root / target-ppc / op_helper.c @ e2542fe2
History | View | Annotate | Download (124 kB)
| 1 |
/*
|
|---|---|
| 2 |
* PowerPC emulation helpers for qemu.
|
| 3 |
*
|
| 4 |
* Copyright (c) 2003-2007 Jocelyn Mayer
|
| 5 |
*
|
| 6 |
* This library is free software; you can redistribute it and/or
|
| 7 |
* modify it under the terms of the GNU Lesser General Public
|
| 8 |
* License as published by the Free Software Foundation; either
|
| 9 |
* version 2 of the License, or (at your option) any later version.
|
| 10 |
*
|
| 11 |
* This library is distributed in the hope that it will be useful,
|
| 12 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 13 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
| 14 |
* Lesser General Public License for more details.
|
| 15 |
*
|
| 16 |
* You should have received a copy of the GNU Lesser General Public
|
| 17 |
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
|
| 18 |
*/
|
| 19 |
#include <string.h> |
| 20 |
#include "exec.h" |
| 21 |
#include "host-utils.h" |
| 22 |
#include "helper.h" |
| 23 |
|
| 24 |
#include "helper_regs.h" |
| 25 |
|
| 26 |
//#define DEBUG_OP
|
| 27 |
//#define DEBUG_EXCEPTIONS
|
| 28 |
//#define DEBUG_SOFTWARE_TLB
|
| 29 |
|
| 30 |
#ifdef DEBUG_SOFTWARE_TLB
|
| 31 |
# define LOG_SWTLB(...) qemu_log(__VA_ARGS__)
|
| 32 |
#else
|
| 33 |
# define LOG_SWTLB(...) do { } while (0) |
| 34 |
#endif
|
| 35 |
|
| 36 |
|
| 37 |
/*****************************************************************************/
|
| 38 |
/* Exceptions processing helpers */
|
| 39 |
|
| 40 |
void helper_raise_exception_err (uint32_t exception, uint32_t error_code)
|
| 41 |
{
|
| 42 |
#if 0
|
| 43 |
printf("Raise exception %3x code : %d\n", exception, error_code);
|
| 44 |
#endif
|
| 45 |
env->exception_index = exception; |
| 46 |
env->error_code = error_code; |
| 47 |
cpu_loop_exit(); |
| 48 |
} |
| 49 |
|
| 50 |
void helper_raise_exception (uint32_t exception)
|
| 51 |
{
|
| 52 |
helper_raise_exception_err(exception, 0);
|
| 53 |
} |
| 54 |
|
| 55 |
/*****************************************************************************/
|
| 56 |
/* SPR accesses */
|
| 57 |
void helper_load_dump_spr (uint32_t sprn)
|
| 58 |
{
|
| 59 |
qemu_log("Read SPR %d %03x => " ADDRX "\n", |
| 60 |
sprn, sprn, env->spr[sprn]); |
| 61 |
} |
| 62 |
|
| 63 |
void helper_store_dump_spr (uint32_t sprn)
|
| 64 |
{
|
| 65 |
qemu_log("Write SPR %d %03x <= " ADDRX "\n", |
| 66 |
sprn, sprn, env->spr[sprn]); |
| 67 |
} |
| 68 |
|
| 69 |
target_ulong helper_load_tbl (void)
|
| 70 |
{
|
| 71 |
return cpu_ppc_load_tbl(env);
|
| 72 |
} |
| 73 |
|
| 74 |
target_ulong helper_load_tbu (void)
|
| 75 |
{
|
| 76 |
return cpu_ppc_load_tbu(env);
|
| 77 |
} |
| 78 |
|
| 79 |
target_ulong helper_load_atbl (void)
|
| 80 |
{
|
| 81 |
return cpu_ppc_load_atbl(env);
|
| 82 |
} |
| 83 |
|
| 84 |
target_ulong helper_load_atbu (void)
|
| 85 |
{
|
| 86 |
return cpu_ppc_load_atbu(env);
|
| 87 |
} |
| 88 |
|
| 89 |
target_ulong helper_load_601_rtcl (void)
|
| 90 |
{
|
| 91 |
return cpu_ppc601_load_rtcl(env);
|
| 92 |
} |
| 93 |
|
| 94 |
target_ulong helper_load_601_rtcu (void)
|
| 95 |
{
|
| 96 |
return cpu_ppc601_load_rtcu(env);
|
| 97 |
} |
| 98 |
|
| 99 |
#if !defined(CONFIG_USER_ONLY)
|
| 100 |
#if defined (TARGET_PPC64)
|
| 101 |
void helper_store_asr (target_ulong val)
|
| 102 |
{
|
| 103 |
ppc_store_asr(env, val); |
| 104 |
} |
| 105 |
#endif
|
| 106 |
|
| 107 |
void helper_store_sdr1 (target_ulong val)
|
| 108 |
{
|
| 109 |
ppc_store_sdr1(env, val); |
| 110 |
} |
| 111 |
|
| 112 |
void helper_store_tbl (target_ulong val)
|
| 113 |
{
|
| 114 |
cpu_ppc_store_tbl(env, val); |
| 115 |
} |
| 116 |
|
| 117 |
void helper_store_tbu (target_ulong val)
|
| 118 |
{
|
| 119 |
cpu_ppc_store_tbu(env, val); |
| 120 |
} |
| 121 |
|
| 122 |
void helper_store_atbl (target_ulong val)
|
| 123 |
{
|
| 124 |
cpu_ppc_store_atbl(env, val); |
| 125 |
} |
| 126 |
|
| 127 |
void helper_store_atbu (target_ulong val)
|
| 128 |
{
|
| 129 |
cpu_ppc_store_atbu(env, val); |
| 130 |
} |
| 131 |
|
| 132 |
void helper_store_601_rtcl (target_ulong val)
|
| 133 |
{
|
| 134 |
cpu_ppc601_store_rtcl(env, val); |
| 135 |
} |
| 136 |
|
| 137 |
void helper_store_601_rtcu (target_ulong val)
|
| 138 |
{
|
| 139 |
cpu_ppc601_store_rtcu(env, val); |
| 140 |
} |
| 141 |
|
| 142 |
target_ulong helper_load_decr (void)
|
| 143 |
{
|
| 144 |
return cpu_ppc_load_decr(env);
|
| 145 |
} |
| 146 |
|
| 147 |
void helper_store_decr (target_ulong val)
|
| 148 |
{
|
| 149 |
cpu_ppc_store_decr(env, val); |
| 150 |
} |
| 151 |
|
| 152 |
void helper_store_hid0_601 (target_ulong val)
|
| 153 |
{
|
| 154 |
target_ulong hid0; |
| 155 |
|
| 156 |
hid0 = env->spr[SPR_HID0]; |
| 157 |
if ((val ^ hid0) & 0x00000008) { |
| 158 |
/* Change current endianness */
|
| 159 |
env->hflags &= ~(1 << MSR_LE);
|
| 160 |
env->hflags_nmsr &= ~(1 << MSR_LE);
|
| 161 |
env->hflags_nmsr |= (1 << MSR_LE) & (((val >> 3) & 1) << MSR_LE); |
| 162 |
env->hflags |= env->hflags_nmsr; |
| 163 |
qemu_log("%s: set endianness to %c => " ADDRX "\n", |
| 164 |
__func__, val & 0x8 ? 'l' : 'b', env->hflags); |
| 165 |
} |
| 166 |
env->spr[SPR_HID0] = (uint32_t)val; |
| 167 |
} |
| 168 |
|
| 169 |
void helper_store_403_pbr (uint32_t num, target_ulong value)
|
| 170 |
{
|
| 171 |
if (likely(env->pb[num] != value)) {
|
| 172 |
env->pb[num] = value; |
| 173 |
/* Should be optimized */
|
| 174 |
tlb_flush(env, 1);
|
| 175 |
} |
| 176 |
} |
| 177 |
|
| 178 |
target_ulong helper_load_40x_pit (void)
|
| 179 |
{
|
| 180 |
return load_40x_pit(env);
|
| 181 |
} |
| 182 |
|
| 183 |
void helper_store_40x_pit (target_ulong val)
|
| 184 |
{
|
| 185 |
store_40x_pit(env, val); |
| 186 |
} |
| 187 |
|
| 188 |
void helper_store_40x_dbcr0 (target_ulong val)
|
| 189 |
{
|
| 190 |
store_40x_dbcr0(env, val); |
| 191 |
} |
| 192 |
|
| 193 |
void helper_store_40x_sler (target_ulong val)
|
| 194 |
{
|
| 195 |
store_40x_sler(env, val); |
| 196 |
} |
| 197 |
|
| 198 |
void helper_store_booke_tcr (target_ulong val)
|
| 199 |
{
|
| 200 |
store_booke_tcr(env, val); |
| 201 |
} |
| 202 |
|
| 203 |
void helper_store_booke_tsr (target_ulong val)
|
| 204 |
{
|
| 205 |
store_booke_tsr(env, val); |
| 206 |
} |
| 207 |
|
| 208 |
void helper_store_ibatu (uint32_t nr, target_ulong val)
|
| 209 |
{
|
| 210 |
ppc_store_ibatu(env, nr, val); |
| 211 |
} |
| 212 |
|
| 213 |
void helper_store_ibatl (uint32_t nr, target_ulong val)
|
| 214 |
{
|
| 215 |
ppc_store_ibatl(env, nr, val); |
| 216 |
} |
| 217 |
|
| 218 |
void helper_store_dbatu (uint32_t nr, target_ulong val)
|
| 219 |
{
|
| 220 |
ppc_store_dbatu(env, nr, val); |
| 221 |
} |
| 222 |
|
| 223 |
void helper_store_dbatl (uint32_t nr, target_ulong val)
|
| 224 |
{
|
| 225 |
ppc_store_dbatl(env, nr, val); |
| 226 |
} |
| 227 |
|
| 228 |
void helper_store_601_batl (uint32_t nr, target_ulong val)
|
| 229 |
{
|
| 230 |
ppc_store_ibatl_601(env, nr, val); |
| 231 |
} |
| 232 |
|
| 233 |
void helper_store_601_batu (uint32_t nr, target_ulong val)
|
| 234 |
{
|
| 235 |
ppc_store_ibatu_601(env, nr, val); |
| 236 |
} |
| 237 |
#endif
|
| 238 |
|
| 239 |
/*****************************************************************************/
|
| 240 |
/* Memory load and stores */
|
| 241 |
|
| 242 |
static always_inline target_ulong addr_add(target_ulong addr, target_long arg)
|
| 243 |
{
|
| 244 |
#if defined(TARGET_PPC64)
|
| 245 |
if (!msr_sf)
|
| 246 |
return (uint32_t)(addr + arg);
|
| 247 |
else
|
| 248 |
#endif
|
| 249 |
return addr + arg;
|
| 250 |
} |
| 251 |
|
| 252 |
void helper_lmw (target_ulong addr, uint32_t reg)
|
| 253 |
{
|
| 254 |
for (; reg < 32; reg++) { |
| 255 |
if (msr_le)
|
| 256 |
env->gpr[reg] = bswap32(ldl(addr)); |
| 257 |
else
|
| 258 |
env->gpr[reg] = ldl(addr); |
| 259 |
addr = addr_add(addr, 4);
|
| 260 |
} |
| 261 |
} |
| 262 |
|
| 263 |
void helper_stmw (target_ulong addr, uint32_t reg)
|
| 264 |
{
|
| 265 |
for (; reg < 32; reg++) { |
| 266 |
if (msr_le)
|
| 267 |
stl(addr, bswap32((uint32_t)env->gpr[reg])); |
| 268 |
else
|
| 269 |
stl(addr, (uint32_t)env->gpr[reg]); |
| 270 |
addr = addr_add(addr, 4);
|
| 271 |
} |
| 272 |
} |
| 273 |
|
| 274 |
void helper_lsw(target_ulong addr, uint32_t nb, uint32_t reg)
|
| 275 |
{
|
| 276 |
int sh;
|
| 277 |
for (; nb > 3; nb -= 4) { |
| 278 |
env->gpr[reg] = ldl(addr); |
| 279 |
reg = (reg + 1) % 32; |
| 280 |
addr = addr_add(addr, 4);
|
| 281 |
} |
| 282 |
if (unlikely(nb > 0)) { |
| 283 |
env->gpr[reg] = 0;
|
| 284 |
for (sh = 24; nb > 0; nb--, sh -= 8) { |
| 285 |
env->gpr[reg] |= ldub(addr) << sh; |
| 286 |
addr = addr_add(addr, 1);
|
| 287 |
} |
| 288 |
} |
| 289 |
} |
| 290 |
/* PPC32 specification says we must generate an exception if
|
| 291 |
* rA is in the range of registers to be loaded.
|
| 292 |
* In an other hand, IBM says this is valid, but rA won't be loaded.
|
| 293 |
* For now, I'll follow the spec...
|
| 294 |
*/
|
| 295 |
void helper_lswx(target_ulong addr, uint32_t reg, uint32_t ra, uint32_t rb)
|
| 296 |
{
|
| 297 |
if (likely(xer_bc != 0)) { |
| 298 |
if (unlikely((ra != 0 && reg < ra && (reg + xer_bc) > ra) || |
| 299 |
(reg < rb && (reg + xer_bc) > rb))) {
|
| 300 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
| 301 |
POWERPC_EXCP_INVAL | |
| 302 |
POWERPC_EXCP_INVAL_LSWX); |
| 303 |
} else {
|
| 304 |
helper_lsw(addr, xer_bc, reg); |
| 305 |
} |
| 306 |
} |
| 307 |
} |
| 308 |
|
| 309 |
void helper_stsw(target_ulong addr, uint32_t nb, uint32_t reg)
|
| 310 |
{
|
| 311 |
int sh;
|
| 312 |
for (; nb > 3; nb -= 4) { |
| 313 |
stl(addr, env->gpr[reg]); |
| 314 |
reg = (reg + 1) % 32; |
| 315 |
addr = addr_add(addr, 4);
|
| 316 |
} |
| 317 |
if (unlikely(nb > 0)) { |
| 318 |
for (sh = 24; nb > 0; nb--, sh -= 8) { |
| 319 |
stb(addr, (env->gpr[reg] >> sh) & 0xFF);
|
| 320 |
addr = addr_add(addr, 1);
|
| 321 |
} |
| 322 |
} |
| 323 |
} |
| 324 |
|
| 325 |
static void do_dcbz(target_ulong addr, int dcache_line_size) |
| 326 |
{
|
| 327 |
addr &= ~(dcache_line_size - 1);
|
| 328 |
int i;
|
| 329 |
for (i = 0 ; i < dcache_line_size ; i += 4) { |
| 330 |
stl(addr + i , 0);
|
| 331 |
} |
| 332 |
if (env->reserve == addr)
|
| 333 |
env->reserve = (target_ulong)-1ULL;
|
| 334 |
} |
| 335 |
|
| 336 |
void helper_dcbz(target_ulong addr)
|
| 337 |
{
|
| 338 |
do_dcbz(addr, env->dcache_line_size); |
| 339 |
} |
| 340 |
|
| 341 |
void helper_dcbz_970(target_ulong addr)
|
| 342 |
{
|
| 343 |
if (((env->spr[SPR_970_HID5] >> 7) & 0x3) == 1) |
| 344 |
do_dcbz(addr, 32);
|
| 345 |
else
|
| 346 |
do_dcbz(addr, env->dcache_line_size); |
| 347 |
} |
| 348 |
|
| 349 |
void helper_icbi(target_ulong addr)
|
| 350 |
{
|
| 351 |
uint32_t tmp; |
| 352 |
|
| 353 |
addr &= ~(env->dcache_line_size - 1);
|
| 354 |
/* Invalidate one cache line :
|
| 355 |
* PowerPC specification says this is to be treated like a load
|
| 356 |
* (not a fetch) by the MMU. To be sure it will be so,
|
| 357 |
* do the load "by hand".
