root / target-arm / translate.c @ 1497c961
History | View | Annotate | Download (260.2 kB)
| 1 |
/*
|
|---|---|
| 2 |
* ARM translation
|
| 3 |
*
|
| 4 |
* Copyright (c) 2003 Fabrice Bellard
|
| 5 |
* Copyright (c) 2005-2007 CodeSourcery
|
| 6 |
* Copyright (c) 2007 OpenedHand, Ltd.
|
| 7 |
*
|
| 8 |
* This library is free software; you can redistribute it and/or
|
| 9 |
* modify it under the terms of the GNU Lesser General Public
|
| 10 |
* License as published by the Free Software Foundation; either
|
| 11 |
* version 2 of the License, or (at your option) any later version.
|
| 12 |
*
|
| 13 |
* This library is distributed in the hope that it will be useful,
|
| 14 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 15 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 16 |
* Lesser General Public License for more details.
|
| 17 |
*
|
| 18 |
* You should have received a copy of the GNU Lesser General Public
|
| 19 |
* License along with this library; if not, write to the Free Software
|
| 20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
| 21 |
*/
|
| 22 |
#include <stdarg.h> |
| 23 |
#include <stdlib.h> |
| 24 |
#include <stdio.h> |
| 25 |
#include <string.h> |
| 26 |
#include <inttypes.h> |
| 27 |
|
| 28 |
#include "cpu.h" |
| 29 |
#include "exec-all.h" |
| 30 |
#include "disas.h" |
| 31 |
#include "tcg-op.h" |
| 32 |
|
| 33 |
#define GEN_HELPER 1 |
| 34 |
#include "helpers.h" |
| 35 |
|
| 36 |
#define ENABLE_ARCH_5J 0 |
| 37 |
#define ENABLE_ARCH_6 arm_feature(env, ARM_FEATURE_V6)
|
| 38 |
#define ENABLE_ARCH_6K arm_feature(env, ARM_FEATURE_V6K)
|
| 39 |
#define ENABLE_ARCH_6T2 arm_feature(env, ARM_FEATURE_THUMB2)
|
| 40 |
#define ENABLE_ARCH_7 arm_feature(env, ARM_FEATURE_V7)
|
| 41 |
|
| 42 |
#define ARCH(x) if (!ENABLE_ARCH_##x) goto illegal_op; |
| 43 |
|
| 44 |
/* internal defines */
|
| 45 |
typedef struct DisasContext { |
| 46 |
target_ulong pc; |
| 47 |
int is_jmp;
|
| 48 |
/* Nonzero if this instruction has been conditionally skipped. */
|
| 49 |
int condjmp;
|
| 50 |
/* The label that will be jumped to when the instruction is skipped. */
|
| 51 |
int condlabel;
|
| 52 |
/* Thumb-2 condtional execution bits. */
|
| 53 |
int condexec_mask;
|
| 54 |
int condexec_cond;
|
| 55 |
struct TranslationBlock *tb;
|
| 56 |
int singlestep_enabled;
|
| 57 |
int thumb;
|
| 58 |
int is_mem;
|
| 59 |
#if !defined(CONFIG_USER_ONLY)
|
| 60 |
int user;
|
| 61 |
#endif
|
| 62 |
} DisasContext; |
| 63 |
|
| 64 |
#if defined(CONFIG_USER_ONLY)
|
| 65 |
#define IS_USER(s) 1 |
| 66 |
#else
|
| 67 |
#define IS_USER(s) (s->user)
|
| 68 |
#endif
|
| 69 |
|
| 70 |
/* These instructions trap after executing, so defer them until after the
|
| 71 |
conditional executions state has been updated. */
|
| 72 |
#define DISAS_WFI 4 |
| 73 |
#define DISAS_SWI 5 |
| 74 |
|
| 75 |
/* XXX: move that elsewhere */
|
| 76 |
extern FILE *logfile;
|
| 77 |
extern int loglevel; |
| 78 |
|
| 79 |
static TCGv cpu_env;
|
| 80 |
/* FIXME: These should be removed. */
|
| 81 |
static TCGv cpu_T[3]; |
| 82 |
|
| 83 |
/* initialize TCG globals. */
|
| 84 |
void arm_translate_init(void) |
| 85 |
{
|
| 86 |
cpu_env = tcg_global_reg_new(TCG_TYPE_PTR, TCG_AREG0, "env");
|
| 87 |
|
| 88 |
cpu_T[0] = tcg_global_reg_new(TCG_TYPE_I32, TCG_AREG1, "T0"); |
| 89 |
cpu_T[1] = tcg_global_reg_new(TCG_TYPE_I32, TCG_AREG2, "T1"); |
| 90 |
cpu_T[2] = tcg_global_reg_new(TCG_TYPE_I32, TCG_AREG3, "T2"); |
| 91 |
} |
| 92 |
|
| 93 |
/* The code generator doesn't like lots of temporaries, so maintain our own
|
| 94 |
cache for reuse within a function. */
|
| 95 |
#define MAX_TEMPS 8 |
| 96 |
static int num_temps; |
| 97 |
static TCGv temps[MAX_TEMPS];
|
| 98 |
|
| 99 |
/* Allocate a temporary variable. */
|
| 100 |
static TCGv new_tmp(void) |
| 101 |
{
|
| 102 |
TCGv tmp; |
| 103 |
if (num_temps == MAX_TEMPS)
|
| 104 |
abort(); |
| 105 |
|
| 106 |
if (GET_TCGV(temps[num_temps]))
|
| 107 |
return temps[num_temps++];
|
| 108 |
|
| 109 |
tmp = tcg_temp_new(TCG_TYPE_I32); |
| 110 |
temps[num_temps++] = tmp; |
| 111 |
return tmp;
|
| 112 |
} |
| 113 |
|
| 114 |
/* Release a temporary variable. */
|
| 115 |
static void dead_tmp(TCGv tmp) |
| 116 |
{
|
| 117 |
int i;
|
| 118 |
num_temps--; |
| 119 |
i = num_temps; |
| 120 |
if (GET_TCGV(temps[i]) == GET_TCGV(tmp))
|
| 121 |
return;
|
| 122 |
|
| 123 |
/* Shuffle this temp to the last slot. */
|
| 124 |
while (GET_TCGV(temps[i]) != GET_TCGV(tmp))
|
| 125 |
i--; |
| 126 |
while (i < num_temps) {
|
| 127 |
temps[i] = temps[i + 1];
|
| 128 |
i++; |
| 129 |
} |
| 130 |
temps[i] = tmp; |
| 131 |
} |
| 132 |
|
| 133 |
/* Set a variable to the value of a CPU register. */
|
| 134 |
static void load_reg_var(DisasContext *s, TCGv var, int reg) |
| 135 |
{
|
| 136 |
if (reg == 15) { |
| 137 |
uint32_t addr; |
| 138 |
/* normaly, since we updated PC, we need only to add one insn */
|
| 139 |
if (s->thumb)
|
| 140 |
addr = (long)s->pc + 2; |
| 141 |
else
|
| 142 |
addr = (long)s->pc + 4; |
| 143 |
tcg_gen_movi_i32(var, addr); |
| 144 |
} else {
|
| 145 |
tcg_gen_ld_i32(var, cpu_env, offsetof(CPUState, regs[reg])); |
| 146 |
} |
| 147 |
} |
| 148 |
|
| 149 |
/* Create a new temporary and set it to the value of a CPU register. */
|
| 150 |
static inline TCGv load_reg(DisasContext *s, int reg) |
| 151 |
{
|
| 152 |
TCGv tmp = new_tmp(); |
| 153 |
load_reg_var(s, tmp, reg); |
| 154 |
return tmp;
|
| 155 |
} |
| 156 |
|
| 157 |
/* Set a CPU register. The source must be a temporary and will be
|
| 158 |
marked as dead. */
|
| 159 |
static void store_reg(DisasContext *s, int reg, TCGv var) |
| 160 |
{
|
| 161 |
if (reg == 15) { |
| 162 |
tcg_gen_andi_i32(var, var, ~1);
|
| 163 |
s->is_jmp = DISAS_JUMP; |
| 164 |
} |
| 165 |
tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, regs[reg])); |
| 166 |
dead_tmp(var); |
| 167 |
} |
| 168 |
|
| 169 |
|
| 170 |
/* Basic operations. */
|
| 171 |
#define gen_op_movl_T0_T1() tcg_gen_mov_i32(cpu_T[0], cpu_T[1]) |
| 172 |
#define gen_op_movl_T0_T2() tcg_gen_mov_i32(cpu_T[0], cpu_T[2]) |
| 173 |
#define gen_op_movl_T1_T0() tcg_gen_mov_i32(cpu_T[1], cpu_T[0]) |
| 174 |
#define gen_op_movl_T1_T2() tcg_gen_mov_i32(cpu_T[1], cpu_T[2]) |
| 175 |
#define gen_op_movl_T2_T0() tcg_gen_mov_i32(cpu_T[2], cpu_T[0]) |
| 176 |
#define gen_op_movl_T0_im(im) tcg_gen_movi_i32(cpu_T[0], im) |
| 177 |
#define gen_op_movl_T1_im(im) tcg_gen_movi_i32(cpu_T[1], im) |
| 178 |
#define gen_op_movl_T2_im(im) tcg_gen_movi_i32(cpu_T[2], im) |
| 179 |
|
| 180 |
#define gen_op_addl_T1_im(im) tcg_gen_addi_i32(cpu_T[1], cpu_T[1], im) |
| 181 |
#define gen_op_addl_T0_T1() tcg_gen_add_i32(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 182 |
#define gen_op_subl_T0_T1() tcg_gen_sub_i32(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 183 |
#define gen_op_rsbl_T0_T1() tcg_gen_sub_i32(cpu_T[0], cpu_T[1], cpu_T[0]) |
| 184 |
|
| 185 |
#define gen_op_andl_T0_T1() tcg_gen_and_i32(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 186 |
#define gen_op_xorl_T0_T1() tcg_gen_xor_i32(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 187 |
#define gen_op_orl_T0_T1() tcg_gen_or_i32(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 188 |
#define gen_op_notl_T0() tcg_gen_not_i32(cpu_T[0], cpu_T[0]) |
| 189 |
#define gen_op_notl_T1() tcg_gen_not_i32(cpu_T[1], cpu_T[1]) |
| 190 |
#define gen_op_logic_T0_cc() gen_logic_CC(cpu_T[0]); |
| 191 |
#define gen_op_logic_T1_cc() gen_logic_CC(cpu_T[1]); |
| 192 |
|
| 193 |
#define gen_op_shll_T0_im(im) tcg_gen_shli_i32(cpu_T[0], cpu_T[0], im) |
| 194 |
#define gen_op_shll_T1_im(im) tcg_gen_shli_i32(cpu_T[1], cpu_T[1], im) |
| 195 |
#define gen_op_shrl_T1_im(im) tcg_gen_shri_i32(cpu_T[1], cpu_T[1], im) |
| 196 |
#define gen_op_sarl_T1_im(im) tcg_gen_sari_i32(cpu_T[1], cpu_T[1], im) |
| 197 |
#define gen_op_rorl_T1_im(im) tcg_gen_rori_i32(cpu_T[1], cpu_T[1], im) |
| 198 |
|
| 199 |
/* Value extensions. */
|
| 200 |
#define gen_uxtb(var) tcg_gen_andi_i32(var, var, 0xff) |
| 201 |
#define gen_uxth(var) tcg_gen_andi_i32(var, var, 0xffff) |
| 202 |
#define gen_sxtb(var) tcg_gen_ext8s_i32(var, var)
|
| 203 |
#define gen_sxth(var) tcg_gen_ext16s_i32(var, var)
|
| 204 |
|
| 205 |
#define gen_sxtb16(var) gen_helper_sxtb16(var, var)
|
| 206 |
#define gen_uxtb16(var) gen_helper_uxtb16(var, var)
|
| 207 |
|
| 208 |
#define gen_op_addl_T0_T1_setq() \
|
| 209 |
gen_helper_add_setq(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 210 |
#define gen_op_addl_T0_T1_saturate() \
|
| 211 |
gen_helper_add_saturate(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 212 |
#define gen_op_subl_T0_T1_saturate() \
|
| 213 |
gen_helper_sub_saturate(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 214 |
#define gen_op_addl_T0_T1_usaturate() \
|
| 215 |
gen_helper_add_usaturate(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 216 |
#define gen_op_subl_T0_T1_usaturate() \
|
| 217 |
gen_helper_sub_usaturate(cpu_T[0], cpu_T[0], cpu_T[1]) |
| 218 |
|
| 219 |
/* Dual 16-bit add. Result placed in t0 and t1 is marked as dead.
|
| 220 |
tmp = (t0 ^ t1) & 0x8000;
|
| 221 |
t0 &= ~0x8000;
|
| 222 |
t1 &= ~0x8000;
|
| 223 |
t0 = (t0 + t1) ^ tmp;
|
| 224 |
*/
|
| 225 |
|
| 226 |
static void gen_add16(TCGv t0, TCGv t1) |
| 227 |
{
|
| 228 |
TCGv tmp = new_tmp(); |
| 229 |
tcg_gen_xor_i32(tmp, t0, t1); |
| 230 |
tcg_gen_andi_i32(tmp, tmp, 0x8000);
|
| 231 |
tcg_gen_andi_i32(t0, t0, ~0x8000);
|
| 232 |
tcg_gen_andi_i32(t1, t1, ~0x8000);
|
| 233 |
tcg_gen_add_i32(t0, t0, t1); |
| 234 |
tcg_gen_xor_i32(t0, t0, tmp); |
| 235 |
dead_tmp(tmp); |
| 236 |
dead_tmp(t1); |
| 237 |
} |
| 238 |
|
| 239 |
#define gen_set_CF(var) tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, CF))
|
| 240 |
|
| 241 |
/* Set CF to the top bit of var. */
|
| 242 |
static void gen_set_CF_bit31(TCGv var) |
| 243 |
{
|
| 244 |
TCGv tmp = new_tmp(); |
| 245 |
tcg_gen_shri_i32(tmp, var, 31);
|
| 246 |
gen_set_CF(var); |
| 247 |
dead_tmp(tmp); |
| 248 |
} |
| 249 |
|
| 250 |
/* Set N and Z flags from var. */
|
| 251 |
static inline void gen_logic_CC(TCGv var) |
| 252 |
{
|
| 253 |
tcg_gen_st_i32(var, cpu_env, offsetof(CPUState, NZF)); |
| 254 |
} |
| 255 |
|
| 256 |
/* T0 += T1 + CF. */
|
| 257 |
static void gen_adc_T0_T1(void) |
| 258 |
{
|
| 259 |
TCGv tmp = new_tmp(); |
| 260 |
gen_op_addl_T0_T1(); |
| 261 |
tcg_gen_ld_i32(tmp, cpu_env, offsetof(CPUState, CF)); |
| 262 |
tcg_gen_add_i32(cpu_T[0], cpu_T[0], tmp); |
| 263 |
dead_tmp(tmp); |
| 264 |
} |
| 265 |
|
| 266 |
/* FIXME: Implement this natively. */
|
| 267 |
static inline void tcg_gen_not_i32(TCGv t0, TCGv t1) |
| 268 |
{
|
| 269 |
tcg_gen_xori_i32(t0, t1, ~0);
|
| 270 |
} |
| 271 |
|
| 272 |
/* T0 &= ~T1. Clobbers T1. */
|
| 273 |
/* FIXME: Implement bic natively. */
|
| 274 |
static inline void gen_op_bicl_T0_T1(void) |
| 275 |
{
|
| 276 |
gen_op_notl_T1(); |
| 277 |
gen_op_andl_T0_T1(); |
| 278 |
} |
| 279 |
|
| 280 |
/* FIXME: Implement this natively. */
|
| 281 |
static void tcg_gen_rori_i32(TCGv t0, TCGv t1, int i) |
| 282 |
{
|
| 283 |
TCGv tmp; |
| 284 |
|
| 285 |
if (i == 0) |
| 286 |
return;
|
| 287 |
|
| 288 |
tmp = new_tmp(); |
| 289 |
tcg_gen_shri_i32(tmp, t1, i); |
| 290 |
tcg_gen_shli_i32(t1, t1, 32 - i);
|
| 291 |
tcg_gen_or_i32(t0, t1, tmp); |
| 292 |
dead_tmp(tmp); |
| 293 |
} |
| 294 |
|
| 295 |
static void shifter_out_im(TCGv var, int shift) |
| 296 |
{
|
| 297 |
TCGv tmp = new_tmp(); |
| 298 |
if (shift == 0) { |
| 299 |
tcg_gen_andi_i32(tmp, var, 1);
|
| 300 |
} else {
|
| 301 |
tcg_gen_shri_i32(tmp, var, shift); |
| 302 |
if (shift != 31); |
| 303 |
tcg_gen_andi_i32(tmp, tmp, 1);
|
| 304 |
} |
| 305 |
gen_set_CF(tmp); |
| 306 |
dead_tmp(tmp); |
| 307 |
} |
| 308 |
|
| 309 |
/* Shift by immediate. Includes special handling for shift == 0. */
|
| 310 |
static inline void gen_arm_shift_im(TCGv var, int shiftop, int shift, int flags) |
| 311 |
{
|
| 312 |
switch (shiftop) {
|
| 313 |
case 0: /* LSL */ |
| 314 |
if (shift != 0) { |
| 315 |
if (flags)
|
| 316 |
shifter_out_im(var, 32 - shift);
|
| 317 |
tcg_gen_shli_i32(var, var, shift); |
| 318 |
} |
| 319 |
break;
|
| 320 |
case 1: /* LSR */ |
| 321 |
if (shift == 0) { |
| 322 |
if (flags) {
|
| 323 |
tcg_gen_shri_i32(var, var, 31);
|
| 324 |
gen_set_CF(var); |
| 325 |
} |
| 326 |
tcg_gen_movi_i32(var, 0);
|
| 327 |
} else {
|
| 328 |
if (flags)
|
| 329 |
shifter_out_im(var, shift - 1);
|
| 330 |
tcg_gen_shri_i32(var, var, shift); |
| 331 |
} |
| 332 |
break;
|
| 333 |
case 2: /* ASR */ |
| 334 |
if (shift == 0) |
| 335 |
shift = 32;
|
| 336 |
if (flags)
|
| 337 |
shifter_out_im(var, shift - 1);
|
| 338 |
if (shift == 32) |
| 339 |
shift = 31;
|
| 340 |
tcg_gen_sari_i32(var, var, shift); |
| 341 |
break;
|
| 342 |
case 3: /* ROR/RRX */ |
| 343 |
if (shift != 0) { |
| 344 |
if (flags)
|
| 345 |
shifter_out_im(var, shift - 1);
|
| 346 |
tcg_gen_rori_i32(var, var, shift); break;
|
| 347 |
} else {
|
| 348 |
TCGv tmp = new_tmp(); |
| 349 |
tcg_gen_ld_i32(tmp, cpu_env, offsetof(CPUState, CF)); |
| 350 |
if (flags)
|
| 351 |
shifter_out_im(var, 0);
|
| 352 |
tcg_gen_shri_i32(var, var, 1);
|
| 353 |
tcg_gen_shli_i32(tmp, tmp, 31);
|
| 354 |
tcg_gen_or_i32(var, var, tmp); |
| 355 |
dead_tmp(tmp); |
| 356 |
} |
| 357 |
} |
| 358 |
}; |
| 359 |
|
| 360 |
#define PAS_OP(pfx) { \
|
| 361 |
gen_op_ ## pfx ## add16_T0_T1, \ |
| 362 |
gen_op_ ## pfx ## addsubx_T0_T1, \ |
| 363 |
gen_op_ ## pfx ## subaddx_T0_T1, \ |
| 364 |
gen_op_ ## pfx ## sub16_T0_T1, \ |
| 365 |
gen_op_ ## pfx ## add8_T0_T1, \ |
| 366 |
NULL, \
|
| 367 |
NULL, \
|
| 368 |
gen_op_ ## pfx ## sub8_T0_T1 } |
| 369 |
|
| 370 |
static GenOpFunc *gen_arm_parallel_addsub[8][8] = { |
| 371 |
{},
|
| 372 |
PAS_OP(s), |
| 373 |
PAS_OP(q), |
| 374 |
PAS_OP(sh), |
| 375 |
{},
|
| 376 |
PAS_OP(u), |
| 377 |
PAS_OP(uq), |
| 378 |
PAS_OP(uh), |
| 379 |
}; |
| 380 |
#undef PAS_OP
|
| 381 |
|
| 382 |
/* For unknown reasons Arm and Thumb-2 use arbitrarily diffenet encodings. */
|
| 383 |
#define PAS_OP(pfx) { \
|
| 384 |
gen_op_ ## pfx ## add8_T0_T1, \ |
| 385 |
gen_op_ ## pfx ## add16_T0_T1, \ |
| 386 |
gen_op_ ## pfx ## addsubx_T0_T1, \ |
| 387 |
NULL, \
|
| 388 |
gen_op_ ## pfx ## sub8_T0_T1, \ |
| 389 |
gen_op_ ## pfx ## sub16_T0_T1, \ |
| 390 |
gen_op_ ## pfx ## subaddx_T0_T1, \ |
| 391 |
NULL }
|
| 392 |
|
| 393 |
static GenOpFunc *gen_thumb2_parallel_addsub[8][8] = { |
| 394 |
PAS_OP(s), |
| 395 |
PAS_OP(q), |
| 396 |
PAS_OP(sh), |
| 397 |
{},
|
| 398 |
PAS_OP(u), |
| 399 |
PAS_OP(uq), |
| 400 |
PAS_OP(uh), |
| 401 |
{}
|
| 402 |
}; |
| 403 |
#undef PAS_OP
|
| 404 |
|
| 405 |
static GenOpFunc1 *gen_test_cc[14] = { |
| 406 |
gen_op_test_eq, |
| 407 |
gen_op_test_ne, |
| 408 |
gen_op_test_cs, |
| 409 |
gen_op_test_cc, |
| 410 |
gen_op_test_mi, |
| 411 |
gen_op_test_pl, |
| 412 |
gen_op_test_vs, |
| 413 |
gen_op_test_vc, |
| 414 |
gen_op_test_hi, |
| 415 |
gen_op_test_ls, |
| 416 |
gen_op_test_ge, |
| 417 |
gen_op_test_lt, |
| 418 |
gen_op_test_gt, |
| 419 |
gen_op_test_le, |
| 420 |
}; |
| 421 |
|
| 422 |
const uint8_t table_logic_cc[16] = { |
| 423 |
1, /* and */ |
| 424 |
1, /* xor */ |
| 425 |
0, /* sub */ |
| 426 |
0, /* rsb */ |
| 427 |
0, /* add */ |
| 428 |
0, /* adc */ |
| 429 |
0, /* sbc */ |
| 430 |
0, /* rsc */ |
| 431 |
1, /* andl */ |
| 432 |
1, /* xorl */ |
| 433 |
0, /* cmp */ |
| 434 |
0, /* cmn */ |
| 435 |
1, /* orr */ |
| 436 |
1, /* mov */ |
| 437 |
1, /* bic */ |
| 438 |
1, /* mvn */ |
| 439 |
}; |
| 440 |
|
| 441 |
static GenOpFunc *gen_shift_T1_T0[4] = { |
| 442 |
gen_op_shll_T1_T0, |
| 443 |
gen_op_shrl_T1_T0, |
| 444 |
gen_op_sarl_T1_T0, |
| 445 |
gen_op_rorl_T1_T0, |
| 446 |
}; |
| 447 |
|
| 448 |
static GenOpFunc *gen_shift_T1_T0_cc[4] = { |
| 449 |
gen_op_shll_T1_T0_cc, |
| 450 |
gen_op_shrl_T1_T0_cc, |
| 451 |
gen_op_sarl_T1_T0_cc, |
| 452 |
gen_op_rorl_T1_T0_cc, |
| 453 |
}; |
| 454 |
|
| 455 |
/* Set PC and thumb state from T0. Clobbers T0. */
|
| 456 |
static inline void gen_bx(DisasContext *s) |
| 457 |
{
|
| 458 |
TCGv tmp; |
| 459 |
|
| 460 |
s->is_jmp = DISAS_UPDATE; |
| 461 |
tmp = new_tmp(); |
| 462 |
tcg_gen_andi_i32(tmp, cpu_T[0], 1); |
| 463 |
tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, thumb)); |
| 464 |
dead_tmp(tmp); |
| 465 |
tcg_gen_andi_i32(cpu_T[0], cpu_T[0], ~1); |
| 466 |
tcg_gen_st_i32(cpu_T[0], cpu_env, offsetof(CPUState, regs[15])); |
| 467 |
} |
| 468 |
|
| 469 |
#if defined(CONFIG_USER_ONLY)
|
| 470 |
#define gen_ldst(name, s) gen_op_##name##_raw() |
| 471 |
#else
|
| 472 |
#define gen_ldst(name, s) do { \ |
| 473 |
s->is_mem = 1; \
|
| 474 |
if (IS_USER(s)) \
|
| 475 |
gen_op_##name##_user(); \ |
| 476 |
else \
|
| 477 |
gen_op_##name##_kernel(); \ |
| 478 |
} while (0) |
| 479 |
#endif
|
| 480 |
|
| 481 |
static inline void gen_movl_T0_reg(DisasContext *s, int reg) |
| 482 |
{
|
| 483 |
load_reg_var(s, cpu_T[0], reg);
|
| 484 |
} |
| 485 |
|
| 486 |
static inline void gen_movl_T1_reg(DisasContext *s, int reg) |
| 487 |
{
|
| 488 |
load_reg_var(s, cpu_T[1], reg);
|
| 489 |
} |
| 490 |
|
| 491 |
static inline void gen_movl_T2_reg(DisasContext *s, int reg) |
| 492 |
{
|
| 493 |
load_reg_var(s, cpu_T[2], reg);
|
| 494 |
} |
| 495 |
|
| 496 |
static inline void gen_set_pc_T0(void) |
| 497 |
{
|
| 498 |
tcg_gen_st_i32(cpu_T[0], cpu_env, offsetof(CPUState, regs[15])); |
| 499 |
} |
| 500 |
|
| 501 |
static inline void gen_movl_reg_TN(DisasContext *s, int reg, int t) |
| 502 |
{
|
| 503 |
TCGv tmp; |
| 504 |
if (reg == 15) { |
| 505 |
tmp = new_tmp(); |
| 506 |
tcg_gen_andi_i32(tmp, cpu_T[t], ~1);
|
| 507 |
} else {
|
| 508 |
tmp = cpu_T[t]; |
| 509 |
} |
| 510 |
tcg_gen_st_i32(tmp, cpu_env, offsetof(CPUState, regs[reg])); |
| 511 |
if (reg == 15) { |
| 512 |
dead_tmp(tmp); |
| 513 |
s->is_jmp = DISAS_JUMP; |
| 514 |
} |
| 515 |
} |
| 516 |
|
| 517 |
static inline void gen_movl_reg_T0(DisasContext *s, int reg) |
| 518 |
{
|
| 519 |
gen_movl_reg_TN(s, reg, 0);
|
| 520 |
} |
| 521 |
|
| 522 |
static inline void gen_movl_reg_T1(DisasContext *s, int reg) |
| 523 |
{
|
| 524 |
gen_movl_reg_TN(s, reg, 1);
|
| 525 |
} |
| 526 |
|
| 527 |
/* Force a TB lookup after an instruction that changes the CPU state. */
|
| 528 |
static inline void gen_lookup_tb(DisasContext *s) |
| 529 |
{
|
| 530 |
gen_op_movl_T0_im(s->pc); |
| 531 |
gen_movl_reg_T0(s, 15);
|
| 532 |
s->is_jmp = DISAS_UPDATE; |
| 533 |
} |
| 534 |
|
| 535 |
static inline void gen_add_data_offset(DisasContext *s, unsigned int insn) |
| 536 |
{
|
| 537 |
int val, rm, shift, shiftop;
|
| 538 |
TCGv offset; |
| 539 |
|
| 540 |
if (!(insn & (1 << 25))) { |
| 541 |
/* immediate */
|
| 542 |
val = insn & 0xfff;
|
| 543 |
if (!(insn & (1 << 23))) |
| 544 |
val = -val; |
| 545 |
if (val != 0) |
| 546 |
gen_op_addl_T1_im(val); |
| 547 |
} else {
|
| 548 |
/* shift/register */
|
| 549 |
rm = (insn) & 0xf;
|
| 550 |
shift = (insn >> 7) & 0x1f; |
| 551 |
shiftop = (insn >> 5) & 3; |
| 552 |
offset = load_reg(s, rm); |
| 553 |
gen_arm_shift_im(offset, shiftop, shift, 0);
|
| 554 |
if (!(insn & (1 << 23))) |
| 555 |
tcg_gen_sub_i32(cpu_T[1], cpu_T[1], offset); |
| 556 |
else
|
| 557 |
tcg_gen_add_i32(cpu_T[1], cpu_T[1], offset); |
| 558 |
dead_tmp(offset); |
| 559 |
} |
| 560 |
} |
| 561 |
|
| 562 |
static inline void gen_add_datah_offset(DisasContext *s, unsigned int insn, |
| 563 |
int extra)
|
| 564 |
{
|
| 565 |
int val, rm;
|
| 566 |
TCGv offset; |
| 567 |
|
| 568 |
if (insn & (1 << 22)) { |
| 569 |
/* immediate */
|
| 570 |
val = (insn & 0xf) | ((insn >> 4) & 0xf0); |
| 571 |
if (!(insn & (1 << 23))) |
| 572 |
val = -val; |
| 573 |
val += extra; |
| 574 |
if (val != 0) |
| 575 |
gen_op_addl_T1_im(val); |
| 576 |
} else {
|
| 577 |
/* register */
|
| 578 |
if (extra)
|
| 579 |
gen_op_addl_T1_im(extra); |
| 580 |
rm = (insn) & 0xf;
|
| 581 |
offset = load_reg(s, rm); |
| 582 |
if (!(insn & (1 << 23))) |
| 583 |
tcg_gen_sub_i32(cpu_T[1], cpu_T[1], offset); |
| 584 |
else
|
| 585 |
tcg_gen_add_i32(cpu_T[1], cpu_T[1], offset); |
| 586 |
dead_tmp(offset); |
| 587 |
} |
| 588 |
} |
| 589 |
|
| 590 |
#define VFP_OP(name) \
|
| 591 |
static inline void gen_vfp_##name(int dp) \ |
| 592 |
{ \
|
| 593 |
if (dp) \
|
| 594 |
gen_op_vfp_##name##d(); \ |
| 595 |
else \
|
| 596 |
gen_op_vfp_##name##s(); \ |
| 597 |
} |
| 598 |
|
| 599 |
#define VFP_OP1(name) \
|
| 600 |
static inline void gen_vfp_##name(int dp, int arg) \ |
| 601 |
{ \
|
| 602 |
if (dp) \
|
| 603 |
gen_op_vfp_##name##d(arg); \ |
| 604 |
else \
|
| 605 |
gen_op_vfp_##name##s(arg); \ |
| 606 |
} |
| 607 |
|
| 608 |
VFP_OP(add) |
| 609 |
VFP_OP(sub) |
| 610 |
VFP_OP(mul) |
| 611 |
VFP_OP(div) |
| 612 |
VFP_OP(neg) |
| 613 |
VFP_OP(abs) |
| 614 |
VFP_OP(sqrt) |
| 615 |
VFP_OP(cmp) |
| 616 |
VFP_OP(cmpe) |
| 617 |
VFP_OP(F1_ld0) |
| 618 |
VFP_OP(uito) |
| 619 |
VFP_OP(sito) |
| 620 |
VFP_OP(toui) |
| 621 |
VFP_OP(touiz) |
| 622 |
VFP_OP(tosi) |
| 623 |
VFP_OP(tosiz) |
| 624 |
VFP_OP1(tosh) |
| 625 |
VFP_OP1(tosl) |
| 626 |
VFP_OP1(touh) |
| 627 |
VFP_OP1(toul) |
| 628 |
VFP_OP1(shto) |
| 629 |
VFP_OP1(slto) |
| 630 |
VFP_OP1(uhto) |
| 631 |
VFP_OP1(ulto) |
| 632 |
|
| 633 |
#undef VFP_OP
|
| 634 |
|
| 635 |
static inline void gen_vfp_fconst(int dp, uint32_t val) |
| 636 |
{
|
| 637 |
if (dp)
|
| 638 |
gen_op_vfp_fconstd(val); |
| 639 |
else
|
| 640 |
gen_op_vfp_fconsts(val); |
| 641 |
} |
| 642 |
|
| 643 |
static inline void gen_vfp_ld(DisasContext *s, int dp) |
| 644 |
{
|
| 645 |
if (dp)
|
| 646 |
gen_ldst(vfp_ldd, s); |
| 647 |
else
|
| 648 |
gen_ldst(vfp_lds, s); |
| 649 |
} |
| 650 |
|
| 651 |
static inline void gen_vfp_st(DisasContext *s, int dp) |
| 652 |
{
|
| 653 |
if (dp)
|
| 654 |
gen_ldst(vfp_std, s); |
| 655 |
else
|
| 656 |
gen_ldst(vfp_sts, s); |
| 657 |
} |
| 658 |
|
| 659 |
static inline long |
| 660 |
vfp_reg_offset (int dp, int reg) |
| 661 |
{
|
| 662 |
if (dp)
|
| 663 |
return offsetof(CPUARMState, vfp.regs[reg]);
|
| 664 |
else if (reg & 1) { |
| 665 |
return offsetof(CPUARMState, vfp.regs[reg >> 1]) |
| 666 |
+ offsetof(CPU_DoubleU, l.upper); |
| 667 |
} else {
|
| 668 |
return offsetof(CPUARMState, vfp.regs[reg >> 1]) |
| 669 |
+ offsetof(CPU_DoubleU, l.lower); |
| 670 |
} |
| 671 |
} |
| 672 |
|
| 673 |
/* Return the offset of a 32-bit piece of a NEON register.
