
* +-----------------------+--------+-------------+------+------+
|
5115 |
* | 0 1 0 1 1 1 1 0 0 0 0 | imm5 | 0 0 0 0 0 1 | Rn | Rd |
|
5116 |
* +-----------------------+--------+-------------+------+------+
|
5117 |
*/
|
5118 |
static void handle_simd_dupes(DisasContext *s, int rd, int rn, |
5119 |
int imm5)
|
5120 |
{ |
5121 |
int size = ctz32(imm5);
|
5122 |
int index;
|
5123 |
TCGv_i64 tmp; |
5124 |
|
5125 |
if (size > 3) { |
5126 |
unallocated_encoding(s); |
5127 |
return;
|
5128 |
} |
5129 |
|
5130 |
index = imm5 >> (size + 1);
|
5131 |
|
5132 |
/* This instruction just extracts the specified element and
|
5133 |
* zero-extends it into the bottom of the destination register.
|
5134 |
*/
|
5135 |
tmp = tcg_temp_new_i64(); |
5136 |
read_vec_element(s, tmp, rn, index, size); |
5137 |
write_fp_dreg(s, rd, tmp); |
5138 |
tcg_temp_free_i64(tmp); |
5139 |
} |
5140 |
|
5141 |
/* C6.3.32 DUP (General)
|
5142 |
*
|
5143 |
* 31 30 29 21 20 16 15 10 9 5 4 0
|
5144 |
* +---+---+-------------------+--------+-------------+------+------+
|
5145 |
* | 0 | Q | 0 0 1 1 1 0 0 0 0 | imm5 | 0 0 0 0 1 1 | Rn | Rd |
|
5146 |
* +---+---+-------------------+--------+-------------+------+------+
|
5147 |
*
|
5148 |
* size: encoded in imm5 (see ARM ARM LowestSetBit())
|
5149 |
*/
|
5150 |
static void handle_simd_dupg(DisasContext *s, int is_q, int rd, int rn, |
5151 |
int imm5)
|
5152 |
{ |
5153 |
int size = ctz32(imm5);
|
5154 |
int esize = 8 << size; |
5155 |
int elements = (is_q ? 128 : 64)/esize; |
5156 |
int i = 0; |
5157 |
|
5158 |
if (size > 3 || ((size == 3) && !is_q)) { |
5159 |
unallocated_encoding(s); |
5160 |
return;
|
5161 |
} |
5162 |
for (i = 0; i < elements; i++) { |
5163 |
write_vec_element(s, cpu_reg(s, rn), rd, i, size); |
5164 |
} |
5165 |
if (!is_q) {
|
5166 |
clear_vec_high(s, rd); |
5167 |
} |
5168 |
} |
5169 |
|
5170 |
/* C6.3.150 INS (Element)
|
5171 |
*
|
5172 |
* 31 21 20 16 15 14 11 10 9 5 4 0
|
5173 |
* +-----------------------+--------+------------+---+------+------+
|
5174 |
* | 0 1 1 0 1 1 1 0 0 0 0 | imm5 | 0 | imm4 | 1 | Rn | Rd |
|
5175 |
* +-----------------------+--------+------------+---+------+------+
|
5176 |
*
|
5177 |
* size: encoded in imm5 (see ARM ARM LowestSetBit())
|
5178 |
* index: encoded in imm5<4:size+1>
|
5179 |
*/
|
5180 |
static void handle_simd_inse(DisasContext *s, int rd, int rn, |
5181 |
int imm4, int imm5) |
5182 |
{ |
5183 |
int size = ctz32(imm5);
|
5184 |
int src_index, dst_index;
|
5185 |
TCGv_i64 tmp; |
5186 |
|
5187 |
if (size > 3) { |
5188 |
unallocated_encoding(s); |
5189 |
return;
|
5190 |
} |
5191 |
dst_index = extract32(imm5, 1+size, 5); |
5192 |
src_index = extract32(imm4, size, 4);
|
5193 |
|
5194 |
tmp = tcg_temp_new_i64(); |
5195 |
|
5196 |
read_vec_element(s, tmp, rn, src_index, size); |
5197 |
write_vec_element(s, tmp, rd, dst_index, size); |
5198 |
|
5199 |
tcg_temp_free_i64(tmp); |
5200 |
} |
5201 |
|
5202 |
|
5203 |
/* C6.3.151 INS (General)
|
5204 |
*
|
5205 |
* 31 21 20 16 15 10 9 5 4 0
|
5206 |
* +-----------------------+--------+-------------+------+------+
|
5207 |
* | 0 1 0 0 1 1 1 0 0 0 0 | imm5 | 0 0 0 1 1 1 | Rn | Rd |
|
5208 |
* +-----------------------+--------+-------------+------+------+
|
5209 |
*
|
5210 |
* size: encoded in imm5 (see ARM ARM LowestSetBit())
|
5211 |
* index: encoded in imm5<4:size+1>
|
5212 |
*/
|
5213 |
static void handle_simd_insg(DisasContext *s, int rd, int rn, int imm5) |
5214 |
{ |
5215 |
int size = ctz32(imm5);
|
5216 |
int idx;
|
5217 |
|
5218 |
if (size > 3) { |
5219 |
unallocated_encoding(s); |
5220 |
return;
|
5221 |
} |
5222 |
|
5223 |
idx = extract32(imm5, 1 + size, 4 - size); |
5224 |
write_vec_element(s, cpu_reg(s, rn), rd, idx, size); |
5225 |
} |
5226 |
|
5227 |
/*
|
5228 |
* C6.3.321 UMOV (General)
|
5229 |
* C6.3.237 SMOV (General)
|
5230 |
*
|
5231 |
* 31 30 29 21 20 16 15 12 10 9 5 4 0
|
5232 |
* +---+---+-------------------+--------+-------------+------+------+
|
5233 |
* | 0 | Q | 0 0 1 1 1 0 0 0 0 | imm5 | 0 0 1 U 1 1 | Rn | Rd |
|
5234 |
* +---+---+-------------------+--------+-------------+------+------+
|
5235 |
*
|
5236 |
* U: unsigned when set
|
5237 |
* size: encoded in imm5 (see ARM ARM LowestSetBit())
|
5238 |
*/
|
5239 |
static void handle_simd_umov_smov(DisasContext *s, int is_q, int is_signed, |
5240 |
int rn, int rd, int imm5) |
5241 |
{ |
5242 |
int size = ctz32(imm5);
|
5243 |
int element;