|
| 358 |
*/
|
| 359 |
tmp = ldl(addr); |
| 360 |
tb_invalidate_page_range(addr, addr + env->icache_line_size); |
| 361 |
} |
| 362 |
|
| 363 |
// XXX: to be tested
|
| 364 |
target_ulong helper_lscbx (target_ulong addr, uint32_t reg, uint32_t ra, uint32_t rb) |
| 365 |
{
|
| 366 |
int i, c, d;
|
| 367 |
d = 24;
|
| 368 |
for (i = 0; i < xer_bc; i++) { |
| 369 |
c = ldub(addr); |
| 370 |
addr = addr_add(addr, 1);
|
| 371 |
/* ra (if not 0) and rb are never modified */
|
| 372 |
if (likely(reg != rb && (ra == 0 || reg != ra))) { |
| 373 |
env->gpr[reg] = (env->gpr[reg] & ~(0xFF << d)) | (c << d);
|
| 374 |
} |
| 375 |
if (unlikely(c == xer_cmp))
|
| 376 |
break;
|
| 377 |
if (likely(d != 0)) { |
| 378 |
d -= 8;
|
| 379 |
} else {
|
| 380 |
d = 24;
|
| 381 |
reg++; |
| 382 |
reg = reg & 0x1F;
|
| 383 |
} |
| 384 |
} |
| 385 |
return i;
|
| 386 |
} |
| 387 |
|
| 388 |
/*****************************************************************************/
|
| 389 |
/* Fixed point operations helpers */
|
| 390 |
#if defined(TARGET_PPC64)
|
| 391 |
|
| 392 |
/* multiply high word */
|
| 393 |
uint64_t helper_mulhd (uint64_t arg1, uint64_t arg2) |
| 394 |
{
|
| 395 |
uint64_t tl, th; |
| 396 |
|
| 397 |
muls64(&tl, &th, arg1, arg2); |
| 398 |
return th;
|
| 399 |
} |
| 400 |
|
| 401 |
/* multiply high word unsigned */
|
| 402 |
uint64_t helper_mulhdu (uint64_t arg1, uint64_t arg2) |
| 403 |
{
|
| 404 |
uint64_t tl, th; |
| 405 |
|
| 406 |
mulu64(&tl, &th, arg1, arg2); |
| 407 |
return th;
|
| 408 |
} |
| 409 |
|
| 410 |
uint64_t helper_mulldo (uint64_t arg1, uint64_t arg2) |
| 411 |
{
|
| 412 |
int64_t th; |
| 413 |
uint64_t tl; |
| 414 |
|
| 415 |
muls64(&tl, (uint64_t *)&th, arg1, arg2); |
| 416 |
/* If th != 0 && th != -1, then we had an overflow */
|
| 417 |
if (likely((uint64_t)(th + 1) <= 1)) { |
| 418 |
env->xer &= ~(1 << XER_OV);
|
| 419 |
} else {
|
| 420 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
| 421 |
} |
| 422 |
return (int64_t)tl;
|
| 423 |
} |
| 424 |
#endif
|
| 425 |
|
| 426 |
target_ulong helper_cntlzw (target_ulong t) |
| 427 |
{
|
| 428 |
return clz32(t);
|
| 429 |
} |
| 430 |
|
| 431 |
#if defined(TARGET_PPC64)
|
| 432 |
target_ulong helper_cntlzd (target_ulong t) |
| 433 |
{
|
| 434 |
return clz64(t);
|
| 435 |
} |
| 436 |
#endif
|
| 437 |
|
| 438 |
/* shift right arithmetic helper */
|
| 439 |
target_ulong helper_sraw (target_ulong value, target_ulong shift) |
| 440 |
{
|
| 441 |
int32_t ret; |
| 442 |
|
| 443 |
if (likely(!(shift & 0x20))) { |
| 444 |
if (likely((uint32_t)shift != 0)) { |
| 445 |
shift &= 0x1f;
|
| 446 |
ret = (int32_t)value >> shift; |
| 447 |
if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) { |
| 448 |
env->xer &= ~(1 << XER_CA);
|
| 449 |
} else {
|
| 450 |
env->xer |= (1 << XER_CA);
|
| 451 |
} |
| 452 |
} else {
|
| 453 |
ret = (int32_t)value; |
| 454 |
env->xer &= ~(1 << XER_CA);
|
| 455 |
} |
| 456 |
} else {
|
| 457 |
ret = (int32_t)value >> 31;
|
| 458 |
if (ret) {
|
| 459 |
env->xer |= (1 << XER_CA);
|
| 460 |
} else {
|
| 461 |
env->xer &= ~(1 << XER_CA);
|
| 462 |
} |
| 463 |
} |
| 464 |
return (target_long)ret;
|
| 465 |
} |
| 466 |
|
| 467 |
#if defined(TARGET_PPC64)
|
| 468 |
target_ulong helper_srad (target_ulong value, target_ulong shift) |
| 469 |
{
|
| 470 |
int64_t ret; |
| 471 |
|
| 472 |
if (likely(!(shift & 0x40))) { |
| 473 |
if (likely((uint64_t)shift != 0)) { |
| 474 |
shift &= 0x3f;
|
| 475 |
ret = (int64_t)value >> shift; |
| 476 |
if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) { |
| 477 |
env->xer &= ~(1 << XER_CA);
|
| 478 |
} else {
|
| 479 |
env->xer |= (1 << XER_CA);
|
| 480 |
} |
| 481 |
} else {
|
| 482 |
ret = (int64_t)value; |
| 483 |
env->xer &= ~(1 << XER_CA);
|
| 484 |
} |
| 485 |
} else {
|
| 486 |
ret = (int64_t)value >> 63;
|
| 487 |
if (ret) {
|
| 488 |
env->xer |= (1 << XER_CA);
|
| 489 |
} else {
|
| 490 |
env->xer &= ~(1 << XER_CA);
|
| 491 |
} |
| 492 |
} |
| 493 |
return ret;
|
| 494 |
} |
| 495 |
#endif
|
| 496 |
|
| 497 |
target_ulong helper_popcntb (target_ulong val) |
| 498 |
{
|
| 499 |
val = (val & 0x55555555) + ((val >> 1) & 0x55555555); |
| 500 |
val = (val & 0x33333333) + ((val >> 2) & 0x33333333); |
| 501 |
val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f); |
| 502 |
return val;
|
| 503 |
} |
| 504 |
|
| 505 |
#if defined(TARGET_PPC64)
|
| 506 |
target_ulong helper_popcntb_64 (target_ulong val) |
| 507 |
{
|
| 508 |
val = (val & 0x5555555555555555ULL) + ((val >> 1) & 0x5555555555555555ULL); |
| 509 |
val = (val & 0x3333333333333333ULL) + ((val >> 2) & 0x3333333333333333ULL); |
| 510 |
val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) & 0x0f0f0f0f0f0f0f0fULL); |
| 511 |
return val;
|
| 512 |
} |
| 513 |
#endif
|
| 514 |
|
| 515 |
/*****************************************************************************/
|
| 516 |
/* Floating point operations helpers */
|
| 517 |
uint64_t helper_float32_to_float64(uint32_t arg) |
| 518 |
{
|
| 519 |
CPU_FloatU f; |
| 520 |
CPU_DoubleU d; |
| 521 |
f.l = arg; |
| 522 |
d.d = float32_to_float64(f.f, &env->fp_status); |
| 523 |
return d.ll;
|
| 524 |
} |
| 525 |
|
| 526 |
uint32_t helper_float64_to_float32(uint64_t arg) |
| 527 |
{
|
| 528 |
CPU_FloatU f; |
| 529 |
CPU_DoubleU d; |
| 530 |
d.ll = arg; |
| 531 |
f.f = float64_to_float32(d.d, &env->fp_status); |
| 532 |
return f.l;
|
| 533 |
} |
| 534 |
|
| 535 |
static always_inline int isden (float64 d) |
| 536 |
{
|
| 537 |
CPU_DoubleU u; |
| 538 |
|
| 539 |
u.d = d; |
| 540 |
|
| 541 |
return ((u.ll >> 52) & 0x7FF) == 0; |
| 542 |
} |
| 543 |
|
| 544 |
uint32_t helper_compute_fprf (uint64_t arg, uint32_t set_fprf) |
| 545 |
{
|
| 546 |
CPU_DoubleU farg; |
| 547 |
int isneg;
|
| 548 |
int ret;
|
| 549 |
farg.ll = arg; |
| 550 |
isneg = float64_is_neg(farg.d); |
| 551 |
if (unlikely(float64_is_nan(farg.d))) {
|
| 552 |
if (float64_is_signaling_nan(farg.d)) {
|
| 553 |
/* Signaling NaN: flags are undefined */
|
| 554 |
ret = 0x00;
|
| 555 |
} else {
|
| 556 |
/* Quiet NaN */
|
| 557 |
ret = 0x11;
|
| 558 |
} |
| 559 |
} else if (unlikely(float64_is_infinity(farg.d))) { |
| 560 |
/* +/- infinity */
|
| 561 |
if (isneg)
|
| 562 |
ret = 0x09;
|
| 563 |
else
|
| 564 |
ret = 0x05;
|
| 565 |
} else {
|
| 566 |
if (float64_is_zero(farg.d)) {
|
| 567 |
/* +/- zero */
|
| 568 |
if (isneg)
|
| 569 |
ret = 0x12;
|
| 570 |
else
|
| 571 |
ret = 0x02;
|
| 572 |
} else {
|
| 573 |
if (isden(farg.d)) {
|
| 574 |
/* Denormalized numbers */
|
| 575 |
ret = 0x10;
|
| 576 |
} else {
|
| 577 |
/* Normalized numbers */
|
| 578 |
ret = 0x00;
|
| 579 |
} |
| 580 |
if (isneg) {
|
| 581 |
ret |= 0x08;
|
| 582 |
} else {
|
| 583 |
ret |= 0x04;
|
| 584 |
} |
| 585 |
} |
| 586 |
} |
| 587 |
if (set_fprf) {
|
| 588 |
/* We update FPSCR_FPRF */
|
| 589 |
env->fpscr &= ~(0x1F << FPSCR_FPRF);
|
| 590 |
env->fpscr |= ret << FPSCR_FPRF; |
| 591 |
} |
| 592 |
/* We just need fpcc to update Rc1 */
|
| 593 |
return ret & 0xF; |
| 594 |
} |
| 595 |
|
| 596 |
/* Floating-point invalid operations exception */
|
| 597 |
static always_inline uint64_t fload_invalid_op_excp (int op) |
| 598 |
{
|
| 599 |
uint64_t ret = 0;
|
| 600 |
int ve;
|
| 601 |
|
| 602 |
ve = fpscr_ve; |
| 603 |
switch (op) {
|
| 604 |
case POWERPC_EXCP_FP_VXSNAN:
|
| 605 |
env->fpscr |= 1 << FPSCR_VXSNAN;
|
| 606 |
break;
|
| 607 |
case POWERPC_EXCP_FP_VXSOFT:
|
| 608 |
env->fpscr |= 1 << FPSCR_VXSOFT;
|
| 609 |
break;
|
| 610 |
case POWERPC_EXCP_FP_VXISI:
|
| 611 |
/* Magnitude subtraction of infinities */
|
| 612 |
env->fpscr |= 1 << FPSCR_VXISI;
|
| 613 |
goto update_arith;
|
| 614 |
case POWERPC_EXCP_FP_VXIDI:
|
| 615 |
/* Division of infinity by infinity */
|
| 616 |
env->fpscr |= 1 << FPSCR_VXIDI;
|
| 617 |
goto update_arith;
|
| 618 |
case POWERPC_EXCP_FP_VXZDZ:
|
| 619 |
/* Division of zero by zero */
|
| 620 |
env->fpscr |= 1 << FPSCR_VXZDZ;
|
| 621 |
goto update_arith;
|
| 622 |
case POWERPC_EXCP_FP_VXIMZ:
|
| 623 |
/* Multiplication of zero by infinity */
|
| 624 |
env->fpscr |= 1 << FPSCR_VXIMZ;
|
| 625 |
goto update_arith;
|
| 626 |
case POWERPC_EXCP_FP_VXVC:
|
| 627 |
/* Ordered comparison of NaN */
|
| 628 |
env->fpscr |= 1 << FPSCR_VXVC;
|
| 629 |
env->fpscr &= ~(0xF << FPSCR_FPCC);
|
| 630 |
env->fpscr |= 0x11 << FPSCR_FPCC;
|
| 631 |
/* We must update the target FPR before raising the exception */
|
| 632 |
if (ve != 0) { |
| 633 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
| 634 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_VXVC; |
| 635 |
/* Update the floating-point enabled exception summary */
|
| 636 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 637 |
/* Exception is differed */
|
| 638 |
ve = 0;
|
| 639 |
} |
| 640 |
break;
|
| 641 |
case POWERPC_EXCP_FP_VXSQRT:
|
| 642 |
/* Square root of a negative number */
|
| 643 |
env->fpscr |= 1 << FPSCR_VXSQRT;
|
| 644 |
update_arith:
|
| 645 |
env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI)); |
| 646 |
if (ve == 0) { |
| 647 |
/* Set the result to quiet NaN */
|
| 648 |
ret = 0xFFF8000000000000ULL;
|
| 649 |
env->fpscr &= ~(0xF << FPSCR_FPCC);
|
| 650 |
env->fpscr |= 0x11 << FPSCR_FPCC;
|
| 651 |
} |
| 652 |
break;
|
| 653 |
case POWERPC_EXCP_FP_VXCVI:
|
| 654 |
/* Invalid conversion */
|
| 655 |
env->fpscr |= 1 << FPSCR_VXCVI;
|
| 656 |
env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI)); |
| 657 |
if (ve == 0) { |
| 658 |
/* Set the result to quiet NaN */
|
| 659 |
ret = 0xFFF8000000000000ULL;
|
| 660 |
env->fpscr &= ~(0xF << FPSCR_FPCC);
|
| 661 |
env->fpscr |= 0x11 << FPSCR_FPCC;
|
| 662 |
} |
| 663 |
break;
|
| 664 |
} |
| 665 |
/* Update the floating-point invalid operation summary */
|
| 666 |
env->fpscr |= 1 << FPSCR_VX;
|
| 667 |
/* Update the floating-point exception summary */
|
| 668 |
env->fpscr |= 1 << FPSCR_FX;
|
| 669 |
if (ve != 0) { |
| 670 |
/* Update the floating-point enabled exception summary */
|
| 671 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 672 |
if (msr_fe0 != 0 || msr_fe1 != 0) |
| 673 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_FP | op); |
| 674 |
} |
| 675 |
return ret;
|
| 676 |
} |
| 677 |
|
| 678 |
static always_inline void float_zero_divide_excp (void) |
| 679 |
{
|
| 680 |
env->fpscr |= 1 << FPSCR_ZX;
|
| 681 |
env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI)); |
| 682 |
/* Update the floating-point exception summary */
|
| 683 |
env->fpscr |= 1 << FPSCR_FX;
|
| 684 |
if (fpscr_ze != 0) { |
| 685 |
/* Update the floating-point enabled exception summary */
|
| 686 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 687 |
if (msr_fe0 != 0 || msr_fe1 != 0) { |
| 688 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
| 689 |
POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX); |
| 690 |
} |
| 691 |
} |
| 692 |
} |
| 693 |
|
| 694 |
static always_inline void float_overflow_excp (void) |
| 695 |
{
|
| 696 |
env->fpscr |= 1 << FPSCR_OX;
|
| 697 |
/* Update the floating-point exception summary */
|
| 698 |
env->fpscr |= 1 << FPSCR_FX;
|
| 699 |
if (fpscr_oe != 0) { |
| 700 |
/* XXX: should adjust the result */
|
| 701 |
/* Update the floating-point enabled exception summary */
|
| 702 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 703 |
/* We must update the target FPR before raising the exception */
|
| 704 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
| 705 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX; |
| 706 |
} else {
|
| 707 |
env->fpscr |= 1 << FPSCR_XX;
|
| 708 |
env->fpscr |= 1 << FPSCR_FI;
|
| 709 |
} |
| 710 |
} |
| 711 |
|
| 712 |
static always_inline void float_underflow_excp (void) |
| 713 |
{
|
| 714 |
env->fpscr |= 1 << FPSCR_UX;
|
| 715 |
/* Update the floating-point exception summary */
|
| 716 |
env->fpscr |= 1 << FPSCR_FX;
|
| 717 |
if (fpscr_ue != 0) { |
| 718 |
/* XXX: should adjust the result */
|
| 719 |
/* Update the floating-point enabled exception summary */
|
| 720 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 721 |
/* We must update the target FPR before raising the exception */
|
| 722 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
| 723 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX; |
| 724 |
} |
| 725 |
} |
| 726 |
|
| 727 |
static always_inline void float_inexact_excp (void) |
| 728 |
{
|
| 729 |
env->fpscr |= 1 << FPSCR_XX;
|
| 730 |
/* Update the floating-point exception summary */
|
| 731 |
env->fpscr |= 1 << FPSCR_FX;
|
| 732 |
if (fpscr_xe != 0) { |
| 733 |
/* Update the floating-point enabled exception summary */
|
| 734 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 735 |
/* We must update the target FPR before raising the exception */
|
| 736 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
| 737 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX; |
| 738 |
} |
| 739 |
} |
| 740 |
|
| 741 |
static always_inline void fpscr_set_rounding_mode (void) |
| 742 |
{
|
| 743 |
int rnd_type;
|
| 744 |
|
| 745 |
/* Set rounding mode */
|
| 746 |
switch (fpscr_rn) {
|
| 747 |
case 0: |
| 748 |
/* Best approximation (round to nearest) */
|
| 749 |
rnd_type = float_round_nearest_even; |