|
| 674 |
zero is the least significant end of the register. */
|
| 675 |
static inline long |
| 676 |
neon_reg_offset (int reg, int n) |
| 677 |
{
|
| 678 |
int sreg;
|
| 679 |
sreg = reg * 2 + n;
|
| 680 |
return vfp_reg_offset(0, sreg); |
| 681 |
} |
| 682 |
|
| 683 |
#define NEON_GET_REG(T, reg, n) gen_op_neon_getreg_##T(neon_reg_offset(reg, n)) |
| 684 |
#define NEON_SET_REG(T, reg, n) gen_op_neon_setreg_##T(neon_reg_offset(reg, n)) |
| 685 |
|
| 686 |
static inline void gen_mov_F0_vreg(int dp, int reg) |
| 687 |
{
|
| 688 |
if (dp)
|
| 689 |
gen_op_vfp_getreg_F0d(vfp_reg_offset(dp, reg)); |
| 690 |
else
|
| 691 |
gen_op_vfp_getreg_F0s(vfp_reg_offset(dp, reg)); |
| 692 |
} |
| 693 |
|
| 694 |
static inline void gen_mov_F1_vreg(int dp, int reg) |
| 695 |
{
|
| 696 |
if (dp)
|
| 697 |
gen_op_vfp_getreg_F1d(vfp_reg_offset(dp, reg)); |
| 698 |
else
|
| 699 |
gen_op_vfp_getreg_F1s(vfp_reg_offset(dp, reg)); |
| 700 |
} |
| 701 |
|
| 702 |
static inline void gen_mov_vreg_F0(int dp, int reg) |
| 703 |
{
|
| 704 |
if (dp)
|
| 705 |
gen_op_vfp_setreg_F0d(vfp_reg_offset(dp, reg)); |
| 706 |
else
|
| 707 |
gen_op_vfp_setreg_F0s(vfp_reg_offset(dp, reg)); |
| 708 |
} |
| 709 |
|
| 710 |
#define ARM_CP_RW_BIT (1 << 20) |
| 711 |
|
| 712 |
static inline int gen_iwmmxt_address(DisasContext *s, uint32_t insn) |
| 713 |
{
|
| 714 |
int rd;
|
| 715 |
uint32_t offset; |
| 716 |
|
| 717 |
rd = (insn >> 16) & 0xf; |
| 718 |
gen_movl_T1_reg(s, rd); |
| 719 |
|
| 720 |
offset = (insn & 0xff) << ((insn >> 7) & 2); |
| 721 |
if (insn & (1 << 24)) { |
| 722 |
/* Pre indexed */
|
| 723 |
if (insn & (1 << 23)) |
| 724 |
gen_op_addl_T1_im(offset); |
| 725 |
else
|
| 726 |
gen_op_addl_T1_im(-offset); |
| 727 |
|
| 728 |
if (insn & (1 << 21)) |
| 729 |
gen_movl_reg_T1(s, rd); |
| 730 |
} else if (insn & (1 << 21)) { |
| 731 |
/* Post indexed */
|
| 732 |
if (insn & (1 << 23)) |
| 733 |
gen_op_movl_T0_im(offset); |
| 734 |
else
|
| 735 |
gen_op_movl_T0_im(- offset); |
| 736 |
gen_op_addl_T0_T1(); |
| 737 |
gen_movl_reg_T0(s, rd); |
| 738 |
} else if (!(insn & (1 << 23))) |
| 739 |
return 1; |
| 740 |
return 0; |
| 741 |
} |
| 742 |
|
| 743 |
static inline int gen_iwmmxt_shift(uint32_t insn, uint32_t mask) |
| 744 |
{
|
| 745 |
int rd = (insn >> 0) & 0xf; |
| 746 |
|
| 747 |
if (insn & (1 << 8)) |
| 748 |
if (rd < ARM_IWMMXT_wCGR0 || rd > ARM_IWMMXT_wCGR3)
|
| 749 |
return 1; |
| 750 |
else
|
| 751 |
gen_op_iwmmxt_movl_T0_wCx(rd); |
| 752 |
else
|
| 753 |
gen_op_iwmmxt_movl_T0_T1_wRn(rd); |
| 754 |
|
| 755 |
gen_op_movl_T1_im(mask); |
| 756 |
gen_op_andl_T0_T1(); |
| 757 |
return 0; |
| 758 |
} |
| 759 |
|
| 760 |
/* Disassemble an iwMMXt instruction. Returns nonzero if an error occured
|
| 761 |
(ie. an undefined instruction). */
|
| 762 |
static int disas_iwmmxt_insn(CPUState *env, DisasContext *s, uint32_t insn) |
| 763 |
{
|
| 764 |
int rd, wrd;
|
| 765 |
int rdhi, rdlo, rd0, rd1, i;
|
| 766 |
|
| 767 |
if ((insn & 0x0e000e00) == 0x0c000000) { |
| 768 |
if ((insn & 0x0fe00ff0) == 0x0c400000) { |
| 769 |
wrd = insn & 0xf;
|
| 770 |
rdlo = (insn >> 12) & 0xf; |
| 771 |
rdhi = (insn >> 16) & 0xf; |
| 772 |
if (insn & ARM_CP_RW_BIT) { /* TMRRC */ |
| 773 |
gen_op_iwmmxt_movl_T0_T1_wRn(wrd); |
| 774 |
gen_movl_reg_T0(s, rdlo); |
| 775 |
gen_movl_reg_T1(s, rdhi); |
| 776 |
} else { /* TMCRR */ |
| 777 |
gen_movl_T0_reg(s, rdlo); |
| 778 |
gen_movl_T1_reg(s, rdhi); |
| 779 |
gen_op_iwmmxt_movl_wRn_T0_T1(wrd); |
| 780 |
gen_op_iwmmxt_set_mup(); |
| 781 |
} |
| 782 |
return 0; |
| 783 |
} |
| 784 |
|
| 785 |
wrd = (insn >> 12) & 0xf; |
| 786 |
if (gen_iwmmxt_address(s, insn))
|
| 787 |
return 1; |
| 788 |
if (insn & ARM_CP_RW_BIT) {
|
| 789 |
if ((insn >> 28) == 0xf) { /* WLDRW wCx */ |
| 790 |
gen_ldst(ldl, s); |
| 791 |
gen_op_iwmmxt_movl_wCx_T0(wrd); |
| 792 |
} else {
|
| 793 |
if (insn & (1 << 8)) |
| 794 |
if (insn & (1 << 22)) /* WLDRD */ |
| 795 |
gen_ldst(iwmmxt_ldq, s); |
| 796 |
else /* WLDRW wRd */ |
| 797 |
gen_ldst(iwmmxt_ldl, s); |
| 798 |
else
|
| 799 |
if (insn & (1 << 22)) /* WLDRH */ |
| 800 |
gen_ldst(iwmmxt_ldw, s); |
| 801 |
else /* WLDRB */ |
| 802 |
gen_ldst(iwmmxt_ldb, s); |
| 803 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 804 |
} |
| 805 |
} else {
|
| 806 |
if ((insn >> 28) == 0xf) { /* WSTRW wCx */ |
| 807 |
gen_op_iwmmxt_movl_T0_wCx(wrd); |
| 808 |
gen_ldst(stl, s); |
| 809 |
} else {
|
| 810 |
gen_op_iwmmxt_movq_M0_wRn(wrd); |
| 811 |
if (insn & (1 << 8)) |
| 812 |
if (insn & (1 << 22)) /* WSTRD */ |
| 813 |
gen_ldst(iwmmxt_stq, s); |
| 814 |
else /* WSTRW wRd */ |
| 815 |
gen_ldst(iwmmxt_stl, s); |
| 816 |
else
|
| 817 |
if (insn & (1 << 22)) /* WSTRH */ |
| 818 |
gen_ldst(iwmmxt_ldw, s); |
| 819 |
else /* WSTRB */ |
| 820 |
gen_ldst(iwmmxt_stb, s); |
| 821 |
} |
| 822 |
} |
| 823 |
return 0; |
| 824 |
} |
| 825 |
|
| 826 |
if ((insn & 0x0f000000) != 0x0e000000) |
| 827 |
return 1; |
| 828 |
|
| 829 |
switch (((insn >> 12) & 0xf00) | ((insn >> 4) & 0xff)) { |
| 830 |
case 0x000: /* WOR */ |
| 831 |
wrd = (insn >> 12) & 0xf; |
| 832 |
rd0 = (insn >> 0) & 0xf; |
| 833 |
rd1 = (insn >> 16) & 0xf; |
| 834 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 835 |
gen_op_iwmmxt_orq_M0_wRn(rd1); |
| 836 |
gen_op_iwmmxt_setpsr_nz(); |
| 837 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 838 |
gen_op_iwmmxt_set_mup(); |
| 839 |
gen_op_iwmmxt_set_cup(); |
| 840 |
break;
|
| 841 |
case 0x011: /* TMCR */ |
| 842 |
if (insn & 0xf) |
| 843 |
return 1; |
| 844 |
rd = (insn >> 12) & 0xf; |
| 845 |
wrd = (insn >> 16) & 0xf; |
| 846 |
switch (wrd) {
|
| 847 |
case ARM_IWMMXT_wCID:
|
| 848 |
case ARM_IWMMXT_wCASF:
|
| 849 |
break;
|
| 850 |
case ARM_IWMMXT_wCon:
|
| 851 |
gen_op_iwmmxt_set_cup(); |
| 852 |
/* Fall through. */
|
| 853 |
case ARM_IWMMXT_wCSSF:
|
| 854 |
gen_op_iwmmxt_movl_T0_wCx(wrd); |
| 855 |
gen_movl_T1_reg(s, rd); |
| 856 |
gen_op_bicl_T0_T1(); |
| 857 |
gen_op_iwmmxt_movl_wCx_T0(wrd); |
| 858 |
break;
|
| 859 |
case ARM_IWMMXT_wCGR0:
|
| 860 |
case ARM_IWMMXT_wCGR1:
|
| 861 |
case ARM_IWMMXT_wCGR2:
|
| 862 |
case ARM_IWMMXT_wCGR3:
|
| 863 |
gen_op_iwmmxt_set_cup(); |
| 864 |
gen_movl_reg_T0(s, rd); |
| 865 |
gen_op_iwmmxt_movl_wCx_T0(wrd); |
| 866 |
break;
|
| 867 |
default:
|
| 868 |
return 1; |
| 869 |
} |
| 870 |
break;
|
| 871 |
case 0x100: /* WXOR */ |
| 872 |
wrd = (insn >> 12) & 0xf; |
| 873 |
rd0 = (insn >> 0) & 0xf; |
| 874 |
rd1 = (insn >> 16) & 0xf; |
| 875 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 876 |
gen_op_iwmmxt_xorq_M0_wRn(rd1); |
| 877 |
gen_op_iwmmxt_setpsr_nz(); |
| 878 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 879 |
gen_op_iwmmxt_set_mup(); |
| 880 |
gen_op_iwmmxt_set_cup(); |
| 881 |
break;
|
| 882 |
case 0x111: /* TMRC */ |
| 883 |
if (insn & 0xf) |
| 884 |
return 1; |
| 885 |
rd = (insn >> 12) & 0xf; |
| 886 |
wrd = (insn >> 16) & 0xf; |
| 887 |
gen_op_iwmmxt_movl_T0_wCx(wrd); |
| 888 |
gen_movl_reg_T0(s, rd); |
| 889 |
break;
|
| 890 |
case 0x300: /* WANDN */ |
| 891 |
wrd = (insn >> 12) & 0xf; |
| 892 |
rd0 = (insn >> 0) & 0xf; |
| 893 |
rd1 = (insn >> 16) & 0xf; |
| 894 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 895 |
gen_op_iwmmxt_negq_M0(); |
| 896 |
gen_op_iwmmxt_andq_M0_wRn(rd1); |
| 897 |
gen_op_iwmmxt_setpsr_nz(); |
| 898 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 899 |
gen_op_iwmmxt_set_mup(); |
| 900 |
gen_op_iwmmxt_set_cup(); |
| 901 |
break;
|
| 902 |
case 0x200: /* WAND */ |
| 903 |
wrd = (insn >> 12) & 0xf; |
| 904 |
rd0 = (insn >> 0) & 0xf; |
| 905 |
rd1 = (insn >> 16) & 0xf; |
| 906 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 907 |
gen_op_iwmmxt_andq_M0_wRn(rd1); |
| 908 |
gen_op_iwmmxt_setpsr_nz(); |
| 909 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 910 |
gen_op_iwmmxt_set_mup(); |
| 911 |
gen_op_iwmmxt_set_cup(); |
| 912 |
break;
|
| 913 |
case 0x810: case 0xa10: /* WMADD */ |
| 914 |
wrd = (insn >> 12) & 0xf; |
| 915 |
rd0 = (insn >> 0) & 0xf; |
| 916 |
rd1 = (insn >> 16) & 0xf; |
| 917 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 918 |
if (insn & (1 << 21)) |
| 919 |
gen_op_iwmmxt_maddsq_M0_wRn(rd1); |
| 920 |
else
|
| 921 |
gen_op_iwmmxt_madduq_M0_wRn(rd1); |
| 922 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 923 |
gen_op_iwmmxt_set_mup(); |
| 924 |
break;
|
| 925 |
case 0x10e: case 0x50e: case 0x90e: case 0xd0e: /* WUNPCKIL */ |
| 926 |
wrd = (insn >> 12) & 0xf; |
| 927 |
rd0 = (insn >> 16) & 0xf; |
| 928 |
rd1 = (insn >> 0) & 0xf; |
| 929 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 930 |
switch ((insn >> 22) & 3) { |
| 931 |
case 0: |
| 932 |
gen_op_iwmmxt_unpacklb_M0_wRn(rd1); |
| 933 |
break;
|
| 934 |
case 1: |
| 935 |
gen_op_iwmmxt_unpacklw_M0_wRn(rd1); |
| 936 |
break;
|
| 937 |
case 2: |
| 938 |
gen_op_iwmmxt_unpackll_M0_wRn(rd1); |
| 939 |
break;
|
| 940 |
case 3: |
| 941 |
return 1; |
| 942 |
} |
| 943 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 944 |
gen_op_iwmmxt_set_mup(); |
| 945 |
gen_op_iwmmxt_set_cup(); |
| 946 |
break;
|
| 947 |
case 0x10c: case 0x50c: case 0x90c: case 0xd0c: /* WUNPCKIH */ |
| 948 |
wrd = (insn >> 12) & 0xf; |
| 949 |
rd0 = (insn >> 16) & 0xf; |
| 950 |
rd1 = (insn >> 0) & 0xf; |
| 951 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 952 |
switch ((insn >> 22) & 3) { |
| 953 |
case 0: |
| 954 |
gen_op_iwmmxt_unpackhb_M0_wRn(rd1); |
| 955 |
break;
|
| 956 |
case 1: |
| 957 |
gen_op_iwmmxt_unpackhw_M0_wRn(rd1); |
| 958 |
break;
|
| 959 |
case 2: |
| 960 |
gen_op_iwmmxt_unpackhl_M0_wRn(rd1); |
| 961 |
break;
|
| 962 |
case 3: |
| 963 |
return 1; |
| 964 |
} |
| 965 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 966 |
gen_op_iwmmxt_set_mup(); |
| 967 |
gen_op_iwmmxt_set_cup(); |
| 968 |
break;
|
| 969 |
case 0x012: case 0x112: case 0x412: case 0x512: /* WSAD */ |
| 970 |
wrd = (insn >> 12) & 0xf; |
| 971 |
rd0 = (insn >> 16) & 0xf; |
| 972 |
rd1 = (insn >> 0) & 0xf; |
| 973 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 974 |
if (insn & (1 << 22)) |
| 975 |
gen_op_iwmmxt_sadw_M0_wRn(rd1); |
| 976 |
else
|
| 977 |
gen_op_iwmmxt_sadb_M0_wRn(rd1); |
| 978 |
if (!(insn & (1 << 20))) |
| 979 |
gen_op_iwmmxt_addl_M0_wRn(wrd); |
| 980 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 981 |
gen_op_iwmmxt_set_mup(); |
| 982 |
break;
|
| 983 |
case 0x010: case 0x110: case 0x210: case 0x310: /* WMUL */ |
| 984 |
wrd = (insn >> 12) & 0xf; |
| 985 |
rd0 = (insn >> 16) & 0xf; |
| 986 |
rd1 = (insn >> 0) & 0xf; |
| 987 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 988 |
if (insn & (1 << 21)) |
| 989 |
gen_op_iwmmxt_mulsw_M0_wRn(rd1, (insn & (1 << 20)) ? 16 : 0); |
| 990 |
else
|
| 991 |
gen_op_iwmmxt_muluw_M0_wRn(rd1, (insn & (1 << 20)) ? 16 : 0); |
| 992 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 993 |
gen_op_iwmmxt_set_mup(); |
| 994 |
break;
|
| 995 |
case 0x410: case 0x510: case 0x610: case 0x710: /* WMAC */ |
| 996 |
wrd = (insn >> 12) & 0xf; |
| 997 |
rd0 = (insn >> 16) & 0xf; |
| 998 |
rd1 = (insn >> 0) & 0xf; |
| 999 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1000 |
if (insn & (1 << 21)) |
| 1001 |
gen_op_iwmmxt_macsw_M0_wRn(rd1); |
| 1002 |
else
|
| 1003 |
gen_op_iwmmxt_macuw_M0_wRn(rd1); |
| 1004 |
if (!(insn & (1 << 20))) { |
| 1005 |
if (insn & (1 << 21)) |
| 1006 |
gen_op_iwmmxt_addsq_M0_wRn(wrd); |
| 1007 |
else
|
| 1008 |
gen_op_iwmmxt_adduq_M0_wRn(wrd); |
| 1009 |
} |
| 1010 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1011 |
gen_op_iwmmxt_set_mup(); |
| 1012 |
break;
|
| 1013 |
case 0x006: case 0x406: case 0x806: case 0xc06: /* WCMPEQ */ |
| 1014 |
wrd = (insn >> 12) & 0xf; |
| 1015 |
rd0 = (insn >> 16) & 0xf; |
| 1016 |
rd1 = (insn >> 0) & 0xf; |
| 1017 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1018 |
switch ((insn >> 22) & 3) { |
| 1019 |
case 0: |
| 1020 |
gen_op_iwmmxt_cmpeqb_M0_wRn(rd1); |
| 1021 |
break;
|
| 1022 |
case 1: |
| 1023 |
gen_op_iwmmxt_cmpeqw_M0_wRn(rd1); |
| 1024 |
break;
|
| 1025 |
case 2: |
| 1026 |
gen_op_iwmmxt_cmpeql_M0_wRn(rd1); |
| 1027 |
break;
|
| 1028 |
case 3: |
| 1029 |
return 1; |
| 1030 |
} |
| 1031 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1032 |
gen_op_iwmmxt_set_mup(); |
| 1033 |
gen_op_iwmmxt_set_cup(); |
| 1034 |
break;
|
| 1035 |
case 0x800: case 0x900: case 0xc00: case 0xd00: /* WAVG2 */ |
| 1036 |
wrd = (insn >> 12) & 0xf; |
| 1037 |
rd0 = (insn >> 16) & 0xf; |
| 1038 |
rd1 = (insn >> 0) & 0xf; |
| 1039 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1040 |
if (insn & (1 << 22)) |
| 1041 |
gen_op_iwmmxt_avgw_M0_wRn(rd1, (insn >> 20) & 1); |
| 1042 |
else
|
| 1043 |
gen_op_iwmmxt_avgb_M0_wRn(rd1, (insn >> 20) & 1); |
| 1044 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1045 |
gen_op_iwmmxt_set_mup(); |
| 1046 |
gen_op_iwmmxt_set_cup(); |
| 1047 |
break;
|
| 1048 |
case 0x802: case 0x902: case 0xa02: case 0xb02: /* WALIGNR */ |
| 1049 |
wrd = (insn >> 12) & 0xf; |
| 1050 |
rd0 = (insn >> 16) & 0xf; |
| 1051 |
rd1 = (insn >> 0) & 0xf; |
| 1052 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1053 |
gen_op_iwmmxt_movl_T0_wCx(ARM_IWMMXT_wCGR0 + ((insn >> 20) & 3)); |
| 1054 |
gen_op_movl_T1_im(7);
|
| 1055 |
gen_op_andl_T0_T1(); |
| 1056 |
gen_op_iwmmxt_align_M0_T0_wRn(rd1); |
| 1057 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1058 |
gen_op_iwmmxt_set_mup(); |
| 1059 |
break;
|
| 1060 |
case 0x601: case 0x605: case 0x609: case 0x60d: /* TINSR */ |
| 1061 |
rd = (insn >> 12) & 0xf; |
| 1062 |
wrd = (insn >> 16) & 0xf; |
| 1063 |
gen_movl_T0_reg(s, rd); |
| 1064 |
gen_op_iwmmxt_movq_M0_wRn(wrd); |
| 1065 |
switch ((insn >> 6) & 3) { |
| 1066 |
case 0: |
| 1067 |
gen_op_movl_T1_im(0xff);
|
| 1068 |
gen_op_iwmmxt_insr_M0_T0_T1((insn & 7) << 3); |
| 1069 |
break;
|
| 1070 |
case 1: |
| 1071 |
gen_op_movl_T1_im(0xffff);
|
| 1072 |
gen_op_iwmmxt_insr_M0_T0_T1((insn & 3) << 4); |
| 1073 |
break;
|
| 1074 |
case 2: |
| 1075 |
gen_op_movl_T1_im(0xffffffff);
|
| 1076 |
gen_op_iwmmxt_insr_M0_T0_T1((insn & 1) << 5); |
| 1077 |
break;
|
| 1078 |
case 3: |
| 1079 |
return 1; |
| 1080 |
} |
| 1081 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1082 |
gen_op_iwmmxt_set_mup(); |
| 1083 |
break;
|
| 1084 |
case 0x107: case 0x507: case 0x907: case 0xd07: /* TEXTRM */ |
| 1085 |
rd = (insn >> 12) & 0xf; |
| 1086 |
wrd = (insn >> 16) & 0xf; |
| 1087 |
if (rd == 15) |
| 1088 |
return 1; |
| 1089 |
gen_op_iwmmxt_movq_M0_wRn(wrd); |
| 1090 |
switch ((insn >> 22) & 3) { |
| 1091 |
case 0: |
| 1092 |
if (insn & 8) |
| 1093 |
gen_op_iwmmxt_extrsb_T0_M0((insn & 7) << 3); |
| 1094 |
else {
|
| 1095 |
gen_op_movl_T1_im(0xff);
|
| 1096 |
gen_op_iwmmxt_extru_T0_M0_T1((insn & 7) << 3); |
| 1097 |
} |
| 1098 |
break;
|
| 1099 |
case 1: |
| 1100 |
if (insn & 8) |
| 1101 |
gen_op_iwmmxt_extrsw_T0_M0((insn & 3) << 4); |
| 1102 |
else {
|
| 1103 |
gen_op_movl_T1_im(0xffff);
|
| 1104 |
gen_op_iwmmxt_extru_T0_M0_T1((insn & 3) << 4); |
| 1105 |
} |
| 1106 |
break;
|
| 1107 |
case 2: |
| 1108 |
gen_op_movl_T1_im(0xffffffff);
|
| 1109 |
gen_op_iwmmxt_extru_T0_M0_T1((insn & 1) << 5); |
| 1110 |
break;
|
| 1111 |
case 3: |
| 1112 |
return 1; |
| 1113 |
} |
| 1114 |
gen_movl_reg_T0(s, rd); |
| 1115 |
break;
|
| 1116 |
case 0x117: case 0x517: case 0x917: case 0xd17: /* TEXTRC */ |
| 1117 |
if ((insn & 0x000ff008) != 0x0003f000) |
| 1118 |
return 1; |
| 1119 |
gen_op_iwmmxt_movl_T1_wCx(ARM_IWMMXT_wCASF); |
| 1120 |
switch ((insn >> 22) & 3) { |
| 1121 |
case 0: |
| 1122 |
gen_op_shrl_T1_im(((insn & 7) << 2) + 0); |
| 1123 |
break;
|
| 1124 |
case 1: |
| 1125 |
gen_op_shrl_T1_im(((insn & 3) << 3) + 4); |
| 1126 |
break;
|
| 1127 |
case 2: |
| 1128 |
gen_op_shrl_T1_im(((insn & 1) << 4) + 12); |
| 1129 |
break;
|
| 1130 |
case 3: |
| 1131 |
return 1; |
| 1132 |
} |
| 1133 |
gen_op_shll_T1_im(28);
|
| 1134 |
gen_op_movl_T0_T1(); |
| 1135 |
gen_op_movl_cpsr_T0(0xf0000000);
|
| 1136 |
break;
|
| 1137 |
case 0x401: case 0x405: case 0x409: case 0x40d: /* TBCST */ |
| 1138 |
rd = (insn >> 12) & 0xf; |
| 1139 |
wrd = (insn >> 16) & 0xf; |
| 1140 |
gen_movl_T0_reg(s, rd); |
| 1141 |
switch ((insn >> 6) & 3) { |
| 1142 |
case 0: |
| 1143 |
gen_op_iwmmxt_bcstb_M0_T0(); |
| 1144 |
break;
|
| 1145 |
case 1: |
| 1146 |
gen_op_iwmmxt_bcstw_M0_T0(); |
| 1147 |
break;
|
| 1148 |
case 2: |
| 1149 |
gen_op_iwmmxt_bcstl_M0_T0(); |
| 1150 |
break;
|
| 1151 |
case 3: |
| 1152 |
return 1; |
| 1153 |
} |
| 1154 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1155 |
gen_op_iwmmxt_set_mup(); |
| 1156 |
break;
|
| 1157 |
case 0x113: case 0x513: case 0x913: case 0xd13: /* TANDC */ |
| 1158 |
if ((insn & 0x000ff00f) != 0x0003f000) |
| 1159 |
return 1; |
| 1160 |
gen_op_iwmmxt_movl_T1_wCx(ARM_IWMMXT_wCASF); |
| 1161 |
switch ((insn >> 22) & 3) { |
| 1162 |
case 0: |
| 1163 |
for (i = 0; i < 7; i ++) { |
| 1164 |
gen_op_shll_T1_im(4);
|
| 1165 |
gen_op_andl_T0_T1(); |
| 1166 |
} |
| 1167 |
break;
|
| 1168 |
case 1: |
| 1169 |
for (i = 0; i < 3; i ++) { |
| 1170 |
gen_op_shll_T1_im(8);
|
| 1171 |
gen_op_andl_T0_T1(); |
| 1172 |
} |
| 1173 |
break;
|
| 1174 |
case 2: |
| 1175 |
gen_op_shll_T1_im(16);
|
| 1176 |
gen_op_andl_T0_T1(); |
| 1177 |
break;
|
| 1178 |
case 3: |
| 1179 |
return 1; |
| 1180 |
} |
| 1181 |
gen_op_movl_cpsr_T0(0xf0000000);
|
| 1182 |
break;
|
| 1183 |
case 0x01c: case 0x41c: case 0x81c: case 0xc1c: /* WACC */ |
| 1184 |
wrd = (insn >> 12) & 0xf; |
| 1185 |
rd0 = (insn >> 16) & 0xf; |