|
5244 |
TCGv_i64 tcg_rd; |
5245 |
|
5246 |
/* Check for UnallocatedEncodings */
|
5247 |
if (is_signed) {
|
5248 |
if (size > 2 || (size == 2 && !is_q)) { |
5249 |
unallocated_encoding(s); |
5250 |
return;
|
5251 |
} |
5252 |
} else {
|
5253 |
if (size > 3 |
5254 |
|| (size < 3 && is_q)
|
5255 |
|| (size == 3 && !is_q)) {
|
5256 |
unallocated_encoding(s); |
5257 |
return;
|
5258 |
} |
5259 |
} |
5260 |
element = extract32(imm5, 1+size, 4); |
5261 |
|
5262 |
tcg_rd = cpu_reg(s, rd); |
5263 |
read_vec_element(s, tcg_rd, rn, element, size | (is_signed ? MO_SIGN : 0));
|
5264 |
if (is_signed && !is_q) {
|
5265 |
tcg_gen_ext32u_i64(tcg_rd, tcg_rd); |
5266 |
} |
5267 |
} |
5268 |
|
5269 |
/* C3.6.5 AdvSIMD copy
|
5270 |
* 31 30 29 28 21 20 16 15 14 11 10 9 5 4 0
|
5271 |
* +---+---+----+-----------------+------+---+------+---+------+------+
|
5272 |
* | 0 | Q | op | 0 1 1 1 0 0 0 0 | imm5 | 0 | imm4 | 1 | Rn | Rd |
|
5273 |
* +---+---+----+-----------------+------+---+------+---+------+------+
|
5274 |
*/
|
5275 |
static void disas_simd_copy(DisasContext *s, uint32_t insn) |
5276 |
{ |
5277 |
int rd = extract32(insn, 0, 5); |
5278 |
int rn = extract32(insn, 5, 5); |
5279 |
int imm4 = extract32(insn, 11, 4); |
5280 |
int op = extract32(insn, 29, 1); |
5281 |
int is_q = extract32(insn, 30, 1); |
5282 |
int imm5 = extract32(insn, 16, 5); |
5283 |
|
5284 |
if (op) {
|
5285 |
if (is_q) {
|
5286 |
/* INS (element) */
|
5287 |
handle_simd_inse(s, rd, rn, imm4, imm5); |
5288 |
} else {
|
5289 |
unallocated_encoding(s); |
5290 |
} |
5291 |
} else {
|
5292 |
switch (imm4) {
|
5293 |
case 0: |
5294 |
/* DUP (element - vector) */
|
5295 |
handle_simd_dupe(s, is_q, rd, rn, imm5); |
5296 |
break;
|
5297 |
case 1: |
5298 |
/* DUP (general) */
|
5299 |
handle_simd_dupg(s, is_q, rd, rn, imm5); |
5300 |
break;
|
5301 |
case 3: |
5302 |
if (is_q) {
|
5303 |
/* INS (general) */
|
5304 |
handle_simd_insg(s, rd, rn, imm5); |
5305 |
} else {
|
5306 |
unallocated_encoding(s); |
5307 |
} |
5308 |
break;
|
5309 |
case 5: |
5310 |
case 7: |
5311 |
/* UMOV/SMOV (is_q indicates 32/64; imm4 indicates signedness) */
|
5312 |
handle_simd_umov_smov(s, is_q, (imm4 == 5), rn, rd, imm5);
|
5313 |
break;
|
5314 |
default:
|
5315 |
unallocated_encoding(s); |
5316 |
break;
|
5317 |
} |
5318 |
} |
5319 |
} |
5320 |
|
5321 |
/* C3.6.6 AdvSIMD modified immediate
|
5322 |
* 31 30 29 28 19 18 16 15 12 11 10 9 5 4 0
|
5323 |
* +---+---+----+---------------------+-----+-------+----+---+-------+------+
|
5324 |
* | 0 | Q | op | 0 1 1 1 1 0 0 0 0 0 | abc | cmode | o2 | 1 | defgh | Rd |
|
5325 |
* +---+---+----+---------------------+-----+-------+----+---+-------+------+
|
5326 |
*
|
5327 |
* There are a number of operations that can be carried out here:
|
5328 |
* MOVI - move (shifted) imm into register
|
5329 |
* MVNI - move inverted (shifted) imm into register
|
5330 |
* ORR - bitwise OR of (shifted) imm with register
|
5331 |
* BIC - bitwise clear of (shifted) imm with register
|
5332 |
*/
|
5333 |
static void disas_simd_mod_imm(DisasContext *s, uint32_t insn) |
5334 |
{ |
5335 |
int rd = extract32(insn, 0, 5); |
5336 |
int cmode = extract32(insn, 12, 4); |
5337 |
int cmode_3_1 = extract32(cmode, 1, 3); |
5338 |
int cmode_0 = extract32(cmode, 0, 1); |
5339 |
int o2 = extract32(insn, 11, 1); |
5340 |
uint64_t abcdefgh = extract32(insn, 5, 5) | (extract32(insn, 16, 3) << 5); |
5341 |
bool is_neg = extract32(insn, 29, 1); |
5342 |
bool is_q = extract32(insn, 30, 1); |
5343 |
uint64_t imm = 0;
|
5344 |
TCGv_i64 tcg_rd, tcg_imm; |
5345 |
int i;
|
5346 |
|
5347 |
if (o2 != 0 || ((cmode == 0xf) && is_neg && !is_q)) { |
5348 |
unallocated_encoding(s); |
5349 |
return;
|
5350 |
} |
5351 |
|
5352 |
/* See AdvSIMDExpandImm() in ARM ARM */
|
5353 |
switch (cmode_3_1) {
|
5354 |
case 0: /* Replicate(Zeros(24):imm8, 2) */ |
5355 |
case 1: /* Replicate(Zeros(16):imm8:Zeros(8), 2) */ |
5356 |
case 2: /* Replicate(Zeros(8):imm8:Zeros(16), 2) */ |
5357 |
case 3: /* Replicate(imm8:Zeros(24), 2) */ |
5358 |
{ |
5359 |
int shift = cmode_3_1 * 8; |
5360 |
imm = bitfield_replicate(abcdefgh << shift, 32);
|
5361 |
break;
|
5362 |
} |
5363 |
case 4: /* Replicate(Zeros(8):imm8, 4) */ |
5364 |
case 5: /* Replicate(imm8:Zeros(8), 4) */ |
5365 |
{ |
5366 |
int shift = (cmode_3_1 & 0x1) * 8; |
5367 |
imm = bitfield_replicate(abcdefgh << shift, 16);
|
5368 |
break;
|
5369 |
} |
5370 |
case 6: |