| 750 |
break;
|
| 751 |
case 1: |
| 752 |
/* Smaller magnitude (round toward zero) */
|
| 753 |
rnd_type = float_round_to_zero; |
| 754 |
break;
|
| 755 |
case 2: |
| 756 |
/* Round toward +infinite */
|
| 757 |
rnd_type = float_round_up; |
| 758 |
break;
|
| 759 |
default:
|
| 760 |
case 3: |
| 761 |
/* Round toward -infinite */
|
| 762 |
rnd_type = float_round_down; |
| 763 |
break;
|
| 764 |
} |
| 765 |
set_float_rounding_mode(rnd_type, &env->fp_status); |
| 766 |
} |
| 767 |
|
| 768 |
void helper_fpscr_clrbit (uint32_t bit)
|
| 769 |
{
|
| 770 |
int prev;
|
| 771 |
|
| 772 |
prev = (env->fpscr >> bit) & 1;
|
| 773 |
env->fpscr &= ~(1 << bit);
|
| 774 |
if (prev == 1) { |
| 775 |
switch (bit) {
|
| 776 |
case FPSCR_RN1:
|
| 777 |
case FPSCR_RN:
|
| 778 |
fpscr_set_rounding_mode(); |
| 779 |
break;
|
| 780 |
default:
|
| 781 |
break;
|
| 782 |
} |
| 783 |
} |
| 784 |
} |
| 785 |
|
| 786 |
void helper_fpscr_setbit (uint32_t bit)
|
| 787 |
{
|
| 788 |
int prev;
|
| 789 |
|
| 790 |
prev = (env->fpscr >> bit) & 1;
|
| 791 |
env->fpscr |= 1 << bit;
|
| 792 |
if (prev == 0) { |
| 793 |
switch (bit) {
|
| 794 |
case FPSCR_VX:
|
| 795 |
env->fpscr |= 1 << FPSCR_FX;
|
| 796 |
if (fpscr_ve)
|
| 797 |
goto raise_ve;
|
| 798 |
case FPSCR_OX:
|
| 799 |
env->fpscr |= 1 << FPSCR_FX;
|
| 800 |
if (fpscr_oe)
|
| 801 |
goto raise_oe;
|
| 802 |
break;
|
| 803 |
case FPSCR_UX:
|
| 804 |
env->fpscr |= 1 << FPSCR_FX;
|
| 805 |
if (fpscr_ue)
|
| 806 |
goto raise_ue;
|
| 807 |
break;
|
| 808 |
case FPSCR_ZX:
|
| 809 |
env->fpscr |= 1 << FPSCR_FX;
|
| 810 |
if (fpscr_ze)
|
| 811 |
goto raise_ze;
|
| 812 |
break;
|
| 813 |
case FPSCR_XX:
|
| 814 |
env->fpscr |= 1 << FPSCR_FX;
|
| 815 |
if (fpscr_xe)
|
| 816 |
goto raise_xe;
|
| 817 |
break;
|
| 818 |
case FPSCR_VXSNAN:
|
| 819 |
case FPSCR_VXISI:
|
| 820 |
case FPSCR_VXIDI:
|
| 821 |
case FPSCR_VXZDZ:
|
| 822 |
case FPSCR_VXIMZ:
|
| 823 |
case FPSCR_VXVC:
|
| 824 |
case FPSCR_VXSOFT:
|
| 825 |
case FPSCR_VXSQRT:
|
| 826 |
case FPSCR_VXCVI:
|
| 827 |
env->fpscr |= 1 << FPSCR_VX;
|
| 828 |
env->fpscr |= 1 << FPSCR_FX;
|
| 829 |
if (fpscr_ve != 0) |
| 830 |
goto raise_ve;
|
| 831 |
break;
|
| 832 |
case FPSCR_VE:
|
| 833 |
if (fpscr_vx != 0) { |
| 834 |
raise_ve:
|
| 835 |
env->error_code = POWERPC_EXCP_FP; |
| 836 |
if (fpscr_vxsnan)
|
| 837 |
env->error_code |= POWERPC_EXCP_FP_VXSNAN; |
| 838 |
if (fpscr_vxisi)
|
| 839 |
env->error_code |= POWERPC_EXCP_FP_VXISI; |
| 840 |
if (fpscr_vxidi)
|
| 841 |
env->error_code |= POWERPC_EXCP_FP_VXIDI; |
| 842 |
if (fpscr_vxzdz)
|
| 843 |
env->error_code |= POWERPC_EXCP_FP_VXZDZ; |
| 844 |
if (fpscr_vximz)
|
| 845 |
env->error_code |= POWERPC_EXCP_FP_VXIMZ; |
| 846 |
if (fpscr_vxvc)
|
| 847 |
env->error_code |= POWERPC_EXCP_FP_VXVC; |
| 848 |
if (fpscr_vxsoft)
|
| 849 |
env->error_code |= POWERPC_EXCP_FP_VXSOFT; |
| 850 |
if (fpscr_vxsqrt)
|
| 851 |
env->error_code |= POWERPC_EXCP_FP_VXSQRT; |
| 852 |
if (fpscr_vxcvi)
|
| 853 |
env->error_code |= POWERPC_EXCP_FP_VXCVI; |
| 854 |
goto raise_excp;
|
| 855 |
} |
| 856 |
break;
|
| 857 |
case FPSCR_OE:
|
| 858 |
if (fpscr_ox != 0) { |
| 859 |
raise_oe:
|
| 860 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX; |
| 861 |
goto raise_excp;
|
| 862 |
} |
| 863 |
break;
|
| 864 |
case FPSCR_UE:
|
| 865 |
if (fpscr_ux != 0) { |
| 866 |
raise_ue:
|
| 867 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX; |
| 868 |
goto raise_excp;
|
| 869 |
} |
| 870 |
break;
|
| 871 |
case FPSCR_ZE:
|
| 872 |
if (fpscr_zx != 0) { |
| 873 |
raise_ze:
|
| 874 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX; |
| 875 |
goto raise_excp;
|
| 876 |
} |
| 877 |
break;
|
| 878 |
case FPSCR_XE:
|
| 879 |
if (fpscr_xx != 0) { |
| 880 |
raise_xe:
|
| 881 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX; |
| 882 |
goto raise_excp;
|
| 883 |
} |
| 884 |
break;
|
| 885 |
case FPSCR_RN1:
|
| 886 |
case FPSCR_RN:
|
| 887 |
fpscr_set_rounding_mode(); |
| 888 |
break;
|
| 889 |
default:
|
| 890 |
break;
|
| 891 |
raise_excp:
|
| 892 |
/* Update the floating-point enabled exception summary */
|
| 893 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 894 |
/* We have to update Rc1 before raising the exception */
|
| 895 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
| 896 |
break;
|
| 897 |
} |
| 898 |
} |
| 899 |
} |
| 900 |
|
| 901 |
void helper_store_fpscr (uint64_t arg, uint32_t mask)
|
| 902 |
{
|
| 903 |
/*
|
| 904 |
* We use only the 32 LSB of the incoming fpr
|
| 905 |
*/
|
| 906 |
uint32_t prev, new; |
| 907 |
int i;
|
| 908 |
|
| 909 |
prev = env->fpscr; |
| 910 |
new = (uint32_t)arg; |
| 911 |
new &= ~0x60000000;
|
| 912 |
new |= prev & 0x60000000;
|
| 913 |
for (i = 0; i < 8; i++) { |
| 914 |
if (mask & (1 << i)) { |
| 915 |
env->fpscr &= ~(0xF << (4 * i)); |
| 916 |
env->fpscr |= new & (0xF << (4 * i)); |
| 917 |
} |
| 918 |
} |
| 919 |
/* Update VX and FEX */
|
| 920 |
if (fpscr_ix != 0) |
| 921 |
env->fpscr |= 1 << FPSCR_VX;
|
| 922 |
else
|
| 923 |
env->fpscr &= ~(1 << FPSCR_VX);
|
| 924 |
if ((fpscr_ex & fpscr_eex) != 0) { |
| 925 |
env->fpscr |= 1 << FPSCR_FEX;
|
| 926 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
| 927 |
/* XXX: we should compute it properly */
|
| 928 |
env->error_code = POWERPC_EXCP_FP; |
| 929 |
} |
| 930 |
else
|
| 931 |
env->fpscr &= ~(1 << FPSCR_FEX);
|
| 932 |
fpscr_set_rounding_mode(); |
| 933 |
} |
| 934 |
|
| 935 |
void helper_float_check_status (void) |
| 936 |
{
|
| 937 |
#ifdef CONFIG_SOFTFLOAT
|
| 938 |
if (env->exception_index == POWERPC_EXCP_PROGRAM &&
|
| 939 |
(env->error_code & POWERPC_EXCP_FP)) {
|
| 940 |
/* Differred floating-point exception after target FPR update */
|
| 941 |
if (msr_fe0 != 0 || msr_fe1 != 0) |
| 942 |
helper_raise_exception_err(env->exception_index, env->error_code); |
| 943 |
} else {
|
| 944 |
int status = get_float_exception_flags(&env->fp_status);
|
| 945 |
if (status & float_flag_divbyzero) {
|
| 946 |
float_zero_divide_excp(); |
| 947 |
} else if (status & float_flag_overflow) { |
| 948 |
float_overflow_excp(); |
| 949 |
} else if (status & float_flag_underflow) { |
| 950 |
float_underflow_excp(); |
| 951 |
} else if (status & float_flag_inexact) { |
| 952 |
float_inexact_excp(); |
| 953 |
} |
| 954 |
} |
| 955 |
#else
|
| 956 |
if (env->exception_index == POWERPC_EXCP_PROGRAM &&
|
| 957 |
(env->error_code & POWERPC_EXCP_FP)) {
|
| 958 |
/* Differred floating-point exception after target FPR update */
|
| 959 |
if (msr_fe0 != 0 || msr_fe1 != 0) |
| 960 |
helper_raise_exception_err(env->exception_index, env->error_code); |
| 961 |
} |
| 962 |
#endif
|
| 963 |
} |
| 964 |
|
| 965 |
#ifdef CONFIG_SOFTFLOAT
|
| 966 |
void helper_reset_fpstatus (void) |
| 967 |
{
|
| 968 |
set_float_exception_flags(0, &env->fp_status);
|
| 969 |
} |
| 970 |
#endif
|
| 971 |
|
| 972 |
/* fadd - fadd. */
|
| 973 |
uint64_t helper_fadd (uint64_t arg1, uint64_t arg2) |
| 974 |
{
|
| 975 |
CPU_DoubleU farg1, farg2; |
| 976 |
|
| 977 |
farg1.ll = arg1; |
| 978 |
farg2.ll = arg2; |
| 979 |
#if USE_PRECISE_EMULATION
|
| 980 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 981 |
float64_is_signaling_nan(farg2.d))) {
|
| 982 |
/* sNaN addition */
|
| 983 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 984 |
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) && |
| 985 |
float64_is_neg(farg1.d) != float64_is_neg(farg2.d))) {
|
| 986 |
/* Magnitude subtraction of infinities */
|
| 987 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
| 988 |
} else {
|
| 989 |
farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status); |
| 990 |
} |
| 991 |
#else
|
| 992 |
farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status); |
| 993 |
#endif
|
| 994 |
return farg1.ll;
|
| 995 |
} |
| 996 |
|
| 997 |
/* fsub - fsub. */
|
| 998 |
uint64_t helper_fsub (uint64_t arg1, uint64_t arg2) |
| 999 |
{
|
| 1000 |
CPU_DoubleU farg1, farg2; |
| 1001 |
|
| 1002 |
farg1.ll = arg1; |
| 1003 |
farg2.ll = arg2; |
| 1004 |
#if USE_PRECISE_EMULATION
|
| 1005 |
{
|
| 1006 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1007 |
float64_is_signaling_nan(farg2.d))) {
|
| 1008 |
/* sNaN subtraction */
|
| 1009 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1010 |
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) && |
| 1011 |
float64_is_neg(farg1.d) == float64_is_neg(farg2.d))) {
|
| 1012 |
/* Magnitude subtraction of infinities */
|
| 1013 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
| 1014 |
} else {
|
| 1015 |
farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status); |
| 1016 |
} |
| 1017 |
} |
| 1018 |
#else
|
| 1019 |
farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status); |
| 1020 |
#endif
|
| 1021 |
return farg1.ll;
|
| 1022 |
} |
| 1023 |
|
| 1024 |
/* fmul - fmul. */
|
| 1025 |
uint64_t helper_fmul (uint64_t arg1, uint64_t arg2) |
| 1026 |
{
|
| 1027 |
CPU_DoubleU farg1, farg2; |
| 1028 |
|
| 1029 |
farg1.ll = arg1; |
| 1030 |
farg2.ll = arg2; |
| 1031 |
#if USE_PRECISE_EMULATION
|
| 1032 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1033 |
float64_is_signaling_nan(farg2.d))) {
|
| 1034 |
/* sNaN multiplication */
|
| 1035 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1036 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
| 1037 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
| 1038 |
/* Multiplication of zero by infinity */
|
| 1039 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
| 1040 |
} else {
|
| 1041 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
| 1042 |
} |
| 1043 |
#else
|
| 1044 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
| 1045 |
#endif
|
| 1046 |
return farg1.ll;
|
| 1047 |
} |
| 1048 |
|
| 1049 |
/* fdiv - fdiv. */
|
| 1050 |
uint64_t helper_fdiv (uint64_t arg1, uint64_t arg2) |
| 1051 |
{
|
| 1052 |
CPU_DoubleU farg1, farg2; |
| 1053 |
|
| 1054 |
farg1.ll = arg1; |
| 1055 |
farg2.ll = arg2; |
| 1056 |
#if USE_PRECISE_EMULATION
|
| 1057 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1058 |
float64_is_signaling_nan(farg2.d))) {
|
| 1059 |
/* sNaN division */
|
| 1060 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1061 |
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d))) { |
| 1062 |
/* Division of infinity by infinity */
|
| 1063 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIDI); |
| 1064 |
} else if (unlikely(float64_is_zero(farg1.d) && float64_is_zero(farg2.d))) { |
| 1065 |
/* Division of zero by zero */
|
| 1066 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXZDZ); |
| 1067 |
} else {
|
| 1068 |
farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status); |
| 1069 |
} |
| 1070 |
#else
|
| 1071 |
farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status); |
| 1072 |
#endif
|
| 1073 |
return farg1.ll;
|
| 1074 |
} |
| 1075 |
|
| 1076 |
/* fabs */
|
| 1077 |
uint64_t helper_fabs (uint64_t arg) |
| 1078 |
{
|
| 1079 |
CPU_DoubleU farg; |
| 1080 |
|
| 1081 |
farg.ll = arg; |
| 1082 |
farg.d = float64_abs(farg.d); |
| 1083 |
return farg.ll;
|
| 1084 |
} |
| 1085 |
|
| 1086 |
/* fnabs */
|
| 1087 |
uint64_t helper_fnabs (uint64_t arg) |
| 1088 |
{
|
| 1089 |
CPU_DoubleU farg; |
| 1090 |
|
| 1091 |
farg.ll = arg; |
| 1092 |
farg.d = float64_abs(farg.d); |
| 1093 |
farg.d = float64_chs(farg.d); |
| 1094 |
return farg.ll;
|
| 1095 |
} |
| 1096 |
|
| 1097 |
/* fneg */
|
| 1098 |
uint64_t helper_fneg (uint64_t arg) |
| 1099 |
{
|
| 1100 |
CPU_DoubleU farg; |
| 1101 |
|
| 1102 |
farg.ll = arg; |
| 1103 |
farg.d = float64_chs(farg.d); |
| 1104 |
return farg.ll;
|
| 1105 |
} |
| 1106 |
|
| 1107 |
/* fctiw - fctiw. */
|
| 1108 |
uint64_t helper_fctiw (uint64_t arg) |
| 1109 |
{
|
| 1110 |
CPU_DoubleU farg; |
| 1111 |
farg.ll = arg; |
| 1112 |
|
| 1113 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1114 |
/* sNaN conversion */
|
| 1115 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
| 1116 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
| 1117 |
/* qNan / infinity conversion */
|
| 1118 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
| 1119 |
} else {
|
| 1120 |
farg.ll = float64_to_int32(farg.d, &env->fp_status); |
| 1121 |
#if USE_PRECISE_EMULATION
|
| 1122 |
/* XXX: higher bits are not supposed to be significant.
|
| 1123 |
* to make tests easier, return the same as a real PowerPC 750
|
| 1124 |
*/
|
| 1125 |
farg.ll |= 0xFFF80000ULL << 32; |
| 1126 |
#endif
|
| 1127 |
} |
| 1128 |
return farg.ll;
|
| 1129 |
} |
| 1130 |
|
| 1131 |
/* fctiwz - fctiwz. */
|
| 1132 |
uint64_t helper_fctiwz (uint64_t arg) |
| 1133 |
{
|
| 1134 |
CPU_DoubleU farg; |
| 1135 |
farg.ll = arg; |
| 1136 |
|
| 1137 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1138 |
/* sNaN conversion */
|
| 1139 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
| 1140 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
| 1141 |
/* qNan / infinity conversion */
|
| 1142 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
| 1143 |
} else {
|
| 1144 |
farg.ll = float64_to_int32_round_to_zero(farg.d, &env->fp_status); |
| 1145 |
#if USE_PRECISE_EMULATION
|
| 1146 |
/* XXX: higher bits are not supposed to be significant.