| 1186 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1187 |
switch ((insn >> 22) & 3) { |
| 1188 |
case 0: |
| 1189 |
gen_op_iwmmxt_addcb_M0(); |
| 1190 |
break;
|
| 1191 |
case 1: |
| 1192 |
gen_op_iwmmxt_addcw_M0(); |
| 1193 |
break;
|
| 1194 |
case 2: |
| 1195 |
gen_op_iwmmxt_addcl_M0(); |
| 1196 |
break;
|
| 1197 |
case 3: |
| 1198 |
return 1; |
| 1199 |
} |
| 1200 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1201 |
gen_op_iwmmxt_set_mup(); |
| 1202 |
break;
|
| 1203 |
case 0x115: case 0x515: case 0x915: case 0xd15: /* TORC */ |
| 1204 |
if ((insn & 0x000ff00f) != 0x0003f000) |
| 1205 |
return 1; |
| 1206 |
gen_op_iwmmxt_movl_T1_wCx(ARM_IWMMXT_wCASF); |
| 1207 |
switch ((insn >> 22) & 3) { |
| 1208 |
case 0: |
| 1209 |
for (i = 0; i < 7; i ++) { |
| 1210 |
gen_op_shll_T1_im(4);
|
| 1211 |
gen_op_orl_T0_T1(); |
| 1212 |
} |
| 1213 |
break;
|
| 1214 |
case 1: |
| 1215 |
for (i = 0; i < 3; i ++) { |
| 1216 |
gen_op_shll_T1_im(8);
|
| 1217 |
gen_op_orl_T0_T1(); |
| 1218 |
} |
| 1219 |
break;
|
| 1220 |
case 2: |
| 1221 |
gen_op_shll_T1_im(16);
|
| 1222 |
gen_op_orl_T0_T1(); |
| 1223 |
break;
|
| 1224 |
case 3: |
| 1225 |
return 1; |
| 1226 |
} |
| 1227 |
gen_op_movl_T1_im(0xf0000000);
|
| 1228 |
gen_op_andl_T0_T1(); |
| 1229 |
gen_op_movl_cpsr_T0(0xf0000000);
|
| 1230 |
break;
|
| 1231 |
case 0x103: case 0x503: case 0x903: case 0xd03: /* TMOVMSK */ |
| 1232 |
rd = (insn >> 12) & 0xf; |
| 1233 |
rd0 = (insn >> 16) & 0xf; |
| 1234 |
if ((insn & 0xf) != 0) |
| 1235 |
return 1; |
| 1236 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1237 |
switch ((insn >> 22) & 3) { |
| 1238 |
case 0: |
| 1239 |
gen_op_iwmmxt_msbb_T0_M0(); |
| 1240 |
break;
|
| 1241 |
case 1: |
| 1242 |
gen_op_iwmmxt_msbw_T0_M0(); |
| 1243 |
break;
|
| 1244 |
case 2: |
| 1245 |
gen_op_iwmmxt_msbl_T0_M0(); |
| 1246 |
break;
|
| 1247 |
case 3: |
| 1248 |
return 1; |
| 1249 |
} |
| 1250 |
gen_movl_reg_T0(s, rd); |
| 1251 |
break;
|
| 1252 |
case 0x106: case 0x306: case 0x506: case 0x706: /* WCMPGT */ |
| 1253 |
case 0x906: case 0xb06: case 0xd06: case 0xf06: |
| 1254 |
wrd = (insn >> 12) & 0xf; |
| 1255 |
rd0 = (insn >> 16) & 0xf; |
| 1256 |
rd1 = (insn >> 0) & 0xf; |
| 1257 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1258 |
switch ((insn >> 22) & 3) { |
| 1259 |
case 0: |
| 1260 |
if (insn & (1 << 21)) |
| 1261 |
gen_op_iwmmxt_cmpgtsb_M0_wRn(rd1); |
| 1262 |
else
|
| 1263 |
gen_op_iwmmxt_cmpgtub_M0_wRn(rd1); |
| 1264 |
break;
|
| 1265 |
case 1: |
| 1266 |
if (insn & (1 << 21)) |
| 1267 |
gen_op_iwmmxt_cmpgtsw_M0_wRn(rd1); |
| 1268 |
else
|
| 1269 |
gen_op_iwmmxt_cmpgtuw_M0_wRn(rd1); |
| 1270 |
break;
|
| 1271 |
case 2: |
| 1272 |
if (insn & (1 << 21)) |
| 1273 |
gen_op_iwmmxt_cmpgtsl_M0_wRn(rd1); |
| 1274 |
else
|
| 1275 |
gen_op_iwmmxt_cmpgtul_M0_wRn(rd1); |
| 1276 |
break;
|
| 1277 |
case 3: |
| 1278 |
return 1; |
| 1279 |
} |
| 1280 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1281 |
gen_op_iwmmxt_set_mup(); |
| 1282 |
gen_op_iwmmxt_set_cup(); |
| 1283 |
break;
|
| 1284 |
case 0x00e: case 0x20e: case 0x40e: case 0x60e: /* WUNPCKEL */ |
| 1285 |
case 0x80e: case 0xa0e: case 0xc0e: case 0xe0e: |
| 1286 |
wrd = (insn >> 12) & 0xf; |
| 1287 |
rd0 = (insn >> 16) & 0xf; |
| 1288 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1289 |
switch ((insn >> 22) & 3) { |
| 1290 |
case 0: |
| 1291 |
if (insn & (1 << 21)) |
| 1292 |
gen_op_iwmmxt_unpacklsb_M0(); |
| 1293 |
else
|
| 1294 |
gen_op_iwmmxt_unpacklub_M0(); |
| 1295 |
break;
|
| 1296 |
case 1: |
| 1297 |
if (insn & (1 << 21)) |
| 1298 |
gen_op_iwmmxt_unpacklsw_M0(); |
| 1299 |
else
|
| 1300 |
gen_op_iwmmxt_unpackluw_M0(); |
| 1301 |
break;
|
| 1302 |
case 2: |
| 1303 |
if (insn & (1 << 21)) |
| 1304 |
gen_op_iwmmxt_unpacklsl_M0(); |
| 1305 |
else
|
| 1306 |
gen_op_iwmmxt_unpacklul_M0(); |
| 1307 |
break;
|
| 1308 |
case 3: |
| 1309 |
return 1; |
| 1310 |
} |
| 1311 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1312 |
gen_op_iwmmxt_set_mup(); |
| 1313 |
gen_op_iwmmxt_set_cup(); |
| 1314 |
break;
|
| 1315 |
case 0x00c: case 0x20c: case 0x40c: case 0x60c: /* WUNPCKEH */ |
| 1316 |
case 0x80c: case 0xa0c: case 0xc0c: case 0xe0c: |
| 1317 |
wrd = (insn >> 12) & 0xf; |
| 1318 |
rd0 = (insn >> 16) & 0xf; |
| 1319 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1320 |
switch ((insn >> 22) & 3) { |
| 1321 |
case 0: |
| 1322 |
if (insn & (1 << 21)) |
| 1323 |
gen_op_iwmmxt_unpackhsb_M0(); |
| 1324 |
else
|
| 1325 |
gen_op_iwmmxt_unpackhub_M0(); |
| 1326 |
break;
|
| 1327 |
case 1: |
| 1328 |
if (insn & (1 << 21)) |
| 1329 |
gen_op_iwmmxt_unpackhsw_M0(); |
| 1330 |
else
|
| 1331 |
gen_op_iwmmxt_unpackhuw_M0(); |
| 1332 |
break;
|
| 1333 |
case 2: |
| 1334 |
if (insn & (1 << 21)) |
| 1335 |
gen_op_iwmmxt_unpackhsl_M0(); |
| 1336 |
else
|
| 1337 |
gen_op_iwmmxt_unpackhul_M0(); |
| 1338 |
break;
|
| 1339 |
case 3: |
| 1340 |
return 1; |
| 1341 |
} |
| 1342 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1343 |
gen_op_iwmmxt_set_mup(); |
| 1344 |
gen_op_iwmmxt_set_cup(); |
| 1345 |
break;
|
| 1346 |
case 0x204: case 0x604: case 0xa04: case 0xe04: /* WSRL */ |
| 1347 |
case 0x214: case 0x614: case 0xa14: case 0xe14: |
| 1348 |
wrd = (insn >> 12) & 0xf; |
| 1349 |
rd0 = (insn >> 16) & 0xf; |
| 1350 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1351 |
if (gen_iwmmxt_shift(insn, 0xff)) |
| 1352 |
return 1; |
| 1353 |
switch ((insn >> 22) & 3) { |
| 1354 |
case 0: |
| 1355 |
return 1; |
| 1356 |
case 1: |
| 1357 |
gen_op_iwmmxt_srlw_M0_T0(); |
| 1358 |
break;
|
| 1359 |
case 2: |
| 1360 |
gen_op_iwmmxt_srll_M0_T0(); |
| 1361 |
break;
|
| 1362 |
case 3: |
| 1363 |
gen_op_iwmmxt_srlq_M0_T0(); |
| 1364 |
break;
|
| 1365 |
} |
| 1366 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1367 |
gen_op_iwmmxt_set_mup(); |
| 1368 |
gen_op_iwmmxt_set_cup(); |
| 1369 |
break;
|
| 1370 |
case 0x004: case 0x404: case 0x804: case 0xc04: /* WSRA */ |
| 1371 |
case 0x014: case 0x414: case 0x814: case 0xc14: |
| 1372 |
wrd = (insn >> 12) & 0xf; |
| 1373 |
rd0 = (insn >> 16) & 0xf; |
| 1374 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1375 |
if (gen_iwmmxt_shift(insn, 0xff)) |
| 1376 |
return 1; |
| 1377 |
switch ((insn >> 22) & 3) { |
| 1378 |
case 0: |
| 1379 |
return 1; |
| 1380 |
case 1: |
| 1381 |
gen_op_iwmmxt_sraw_M0_T0(); |
| 1382 |
break;
|
| 1383 |
case 2: |
| 1384 |
gen_op_iwmmxt_sral_M0_T0(); |
| 1385 |
break;
|
| 1386 |
case 3: |
| 1387 |
gen_op_iwmmxt_sraq_M0_T0(); |
| 1388 |
break;
|
| 1389 |
} |
| 1390 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1391 |
gen_op_iwmmxt_set_mup(); |
| 1392 |
gen_op_iwmmxt_set_cup(); |
| 1393 |
break;
|
| 1394 |
case 0x104: case 0x504: case 0x904: case 0xd04: /* WSLL */ |
| 1395 |
case 0x114: case 0x514: case 0x914: case 0xd14: |
| 1396 |
wrd = (insn >> 12) & 0xf; |
| 1397 |
rd0 = (insn >> 16) & 0xf; |
| 1398 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1399 |
if (gen_iwmmxt_shift(insn, 0xff)) |
| 1400 |
return 1; |
| 1401 |
switch ((insn >> 22) & 3) { |
| 1402 |
case 0: |
| 1403 |
return 1; |
| 1404 |
case 1: |
| 1405 |
gen_op_iwmmxt_sllw_M0_T0(); |
| 1406 |
break;
|
| 1407 |
case 2: |
| 1408 |
gen_op_iwmmxt_slll_M0_T0(); |
| 1409 |
break;
|
| 1410 |
case 3: |
| 1411 |
gen_op_iwmmxt_sllq_M0_T0(); |
| 1412 |
break;
|
| 1413 |
} |
| 1414 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1415 |
gen_op_iwmmxt_set_mup(); |
| 1416 |
gen_op_iwmmxt_set_cup(); |
| 1417 |
break;
|
| 1418 |
case 0x304: case 0x704: case 0xb04: case 0xf04: /* WROR */ |
| 1419 |
case 0x314: case 0x714: case 0xb14: case 0xf14: |
| 1420 |
wrd = (insn >> 12) & 0xf; |
| 1421 |
rd0 = (insn >> 16) & 0xf; |
| 1422 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1423 |
switch ((insn >> 22) & 3) { |
| 1424 |
case 0: |
| 1425 |
return 1; |
| 1426 |
case 1: |
| 1427 |
if (gen_iwmmxt_shift(insn, 0xf)) |
| 1428 |
return 1; |
| 1429 |
gen_op_iwmmxt_rorw_M0_T0(); |
| 1430 |
break;
|
| 1431 |
case 2: |
| 1432 |
if (gen_iwmmxt_shift(insn, 0x1f)) |
| 1433 |
return 1; |
| 1434 |
gen_op_iwmmxt_rorl_M0_T0(); |
| 1435 |
break;
|
| 1436 |
case 3: |
| 1437 |
if (gen_iwmmxt_shift(insn, 0x3f)) |
| 1438 |
return 1; |
| 1439 |
gen_op_iwmmxt_rorq_M0_T0(); |
| 1440 |
break;
|
| 1441 |
} |
| 1442 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1443 |
gen_op_iwmmxt_set_mup(); |
| 1444 |
gen_op_iwmmxt_set_cup(); |
| 1445 |
break;
|
| 1446 |
case 0x116: case 0x316: case 0x516: case 0x716: /* WMIN */ |
| 1447 |
case 0x916: case 0xb16: case 0xd16: case 0xf16: |
| 1448 |
wrd = (insn >> 12) & 0xf; |
| 1449 |
rd0 = (insn >> 16) & 0xf; |
| 1450 |
rd1 = (insn >> 0) & 0xf; |
| 1451 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1452 |
switch ((insn >> 22) & 3) { |
| 1453 |
case 0: |
| 1454 |
if (insn & (1 << 21)) |
| 1455 |
gen_op_iwmmxt_minsb_M0_wRn(rd1); |
| 1456 |
else
|
| 1457 |
gen_op_iwmmxt_minub_M0_wRn(rd1); |
| 1458 |
break;
|
| 1459 |
case 1: |
| 1460 |
if (insn & (1 << 21)) |
| 1461 |
gen_op_iwmmxt_minsw_M0_wRn(rd1); |
| 1462 |
else
|
| 1463 |
gen_op_iwmmxt_minuw_M0_wRn(rd1); |
| 1464 |
break;
|
| 1465 |
case 2: |
| 1466 |
if (insn & (1 << 21)) |
| 1467 |
gen_op_iwmmxt_minsl_M0_wRn(rd1); |
| 1468 |
else
|
| 1469 |
gen_op_iwmmxt_minul_M0_wRn(rd1); |
| 1470 |
break;
|
| 1471 |
case 3: |
| 1472 |
return 1; |
| 1473 |
} |
| 1474 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1475 |
gen_op_iwmmxt_set_mup(); |
| 1476 |
break;
|
| 1477 |
case 0x016: case 0x216: case 0x416: case 0x616: /* WMAX */ |
| 1478 |
case 0x816: case 0xa16: case 0xc16: case 0xe16: |
| 1479 |
wrd = (insn >> 12) & 0xf; |
| 1480 |
rd0 = (insn >> 16) & 0xf; |
| 1481 |
rd1 = (insn >> 0) & 0xf; |
| 1482 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1483 |
switch ((insn >> 22) & 3) { |
| 1484 |
case 0: |
| 1485 |
if (insn & (1 << 21)) |
| 1486 |
gen_op_iwmmxt_maxsb_M0_wRn(rd1); |
| 1487 |
else
|
| 1488 |
gen_op_iwmmxt_maxub_M0_wRn(rd1); |
| 1489 |
break;
|
| 1490 |
case 1: |
| 1491 |
if (insn & (1 << 21)) |
| 1492 |
gen_op_iwmmxt_maxsw_M0_wRn(rd1); |
| 1493 |
else
|
| 1494 |
gen_op_iwmmxt_maxuw_M0_wRn(rd1); |
| 1495 |
break;
|
| 1496 |
case 2: |
| 1497 |
if (insn & (1 << 21)) |
| 1498 |
gen_op_iwmmxt_maxsl_M0_wRn(rd1); |
| 1499 |
else
|
| 1500 |
gen_op_iwmmxt_maxul_M0_wRn(rd1); |
| 1501 |
break;
|
| 1502 |
case 3: |
| 1503 |
return 1; |
| 1504 |
} |
| 1505 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1506 |
gen_op_iwmmxt_set_mup(); |
| 1507 |
break;
|
| 1508 |
case 0x002: case 0x102: case 0x202: case 0x302: /* WALIGNI */ |
| 1509 |
case 0x402: case 0x502: case 0x602: case 0x702: |
| 1510 |
wrd = (insn >> 12) & 0xf; |
| 1511 |
rd0 = (insn >> 16) & 0xf; |
| 1512 |
rd1 = (insn >> 0) & 0xf; |
| 1513 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1514 |
gen_op_movl_T0_im((insn >> 20) & 3); |
| 1515 |
gen_op_iwmmxt_align_M0_T0_wRn(rd1); |
| 1516 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1517 |
gen_op_iwmmxt_set_mup(); |
| 1518 |
break;
|
| 1519 |
case 0x01a: case 0x11a: case 0x21a: case 0x31a: /* WSUB */ |
| 1520 |
case 0x41a: case 0x51a: case 0x61a: case 0x71a: |
| 1521 |
case 0x81a: case 0x91a: case 0xa1a: case 0xb1a: |
| 1522 |
case 0xc1a: case 0xd1a: case 0xe1a: case 0xf1a: |
| 1523 |
wrd = (insn >> 12) & 0xf; |
| 1524 |
rd0 = (insn >> 16) & 0xf; |
| 1525 |
rd1 = (insn >> 0) & 0xf; |
| 1526 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1527 |
switch ((insn >> 20) & 0xf) { |
| 1528 |
case 0x0: |
| 1529 |
gen_op_iwmmxt_subnb_M0_wRn(rd1); |
| 1530 |
break;
|
| 1531 |
case 0x1: |
| 1532 |
gen_op_iwmmxt_subub_M0_wRn(rd1); |
| 1533 |
break;
|
| 1534 |
case 0x3: |
| 1535 |
gen_op_iwmmxt_subsb_M0_wRn(rd1); |
| 1536 |
break;
|
| 1537 |
case 0x4: |
| 1538 |
gen_op_iwmmxt_subnw_M0_wRn(rd1); |
| 1539 |
break;
|
| 1540 |
case 0x5: |
| 1541 |
gen_op_iwmmxt_subuw_M0_wRn(rd1); |
| 1542 |
break;
|
| 1543 |
case 0x7: |
| 1544 |
gen_op_iwmmxt_subsw_M0_wRn(rd1); |
| 1545 |
break;
|
| 1546 |
case 0x8: |
| 1547 |
gen_op_iwmmxt_subnl_M0_wRn(rd1); |
| 1548 |
break;
|
| 1549 |
case 0x9: |
| 1550 |
gen_op_iwmmxt_subul_M0_wRn(rd1); |
| 1551 |
break;
|
| 1552 |
case 0xb: |
| 1553 |
gen_op_iwmmxt_subsl_M0_wRn(rd1); |
| 1554 |
break;
|
| 1555 |
default:
|
| 1556 |
return 1; |
| 1557 |
} |
| 1558 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1559 |
gen_op_iwmmxt_set_mup(); |
| 1560 |
gen_op_iwmmxt_set_cup(); |
| 1561 |
break;
|
| 1562 |
case 0x01e: case 0x11e: case 0x21e: case 0x31e: /* WSHUFH */ |
| 1563 |
case 0x41e: case 0x51e: case 0x61e: case 0x71e: |
| 1564 |
case 0x81e: case 0x91e: case 0xa1e: case 0xb1e: |
| 1565 |
case 0xc1e: case 0xd1e: case 0xe1e: case 0xf1e: |
| 1566 |
wrd = (insn >> 12) & 0xf; |
| 1567 |
rd0 = (insn >> 16) & 0xf; |
| 1568 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1569 |
gen_op_movl_T0_im(((insn >> 16) & 0xf0) | (insn & 0x0f)); |
| 1570 |
gen_op_iwmmxt_shufh_M0_T0(); |
| 1571 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1572 |
gen_op_iwmmxt_set_mup(); |
| 1573 |
gen_op_iwmmxt_set_cup(); |
| 1574 |
break;
|
| 1575 |
case 0x018: case 0x118: case 0x218: case 0x318: /* WADD */ |
| 1576 |
case 0x418: case 0x518: case 0x618: case 0x718: |
| 1577 |
case 0x818: case 0x918: case 0xa18: case 0xb18: |
| 1578 |
case 0xc18: case 0xd18: case 0xe18: case 0xf18: |
| 1579 |
wrd = (insn >> 12) & 0xf; |
| 1580 |
rd0 = (insn >> 16) & 0xf; |
| 1581 |
rd1 = (insn >> 0) & 0xf; |
| 1582 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1583 |
switch ((insn >> 20) & 0xf) { |
| 1584 |
case 0x0: |
| 1585 |
gen_op_iwmmxt_addnb_M0_wRn(rd1); |
| 1586 |
break;
|
| 1587 |
case 0x1: |
| 1588 |
gen_op_iwmmxt_addub_M0_wRn(rd1); |
| 1589 |
break;
|
| 1590 |
case 0x3: |
| 1591 |
gen_op_iwmmxt_addsb_M0_wRn(rd1); |
| 1592 |
break;
|
| 1593 |
case 0x4: |
| 1594 |
gen_op_iwmmxt_addnw_M0_wRn(rd1); |
| 1595 |
break;
|
| 1596 |
case 0x5: |
| 1597 |
gen_op_iwmmxt_adduw_M0_wRn(rd1); |
| 1598 |
break;
|
| 1599 |
case 0x7: |
| 1600 |
gen_op_iwmmxt_addsw_M0_wRn(rd1); |
| 1601 |
break;
|
| 1602 |
case 0x8: |
| 1603 |
gen_op_iwmmxt_addnl_M0_wRn(rd1); |
| 1604 |
break;
|
| 1605 |
case 0x9: |
| 1606 |
gen_op_iwmmxt_addul_M0_wRn(rd1); |
| 1607 |
break;
|
| 1608 |
case 0xb: |
| 1609 |
gen_op_iwmmxt_addsl_M0_wRn(rd1); |
| 1610 |
break;
|
| 1611 |
default:
|
| 1612 |
return 1; |
| 1613 |
} |
| 1614 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1615 |
gen_op_iwmmxt_set_mup(); |
| 1616 |
gen_op_iwmmxt_set_cup(); |
| 1617 |
break;
|
| 1618 |
case 0x008: case 0x108: case 0x208: case 0x308: /* WPACK */ |
| 1619 |
case 0x408: case 0x508: case 0x608: case 0x708: |
| 1620 |
case 0x808: case 0x908: case 0xa08: case 0xb08: |
| 1621 |
case 0xc08: case 0xd08: case 0xe08: case 0xf08: |
| 1622 |
wrd = (insn >> 12) & 0xf; |
| 1623 |
rd0 = (insn >> 16) & 0xf; |
| 1624 |
rd1 = (insn >> 0) & 0xf; |
| 1625 |
gen_op_iwmmxt_movq_M0_wRn(rd0); |
| 1626 |
if (!(insn & (1 << 20))) |
| 1627 |
return 1; |
| 1628 |
switch ((insn >> 22) & 3) { |
| 1629 |
case 0: |
| 1630 |
return 1; |
| 1631 |
case 1: |
| 1632 |
if (insn & (1 << 21)) |
| 1633 |
gen_op_iwmmxt_packsw_M0_wRn(rd1); |
| 1634 |
else
|
| 1635 |
gen_op_iwmmxt_packuw_M0_wRn(rd1); |
| 1636 |
break;
|
| 1637 |
case 2: |
| 1638 |
if (insn & (1 << 21)) |
| 1639 |
gen_op_iwmmxt_packsl_M0_wRn(rd1); |
| 1640 |
else
|
| 1641 |
gen_op_iwmmxt_packul_M0_wRn(rd1); |
| 1642 |
break;
|
| 1643 |
case 3: |
| 1644 |
if (insn & (1 << 21)) |
| 1645 |
gen_op_iwmmxt_packsq_M0_wRn(rd1); |
| 1646 |
else
|
| 1647 |
gen_op_iwmmxt_packuq_M0_wRn(rd1); |
| 1648 |
break;
|
| 1649 |
} |
| 1650 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1651 |
gen_op_iwmmxt_set_mup(); |
| 1652 |
gen_op_iwmmxt_set_cup(); |
| 1653 |
break;
|
| 1654 |
case 0x201: case 0x203: case 0x205: case 0x207: |
| 1655 |
case 0x209: case 0x20b: case 0x20d: case 0x20f: |
| 1656 |
case 0x211: case 0x213: case 0x215: case 0x217: |
| 1657 |
case 0x219: case 0x21b: case 0x21d: case 0x21f: |
| 1658 |
wrd = (insn >> 5) & 0xf; |
| 1659 |
rd0 = (insn >> 12) & 0xf; |
| 1660 |
rd1 = (insn >> 0) & 0xf; |
| 1661 |
if (rd0 == 0xf || rd1 == 0xf) |
| 1662 |
return 1; |
| 1663 |
gen_op_iwmmxt_movq_M0_wRn(wrd); |
| 1664 |
switch ((insn >> 16) & 0xf) { |
| 1665 |
case 0x0: /* TMIA */ |
| 1666 |
gen_movl_T0_reg(s, rd0); |
| 1667 |
gen_movl_T1_reg(s, rd1); |
| 1668 |
gen_op_iwmmxt_muladdsl_M0_T0_T1(); |
| 1669 |
break;
|
| 1670 |
case 0x8: /* TMIAPH */ |
| 1671 |
gen_movl_T0_reg(s, rd0); |
| 1672 |
gen_movl_T1_reg(s, rd1); |
| 1673 |
gen_op_iwmmxt_muladdsw_M0_T0_T1(); |
| 1674 |
break;
|
| 1675 |
case 0xc: case 0xd: case 0xe: case 0xf: /* TMIAxy */ |
| 1676 |
gen_movl_T1_reg(s, rd0); |
| 1677 |
if (insn & (1 << 16)) |
| 1678 |
gen_op_shrl_T1_im(16);
|
| 1679 |
gen_op_movl_T0_T1(); |
| 1680 |
gen_movl_T1_reg(s, rd1); |
| 1681 |
if (insn & (1 << 17)) |
| 1682 |
gen_op_shrl_T1_im(16);
|
| 1683 |
gen_op_iwmmxt_muladdswl_M0_T0_T1(); |
| 1684 |
break;
|
| 1685 |
default:
|
| 1686 |
return 1; |
| 1687 |
} |
| 1688 |
gen_op_iwmmxt_movq_wRn_M0(wrd); |
| 1689 |
gen_op_iwmmxt_set_mup(); |
| 1690 |
break;
|
| 1691 |
default:
|
| 1692 |
return 1; |
| 1693 |
} |
| 1694 |
|
| 1695 |
return 0; |
| 1696 |
} |
| 1697 |
|
| 1698 |
/* Disassemble an XScale DSP instruction. Returns nonzero if an error occured
|
| 1699 |
(ie. an undefined instruction). */
|
| 1700 |
static int disas_dsp_insn(CPUState *env, DisasContext *s, uint32_t insn) |
| 1701 |
{
|
| 1702 |