5371 |
if (cmode_0) {
|
5372 |
/* Replicate(Zeros(8):imm8:Ones(16), 2) */
|
5373 |
imm = (abcdefgh << 16) | 0xffff; |
5374 |
} else {
|
5375 |
/* Replicate(Zeros(16):imm8:Ones(8), 2) */
|
5376 |
imm = (abcdefgh << 8) | 0xff; |
5377 |
} |
5378 |
imm = bitfield_replicate(imm, 32);
|
5379 |
break;
|
5380 |
case 7: |
5381 |
if (!cmode_0 && !is_neg) {
|
5382 |
imm = bitfield_replicate(abcdefgh, 8);
|
5383 |
} else if (!cmode_0 && is_neg) { |
5384 |
int i;
|
5385 |
imm = 0;
|
5386 |
for (i = 0; i < 8; i++) { |
5387 |
if ((abcdefgh) & (1 << i)) { |
5388 |
imm |= 0xffULL << (i * 8); |
5389 |
} |
5390 |
} |
5391 |
} else if (cmode_0) { |
5392 |
if (is_neg) {
|
5393 |
imm = (abcdefgh & 0x3f) << 48; |
5394 |
if (abcdefgh & 0x80) { |
5395 |
imm |= 0x8000000000000000ULL;
|
5396 |
} |
5397 |
if (abcdefgh & 0x40) { |
5398 |
imm |= 0x3fc0000000000000ULL;
|
5399 |
} else {
|
5400 |
imm |= 0x4000000000000000ULL;
|
5401 |
} |
5402 |
} else {
|
5403 |
imm = (abcdefgh & 0x3f) << 19; |
5404 |
if (abcdefgh & 0x80) { |
5405 |
imm |= 0x80000000;
|
5406 |
} |
5407 |
if (abcdefgh & 0x40) { |
5408 |
imm |= 0x3e000000;
|
5409 |
} else {
|
5410 |
imm |= 0x40000000;
|
5411 |
} |
5412 |
imm |= (imm << 32);
|
5413 |
} |
5414 |
} |
5415 |
break;
|
5416 |
} |
5417 |
|
5418 |
if (cmode_3_1 != 7 && is_neg) { |
5419 |
imm = ~imm; |
5420 |
} |
5421 |
|
5422 |
tcg_imm = tcg_const_i64(imm); |
5423 |
tcg_rd = new_tmp_a64(s); |
5424 |
|
5425 |
for (i = 0; i < 2; i++) { |
5426 |
int foffs = i ? fp_reg_hi_offset(rd) : fp_reg_offset(rd, MO_64);
|
5427 |
|
5428 |
if (i == 1 && !is_q) { |
5429 |
/* non-quad ops clear high half of vector */
|
5430 |
tcg_gen_movi_i64(tcg_rd, 0);
|
5431 |
} else if ((cmode & 0x9) == 0x1 || (cmode & 0xd) == 0x9) { |
5432 |
tcg_gen_ld_i64(tcg_rd, cpu_env, foffs); |
5433 |
if (is_neg) {
|
5434 |
/* AND (BIC) */
|
5435 |
tcg_gen_and_i64(tcg_rd, tcg_rd, tcg_imm); |
5436 |
} else {
|
5437 |
/* ORR */
|
5438 |
tcg_gen_or_i64(tcg_rd, tcg_rd, tcg_imm); |
5439 |
} |
5440 |
} else {
|
5441 |
/* MOVI */
|
5442 |
tcg_gen_mov_i64(tcg_rd, tcg_imm); |
5443 |
} |
5444 |
tcg_gen_st_i64(tcg_rd, cpu_env, foffs); |
5445 |
} |
5446 |
|
5447 |
tcg_temp_free_i64(tcg_imm); |
5448 |
} |
5449 |
|
5450 |
/* C3.6.7 AdvSIMD scalar copy
|
5451 |
* 31 30 29 28 21 20 16 15 14 11 10 9 5 4 0
|
5452 |
* +-----+----+-----------------+------+---+------+---+------+------+
|
5453 |
* | 0 1 | op | 1 1 1 1 0 0 0 0 | imm5 | 0 | imm4 | 1 | Rn | Rd |
|
5454 |
* +-----+----+-----------------+------+---+------+---+------+------+
|
5455 |
*/
|
5456 |
static void disas_simd_scalar_copy(DisasContext *s, uint32_t insn) |
5457 |
{ |
5458 |
int rd = extract32(insn, 0, 5); |
5459 |
int rn = extract32(insn, 5, 5); |
5460 |
int imm4 = extract32(insn, 11, 4); |
5461 |
int imm5 = extract32(insn, 16, 5); |
5462 |
int op = extract32(insn, 29, 1); |
5463 |
|
5464 |
if (op != 0 || imm4 != 0) { |
5465 |
unallocated_encoding(s); |
5466 |
return;
|
5467 |
} |
5468 |
|
5469 |
/* DUP (element, scalar) */
|
5470 |
handle_simd_dupes(s, rd, rn, imm5); |
5471 |
} |
5472 |
|
5473 |
/* C3.6.8 AdvSIMD scalar pairwise
|
5474 |
* 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
|
5475 |
* +-----+---+-----------+------+-----------+--------+-----+------+------+
|
5476 |
* | 0 1 | U | 1 1 1 1 0 | size | 1 1 0 0 0 | opcode | 1 0 | Rn | Rd |
|
5477 |
* +-----+---+-----------+------+-----------+--------+-----+------+------+
|
5478 |
*/
|
5479 |
static void disas_simd_scalar_pairwise(DisasContext *s, uint32_t insn) |
5480 |
{ |
5481 |
unsupported_encoding(s, insn); |
5482 |
} |
5483 |
|
5484 |
/* C3.6.9 AdvSIMD scalar shift by immediate
|
5485 |
* 31 30 29 28 23 22 19 18 16 15 11 10 9 5 4 0
|
5486 |
* +-----+---+-------------+------+------+--------+---+------+------+
|
5487 |
* | 0 1 | U | 1 1 1 1 1 0 | immh | immb | opcode | 1 | Rn | Rd |
|
5488 |
* +-----+---+-------------+------+------+--------+---+------+------+
|
5489 |
*/
|
5490 |
static void disas_simd_scalar_shift_imm(DisasContext *s, uint32_t insn) |
5491 |
{ |
5492 |
unsupported_encoding(s, insn); |
5493 |
} |
5494 |
|
5495 |
/* C3.6.10 AdvSIMD scalar three different
|
5496 |
* 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
|
5497 |
* +-----+---+-----------+------+---+------+--------+-----+------+------+
|
5498 |
* | 0 1 | U | 1 1 1 1 0 | size | 1 | Rm | opcode | 0 0 | Rn | Rd |
|
5499 |
* +-----+---+-----------+------+---+------+--------+-----+------+------+