|
| 1147 |
* to make tests easier, return the same as a real PowerPC 750
|
| 1148 |
*/
|
| 1149 |
farg.ll |= 0xFFF80000ULL << 32; |
| 1150 |
#endif
|
| 1151 |
} |
| 1152 |
return farg.ll;
|
| 1153 |
} |
| 1154 |
|
| 1155 |
#if defined(TARGET_PPC64)
|
| 1156 |
/* fcfid - fcfid. */
|
| 1157 |
uint64_t helper_fcfid (uint64_t arg) |
| 1158 |
{
|
| 1159 |
CPU_DoubleU farg; |
| 1160 |
farg.d = int64_to_float64(arg, &env->fp_status); |
| 1161 |
return farg.ll;
|
| 1162 |
} |
| 1163 |
|
| 1164 |
/* fctid - fctid. */
|
| 1165 |
uint64_t helper_fctid (uint64_t arg) |
| 1166 |
{
|
| 1167 |
CPU_DoubleU farg; |
| 1168 |
farg.ll = arg; |
| 1169 |
|
| 1170 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1171 |
/* sNaN conversion */
|
| 1172 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
| 1173 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
| 1174 |
/* qNan / infinity conversion */
|
| 1175 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
| 1176 |
} else {
|
| 1177 |
farg.ll = float64_to_int64(farg.d, &env->fp_status); |
| 1178 |
} |
| 1179 |
return farg.ll;
|
| 1180 |
} |
| 1181 |
|
| 1182 |
/* fctidz - fctidz. */
|
| 1183 |
uint64_t helper_fctidz (uint64_t arg) |
| 1184 |
{
|
| 1185 |
CPU_DoubleU farg; |
| 1186 |
farg.ll = arg; |
| 1187 |
|
| 1188 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1189 |
/* sNaN conversion */
|
| 1190 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
| 1191 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
| 1192 |
/* qNan / infinity conversion */
|
| 1193 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
| 1194 |
} else {
|
| 1195 |
farg.ll = float64_to_int64_round_to_zero(farg.d, &env->fp_status); |
| 1196 |
} |
| 1197 |
return farg.ll;
|
| 1198 |
} |
| 1199 |
|
| 1200 |
#endif
|
| 1201 |
|
| 1202 |
static always_inline uint64_t do_fri (uint64_t arg, int rounding_mode) |
| 1203 |
{
|
| 1204 |
CPU_DoubleU farg; |
| 1205 |
farg.ll = arg; |
| 1206 |
|
| 1207 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1208 |
/* sNaN round */
|
| 1209 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
| 1210 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
| 1211 |
/* qNan / infinity round */
|
| 1212 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
| 1213 |
} else {
|
| 1214 |
set_float_rounding_mode(rounding_mode, &env->fp_status); |
| 1215 |
farg.ll = float64_round_to_int(farg.d, &env->fp_status); |
| 1216 |
/* Restore rounding mode from FPSCR */
|
| 1217 |
fpscr_set_rounding_mode(); |
| 1218 |
} |
| 1219 |
return farg.ll;
|
| 1220 |
} |
| 1221 |
|
| 1222 |
uint64_t helper_frin (uint64_t arg) |
| 1223 |
{
|
| 1224 |
return do_fri(arg, float_round_nearest_even);
|
| 1225 |
} |
| 1226 |
|
| 1227 |
uint64_t helper_friz (uint64_t arg) |
| 1228 |
{
|
| 1229 |
return do_fri(arg, float_round_to_zero);
|
| 1230 |
} |
| 1231 |
|
| 1232 |
uint64_t helper_frip (uint64_t arg) |
| 1233 |
{
|
| 1234 |
return do_fri(arg, float_round_up);
|
| 1235 |
} |
| 1236 |
|
| 1237 |
uint64_t helper_frim (uint64_t arg) |
| 1238 |
{
|
| 1239 |
return do_fri(arg, float_round_down);
|
| 1240 |
} |
| 1241 |
|
| 1242 |
/* fmadd - fmadd. */
|
| 1243 |
uint64_t helper_fmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
| 1244 |
{
|
| 1245 |
CPU_DoubleU farg1, farg2, farg3; |
| 1246 |
|
| 1247 |
farg1.ll = arg1; |
| 1248 |
farg2.ll = arg2; |
| 1249 |
farg3.ll = arg3; |
| 1250 |
#if USE_PRECISE_EMULATION
|
| 1251 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1252 |
float64_is_signaling_nan(farg2.d) || |
| 1253 |
float64_is_signaling_nan(farg3.d))) {
|
| 1254 |
/* sNaN operation */
|
| 1255 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1256 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
| 1257 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
| 1258 |
/* Multiplication of zero by infinity */
|
| 1259 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
| 1260 |
} else {
|
| 1261 |
#ifdef FLOAT128
|
| 1262 |
/* This is the way the PowerPC specification defines it */
|
| 1263 |
float128 ft0_128, ft1_128; |
| 1264 |
|
| 1265 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
| 1266 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
| 1267 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
| 1268 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
| 1269 |
float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) {
|
| 1270 |
/* Magnitude subtraction of infinities */
|
| 1271 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
| 1272 |
} else {
|
| 1273 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
| 1274 |
ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status); |
| 1275 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
| 1276 |
} |
| 1277 |
#else
|
| 1278 |
/* This is OK on x86 hosts */
|
| 1279 |
farg1.d = (farg1.d * farg2.d) + farg3.d; |
| 1280 |
#endif
|
| 1281 |
} |
| 1282 |
#else
|
| 1283 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
| 1284 |
farg1.d = float64_add(farg1.d, farg3.d, &env->fp_status); |
| 1285 |
#endif
|
| 1286 |
return farg1.ll;
|
| 1287 |
} |
| 1288 |
|
| 1289 |
/* fmsub - fmsub. */
|
| 1290 |
uint64_t helper_fmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
| 1291 |
{
|
| 1292 |
CPU_DoubleU farg1, farg2, farg3; |
| 1293 |
|
| 1294 |
farg1.ll = arg1; |
| 1295 |
farg2.ll = arg2; |
| 1296 |
farg3.ll = arg3; |
| 1297 |
#if USE_PRECISE_EMULATION
|
| 1298 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1299 |
float64_is_signaling_nan(farg2.d) || |
| 1300 |
float64_is_signaling_nan(farg3.d))) {
|
| 1301 |
/* sNaN operation */
|
| 1302 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1303 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
| 1304 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
| 1305 |
/* Multiplication of zero by infinity */
|
| 1306 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
| 1307 |
} else {
|
| 1308 |
#ifdef FLOAT128
|
| 1309 |
/* This is the way the PowerPC specification defines it */
|
| 1310 |
float128 ft0_128, ft1_128; |
| 1311 |
|
| 1312 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
| 1313 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
| 1314 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
| 1315 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
| 1316 |
float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) {
|
| 1317 |
/* Magnitude subtraction of infinities */
|
| 1318 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
| 1319 |
} else {
|
| 1320 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
| 1321 |
ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status); |
| 1322 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
| 1323 |
} |
| 1324 |
#else
|
| 1325 |
/* This is OK on x86 hosts */
|
| 1326 |
farg1.d = (farg1.d * farg2.d) - farg3.d; |
| 1327 |
#endif
|
| 1328 |
} |
| 1329 |
#else
|
| 1330 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
| 1331 |
farg1.d = float64_sub(farg1.d, farg3.d, &env->fp_status); |
| 1332 |
#endif
|
| 1333 |
return farg1.ll;
|
| 1334 |
} |
| 1335 |
|
| 1336 |
/* fnmadd - fnmadd. */
|
| 1337 |
uint64_t helper_fnmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
| 1338 |
{
|
| 1339 |
CPU_DoubleU farg1, farg2, farg3; |
| 1340 |
|
| 1341 |
farg1.ll = arg1; |
| 1342 |
farg2.ll = arg2; |
| 1343 |
farg3.ll = arg3; |
| 1344 |
|
| 1345 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1346 |
float64_is_signaling_nan(farg2.d) || |
| 1347 |
float64_is_signaling_nan(farg3.d))) {
|
| 1348 |
/* sNaN operation */
|
| 1349 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1350 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
| 1351 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
| 1352 |
/* Multiplication of zero by infinity */
|
| 1353 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
| 1354 |
} else {
|
| 1355 |
#if USE_PRECISE_EMULATION
|
| 1356 |
#ifdef FLOAT128
|
| 1357 |
/* This is the way the PowerPC specification defines it */
|
| 1358 |
float128 ft0_128, ft1_128; |
| 1359 |
|
| 1360 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
| 1361 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
| 1362 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
| 1363 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
| 1364 |
float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) {
|
| 1365 |
/* Magnitude subtraction of infinities */
|
| 1366 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
| 1367 |
} else {
|
| 1368 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
| 1369 |
ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status); |
| 1370 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
| 1371 |
} |
| 1372 |
#else
|
| 1373 |
/* This is OK on x86 hosts */
|
| 1374 |
farg1.d = (farg1.d * farg2.d) + farg3.d; |
| 1375 |
#endif
|
| 1376 |
#else
|
| 1377 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
| 1378 |
farg1.d = float64_add(farg1.d, farg3.d, &env->fp_status); |
| 1379 |
#endif
|
| 1380 |
if (likely(!float64_is_nan(farg1.d)))
|
| 1381 |
farg1.d = float64_chs(farg1.d); |
| 1382 |
} |
| 1383 |
return farg1.ll;
|
| 1384 |
} |
| 1385 |
|
| 1386 |
/* fnmsub - fnmsub. */
|
| 1387 |
uint64_t helper_fnmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
| 1388 |
{
|
| 1389 |
CPU_DoubleU farg1, farg2, farg3; |
| 1390 |
|
| 1391 |
farg1.ll = arg1; |
| 1392 |
farg2.ll = arg2; |
| 1393 |
farg3.ll = arg3; |
| 1394 |
|
| 1395 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
| 1396 |
float64_is_signaling_nan(farg2.d) || |
| 1397 |
float64_is_signaling_nan(farg3.d))) {
|
| 1398 |
/* sNaN operation */
|
| 1399 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1400 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
| 1401 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) {
|
| 1402 |
/* Multiplication of zero by infinity */
|
| 1403 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
| 1404 |
} else {
|
| 1405 |
#if USE_PRECISE_EMULATION
|
| 1406 |
#ifdef FLOAT128
|
| 1407 |
/* This is the way the PowerPC specification defines it */
|
| 1408 |
float128 ft0_128, ft1_128; |
| 1409 |
|
| 1410 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
| 1411 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
| 1412 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
| 1413 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
| 1414 |
float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) {
|
| 1415 |
/* Magnitude subtraction of infinities */
|
| 1416 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
| 1417 |
} else {
|
| 1418 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
| 1419 |
ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status); |
| 1420 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
| 1421 |
} |
| 1422 |
#else
|
| 1423 |
/* This is OK on x86 hosts */
|
| 1424 |
farg1.d = (farg1.d * farg2.d) - farg3.d; |
| 1425 |
#endif
|
| 1426 |
#else
|
| 1427 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
| 1428 |
farg1.d = float64_sub(farg1.d, farg3.d, &env->fp_status); |
| 1429 |
#endif
|
| 1430 |
if (likely(!float64_is_nan(farg1.d)))
|
| 1431 |
farg1.d = float64_chs(farg1.d); |
| 1432 |
} |
| 1433 |
return farg1.ll;
|
| 1434 |
} |
| 1435 |
|
| 1436 |
/* frsp - frsp. */
|
| 1437 |
uint64_t helper_frsp (uint64_t arg) |
| 1438 |
{
|
| 1439 |
CPU_DoubleU farg; |
| 1440 |
float32 f32; |
| 1441 |
farg.ll = arg; |
| 1442 |
|
| 1443 |
#if USE_PRECISE_EMULATION
|
| 1444 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1445 |
/* sNaN square root */
|
| 1446 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1447 |
} else {
|
| 1448 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
| 1449 |
farg.d = float32_to_float64(f32, &env->fp_status); |
| 1450 |
} |
| 1451 |
#else
|
| 1452 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
| 1453 |
farg.d = float32_to_float64(f32, &env->fp_status); |
| 1454 |
#endif
|
| 1455 |
return farg.ll;
|
| 1456 |
} |
| 1457 |
|
| 1458 |
/* fsqrt - fsqrt. */
|
| 1459 |
uint64_t helper_fsqrt (uint64_t arg) |
| 1460 |
{
|
| 1461 |
CPU_DoubleU farg; |
| 1462 |
farg.ll = arg; |
| 1463 |
|
| 1464 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1465 |
/* sNaN square root */
|
| 1466 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1467 |
} else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) { |
| 1468 |
/* Square root of a negative nonzero number */
|
| 1469 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT); |
| 1470 |
} else {
|
| 1471 |
farg.d = float64_sqrt(farg.d, &env->fp_status); |
| 1472 |
} |
| 1473 |
return farg.ll;
|
| 1474 |
} |
| 1475 |
|
| 1476 |
/* fre - fre. */
|
| 1477 |
uint64_t helper_fre (uint64_t arg) |
| 1478 |
{
|
| 1479 |
CPU_DoubleU farg; |
| 1480 |
farg.ll = arg; |
| 1481 |
|
| 1482 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1483 |
/* sNaN reciprocal */
|
| 1484 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1485 |
} else {
|
| 1486 |
farg.d = float64_div(float64_one, farg.d, &env->fp_status); |
| 1487 |
} |
| 1488 |
return farg.d;
|
| 1489 |
} |
| 1490 |
|
| 1491 |
/* fres - fres. */
|
| 1492 |
uint64_t helper_fres (uint64_t arg) |
| 1493 |
{
|
| 1494 |
CPU_DoubleU farg; |
| 1495 |
float32 f32; |
| 1496 |
farg.ll = arg; |
| 1497 |
|
| 1498 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1499 |
/* sNaN reciprocal */
|
| 1500 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1501 |
} else {
|
| 1502 |
farg.d = float64_div(float64_one, farg.d, &env->fp_status); |
| 1503 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
| 1504 |
farg.d = float32_to_float64(f32, &env->fp_status); |
| 1505 |
} |
| 1506 |
return farg.ll;
|
| 1507 |
} |
| 1508 |
|
| 1509 |
/* frsqrte - frsqrte. */
|
| 1510 |
uint64_t helper_frsqrte (uint64_t arg) |
| 1511 |
{
|
| 1512 |
CPU_DoubleU farg; |
| 1513 |
float32 f32; |
| 1514 |
farg.ll = arg; |
| 1515 |
|
| 1516 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
| 1517 |
/* sNaN reciprocal square root */
|
| 1518 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1519 |
} else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) { |
| 1520 |
/* Reciprocal square root of a negative nonzero number */
|
| 1521 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT); |
| 1522 |
} else {
|
| 1523 |
farg.d = float64_sqrt(farg.d, &env->fp_status); |
| 1524 |
farg.d = float64_div(float64_one, farg.d, &env->fp_status); |
| 1525 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
| 1526 |
farg.d = float32_to_float64(f32, &env->fp_status); |
| 1527 |
} |
| 1528 |
return farg.ll;
|
| 1529 |
} |
| 1530 |
|
| 1531 |
/* fsel - fsel. */
|
| 1532 |
uint64_t helper_fsel (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
| 1533 |
{
|
| 1534 |
CPU_DoubleU farg1; |
| 1535 |
|
| 1536 |
farg1.ll = arg1; |
| 1537 |
|
| 1538 |
if ((!float64_is_neg(farg1.d) || float64_is_zero(farg1.d)) && !float64_is_nan(farg1.d))
|
| 1539 |
return arg2;
|
| 1540 |
else
|
| 1541 |
return arg3;
|
| 1542 |
} |
| 1543 |
|
| 1544 |
void helper_fcmpu (uint64_t arg1, uint64_t arg2, uint32_t crfD)
|
| 1545 |
{
|
| 1546 |
CPU_DoubleU farg1, farg2; |
| 1547 |
uint32_t ret = 0;
|
| 1548 |
farg1.ll = arg1; |
| 1549 |
farg2.ll = arg2; |
| 1550 |
|
| 1551 |
if (unlikely(float64_is_nan(farg1.d) ||
|
| 1552 |
float64_is_nan(farg2.d))) {
|
| 1553 |
ret = 0x01UL;
|
| 1554 |
} else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) { |
| 1555 |
ret = 0x08UL;
|
| 1556 |
} else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) { |
| 1557 |
ret = 0x04UL;
|
| 1558 |
} else {
|
| 1559 |
ret = 0x02UL;
|
| 1560 |
} |
| 1561 |
|
| 1562 |
env->fpscr &= ~(0x0F << FPSCR_FPRF);
|
| 1563 |
env->fpscr |= ret << FPSCR_FPRF; |
| 1564 |
env->crf[crfD] = ret; |
| 1565 |
if (unlikely(ret == 0x01UL |
| 1566 |
&& (float64_is_signaling_nan(farg1.d) || |
| 1567 |
float64_is_signaling_nan(farg2.d)))) {
|
| 1568 |
/* sNaN comparison */
|
| 1569 |
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
| 1570 |
} |
| 1571 |
} |
| 1572 |
|
| 1573 |
void helper_fcmpo (uint64_t arg1, uint64_t arg2, uint32_t crfD)
|
| 1574 |
{
|
| 1575 |
CPU_DoubleU farg1, farg2; |
| 1576 |
uint32_t ret = 0;
|
| 1577 |
farg1.ll = arg1; |
| 1578 |
farg2.ll = arg2; |
| 1579 |
|
| 1580 |
if (unlikely(float64_is_nan(farg1.d) ||
|
| 1581 |
float64_is_nan(farg2.d))) {
|
| 1582 |
ret = 0x01UL;
|
| 1583 |
} else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) { |
| 1584 |
ret = 0x08UL;
|
| 1585 |
} else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) { |
| 1586 |
ret = 0x04UL;
|
| 1587 |
} else {
|
| 1588 |
ret = 0x02UL;
|
| 1589 |
} |
| 1590 |
|
| 1591 |
env->fpscr &= ~(0x0F << FPSCR_FPRF);
|
| 1592 |
env->fpscr |= ret << FPSCR_FPRF; |
| 1593 |
env->crf[crfD] = ret; |
| 1594 |
if (unlikely (ret == 0x01UL)) { |
| 1595 |
if (float64_is_signaling_nan(farg1.d) ||
|
| 1596 |
float64_is_signaling_nan(farg2.d)) {
|
| 1597 |
/* sNaN comparison */
|
| 1598 |
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | |
| 1599 |
POWERPC_EXCP_FP_VXVC); |
| 1600 |
} else {
|
| 1601 |
/* qNaN comparison */
|
| 1602 |
fload_invalid_op_excp(POWERPC_EXCP_FP_VXVC); |
| 1603 |
} |
| 1604 |
} |
| 1605 |
} |
| 1606 |
|
| 1607 |
#if !defined (CONFIG_USER_ONLY)
|
| 1608 |
void helper_store_msr (target_ulong val)
|
| 1609 |
{
|
| 1610 |
val = hreg_store_msr(env, val, 0);
|
| 1611 |
if (val != 0) { |
| 1612 |
env->interrupt_request |= CPU_INTERRUPT_EXITTB; |
| 1613 |
helper_raise_exception(val); |
| 1614 |
} |
| 1615 |
} |
| 1616 |
|
| 1617 |
static always_inline void do_rfi (target_ulong nip, target_ulong msr, |
| 1618 |
target_ulong msrm, int keep_msrh)
|
| 1619 |
{
|
| 1620 |
#if defined(TARGET_PPC64)
|
| 1621 |
if (msr & (1ULL << MSR_SF)) { |
| 1622 |
nip = (uint64_t)nip; |
| 1623 |
msr &= (uint64_t)msrm; |
| 1624 |
} else {
|
| 1625 |
nip = (uint32_t)nip; |
| 1626 |
msr = (uint32_t)(msr & msrm); |
| 1627 |
if (keep_msrh)
|
| 1628 |
msr |= env->msr & ~((uint64_t)0xFFFFFFFF);
|
| 1629 |
} |
| 1630 |
#else
|
| 1631 |
nip = (uint32_t)nip; |
| 1632 |
msr &= (uint32_t)msrm; |
| 1633 |
#endif
|
| 1634 |
/* XXX: beware: this is false if VLE is supported */
|
| 1635 |
env->nip = nip & ~((target_ulong)0x00000003);
|
| 1636 |
hreg_store_msr(env, msr, 1);
|
| 1637 |
#if defined (DEBUG_OP)
|
| 1638 |
cpu_dump_rfi(env->nip, env->msr); |
| 1639 |
#endif
|
| 1640 |
/* No need to raise an exception here,
|
| 1641 |
* as rfi is always the last insn of a TB
|
| 1642 |
*/
|
| 1643 |
env->interrupt_request |= CPU_INTERRUPT_EXITTB; |
| 1644 |
} |
| 1645 |
|
| 1646 |
void helper_rfi (void) |
| 1647 |
{
|
| 1648 |
do_rfi(env->spr[SPR_SRR0], env->spr[SPR_SRR1], |
| 1649 |
~((target_ulong)0x0), 1); |
| 1650 |
} |
| 1651 |
|
| 1652 |
#if defined(TARGET_PPC64)
|
| 1653 |
void helper_rfid (void) |
| 1654 |
{
|
| 1655 |
do_rfi(env->spr[SPR_SRR0], env->spr[SPR_SRR1], |
| 1656 |
~((target_ulong)0x0), 0); |
| 1657 |
} |
| 1658 |
|
| 1659 |
void helper_hrfid (void) |
| 1660 |
{
|
| 1661 |
do_rfi(env->spr[SPR_HSRR0], env->spr[SPR_HSRR1], |
| 1662 |
~((target_ulong)0x0), 0); |
| 1663 |
} |
| 1664 |
#endif
|
| 1665 |
#endif
|
| 1666 |
|
| 1667 |
void helper_tw (target_ulong arg1, target_ulong arg2, uint32_t flags)
|
| 1668 |
{
|
| 1669 |
if (!likely(!(((int32_t)arg1 < (int32_t)arg2 && (flags & 0x10)) || |
| 1670 |
((int32_t)arg1 > (int32_t)arg2 && (flags & 0x08)) ||
|
| 1671 |
((int32_t)arg1 == (int32_t)arg2 && (flags & 0x04)) ||
|
| 1672 |
((uint32_t)arg1 < (uint32_t)arg2 && (flags & 0x02)) ||
|
| 1673 |
((uint32_t)arg1 > (uint32_t)arg2 && (flags & 0x01))))) {
|
| 1674 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_TRAP); |
| 1675 |
} |
| 1676 |
} |
| 1677 |
|
| 1678 |
#if defined(TARGET_PPC64)
|
| 1679 |
void helper_td (target_ulong arg1, target_ulong arg2, uint32_t flags)
|
| 1680 |
{
|
| 1681 |
if (!likely(!(((int64_t)arg1 < (int64_t)arg2 && (flags & 0x10)) || |
| 1682 |
((int64_t)arg1 > (int64_t)arg2 && (flags & 0x08)) ||
|
| 1683 |
((int64_t)arg1 == (int64_t)arg2 && (flags & 0x04)) ||
|
| 1684 |
((uint64_t)arg1 < (uint64_t)arg2 && (flags & 0x02)) ||
|
| 1685 |
((uint64_t)arg1 > (uint64_t)arg2 && (flags & 0x01)))))
|
| 1686 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_TRAP); |
| 1687 |
} |
| 1688 |
#endif
|
| 1689 |
|
| 1690 |
/*****************************************************************************/
|
| 1691 |
/* PowerPC 601 specific instructions (POWER bridge) */
|
| 1692 |
|
| 1693 |
target_ulong helper_clcs (uint32_t arg) |
| 1694 |
{
|
| 1695 |
switch (arg) {
|
| 1696 |
case 0x0CUL: |
| 1697 |
/* Instruction cache line size */
|
| 1698 |
return env->icache_line_size;
|
| 1699 |
break;
|
| 1700 |
case 0x0DUL: |
| 1701 |
/* Data cache line size */
|
| 1702 |
return env->dcache_line_size;
|
| 1703 |
break;
|
| 1704 |
case 0x0EUL: |
| 1705 |
/* Minimum cache line size */
|
| 1706 |
return (env->icache_line_size < env->dcache_line_size) ?