int acc, rd0, rd1, rdhi, rdlo;
|
| 1703 |
|
| 1704 |
if ((insn & 0x0ff00f10) == 0x0e200010) { |
| 1705 |
/* Multiply with Internal Accumulate Format */
|
| 1706 |
rd0 = (insn >> 12) & 0xf; |
| 1707 |
rd1 = insn & 0xf;
|
| 1708 |
acc = (insn >> 5) & 7; |
| 1709 |
|
| 1710 |
if (acc != 0) |
| 1711 |
return 1; |
| 1712 |
|
| 1713 |
switch ((insn >> 16) & 0xf) { |
| 1714 |
case 0x0: /* MIA */ |
| 1715 |
gen_movl_T0_reg(s, rd0); |
| 1716 |
gen_movl_T1_reg(s, rd1); |
| 1717 |
gen_op_iwmmxt_muladdsl_M0_T0_T1(); |
| 1718 |
break;
|
| 1719 |
case 0x8: /* MIAPH */ |
| 1720 |
gen_movl_T0_reg(s, rd0); |
| 1721 |
gen_movl_T1_reg(s, rd1); |
| 1722 |
gen_op_iwmmxt_muladdsw_M0_T0_T1(); |
| 1723 |
break;
|
| 1724 |
case 0xc: /* MIABB */ |
| 1725 |
case 0xd: /* MIABT */ |
| 1726 |
case 0xe: /* MIATB */ |
| 1727 |
case 0xf: /* MIATT */ |
| 1728 |
gen_movl_T1_reg(s, rd0); |
| 1729 |
if (insn & (1 << 16)) |
| 1730 |
gen_op_shrl_T1_im(16);
|
| 1731 |
gen_op_movl_T0_T1(); |
| 1732 |
gen_movl_T1_reg(s, rd1); |
| 1733 |
if (insn & (1 << 17)) |
| 1734 |
gen_op_shrl_T1_im(16);
|
| 1735 |
gen_op_iwmmxt_muladdswl_M0_T0_T1(); |
| 1736 |
break;
|
| 1737 |
default:
|
| 1738 |
return 1; |
| 1739 |
} |
| 1740 |
|
| 1741 |
gen_op_iwmmxt_movq_wRn_M0(acc); |
| 1742 |
return 0; |
| 1743 |
} |
| 1744 |
|
| 1745 |
if ((insn & 0x0fe00ff8) == 0x0c400000) { |
| 1746 |
/* Internal Accumulator Access Format */
|
| 1747 |
rdhi = (insn >> 16) & 0xf; |
| 1748 |
rdlo = (insn >> 12) & 0xf; |
| 1749 |
acc = insn & 7;
|
| 1750 |
|
| 1751 |
if (acc != 0) |
| 1752 |
return 1; |
| 1753 |
|
| 1754 |
if (insn & ARM_CP_RW_BIT) { /* MRA */ |
| 1755 |
gen_op_iwmmxt_movl_T0_T1_wRn(acc); |
| 1756 |
gen_movl_reg_T0(s, rdlo); |
| 1757 |
gen_op_movl_T0_im((1 << (40 - 32)) - 1); |
| 1758 |
gen_op_andl_T0_T1(); |
| 1759 |
gen_movl_reg_T0(s, rdhi); |
| 1760 |
} else { /* MAR */ |
| 1761 |
gen_movl_T0_reg(s, rdlo); |
| 1762 |
gen_movl_T1_reg(s, rdhi); |
| 1763 |
gen_op_iwmmxt_movl_wRn_T0_T1(acc); |
| 1764 |
} |
| 1765 |
return 0; |
| 1766 |
} |
| 1767 |
|
| 1768 |
return 1; |
| 1769 |
} |
| 1770 |
|
| 1771 |
/* Disassemble system coprocessor instruction. Return nonzero if
|
| 1772 |
instruction is not defined. */
|
| 1773 |
static int disas_cp_insn(CPUState *env, DisasContext *s, uint32_t insn) |
| 1774 |
{
|
| 1775 |
uint32_t rd = (insn >> 12) & 0xf; |
| 1776 |
uint32_t cp = (insn >> 8) & 0xf; |
| 1777 |
if (IS_USER(s)) {
|
| 1778 |
return 1; |
| 1779 |
} |
| 1780 |
|
| 1781 |
if (insn & ARM_CP_RW_BIT) {
|
| 1782 |
if (!env->cp[cp].cp_read)
|
| 1783 |
return 1; |
| 1784 |
gen_op_movl_T0_im((uint32_t) s->pc); |
| 1785 |
gen_set_pc_T0(); |
| 1786 |
gen_op_movl_T0_cp(insn); |
| 1787 |
gen_movl_reg_T0(s, rd); |
| 1788 |
} else {
|
| 1789 |
if (!env->cp[cp].cp_write)
|
| 1790 |
return 1; |
| 1791 |
gen_op_movl_T0_im((uint32_t) s->pc); |
| 1792 |
gen_set_pc_T0(); |
| 1793 |
gen_movl_T0_reg(s, rd); |
| 1794 |
gen_op_movl_cp_T0(insn); |
| 1795 |
} |
| 1796 |
return 0; |
| 1797 |
} |
| 1798 |
|
| 1799 |
static int cp15_user_ok(uint32_t insn) |
| 1800 |
{
|
| 1801 |
int cpn = (insn >> 16) & 0xf; |
| 1802 |
int cpm = insn & 0xf; |
| 1803 |
int op = ((insn >> 5) & 7) | ((insn >> 18) & 0x38); |
| 1804 |
|
| 1805 |
if (cpn == 13 && cpm == 0) { |
| 1806 |
/* TLS register. */
|
| 1807 |
if (op == 2 || (op == 3 && (insn & ARM_CP_RW_BIT))) |
| 1808 |
return 1; |
| 1809 |
} |
| 1810 |
if (cpn == 7) { |
| 1811 |
/* ISB, DSB, DMB. */
|
| 1812 |
if ((cpm == 5 && op == 4) |
| 1813 |
|| (cpm == 10 && (op == 4 || op == 5))) |
| 1814 |
return 1; |
| 1815 |
} |
| 1816 |
return 0; |
| 1817 |
} |
| 1818 |
|
| 1819 |
/* Disassemble system coprocessor (cp15) instruction. Return nonzero if
|
| 1820 |
instruction is not defined. */
|
| 1821 |
static int disas_cp15_insn(CPUState *env, DisasContext *s, uint32_t insn) |
| 1822 |
{
|
| 1823 |
uint32_t rd; |
| 1824 |
|
| 1825 |
/* M profile cores use memory mapped registers instead of cp15. */
|
| 1826 |
if (arm_feature(env, ARM_FEATURE_M))
|
| 1827 |
return 1; |
| 1828 |
|
| 1829 |
if ((insn & (1 << 25)) == 0) { |
| 1830 |
if (insn & (1 << 20)) { |
| 1831 |
/* mrrc */
|
| 1832 |
return 1; |
| 1833 |
} |
| 1834 |
/* mcrr. Used for block cache operations, so implement as no-op. */
|
| 1835 |
return 0; |
| 1836 |
} |
| 1837 |
if ((insn & (1 << 4)) == 0) { |
| 1838 |
/* cdp */
|
| 1839 |
return 1; |
| 1840 |
} |
| 1841 |
if (IS_USER(s) && !cp15_user_ok(insn)) {
|
| 1842 |
return 1; |
| 1843 |
} |
| 1844 |
if ((insn & 0x0fff0fff) == 0x0e070f90 |
| 1845 |
|| (insn & 0x0fff0fff) == 0x0e070f58) { |
| 1846 |
/* Wait for interrupt. */
|
| 1847 |
gen_op_movl_T0_im((long)s->pc);
|
| 1848 |
gen_set_pc_T0(); |
| 1849 |
s->is_jmp = DISAS_WFI; |
| 1850 |
return 0; |
| 1851 |
} |
| 1852 |
rd = (insn >> 12) & 0xf; |
| 1853 |
if (insn & ARM_CP_RW_BIT) {
|
| 1854 |
gen_op_movl_T0_cp15(insn); |
| 1855 |
/* If the destination register is r15 then sets condition codes. */
|
| 1856 |
if (rd != 15) |
| 1857 |
gen_movl_reg_T0(s, rd); |
| 1858 |
} else {
|
| 1859 |
gen_movl_T0_reg(s, rd); |
| 1860 |
gen_op_movl_cp15_T0(insn); |
| 1861 |
/* Normally we would always end the TB here, but Linux
|
| 1862 |
* arch/arm/mach-pxa/sleep.S expects two instructions following
|
| 1863 |
* an MMU enable to execute from cache. Imitate this behaviour. */
|
| 1864 |
if (!arm_feature(env, ARM_FEATURE_XSCALE) ||
|
| 1865 |
(insn & 0x0fff0fff) != 0x0e010f10) |
| 1866 |
gen_lookup_tb(s); |
| 1867 |
} |
| 1868 |
return 0; |
| 1869 |
} |
| 1870 |
|
| 1871 |
#define VFP_REG_SHR(x, n) (((n) > 0) ? (x) >> (n) : (x) << -(n)) |
| 1872 |
#define VFP_SREG(insn, bigbit, smallbit) \
|
| 1873 |
((VFP_REG_SHR(insn, bigbit - 1) & 0x1e) | (((insn) >> (smallbit)) & 1)) |
| 1874 |
#define VFP_DREG(reg, insn, bigbit, smallbit) do { \ |
| 1875 |
if (arm_feature(env, ARM_FEATURE_VFP3)) { \
|
| 1876 |
reg = (((insn) >> (bigbit)) & 0x0f) \
|
| 1877 |
| (((insn) >> ((smallbit) - 4)) & 0x10); \ |
| 1878 |
} else { \
|
| 1879 |
if (insn & (1 << (smallbit))) \ |
| 1880 |
return 1; \ |
| 1881 |
reg = ((insn) >> (bigbit)) & 0x0f; \
|
| 1882 |
}} while (0) |
| 1883 |
|
| 1884 |
#define VFP_SREG_D(insn) VFP_SREG(insn, 12, 22) |
| 1885 |
#define VFP_DREG_D(reg, insn) VFP_DREG(reg, insn, 12, 22) |
| 1886 |
#define VFP_SREG_N(insn) VFP_SREG(insn, 16, 7) |
| 1887 |
#define VFP_DREG_N(reg, insn) VFP_DREG(reg, insn, 16, 7) |
| 1888 |
#define VFP_SREG_M(insn) VFP_SREG(insn, 0, 5) |
| 1889 |
#define VFP_DREG_M(reg, insn) VFP_DREG(reg, insn, 0, 5) |
| 1890 |
|
| 1891 |
static inline int |
| 1892 |
vfp_enabled(CPUState * env) |
| 1893 |
{
|
| 1894 |
return ((env->vfp.xregs[ARM_VFP_FPEXC] & (1 << 30)) != 0); |
| 1895 |
} |
| 1896 |
|
| 1897 |
/* Disassemble a VFP instruction. Returns nonzero if an error occured
|
| 1898 |
(ie. an undefined instruction). */
|
| 1899 |
static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn) |
| 1900 |
{
|
| 1901 |
uint32_t rd, rn, rm, op, i, n, offset, delta_d, delta_m, bank_mask; |
| 1902 |
int dp, veclen;
|
| 1903 |
|
| 1904 |
if (!arm_feature(env, ARM_FEATURE_VFP))
|
| 1905 |
return 1; |
| 1906 |
|
| 1907 |
if (!vfp_enabled(env)) {
|
| 1908 |
/* VFP disabled. Only allow fmxr/fmrx to/from some control regs. */
|
| 1909 |
if ((insn & 0x0fe00fff) != 0x0ee00a10) |
| 1910 |
return 1; |
| 1911 |
rn = (insn >> 16) & 0xf; |
| 1912 |
if (rn != ARM_VFP_FPSID && rn != ARM_VFP_FPEXC
|
| 1913 |
&& rn != ARM_VFP_MVFR1 && rn != ARM_VFP_MVFR0) |
| 1914 |
return 1; |
| 1915 |
} |
| 1916 |
dp = ((insn & 0xf00) == 0xb00); |
| 1917 |
switch ((insn >> 24) & 0xf) { |
| 1918 |
case 0xe: |
| 1919 |
if (insn & (1 << 4)) { |
| 1920 |
/* single register transfer */
|
| 1921 |
rd = (insn >> 12) & 0xf; |
| 1922 |
if (dp) {
|
| 1923 |
int size;
|
| 1924 |
int pass;
|
| 1925 |
|
| 1926 |
VFP_DREG_N(rn, insn); |
| 1927 |
if (insn & 0xf) |
| 1928 |
return 1; |
| 1929 |
if (insn & 0x00c00060 |
| 1930 |
&& !arm_feature(env, ARM_FEATURE_NEON)) |
| 1931 |
return 1; |
| 1932 |
|
| 1933 |
pass = (insn >> 21) & 1; |
| 1934 |
if (insn & (1 << 22)) { |
| 1935 |
size = 0;
|
| 1936 |
offset = ((insn >> 5) & 3) * 8; |
| 1937 |
} else if (insn & (1 << 5)) { |
| 1938 |
size = 1;
|
| 1939 |
offset = (insn & (1 << 6)) ? 16 : 0; |
| 1940 |
} else {
|
| 1941 |
size = 2;
|
| 1942 |
offset = 0;
|
| 1943 |
} |
| 1944 |
if (insn & ARM_CP_RW_BIT) {
|
| 1945 |
/* vfp->arm */
|
| 1946 |
switch (size) {
|
| 1947 |
case 0: |
| 1948 |
NEON_GET_REG(T1, rn, pass); |
| 1949 |
if (offset)
|
| 1950 |
gen_op_shrl_T1_im(offset); |
| 1951 |
if (insn & (1 << 23)) |
| 1952 |
gen_uxtb(cpu_T[1]);
|
| 1953 |
else
|
| 1954 |
gen_sxtb(cpu_T[1]);
|
| 1955 |
break;
|
| 1956 |
case 1: |
| 1957 |
NEON_GET_REG(T1, rn, pass); |
| 1958 |
if (insn & (1 << 23)) { |
| 1959 |
if (offset) {
|
| 1960 |
gen_op_shrl_T1_im(16);
|
| 1961 |
} else {
|
| 1962 |
gen_uxth(cpu_T[1]);
|
| 1963 |
} |
| 1964 |
} else {
|
| 1965 |
if (offset) {
|
| 1966 |
gen_op_sarl_T1_im(16);
|
| 1967 |
} else {
|
| 1968 |
gen_sxth(cpu_T[1]);
|
| 1969 |
} |
| 1970 |
} |
| 1971 |
break;
|
| 1972 |
case 2: |
| 1973 |
NEON_GET_REG(T1, rn, pass); |
| 1974 |
break;
|
| 1975 |
} |
| 1976 |
gen_movl_reg_T1(s, rd); |
| 1977 |
} else {
|
| 1978 |
/* arm->vfp */
|
| 1979 |
gen_movl_T0_reg(s, rd); |
| 1980 |
if (insn & (1 << 23)) { |
| 1981 |
/* VDUP */
|
| 1982 |
if (size == 0) { |
| 1983 |
gen_op_neon_dup_u8(0);
|
| 1984 |
} else if (size == 1) { |
| 1985 |
gen_op_neon_dup_low16(); |
| 1986 |
} |
| 1987 |
NEON_SET_REG(T0, rn, 0);
|
| 1988 |
NEON_SET_REG(T0, rn, 1);
|
| 1989 |
} else {
|
| 1990 |
/* VMOV */
|
| 1991 |
switch (size) {
|
| 1992 |
case 0: |
| 1993 |
NEON_GET_REG(T2, rn, pass); |
| 1994 |
gen_op_movl_T1_im(0xff);
|
| 1995 |
gen_op_andl_T0_T1(); |
| 1996 |
gen_op_neon_insert_elt(offset, ~(0xff << offset));
|
| 1997 |
NEON_SET_REG(T2, rn, pass); |
| 1998 |
break;
|
| 1999 |
case 1: |
| 2000 |
NEON_GET_REG(T2, rn, pass); |
| 2001 |
gen_op_movl_T1_im(0xffff);
|
| 2002 |
gen_op_andl_T0_T1(); |
| 2003 |
bank_mask = offset ? 0xffff : 0xffff0000; |
| 2004 |
gen_op_neon_insert_elt(offset, bank_mask); |
| 2005 |
NEON_SET_REG(T2, rn, pass); |
| 2006 |
break;
|
| 2007 |
case 2: |
| 2008 |
NEON_SET_REG(T0, rn, pass); |
| 2009 |
break;
|
| 2010 |
} |
| 2011 |
} |
| 2012 |
} |
| 2013 |
} else { /* !dp */ |
| 2014 |
if ((insn & 0x6f) != 0x00) |
| 2015 |
return 1; |
| 2016 |
rn = VFP_SREG_N(insn); |
| 2017 |
if (insn & ARM_CP_RW_BIT) {
|
| 2018 |
/* vfp->arm */
|
| 2019 |
if (insn & (1 << 21)) { |
| 2020 |
/* system register */
|
| 2021 |
rn >>= 1;
|
| 2022 |
|
| 2023 |
switch (rn) {
|
| 2024 |
case ARM_VFP_FPSID:
|
| 2025 |
/* VFP2 allows access for FSID from userspace.
|
| 2026 |
VFP3 restricts all id registers to privileged
|
| 2027 |
accesses. */
|
| 2028 |
if (IS_USER(s)
|
| 2029 |
&& arm_feature(env, ARM_FEATURE_VFP3)) |
| 2030 |
return 1; |
| 2031 |
gen_op_vfp_movl_T0_xreg(rn); |
| 2032 |
break;
|
| 2033 |
case ARM_VFP_FPEXC:
|
| 2034 |
if (IS_USER(s))
|
| 2035 |
return 1; |
| 2036 |
gen_op_vfp_movl_T0_xreg(rn); |
| 2037 |
break;
|
| 2038 |
case ARM_VFP_FPINST:
|
| 2039 |
case ARM_VFP_FPINST2:
|
| 2040 |
/* Not present in VFP3. */
|
| 2041 |
if (IS_USER(s)
|
| 2042 |
|| arm_feature(env, ARM_FEATURE_VFP3)) |
| 2043 |
return 1; |
| 2044 |
gen_op_vfp_movl_T0_xreg(rn); |
| 2045 |
break;
|
| 2046 |
case ARM_VFP_FPSCR:
|
| 2047 |
if (rd == 15) |
| 2048 |
gen_op_vfp_movl_T0_fpscr_flags(); |
| 2049 |
else
|
| 2050 |
gen_op_vfp_movl_T0_fpscr(); |
| 2051 |
break;
|
| 2052 |
case ARM_VFP_MVFR0:
|
| 2053 |
case ARM_VFP_MVFR1:
|
| 2054 |
if (IS_USER(s)
|
| 2055 |
|| !arm_feature(env, ARM_FEATURE_VFP3)) |
| 2056 |
return 1; |
| 2057 |
gen_op_vfp_movl_T0_xreg(rn); |
| 2058 |
break;
|
| 2059 |
default:
|
| 2060 |
return 1; |
| 2061 |
} |
| 2062 |
} else {
|
| 2063 |
gen_mov_F0_vreg(0, rn);
|
| 2064 |
gen_op_vfp_mrs(); |
| 2065 |
} |
| 2066 |
if (rd == 15) { |
| 2067 |
/* Set the 4 flag bits in the CPSR. */
|
| 2068 |
gen_op_movl_cpsr_T0(0xf0000000);
|
| 2069 |
} else
|
| 2070 |
gen_movl_reg_T0(s, rd); |
| 2071 |
} else {
|
| 2072 |
/* arm->vfp */
|
| 2073 |
gen_movl_T0_reg(s, rd); |
| 2074 |
if (insn & (1 << 21)) { |
| 2075 |
rn >>= 1;
|
| 2076 |
/* system register */
|
| 2077 |
switch (rn) {
|
| 2078 |
case ARM_VFP_FPSID:
|
| 2079 |
case ARM_VFP_MVFR0:
|
| 2080 |
case ARM_VFP_MVFR1:
|
| 2081 |
/* Writes are ignored. */
|
| 2082 |
break;
|
| 2083 |
case ARM_VFP_FPSCR:
|
| 2084 |
gen_op_vfp_movl_fpscr_T0(); |
| 2085 |
gen_lookup_tb(s); |
| 2086 |
break;
|
| 2087 |
case ARM_VFP_FPEXC:
|
| 2088 |
if (IS_USER(s))
|
| 2089 |
return 1; |
| 2090 |
gen_op_vfp_movl_xreg_T0(rn); |
| 2091 |
gen_lookup_tb(s); |
| 2092 |
break;
|
| 2093 |
case ARM_VFP_FPINST:
|
| 2094 |
case ARM_VFP_FPINST2:
|
| 2095 |
gen_op_vfp_movl_xreg_T0(rn); |
| 2096 |
break;
|
| 2097 |
default:
|
| 2098 |
return 1; |
| 2099 |
} |
| 2100 |
} else {
|
| 2101 |
gen_op_vfp_msr(); |
| 2102 |
gen_mov_vreg_F0(0, rn);
|
| 2103 |
} |
| 2104 |
} |
| 2105 |
} |
| 2106 |
} else {
|
| 2107 |
/* data processing */
|
| 2108 |
/* The opcode is in bits 23, 21, 20 and 6. */
|
| 2109 |
op = ((insn >> 20) & 8) | ((insn >> 19) & 6) | ((insn >> 6) & 1); |
| 2110 |
if (dp) {
|
| 2111 |
if (op == 15) { |
| 2112 |
/* rn is opcode */
|
| 2113 |
rn = ((insn >> 15) & 0x1e) | ((insn >> 7) & 1); |
| 2114 |
} else {
|
| 2115 |
/* rn is register number */
|
| 2116 |
VFP_DREG_N(rn, insn); |
| 2117 |
} |
| 2118 |
|
| 2119 |
if (op == 15 && (rn == 15 || rn > 17)) { |
| 2120 |
/* Integer or single precision destination. */
|
| 2121 |
rd = VFP_SREG_D(insn); |
| 2122 |
} else {
|
| 2123 |
VFP_DREG_D(rd, insn); |
| 2124 |
} |
| 2125 |
|
| 2126 |
if (op == 15 && (rn == 16 || rn == 17)) { |
| 2127 |
/* Integer source. */
|
| 2128 |
rm = ((insn << 1) & 0x1e) | ((insn >> 5) & 1); |
| 2129 |
} else {
|
| 2130 |
VFP_DREG_M(rm, insn); |
| 2131 |
} |
| 2132 |
} else {
|
| 2133 |
rn = VFP_SREG_N(insn); |
| 2134 |
if (op == 15 && rn == 15) { |
| 2135 |
/* Double precision destination. */
|
| 2136 |
VFP_DREG_D(rd, insn); |
| 2137 |
} else {
|
| 2138 |
rd = VFP_SREG_D(insn); |
| 2139 |
} |
| 2140 |
rm = VFP_SREG_M(insn); |
| 2141 |
} |
| 2142 |
|
| 2143 |
veclen = env->vfp.vec_len; |
| 2144 |
if (op == 15 && rn > 3) |
| 2145 |
veclen = 0;
|
| 2146 |
|
| 2147 |
/* Shut up compiler warnings. */
|
| 2148 |
delta_m = 0;
|
| 2149 |
delta_d = 0;
|
| 2150 |
bank_mask = 0;
|
| 2151 |
|
| 2152 |
if (veclen > 0) { |
| 2153 |
if (dp)
|
| 2154 |
bank_mask = 0xc;
|
| 2155 |
else
|
| 2156 |
bank_mask = 0x18;
|
| 2157 |
|
| 2158 |
/* Figure out what type of vector operation this is. */
|
| 2159 |
if ((rd & bank_mask) == 0) { |
| 2160 |
/* scalar */
|
| 2161 |
veclen = 0;
|
| 2162 |
} else {
|
| 2163 |
if (dp)
|
| 2164 |
delta_d = (env->vfp.vec_stride >> 1) + 1; |
| 2165 |
else
|
| 2166 |
delta_d = env->vfp.vec_stride + 1;
|
| 2167 |
|
| 2168 |
if ((rm & bank_mask) == 0) { |
| 2169 |
/* mixed scalar/vector */
|
| 2170 |
delta_m = 0;
|
| 2171 |
} else {
|
| 2172 |
/* vector */
|
| 2173 |
delta_m = delta_d; |
| 2174 |
} |
| 2175 |
} |
| 2176 |
} |
| 2177 |
|
| 2178 |
/* Load the initial operands. */
|
| 2179 |
if (op == 15) { |
| 2180 |
switch (rn) {
|
| 2181 |
case 16: |
| 2182 |
case 17: |
| 2183 |
/* Integer source */
|
| 2184 |
gen_mov_F0_vreg(0, rm);
|
| 2185 |
break;
|
| 2186 |
case 8: |
| 2187 |
case 9: |
| 2188 |
/* Compare */
|
| 2189 |
gen_mov_F0_vreg(dp, rd); |
| 2190 |
gen_mov_F1_vreg(dp, rm); |
| 2191 |
break;
|
| 2192 |
case 10: |
| 2193 |
case 11: |
| 2194 |
/* Compare with zero */
|
| 2195 |
gen_mov_F0_vreg(dp, rd); |
| 2196 |
gen_vfp_F1_ld0(dp); |
| 2197 |
break;
|
| 2198 |
case 20: |
| 2199 |
case 21: |
| 2200 |
case 22: |
| 2201 |
case 23: |
| 2202 |
/* Source and destination the same. */
|
| 2203 |
gen_mov_F0_vreg(dp, rd); |
| 2204 |
break;
|
| 2205 |
default:
|
| 2206 |
/* One source operand. */
|
| 2207 |
gen_mov_F0_vreg(dp, rm); |
| 2208 |
break;
|
| 2209 |
} |
| 2210 |
} else {
|
| 2211 |
/* Two source operands. */
|
| 2212 |
gen_mov_F0_vreg(dp, rn); |
| 2213 |
gen_mov_F1_vreg(dp, rm); |
| 2214 |
} |
| 2215 |
|
| 2216 |
for (;;) {
|
| 2217 |
/* Perform the calculation. */
|
| 2218 |
switch (op) {
|
| 2219 |
case 0: /* mac: fd + (fn * fm) */ |
| 2220 |
gen_vfp_mul(dp); |
| 2221 |
gen_mov_F1_vreg(dp, rd); |
| 2222 |
gen_vfp_add(dp); |
| 2223 |
break;
|
| 2224 |
case 1: /* nmac: fd - (fn * fm) */ |
| 2225 |
gen_vfp_mul(dp); |
| 2226 |
gen_vfp_neg(dp); |
| 2227 |
gen_mov_F1_vreg(dp, rd); |
| 2228 |
gen_vfp_add(dp); |
| 2229 |
break;
|
| 2230 |
case 2: /* msc: -fd + (fn * fm) */ |
| 2231 |
gen_vfp_mul(dp); |
| 2232 |
gen_mov_F1_vreg(dp, rd); |
| 2233 |
gen_vfp_sub(dp); |
| 2234 |
break;
|
| 2235 |
case 3: /* nmsc: -fd - (fn * fm) */ |
| 2236 |
gen_vfp_mul(dp); |
| 2237 |
gen_mov_F1_vreg(dp, rd); |
| 2238 |
gen_vfp_add(dp); |
| 2239 |
gen_vfp_neg(dp); |
| 2240 |
break;
|
| 2241 |
case 4: /* mul: fn * fm */ |
| 2242 |
gen_vfp_mul(dp); |
| 2243 |
break;
|
| 2244 |
case 5: /* nmul: -(fn * fm) */ |
| 2245 |
gen_vfp_mul(dp); |
| 2246 |
gen_vfp_neg(dp); |
| 2247 |
break;
|
| 2248 |
case 6: /* add: fn + fm */ |
| 2249 |
gen_vfp_add(dp); |
| 2250 |
break;
|
| 2251 |
case 7: /* sub: fn - fm */ |
| 2252 |
gen_vfp_sub(dp); |
| 2253 |
break;
|
| 2254 |
case 8: /* div: fn / fm */ |
| 2255 |
gen_vfp_div(dp); |
| 2256 |
break;
|
| 2257 |
case 14: /* fconst */ |
| 2258 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2259 |
return 1; |
| 2260 |
|
| 2261 |
n = (insn << 12) & 0x80000000; |
| 2262 |