|
5500 |
*/
|
5501 |
static void disas_simd_scalar_three_reg_diff(DisasContext *s, uint32_t insn) |
5502 |
{ |
5503 |
unsupported_encoding(s, insn); |
5504 |
} |
5505 |
|
5506 |
/* C3.6.11 AdvSIMD scalar three same
|
5507 |
* 31 30 29 28 24 23 22 21 20 16 15 11 10 9 5 4 0
|
5508 |
* +-----+---+-----------+------+---+------+--------+---+------+------+
|
5509 |
* | 0 1 | U | 1 1 1 1 0 | size | 1 | Rm | opcode | 1 | Rn | Rd |
|
5510 |
* +-----+---+-----------+------+---+------+--------+---+------+------+
|
5511 |
*/
|
5512 |
static void disas_simd_scalar_three_reg_same(DisasContext *s, uint32_t insn) |
5513 |
{ |
5514 |
unsupported_encoding(s, insn); |
5515 |
} |
5516 |
|
5517 |
/* C3.6.12 AdvSIMD scalar two reg misc
|
5518 |
* 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
|
5519 |
* +-----+---+-----------+------+-----------+--------+-----+------+------+
|
5520 |
* | 0 1 | U | 1 1 1 1 0 | size | 1 0 0 0 0 | opcode | 1 0 | Rn | Rd |
|
5521 |
* +-----+---+-----------+------+-----------+--------+-----+------+------+
|
5522 |
*/
|
5523 |
static void disas_simd_scalar_two_reg_misc(DisasContext *s, uint32_t insn) |
5524 |
{ |
5525 |
unsupported_encoding(s, insn); |
5526 |
} |
5527 |
|
5528 |
/* C3.6.13 AdvSIMD scalar x indexed element
|
5529 |
* 31 30 29 28 24 23 22 21 20 19 16 15 12 11 10 9 5 4 0
|
5530 |
* +-----+---+-----------+------+---+---+------+-----+---+---+------+------+
|
5531 |
* | 0 1 | U | 1 1 1 1 1 | size | L | M | Rm | opc | H | 0 | Rn | Rd |
|
5532 |
* +-----+---+-----------+------+---+---+------+-----+---+---+------+------+
|
5533 |
*/
|
5534 |
static void disas_simd_scalar_indexed(DisasContext *s, uint32_t insn) |
5535 |
{ |
5536 |
unsupported_encoding(s, insn); |
5537 |
} |
5538 |
|
5539 |
/* C3.6.14 AdvSIMD shift by immediate
|
5540 |
* 31 30 29 28 23 22 19 18 16 15 11 10 9 5 4 0
|
5541 |
* +---+---+---+-------------+------+------+--------+---+------+------+
|
5542 |
* | 0 | Q | U | 0 1 1 1 1 0 | immh | immb | opcode | 1 | Rn | Rd |
|
5543 |
* +---+---+---+-------------+------+------+--------+---+------+------+
|
5544 |
*/
|
5545 |
static void disas_simd_shift_imm(DisasContext *s, uint32_t insn) |
5546 |
{ |
5547 |
unsupported_encoding(s, insn); |
5548 |
} |
5549 |
|
5550 |
static void handle_3rd_widening(DisasContext *s, int is_q, int is_u, int size, |
5551 |
int opcode, int rd, int rn, int rm) |
5552 |
{ |
5553 |
/* 3-reg-different widening insns: 64 x 64 -> 128 */
|
5554 |
TCGv_i64 tcg_res[2];
|
5555 |
int pass, accop;
|
5556 |
|
5557 |
tcg_res[0] = tcg_temp_new_i64();
|
5558 |
tcg_res[1] = tcg_temp_new_i64();
|
5559 |
|
5560 |
/* Does this op do an adding accumulate, a subtracting accumulate,
|
5561 |
* or no accumulate at all?
|
5562 |
*/
|
5563 |
switch (opcode) {
|
5564 |
case 5: |
5565 |
case 8: |
5566 |
case 9: |
5567 |
accop = 1;
|
5568 |
break;
|
5569 |
case 10: |
5570 |
case 11: |
5571 |
accop = -1;
|
5572 |
break;
|
5573 |
default:
|
5574 |
accop = 0;
|
5575 |
break;
|
5576 |
} |
5577 |
|
5578 |
if (accop != 0) { |
5579 |
read_vec_element(s, tcg_res[0], rd, 0, MO_64); |
5580 |
read_vec_element(s, tcg_res[1], rd, 1, MO_64); |
5581 |
} |
5582 |
|
5583 |
/* size == 2 means two 32x32->64 operations; this is worth special
|
5584 |
* casing because we can generally handle it inline.
|
5585 |
*/
|
5586 |
if (size == 2) { |
5587 |
for (pass = 0; pass < 2; pass++) { |
5588 |
TCGv_i64 tcg_op1 = tcg_temp_new_i64(); |
5589 |
TCGv_i64 tcg_op2 = tcg_temp_new_i64(); |
5590 |
TCGv_i64 tcg_passres; |
5591 |
TCGMemOp memop = MO_32 | (is_u ? 0 : MO_SIGN);
|
5592 |
|
5593 |
int elt = pass + is_q * 2; |
5594 |
|
5595 |
read_vec_element(s, tcg_op1, rn, elt, memop); |
5596 |
read_vec_element(s, tcg_op2, rm, elt, memop); |
5597 |
|
5598 |
if (accop == 0) { |
5599 |
tcg_passres = tcg_res[pass]; |
5600 |
} else {
|
5601 |
tcg_passres = tcg_temp_new_i64(); |
5602 |
} |
5603 |
|
5604 |
switch (opcode) {
|
5605 |
case 5: /* SABAL, SABAL2, UABAL, UABAL2 */ |
5606 |
case 7: /* SABDL, SABDL2, UABDL, UABDL2 */ |
5607 |
{ |
5608 |
TCGv_i64 tcg_tmp1 = tcg_temp_new_i64(); |
5609 |
TCGv_i64 tcg_tmp2 = tcg_temp_new_i64(); |
5610 |
|
5611 |
tcg_gen_sub_i64(tcg_tmp1, tcg_op1, tcg_op2); |
5612 |
tcg_gen_sub_i64(tcg_tmp2, tcg_op2, tcg_op1); |
5613 |
tcg_gen_movcond_i64(is_u ? TCG_COND_GEU : TCG_COND_GE, |
5614 |
tcg_passres, |
5615 |
tcg_op1, tcg_op2, tcg_tmp1, tcg_tmp2); |
5616 |
tcg_temp_free_i64(tcg_tmp1); |