|
| 1707 |
env->icache_line_size : env->dcache_line_size; |
| 1708 |
break;
|
| 1709 |
case 0x0FUL: |
| 1710 |
/* Maximum cache line size */
|
| 1711 |
return (env->icache_line_size > env->dcache_line_size) ?
|
| 1712 |
env->icache_line_size : env->dcache_line_size; |
| 1713 |
break;
|
| 1714 |
default:
|
| 1715 |
/* Undefined */
|
| 1716 |
return 0; |
| 1717 |
break;
|
| 1718 |
} |
| 1719 |
} |
| 1720 |
|
| 1721 |
target_ulong helper_div (target_ulong arg1, target_ulong arg2) |
| 1722 |
{
|
| 1723 |
uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
|
| 1724 |
|
| 1725 |
if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
| 1726 |
(int32_t)arg2 == 0) {
|
| 1727 |
env->spr[SPR_MQ] = 0;
|
| 1728 |
return INT32_MIN;
|
| 1729 |
} else {
|
| 1730 |
env->spr[SPR_MQ] = tmp % arg2; |
| 1731 |
return tmp / (int32_t)arg2;
|
| 1732 |
} |
| 1733 |
} |
| 1734 |
|
| 1735 |
target_ulong helper_divo (target_ulong arg1, target_ulong arg2) |
| 1736 |
{
|
| 1737 |
uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
|
| 1738 |
|
| 1739 |
if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
| 1740 |
(int32_t)arg2 == 0) {
|
| 1741 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
| 1742 |
env->spr[SPR_MQ] = 0;
|
| 1743 |
return INT32_MIN;
|
| 1744 |
} else {
|
| 1745 |
env->spr[SPR_MQ] = tmp % arg2; |
| 1746 |
tmp /= (int32_t)arg2; |
| 1747 |
if ((int32_t)tmp != tmp) {
|
| 1748 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
| 1749 |
} else {
|
| 1750 |
env->xer &= ~(1 << XER_OV);
|
| 1751 |
} |
| 1752 |
return tmp;
|
| 1753 |
} |
| 1754 |
} |
| 1755 |
|
| 1756 |
target_ulong helper_divs (target_ulong arg1, target_ulong arg2) |
| 1757 |
{
|
| 1758 |
if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
| 1759 |
(int32_t)arg2 == 0) {
|
| 1760 |
env->spr[SPR_MQ] = 0;
|
| 1761 |
return INT32_MIN;
|
| 1762 |
} else {
|
| 1763 |
env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2; |
| 1764 |
return (int32_t)arg1 / (int32_t)arg2;
|
| 1765 |
} |
| 1766 |
} |
| 1767 |
|
| 1768 |
target_ulong helper_divso (target_ulong arg1, target_ulong arg2) |
| 1769 |
{
|
| 1770 |
if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
| 1771 |
(int32_t)arg2 == 0) {
|
| 1772 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
| 1773 |
env->spr[SPR_MQ] = 0;
|
| 1774 |
return INT32_MIN;
|
| 1775 |
} else {
|
| 1776 |
env->xer &= ~(1 << XER_OV);
|
| 1777 |
env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2; |
| 1778 |
return (int32_t)arg1 / (int32_t)arg2;
|
| 1779 |
} |
| 1780 |
} |
| 1781 |
|
| 1782 |
#if !defined (CONFIG_USER_ONLY)
|
| 1783 |
target_ulong helper_rac (target_ulong addr) |
| 1784 |
{
|
| 1785 |
mmu_ctx_t ctx; |
| 1786 |
int nb_BATs;
|
| 1787 |
target_ulong ret = 0;
|
| 1788 |
|
| 1789 |
/* We don't have to generate many instances of this instruction,
|
| 1790 |
* as rac is supervisor only.
|
| 1791 |
*/
|
| 1792 |
/* XXX: FIX THIS: Pretend we have no BAT */
|
| 1793 |
nb_BATs = env->nb_BATs; |
| 1794 |
env->nb_BATs = 0;
|
| 1795 |
if (get_physical_address(env, &ctx, addr, 0, ACCESS_INT) == 0) |
| 1796 |
ret = ctx.raddr; |
| 1797 |
env->nb_BATs = nb_BATs; |
| 1798 |
return ret;
|
| 1799 |
} |
| 1800 |
|
| 1801 |
void helper_rfsvc (void) |
| 1802 |
{
|
| 1803 |
do_rfi(env->lr, env->ctr, 0x0000FFFF, 0); |
| 1804 |
} |
| 1805 |
#endif
|
| 1806 |
|
| 1807 |
/*****************************************************************************/
|
| 1808 |
/* 602 specific instructions */
|
| 1809 |
/* mfrom is the most crazy instruction ever seen, imho ! */
|
| 1810 |
/* Real implementation uses a ROM table. Do the same */
|
| 1811 |
/* Extremly decomposed:
|
| 1812 |
* -arg / 256
|
| 1813 |
* return 256 * log10(10 + 1.0) + 0.5
|
| 1814 |
*/
|
| 1815 |
#if !defined (CONFIG_USER_ONLY)
|
| 1816 |
target_ulong helper_602_mfrom (target_ulong arg) |
| 1817 |
{
|
| 1818 |
if (likely(arg < 602)) { |
| 1819 |
#include "mfrom_table.c" |
| 1820 |
return mfrom_ROM_table[arg];
|
| 1821 |
} else {
|
| 1822 |
return 0; |
| 1823 |
} |
| 1824 |
} |
| 1825 |
#endif
|
| 1826 |
|
| 1827 |
/*****************************************************************************/
|
| 1828 |
/* Embedded PowerPC specific helpers */
|
| 1829 |
|
| 1830 |
/* XXX: to be improved to check access rights when in user-mode */
|
| 1831 |
target_ulong helper_load_dcr (target_ulong dcrn) |
| 1832 |
{
|
| 1833 |
target_ulong val = 0;
|
| 1834 |
|
| 1835 |
if (unlikely(env->dcr_env == NULL)) { |
| 1836 |
qemu_log("No DCR environment\n");
|
| 1837 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
| 1838 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_INVAL_INVAL); |
| 1839 |
} else if (unlikely(ppc_dcr_read(env->dcr_env, dcrn, &val) != 0)) { |
| 1840 |
qemu_log("DCR read error %d %03x\n", (int)dcrn, (int)dcrn); |
| 1841 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
| 1842 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_PRIV_REG); |
| 1843 |
} |
| 1844 |
return val;
|
| 1845 |
} |
| 1846 |
|
| 1847 |
void helper_store_dcr (target_ulong dcrn, target_ulong val)
|
| 1848 |
{
|
| 1849 |
if (unlikely(env->dcr_env == NULL)) { |
| 1850 |
qemu_log("No DCR environment\n");
|
| 1851 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
| 1852 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_INVAL_INVAL); |
| 1853 |
} else if (unlikely(ppc_dcr_write(env->dcr_env, dcrn, val) != 0)) { |
| 1854 |
qemu_log("DCR write error %d %03x\n", (int)dcrn, (int)dcrn); |
| 1855 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
| 1856 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_PRIV_REG); |
| 1857 |
} |
| 1858 |
} |
| 1859 |
|
| 1860 |
#if !defined(CONFIG_USER_ONLY)
|
| 1861 |
void helper_40x_rfci (void) |
| 1862 |
{
|
| 1863 |
do_rfi(env->spr[SPR_40x_SRR2], env->spr[SPR_40x_SRR3], |
| 1864 |
~((target_ulong)0xFFFF0000), 0); |
| 1865 |
} |
| 1866 |
|
| 1867 |
void helper_rfci (void) |
| 1868 |
{
|
| 1869 |
do_rfi(env->spr[SPR_BOOKE_CSRR0], SPR_BOOKE_CSRR1, |
| 1870 |
~((target_ulong)0x3FFF0000), 0); |
| 1871 |
} |
| 1872 |
|
| 1873 |
void helper_rfdi (void) |
| 1874 |
{
|
| 1875 |
do_rfi(env->spr[SPR_BOOKE_DSRR0], SPR_BOOKE_DSRR1, |
| 1876 |
~((target_ulong)0x3FFF0000), 0); |
| 1877 |
} |
| 1878 |
|
| 1879 |
void helper_rfmci (void) |
| 1880 |
{
|
| 1881 |
do_rfi(env->spr[SPR_BOOKE_MCSRR0], SPR_BOOKE_MCSRR1, |
| 1882 |
~((target_ulong)0x3FFF0000), 0); |
| 1883 |
} |
| 1884 |
#endif
|
| 1885 |
|
| 1886 |
/* 440 specific */
|
| 1887 |
target_ulong helper_dlmzb (target_ulong high, target_ulong low, uint32_t update_Rc) |
| 1888 |
{
|
| 1889 |
target_ulong mask; |
| 1890 |
int i;
|
| 1891 |
|
| 1892 |
i = 1;
|
| 1893 |
for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { |
| 1894 |
if ((high & mask) == 0) { |
| 1895 |
if (update_Rc) {
|
| 1896 |
env->crf[0] = 0x4; |
| 1897 |
} |
| 1898 |
goto done;
|
| 1899 |
} |
| 1900 |
i++; |
| 1901 |
} |
| 1902 |
for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { |
| 1903 |
if ((low & mask) == 0) { |
| 1904 |
if (update_Rc) {
|
| 1905 |
env->crf[0] = 0x8; |
| 1906 |
} |
| 1907 |
goto done;
|
| 1908 |
} |
| 1909 |
i++; |
| 1910 |
} |
| 1911 |
if (update_Rc) {
|
| 1912 |
env->crf[0] = 0x2; |
| 1913 |
} |
| 1914 |
done:
|
| 1915 |
env->xer = (env->xer & ~0x7F) | i;
|
| 1916 |
if (update_Rc) {
|
| 1917 |
env->crf[0] |= xer_so;
|
| 1918 |
} |
| 1919 |
return i;
|
| 1920 |
} |
| 1921 |
|
| 1922 |
/*****************************************************************************/
|
| 1923 |
/* Altivec extension helpers */
|
| 1924 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 1925 |
#define HI_IDX 0 |
| 1926 |
#define LO_IDX 1 |
| 1927 |
#else
|
| 1928 |
#define HI_IDX 1 |
| 1929 |
#define LO_IDX 0 |
| 1930 |
#endif
|
| 1931 |
|
| 1932 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 1933 |
#define VECTOR_FOR_INORDER_I(index, element) \
|
| 1934 |
for (index = 0; index < ARRAY_SIZE(r->element); index++) |
| 1935 |
#else
|
| 1936 |
#define VECTOR_FOR_INORDER_I(index, element) \
|
| 1937 |
for (index = ARRAY_SIZE(r->element)-1; index >= 0; index--) |
| 1938 |
#endif
|
| 1939 |
|
| 1940 |
/* If X is a NaN, store the corresponding QNaN into RESULT. Otherwise,
|
| 1941 |
* execute the following block. */
|
| 1942 |
#define DO_HANDLE_NAN(result, x) \
|
| 1943 |
if (float32_is_nan(x) || float32_is_signaling_nan(x)) { \
|
| 1944 |
CPU_FloatU __f; \ |
| 1945 |
__f.f = x; \ |
| 1946 |
__f.l = __f.l | (1 << 22); /* Set QNaN bit. */ \ |
| 1947 |
result = __f.f; \ |
| 1948 |
} else
|
| 1949 |
|
| 1950 |
#define HANDLE_NAN1(result, x) \
|
| 1951 |
DO_HANDLE_NAN(result, x) |
| 1952 |
#define HANDLE_NAN2(result, x, y) \
|
| 1953 |
DO_HANDLE_NAN(result, x) DO_HANDLE_NAN(result, y) |
| 1954 |
#define HANDLE_NAN3(result, x, y, z) \
|
| 1955 |
DO_HANDLE_NAN(result, x) DO_HANDLE_NAN(result, y) DO_HANDLE_NAN(result, z) |
| 1956 |
|
| 1957 |
/* Saturating arithmetic helpers. */
|
| 1958 |
#define SATCVT(from, to, from_type, to_type, min, max, use_min, use_max) \
|
| 1959 |
static always_inline to_type cvt##from##to (from_type x, int *sat) \ |
| 1960 |
{ \
|
| 1961 |
to_type r; \ |
| 1962 |
if (use_min && x < min) { \
|
| 1963 |
r = min; \ |
| 1964 |
*sat = 1; \
|
| 1965 |
} else if (use_max && x > max) { \ |
| 1966 |
r = max; \ |
| 1967 |
*sat = 1; \
|
| 1968 |
} else { \
|
| 1969 |
r = x; \ |
| 1970 |
} \ |
| 1971 |
return r; \
|
| 1972 |
} |
| 1973 |
SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX, 1, 1) |
| 1974 |
SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX, 1, 1) |
| 1975 |
SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX, 1, 1) |
| 1976 |
|
| 1977 |
/* Work around gcc problems with the macro version */
|
| 1978 |
static always_inline uint8_t cvtuhub(uint16_t x, int *sat) |
| 1979 |
{
|
| 1980 |
uint8_t r; |
| 1981 |
|
| 1982 |
if (x > UINT8_MAX) {
|
| 1983 |
r = UINT8_MAX; |
| 1984 |
*sat = 1;
|
| 1985 |
} else {
|
| 1986 |
r = x; |
| 1987 |
} |
| 1988 |
return r;
|
| 1989 |
} |
| 1990 |
//SATCVT(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX, 0, 1)
|
| 1991 |
SATCVT(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX, 0, 1) |
| 1992 |
SATCVT(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX, 0, 1) |
| 1993 |
SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX, 1, 1) |
| 1994 |
SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX, 1, 1) |
| 1995 |
SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX, 1, 1) |
| 1996 |
#undef SATCVT
|
| 1997 |
|
| 1998 |
#define LVE(name, access, swap, element) \
|
| 1999 |
void helper_##name (ppc_avr_t *r, target_ulong addr) \ |
| 2000 |
{ \
|
| 2001 |
size_t n_elems = ARRAY_SIZE(r->element); \ |
| 2002 |
int adjust = HI_IDX*(n_elems-1); \ |
| 2003 |
int sh = sizeof(r->element[0]) >> 1; \ |
| 2004 |
int index = (addr & 0xf) >> sh; \ |
| 2005 |
if(msr_le) { \
|
| 2006 |
r->element[LO_IDX ? index : (adjust - index)] = swap(access(addr)); \ |
| 2007 |
} else { \
|
| 2008 |
r->element[LO_IDX ? index : (adjust - index)] = access(addr); \ |
| 2009 |
} \ |
| 2010 |
} |
| 2011 |
#define I(x) (x)
|
| 2012 |
LVE(lvebx, ldub, I, u8) |
| 2013 |
LVE(lvehx, lduw, bswap16, u16) |
| 2014 |
LVE(lvewx, ldl, bswap32, u32) |
| 2015 |
#undef I
|
| 2016 |
#undef LVE
|
| 2017 |
|
| 2018 |
void helper_lvsl (ppc_avr_t *r, target_ulong sh)
|
| 2019 |
{
|
| 2020 |
int i, j = (sh & 0xf); |
| 2021 |
|
| 2022 |
VECTOR_FOR_INORDER_I (i, u8) {
|
| 2023 |
r->u8[i] = j++; |
| 2024 |
} |
| 2025 |
} |
| 2026 |
|
| 2027 |
void helper_lvsr (ppc_avr_t *r, target_ulong sh)
|
| 2028 |
{
|
| 2029 |
int i, j = 0x10 - (sh & 0xf); |
| 2030 |
|
| 2031 |
VECTOR_FOR_INORDER_I (i, u8) {
|
| 2032 |
r->u8[i] = j++; |
| 2033 |
} |
| 2034 |
} |
| 2035 |
|
| 2036 |
#define STVE(name, access, swap, element) \
|
| 2037 |
void helper_##name (ppc_avr_t *r, target_ulong addr) \ |
| 2038 |
{ \
|
| 2039 |
size_t n_elems = ARRAY_SIZE(r->element); \ |
| 2040 |
int adjust = HI_IDX*(n_elems-1); \ |
| 2041 |
int sh = sizeof(r->element[0]) >> 1; \ |
| 2042 |
int index = (addr & 0xf) >> sh; \ |
| 2043 |
if(msr_le) { \
|
| 2044 |
access(addr, swap(r->element[LO_IDX ? index : (adjust - index)])); \ |
| 2045 |
} else { \
|
| 2046 |
access(addr, r->element[LO_IDX ? index : (adjust - index)]); \ |
| 2047 |
} \ |
| 2048 |
} |
| 2049 |
#define I(x) (x)
|
| 2050 |
STVE(stvebx, stb, I, u8) |
| 2051 |
STVE(stvehx, stw, bswap16, u16) |
| 2052 |
STVE(stvewx, stl, bswap32, u32) |
| 2053 |
#undef I
|
| 2054 |
#undef LVE
|
| 2055 |
|
| 2056 |
void helper_mtvscr (ppc_avr_t *r)
|
| 2057 |
{
|
| 2058 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2059 |
env->vscr = r->u32[3];
|
| 2060 |
#else
|
| 2061 |
env->vscr = r->u32[0];
|
| 2062 |
#endif
|
| 2063 |
set_flush_to_zero(vscr_nj, &env->vec_status); |
| 2064 |
} |
| 2065 |
|
| 2066 |
void helper_vaddcuw (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2067 |
{
|
| 2068 |
int i;
|
| 2069 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { |
| 2070 |
r->u32[i] = ~a->u32[i] < b->u32[i]; |
| 2071 |
} |
| 2072 |
} |
| 2073 |
|
| 2074 |
#define VARITH_DO(name, op, element) \
|
| 2075 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2076 |
{ \
|
| 2077 |
int i; \
|
| 2078 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2079 |
r->element[i] = a->element[i] op b->element[i]; \ |
| 2080 |
} \ |
| 2081 |
} |
| 2082 |
#define VARITH(suffix, element) \
|
| 2083 |
VARITH_DO(add##suffix, +, element) \ |
| 2084 |
VARITH_DO(sub##suffix, -, element) |
| 2085 |
VARITH(ubm, u8) |
| 2086 |
VARITH(uhm, u16) |
| 2087 |
VARITH(uwm, u32) |
| 2088 |
#undef VARITH_DO
|
| 2089 |
#undef VARITH
|
| 2090 |
|
| 2091 |
#define VARITHFP(suffix, func) \
|
| 2092 |
void helper_v##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2093 |
{ \
|
| 2094 |
int i; \
|
| 2095 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
| 2096 |
HANDLE_NAN2(r->f[i], a->f[i], b->f[i]) { \
|
| 2097 |
r->f[i] = func(a->f[i], b->f[i], &env->vec_status); \ |
| 2098 |
} \ |
| 2099 |
} \ |
| 2100 |
} |
| 2101 |
VARITHFP(addfp, float32_add) |
| 2102 |
VARITHFP(subfp, float32_sub) |
| 2103 |
#undef VARITHFP
|
| 2104 |
|
| 2105 |
#define VARITHSAT_CASE(type, op, cvt, element) \
|
| 2106 |
{ \
|
| 2107 |
type result = (type)a->element[i] op (type)b->element[i]; \ |
| 2108 |
r->element[i] = cvt(result, &sat); \ |
| 2109 |
} |
| 2110 |
|
| 2111 |
#define VARITHSAT_DO(name, op, optype, cvt, element) \
|
| 2112 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2113 |
{ \
|
| 2114 |
int sat = 0; \ |
| 2115 |
int i; \
|
| 2116 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2117 |
switch (sizeof(r->element[0])) { \ |
| 2118 |
case 1: VARITHSAT_CASE(optype, op, cvt, element); break; \ |
| 2119 |
case 2: VARITHSAT_CASE(optype, op, cvt, element); break; \ |
| 2120 |
case 4: VARITHSAT_CASE(optype, op, cvt, element); break; \ |
| 2121 |
} \ |
| 2122 |
} \ |
| 2123 |
if (sat) { \
|
| 2124 |
env->vscr |= (1 << VSCR_SAT); \
|
| 2125 |
} \ |
| 2126 |
} |
| 2127 |
#define VARITHSAT_SIGNED(suffix, element, optype, cvt) \
|
| 2128 |
VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \ |
| 2129 |
VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element) |
| 2130 |
#define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \
|
| 2131 |
VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \ |
| 2132 |
VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element) |
| 2133 |
VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb) |
| 2134 |
VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh) |
| 2135 |
VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw) |
| 2136 |
VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub) |
| 2137 |
VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh) |
| 2138 |
VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw) |
| 2139 |
#undef VARITHSAT_CASE
|
| 2140 |
#undef VARITHSAT_DO
|
| 2141 |
#undef VARITHSAT_SIGNED
|
| 2142 |
#undef VARITHSAT_UNSIGNED
|
| 2143 |
|
| 2144 |
#define VAVG_DO(name, element, etype) \
|
| 2145 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2146 |
{ \
|
| 2147 |
int i; \
|
| 2148 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2149 |
etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \
|
| 2150 |
r->element[i] = x >> 1; \
|
| 2151 |
} \ |
| 2152 |
} |
| 2153 |
|
| 2154 |
#define VAVG(type, signed_element, signed_type, unsigned_element, unsigned_type) \
|
| 2155 |
VAVG_DO(avgs##type, signed_element, signed_type) \ |
| 2156 |
VAVG_DO(avgu##type, unsigned_element, unsigned_type) |
| 2157 |
VAVG(b, s8, int16_t, u8, uint16_t) |
| 2158 |
VAVG(h, s16, int32_t, u16, uint32_t) |
| 2159 |
VAVG(w, s32, int64_t, u32, uint64_t) |
| 2160 |
#undef VAVG_DO
|
| 2161 |
#undef VAVG
|
| 2162 |
|
| 2163 |
#define VCF(suffix, cvt, element) \
|
| 2164 |
void helper_vcf##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t uim) \ |
| 2165 |
{ \
|
| 2166 |
int i; \
|
| 2167 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
| 2168 |
float32 t = cvt(b->element[i], &env->vec_status); \ |
| 2169 |
r->f[i] = float32_scalbn (t, -uim, &env->vec_status); \ |
| 2170 |
} \ |
| 2171 |
} |
| 2172 |
VCF(ux, uint32_to_float32, u32) |
| 2173 |
VCF(sx, int32_to_float32, s32) |
| 2174 |
#undef VCF
|
| 2175 |
|
| 2176 |
#define VCMP_DO(suffix, compare, element, record) \
|
| 2177 |
void helper_vcmp##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2178 |
{ \
|
| 2179 |
uint32_t ones = (uint32_t)-1; \
|
| 2180 |
uint32_t all = ones; \ |
| 2181 |
uint32_t none = 0; \
|
| 2182 |
int i; \
|
| 2183 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2184 |
uint32_t result = (a->element[i] compare b->element[i] ? ones : 0x0); \
|
| 2185 |
switch (sizeof (a->element[0])) { \ |
| 2186 |
case 4: r->u32[i] = result; break; \ |
| 2187 |
case 2: r->u16[i] = result; break; \ |
| 2188 |
case 1: r->u8[i] = result; break; \ |
| 2189 |
} \ |
| 2190 |
all &= result; \ |
| 2191 |
none |= result; \ |
| 2192 |
} \ |
| 2193 |
if (record) { \
|
| 2194 |
env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ |
| 2195 |
} \ |
| 2196 |
} |
| 2197 |
#define VCMP(suffix, compare, element) \
|
| 2198 |
VCMP_DO(suffix, compare, element, 0) \
|
| 2199 |
VCMP_DO(suffix##_dot, compare, element, 1) |
| 2200 |
VCMP(equb, ==, u8) |
| 2201 |
VCMP(equh, ==, u16) |
| 2202 |
VCMP(equw, ==, u32) |
| 2203 |
VCMP(gtub, >, u8) |
| 2204 |
VCMP(gtuh, >, u16) |
| 2205 |
VCMP(gtuw, >, u32) |
| 2206 |
VCMP(gtsb, >, s8) |
| 2207 |
VCMP(gtsh, >, s16) |
| 2208 |
VCMP(gtsw, >, s32) |
| 2209 |
#undef VCMP_DO
|
| 2210 |
#undef VCMP
|
| 2211 |
|
| 2212 |
#define VCMPFP_DO(suffix, compare, order, record) \
|
| 2213 |
void helper_vcmp##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2214 |
{ \
|
| 2215 |
uint32_t ones = (uint32_t)-1; \
|
| 2216 |
uint32_t all = ones; \ |
| 2217 |
uint32_t none = 0; \
|
| 2218 |
int i; \
|
| 2219 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
| 2220 |
uint32_t result; \ |
| 2221 |
int rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status); \
|
| 2222 |
if (rel == float_relation_unordered) { \
|
| 2223 |
result = 0; \
|
| 2224 |
} else if (rel compare order) { \ |
| 2225 |
result = ones; \ |
| 2226 |
} else { \
|
| 2227 |
result = 0; \
|
| 2228 |
} \ |
| 2229 |
r->u32[i] = result; \ |
| 2230 |
all &= result; \ |
| 2231 |
none |= result; \ |
| 2232 |
} \ |
| 2233 |
if (record) { \
|
| 2234 |
env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ |
| 2235 |
} \ |
| 2236 |
} |
| 2237 |
#define VCMPFP(suffix, compare, order) \
|
| 2238 |
VCMPFP_DO(suffix, compare, order, 0) \
|
| 2239 |
VCMPFP_DO(suffix##_dot, compare, order, 1) |
| 2240 |
VCMPFP(eqfp, ==, float_relation_equal) |
| 2241 |
VCMPFP(gefp, !=, float_relation_less) |
| 2242 |
VCMPFP(gtfp, ==, float_relation_greater) |
| 2243 |
#undef VCMPFP_DO
|
| 2244 |
#undef VCMPFP
|
| 2245 |
|
| 2246 |
static always_inline void vcmpbfp_internal (ppc_avr_t *r, ppc_avr_t *a, |
| 2247 |
ppc_avr_t *b, int record)
|
| 2248 |
{
|
| 2249 |
int i;
|
| 2250 |
int all_in = 0; |
| 2251 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
| 2252 |
int le_rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status);
|
| 2253 |
if (le_rel == float_relation_unordered) {
|
| 2254 |
r->u32[i] = 0xc0000000;
|
| 2255 |
/* ALL_IN does not need to be updated here. */
|
| 2256 |
} else {
|
| 2257 |
float32 bneg = float32_chs(b->f[i]); |
| 2258 |
int ge_rel = float32_compare_quiet(a->f[i], bneg, &env->vec_status);
|
| 2259 |
int le = le_rel != float_relation_greater;
|
| 2260 |
int ge = ge_rel != float_relation_less;
|
| 2261 |
r->u32[i] = ((!le) << 31) | ((!ge) << 30); |
| 2262 |
all_in |= (!le | !ge); |
| 2263 |
} |
| 2264 |
} |
| 2265 |
if (record) {
|
| 2266 |
env->crf[6] = (all_in == 0) << 1; |
| 2267 |
} |
| 2268 |
} |
| 2269 |
|
| 2270 |
void helper_vcmpbfp (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2271 |
{
|
| 2272 |
vcmpbfp_internal(r, a, b, 0);
|
| 2273 |
} |
| 2274 |
|
| 2275 |
void helper_vcmpbfp_dot (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2276 |
{
|
| 2277 |
vcmpbfp_internal(r, a, b, 1);
|
| 2278 |
} |
| 2279 |
|
| 2280 |
#define VCT(suffix, satcvt, element) \
|
| 2281 |
void helper_vct##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t uim) \ |
| 2282 |
{ \
|
| 2283 |
int i; \
|
| 2284 |
int sat = 0; \ |
| 2285 |
float_status s = env->vec_status; \ |
| 2286 |
set_float_rounding_mode(float_round_to_zero, &s); \ |
| 2287 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
| 2288 |
if (float32_is_nan(b->f[i]) || \
|
| 2289 |
float32_is_signaling_nan(b->f[i])) { \
|
| 2290 |
r->element[i] = 0; \
|
| 2291 |
} else { \
|
| 2292 |
float64 t = float32_to_float64(b->f[i], &s); \ |
| 2293 |
int64_t j; \ |
| 2294 |
t = float64_scalbn(t, uim, &s); \ |
| 2295 |
j = float64_to_int64(t, &s); \ |
| 2296 |
r->element[i] = satcvt(j, &sat); \ |
| 2297 |
} \ |
| 2298 |
} \ |
| 2299 |
if (sat) { \
|
| 2300 |
env->vscr |= (1 << VSCR_SAT); \
|
| 2301 |
} \ |
| 2302 |
} |
| 2303 |
VCT(uxs, cvtsduw, u32) |
| 2304 |
VCT(sxs, cvtsdsw, s32) |
| 2305 |
#undef VCT
|
| 2306 |
|
| 2307 |
void helper_vmaddfp (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2308 |
{
|
| 2309 |
int i;
|
| 2310 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
| 2311 |
HANDLE_NAN3(r->f[i], a->f[i], b->f[i], c->f[i]) {
|
| 2312 |
/* Need to do the computation in higher precision and round
|
| 2313 |
* once at the end. */
|
| 2314 |
float64 af, bf, cf, t; |
| 2315 |
af = float32_to_float64(a->f[i], &env->vec_status); |
| 2316 |
bf = float32_to_float64(b->f[i], &env->vec_status); |
| 2317 |
cf = float32_to_float64(c->f[i], &env->vec_status); |
| 2318 |
t = float64_mul(af, cf, &env->vec_status); |
| 2319 |
t = float64_add(t, bf, &env->vec_status); |
| 2320 |
r->f[i] = float64_to_float32(t, &env->vec_status); |
| 2321 |
} |
| 2322 |
} |
| 2323 |
} |
| 2324 |
|
| 2325 |
void helper_vmhaddshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2326 |
{
|
| 2327 |
int sat = 0; |
| 2328 |
int i;
|
| 2329 |
|
| 2330 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
| 2331 |
int32_t prod = a->s16[i] * b->s16[i]; |
| 2332 |
int32_t t = (int32_t)c->s16[i] + (prod >> 15);
|
| 2333 |
r->s16[i] = cvtswsh (t, &sat); |
| 2334 |
} |
| 2335 |
|
| 2336 |
if (sat) {
|
| 2337 |
env->vscr |= (1 << VSCR_SAT);
|
| 2338 |
} |
| 2339 |
} |
| 2340 |
|
| 2341 |
void helper_vmhraddshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2342 |
{
|
| 2343 |
int sat = 0; |
| 2344 |
int i;
|
| 2345 |
|
| 2346 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
| 2347 |
int32_t prod = a->s16[i] * b->s16[i] + 0x00004000;
|
| 2348 |
int32_t t = (int32_t)c->s16[i] + (prod >> 15);
|
| 2349 |
r->s16[i] = cvtswsh (t, &sat); |
| 2350 |
} |
| 2351 |
|
| 2352 |
if (sat) {
|
| 2353 |
env->vscr |= (1 << VSCR_SAT);
|
| 2354 |
} |
| 2355 |
} |
| 2356 |
|
| 2357 |
#define VMINMAX_DO(name, compare, element) \
|
| 2358 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2359 |
{ \
|
| 2360 |
int i; \
|
| 2361 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2362 |
if (a->element[i] compare b->element[i]) { \
|
| 2363 |
r->element[i] = b->element[i]; \ |
| 2364 |
} else { \
|
| 2365 |
r->element[i] = a->element[i]; \ |
| 2366 |
} \ |
| 2367 |
} \ |
| 2368 |
} |
| 2369 |
#define VMINMAX(suffix, element) \
|
| 2370 |
VMINMAX_DO(min##suffix, >, element) \ |
| 2371 |
VMINMAX_DO(max##suffix, <, element) |
| 2372 |
VMINMAX(sb, s8) |
| 2373 |
VMINMAX(sh, s16) |
| 2374 |
VMINMAX(sw, s32) |
| 2375 |
VMINMAX(ub, u8) |
| 2376 |
VMINMAX(uh, u16) |
| 2377 |
VMINMAX(uw, u32) |
| 2378 |
#undef VMINMAX_DO
|
| 2379 |
#undef VMINMAX
|
| 2380 |
|
| 2381 |
#define VMINMAXFP(suffix, rT, rF) \
|
| 2382 |
void helper_v##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2383 |
{ \
|
| 2384 |
int i; \
|
| 2385 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
| 2386 |
HANDLE_NAN2(r->f[i], a->f[i], b->f[i]) { \
|
| 2387 |
if (float32_lt_quiet(a->f[i], b->f[i], &env->vec_status)) { \
|
| 2388 |
r->f[i] = rT->f[i]; \ |
| 2389 |
} else { \
|
| 2390 |
r->f[i] = rF->f[i]; \ |
| 2391 |
} \ |
| 2392 |
} \ |
| 2393 |
} \ |
| 2394 |
} |
| 2395 |
VMINMAXFP(minfp, a, b) |
| 2396 |
VMINMAXFP(maxfp, b, a) |
| 2397 |
#undef VMINMAXFP
|
| 2398 |
|
| 2399 |
void helper_vmladduhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2400 |
{
|
| 2401 |
int i;
|
| 2402 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
| 2403 |
int32_t prod = a->s16[i] * b->s16[i]; |
| 2404 |
r->s16[i] = (int16_t) (prod + c->s16[i]); |
| 2405 |
} |
| 2406 |
} |
| 2407 |
|
| 2408 |
#define VMRG_DO(name, element, highp) \
|
| 2409 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2410 |
{ \
|
| 2411 |
ppc_avr_t result; \ |
| 2412 |
int i; \
|
| 2413 |
size_t n_elems = ARRAY_SIZE(r->element); \ |
| 2414 |
for (i = 0; i < n_elems/2; i++) { \ |
| 2415 |
if (highp) { \
|
| 2416 |
result.element[i*2+HI_IDX] = a->element[i]; \