i = ((insn >> 12) & 0x70) | (insn & 0xf); |
| 2263 |
if (dp) {
|
| 2264 |
if (i & 0x40) |
| 2265 |
i |= 0x3f80;
|
| 2266 |
else
|
| 2267 |
i |= 0x4000;
|
| 2268 |
n |= i << 16;
|
| 2269 |
} else {
|
| 2270 |
if (i & 0x40) |
| 2271 |
i |= 0x780;
|
| 2272 |
else
|
| 2273 |
i |= 0x800;
|
| 2274 |
n |= i << 19;
|
| 2275 |
} |
| 2276 |
gen_vfp_fconst(dp, n); |
| 2277 |
break;
|
| 2278 |
case 15: /* extension space */ |
| 2279 |
switch (rn) {
|
| 2280 |
case 0: /* cpy */ |
| 2281 |
/* no-op */
|
| 2282 |
break;
|
| 2283 |
case 1: /* abs */ |
| 2284 |
gen_vfp_abs(dp); |
| 2285 |
break;
|
| 2286 |
case 2: /* neg */ |
| 2287 |
gen_vfp_neg(dp); |
| 2288 |
break;
|
| 2289 |
case 3: /* sqrt */ |
| 2290 |
gen_vfp_sqrt(dp); |
| 2291 |
break;
|
| 2292 |
case 8: /* cmp */ |
| 2293 |
gen_vfp_cmp(dp); |
| 2294 |
break;
|
| 2295 |
case 9: /* cmpe */ |
| 2296 |
gen_vfp_cmpe(dp); |
| 2297 |
break;
|
| 2298 |
case 10: /* cmpz */ |
| 2299 |
gen_vfp_cmp(dp); |
| 2300 |
break;
|
| 2301 |
case 11: /* cmpez */ |
| 2302 |
gen_vfp_F1_ld0(dp); |
| 2303 |
gen_vfp_cmpe(dp); |
| 2304 |
break;
|
| 2305 |
case 15: /* single<->double conversion */ |
| 2306 |
if (dp)
|
| 2307 |
gen_op_vfp_fcvtsd(); |
| 2308 |
else
|
| 2309 |
gen_op_vfp_fcvtds(); |
| 2310 |
break;
|
| 2311 |
case 16: /* fuito */ |
| 2312 |
gen_vfp_uito(dp); |
| 2313 |
break;
|
| 2314 |
case 17: /* fsito */ |
| 2315 |
gen_vfp_sito(dp); |
| 2316 |
break;
|
| 2317 |
case 20: /* fshto */ |
| 2318 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2319 |
return 1; |
| 2320 |
gen_vfp_shto(dp, rm); |
| 2321 |
break;
|
| 2322 |
case 21: /* fslto */ |
| 2323 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2324 |
return 1; |
| 2325 |
gen_vfp_slto(dp, rm); |
| 2326 |
break;
|
| 2327 |
case 22: /* fuhto */ |
| 2328 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2329 |
return 1; |
| 2330 |
gen_vfp_uhto(dp, rm); |
| 2331 |
break;
|
| 2332 |
case 23: /* fulto */ |
| 2333 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2334 |
return 1; |
| 2335 |
gen_vfp_ulto(dp, rm); |
| 2336 |
break;
|
| 2337 |
case 24: /* ftoui */ |
| 2338 |
gen_vfp_toui(dp); |
| 2339 |
break;
|
| 2340 |
case 25: /* ftouiz */ |
| 2341 |
gen_vfp_touiz(dp); |
| 2342 |
break;
|
| 2343 |
case 26: /* ftosi */ |
| 2344 |
gen_vfp_tosi(dp); |
| 2345 |
break;
|
| 2346 |
case 27: /* ftosiz */ |
| 2347 |
gen_vfp_tosiz(dp); |
| 2348 |
break;
|
| 2349 |
case 28: /* ftosh */ |
| 2350 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2351 |
return 1; |
| 2352 |
gen_vfp_tosh(dp, rm); |
| 2353 |
break;
|
| 2354 |
case 29: /* ftosl */ |
| 2355 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2356 |
return 1; |
| 2357 |
gen_vfp_tosl(dp, rm); |
| 2358 |
break;
|
| 2359 |
case 30: /* ftouh */ |
| 2360 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2361 |
return 1; |
| 2362 |
gen_vfp_touh(dp, rm); |
| 2363 |
break;
|
| 2364 |
case 31: /* ftoul */ |
| 2365 |
if (!arm_feature(env, ARM_FEATURE_VFP3))
|
| 2366 |
return 1; |
| 2367 |
gen_vfp_toul(dp, rm); |
| 2368 |
break;
|
| 2369 |
default: /* undefined */ |
| 2370 |
printf ("rn:%d\n", rn);
|
| 2371 |
return 1; |
| 2372 |
} |
| 2373 |
break;
|
| 2374 |
default: /* undefined */ |
| 2375 |
printf ("op:%d\n", op);
|
| 2376 |
return 1; |
| 2377 |
} |
| 2378 |
|
| 2379 |
/* Write back the result. */
|
| 2380 |
if (op == 15 && (rn >= 8 && rn <= 11)) |
| 2381 |
; /* Comparison, do nothing. */
|
| 2382 |
else if (op == 15 && rn > 17) |
| 2383 |
/* Integer result. */
|
| 2384 |
gen_mov_vreg_F0(0, rd);
|
| 2385 |
else if (op == 15 && rn == 15) |
| 2386 |
/* conversion */
|
| 2387 |
gen_mov_vreg_F0(!dp, rd); |
| 2388 |
else
|
| 2389 |
gen_mov_vreg_F0(dp, rd); |
| 2390 |
|
| 2391 |
/* break out of the loop if we have finished */
|
| 2392 |
if (veclen == 0) |
| 2393 |
break;
|
| 2394 |
|
| 2395 |
if (op == 15 && delta_m == 0) { |
| 2396 |
/* single source one-many */
|
| 2397 |
while (veclen--) {
|
| 2398 |
rd = ((rd + delta_d) & (bank_mask - 1))
|
| 2399 |
| (rd & bank_mask); |
| 2400 |
gen_mov_vreg_F0(dp, rd); |
| 2401 |
} |
| 2402 |
break;
|
| 2403 |
} |
| 2404 |
/* Setup the next operands. */
|
| 2405 |
veclen--; |
| 2406 |
rd = ((rd + delta_d) & (bank_mask - 1))
|
| 2407 |
| (rd & bank_mask); |
| 2408 |
|
| 2409 |
if (op == 15) { |
| 2410 |
/* One source operand. */
|
| 2411 |
rm = ((rm + delta_m) & (bank_mask - 1))
|
| 2412 |
| (rm & bank_mask); |
| 2413 |
gen_mov_F0_vreg(dp, rm); |
| 2414 |
} else {
|
| 2415 |
/* Two source operands. */
|
| 2416 |
rn = ((rn + delta_d) & (bank_mask - 1))
|
| 2417 |
| (rn & bank_mask); |
| 2418 |
gen_mov_F0_vreg(dp, rn); |
| 2419 |
if (delta_m) {
|
| 2420 |
rm = ((rm + delta_m) & (bank_mask - 1))
|
| 2421 |
| (rm & bank_mask); |
| 2422 |
gen_mov_F1_vreg(dp, rm); |
| 2423 |
} |
| 2424 |
} |
| 2425 |
} |
| 2426 |
} |
| 2427 |
break;
|
| 2428 |
case 0xc: |
| 2429 |
case 0xd: |
| 2430 |
if (dp && (insn & 0x03e00000) == 0x00400000) { |
| 2431 |
/* two-register transfer */
|
| 2432 |
rn = (insn >> 16) & 0xf; |
| 2433 |
rd = (insn >> 12) & 0xf; |
| 2434 |
if (dp) {
|
| 2435 |
VFP_DREG_M(rm, insn); |
| 2436 |
} else {
|
| 2437 |
rm = VFP_SREG_M(insn); |
| 2438 |
} |
| 2439 |
|
| 2440 |
if (insn & ARM_CP_RW_BIT) {
|
| 2441 |
/* vfp->arm */
|
| 2442 |
if (dp) {
|
| 2443 |
gen_mov_F0_vreg(1, rm);
|
| 2444 |
gen_op_vfp_mrrd(); |
| 2445 |
gen_movl_reg_T0(s, rd); |
| 2446 |
gen_movl_reg_T1(s, rn); |
| 2447 |
} else {
|
| 2448 |
gen_mov_F0_vreg(0, rm);
|
| 2449 |
gen_op_vfp_mrs(); |
| 2450 |
gen_movl_reg_T0(s, rn); |
| 2451 |
gen_mov_F0_vreg(0, rm + 1); |
| 2452 |
gen_op_vfp_mrs(); |
| 2453 |
gen_movl_reg_T0(s, rd); |
| 2454 |
} |
| 2455 |
} else {
|
| 2456 |
/* arm->vfp */
|
| 2457 |
if (dp) {
|
| 2458 |
gen_movl_T0_reg(s, rd); |
| 2459 |
gen_movl_T1_reg(s, rn); |
| 2460 |
gen_op_vfp_mdrr(); |
| 2461 |
gen_mov_vreg_F0(1, rm);
|
| 2462 |
} else {
|
| 2463 |
gen_movl_T0_reg(s, rn); |
| 2464 |
gen_op_vfp_msr(); |
| 2465 |
gen_mov_vreg_F0(0, rm);
|
| 2466 |
gen_movl_T0_reg(s, rd); |
| 2467 |
gen_op_vfp_msr(); |
| 2468 |
gen_mov_vreg_F0(0, rm + 1); |
| 2469 |
} |
| 2470 |
} |
| 2471 |
} else {
|
| 2472 |
/* Load/store */
|
| 2473 |
rn = (insn >> 16) & 0xf; |
| 2474 |
if (dp)
|
| 2475 |
VFP_DREG_D(rd, insn); |
| 2476 |
else
|
| 2477 |
rd = VFP_SREG_D(insn); |
| 2478 |
if (s->thumb && rn == 15) { |
| 2479 |
gen_op_movl_T1_im(s->pc & ~2);
|
| 2480 |
} else {
|
| 2481 |
gen_movl_T1_reg(s, rn); |
| 2482 |
} |
| 2483 |
if ((insn & 0x01200000) == 0x01000000) { |
| 2484 |
/* Single load/store */
|
| 2485 |
offset = (insn & 0xff) << 2; |
| 2486 |
if ((insn & (1 << 23)) == 0) |
| 2487 |
offset = -offset; |
| 2488 |
gen_op_addl_T1_im(offset); |
| 2489 |
if (insn & (1 << 20)) { |
| 2490 |
gen_vfp_ld(s, dp); |
| 2491 |
gen_mov_vreg_F0(dp, rd); |
| 2492 |
} else {
|
| 2493 |
gen_mov_F0_vreg(dp, rd); |
| 2494 |
gen_vfp_st(s, dp); |
| 2495 |
} |
| 2496 |
} else {
|
| 2497 |
/* load/store multiple */
|
| 2498 |
if (dp)
|
| 2499 |
n = (insn >> 1) & 0x7f; |
| 2500 |
else
|
| 2501 |
n = insn & 0xff;
|
| 2502 |
|
| 2503 |
if (insn & (1 << 24)) /* pre-decrement */ |
| 2504 |
gen_op_addl_T1_im(-((insn & 0xff) << 2)); |
| 2505 |
|
| 2506 |
if (dp)
|
| 2507 |
offset = 8;
|
| 2508 |
else
|
| 2509 |
offset = 4;
|
| 2510 |
for (i = 0; i < n; i++) { |
| 2511 |
if (insn & ARM_CP_RW_BIT) {
|
| 2512 |
/* load */
|
| 2513 |
gen_vfp_ld(s, dp); |
| 2514 |
gen_mov_vreg_F0(dp, rd + i); |
| 2515 |
} else {
|
| 2516 |
/* store */
|
| 2517 |
gen_mov_F0_vreg(dp, rd + i); |
| 2518 |
gen_vfp_st(s, dp); |
| 2519 |
} |
| 2520 |
gen_op_addl_T1_im(offset); |
| 2521 |
} |
| 2522 |
if (insn & (1 << 21)) { |
| 2523 |
/* writeback */
|
| 2524 |
if (insn & (1 << 24)) |
| 2525 |
offset = -offset * n; |
| 2526 |
else if (dp && (insn & 1)) |
| 2527 |
offset = 4;
|
| 2528 |
else
|
| 2529 |
offset = 0;
|
| 2530 |
|
| 2531 |
if (offset != 0) |
| 2532 |
gen_op_addl_T1_im(offset); |
| 2533 |
gen_movl_reg_T1(s, rn); |
| 2534 |
} |
| 2535 |
} |
| 2536 |
} |
| 2537 |
break;
|
| 2538 |
default:
|
| 2539 |
/* Should never happen. */
|
| 2540 |
return 1; |
| 2541 |
} |
| 2542 |
return 0; |
| 2543 |
} |
| 2544 |
|
| 2545 |
static inline void gen_goto_tb(DisasContext *s, int n, uint32_t dest) |
| 2546 |
{
|
| 2547 |
TranslationBlock *tb; |
| 2548 |
|
| 2549 |
tb = s->tb; |
| 2550 |
if ((tb->pc & TARGET_PAGE_MASK) == (dest & TARGET_PAGE_MASK)) {
|
| 2551 |
tcg_gen_goto_tb(n); |
| 2552 |
gen_op_movl_T0_im(dest); |
| 2553 |
gen_set_pc_T0(); |
| 2554 |
tcg_gen_exit_tb((long)tb + n);
|
| 2555 |
} else {
|
| 2556 |
gen_op_movl_T0_im(dest); |
| 2557 |
gen_set_pc_T0(); |
| 2558 |
tcg_gen_exit_tb(0);
|
| 2559 |
} |
| 2560 |
} |
| 2561 |
|
| 2562 |
static inline void gen_jmp (DisasContext *s, uint32_t dest) |
| 2563 |
{
|
| 2564 |
if (__builtin_expect(s->singlestep_enabled, 0)) { |
| 2565 |
/* An indirect jump so that we still trigger the debug exception. */
|
| 2566 |
if (s->thumb)
|
| 2567 |
dest |= 1;
|
| 2568 |
gen_op_movl_T0_im(dest); |
| 2569 |
gen_bx(s); |
| 2570 |
} else {
|
| 2571 |
gen_goto_tb(s, 0, dest);
|
| 2572 |
s->is_jmp = DISAS_TB_JUMP; |
| 2573 |
} |
| 2574 |
} |
| 2575 |
|
| 2576 |
static inline void gen_mulxy(int x, int y) |
| 2577 |
{
|
| 2578 |
if (x)
|
| 2579 |
tcg_gen_sari_i32(cpu_T[0], cpu_T[0], 16); |
| 2580 |
else
|
| 2581 |
gen_sxth(cpu_T[0]);
|
| 2582 |
if (y)
|
| 2583 |
gen_op_sarl_T1_im(16);
|
| 2584 |
else
|
| 2585 |
gen_sxth(cpu_T[1]);
|
| 2586 |
gen_op_mul_T0_T1(); |
| 2587 |
} |
| 2588 |
|
| 2589 |
/* Return the mask of PSR bits set by a MSR instruction. */
|
| 2590 |
static uint32_t msr_mask(CPUState *env, DisasContext *s, int flags, int spsr) { |
| 2591 |
uint32_t mask; |
| 2592 |
|
| 2593 |
mask = 0;
|
| 2594 |
if (flags & (1 << 0)) |
| 2595 |
mask |= 0xff;
|
| 2596 |
if (flags & (1 << 1)) |
| 2597 |
mask |= 0xff00;
|
| 2598 |
if (flags & (1 << 2)) |
| 2599 |
mask |= 0xff0000;
|
| 2600 |
if (flags & (1 << 3)) |
| 2601 |
mask |= 0xff000000;
|
| 2602 |
|
| 2603 |
/* Mask out undefined bits. */
|
| 2604 |
mask &= ~CPSR_RESERVED; |
| 2605 |
if (!arm_feature(env, ARM_FEATURE_V6))
|
| 2606 |
mask &= ~(CPSR_E | CPSR_GE); |
| 2607 |
if (!arm_feature(env, ARM_FEATURE_THUMB2))
|
| 2608 |
mask &= ~CPSR_IT; |
| 2609 |
/* Mask out execution state bits. */
|
| 2610 |
if (!spsr)
|
| 2611 |
mask &= ~CPSR_EXEC; |
| 2612 |
/* Mask out privileged bits. */
|
| 2613 |
if (IS_USER(s))
|
| 2614 |
mask &= CPSR_USER; |
| 2615 |
return mask;
|
| 2616 |
} |
| 2617 |
|
| 2618 |
/* Returns nonzero if access to the PSR is not permitted. */
|
| 2619 |
static int gen_set_psr_T0(DisasContext *s, uint32_t mask, int spsr) |
| 2620 |
{
|
| 2621 |
if (spsr) {
|
| 2622 |
/* ??? This is also undefined in system mode. */
|
| 2623 |
if (IS_USER(s))
|
| 2624 |
return 1; |
| 2625 |
gen_op_movl_spsr_T0(mask); |
| 2626 |
} else {
|
| 2627 |
gen_op_movl_cpsr_T0(mask); |
| 2628 |
} |
| 2629 |
gen_lookup_tb(s); |
| 2630 |
return 0; |
| 2631 |
} |
| 2632 |
|
| 2633 |
/* Generate an old-style exception return. */
|
| 2634 |
static void gen_exception_return(DisasContext *s) |
| 2635 |
{
|
| 2636 |
gen_set_pc_T0(); |
| 2637 |
gen_op_movl_T0_spsr(); |
| 2638 |
gen_op_movl_cpsr_T0(0xffffffff);
|
| 2639 |
s->is_jmp = DISAS_UPDATE; |
| 2640 |
} |
| 2641 |
|
| 2642 |
/* Generate a v6 exception return. */
|
| 2643 |
static void gen_rfe(DisasContext *s) |
| 2644 |
{
|
| 2645 |
gen_op_movl_cpsr_T0(0xffffffff);
|
| 2646 |
gen_op_movl_T0_T2(); |
| 2647 |
gen_set_pc_T0(); |
| 2648 |
s->is_jmp = DISAS_UPDATE; |
| 2649 |
} |
| 2650 |
|
| 2651 |
static inline void |
| 2652 |
gen_set_condexec (DisasContext *s) |
| 2653 |
{
|
| 2654 |
if (s->condexec_mask) {
|
| 2655 |
gen_op_set_condexec((s->condexec_cond << 4) | (s->condexec_mask >> 1)); |
| 2656 |
} |
| 2657 |
} |
| 2658 |
|
| 2659 |
static void gen_nop_hint(DisasContext *s, int val) |
| 2660 |
{
|
| 2661 |
switch (val) {
|
| 2662 |
case 3: /* wfi */ |
| 2663 |
gen_op_movl_T0_im((long)s->pc);
|
| 2664 |
gen_set_pc_T0(); |
| 2665 |
s->is_jmp = DISAS_WFI; |
| 2666 |
break;
|
| 2667 |
case 2: /* wfe */ |
| 2668 |
case 4: /* sev */ |
| 2669 |
/* TODO: Implement SEV and WFE. May help SMP performance. */
|
| 2670 |
default: /* nop */ |
| 2671 |
break;
|
| 2672 |
} |
| 2673 |
} |
| 2674 |
|
| 2675 |
/* Neon shift by constant. The actual ops are the same as used for variable
|
| 2676 |
shifts. [OP][U][SIZE] */
|
| 2677 |
static GenOpFunc *gen_neon_shift_im[8][2][4] = { |
| 2678 |
{ /* 0 */ /* VSHR */
|
| 2679 |
{
|
| 2680 |
gen_op_neon_shl_u8, |
| 2681 |
gen_op_neon_shl_u16, |
| 2682 |
gen_op_neon_shl_u32, |
| 2683 |
gen_op_neon_shl_u64 |
| 2684 |
}, {
|
| 2685 |
gen_op_neon_shl_s8, |
| 2686 |
gen_op_neon_shl_s16, |
| 2687 |
gen_op_neon_shl_s32, |
| 2688 |
gen_op_neon_shl_s64 |
| 2689 |
} |
| 2690 |
}, { /* 1 */ /* VSRA */
|
| 2691 |
{
|
| 2692 |
gen_op_neon_shl_u8, |
| 2693 |
gen_op_neon_shl_u16, |
| 2694 |
gen_op_neon_shl_u32, |
| 2695 |
gen_op_neon_shl_u64 |
| 2696 |
}, {
|
| 2697 |
gen_op_neon_shl_s8, |
| 2698 |
gen_op_neon_shl_s16, |
| 2699 |
gen_op_neon_shl_s32, |
| 2700 |
gen_op_neon_shl_s64 |
| 2701 |
} |
| 2702 |
}, { /* 2 */ /* VRSHR */
|
| 2703 |
{
|
| 2704 |
gen_op_neon_rshl_u8, |
| 2705 |
gen_op_neon_rshl_u16, |
| 2706 |
gen_op_neon_rshl_u32, |
| 2707 |
gen_op_neon_rshl_u64 |
| 2708 |
}, {
|
| 2709 |
gen_op_neon_rshl_s8, |
| 2710 |
gen_op_neon_rshl_s16, |
| 2711 |
gen_op_neon_rshl_s32, |
| 2712 |
gen_op_neon_rshl_s64 |
| 2713 |
} |
| 2714 |
}, { /* 3 */ /* VRSRA */
|
| 2715 |
{
|
| 2716 |
gen_op_neon_rshl_u8, |
| 2717 |
gen_op_neon_rshl_u16, |
| 2718 |
gen_op_neon_rshl_u32, |
| 2719 |
gen_op_neon_rshl_u64 |
| 2720 |
}, {
|
| 2721 |
gen_op_neon_rshl_s8, |
| 2722 |
gen_op_neon_rshl_s16, |
| 2723 |
gen_op_neon_rshl_s32, |
| 2724 |
gen_op_neon_rshl_s64 |
| 2725 |
} |
| 2726 |
}, { /* 4 */
|
| 2727 |
{
|
| 2728 |
NULL, NULL, NULL, NULL |
| 2729 |
}, { /* VSRI */
|
| 2730 |
gen_op_neon_shl_u8, |
| 2731 |
gen_op_neon_shl_u16, |
| 2732 |
gen_op_neon_shl_u32, |
| 2733 |
gen_op_neon_shl_u64, |
| 2734 |
} |
| 2735 |
}, { /* 5 */
|
| 2736 |
{ /* VSHL */
|
| 2737 |
gen_op_neon_shl_u8, |
| 2738 |
gen_op_neon_shl_u16, |
| 2739 |
gen_op_neon_shl_u32, |
| 2740 |
gen_op_neon_shl_u64, |
| 2741 |
}, { /* VSLI */
|
| 2742 |
gen_op_neon_shl_u8, |
| 2743 |
gen_op_neon_shl_u16, |
| 2744 |
gen_op_neon_shl_u32, |
| 2745 |
gen_op_neon_shl_u64, |
| 2746 |
} |
| 2747 |
}, { /* 6 */ /* VQSHL */
|
| 2748 |
{
|
| 2749 |
gen_op_neon_qshl_u8, |
| 2750 |
gen_op_neon_qshl_u16, |
| 2751 |
gen_op_neon_qshl_u32, |
| 2752 |
gen_op_neon_qshl_u64 |
| 2753 |
}, {
|
| 2754 |
gen_op_neon_qshl_s8, |
| 2755 |
gen_op_neon_qshl_s16, |
| 2756 |
gen_op_neon_qshl_s32, |
| 2757 |
gen_op_neon_qshl_s64 |
| 2758 |
} |
| 2759 |
}, { /* 7 */ /* VQSHLU */
|
| 2760 |
{
|
| 2761 |
gen_op_neon_qshl_u8, |
| 2762 |
gen_op_neon_qshl_u16, |
| 2763 |
gen_op_neon_qshl_u32, |
| 2764 |
gen_op_neon_qshl_u64 |
| 2765 |
}, {
|
| 2766 |
gen_op_neon_qshl_u8, |
| 2767 |
gen_op_neon_qshl_u16, |
| 2768 |
gen_op_neon_qshl_u32, |
| 2769 |
gen_op_neon_qshl_u64 |
| 2770 |
} |
| 2771 |
} |
| 2772 |
}; |
| 2773 |
|
| 2774 |
/* [R][U][size - 1] */
|
| 2775 |
static GenOpFunc *gen_neon_shift_im_narrow[2][2][3] = { |
| 2776 |
{
|
| 2777 |
{
|
| 2778 |
gen_op_neon_shl_u16, |
| 2779 |
gen_op_neon_shl_u32, |
| 2780 |
gen_op_neon_shl_u64 |
| 2781 |
}, {
|
| 2782 |
gen_op_neon_shl_s16, |
| 2783 |
gen_op_neon_shl_s32, |
| 2784 |
gen_op_neon_shl_s64 |
| 2785 |
} |
| 2786 |
}, {
|
| 2787 |
{
|
| 2788 |
gen_op_neon_rshl_u16, |
| 2789 |
gen_op_neon_rshl_u32, |
| 2790 |
gen_op_neon_rshl_u64 |
| 2791 |
}, {
|
| 2792 |
gen_op_neon_rshl_s16, |
| 2793 |
gen_op_neon_rshl_s32, |
| 2794 |
gen_op_neon_rshl_s64 |
| 2795 |
} |
| 2796 |
} |
| 2797 |
}; |
| 2798 |
|
| 2799 |
static inline void |
| 2800 |
gen_op_neon_narrow_u32 () |
| 2801 |
{
|
| 2802 |
/* No-op. */
|
| 2803 |
} |
| 2804 |
|
| 2805 |
static GenOpFunc *gen_neon_narrow[3] = { |
| 2806 |
gen_op_neon_narrow_u8, |
| 2807 |
gen_op_neon_narrow_u16, |
| 2808 |
gen_op_neon_narrow_u32 |
| 2809 |
}; |
| 2810 |
|
| 2811 |
static GenOpFunc *gen_neon_narrow_satu[3] = { |
| 2812 |
gen_op_neon_narrow_sat_u8, |
| 2813 |
gen_op_neon_narrow_sat_u16, |
| 2814 |
gen_op_neon_narrow_sat_u32 |
| 2815 |
}; |
| 2816 |
|
| 2817 |
static GenOpFunc *gen_neon_narrow_sats[3] = { |
| 2818 |
gen_op_neon_narrow_sat_s8, |
| 2819 |
gen_op_neon_narrow_sat_s16, |
| 2820 |
gen_op_neon_narrow_sat_s32 |
| 2821 |
}; |
| 2822 |
|
| 2823 |
static inline int gen_neon_add(int size) |
| 2824 |
{
|
| 2825 |
switch (size) {
|
| 2826 |
case 0: gen_op_neon_add_u8(); break; |
| 2827 |
case 1: gen_op_neon_add_u16(); break; |
| 2828 |
case 2: gen_op_addl_T0_T1(); break; |
| 2829 |
default: return 1; |
| 2830 |
} |
| 2831 |
return 0; |
| 2832 |
} |
| 2833 |
|
| 2834 |
/* 32-bit pairwise ops end up the same as the elementsise versions. */
|
| 2835 |
#define gen_op_neon_pmax_s32 gen_op_neon_max_s32
|
| 2836 |
#define gen_op_neon_pmax_u32 gen_op_neon_max_u32
|
| 2837 |
#define gen_op_neon_pmin_s32 gen_op_neon_min_s32
|
| 2838 |
#define gen_op_neon_pmin_u32 gen_op_neon_min_u32
|
| 2839 |
|
| 2840 |
#define GEN_NEON_INTEGER_OP(name) do { \ |
| 2841 |
switch ((size << 1) | u) { \ |
| 2842 |
case 0: gen_op_neon_##name##_s8(); break; \ |
| 2843 |
case 1: gen_op_neon_##name##_u8(); break; \ |
| 2844 |
case 2: gen_op_neon_##name##_s16(); break; \ |
| 2845 |
case 3: gen_op_neon_##name##_u16(); break; \ |
| 2846 |
case 4: gen_op_neon_##name##_s32(); break; \ |
| 2847 |