5617 |
tcg_temp_free_i64(tcg_tmp2); |
5618 |
break;
|
5619 |
} |
5620 |
case 8: /* SMLAL, SMLAL2, UMLAL, UMLAL2 */ |
5621 |
case 10: /* SMLSL, SMLSL2, UMLSL, UMLSL2 */ |
5622 |
case 12: /* UMULL, UMULL2, SMULL, SMULL2 */ |
5623 |
tcg_gen_mul_i64(tcg_passres, tcg_op1, tcg_op2); |
5624 |
break;
|
5625 |
default:
|
5626 |
g_assert_not_reached(); |
5627 |
} |
5628 |
|
5629 |
if (accop > 0) { |
5630 |
tcg_gen_add_i64(tcg_res[pass], tcg_res[pass], tcg_passres); |
5631 |
tcg_temp_free_i64(tcg_passres); |
5632 |
} else if (accop < 0) { |
5633 |
tcg_gen_sub_i64(tcg_res[pass], tcg_res[pass], tcg_passres); |
5634 |
tcg_temp_free_i64(tcg_passres); |
5635 |
} |
5636 |
|
5637 |
tcg_temp_free_i64(tcg_op1); |
5638 |
tcg_temp_free_i64(tcg_op2); |
5639 |
} |
5640 |
} else {
|
5641 |
/* size 0 or 1, generally helper functions */
|
5642 |
for (pass = 0; pass < 2; pass++) { |
5643 |
TCGv_i32 tcg_op1 = tcg_temp_new_i32(); |
5644 |
TCGv_i32 tcg_op2 = tcg_temp_new_i32(); |
5645 |
TCGv_i64 tcg_passres; |
5646 |
int elt = pass + is_q * 2; |
5647 |
|
5648 |
read_vec_element_i32(s, tcg_op1, rn, elt, MO_32); |
5649 |
read_vec_element_i32(s, tcg_op2, rm, elt, MO_32); |
5650 |
|
5651 |
if (accop == 0) { |
5652 |
tcg_passres = tcg_res[pass]; |
5653 |
} else {
|
5654 |
tcg_passres = tcg_temp_new_i64(); |
5655 |
} |
5656 |
|
5657 |
switch (opcode) {
|
5658 |
case 5: /* SABAL, SABAL2, UABAL, UABAL2 */ |
5659 |
case 7: /* SABDL, SABDL2, UABDL, UABDL2 */ |
5660 |
if (size == 0) { |
5661 |
if (is_u) {
|
5662 |
gen_helper_neon_abdl_u16(tcg_passres, tcg_op1, tcg_op2); |
5663 |
} else {
|
5664 |
gen_helper_neon_abdl_s16(tcg_passres, tcg_op1, tcg_op2); |
5665 |
} |
5666 |
} else {
|
5667 |
if (is_u) {
|
5668 |
gen_helper_neon_abdl_u32(tcg_passres, tcg_op1, tcg_op2); |
5669 |
} else {
|
5670 |
gen_helper_neon_abdl_s32(tcg_passres, tcg_op1, tcg_op2); |
5671 |
} |
5672 |
} |
5673 |
break;
|
5674 |
case 8: /* SMLAL, SMLAL2, UMLAL, UMLAL2 */ |
5675 |
case 10: /* SMLSL, SMLSL2, UMLSL, UMLSL2 */ |
5676 |
case 12: /* UMULL, UMULL2, SMULL, SMULL2 */ |
5677 |
if (size == 0) { |
5678 |
if (is_u) {
|
5679 |
gen_helper_neon_mull_u8(tcg_passres, tcg_op1, tcg_op2); |
5680 |
} else {
|
5681 |
gen_helper_neon_mull_s8(tcg_passres, tcg_op1, tcg_op2); |
5682 |
} |
5683 |
} else {
|
5684 |
if (is_u) {
|
5685 |
gen_helper_neon_mull_u16(tcg_passres, tcg_op1, tcg_op2); |
5686 |
} else {
|
5687 |
gen_helper_neon_mull_s16(tcg_passres, tcg_op1, tcg_op2); |
5688 |
} |
5689 |
} |
5690 |
break;
|
5691 |
default:
|
5692 |
g_assert_not_reached(); |
5693 |
} |
5694 |
tcg_temp_free_i32(tcg_op1); |
5695 |
tcg_temp_free_i32(tcg_op2); |
5696 |
|
5697 |
if (accop > 0) { |
5698 |
if (size == 0) { |
5699 |
gen_helper_neon_addl_u16(tcg_res[pass], tcg_res[pass], |
5700 |
tcg_passres); |
5701 |
} else {
|
5702 |
gen_helper_neon_addl_u32(tcg_res[pass], tcg_res[pass], |
5703 |
tcg_passres); |
5704 |
} |
5705 |
tcg_temp_free_i64(tcg_passres); |
5706 |
} else if (accop < 0) { |
5707 |
if (size == 0) { |
5708 |
gen_helper_neon_subl_u16(tcg_res[pass], tcg_res[pass], |
5709 |
tcg_passres); |
5710 |
} else {
|
5711 |
gen_helper_neon_subl_u32(tcg_res[pass], tcg_res[pass], |
5712 |
tcg_passres); |
5713 |
} |
5714 |
tcg_temp_free_i64(tcg_passres); |
5715 |
} |
5716 |
} |
5717 |
} |
5718 |
|
5719 |
write_vec_element(s, tcg_res[0], rd, 0, MO_64); |
5720 |
write_vec_element(s, tcg_res[1], rd, 1, MO_64); |
5721 |
tcg_temp_free_i64(tcg_res[0]);
|
5722 |
tcg_temp_free_i64(tcg_res[1]);
|
5723 |
} |
5724 |
|
5725 |
/* C3.6.15 AdvSIMD three different
|
5726 |
* 31 30 29 28 24 23 22 21 20 16 15 12 11 10 9 5 4 0
|
5727 |
* +---+---+---+-----------+------+---+------+--------+-----+------+------+
|
5728 |
* | 0 | Q | U | 0 1 1 1 0 | size | 1 | Rm | opcode | 0 0 | Rn | Rd |
|
5729 |
* +---+---+---+-----------+------+---+------+--------+-----+------+------+
|
5730 |
*/
|
5731 |
static void disas_simd_three_reg_diff(DisasContext *s, uint32_t insn) |
5732 |
{ |
5733 |
/* Instructions in this group fall into three basic classes
|
5734 |
* (in each case with the operation working on each element in
|
5735 |
* the input vectors):
|
5736 |
* (1) widening 64 x 64 -> 128 (with possibly Vd as an extra
|
5737 |
* 128 bit input)
|
5738 |
* (2) wide 64 x 128 -> 128
|
5739 |
* (3) narrowing 128 x 128 -> 64
|
5740 |
* Here we do initial decode, catch unallocated cases and
|
5741 |
* dispatch to separate functions for each class.