|
| 2417 |
result.element[i*2+LO_IDX] = b->element[i]; \
|
| 2418 |
} else { \
|
| 2419 |
result.element[n_elems - i*2 - (1+HI_IDX)] = b->element[n_elems - i - 1]; \ |
| 2420 |
result.element[n_elems - i*2 - (1+LO_IDX)] = a->element[n_elems - i - 1]; \ |
| 2421 |
} \ |
| 2422 |
} \ |
| 2423 |
*r = result; \ |
| 2424 |
} |
| 2425 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2426 |
#define MRGHI 0 |
| 2427 |
#define MRGLO 1 |
| 2428 |
#else
|
| 2429 |
#define MRGHI 1 |
| 2430 |
#define MRGLO 0 |
| 2431 |
#endif
|
| 2432 |
#define VMRG(suffix, element) \
|
| 2433 |
VMRG_DO(mrgl##suffix, element, MRGHI) \ |
| 2434 |
VMRG_DO(mrgh##suffix, element, MRGLO) |
| 2435 |
VMRG(b, u8) |
| 2436 |
VMRG(h, u16) |
| 2437 |
VMRG(w, u32) |
| 2438 |
#undef VMRG_DO
|
| 2439 |
#undef VMRG
|
| 2440 |
#undef MRGHI
|
| 2441 |
#undef MRGLO
|
| 2442 |
|
| 2443 |
void helper_vmsummbm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2444 |
{
|
| 2445 |
int32_t prod[16];
|
| 2446 |
int i;
|
| 2447 |
|
| 2448 |
for (i = 0; i < ARRAY_SIZE(r->s8); i++) { |
| 2449 |
prod[i] = (int32_t)a->s8[i] * b->u8[i]; |
| 2450 |
} |
| 2451 |
|
| 2452 |
VECTOR_FOR_INORDER_I(i, s32) {
|
| 2453 |
r->s32[i] = c->s32[i] + prod[4*i] + prod[4*i+1] + prod[4*i+2] + prod[4*i+3]; |
| 2454 |
} |
| 2455 |
} |
| 2456 |
|
| 2457 |
void helper_vmsumshm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2458 |
{
|
| 2459 |
int32_t prod[8];
|
| 2460 |
int i;
|
| 2461 |
|
| 2462 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
| 2463 |
prod[i] = a->s16[i] * b->s16[i]; |
| 2464 |
} |
| 2465 |
|
| 2466 |
VECTOR_FOR_INORDER_I(i, s32) {
|
| 2467 |
r->s32[i] = c->s32[i] + prod[2*i] + prod[2*i+1]; |
| 2468 |
} |
| 2469 |
} |
| 2470 |
|
| 2471 |
void helper_vmsumshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2472 |
{
|
| 2473 |
int32_t prod[8];
|
| 2474 |
int i;
|
| 2475 |
int sat = 0; |
| 2476 |
|
| 2477 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
| 2478 |
prod[i] = (int32_t)a->s16[i] * b->s16[i]; |
| 2479 |
} |
| 2480 |
|
| 2481 |
VECTOR_FOR_INORDER_I (i, s32) {
|
| 2482 |
int64_t t = (int64_t)c->s32[i] + prod[2*i] + prod[2*i+1]; |
| 2483 |
r->u32[i] = cvtsdsw(t, &sat); |
| 2484 |
} |
| 2485 |
|
| 2486 |
if (sat) {
|
| 2487 |
env->vscr |= (1 << VSCR_SAT);
|
| 2488 |
} |
| 2489 |
} |
| 2490 |
|
| 2491 |
void helper_vmsumubm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2492 |
{
|
| 2493 |
uint16_t prod[16];
|
| 2494 |
int i;
|
| 2495 |
|
| 2496 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { |
| 2497 |
prod[i] = a->u8[i] * b->u8[i]; |
| 2498 |
} |
| 2499 |
|
| 2500 |
VECTOR_FOR_INORDER_I(i, u32) {
|
| 2501 |
r->u32[i] = c->u32[i] + prod[4*i] + prod[4*i+1] + prod[4*i+2] + prod[4*i+3]; |
| 2502 |
} |
| 2503 |
} |
| 2504 |
|
| 2505 |
void helper_vmsumuhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2506 |
{
|
| 2507 |
uint32_t prod[8];
|
| 2508 |
int i;
|
| 2509 |
|
| 2510 |
for (i = 0; i < ARRAY_SIZE(r->u16); i++) { |
| 2511 |
prod[i] = a->u16[i] * b->u16[i]; |
| 2512 |
} |
| 2513 |
|
| 2514 |
VECTOR_FOR_INORDER_I(i, u32) {
|
| 2515 |
r->u32[i] = c->u32[i] + prod[2*i] + prod[2*i+1]; |
| 2516 |
} |
| 2517 |
} |
| 2518 |
|
| 2519 |
void helper_vmsumuhs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2520 |
{
|
| 2521 |
uint32_t prod[8];
|
| 2522 |
int i;
|
| 2523 |
int sat = 0; |
| 2524 |
|
| 2525 |
for (i = 0; i < ARRAY_SIZE(r->u16); i++) { |
| 2526 |
prod[i] = a->u16[i] * b->u16[i]; |
| 2527 |
} |
| 2528 |
|
| 2529 |
VECTOR_FOR_INORDER_I (i, s32) {
|
| 2530 |
uint64_t t = (uint64_t)c->u32[i] + prod[2*i] + prod[2*i+1]; |
| 2531 |
r->u32[i] = cvtuduw(t, &sat); |
| 2532 |
} |
| 2533 |
|
| 2534 |
if (sat) {
|
| 2535 |
env->vscr |= (1 << VSCR_SAT);
|
| 2536 |
} |
| 2537 |
} |
| 2538 |
|
| 2539 |
#define VMUL_DO(name, mul_element, prod_element, evenp) \
|
| 2540 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2541 |
{ \
|
| 2542 |
int i; \
|
| 2543 |
VECTOR_FOR_INORDER_I(i, prod_element) { \
|
| 2544 |
if (evenp) { \
|
| 2545 |
r->prod_element[i] = a->mul_element[i*2+HI_IDX] * b->mul_element[i*2+HI_IDX]; \ |
| 2546 |
} else { \
|
| 2547 |
r->prod_element[i] = a->mul_element[i*2+LO_IDX] * b->mul_element[i*2+LO_IDX]; \ |
| 2548 |
} \ |
| 2549 |
} \ |
| 2550 |
} |
| 2551 |
#define VMUL(suffix, mul_element, prod_element) \
|
| 2552 |
VMUL_DO(mule##suffix, mul_element, prod_element, 1) \ |
| 2553 |
VMUL_DO(mulo##suffix, mul_element, prod_element, 0) |
| 2554 |
VMUL(sb, s8, s16) |
| 2555 |
VMUL(sh, s16, s32) |
| 2556 |
VMUL(ub, u8, u16) |
| 2557 |
VMUL(uh, u16, u32) |
| 2558 |
#undef VMUL_DO
|
| 2559 |
#undef VMUL
|
| 2560 |
|
| 2561 |
void helper_vnmsubfp (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2562 |
{
|
| 2563 |
int i;
|
| 2564 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
| 2565 |
HANDLE_NAN3(r->f[i], a->f[i], b->f[i], c->f[i]) {
|
| 2566 |
/* Need to do the computation is higher precision and round
|
| 2567 |
* once at the end. */
|
| 2568 |
float64 af, bf, cf, t; |
| 2569 |
af = float32_to_float64(a->f[i], &env->vec_status); |
| 2570 |
bf = float32_to_float64(b->f[i], &env->vec_status); |
| 2571 |
cf = float32_to_float64(c->f[i], &env->vec_status); |
| 2572 |
t = float64_mul(af, cf, &env->vec_status); |
| 2573 |
t = float64_sub(t, bf, &env->vec_status); |
| 2574 |
t = float64_chs(t); |
| 2575 |
r->f[i] = float64_to_float32(t, &env->vec_status); |
| 2576 |
} |
| 2577 |
} |
| 2578 |
} |
| 2579 |
|
| 2580 |
void helper_vperm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2581 |
{
|
| 2582 |
ppc_avr_t result; |
| 2583 |
int i;
|
| 2584 |
VECTOR_FOR_INORDER_I (i, u8) {
|
| 2585 |
int s = c->u8[i] & 0x1f; |
| 2586 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2587 |
int index = s & 0xf; |
| 2588 |
#else
|
| 2589 |
int index = 15 - (s & 0xf); |
| 2590 |
#endif
|
| 2591 |
if (s & 0x10) { |
| 2592 |
result.u8[i] = b->u8[index]; |
| 2593 |
} else {
|
| 2594 |
result.u8[i] = a->u8[index]; |
| 2595 |
} |
| 2596 |
} |
| 2597 |
*r = result; |
| 2598 |
} |
| 2599 |
|
| 2600 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2601 |
#define PKBIG 1 |
| 2602 |
#else
|
| 2603 |
#define PKBIG 0 |
| 2604 |
#endif
|
| 2605 |
void helper_vpkpx (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2606 |
{
|
| 2607 |
int i, j;
|
| 2608 |
ppc_avr_t result; |
| 2609 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2610 |
const ppc_avr_t *x[2] = { a, b }; |
| 2611 |
#else
|
| 2612 |
const ppc_avr_t *x[2] = { b, a }; |
| 2613 |
#endif
|
| 2614 |
|
| 2615 |
VECTOR_FOR_INORDER_I (i, u64) {
|
| 2616 |
VECTOR_FOR_INORDER_I (j, u32){
|
| 2617 |
uint32_t e = x[i]->u32[j]; |
| 2618 |
result.u16[4*i+j] = (((e >> 9) & 0xfc00) | |
| 2619 |
((e >> 6) & 0x3e0) | |
| 2620 |
((e >> 3) & 0x1f)); |
| 2621 |
} |
| 2622 |
} |
| 2623 |
*r = result; |
| 2624 |
} |
| 2625 |
|
| 2626 |
#define VPK(suffix, from, to, cvt, dosat) \
|
| 2627 |
void helper_vpk##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2628 |
{ \
|
| 2629 |
int i; \
|
| 2630 |
int sat = 0; \ |
| 2631 |
ppc_avr_t result; \ |
| 2632 |
ppc_avr_t *a0 = PKBIG ? a : b; \ |
| 2633 |
ppc_avr_t *a1 = PKBIG ? b : a; \ |
| 2634 |
VECTOR_FOR_INORDER_I (i, from) { \
|
| 2635 |
result.to[i] = cvt(a0->from[i], &sat); \ |
| 2636 |
result.to[i+ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat); \ |
| 2637 |
} \ |
| 2638 |
*r = result; \ |
| 2639 |
if (dosat && sat) { \
|
| 2640 |
env->vscr |= (1 << VSCR_SAT); \
|
| 2641 |
} \ |
| 2642 |
} |
| 2643 |
#define I(x, y) (x)
|
| 2644 |
VPK(shss, s16, s8, cvtshsb, 1)
|
| 2645 |
VPK(shus, s16, u8, cvtshub, 1)
|
| 2646 |
VPK(swss, s32, s16, cvtswsh, 1)
|
| 2647 |
VPK(swus, s32, u16, cvtswuh, 1)
|
| 2648 |
VPK(uhus, u16, u8, cvtuhub, 1)
|
| 2649 |
VPK(uwus, u32, u16, cvtuwuh, 1)
|
| 2650 |
VPK(uhum, u16, u8, I, 0)
|
| 2651 |
VPK(uwum, u32, u16, I, 0)
|
| 2652 |
#undef I
|
| 2653 |
#undef VPK
|
| 2654 |
#undef PKBIG
|
| 2655 |
|
| 2656 |
void helper_vrefp (ppc_avr_t *r, ppc_avr_t *b)
|
| 2657 |
{
|
| 2658 |
int i;
|
| 2659 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
| 2660 |
HANDLE_NAN1(r->f[i], b->f[i]) {
|
| 2661 |
r->f[i] = float32_div(float32_one, b->f[i], &env->vec_status); |
| 2662 |
} |
| 2663 |
} |
| 2664 |
} |
| 2665 |
|
| 2666 |
#define VRFI(suffix, rounding) \
|
| 2667 |
void helper_vrfi##suffix (ppc_avr_t *r, ppc_avr_t *b) \ |
| 2668 |
{ \
|
| 2669 |
int i; \
|
| 2670 |
float_status s = env->vec_status; \ |
| 2671 |
set_float_rounding_mode(rounding, &s); \ |
| 2672 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
| 2673 |
HANDLE_NAN1(r->f[i], b->f[i]) { \
|
| 2674 |
r->f[i] = float32_round_to_int (b->f[i], &s); \ |
| 2675 |
} \ |
| 2676 |
} \ |
| 2677 |
} |
| 2678 |
VRFI(n, float_round_nearest_even) |
| 2679 |
VRFI(m, float_round_down) |
| 2680 |
VRFI(p, float_round_up) |
| 2681 |
VRFI(z, float_round_to_zero) |
| 2682 |
#undef VRFI
|
| 2683 |
|
| 2684 |
#define VROTATE(suffix, element) \
|
| 2685 |
void helper_vrl##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2686 |
{ \
|
| 2687 |
int i; \
|
| 2688 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2689 |
unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \ |
| 2690 |
unsigned int shift = b->element[i] & mask; \ |
| 2691 |
r->element[i] = (a->element[i] << shift) | (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \ |
| 2692 |
} \ |
| 2693 |
} |
| 2694 |
VROTATE(b, u8) |
| 2695 |
VROTATE(h, u16) |
| 2696 |
VROTATE(w, u32) |
| 2697 |
#undef VROTATE
|
| 2698 |
|
| 2699 |
void helper_vrsqrtefp (ppc_avr_t *r, ppc_avr_t *b)
|
| 2700 |
{
|
| 2701 |
int i;
|
| 2702 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
| 2703 |
HANDLE_NAN1(r->f[i], b->f[i]) {
|
| 2704 |
float32 t = float32_sqrt(b->f[i], &env->vec_status); |
| 2705 |
r->f[i] = float32_div(float32_one, t, &env->vec_status); |
| 2706 |
} |
| 2707 |
} |
| 2708 |
} |
| 2709 |
|
| 2710 |
void helper_vsel (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
| 2711 |
{
|
| 2712 |
r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]); |
| 2713 |
r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]); |
| 2714 |
} |
| 2715 |
|
| 2716 |
void helper_vlogefp (ppc_avr_t *r, ppc_avr_t *b)
|
| 2717 |
{
|
| 2718 |
int i;
|
| 2719 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
| 2720 |
HANDLE_NAN1(r->f[i], b->f[i]) {
|
| 2721 |
r->f[i] = float32_log2(b->f[i], &env->vec_status); |
| 2722 |
} |
| 2723 |
} |
| 2724 |
} |
| 2725 |
|
| 2726 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2727 |
#define LEFT 0 |
| 2728 |
#define RIGHT 1 |
| 2729 |
#else
|
| 2730 |
#define LEFT 1 |
| 2731 |
#define RIGHT 0 |
| 2732 |
#endif
|
| 2733 |
/* The specification says that the results are undefined if all of the
|
| 2734 |
* shift counts are not identical. We check to make sure that they are
|
| 2735 |
* to conform to what real hardware appears to do. */
|
| 2736 |
#define VSHIFT(suffix, leftp) \
|
| 2737 |
void helper_vs##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2738 |
{ \
|
| 2739 |
int shift = b->u8[LO_IDX*15] & 0x7; \ |
| 2740 |
int doit = 1; \ |
| 2741 |
int i; \
|
| 2742 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { \ |
| 2743 |
doit = doit && ((b->u8[i] & 0x7) == shift); \
|
| 2744 |
} \ |
| 2745 |
if (doit) { \
|
| 2746 |
if (shift == 0) { \ |
| 2747 |
*r = *a; \ |
| 2748 |
} else if (leftp) { \ |
| 2749 |
uint64_t carry = a->u64[LO_IDX] >> (64 - shift); \
|
| 2750 |
r->u64[HI_IDX] = (a->u64[HI_IDX] << shift) | carry; \ |
| 2751 |
r->u64[LO_IDX] = a->u64[LO_IDX] << shift; \ |
| 2752 |
} else { \
|
| 2753 |
uint64_t carry = a->u64[HI_IDX] << (64 - shift); \
|
| 2754 |
r->u64[LO_IDX] = (a->u64[LO_IDX] >> shift) | carry; \ |
| 2755 |
r->u64[HI_IDX] = a->u64[HI_IDX] >> shift; \ |
| 2756 |
} \ |
| 2757 |
} \ |
| 2758 |
} |
| 2759 |
VSHIFT(l, LEFT) |
| 2760 |
VSHIFT(r, RIGHT) |
| 2761 |
#undef VSHIFT
|
| 2762 |
#undef LEFT
|
| 2763 |
#undef RIGHT
|
| 2764 |
|
| 2765 |
#define VSL(suffix, element) \
|
| 2766 |
void helper_vsl##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2767 |
{ \
|
| 2768 |
int i; \
|
| 2769 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2770 |
unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \ |
| 2771 |
unsigned int shift = b->element[i] & mask; \ |
| 2772 |
r->element[i] = a->element[i] << shift; \ |
| 2773 |
} \ |
| 2774 |
} |
| 2775 |
VSL(b, u8) |
| 2776 |
VSL(h, u16) |
| 2777 |
VSL(w, u32) |
| 2778 |
#undef VSL
|
| 2779 |
|
| 2780 |
void helper_vsldoi (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift)
|
| 2781 |
{
|
| 2782 |
int sh = shift & 0xf; |
| 2783 |
int i;
|
| 2784 |
ppc_avr_t result; |
| 2785 |
|
| 2786 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2787 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { |
| 2788 |
int index = sh + i;
|
| 2789 |
if (index > 0xf) { |
| 2790 |
result.u8[i] = b->u8[index-0x10];
|
| 2791 |
} else {
|
| 2792 |
result.u8[i] = a->u8[index]; |