case 5: gen_op_neon_##name##_u32(); break; \ |
| 2848 |
default: return 1; \ |
| 2849 |
}} while (0) |
| 2850 |
|
| 2851 |
static inline void |
| 2852 |
gen_neon_movl_scratch_T0(int scratch)
|
| 2853 |
{
|
| 2854 |
uint32_t offset; |
| 2855 |
|
| 2856 |
offset = offsetof(CPUARMState, vfp.scratch[scratch]); |
| 2857 |
gen_op_neon_setreg_T0(offset); |
| 2858 |
} |
| 2859 |
|
| 2860 |
static inline void |
| 2861 |
gen_neon_movl_scratch_T1(int scratch)
|
| 2862 |
{
|
| 2863 |
uint32_t offset; |
| 2864 |
|
| 2865 |
offset = offsetof(CPUARMState, vfp.scratch[scratch]); |
| 2866 |
gen_op_neon_setreg_T1(offset); |
| 2867 |
} |
| 2868 |
|
| 2869 |
static inline void |
| 2870 |
gen_neon_movl_T0_scratch(int scratch)
|
| 2871 |
{
|
| 2872 |
uint32_t offset; |
| 2873 |
|
| 2874 |
offset = offsetof(CPUARMState, vfp.scratch[scratch]); |
| 2875 |
gen_op_neon_getreg_T0(offset); |
| 2876 |
} |
| 2877 |
|
| 2878 |
static inline void |
| 2879 |
gen_neon_movl_T1_scratch(int scratch)
|
| 2880 |
{
|
| 2881 |
uint32_t offset; |
| 2882 |
|
| 2883 |
offset = offsetof(CPUARMState, vfp.scratch[scratch]); |
| 2884 |
gen_op_neon_getreg_T1(offset); |
| 2885 |
} |
| 2886 |
|
| 2887 |
static inline void gen_op_neon_widen_u32(void) |
| 2888 |
{
|
| 2889 |
gen_op_movl_T1_im(0);
|
| 2890 |
} |
| 2891 |
|
| 2892 |
static inline void gen_neon_get_scalar(int size, int reg) |
| 2893 |
{
|
| 2894 |
if (size == 1) { |
| 2895 |
NEON_GET_REG(T0, reg >> 1, reg & 1); |
| 2896 |
} else {
|
| 2897 |
NEON_GET_REG(T0, reg >> 2, (reg >> 1) & 1); |
| 2898 |
if (reg & 1) |
| 2899 |
gen_op_neon_dup_low16(); |
| 2900 |
else
|
| 2901 |
gen_op_neon_dup_high16(); |
| 2902 |
} |
| 2903 |
} |
| 2904 |
|
| 2905 |
static void gen_neon_unzip(int reg, int q, int tmp, int size) |
| 2906 |
{
|
| 2907 |
int n;
|
| 2908 |
|
| 2909 |
for (n = 0; n < q + 1; n += 2) { |
| 2910 |
NEON_GET_REG(T0, reg, n); |
| 2911 |
NEON_GET_REG(T0, reg, n + n); |
| 2912 |
switch (size) {
|
| 2913 |
case 0: gen_op_neon_unzip_u8(); break; |
| 2914 |
case 1: gen_op_neon_zip_u16(); break; /* zip and unzip are the same. */ |
| 2915 |
case 2: /* no-op */; break; |
| 2916 |
default: abort();
|
| 2917 |
} |
| 2918 |
gen_neon_movl_scratch_T0(tmp + n); |
| 2919 |
gen_neon_movl_scratch_T1(tmp + n + 1);
|
| 2920 |
} |
| 2921 |
} |
| 2922 |
|
| 2923 |
static struct { |
| 2924 |
int nregs;
|
| 2925 |
int interleave;
|
| 2926 |
int spacing;
|
| 2927 |
} neon_ls_element_type[11] = {
|
| 2928 |
{4, 4, 1},
|
| 2929 |
{4, 4, 2},
|
| 2930 |
{4, 1, 1},
|
| 2931 |
{4, 2, 1},
|
| 2932 |
{3, 3, 1},
|
| 2933 |
{3, 3, 2},
|
| 2934 |
{3, 1, 1},
|
| 2935 |
{1, 1, 1},
|
| 2936 |
{2, 2, 1},
|
| 2937 |
{2, 2, 2},
|
| 2938 |
{2, 1, 1}
|
| 2939 |
}; |
| 2940 |
|
| 2941 |
/* Translate a NEON load/store element instruction. Return nonzero if the
|
| 2942 |
instruction is invalid. */
|
| 2943 |
static int disas_neon_ls_insn(CPUState * env, DisasContext *s, uint32_t insn) |
| 2944 |
{
|
| 2945 |
int rd, rn, rm;
|
| 2946 |
int op;
|
| 2947 |
int nregs;
|
| 2948 |
int interleave;
|
| 2949 |
int stride;
|
| 2950 |
int size;
|
| 2951 |
int reg;
|
| 2952 |
int pass;
|
| 2953 |
int load;
|
| 2954 |
int shift;
|
| 2955 |
uint32_t mask; |
| 2956 |
int n;
|
| 2957 |
|
| 2958 |
if (!vfp_enabled(env))
|
| 2959 |
return 1; |
| 2960 |
VFP_DREG_D(rd, insn); |
| 2961 |
rn = (insn >> 16) & 0xf; |
| 2962 |
rm = insn & 0xf;
|
| 2963 |
load = (insn & (1 << 21)) != 0; |
| 2964 |
if ((insn & (1 << 23)) == 0) { |
| 2965 |
/* Load store all elements. */
|
| 2966 |
op = (insn >> 8) & 0xf; |
| 2967 |
size = (insn >> 6) & 3; |
| 2968 |
if (op > 10 || size == 3) |
| 2969 |
return 1; |
| 2970 |
nregs = neon_ls_element_type[op].nregs; |
| 2971 |
interleave = neon_ls_element_type[op].interleave; |
| 2972 |
gen_movl_T1_reg(s, rn); |
| 2973 |
stride = (1 << size) * interleave;
|
| 2974 |
for (reg = 0; reg < nregs; reg++) { |
| 2975 |
if (interleave > 2 || (interleave == 2 && nregs == 2)) { |
| 2976 |
gen_movl_T1_reg(s, rn); |
| 2977 |
gen_op_addl_T1_im((1 << size) * reg);
|
| 2978 |
} else if (interleave == 2 && nregs == 4 && reg == 2) { |
| 2979 |
gen_movl_T1_reg(s, rn); |
| 2980 |
gen_op_addl_T1_im(1 << size);
|
| 2981 |
} |
| 2982 |
for (pass = 0; pass < 2; pass++) { |
| 2983 |
if (size == 2) { |
| 2984 |
if (load) {
|
| 2985 |
gen_ldst(ldl, s); |
| 2986 |
NEON_SET_REG(T0, rd, pass); |
| 2987 |
} else {
|
| 2988 |
NEON_GET_REG(T0, rd, pass); |
| 2989 |
gen_ldst(stl, s); |
| 2990 |
} |
| 2991 |
gen_op_addl_T1_im(stride); |
| 2992 |
} else if (size == 1) { |
| 2993 |
if (load) {
|
| 2994 |
gen_ldst(lduw, s); |
| 2995 |
gen_op_addl_T1_im(stride); |
| 2996 |
gen_op_movl_T2_T0(); |
| 2997 |
gen_ldst(lduw, s); |
| 2998 |
gen_op_addl_T1_im(stride); |
| 2999 |
gen_op_neon_insert_elt(16, 0xffff); |
| 3000 |
NEON_SET_REG(T2, rd, pass); |
| 3001 |
} else {
|
| 3002 |
NEON_GET_REG(T2, rd, pass); |
| 3003 |
gen_op_movl_T0_T2(); |
| 3004 |
gen_ldst(stw, s); |
| 3005 |
gen_op_addl_T1_im(stride); |
| 3006 |
gen_op_neon_extract_elt(16, 0xffff0000); |
| 3007 |
gen_ldst(stw, s); |
| 3008 |
gen_op_addl_T1_im(stride); |
| 3009 |
} |
| 3010 |
} else /* size == 0 */ { |
| 3011 |
if (load) {
|
| 3012 |
mask = 0xff;
|
| 3013 |
for (n = 0; n < 4; n++) { |
| 3014 |
gen_ldst(ldub, s); |
| 3015 |
gen_op_addl_T1_im(stride); |
| 3016 |
if (n == 0) { |
| 3017 |
gen_op_movl_T2_T0(); |
| 3018 |
} else {
|
| 3019 |
gen_op_neon_insert_elt(n * 8, ~mask);
|
| 3020 |
} |
| 3021 |
mask <<= 8;
|
| 3022 |
} |
| 3023 |
NEON_SET_REG(T2, rd, pass); |
| 3024 |
} else {
|
| 3025 |
NEON_GET_REG(T2, rd, pass); |
| 3026 |
mask = 0xff;
|
| 3027 |
for (n = 0; n < 4; n++) { |
| 3028 |
if (n == 0) { |
| 3029 |
gen_op_movl_T0_T2(); |
| 3030 |
} else {
|
| 3031 |
gen_op_neon_extract_elt(n * 8, mask);
|
| 3032 |
} |
| 3033 |
gen_ldst(stb, s); |
| 3034 |
gen_op_addl_T1_im(stride); |
| 3035 |
mask <<= 8;
|
| 3036 |
} |
| 3037 |
} |
| 3038 |
} |
| 3039 |
} |
| 3040 |
rd += neon_ls_element_type[op].spacing; |
| 3041 |
} |
| 3042 |
stride = nregs * 8;
|
| 3043 |
} else {
|
| 3044 |
size = (insn >> 10) & 3; |
| 3045 |
if (size == 3) { |
| 3046 |
/* Load single element to all lanes. */
|
| 3047 |
if (!load)
|
| 3048 |
return 1; |
| 3049 |
size = (insn >> 6) & 3; |
| 3050 |
nregs = ((insn >> 8) & 3) + 1; |
| 3051 |
stride = (insn & (1 << 5)) ? 2 : 1; |
| 3052 |
gen_movl_T1_reg(s, rn); |
| 3053 |
for (reg = 0; reg < nregs; reg++) { |
| 3054 |
switch (size) {
|
| 3055 |
case 0: |
| 3056 |
gen_ldst(ldub, s); |
| 3057 |
gen_op_neon_dup_u8(0);
|
| 3058 |
break;
|
| 3059 |
case 1: |
| 3060 |
gen_ldst(lduw, s); |
| 3061 |
gen_op_neon_dup_low16(); |
| 3062 |
break;
|
| 3063 |
case 2: |
| 3064 |
gen_ldst(ldl, s); |
| 3065 |
break;
|
| 3066 |
case 3: |
| 3067 |
return 1; |
| 3068 |
} |
| 3069 |
gen_op_addl_T1_im(1 << size);
|
| 3070 |
NEON_SET_REG(T0, rd, 0);
|
| 3071 |
NEON_SET_REG(T0, rd, 1);
|
| 3072 |
rd += stride; |
| 3073 |
} |
| 3074 |
stride = (1 << size) * nregs;
|
| 3075 |
} else {
|
| 3076 |
/* Single element. */
|
| 3077 |
pass = (insn >> 7) & 1; |
| 3078 |
switch (size) {
|
| 3079 |
case 0: |
| 3080 |
shift = ((insn >> 5) & 3) * 8; |
| 3081 |
mask = 0xff << shift;
|
| 3082 |
stride = 1;
|
| 3083 |
break;
|
| 3084 |
case 1: |
| 3085 |
shift = ((insn >> 6) & 1) * 16; |
| 3086 |
mask = shift ? 0xffff0000 : 0xffff; |
| 3087 |
stride = (insn & (1 << 5)) ? 2 : 1; |
| 3088 |
break;
|
| 3089 |
case 2: |
| 3090 |
shift = 0;
|
| 3091 |
mask = 0xffffffff;
|
| 3092 |
stride = (insn & (1 << 6)) ? 2 : 1; |
| 3093 |
break;
|
| 3094 |
default:
|
| 3095 |
abort(); |
| 3096 |
} |
| 3097 |
nregs = ((insn >> 8) & 3) + 1; |
| 3098 |
gen_movl_T1_reg(s, rn); |
| 3099 |
for (reg = 0; reg < nregs; reg++) { |
| 3100 |
if (load) {
|
| 3101 |
if (size != 2) { |
| 3102 |
NEON_GET_REG(T2, rd, pass); |
| 3103 |
} |
| 3104 |
switch (size) {
|
| 3105 |
case 0: |
| 3106 |
gen_ldst(ldub, s); |
| 3107 |
break;
|
| 3108 |
case 1: |
| 3109 |
gen_ldst(lduw, s); |
| 3110 |
break;
|
| 3111 |
case 2: |
| 3112 |
gen_ldst(ldl, s); |
| 3113 |
NEON_SET_REG(T0, rd, pass); |
| 3114 |
break;
|
| 3115 |
} |
| 3116 |
if (size != 2) { |
| 3117 |
gen_op_neon_insert_elt(shift, ~mask); |
| 3118 |
NEON_SET_REG(T0, rd, pass); |
| 3119 |
} |
| 3120 |
} else { /* Store */ |
| 3121 |
if (size == 2) { |
| 3122 |
NEON_GET_REG(T0, rd, pass); |
| 3123 |
} else {
|
| 3124 |
NEON_GET_REG(T2, rd, pass); |
| 3125 |
gen_op_neon_extract_elt(shift, mask); |
| 3126 |
} |
| 3127 |
switch (size) {
|
| 3128 |
case 0: |
| 3129 |
gen_ldst(stb, s); |
| 3130 |
break;
|
| 3131 |
case 1: |
| 3132 |
gen_ldst(stw, s); |
| 3133 |
break;
|
| 3134 |
case 2: |
| 3135 |
gen_ldst(stl, s); |
| 3136 |
break;
|
| 3137 |
} |
| 3138 |
} |
| 3139 |
rd += stride; |
| 3140 |
gen_op_addl_T1_im(1 << size);
|
| 3141 |
} |
| 3142 |
stride = nregs * (1 << size);
|
| 3143 |
} |
| 3144 |
} |
| 3145 |
if (rm != 15) { |
| 3146 |
TCGv base; |
| 3147 |
|
| 3148 |
base = load_reg(s, rn); |
| 3149 |
if (rm == 13) { |
| 3150 |
tcg_gen_addi_i32(base, base, stride); |
| 3151 |
} else {
|
| 3152 |
TCGv index; |
| 3153 |
index = load_reg(s, rm); |
| 3154 |
tcg_gen_add_i32(base, base, index); |
| 3155 |
dead_tmp(index); |
| 3156 |
} |
| 3157 |
store_reg(s, rn, base); |
| 3158 |
} |
| 3159 |
return 0; |
| 3160 |
} |
| 3161 |
|
| 3162 |
/* Translate a NEON data processing instruction. Return nonzero if the
|
| 3163 |
instruction is invalid.
|
| 3164 |
In general we process vectors in 32-bit chunks. This means we can reuse
|
| 3165 |
some of the scalar ops, and hopefully the code generated for 32-bit
|
| 3166 |
hosts won't be too awful. The downside is that the few 64-bit operations
|
| 3167 |
(mainly shifts) get complicated. */
|
| 3168 |
|
| 3169 |
static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn) |
| 3170 |
{
|
| 3171 |
int op;
|
| 3172 |
int q;
|
| 3173 |
int rd, rn, rm;
|
| 3174 |
int size;
|
| 3175 |
int shift;
|
| 3176 |
int pass;
|
| 3177 |
int count;
|
| 3178 |
int pairwise;
|
| 3179 |
int u;
|
| 3180 |
int n;
|
| 3181 |
uint32_t imm; |
| 3182 |
|
| 3183 |
if (!vfp_enabled(env))
|
| 3184 |
return 1; |
| 3185 |
q = (insn & (1 << 6)) != 0; |
| 3186 |
u = (insn >> 24) & 1; |
| 3187 |
VFP_DREG_D(rd, insn); |
| 3188 |
VFP_DREG_N(rn, insn); |
| 3189 |
VFP_DREG_M(rm, insn); |
| 3190 |
size = (insn >> 20) & 3; |
| 3191 |
if ((insn & (1 << 23)) == 0) { |
| 3192 |
/* Three register same length. */
|
| 3193 |
op = ((insn >> 7) & 0x1e) | ((insn >> 4) & 1); |
| 3194 |
if (size == 3 && (op == 1 || op == 5 || op == 16)) { |
| 3195 |
for (pass = 0; pass < (q ? 2 : 1); pass++) { |
| 3196 |
NEON_GET_REG(T0, rm, pass * 2);
|
| 3197 |
NEON_GET_REG(T1, rm, pass * 2 + 1); |
| 3198 |
gen_neon_movl_scratch_T0(0);
|
| 3199 |
gen_neon_movl_scratch_T1(1);
|
| 3200 |
NEON_GET_REG(T0, rn, pass * 2);
|
| 3201 |
NEON_GET_REG(T1, rn, pass * 2 + 1); |
| 3202 |
switch (op) {
|
| 3203 |
case 1: /* VQADD */ |
| 3204 |
if (u) {
|
| 3205 |
gen_op_neon_addl_saturate_u64(); |
| 3206 |
} else {
|
| 3207 |
gen_op_neon_addl_saturate_s64(); |
| 3208 |
} |
| 3209 |
break;
|
| 3210 |
case 5: /* VQSUB */ |
| 3211 |
if (u) {
|
| 3212 |
gen_op_neon_subl_saturate_u64(); |
| 3213 |
} else {
|
| 3214 |
gen_op_neon_subl_saturate_s64(); |
| 3215 |
} |
| 3216 |
break;
|
| 3217 |
case 16: |
| 3218 |
if (u) {
|
| 3219 |
gen_op_neon_subl_u64(); |
| 3220 |
} else {
|
| 3221 |
gen_op_neon_addl_u64(); |
| 3222 |
} |
| 3223 |
break;
|
| 3224 |
default:
|
| 3225 |
abort(); |
| 3226 |
} |
| 3227 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 3228 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 3229 |
} |
| 3230 |
return 0; |
| 3231 |
} |
| 3232 |
switch (op) {
|
| 3233 |
case 8: /* VSHL */ |
| 3234 |
case 9: /* VQSHL */ |
| 3235 |
case 10: /* VRSHL */ |
| 3236 |
case 11: /* VQSHL */ |
| 3237 |
/* Shift operations have Rn and Rm reversed. */
|
| 3238 |
{
|
| 3239 |
int tmp;
|
| 3240 |
tmp = rn; |
| 3241 |
rn = rm; |
| 3242 |
rm = tmp; |
| 3243 |
pairwise = 0;
|
| 3244 |
} |
| 3245 |
break;
|
| 3246 |
case 20: /* VPMAX */ |
| 3247 |
case 21: /* VPMIN */ |
| 3248 |
case 23: /* VPADD */ |
| 3249 |
pairwise = 1;
|
| 3250 |
break;
|
| 3251 |
case 26: /* VPADD (float) */ |
| 3252 |
pairwise = (u && size < 2);
|
| 3253 |
break;
|
| 3254 |
case 30: /* VPMIN/VPMAX (float) */ |
| 3255 |
pairwise = u; |
| 3256 |
break;
|
| 3257 |
default:
|
| 3258 |
pairwise = 0;
|
| 3259 |
break;
|
| 3260 |
} |
| 3261 |
for (pass = 0; pass < (q ? 4 : 2); pass++) { |
| 3262 |
|
| 3263 |
if (pairwise) {
|
| 3264 |
/* Pairwise. */
|
| 3265 |
if (q)
|
| 3266 |
n = (pass & 1) * 2; |
| 3267 |
else
|
| 3268 |
n = 0;
|
| 3269 |
if (pass < q + 1) { |
| 3270 |
NEON_GET_REG(T0, rn, n); |
| 3271 |
NEON_GET_REG(T1, rn, n + 1);
|
| 3272 |
} else {
|
| 3273 |
NEON_GET_REG(T0, rm, n); |
| 3274 |
NEON_GET_REG(T1, rm, n + 1);
|
| 3275 |
} |
| 3276 |
} else {
|
| 3277 |
/* Elementwise. */
|
| 3278 |
NEON_GET_REG(T0, rn, pass); |
| 3279 |
NEON_GET_REG(T1, rm, pass); |
| 3280 |
} |
| 3281 |
switch (op) {
|
| 3282 |
case 0: /* VHADD */ |
| 3283 |
GEN_NEON_INTEGER_OP(hadd); |
| 3284 |
break;
|
| 3285 |
case 1: /* VQADD */ |
| 3286 |
switch (size << 1| u) { |
| 3287 |
case 0: gen_op_neon_qadd_s8(); break; |
| 3288 |
case 1: gen_op_neon_qadd_u8(); break; |
| 3289 |
case 2: gen_op_neon_qadd_s16(); break; |
| 3290 |
case 3: gen_op_neon_qadd_u16(); break; |
| 3291 |
case 4: gen_op_addl_T0_T1_saturate(); break; |
| 3292 |
case 5: gen_op_addl_T0_T1_usaturate(); break; |
| 3293 |
default: abort();
|
| 3294 |
} |
| 3295 |
break;
|
| 3296 |
case 2: /* VRHADD */ |
| 3297 |
GEN_NEON_INTEGER_OP(rhadd); |
| 3298 |
break;
|
| 3299 |
case 3: /* Logic ops. */ |
| 3300 |
switch ((u << 2) | size) { |
| 3301 |
case 0: /* VAND */ |
| 3302 |
gen_op_andl_T0_T1(); |
| 3303 |
break;
|
| 3304 |
case 1: /* BIC */ |
| 3305 |
gen_op_bicl_T0_T1(); |
| 3306 |
break;
|
| 3307 |
case 2: /* VORR */ |
| 3308 |
gen_op_orl_T0_T1(); |
| 3309 |
break;
|
| 3310 |
case 3: /* VORN */ |
| 3311 |
gen_op_notl_T1(); |
| 3312 |
gen_op_orl_T0_T1(); |
| 3313 |
break;
|
| 3314 |
case 4: /* VEOR */ |
| 3315 |
gen_op_xorl_T0_T1(); |
| 3316 |
break;
|
| 3317 |
case 5: /* VBSL */ |
| 3318 |
NEON_GET_REG(T2, rd, pass); |
| 3319 |
gen_op_neon_bsl(); |
| 3320 |
break;
|
| 3321 |
case 6: /* VBIT */ |
| 3322 |
NEON_GET_REG(T2, rd, pass); |
| 3323 |
gen_op_neon_bit(); |
| 3324 |
break;
|
| 3325 |
case 7: /* VBIF */ |
| 3326 |
NEON_GET_REG(T2, rd, pass); |
| 3327 |
gen_op_neon_bif(); |
| 3328 |
break;
|
| 3329 |
} |
| 3330 |
break;
|
| 3331 |
case 4: /* VHSUB */ |
| 3332 |
GEN_NEON_INTEGER_OP(hsub); |
| 3333 |
break;
|
| 3334 |
case 5: /* VQSUB */ |
| 3335 |
switch ((size << 1) | u) { |
| 3336 |
case 0: gen_op_neon_qsub_s8(); break; |
| 3337 |
case 1: gen_op_neon_qsub_u8(); break; |
| 3338 |
case 2: gen_op_neon_qsub_s16(); break; |
| 3339 |
case 3: gen_op_neon_qsub_u16(); break; |
| 3340 |
case 4: gen_op_subl_T0_T1_saturate(); break; |
| 3341 |
case 5: gen_op_subl_T0_T1_usaturate(); break; |
| 3342 |
default: abort();
|
| 3343 |
} |
| 3344 |
break;
|
| 3345 |
case 6: /* VCGT */ |
| 3346 |
GEN_NEON_INTEGER_OP(cgt); |
| 3347 |
break;
|
| 3348 |
case 7: /* VCGE */ |
| 3349 |
GEN_NEON_INTEGER_OP(cge); |
| 3350 |
break;
|
| 3351 |
case 8: /* VSHL */ |
| 3352 |
switch ((size << 1) | u) { |
| 3353 |
case 0: gen_op_neon_shl_s8(); break; |
| 3354 |
case 1: gen_op_neon_shl_u8(); break; |
| 3355 |
case 2: gen_op_neon_shl_s16(); break; |
| 3356 |
case 3: gen_op_neon_shl_u16(); break; |
| 3357 |
case 4: gen_op_neon_shl_s32(); break; |
| 3358 |
case 5: gen_op_neon_shl_u32(); break; |
| 3359 |
#if 0
|
| 3360 |
/* ??? Implementing these is tricky because the vector ops work
|
| 3361 |
on 32-bit pieces. */
|
| 3362 |
case 6: gen_op_neon_shl_s64(); break;
|
| 3363 |
case 7: gen_op_neon_shl_u64(); break;
|
| 3364 |
#else
|
| 3365 |
case 6: case 7: cpu_abort(env, "VSHL.64 not implemented"); |
| 3366 |
#endif
|
| 3367 |
} |
| 3368 |
break;
|
| 3369 |
case 9: /* VQSHL */ |
| 3370 |
switch ((size << 1) | u) { |
| 3371 |
case 0: gen_op_neon_qshl_s8(); break; |
| 3372 |
case 1: gen_op_neon_qshl_u8(); break; |
| 3373 |
case 2: gen_op_neon_qshl_s16(); break; |
| 3374 |
case 3: gen_op_neon_qshl_u16(); break; |
| 3375 |
case 4: gen_op_neon_qshl_s32(); break; |
| 3376 |
case 5: gen_op_neon_qshl_u32(); break; |
| 3377 |
#if 0
|
| 3378 |
/* ??? Implementing these is tricky because the vector ops work
|
| 3379 |
on 32-bit pieces. */
|
| 3380 |
case 6: gen_op_neon_qshl_s64(); break;
|
| 3381 |
case 7: gen_op_neon_qshl_u64(); break;
|
| 3382 |
#else
|
| 3383 |
case 6: case 7: cpu_abort(env, "VQSHL.64 not implemented"); |
| 3384 |
#endif
|
| 3385 |
} |
| 3386 |
break;
|
| 3387 |
case 10: /* VRSHL */ |
| 3388 |
switch ((size << 1) | u) { |
| 3389 |
case 0: gen_op_neon_rshl_s8(); break; |
| 3390 |
case 1: gen_op_neon_rshl_u8(); break; |
| 3391 |
case 2: gen_op_neon_rshl_s16(); break; |
| 3392 |
case 3: gen_op_neon_rshl_u16(); break; |
| 3393 |
case 4: gen_op_neon_rshl_s32(); break; |
| 3394 |
case 5: gen_op_neon_rshl_u32(); break; |
| 3395 |
#if 0
|
| 3396 |
/* ??? Implementing these is tricky because the vector ops work
|
| 3397 |
on 32-bit pieces. */
|
| 3398 |
case 6: gen_op_neon_rshl_s64(); break;
|
| 3399 |
case 7: gen_op_neon_rshl_u64(); break;
|
| 3400 |
#else
|
| 3401 |
case 6: case 7: cpu_abort(env, "VRSHL.64 not implemented"); |
| 3402 |
#endif
|
| 3403 |
} |
| 3404 |
break;
|
| 3405 |
case 11: /* VQRSHL */ |
| 3406 |
switch ((size << 1) | u) { |
| 3407 |