|
5742 |
*/
|
5743 |
int is_q = extract32(insn, 30, 1); |
5744 |
int is_u = extract32(insn, 29, 1); |
5745 |
int size = extract32(insn, 22, 2); |
5746 |
int opcode = extract32(insn, 12, 4); |
5747 |
int rm = extract32(insn, 16, 5); |
5748 |
int rn = extract32(insn, 5, 5); |
5749 |
int rd = extract32(insn, 0, 5); |
5750 |
|
5751 |
switch (opcode) {
|
5752 |
case 1: /* SADDW, SADDW2, UADDW, UADDW2 */ |
5753 |
case 3: /* SSUBW, SSUBW2, USUBW, USUBW2 */ |
5754 |
/* 64 x 128 -> 128 */
|
5755 |
unsupported_encoding(s, insn); |
5756 |
break;
|
5757 |
case 4: /* ADDHN, ADDHN2, RADDHN, RADDHN2 */ |
5758 |
case 6: /* SUBHN, SUBHN2, RSUBHN, RSUBHN2 */ |
5759 |
/* 128 x 128 -> 64 */
|
5760 |
unsupported_encoding(s, insn); |
5761 |
break;
|
5762 |
case 9: |
5763 |
case 11: |
5764 |
case 13: |
5765 |
case 14: |
5766 |
if (is_u) {
|
5767 |
unallocated_encoding(s); |
5768 |
return;
|
5769 |
} |
5770 |
/* fall through */
|
5771 |
case 0: |
5772 |
case 2: |
5773 |
unsupported_encoding(s, insn); |
5774 |
break;
|
5775 |
case 5: |
5776 |
case 7: |
5777 |
case 8: |
5778 |
case 10: |
5779 |
case 12: |
5780 |
/* 64 x 64 -> 128 */
|
5781 |
if (size == 3) { |
5782 |
unallocated_encoding(s); |
5783 |
return;
|
5784 |
} |
5785 |
handle_3rd_widening(s, is_q, is_u, size, opcode, rd, rn, rm); |
5786 |
break;
|
5787 |
default:
|
5788 |
/* opcode 15 not allocated */
|
5789 |
unallocated_encoding(s); |
5790 |
break;
|
5791 |
} |
5792 |
} |
5793 |
|
5794 |
/* C3.6.16 AdvSIMD three same
|
5795 |
* 31 30 29 28 24 23 22 21 20 16 15 11 10 9 5 4 0
|
5796 |
* +---+---+---+-----------+------+---+------+--------+---+------+------+
|
5797 |
* | 0 | Q | U | 0 1 1 1 0 | size | 1 | Rm | opcode | 1 | Rn | Rd |
|
5798 |
* +---+---+---+-----------+------+---+------+--------+---+------+------+
|
5799 |
*/
|
5800 |
static void disas_simd_three_reg_same(DisasContext *s, uint32_t insn) |
5801 |
{ |
5802 |
unsupported_encoding(s, insn); |
5803 |
} |
5804 |
|
5805 |
/* C3.6.17 AdvSIMD two reg misc
|
5806 |
* 31 30 29 28 24 23 22 21 17 16 12 11 10 9 5 4 0
|
5807 |
* +---+---+---+-----------+------+-----------+--------+-----+------+------+
|
5808 |
* | 0 | Q | U | 0 1 1 1 0 | size | 1 0 0 0 0 | opcode | 1 0 | Rn | Rd |
|
5809 |
* +---+---+---+-----------+------+-----------+--------+-----+------+------+
|
5810 |
*/
|
5811 |
static void disas_simd_two_reg_misc(DisasContext *s, uint32_t insn) |
5812 |
{ |
5813 |
unsupported_encoding(s, insn); |
5814 |
} |
5815 |
|
5816 |
/* C3.6.18 AdvSIMD vector x indexed element
|
5817 |
* 31 30 29 28 24 23 22 21 20 19 16 15 12 11 10 9 5 4 0
|
5818 |
* +---+---+---+-----------+------+---+---+------+-----+---+---+------+------+
|
5819 |
* | 0 | Q | U | 0 1 1 1 1 | size | L | M | Rm | opc | H | 0 | Rn | Rd |
|
5820 |
* +---+---+---+-----------+------+---+---+------+-----+---+---+------+------+
|
5821 |
*/
|
5822 |
static void disas_simd_indexed_vector(DisasContext *s, uint32_t insn) |
5823 |
{ |
5824 |
unsupported_encoding(s, insn); |
5825 |
} |
5826 |
|
5827 |
/* C3.6.19 Crypto AES
|
5828 |
* 31 24 23 22 21 17 16 12 11 10 9 5 4 0
|
5829 |
* +-----------------+------+-----------+--------+-----+------+------+
|
5830 |
* | 0 1 0 0 1 1 1 0 | size | 1 0 1 0 0 | opcode | 1 0 | Rn | Rd |
|
5831 |
* +-----------------+------+-----------+--------+-----+------+------+
|
5832 |
*/
|
5833 |
static void disas_crypto_aes(DisasContext *s, uint32_t insn) |
5834 |
{ |
5835 |
unsupported_encoding(s, insn); |
5836 |
} |
5837 |
|
5838 |
/* C3.6.20 Crypto three-reg SHA
|
5839 |
* 31 24 23 22 21 20 16 15 14 12 11 10 9 5 4 0
|
5840 |
* +-----------------+------+---+------+---+--------+-----+------+------+
|
5841 |
* | 0 1 0 1 1 1 1 0 | size | 0 | Rm | 0 | opcode | 0 0 | Rn | Rd |
|
5842 |
* +-----------------+------+---+------+---+--------+-----+------+------+
|
5843 |
*/
|
5844 |
static void disas_crypto_three_reg_sha(DisasContext *s, uint32_t insn) |
5845 |
{ |
5846 |
unsupported_encoding(s, insn); |
5847 |
} |
5848 |
|
5849 |
/* C3.6.21 Crypto two-reg SHA
|
5850 |
* 31 24 23 22 21 17 16 12 11 10 9 5 4 0
|
5851 |
* +-----------------+------+-----------+--------+-----+------+------+
|
5852 |
* | 0 1 0 1 1 1 1 0 | size | 1 0 1 0 0 | opcode | 1 0 | Rn | Rd |
|
5853 |
* +-----------------+------+-----------+--------+-----+------+------+
|
5854 |
*/
|
5855 |
static void disas_crypto_two_reg_sha(DisasContext *s, uint32_t insn) |
5856 |
{ |
5857 |
unsupported_encoding(s, insn); |
5858 |
} |
5859 |
|
5860 |
/* C3.6 Data processing - SIMD, inc Crypto
|
5861 |
*
|
5862 |
* As the decode gets a little complex we are using a table based
|
5863 |
* approach for this part of the decode.