case 0: gen_op_neon_qrshl_s8(); break; |
| 3408 |
case 1: gen_op_neon_qrshl_u8(); break; |
| 3409 |
case 2: gen_op_neon_qrshl_s16(); break; |
| 3410 |
case 3: gen_op_neon_qrshl_u16(); break; |
| 3411 |
case 4: gen_op_neon_qrshl_s32(); break; |
| 3412 |
case 5: gen_op_neon_qrshl_u32(); break; |
| 3413 |
#if 0
|
| 3414 |
/* ??? Implementing these is tricky because the vector ops work
|
| 3415 |
on 32-bit pieces. */
|
| 3416 |
case 6: gen_op_neon_qrshl_s64(); break;
|
| 3417 |
case 7: gen_op_neon_qrshl_u64(); break;
|
| 3418 |
#else
|
| 3419 |
case 6: case 7: cpu_abort(env, "VQRSHL.64 not implemented"); |
| 3420 |
#endif
|
| 3421 |
} |
| 3422 |
break;
|
| 3423 |
case 12: /* VMAX */ |
| 3424 |
GEN_NEON_INTEGER_OP(max); |
| 3425 |
break;
|
| 3426 |
case 13: /* VMIN */ |
| 3427 |
GEN_NEON_INTEGER_OP(min); |
| 3428 |
break;
|
| 3429 |
case 14: /* VABD */ |
| 3430 |
GEN_NEON_INTEGER_OP(abd); |
| 3431 |
break;
|
| 3432 |
case 15: /* VABA */ |
| 3433 |
GEN_NEON_INTEGER_OP(abd); |
| 3434 |
NEON_GET_REG(T1, rd, pass); |
| 3435 |
gen_neon_add(size); |
| 3436 |
break;
|
| 3437 |
case 16: |
| 3438 |
if (!u) { /* VADD */ |
| 3439 |
if (gen_neon_add(size))
|
| 3440 |
return 1; |
| 3441 |
} else { /* VSUB */ |
| 3442 |
switch (size) {
|
| 3443 |
case 0: gen_op_neon_sub_u8(); break; |
| 3444 |
case 1: gen_op_neon_sub_u16(); break; |
| 3445 |
case 2: gen_op_subl_T0_T1(); break; |
| 3446 |
default: return 1; |
| 3447 |
} |
| 3448 |
} |
| 3449 |
break;
|
| 3450 |
case 17: |
| 3451 |
if (!u) { /* VTST */ |
| 3452 |
switch (size) {
|
| 3453 |
case 0: gen_op_neon_tst_u8(); break; |
| 3454 |
case 1: gen_op_neon_tst_u16(); break; |
| 3455 |
case 2: gen_op_neon_tst_u32(); break; |
| 3456 |
default: return 1; |
| 3457 |
} |
| 3458 |
} else { /* VCEQ */ |
| 3459 |
switch (size) {
|
| 3460 |
case 0: gen_op_neon_ceq_u8(); break; |
| 3461 |
case 1: gen_op_neon_ceq_u16(); break; |
| 3462 |
case 2: gen_op_neon_ceq_u32(); break; |
| 3463 |
default: return 1; |
| 3464 |
} |
| 3465 |
} |
| 3466 |
break;
|
| 3467 |
case 18: /* Multiply. */ |
| 3468 |
switch (size) {
|
| 3469 |
case 0: gen_op_neon_mul_u8(); break; |
| 3470 |
case 1: gen_op_neon_mul_u16(); break; |
| 3471 |
case 2: gen_op_mul_T0_T1(); break; |
| 3472 |
default: return 1; |
| 3473 |
} |
| 3474 |
NEON_GET_REG(T1, rd, pass); |
| 3475 |
if (u) { /* VMLS */ |
| 3476 |
switch (size) {
|
| 3477 |
case 0: gen_op_neon_rsb_u8(); break; |
| 3478 |
case 1: gen_op_neon_rsb_u16(); break; |
| 3479 |
case 2: gen_op_rsbl_T0_T1(); break; |
| 3480 |
default: return 1; |
| 3481 |
} |
| 3482 |
} else { /* VMLA */ |
| 3483 |
gen_neon_add(size); |
| 3484 |
} |
| 3485 |
break;
|
| 3486 |
case 19: /* VMUL */ |
| 3487 |
if (u) { /* polynomial */ |
| 3488 |
gen_op_neon_mul_p8(); |
| 3489 |
} else { /* Integer */ |
| 3490 |
switch (size) {
|
| 3491 |
case 0: gen_op_neon_mul_u8(); break; |
| 3492 |
case 1: gen_op_neon_mul_u16(); break; |
| 3493 |
case 2: gen_op_mul_T0_T1(); break; |
| 3494 |
default: return 1; |
| 3495 |
} |
| 3496 |
} |
| 3497 |
break;
|
| 3498 |
case 20: /* VPMAX */ |
| 3499 |
GEN_NEON_INTEGER_OP(pmax); |
| 3500 |
break;
|
| 3501 |
case 21: /* VPMIN */ |
| 3502 |
GEN_NEON_INTEGER_OP(pmin); |
| 3503 |
break;
|
| 3504 |
case 22: /* Hultiply high. */ |
| 3505 |
if (!u) { /* VQDMULH */ |
| 3506 |
switch (size) {
|
| 3507 |
case 1: gen_op_neon_qdmulh_s16(); break; |
| 3508 |
case 2: gen_op_neon_qdmulh_s32(); break; |
| 3509 |
default: return 1; |
| 3510 |
} |
| 3511 |
} else { /* VQRDHMUL */ |
| 3512 |
switch (size) {
|
| 3513 |
case 1: gen_op_neon_qrdmulh_s16(); break; |
| 3514 |
case 2: gen_op_neon_qrdmulh_s32(); break; |
| 3515 |
default: return 1; |
| 3516 |
} |
| 3517 |
} |
| 3518 |
break;
|
| 3519 |
case 23: /* VPADD */ |
| 3520 |
if (u)
|
| 3521 |
return 1; |
| 3522 |
switch (size) {
|
| 3523 |
case 0: gen_op_neon_padd_u8(); break; |
| 3524 |
case 1: gen_op_neon_padd_u16(); break; |
| 3525 |
case 2: gen_op_addl_T0_T1(); break; |
| 3526 |
default: return 1; |
| 3527 |
} |
| 3528 |
break;
|
| 3529 |
case 26: /* Floating point arithnetic. */ |
| 3530 |
switch ((u << 2) | size) { |
| 3531 |
case 0: /* VADD */ |
| 3532 |
gen_op_neon_add_f32(); |
| 3533 |
break;
|
| 3534 |
case 2: /* VSUB */ |
| 3535 |
gen_op_neon_sub_f32(); |
| 3536 |
break;
|
| 3537 |
case 4: /* VPADD */ |
| 3538 |
gen_op_neon_add_f32(); |
| 3539 |
break;
|
| 3540 |
case 6: /* VABD */ |
| 3541 |
gen_op_neon_abd_f32(); |
| 3542 |
break;
|
| 3543 |
default:
|
| 3544 |
return 1; |
| 3545 |
} |
| 3546 |
break;
|
| 3547 |
case 27: /* Float multiply. */ |
| 3548 |
gen_op_neon_mul_f32(); |
| 3549 |
if (!u) {
|
| 3550 |
NEON_GET_REG(T1, rd, pass); |
| 3551 |
if (size == 0) { |
| 3552 |
gen_op_neon_add_f32(); |
| 3553 |
} else {
|
| 3554 |
gen_op_neon_rsb_f32(); |
| 3555 |
} |
| 3556 |
} |
| 3557 |
break;
|
| 3558 |
case 28: /* Float compare. */ |
| 3559 |
if (!u) {
|
| 3560 |
gen_op_neon_ceq_f32(); |
| 3561 |
} else {
|
| 3562 |
if (size == 0) |
| 3563 |
gen_op_neon_cge_f32(); |
| 3564 |
else
|
| 3565 |
gen_op_neon_cgt_f32(); |
| 3566 |
} |
| 3567 |
break;
|
| 3568 |
case 29: /* Float compare absolute. */ |
| 3569 |
if (!u)
|
| 3570 |
return 1; |
| 3571 |
if (size == 0) |
| 3572 |
gen_op_neon_acge_f32(); |
| 3573 |
else
|
| 3574 |
gen_op_neon_acgt_f32(); |
| 3575 |
break;
|
| 3576 |
case 30: /* Float min/max. */ |
| 3577 |
if (size == 0) |
| 3578 |
gen_op_neon_max_f32(); |
| 3579 |
else
|
| 3580 |
gen_op_neon_min_f32(); |
| 3581 |
break;
|
| 3582 |
case 31: |
| 3583 |
if (size == 0) |
| 3584 |
gen_op_neon_recps_f32(); |
| 3585 |
else
|
| 3586 |
gen_op_neon_rsqrts_f32(); |
| 3587 |
break;
|
| 3588 |
default:
|
| 3589 |
abort(); |
| 3590 |
} |
| 3591 |
/* Save the result. For elementwise operations we can put it
|
| 3592 |
straight into the destination register. For pairwise operations
|
| 3593 |
we have to be careful to avoid clobbering the source operands. */
|
| 3594 |
if (pairwise && rd == rm) {
|
| 3595 |
gen_neon_movl_scratch_T0(pass); |
| 3596 |
} else {
|
| 3597 |
NEON_SET_REG(T0, rd, pass); |
| 3598 |
} |
| 3599 |
|
| 3600 |
} /* for pass */
|
| 3601 |
if (pairwise && rd == rm) {
|
| 3602 |
for (pass = 0; pass < (q ? 4 : 2); pass++) { |
| 3603 |
gen_neon_movl_T0_scratch(pass); |
| 3604 |
NEON_SET_REG(T0, rd, pass); |
| 3605 |
} |
| 3606 |
} |
| 3607 |
} else if (insn & (1 << 4)) { |
| 3608 |
if ((insn & 0x00380080) != 0) { |
| 3609 |
/* Two registers and shift. */
|
| 3610 |
op = (insn >> 8) & 0xf; |
| 3611 |
if (insn & (1 << 7)) { |
| 3612 |
/* 64-bit shift. */
|
| 3613 |
size = 3;
|
| 3614 |
} else {
|
| 3615 |
size = 2;
|
| 3616 |
while ((insn & (1 << (size + 19))) == 0) |
| 3617 |
size--; |
| 3618 |
} |
| 3619 |
shift = (insn >> 16) & ((1 << (3 + size)) - 1); |
| 3620 |
/* To avoid excessive dumplication of ops we implement shift
|
| 3621 |
by immediate using the variable shift operations. */
|
| 3622 |
if (op < 8) { |
| 3623 |
/* Shift by immediate:
|
| 3624 |
VSHR, VSRA, VRSHR, VRSRA, VSRI, VSHL, VQSHL, VQSHLU. */
|
| 3625 |
/* Right shifts are encoded as N - shift, where N is the
|
| 3626 |
element size in bits. */
|
| 3627 |
if (op <= 4) |
| 3628 |
shift = shift - (1 << (size + 3)); |
| 3629 |
else
|
| 3630 |
shift++; |
| 3631 |
if (size == 3) { |
| 3632 |
count = q + 1;
|
| 3633 |
} else {
|
| 3634 |
count = q ? 4: 2; |
| 3635 |
} |
| 3636 |
switch (size) {
|
| 3637 |
case 0: |
| 3638 |
imm = (uint8_t) shift; |
| 3639 |
imm |= imm << 8;
|
| 3640 |
imm |= imm << 16;
|
| 3641 |
break;
|
| 3642 |
case 1: |
| 3643 |
imm = (uint16_t) shift; |
| 3644 |
imm |= imm << 16;
|
| 3645 |
break;
|
| 3646 |
case 2: |
| 3647 |
case 3: |
| 3648 |
imm = shift; |
| 3649 |
break;
|
| 3650 |
default:
|
| 3651 |
abort(); |
| 3652 |
} |
| 3653 |
|
| 3654 |
for (pass = 0; pass < count; pass++) { |
| 3655 |
if (size < 3) { |
| 3656 |
/* Operands in T0 and T1. */
|
| 3657 |
gen_op_movl_T1_im(imm); |
| 3658 |
NEON_GET_REG(T0, rm, pass); |
| 3659 |
} else {
|
| 3660 |
/* Operands in {T0, T1} and env->vfp.scratch. */
|
| 3661 |
gen_op_movl_T0_im(imm); |
| 3662 |
gen_neon_movl_scratch_T0(0);
|
| 3663 |
gen_op_movl_T0_im((int32_t)imm >> 31);
|
| 3664 |
gen_neon_movl_scratch_T0(1);
|
| 3665 |
NEON_GET_REG(T0, rm, pass * 2);
|
| 3666 |
NEON_GET_REG(T1, rm, pass * 2 + 1); |
| 3667 |
} |
| 3668 |
|
| 3669 |
if (gen_neon_shift_im[op][u][size] == NULL) |
| 3670 |
return 1; |
| 3671 |
gen_neon_shift_im[op][u][size](); |
| 3672 |
|
| 3673 |
if (op == 1 || op == 3) { |
| 3674 |
/* Accumulate. */
|
| 3675 |
if (size == 3) { |
| 3676 |
gen_neon_movl_scratch_T0(0);
|
| 3677 |
gen_neon_movl_scratch_T1(1);
|
| 3678 |
NEON_GET_REG(T0, rd, pass * 2);
|
| 3679 |
NEON_GET_REG(T1, rd, pass * 2 + 1); |
| 3680 |
gen_op_neon_addl_u64(); |
| 3681 |
} else {
|
| 3682 |
NEON_GET_REG(T1, rd, pass); |
| 3683 |
gen_neon_add(size); |
| 3684 |
} |
| 3685 |
} else if (op == 4 || (op == 5 && u)) { |
| 3686 |
/* Insert */
|
| 3687 |
if (size == 3) { |
| 3688 |
cpu_abort(env, "VS[LR]I.64 not implemented");
|
| 3689 |
} |
| 3690 |
switch (size) {
|
| 3691 |
case 0: |
| 3692 |
if (op == 4) |
| 3693 |
imm = 0xff >> -shift;
|
| 3694 |
else
|
| 3695 |
imm = (uint8_t)(0xff << shift);
|
| 3696 |
imm |= imm << 8;
|
| 3697 |
imm |= imm << 16;
|
| 3698 |
break;
|
| 3699 |
case 1: |
| 3700 |
if (op == 4) |
| 3701 |
imm = 0xffff >> -shift;
|
| 3702 |
else
|
| 3703 |
imm = (uint16_t)(0xffff << shift);
|
| 3704 |
imm |= imm << 16;
|
| 3705 |
break;
|
| 3706 |
case 2: |
| 3707 |
if (op == 4) |
| 3708 |
imm = 0xffffffffu >> -shift;
|
| 3709 |
else
|
| 3710 |
imm = 0xffffffffu << shift;
|
| 3711 |
break;
|
| 3712 |
default:
|
| 3713 |
abort(); |
| 3714 |
} |
| 3715 |
NEON_GET_REG(T1, rd, pass); |
| 3716 |
gen_op_movl_T2_im(imm); |
| 3717 |
gen_op_neon_bsl(); |
| 3718 |
} |
| 3719 |
if (size == 3) { |
| 3720 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 3721 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 3722 |
} else {
|
| 3723 |
NEON_SET_REG(T0, rd, pass); |
| 3724 |
} |
| 3725 |
} /* for pass */
|
| 3726 |
} else if (op < 10) { |
| 3727 |
/* Shift by immedaiate and narrow:
|
| 3728 |
VSHRN, VRSHRN, VQSHRN, VQRSHRN. */
|
| 3729 |
shift = shift - (1 << (size + 3)); |
| 3730 |
size++; |
| 3731 |
if (size == 3) { |
| 3732 |
count = q + 1;
|
| 3733 |
} else {
|
| 3734 |
count = q ? 4: 2; |
| 3735 |
} |
| 3736 |
switch (size) {
|
| 3737 |
case 1: |
| 3738 |
imm = (uint16_t) shift; |
| 3739 |
imm |= imm << 16;
|
| 3740 |
break;
|
| 3741 |
case 2: |
| 3742 |
case 3: |
| 3743 |
imm = shift; |
| 3744 |
break;
|
| 3745 |
default:
|
| 3746 |
abort(); |
| 3747 |
} |
| 3748 |
|
| 3749 |
/* Processing MSB first means we need to do less shuffling at
|
| 3750 |
the end. */
|
| 3751 |
for (pass = count - 1; pass >= 0; pass--) { |
| 3752 |
/* Avoid clobbering the second operand before it has been
|
| 3753 |
written. */
|
| 3754 |
n = pass; |
| 3755 |
if (rd == rm)
|
| 3756 |
n ^= (count - 1);
|
| 3757 |
else
|
| 3758 |
n = pass; |
| 3759 |
|
| 3760 |
if (size < 3) { |
| 3761 |
/* Operands in T0 and T1. */
|
| 3762 |
gen_op_movl_T1_im(imm); |
| 3763 |
NEON_GET_REG(T0, rm, n); |
| 3764 |
} else {
|
| 3765 |
/* Operands in {T0, T1} and env->vfp.scratch. */
|
| 3766 |
gen_op_movl_T0_im(imm); |
| 3767 |
gen_neon_movl_scratch_T0(0);
|
| 3768 |
gen_op_movl_T0_im((int32_t)imm >> 31);
|
| 3769 |
gen_neon_movl_scratch_T0(1);
|
| 3770 |
NEON_GET_REG(T0, rm, n * 2);
|
| 3771 |
NEON_GET_REG(T0, rm, n * 2 + 1); |
| 3772 |
} |
| 3773 |
|
| 3774 |
gen_neon_shift_im_narrow[q][u][size - 1]();
|
| 3775 |
|
| 3776 |
if (size < 3 && (pass & 1) == 0) { |
| 3777 |
gen_neon_movl_scratch_T0(0);
|
| 3778 |
} else {
|
| 3779 |
uint32_t offset; |
| 3780 |
|
| 3781 |
if (size < 3) |
| 3782 |
gen_neon_movl_T1_scratch(0);
|
| 3783 |
|
| 3784 |
if (op == 8 && !u) { |
| 3785 |
gen_neon_narrow[size - 1]();
|
| 3786 |
} else {
|
| 3787 |
if (op == 8) |
| 3788 |
gen_neon_narrow_sats[size - 2]();
|
| 3789 |
else
|
| 3790 |
gen_neon_narrow_satu[size - 1]();
|
| 3791 |
} |
| 3792 |
if (size == 3) |
| 3793 |
offset = neon_reg_offset(rd, n); |
| 3794 |
else
|
| 3795 |
offset = neon_reg_offset(rd, n >> 1);
|
| 3796 |
gen_op_neon_setreg_T0(offset); |
| 3797 |
} |
| 3798 |
} /* for pass */
|
| 3799 |
} else if (op == 10) { |
| 3800 |
/* VSHLL */
|
| 3801 |
if (q)
|
| 3802 |
return 1; |
| 3803 |
for (pass = 0; pass < 2; pass++) { |
| 3804 |
/* Avoid clobbering the input operand. */
|
| 3805 |
if (rd == rm)
|
| 3806 |
n = 1 - pass;
|
| 3807 |
else
|
| 3808 |
n = pass; |
| 3809 |
|
| 3810 |
NEON_GET_REG(T0, rm, n); |
| 3811 |
GEN_NEON_INTEGER_OP(widen); |
| 3812 |
if (shift != 0) { |
| 3813 |
/* The shift is less than the width of the source
|
| 3814 |
type, so in some cases we can just
|
| 3815 |
shift the whole register. */
|
| 3816 |
if (size == 1 || (size == 0 && u)) { |
| 3817 |
gen_op_shll_T0_im(shift); |
| 3818 |
gen_op_shll_T1_im(shift); |
| 3819 |
} else {
|
| 3820 |
switch (size) {
|
| 3821 |
case 0: gen_op_neon_shll_u16(shift); break; |
| 3822 |
case 2: gen_op_neon_shll_u64(shift); break; |
| 3823 |
default: abort();
|
| 3824 |
} |
| 3825 |
} |
| 3826 |
} |
| 3827 |
NEON_SET_REG(T0, rd, n * 2);
|
| 3828 |
NEON_SET_REG(T1, rd, n * 2 + 1); |
| 3829 |
} |
| 3830 |
} else if (op == 15 || op == 16) { |
| 3831 |
/* VCVT fixed-point. */
|
| 3832 |
for (pass = 0; pass < (q ? 4 : 2); pass++) { |
| 3833 |
gen_op_vfp_getreg_F0s(neon_reg_offset(rm, pass)); |
| 3834 |
if (op & 1) { |
| 3835 |
if (u)
|
| 3836 |
gen_op_vfp_ultos(shift); |
| 3837 |
else
|
| 3838 |
gen_op_vfp_sltos(shift); |
| 3839 |
} else {
|
| 3840 |
if (u)
|
| 3841 |
gen_op_vfp_touls(shift); |
| 3842 |
else
|
| 3843 |
gen_op_vfp_tosls(shift); |
| 3844 |
} |
| 3845 |
gen_op_vfp_setreg_F0s(neon_reg_offset(rd, pass)); |
| 3846 |
} |
| 3847 |
} else {
|
| 3848 |
return 1; |
| 3849 |
} |
| 3850 |
} else { /* (insn & 0x00380080) == 0 */ |
| 3851 |
int invert;
|
| 3852 |
|
| 3853 |
op = (insn >> 8) & 0xf; |
| 3854 |
/* One register and immediate. */
|
| 3855 |
imm = (u << 7) | ((insn >> 12) & 0x70) | (insn & 0xf); |
| 3856 |
invert = (insn & (1 << 5)) != 0; |
| 3857 |
switch (op) {
|
| 3858 |
case 0: case 1: |
| 3859 |
/* no-op */
|
| 3860 |
break;
|
| 3861 |
case 2: case 3: |
| 3862 |
imm <<= 8;
|
| 3863 |
break;
|
| 3864 |
case 4: case 5: |
| 3865 |
imm <<= 16;
|
| 3866 |
break;
|
| 3867 |
case 6: case 7: |
| 3868 |
imm <<= 24;
|
| 3869 |
break;
|
| 3870 |
case 8: case 9: |
| 3871 |
imm |= imm << 16;
|
| 3872 |
break;
|
| 3873 |
case 10: case 11: |
| 3874 |
imm = (imm << 8) | (imm << 24); |
| 3875 |
break;
|
| 3876 |
case 12: |
| 3877 |
imm = (imm < 8) | 0xff; |
| 3878 |
break;
|
| 3879 |
case 13: |
| 3880 |
imm = (imm << 16) | 0xffff; |
| 3881 |
break;
|
| 3882 |
case 14: |
| 3883 |
imm |= (imm << 8) | (imm << 16) | (imm << 24); |
| 3884 |
if (invert)
|
| 3885 |
imm = ~imm; |
| 3886 |
break;
|
| 3887 |
case 15: |
| 3888 |
imm = ((imm & 0x80) << 24) | ((imm & 0x3f) << 19) |
| 3889 |
| ((imm & 0x40) ? (0x1f << 25) : (1 << 30)); |
| 3890 |
break;
|
| 3891 |
} |
| 3892 |
if (invert)
|
| 3893 |
imm = ~imm; |
| 3894 |
|
| 3895 |
if (op != 14 || !invert) |
| 3896 |
gen_op_movl_T1_im(imm); |
| 3897 |
|
| 3898 |
for (pass = 0; pass < (q ? 4 : 2); pass++) { |
| 3899 |
if (op & 1 && op < 12) { |
| 3900 |
NEON_GET_REG(T0, rd, pass); |
| 3901 |
if (invert) {
|
| 3902 |
/* The immediate value has already been inverted, so
|
| 3903 |
BIC becomes AND. */
|
| 3904 |
gen_op_andl_T0_T1(); |
| 3905 |
} else {
|
| 3906 |
gen_op_orl_T0_T1(); |
| 3907 |
} |
| 3908 |
NEON_SET_REG(T0, rd, pass); |
| 3909 |
} else {
|
| 3910 |
if (op == 14 && invert) { |
| 3911 |
uint32_t tmp; |
| 3912 |
tmp = 0;
|
| 3913 |
for (n = 0; n < 4; n++) { |
| 3914 |
if (imm & (1 << (n + (pass & 1) * 4))) |
| 3915 |
tmp |= 0xff << (n * 8); |
| 3916 |
} |
| 3917 |
gen_op_movl_T1_im(tmp); |
| 3918 |
} |
| 3919 |
/* VMOV, VMVN. */
|
| 3920 |
NEON_SET_REG(T1, rd, pass); |
| 3921 |
} |
| 3922 |
} |
| 3923 |
} |
| 3924 |
} else { /* (insn & 0x00800010 == 0x00800010) */ |
| 3925 |
if (size != 3) { |
| 3926 |
op = (insn >> 8) & 0xf; |
| 3927 |
if ((insn & (1 << 6)) == 0) { |
| 3928 |
/* Three registers of different lengths. */
|
| 3929 |
int src1_wide;
|
| 3930 |
int src2_wide;
|
| 3931 |
int prewiden;
|
| 3932 |
/* prewiden, src1_wide, src2_wide */
|
| 3933 |
static const int neon_3reg_wide[16][3] = { |
| 3934 |
{1, 0, 0}, /* VADDL */
|
| 3935 |
{1, 1, 0}, /* VADDW */
|
| 3936 |
{1, 0, 0}, /* VSUBL */
|
| 3937 |
{1, 1, 0}, /* VSUBW */
|
| 3938 |
{0, 1, 1}, /* VADDHN */
|
| 3939 |
{0, 0, 0}, /* VABAL */
|
| 3940 |
{0, 1, 1}, /* VSUBHN */
|
| 3941 |
{0, 0, 0}, /* VABDL */
|
| 3942 |
{0, 0, 0}, /* VMLAL */
|
| 3943 |
{0, 0, 0}, /* VQDMLAL */
|
| 3944 |
{0, 0, 0}, /* VMLSL */
|
| 3945 |
{0, 0, 0}, /* VQDMLSL */
|
| 3946 |
{0, 0, 0}, /* Integer VMULL */
|
| 3947 |
{0, 0, 0}, /* VQDMULL */
|
| 3948 |
{0, 0, 0} /* Polynomial VMULL */
|
| 3949 |
}; |
| 3950 |
|
| 3951 |
prewiden = neon_3reg_wide[op][0];
|
| 3952 |
src1_wide = neon_3reg_wide[op][1];
|
| 3953 |
src2_wide = neon_3reg_wide[op][2];
|
| 3954 |
|
| 3955 |
/* Avoid overlapping operands. Wide source operands are
|
| 3956 |
always aligned so will never overlap with wide
|
| 3957 |
destinations in problematic ways. */
|
| 3958 |
if (rd == rm) {
|
| 3959 |
NEON_GET_REG(T2, rm, 1);
|
| 3960 |
} else if (rd == rn) { |
| 3961 |
NEON_GET_REG(T2, rn, 1);
|
| 3962 |
} |
| 3963 |
for (pass = 0; pass < 2; pass++) { |
| 3964 |
/* Load the second operand into env->vfp.scratch.
|
| 3965 |
Also widen narrow operands. */
|
| 3966 |
if (pass == 1 && rd == rm) { |
| 3967 |
if (prewiden) {
|
| 3968 |
gen_op_movl_T0_T2(); |
| 3969 |
} else {
|
| 3970 |
gen_op_movl_T1_T2(); |
| 3971 |
} |
| 3972 |
} else {
|
| 3973 |
if (src2_wide) {
|
| 3974 |
NEON_GET_REG(T0, rm, pass * 2);
|
| 3975 |
NEON_GET_REG(T1, rm, pass * 2 + 1); |
| 3976 |
} else {
|
| 3977 |
if (prewiden) {
|
| 3978 |
NEON_GET_REG(T0, rm, pass); |
| 3979 |
} else {
|
| 3980 |
NEON_GET_REG(T1, rm, pass); |
| 3981 |
} |
| 3982 |
} |
| 3983 |
} |
| 3984 |
if (prewiden && !src2_wide) {
|
| 3985 |
GEN_NEON_INTEGER_OP(widen); |
| 3986 |
} |
| 3987 |
if (prewiden || src2_wide) {
|
| 3988 |
gen_neon_movl_scratch_T0(0);
|
| 3989 |
gen_neon_movl_scratch_T1(1);
|
| 3990 |
} |
| 3991 |
|
| 3992 |
/* Load the first operand. */
|
| 3993 |
if (pass == 1 && rd == rn) { |
| 3994 |
gen_op_movl_T0_T2(); |
| 3995 |
} else {
|
| 3996 |
if (src1_wide) {
|
| 3997 |
NEON_GET_REG(T0, rn, pass * 2);
|
| 3998 |
NEON_GET_REG(T1, rn, pass * 2 + 1); |
| 3999 |
} else {
|
| 4000 |
NEON_GET_REG(T0, rn, pass); |
| 4001 |
} |
| 4002 |
} |
| 4003 |
if (prewiden && !src1_wide) {
|
| 4004 |
GEN_NEON_INTEGER_OP(widen); |
| 4005 |
} |
| 4006 |
switch (op) {
|
| 4007 |
case 0: case 1: case 4: /* VADDL, VADDW, VADDHN, VRADDHN */ |
| 4008 |
switch (size) {
|
| 4009 |
case 0: gen_op_neon_addl_u16(); break; |
| 4010 |
case 1: gen_op_neon_addl_u32(); break; |
| 4011 |
case 2: gen_op_neon_addl_u64(); break; |
| 4012 |
default: abort();
|
| 4013 |
} |
| 4014 |
break;
|
| 4015 |
case 2: case 3: case 6: /* VSUBL, VSUBW, VSUBHL, VRSUBHL */ |
| 4016 |
switch (size) {
|
| 4017 |
case 0: gen_op_neon_subl_u16(); break; |
| 4018 |
case 1: gen_op_neon_subl_u32(); break; |
| 4019 |
case 2: gen_op_neon_subl_u64(); break; |
| 4020 |
default: abort();
|
| 4021 |
} |
| 4022 |
break;
|
| 4023 |
case 5: case 7: /* VABAL, VABDL */ |
| 4024 |
switch ((size << 1) | u) { |
| 4025 |
case 0: gen_op_neon_abdl_s16(); break; |
| 4026 |
case 1: gen_op_neon_abdl_u16(); break; |
| 4027 |
case 2: gen_op_neon_abdl_s32(); break; |
| 4028 |
case 3: gen_op_neon_abdl_u32(); break; |
| 4029 |
case 4: gen_op_neon_abdl_s64(); break; |
| 4030 |
case 5: gen_op_neon_abdl_u64(); break; |
| 4031 |
default: abort();
|
| 4032 |
} |
| 4033 |
break;
|
| 4034 |
case 8: case 9: case 10: case 11: case 12: case 13: |
| 4035 |
/* VMLAL, VQDMLAL, VMLSL, VQDMLSL, VMULL, VQDMULL */
|
| 4036 |
switch ((size << 1) | u) { |
| 4037 |
case 0: gen_op_neon_mull_s8(); break; |
| 4038 |
case 1: gen_op_neon_mull_u8(); break; |
| 4039 |
case 2: gen_op_neon_mull_s16(); break; |
| 4040 |
case 3: gen_op_neon_mull_u16(); break; |
| 4041 |
case 4: gen_op_imull_T0_T1(); break; |
| 4042 |
case 5: gen_op_mull_T0_T1(); break; |
| 4043 |
default: abort();
|
| 4044 |
} |
| 4045 |
break;
|
| 4046 |
case 14: /* Polynomial VMULL */ |
| 4047 |
cpu_abort(env, "Polynomial VMULL not implemented");
|
| 4048 |
|
| 4049 |
default: /* 15 is RESERVED. */ |
| 4050 |
return 1; |
| 4051 |
} |
| 4052 |
if (op == 5 || op == 13 || (op >= 8 && op <= 11)) { |
| 4053 |
/* Accumulate. */
|
| 4054 |
if (op == 10 || op == 11) { |
| 4055 |
switch (size) {
|
| 4056 |
case 0: gen_op_neon_negl_u16(); break; |
| 4057 |
case 1: gen_op_neon_negl_u32(); break; |
| 4058 |
case 2: gen_op_neon_negl_u64(); break; |
| 4059 |
default: abort();
|
| 4060 |
} |
| 4061 |
} |
| 4062 |
|
| 4063 |
gen_neon_movl_scratch_T0(0);
|
| 4064 |
gen_neon_movl_scratch_T1(1);
|
| 4065 |
|
| 4066 |
if (op != 13) { |
| 4067 |
NEON_GET_REG(T0, rd, pass * 2);
|
| 4068 |
NEON_GET_REG(T1, rd, pass * 2 + 1); |
| 4069 |
} |
| 4070 |
|
| 4071 |
switch (op) {
|
| 4072 |
case 5: case 8: case 10: /* VABAL, VMLAL, VMLSL */ |
| 4073 |
switch (size) {
|
| 4074 |
case 0: gen_op_neon_addl_u16(); break; |
| 4075 |
case 1: gen_op_neon_addl_u32(); break; |
| 4076 |
case 2: gen_op_neon_addl_u64(); break; |
| 4077 |
default: abort();
|
| 4078 |
} |
| 4079 |
break;
|
| 4080 |
case 9: case 11: /* VQDMLAL, VQDMLSL */ |
| 4081 |
switch (size) {
|
| 4082 |
case 1: gen_op_neon_addl_saturate_s32(); break; |
| 4083 |
case 2: gen_op_neon_addl_saturate_s64(); break; |
| 4084 |
default: abort();
|
| 4085 |
} |
| 4086 |
/* Fall through. */
|
| 4087 |
case 13: /* VQDMULL */ |
| 4088 |
switch (size) {
|
| 4089 |
case 1: gen_op_neon_addl_saturate_s32(); break; |
| 4090 |
case 2: gen_op_neon_addl_saturate_s64(); break; |
| 4091 |
default: abort();
|
| 4092 |
} |
| 4093 |
break;
|
| 4094 |
default:
|
| 4095 |
abort(); |
| 4096 |
} |
| 4097 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 4098 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 4099 |
} else if (op == 4 || op == 6) { |
| 4100 |
/* Narrowing operation. */
|
| 4101 |
if (u) {
|
| 4102 |
switch (size) {
|
| 4103 |
case 0: gen_op_neon_narrow_high_u8(); break; |
| 4104 |
case 1: gen_op_neon_narrow_high_u16(); break; |
| 4105 |
case 2: gen_op_movl_T0_T1(); break; |
| 4106 |
default: abort();
|
| 4107 |
} |
| 4108 |
} else {
|
| 4109 |
switch (size) {
|
| 4110 |
case 0: gen_op_neon_narrow_high_round_u8(); break; |
| 4111 |
case 1: gen_op_neon_narrow_high_round_u16(); break; |
| 4112 |
case 2: gen_op_neon_narrow_high_round_u32(); break; |
| 4113 |
default: abort();
|
| 4114 |
} |
| 4115 |
} |
| 4116 |
NEON_SET_REG(T0, rd, pass); |
| 4117 |
} else {
|
| 4118 |
/* Write back the result. */
|
| 4119 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 4120 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 4121 |
} |
| 4122 |
} |
| 4123 |
} else {
|
| 4124 |
/* Two registers and a scalar. */
|
| 4125 |
switch (op) {
|
| 4126 |
case 0: /* Integer VMLA scalar */ |
| 4127 |
case 1: /* Float VMLA scalar */ |
| 4128 |
case 4: /* Integer VMLS scalar */ |
| 4129 |
case 5: /* Floating point VMLS scalar */ |
| 4130 |
case 8: /* Integer VMUL scalar */ |
| 4131 |
case 9: /* Floating point VMUL scalar */ |
| 4132 |
case 12: /* VQDMULH scalar */ |
| 4133 |
case 13: /* VQRDMULH scalar */ |
| 4134 |
gen_neon_get_scalar(size, rm); |
| 4135 |
gen_op_movl_T2_T0(); |
| 4136 |
for (pass = 0; pass < (u ? 4 : 2); pass++) { |
| 4137 |
if (pass != 0) |
| 4138 |
gen_op_movl_T0_T2(); |
| 4139 |
NEON_GET_REG(T1, rn, pass); |
| 4140 |
if (op == 12) { |
| 4141 |
if (size == 1) { |
| 4142 |
gen_op_neon_qdmulh_s16(); |
| 4143 |
} else {
|
| 4144 |
gen_op_neon_qdmulh_s32(); |
| 4145 |
} |
| 4146 |
} else if (op == 13) { |
| 4147 |
if (size == 1) { |
| 4148 |
gen_op_neon_qrdmulh_s16(); |
| 4149 |
} else {
|
| 4150 |
gen_op_neon_qrdmulh_s32(); |
| 4151 |
} |
| 4152 |
} else if (op & 1) { |
| 4153 |
gen_op_neon_mul_f32(); |
| 4154 |
} else {
|
| 4155 |
switch (size) {
|
| 4156 |
case 0: gen_op_neon_mul_u8(); break; |
| 4157 |
case 1: gen_op_neon_mul_u16(); break; |
| 4158 |
case 2: gen_op_mul_T0_T1(); break; |
| 4159 |
default: return 1; |
| 4160 |
} |
| 4161 |
} |
| 4162 |
if (op < 8) { |
| 4163 |
/* Accumulate. */
|
| 4164 |
NEON_GET_REG(T1, rd, pass); |
| 4165 |
switch (op) {
|
| 4166 |
case 0: |
| 4167 |
gen_neon_add(size); |
| 4168 |
break;
|
| 4169 |
case 1: |
| 4170 |
gen_op_neon_add_f32(); |
| 4171 |
break;
|
| 4172 |
case 4: |
| 4173 |
switch (size) {
|
| 4174 |
case 0: gen_op_neon_rsb_u8(); break; |
| 4175 |
case 1: gen_op_neon_rsb_u16(); break; |
| 4176 |
case 2: gen_op_rsbl_T0_T1(); break; |
| 4177 |
default: return 1; |
| 4178 |
} |
| 4179 |
break;
|
| 4180 |
case 5: |
| 4181 |
gen_op_neon_rsb_f32(); |
| 4182 |
break;
|
| 4183 |
default:
|
| 4184 |
abort(); |
| 4185 |
} |
| 4186 |
} |
| 4187 |
NEON_SET_REG(T0, rd, pass); |
| 4188 |
} |
| 4189 |
break;
|
| 4190 |
case 2: /* VMLAL sclar */ |
| 4191 |
case 3: /* VQDMLAL scalar */ |
| 4192 |
case 6: /* VMLSL scalar */ |
| 4193 |
case 7: /* VQDMLSL scalar */ |
| 4194 |
case 10: /* VMULL scalar */ |
| 4195 |
case 11: /* VQDMULL scalar */ |
| 4196 |
if (rd == rn) {
|
| 4197 |
/* Save overlapping operands before they are
|
| 4198 |
clobbered. */
|
| 4199 |
NEON_GET_REG(T0, rn, 1);
|
| 4200 |
gen_neon_movl_scratch_T0(2);
|
| 4201 |
} |
| 4202 |
gen_neon_get_scalar(size, rm); |
| 4203 |
gen_op_movl_T2_T0(); |
| 4204 |
for (pass = 0; pass < 2; pass++) { |
| 4205 |
if (pass != 0) { |
| 4206 |
gen_op_movl_T0_T2(); |
| 4207 |
} |
| 4208 |
if (pass != 0 && rd == rn) { |
| 4209 |
gen_neon_movl_T1_scratch(2);
|
| 4210 |
} else {
|
| 4211 |
NEON_GET_REG(T1, rn, pass); |
| 4212 |
} |
| 4213 |
switch ((size << 1) | u) { |
| 4214 |
case 0: gen_op_neon_mull_s8(); break; |
| 4215 |
case 1: gen_op_neon_mull_u8(); break; |
| 4216 |
case 2: gen_op_neon_mull_s16(); break; |
| 4217 |
case 3: gen_op_neon_mull_u16(); break; |
| 4218 |
case 4: gen_op_imull_T0_T1(); break; |
| 4219 |
case 5: gen_op_mull_T0_T1(); break; |
| 4220 |
default: abort();
|
| 4221 |
} |
| 4222 |
if (op == 6 || op == 7) { |
| 4223 |
switch (size) {
|
| 4224 |
case 0: gen_op_neon_negl_u16(); break; |
| 4225 |
case 1: gen_op_neon_negl_u32(); break; |
| 4226 |
case 2: gen_op_neon_negl_u64(); break; |
| 4227 |
default: abort();
|
| 4228 |
} |
| 4229 |
} |
| 4230 |
gen_neon_movl_scratch_T0(0);
|
| 4231 |
gen_neon_movl_scratch_T1(1);
|
| 4232 |
NEON_GET_REG(T0, rd, pass * 2);
|
| 4233 |
NEON_GET_REG(T1, rd, pass * 2 + 1); |
| 4234 |
switch (op) {
|
| 4235 |
case 2: case 6: |
| 4236 |
switch (size) {
|
| 4237 |
case 0: gen_op_neon_addl_u16(); break; |
| 4238 |
case 1: gen_op_neon_addl_u32(); break; |
| 4239 |
case 2: gen_op_neon_addl_u64(); break; |
| 4240 |
default: abort();
|
| 4241 |
} |
| 4242 |
break;
|
| 4243 |
case 3: case 7: |
| 4244 |
switch (size) {
|
| 4245 |
case 1: |
| 4246 |
gen_op_neon_addl_saturate_s32(); |
| 4247 |
gen_op_neon_addl_saturate_s32(); |
| 4248 |
break;
|
| 4249 |
case 2: |
| 4250 |
gen_op_neon_addl_saturate_s64(); |
| 4251 |
gen_op_neon_addl_saturate_s64(); |
| 4252 |
break;
|
| 4253 |
default: abort();
|
| 4254 |
} |
| 4255 |
break;
|
| 4256 |
case 10: |
| 4257 |
/* no-op */
|
| 4258 |
break;
|
| 4259 |
case 11: |
| 4260 |
switch (size) {
|
| 4261 |
case 1: gen_op_neon_addl_saturate_s32(); break; |
| 4262 |
case 2: gen_op_neon_addl_saturate_s64(); break; |
| 4263 |
default: abort();
|
| 4264 |
} |
| 4265 |
break;
|
| 4266 |
default:
|
| 4267 |
abort(); |
| 4268 |
} |
| 4269 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 4270 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 4271 |
} |
| 4272 |
break;
|
| 4273 |
default: /* 14 and 15 are RESERVED */ |
| 4274 |
return 1; |
| 4275 |
} |
| 4276 |
} |
| 4277 |
} else { /* size == 3 */ |
| 4278 |
if (!u) {
|
| 4279 |
/* Extract. */
|
| 4280 |
int reg;
|
| 4281 |
imm = (insn >> 8) & 0xf; |
| 4282 |
reg = rn; |
| 4283 |
count = q ? 4 : 2; |
| 4284 |
n = imm >> 2;
|
| 4285 |
NEON_GET_REG(T0, reg, n); |
| 4286 |
for (pass = 0; pass < count; pass++) { |
| 4287 |
n++; |
| 4288 |
if (n > count) {
|
| 4289 |
reg = rm; |
| 4290 |
n -= count; |
| 4291 |
} |
| 4292 |
if (imm & 3) { |
| 4293 |
NEON_GET_REG(T1, reg, n); |
| 4294 |
gen_op_neon_extract((insn << 3) & 0x1f); |
| 4295 |
} |
| 4296 |
/* ??? This is broken if rd and rm overlap */
|
| 4297 |
NEON_SET_REG(T0, rd, pass); |
| 4298 |
if (imm & 3) { |
| 4299 |
gen_op_movl_T0_T1(); |
| 4300 |
} else {
|
| 4301 |
NEON_GET_REG(T0, reg, n); |
| 4302 |
} |
| 4303 |
} |
| 4304 |
} else if ((insn & (1 << 11)) == 0) { |
| 4305 |
/* Two register misc. */
|
| 4306 |
op = ((insn >> 12) & 0x30) | ((insn >> 7) & 0xf); |
| 4307 |
size = (insn >> 18) & 3; |
| 4308 |
switch (op) {
|
| 4309 |
case 0: /* VREV64 */ |
| 4310 |
if (size == 3) |
| 4311 |
return 1; |
| 4312 |
for (pass = 0; pass < (q ? 2 : 1); pass++) { |
| 4313 |
NEON_GET_REG(T0, rm, pass * 2);
|
| 4314 |
NEON_GET_REG(T1, rm, pass * 2 + 1); |
| 4315 |
switch (size) {
|
| 4316 |
case 0: gen_op_rev_T0(); break; |
| 4317 |
case 1: gen_op_revh_T0(); break; |
| 4318 |
case 2: /* no-op */ break; |
| 4319 |
default: abort();
|
| 4320 |
} |
| 4321 |
NEON_SET_REG(T0, rd, pass * 2 + 1); |
| 4322 |
if (size == 2) { |
| 4323 |
NEON_SET_REG(T1, rd, pass * 2);
|
| 4324 |
} else {
|
| 4325 |
gen_op_movl_T0_T1(); |
| 4326 |
switch (size) {
|
| 4327 |
case 0: gen_op_rev_T0(); break; |
| 4328 |
case 1: gen_op_revh_T0(); break; |
| 4329 |
default: abort();
|
| 4330 |
} |
| 4331 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 4332 |
} |
| 4333 |
} |
| 4334 |
break;
|
| 4335 |
case 4: case 5: /* VPADDL */ |
| 4336 |
case 12: case 13: /* VPADAL */ |
| 4337 |
if (size < 2) |
| 4338 |
goto elementwise;
|
| 4339 |
if (size == 3) |
| 4340 |
return 1; |
| 4341 |
for (pass = 0; pass < (q ? 2 : 1); pass++) { |
| 4342 |
NEON_GET_REG(T0, rm, pass * 2);
|
| 4343 |
NEON_GET_REG(T1, rm, pass * 2 + 1); |
| 4344 |
if (op & 1) |
| 4345 |
gen_op_neon_paddl_u32(); |
| 4346 |
else
|
| 4347 |
gen_op_neon_paddl_s32(); |
| 4348 |
if (op >= 12) { |
| 4349 |
/* Accumulate. */
|
| 4350 |
gen_neon_movl_scratch_T0(0);
|
| 4351 |
gen_neon_movl_scratch_T1(1);
|
| 4352 |
|
| 4353 |
NEON_GET_REG(T0, rd, pass * 2);
|
| 4354 |
NEON_GET_REG(T1, rd, pass * 2 + 1); |
| 4355 |
gen_op_neon_addl_u64(); |
| 4356 |
} |
| 4357 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 4358 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 4359 |
} |
| 4360 |
break;
|
| 4361 |
case 33: /* VTRN */ |
| 4362 |
if (size == 2) { |
| 4363 |
for (n = 0; n < (q ? 4 : 2); n += 2) { |
| 4364 |
NEON_GET_REG(T0, rm, n); |
| 4365 |
NEON_GET_REG(T1, rd, n + 1);
|
| 4366 |
NEON_SET_REG(T1, rm, n); |
| 4367 |
NEON_SET_REG(T0, rd, n + 1);
|
| 4368 |
} |
| 4369 |
} else {
|
| 4370 |
goto elementwise;
|
| 4371 |
} |
| 4372 |
break;
|
| 4373 |
case 34: /* VUZP */ |
| 4374 |
/* Reg Before After
|
| 4375 |
Rd A3 A2 A1 A0 B2 B0 A2 A0
|
| 4376 |
Rm B3 B2 B1 B0 B3 B1 A3 A1
|
| 4377 |
*/
|
| 4378 |
if (size == 3) |
| 4379 |
return 1; |
| 4380 |
gen_neon_unzip(rd, q, 0, size);
|
| 4381 |
gen_neon_unzip(rm, q, 4, size);
|
| 4382 |
if (q) {
|
| 4383 |
static int unzip_order_q[8] = |
| 4384 |
{0, 2, 4, 6, 1, 3, 5, 7};
|
| 4385 |
for (n = 0; n < 8; n++) { |
| 4386 |
int reg = (n < 4) ? rd : rm; |
| 4387 |
gen_neon_movl_T0_scratch(unzip_order_q[n]); |
| 4388 |
NEON_SET_REG(T0, reg, n % 4);
|
| 4389 |
} |
| 4390 |
} else {
|
| 4391 |
static int unzip_order[4] = |
| 4392 |
{0, 4, 1, 5};
|
| 4393 |
for (n = 0; n < 4; n++) { |
| 4394 |
int reg = (n < 2) ? rd : rm; |
| 4395 |
gen_neon_movl_T0_scratch(unzip_order[n]); |
| 4396 |
NEON_SET_REG(T0, reg, n % 2);
|
| 4397 |
} |
| 4398 |
} |
| 4399 |
break;
|
| 4400 |
case 35: /* VZIP */ |
| 4401 |
/* Reg Before After
|
| 4402 |
Rd A3 A2 A1 A0 B1 A1 B0 A0
|
| 4403 |
Rm B3 B2 B1 B0 B3 A3 B2 A2
|
| 4404 |
*/
|
| 4405 |
if (size == 3) |
| 4406 |
return 1; |
| 4407 |
count = (q ? 4 : 2); |
| 4408 |
for (n = 0; n < count; n++) { |
| 4409 |
NEON_GET_REG(T0, rd, n); |
| 4410 |
NEON_GET_REG(T1, rd, n); |
| 4411 |
switch (size) {
|
| 4412 |
case 0: gen_op_neon_zip_u8(); break; |
| 4413 |
case 1: gen_op_neon_zip_u16(); break; |
| 4414 |
case 2: /* no-op */; break; |
| 4415 |
default: abort();
|
| 4416 |
} |
| 4417 |
gen_neon_movl_scratch_T0(n * 2);
|
| 4418 |
gen_neon_movl_scratch_T1(n * 2 + 1); |
| 4419 |
} |
| 4420 |
for (n = 0; n < count * 2; n++) { |
| 4421 |
int reg = (n < count) ? rd : rm;
|
| 4422 |
gen_neon_movl_T0_scratch(n); |
| 4423 |
NEON_SET_REG(T0, reg, n % count); |
| 4424 |
} |
| 4425 |
break;
|
| 4426 |
case 36: case 37: /* VMOVN, VQMOVUN, VQMOVN */ |
| 4427 |
for (pass = 0; pass < 2; pass++) { |
| 4428 |
if (rd == rm + 1) { |
| 4429 |
n = 1 - pass;
|
| 4430 |
} else {
|
| 4431 |
n = pass; |
| 4432 |
} |
| 4433 |
NEON_GET_REG(T0, rm, n * 2);
|
| 4434 |
NEON_GET_REG(T1, rm, n * 2 + 1); |
| 4435 |
if (op == 36 && q == 0) { |
| 4436 |
switch (size) {
|
| 4437 |
case 0: gen_op_neon_narrow_u8(); break; |
| 4438 |
case 1: gen_op_neon_narrow_u16(); break; |
| 4439 |
case 2: /* no-op */ break; |
| 4440 |
default: return 1; |
| 4441 |
} |
| 4442 |
} else if (q) { |
| 4443 |
switch (size) {
|
| 4444 |
case 0: gen_op_neon_narrow_sat_u8(); break; |
| 4445 |
case 1: gen_op_neon_narrow_sat_u16(); break; |
| 4446 |
case 2: gen_op_neon_narrow_sat_u32(); break; |
| 4447 |
default: return 1; |
| 4448 |
} |
| 4449 |
} else {
|
| 4450 |
switch (size) {
|
| 4451 |
case 0: gen_op_neon_narrow_sat_s8(); break; |
| 4452 |
case 1: gen_op_neon_narrow_sat_s16(); break; |
| 4453 |
case 2: gen_op_neon_narrow_sat_s32(); break; |
| 4454 |
default: return 1; |
| 4455 |
} |
| 4456 |
} |
| 4457 |
NEON_SET_REG(T0, rd, n); |
| 4458 |
} |
| 4459 |
break;
|
| 4460 |
case 38: /* VSHLL */ |
| 4461 |
if (q)
|
| 4462 |
return 1; |
| 4463 |
if (rm == rd) {
|
| 4464 |
NEON_GET_REG(T2, rm, 1);
|
| 4465 |
} |
| 4466 |
for (pass = 0; pass < 2; pass++) { |
| 4467 |
if (pass == 1 && rm == rd) { |
| 4468 |
gen_op_movl_T0_T2(); |
| 4469 |
} else {
|
| 4470 |
NEON_GET_REG(T0, rm, pass); |
| 4471 |
} |
| 4472 |
switch (size) {
|
| 4473 |
case 0: gen_op_neon_widen_high_u8(); break; |
| 4474 |
case 1: gen_op_neon_widen_high_u16(); break; |
| 4475 |
case 2: |
| 4476 |
gen_op_movl_T1_T0(); |
| 4477 |
gen_op_movl_T0_im(0);
|
| 4478 |
break;
|
| 4479 |
default: return 1; |
| 4480 |
} |
| 4481 |
NEON_SET_REG(T0, rd, pass * 2);
|
| 4482 |
NEON_SET_REG(T1, rd, pass * 2 + 1); |
| 4483 |
} |
| 4484 |
break;
|
| 4485 |
default:
|
| 4486 |
elementwise:
|
| 4487 |
for (pass = 0; pass < (q ? 4 : 2); pass++) { |
| 4488 |
if (op == 30 || op == 31 || op >= 58) { |
| 4489 |
gen_op_vfp_getreg_F0s(neon_reg_offset(rm, pass)); |
| 4490 |
} else {
|
| 4491 |
NEON_GET_REG(T0, rm, pass); |
| 4492 |
} |
| 4493 |
switch (op) {
|
| 4494 |
case 1: /* VREV32 */ |
| 4495 |
switch (size) {
|
| 4496 |
case 0: gen_op_rev_T0(); break; |
| 4497 |
case 1: gen_op_revh_T0(); break; |
| 4498 |
default: return 1; |
| 4499 |
} |
| 4500 |
break;
|
| 4501 |
case 2: /* VREV16 */ |
| 4502 |
if (size != 0) |
| 4503 |
return 1; |
| 4504 |
gen_op_rev16_T0(); |
| 4505 |
break;
|
| 4506 |
case 4: case 5: /* VPADDL */ |
| 4507 |
case 12: case 13: /* VPADAL */ |
| 4508 |
switch ((size << 1) | (op & 1)) { |
| 4509 |
case 0: gen_op_neon_paddl_s8(); break; |
| 4510 |
case 1: gen_op_neon_paddl_u8(); break; |
| 4511 |
case 2: gen_op_neon_paddl_s16(); break; |
| 4512 |
case 3: gen_op_neon_paddl_u16(); break; |
| 4513 |
default: abort();
|
| 4514 |
} |
| 4515 |
if (op >= 12) { |
| 4516 |
/* Accumulate */
|
| 4517 |
NEON_GET_REG(T1, rd, pass); |
| 4518 |
switch (size) {
|
| 4519 |
case 0: gen_op_neon_add_u16(); break; |
| 4520 |
case 1: gen_op_addl_T0_T1(); break; |
| 4521 |
default: abort();
|
| 4522 |
} |
| 4523 |
} |
| 4524 |
break;
|
| 4525 |
case 8: /* CLS */ |
| 4526 |
switch (size) {
|
| 4527 |
case 0: gen_op_neon_cls_s8(); break; |
| 4528 |
case 1: gen_op_neon_cls_s16(); break; |
| 4529 |
case 2: gen_op_neon_cls_s32(); break; |
| 4530 |
default: return 1; |
| 4531 |
} |
| 4532 |
break;
|
| 4533 |
case 9: /* CLZ */ |
| 4534 |
switch (size) {
|
| 4535 |
case 0: gen_op_neon_clz_u8(); break; |
| 4536 |
case 1: gen_op_neon_clz_u16(); break; |
| 4537 |
case 2: gen_helper_clz(cpu_T[0], cpu_T[0]); break; |
| 4538 |
default: return 1; |
| 4539 |
} |
| 4540 |
break;
|
| 4541 |
case 10: /* CNT */ |
| 4542 |
if (size != 0) |
| 4543 |
return 1; |
| 4544 |
gen_op_neon_cnt_u8(); |
| 4545 |
break;
|
| 4546 |
case 11: /* VNOT */ |
| 4547 |
if (size != 0) |
| 4548 |
return 1; |
| 4549 |
gen_op_notl_T0(); |
| 4550 |
break;
|
| 4551 |
case 14: /* VQABS */ |
| 4552 |
switch (size) {
|
| 4553 |
case 0: gen_op_neon_qabs_s8(); break; |
| 4554 |
case 1: gen_op_neon_qabs_s16(); break; |
| 4555 |
case 2: gen_op_neon_qabs_s32(); break; |
| 4556 |
default: return 1; |
| 4557 |
} |
| 4558 |
break;
|
| 4559 |
case 15: /* VQNEG */ |
| 4560 |
switch (size) {
|
| 4561 |
case 0: gen_op_neon_qneg_s8(); break; |
| 4562 |
case 1: gen_op_neon_qneg_s16(); break; |
| 4563 |
case 2: gen_op_neon_qneg_s32(); break; |
| 4564 |
default: return 1; |
| 4565 |
} |
| 4566 |
break;
|
| 4567 |
case 16: case 19: /* VCGT #0, VCLE #0 */ |
| 4568 |
gen_op_movl_T1_im(0);
|
| 4569 |
switch(size) {
|
| 4570 |
case 0: gen_op_neon_cgt_s8(); break; |
| 4571 |
case 1: gen_op_neon_cgt_s16(); break; |
| 4572 |
case 2: gen_op_neon_cgt_s32(); break; |
| 4573 |
default: return 1; |
| 4574 |
} |
| 4575 |
if (op == 19) |
| 4576 |
gen_op_notl_T0(); |
| 4577 |
break;
|
| 4578 |
case 17: case 20: /* VCGE #0, VCLT #0 */ |
| 4579 |
gen_op_movl_T1_im(0);
|
| 4580 |
switch(size) {
|
| 4581 |
case 0: gen_op_neon_cge_s8(); break; |
| 4582 |
case 1: gen_op_neon_cge_s16(); break; |
| 4583 |
case 2: gen_op_neon_cge_s32(); break; |
| 4584 |
default: return 1; |
| 4585 |
} |
| 4586 |
if (op == 20) |
| 4587 |
gen_op_notl_T0(); |
| 4588 |
break;
|
| 4589 |
case 18: /* VCEQ #0 */ |
| 4590 |
gen_op_movl_T1_im(0);
|
| 4591 |
switch(size) {
|
| 4592 |
case 0: gen_op_neon_ceq_u8(); break; |
| 4593 |
case 1: gen_op_neon_ceq_u16(); break; |
| 4594 |
case 2: gen_op_neon_ceq_u32(); break; |
| 4595 |
default: return 1; |
| 4596 |
} |
| 4597 |
break;
|
| 4598 |
case 22: /* VABS */ |
| 4599 |
switch(size) {
|
| 4600 |
case 0: gen_op_neon_abs_s8(); break; |
| 4601 |
case 1: gen_op_neon_abs_s16(); break; |
| 4602 |
case 2: gen_op_neon_abs_s32(); break; |
| 4603 |
default: return 1; |
| 4604 |
} |
| 4605 |
break;
|
| 4606 |
case 23: /* VNEG */ |
| 4607 |
gen_op_movl_T1_im(0);
|
| 4608 |
switch(size) {
|
| 4609 |
case 0: gen_op_neon_rsb_u8(); break; |
| 4610 |
case 1: gen_op_neon_rsb_u16(); break; |
| 4611 |
case 2: gen_op_rsbl_T0_T1(); break; |
| 4612 |
default: return 1; |
| 4613 |
} |
| 4614 |
break;
|
| 4615 |
case 24: case 27: /* Float VCGT #0, Float VCLE #0 */ |
| 4616 |
gen_op_movl_T1_im(0);
|
| 4617 |
gen_op_neon_cgt_f32(); |
| 4618 |
if (op == 27) |
| 4619 |
gen_op_notl_T0(); |
| 4620 |
break;
|
| 4621 |
case 25: case 28: /* Float VCGE #0, Float VCLT #0 */ |
| 4622 |
gen_op_movl_T1_im(0);
|
| 4623 |
gen_op_neon_cge_f32(); |
| 4624 |
if (op == 28) |
| 4625 |
gen_op_notl_T0(); |
| 4626 |
break;
|
| 4627 |
case 26: /* Float VCEQ #0 */ |
| 4628 |
gen_op_movl_T1_im(0);
|
| 4629 |
gen_op_neon_ceq_f32(); |
| 4630 |
break;
|
| 4631 |
case 30: /* Float VABS */ |
| 4632 |
gen_op_vfp_abss(); |
| 4633 |
break;
|
| 4634 |
case 31: /* Float VNEG */ |
| 4635 |
gen_op_vfp_negs(); |
| 4636 |
break;
|
| 4637 |
case 32: /* VSWP */ |
| 4638 |
NEON_GET_REG(T1, rd, pass); |
| 4639 |
NEON_SET_REG(T1, rm, pass); |
| 4640 |
break;
|
| 4641 |
case 33: /* VTRN */ |
| 4642 |
NEON_GET_REG(T1, rd, pass); |
| 4643 |
switch (size) {
|
| 4644 |
case 0: gen_op_neon_trn_u8(); break; |
| 4645 |
case 1: gen_op_neon_trn_u16(); break; |
| 4646 |
case 2: abort(); |
| 4647 |
default: return 1; |
| 4648< |