|
5864 |
*/
|
5865 |
static const AArch64DecodeTable data_proc_simd[] = { |
5866 |
/* pattern , mask , fn */
|
5867 |
{ 0x0e200400, 0x9f200400, disas_simd_three_reg_same }, |
5868 |
{ 0x0e200000, 0x9f200c00, disas_simd_three_reg_diff }, |
5869 |
{ 0x0e200800, 0x9f3e0c00, disas_simd_two_reg_misc }, |
5870 |
{ 0x0e300800, 0x9f3e0c00, disas_simd_across_lanes }, |
5871 |
{ 0x0e000400, 0x9fe08400, disas_simd_copy }, |
5872 |
{ 0x0f000000, 0x9f000400, disas_simd_indexed_vector }, |
5873 |
/* simd_mod_imm decode is a subset of simd_shift_imm, so must precede it */
|
5874 |
{ 0x0f000400, 0x9ff80400, disas_simd_mod_imm }, |
5875 |
{ 0x0f000400, 0x9f800400, disas_simd_shift_imm }, |
5876 |
{ 0x0e000000, 0xbf208c00, disas_simd_tb }, |
5877 |
{ 0x0e000800, 0xbf208c00, disas_simd_zip_trn }, |
5878 |
{ 0x2e000000, 0xbf208400, disas_simd_ext }, |
5879 |
{ 0x5e200400, 0xdf200400, disas_simd_scalar_three_reg_same }, |
5880 |
{ 0x5e200000, 0xdf200c00, disas_simd_scalar_three_reg_diff }, |
5881 |
{ 0x5e200800, 0xdf3e0c00, disas_simd_scalar_two_reg_misc }, |
5882 |
{ 0x5e300800, 0xdf3e0c00, disas_simd_scalar_pairwise }, |
5883 |
{ 0x5e000400, 0xdfe08400, disas_simd_scalar_copy }, |
5884 |
{ 0x5f000000, 0xdf000400, disas_simd_scalar_indexed }, |
5885 |
{ 0x5f000400, 0xdf800400, disas_simd_scalar_shift_imm }, |
5886 |
{ 0x4e280800, 0xff3e0c00, disas_crypto_aes }, |
5887 |
{ 0x5e000000, 0xff208c00, disas_crypto_three_reg_sha }, |
5888 |
{ 0x5e280800, 0xff3e0c00, disas_crypto_two_reg_sha }, |
5889 |
{ 0x00000000, 0x00000000, NULL } |
5890 |
}; |
5891 |
|
5892 |
static void disas_data_proc_simd(DisasContext *s, uint32_t insn) |
5893 |
{ |
5894 |
/* Note that this is called with all non-FP cases from
|
5895 |
* table C3-6 so it must UNDEF for entries not specifically
|
5896 |
* allocated to instructions in that table.
|
5897 |
*/
|
5898 |
AArch64DecodeFn *fn = lookup_disas_fn(&data_proc_simd[0], insn);
|
5899 |
if (fn) {
|
5900 |
fn(s, insn); |
5901 |
} else {
|
5902 |
unallocated_encoding(s); |
5903 |
} |
5904 |
} |
5905 |
|
5906 |
/* C3.6 Data processing - SIMD and floating point */
|
5907 |
static void disas_data_proc_simd_fp(DisasContext *s, uint32_t insn) |
5908 |
{ |
5909 |
if (extract32(insn, 28, 1) == 1 && extract32(insn, 30, 1) == 0) { |
5910 |
disas_data_proc_fp(s, insn); |
5911 |
} else {
|
5912 |
/* SIMD, including crypto */
|
5913 |
disas_data_proc_simd(s, insn); |
5914 |
} |
5915 |
} |
5916 |
|
5917 |
/* C3.1 A64 instruction index by encoding */
|
5918 |
static void disas_a64_insn(CPUARMState *env, DisasContext *s) |
5919 |
{ |
5920 |
uint32_t insn; |
5921 |
|
5922 |
insn = arm_ldl_code(env, s->pc, s->bswap_code); |
5923 |
s->insn = insn; |
5924 |
s->pc += 4;
|
5925 |
|
5926 |
switch (extract32(insn, 25, 4)) { |
5927 |
case 0x0: case 0x1: case 0x2: case 0x3: /* UNALLOCATED */ |
5928 |
unallocated_encoding(s); |
5929 |
break;
|
5930 |
case 0x8: case 0x9: /* Data processing - immediate */ |
5931 |
disas_data_proc_imm(s, insn); |
5932 |
break;
|
5933 |
case 0xa: case 0xb: /* Branch, exception generation and system insns */ |
5934 |
disas_b_exc_sys(s, insn); |
5935 |
break;
|
5936 |
case 0x4: |
5937 |
case 0x6: |
5938 |
case 0xc: |
5939 |
case 0xe: /* Loads and stores */ |
5940 |
disas_ldst(s, insn); |
5941 |
break;
|
5942 |
case 0x5: |
5943 |
case 0xd: /* Data processing - register */ |
5944 |
disas_data_proc_reg(s, insn); |
5945 |
break;
|
5946 |
case 0x7: |
5947 |
case 0xf: /* Data processing - SIMD and floating point */ |
5948 |
disas_data_proc_simd_fp(s, insn); |
5949 |
break;
|
5950 |
default:
|
5951 |
assert(FALSE); /* all 15 cases should be handled above */
|
5952 |
break;
|
5953 |
} |
5954 |
|
5955 |
/* if we allocated any temporaries, free them here */
|
5956 |
free_tmp_a64(s); |
5957 |
} |
5958 |
|
5959 |
void gen_intermediate_code_internal_a64(ARMCPU *cpu,
|
5960 |
TranslationBlock *tb, |
5961 |
bool search_pc)
|
5962 |
{ |
5963 |
CPUState *cs = CPU(cpu); |
5964 |
CPUARMState *env = &cpu->env; |
5965 |
DisasContext dc1, *dc = &dc1; |
5966 |
CPUBreakpoint *bp; |
5967 |
uint16_t *gen_opc_end; |
5968 |
int j, lj;
|
5969 |
target_ulong pc_start; |
5970 |
target_ulong next_page_start; |
5971 |
int num_insns;
|
5972 |
int max_insns;
|
5973 |
|
5974 |
pc_start = tb->pc; |
5975 |
|
5976 |
dc->tb = tb; |
5977 |
|
5978 |
gen_opc_end = tcg_ctx.gen_opc_buf + OPC_MAX_SIZE; |
5979 |
|
5980 |
dc->is_jmp = DISAS_NEXT; |
5981 |
dc->pc = pc_start; |
5982 |
dc->singlestep_enabled = cs->singlestep_enabled; |
5983 |
dc->condjmp = 0;
|
5984 |
|
5985 |
dc->aarch64 = 1;
|
5986 |
dc->thumb = 0;
|
5987 |
dc->bswap_code = 0;
|
5988 |
dc->condexec_mask = 0;
|
5989 |
dc->condexec_cond = 0;
|
5990 |
#if !defined(CONFIG_USER_ONLY)
|
5991 |
dc->user = 0;
|
5992 |
#endif
|
5993 |
dc->vfp_enabled = 0;
|
5994 |
dc->vec_len = 0;
|
5995 |
dc->vec_stride = 0;
|
5996 |
dc->cp_regs = cpu->cp_regs; |
5997 |
dc->current_pl = arm_current_pl(env); |
5998 |
|
5999 |
init_tmp_a64_array(dc); |
6000 |
|
6001 |
next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE; |
6002 |
lj = -1;
|
6003 |
num_insns = 0;
|
6004 |
max_insns = tb->cflags & CF_COUNT_MASK; |
6005 |
if (max_insns == 0) { |
6006 |
max_insns = CF_COUNT_MASK; |
6007 |
} |
6008 |
|
6009 |
gen_tb_start(); |
6010 |
|
6011 |
tcg_clear_temp_count(); |
6012 |
|
6013 |
do {
|
6014 |
if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
|
6015 |
QTAILQ_FOREACH(bp, &env->breakpoints, entry) { |
6016 |
if (bp->pc == dc->pc) {
|
6017 |
gen_exception_insn(dc, 0, EXCP_DEBUG);
|
6018 |
/* Advance PC so that clearing the breakpoint will
|
6019 |
invalidate this TB. */
|
6020 |
dc->pc += 2;
|
6021 |
goto done_generating;
|
6022 |
} |
6023 |
} |
6024 |
} |
6025 |
|
6026 |
if (search_pc) {
|
6027 |
j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf; |
6028 |
if (lj < j) {
|
6029 |
lj++; |
6030 |
while (lj < j) {
|
6031 |
tcg_ctx.gen_opc_instr_start[lj++] = 0;
|
6032 |
} |
6033 |
} |
6034 |
tcg_ctx.gen_opc_pc[lj] = dc->pc; |
6035 |
tcg_ctx.gen_opc_instr_start[lj] = 1;
|
6036 |
tcg_ctx.gen_opc_icount[lj] = num_insns; |
6037 |
} |
6038 |
|
6039 |
if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO)) { |
6040 |
gen_io_start(); |
6041 |
} |
6042 |
|
6043 |
if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP | CPU_LOG_TB_OP_OPT))) {
|
6044 |
tcg_gen_debug_insn_start(dc->pc); |
6045 |
} |
6046 |
|
6047 |
disas_a64_insn(env, dc); |
6048 |
|
6049 |
if (tcg_check_temp_count()) {
|
6050 |
fprintf(stderr, "TCG temporary leak before "TARGET_FMT_lx"\n", |
6051 |
dc->pc); |
6052 |
} |
6053 |
|
6054 |
/* Translation stops when a conditional branch is encountered.
|
6055 |
* Otherwise the subsequent code could get translated several times.
|
6056 |
* Also stop translation when a page boundary is reached. This
|
6057 |
* ensures prefetch aborts occur at the right place.
|
6058 |
*/
|
6059 |
num_insns++; |
6060 |
} while (!dc->is_jmp && tcg_ctx.gen_opc_ptr < gen_opc_end &&
|
6061 |
!cs->singlestep_enabled && |
6062 |
!singlestep && |
6063 |
dc->pc < next_page_start && |
6064 |
num_insns < max_insns); |
6065 |
|
6066 |
if (tb->cflags & CF_LAST_IO) {
|
6067 |
gen_io_end(); |
6068 |
} |
6069 |
|
6070 |
if (unlikely(cs->singlestep_enabled) && dc->is_jmp != DISAS_EXC) {
|
6071 |
/* Note that this means single stepping WFI doesn't halt the CPU.
|
6072 |
* For conditional branch insns this is harmless unreachable code as
|
6073 |
* gen_goto_tb() has already handled emitting the debug exception
|
6074 |
* (and thus a tb-jump is not possible when singlestepping).
|
6075 |
*/
|
6076 |
assert(dc->is_jmp != DISAS_TB_JUMP); |
6077 |
if (dc->is_jmp != DISAS_JUMP) {
|
6078 |
gen_a64_set_pc_im(dc->pc); |
6079 |
} |
6080 |
gen_exception(EXCP_DEBUG); |
6081 |
} else {
|
6082 |
switch (dc->is_jmp) {
|
6083 |
case DISAS_NEXT:
|
6084 |
gen_goto_tb(dc, 1, dc->pc);
|
6085 |
break;
|
6086 |
default:
|
6087 |
case DISAS_UPDATE:
|
6088 |
gen_a64_set_pc_im(dc->pc); |
6089 |
/* fall through */
|
6090 |
case DISAS_JUMP:
|
6091 |
/* indicate that the hash table must be used to find the next TB */
|
6092 |
tcg_gen_exit_tb(0);
|
6093 |
break;
|
6094 |
case DISAS_TB_JUMP:
|
6095 |
case DISAS_EXC:
|
6096 |
case DISAS_SWI:
|
6097 |
break;
|
6098 |
case DISAS_WFI:
|
6099 |
/* This is a special case because we don't want to just halt the CPU
|
6100 |
* if trying to debug across a WFI.
|
6101 |
*/
|
6102 |
gen_helper_wfi(cpu_env); |
6103 |
break;
|
6104 |
} |
6105 |
} |
6106 |
|
6107 |
done_generating:
|
6108 |
gen_tb_end(tb, num_insns); |
6109 |
*tcg_ctx.gen_opc_ptr = INDEX_op_end; |
6110 |
|
6111 |
#ifdef DEBUG_DISAS
|
6112 |
if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
|
6113 |
qemu_log("----------------\n");
|
6114 |
qemu_log("IN: %s\n", lookup_symbol(pc_start));
|
6115 |
log_target_disas(env, pc_start, dc->pc - pc_start, |
6116 |
dc->thumb | (dc->bswap_code << 1));
|
6117 |
qemu_log("\n");
|
6118 |
} |
6119 |
#endif
|
6120 |
if (search_pc) {
|
6121 |
j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf; |
6122 |
lj++; |
6123 |
while (lj <= j) {
|
6124 |
tcg_ctx.gen_opc_instr_start[lj++] = 0;
|
6125 |
} |
6126 |
} else {
|
6127 |
tb->size = dc->pc - pc_start; |
6128 |
tb->icount = num_insns; |
6129 |
} |
6130 |
} |