| 2793 |
} |
| 2794 |
} |
| 2795 |
#else
|
| 2796 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { |
| 2797 |
int index = (16 - sh) + i; |
| 2798 |
if (index > 0xf) { |
| 2799 |
result.u8[i] = a->u8[index-0x10];
|
| 2800 |
} else {
|
| 2801 |
result.u8[i] = b->u8[index]; |
| 2802 |
} |
| 2803 |
} |
| 2804 |
#endif
|
| 2805 |
*r = result; |
| 2806 |
} |
| 2807 |
|
| 2808 |
void helper_vslo (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2809 |
{
|
| 2810 |
int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf; |
| 2811 |
|
| 2812 |
#if defined (HOST_WORDS_BIGENDIAN)
|
| 2813 |
memmove (&r->u8[0], &a->u8[sh], 16-sh); |
| 2814 |
memset (&r->u8[16-sh], 0, sh); |
| 2815 |
#else
|
| 2816 |
memmove (&r->u8[sh], &a->u8[0], 16-sh); |
| 2817 |
memset (&r->u8[0], 0, sh); |
| 2818 |
#endif
|
| 2819 |
} |
| 2820 |
|
| 2821 |
/* Experimental testing shows that hardware masks the immediate. */
|
| 2822 |
#define _SPLAT_MASKED(element) (splat & (ARRAY_SIZE(r->element) - 1)) |
| 2823 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2824 |
#define SPLAT_ELEMENT(element) _SPLAT_MASKED(element)
|
| 2825 |
#else
|
| 2826 |
#define SPLAT_ELEMENT(element) (ARRAY_SIZE(r->element)-1 - _SPLAT_MASKED(element)) |
| 2827 |
#endif
|
| 2828 |
#define VSPLT(suffix, element) \
|
| 2829 |
void helper_vsplt##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t splat) \ |
| 2830 |
{ \
|
| 2831 |
uint32_t s = b->element[SPLAT_ELEMENT(element)]; \ |
| 2832 |
int i; \
|
| 2833 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2834 |
r->element[i] = s; \ |
| 2835 |
} \ |
| 2836 |
} |
| 2837 |
VSPLT(b, u8) |
| 2838 |
VSPLT(h, u16) |
| 2839 |
VSPLT(w, u32) |
| 2840 |
#undef VSPLT
|
| 2841 |
#undef SPLAT_ELEMENT
|
| 2842 |
#undef _SPLAT_MASKED
|
| 2843 |
|
| 2844 |
#define VSPLTI(suffix, element, splat_type) \
|
| 2845 |
void helper_vspltis##suffix (ppc_avr_t *r, uint32_t splat) \ |
| 2846 |
{ \
|
| 2847 |
splat_type x = (int8_t)(splat << 3) >> 3; \ |
| 2848 |
int i; \
|
| 2849 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2850 |
r->element[i] = x; \ |
| 2851 |
} \ |
| 2852 |
} |
| 2853 |
VSPLTI(b, s8, int8_t) |
| 2854 |
VSPLTI(h, s16, int16_t) |
| 2855 |
VSPLTI(w, s32, int32_t) |
| 2856 |
#undef VSPLTI
|
| 2857 |
|
| 2858 |
#define VSR(suffix, element) \
|
| 2859 |
void helper_vsr##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
| 2860 |
{ \
|
| 2861 |
int i; \
|
| 2862 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
| 2863 |
unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \ |
| 2864 |
unsigned int shift = b->element[i] & mask; \ |
| 2865 |
r->element[i] = a->element[i] >> shift; \ |
| 2866 |
} \ |
| 2867 |
} |
| 2868 |
VSR(ab, s8) |
| 2869 |
VSR(ah, s16) |
| 2870 |
VSR(aw, s32) |
| 2871 |
VSR(b, u8) |
| 2872 |
VSR(h, u16) |
| 2873 |
VSR(w, u32) |
| 2874 |
#undef VSR
|
| 2875 |
|
| 2876 |
void helper_vsro (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2877 |
{
|
| 2878 |
int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf; |
| 2879 |
|
| 2880 |
#if defined (HOST_WORDS_BIGENDIAN)
|
| 2881 |
memmove (&r->u8[sh], &a->u8[0], 16-sh); |
| 2882 |
memset (&r->u8[0], 0, sh); |
| 2883 |
#else
|
| 2884 |
memmove (&r->u8[0], &a->u8[sh], 16-sh); |
| 2885 |
memset (&r->u8[16-sh], 0, sh); |
| 2886 |
#endif
|
| 2887 |
} |
| 2888 |
|
| 2889 |
void helper_vsubcuw (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2890 |
{
|
| 2891 |
int i;
|
| 2892 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { |
| 2893 |
r->u32[i] = a->u32[i] >= b->u32[i]; |
| 2894 |
} |
| 2895 |
} |
| 2896 |
|
| 2897 |
void helper_vsumsws (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2898 |
{
|
| 2899 |
int64_t t; |
| 2900 |
int i, upper;
|
| 2901 |
ppc_avr_t result; |
| 2902 |
int sat = 0; |
| 2903 |
|
| 2904 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2905 |
upper = ARRAY_SIZE(r->s32)-1;
|
| 2906 |
#else
|
| 2907 |
upper = 0;
|
| 2908 |
#endif
|
| 2909 |
t = (int64_t)b->s32[upper]; |
| 2910 |
for (i = 0; i < ARRAY_SIZE(r->s32); i++) { |
| 2911 |
t += a->s32[i]; |
| 2912 |
result.s32[i] = 0;
|
| 2913 |
} |
| 2914 |
result.s32[upper] = cvtsdsw(t, &sat); |
| 2915 |
*r = result; |
| 2916 |
|
| 2917 |
if (sat) {
|
| 2918 |
env->vscr |= (1 << VSCR_SAT);
|
| 2919 |
} |
| 2920 |
} |
| 2921 |
|
| 2922 |
void helper_vsum2sws (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2923 |
{
|
| 2924 |
int i, j, upper;
|
| 2925 |
ppc_avr_t result; |
| 2926 |
int sat = 0; |
| 2927 |
|
| 2928 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 2929 |
upper = 1;
|
| 2930 |
#else
|
| 2931 |
upper = 0;
|
| 2932 |
#endif
|
| 2933 |
for (i = 0; i < ARRAY_SIZE(r->u64); i++) { |
| 2934 |
int64_t t = (int64_t)b->s32[upper+i*2];
|
| 2935 |
result.u64[i] = 0;
|
| 2936 |
for (j = 0; j < ARRAY_SIZE(r->u64); j++) { |
| 2937 |
t += a->s32[2*i+j];
|
| 2938 |
} |
| 2939 |
result.s32[upper+i*2] = cvtsdsw(t, &sat);
|
| 2940 |
} |
| 2941 |
|
| 2942 |
*r = result; |
| 2943 |
if (sat) {
|
| 2944 |
env->vscr |= (1 << VSCR_SAT);
|
| 2945 |
} |
| 2946 |
} |
| 2947 |
|
| 2948 |
void helper_vsum4sbs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2949 |
{
|
| 2950 |
int i, j;
|
| 2951 |
int sat = 0; |
| 2952 |
|
| 2953 |
for (i = 0; i < ARRAY_SIZE(r->s32); i++) { |
| 2954 |
int64_t t = (int64_t)b->s32[i]; |
| 2955 |
for (j = 0; j < ARRAY_SIZE(r->s32); j++) { |
| 2956 |
t += a->s8[4*i+j];
|
| 2957 |
} |
| 2958 |
r->s32[i] = cvtsdsw(t, &sat); |
| 2959 |
} |
| 2960 |
|
| 2961 |
if (sat) {
|
| 2962 |
env->vscr |= (1 << VSCR_SAT);
|
| 2963 |
} |
| 2964 |
} |
| 2965 |
|
| 2966 |
void helper_vsum4shs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2967 |
{
|
| 2968 |
int sat = 0; |
| 2969 |
int i;
|
| 2970 |
|
| 2971 |
for (i = 0; i < ARRAY_SIZE(r->s32); i++) { |
| 2972 |
int64_t t = (int64_t)b->s32[i]; |
| 2973 |
t += a->s16[2*i] + a->s16[2*i+1]; |
| 2974 |
r->s32[i] = cvtsdsw(t, &sat); |
| 2975 |
} |
| 2976 |
|
| 2977 |
if (sat) {
|
| 2978 |
env->vscr |= (1 << VSCR_SAT);
|
| 2979 |
} |
| 2980 |
} |
| 2981 |
|
| 2982 |
void helper_vsum4ubs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
| 2983 |
{
|
| 2984 |
int i, j;
|
| 2985 |
int sat = 0; |
| 2986 |
|
| 2987 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { |
| 2988 |
uint64_t t = (uint64_t)b->u32[i]; |
| 2989 |
for (j = 0; j < ARRAY_SIZE(r->u32); j++) { |
| 2990 |
t += a->u8[4*i+j];
|
| 2991 |
} |
| 2992 |
r->u32[i] = cvtuduw(t, &sat); |
| 2993 |
} |
| 2994 |
|
| 2995 |
if (sat) {
|
| 2996 |
env->vscr |= (1 << VSCR_SAT);
|
| 2997 |
} |
| 2998 |
} |
| 2999 |
|
| 3000 |
#if defined(HOST_WORDS_BIGENDIAN)
|
| 3001 |
#define UPKHI 1 |
| 3002 |
#define UPKLO 0 |
| 3003 |
#else
|
| 3004 |
#define UPKHI 0 |
| 3005 |
#define UPKLO 1 |
| 3006 |
#endif
|
| 3007 |
#define VUPKPX(suffix, hi) \
|
| 3008 |
void helper_vupk##suffix (ppc_avr_t *r, ppc_avr_t *b) \ |
| 3009 |
{ \
|
| 3010 |
int i; \
|
| 3011 |
ppc_avr_t result; \ |
| 3012 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \ |
| 3013 |
uint16_t e = b->u16[hi ? i : i+4]; \
|
| 3014 |
uint8_t a = (e >> 15) ? 0xff : 0; \ |
| 3015 |
uint8_t r = (e >> 10) & 0x1f; \ |
| 3016 |
uint8_t g = (e >> 5) & 0x1f; \ |
| 3017 |
uint8_t b = e & 0x1f; \
|
| 3018 |
result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \ |
| 3019 |
} \ |
| 3020 |
*r = result; \ |
| 3021 |
} |
| 3022 |
VUPKPX(lpx, UPKLO) |
| 3023 |
VUPKPX(hpx, UPKHI) |
| 3024 |
#undef VUPKPX
|
| 3025 |
|
| 3026 |
#define VUPK(suffix, unpacked, packee, hi) \
|
| 3027 |
void helper_vupk##suffix (ppc_avr_t *r, ppc_avr_t *b) \ |
| 3028 |
{ \
|
| 3029 |
int i; \
|
| 3030 |
ppc_avr_t result; \ |
| 3031 |
if (hi) { \
|
| 3032 |
for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \ |
| 3033 |
result.unpacked[i] = b->packee[i]; \ |
| 3034 |
} \ |
| 3035 |
} else { \
|
| 3036 |
for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); i++) { \
|
| 3037 |
result.unpacked[i-ARRAY_SIZE(r->unpacked)] = b->packee[i]; \ |
| 3038 |
} \ |
| 3039 |
} \ |
| 3040 |
*r = result; \ |
| 3041 |
} |
| 3042 |
VUPK(hsb, s16, s8, UPKHI) |
| 3043 |
VUPK(hsh, s32, s16, UPKHI) |
| 3044 |
VUPK(lsb, s16, s8, UPKLO) |
| 3045 |
VUPK(lsh, s32, s16, UPKLO) |
| 3046 |
#undef VUPK
|
| 3047 |
#undef UPKHI
|
| 3048 |
#undef UPKLO
|
| 3049 |
|
| 3050 |
#undef DO_HANDLE_NAN
|
| 3051 |
#undef HANDLE_NAN1
|
| 3052 |
#undef HANDLE_NAN2
|
| 3053 |
#undef HANDLE_NAN3
|
| 3054 |
#undef VECTOR_FOR_INORDER_I
|
| 3055 |
#undef HI_IDX
|
| 3056 |
#undef LO_IDX
|
| 3057 |
|
| 3058 |
/*****************************************************************************/
|
| 3059 |
/* SPE extension helpers */
|
| 3060 |
/* Use a table to make this quicker */
|
| 3061 |
static uint8_t hbrev[16] = { |
| 3062 |
0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE, |
| 3063 |
0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF, |
| 3064 |
}; |
| 3065 |
|
| 3066 |
static always_inline uint8_t byte_reverse (uint8_t val)
|
| 3067 |
{
|
| 3068 |
return hbrev[val >> 4] | (hbrev[val & 0xF] << 4); |
| 3069 |
} |
| 3070 |
|
| 3071 |
static always_inline uint32_t word_reverse (uint32_t val)
|
| 3072 |
{
|
| 3073 |
return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) | |
| 3074 |
(byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24); |
| 3075 |
} |
| 3076 |
|
| 3077 |
#define MASKBITS 16 // Random value - to be fixed (implementation dependant) |
| 3078 |
target_ulong helper_brinc (target_ulong arg1, target_ulong arg2) |
| 3079 |
{
|
| 3080 |
uint32_t a, b, d, mask; |
| 3081 |
|
| 3082 |
mask = UINT32_MAX >> (32 - MASKBITS);
|
| 3083 |
a = arg1 & mask; |
| 3084 |
b = arg2 & mask; |
| 3085 |
d = word_reverse(1 + word_reverse(a | ~b));
|
| 3086 |
return (arg1 & ~mask) | (d & b);
|
| 3087 |
} |
| 3088 |
|
| 3089 |
uint32_t helper_cntlsw32 (uint32_t val) |
| 3090 |
{
|
| 3091 |
if (val & 0x80000000) |
| 3092 |
return clz32(~val);
|
| 3093 |
else
|
| 3094 |
return clz32(val);
|
| 3095 |
} |
| 3096 |
|
| 3097 |
uint32_t helper_cntlzw32 (uint32_t val) |
| 3098 |
{
|
| 3099 |
return clz32(val);
|
| 3100 |
} |
| 3101 |
|
| 3102 |
/* Single-precision floating-point conversions */
|
| 3103 |
static always_inline uint32_t efscfsi (uint32_t val)
|
| 3104 |
{
|
| 3105 |
CPU_FloatU u; |
| 3106 |
|
| 3107 |
u.f = int32_to_float32(val, &env->vec_status); |
| 3108 |
|
| 3109 |
return u.l;
|
| 3110 |
} |
| 3111 |
|
| 3112 |
static always_inline uint32_t efscfui (uint32_t val)
|
| 3113 |
{
|
| 3114 |
CPU_FloatU u; |
| 3115 |
|
| 3116 |
u.f = uint32_to_float32(val, &env->vec_status); |
| 3117 |
|
| 3118 |
return u.l;
|
| 3119 |
} |
| 3120 |
|
| 3121 |
static always_inline int32_t efsctsi (uint32_t val)
|
| 3122 |
{
|
| 3123 |
CPU_FloatU u; |
| 3124 |
|
| 3125 |
u.l = val; |
| 3126 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3127 |
if (unlikely(float32_is_nan(u.f)))
|
| 3128 |
return 0; |
| 3129 |
|
| 3130 |
return float32_to_int32(u.f, &env->vec_status);
|
| 3131 |
} |
| 3132 |
|
| 3133 |
static always_inline uint32_t efsctui (uint32_t val)
|
| 3134 |
{
|
| 3135 |
CPU_FloatU u; |
| 3136 |
|
| 3137 |
u.l = val; |
| 3138 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3139 |
if (unlikely(float32_is_nan(u.f)))
|
| 3140 |
return 0; |
| 3141 |
|
| 3142 |
return float32_to_uint32(u.f, &env->vec_status);
|
| 3143 |
} |
| 3144 |
|
| 3145 |
static always_inline uint32_t efsctsiz (uint32_t val)
|
| 3146 |
{
|
| 3147 |
CPU_FloatU u; |
| 3148 |
|
| 3149 |
u.l = val; |
| 3150 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3151 |
if (unlikely(float32_is_nan(u.f)))
|
| 3152 |
return 0; |
| 3153 |
|
| 3154 |
return float32_to_int32_round_to_zero(u.f, &env->vec_status);
|
| 3155 |
} |
| 3156 |
|
| 3157 |
static always_inline uint32_t efsctuiz (uint32_t val)
|
| 3158 |
{
|
| 3159 |
CPU_FloatU u; |
| 3160 |
|
| 3161 |
u.l = val; |
| 3162 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3163 |
if (unlikely(float32_is_nan(u.f)))
|
| 3164 |
return 0; |
| 3165 |
|
| 3166 |
return float32_to_uint32_round_to_zero(u.f, &env->vec_status);
|
| 3167 |
} |
| 3168 |
|
| 3169 |
static always_inline uint32_t efscfsf (uint32_t val)
|
| 3170 |
{
|
| 3171 |
CPU_FloatU u; |
| 3172 |
float32 tmp; |
| 3173 |
|
| 3174 |
u.f = int32_to_float32(val, &env->vec_status); |
| 3175 |
tmp = int64_to_float32(1ULL << 32, &env->vec_status); |
| 3176 |
u.f = float32_div(u.f, tmp, &env->vec_status); |
| 3177 |
|
| 3178 |
return u.l;
|
| 3179 |
} |
| 3180 |
|
| 3181 |
static always_inline uint32_t efscfuf (uint32_t val)
|
| 3182 |
{
|
| 3183 |
CPU_FloatU u; |
| 3184 |
float32 tmp; |
| 3185 |
|
| 3186 |
u.f = uint32_to_float32(val, &env->vec_status); |
| 3187 |
tmp = uint64_to_float32(1ULL << 32, &env->vec_status); |
| 3188 |
u.f = float32_div(u.f, tmp, &env->vec_status); |
| 3189 |
|
| 3190 |
return u.l;
|
| 3191 |
} |
| 3192 |
|
| 3193 |
static always_inline uint32_t efsctsf (uint32_t val)
|
| 3194 |
{
|
| 3195 |
CPU_FloatU u; |
| 3196 |
float32 tmp; |
| 3197 |
|
| 3198 |
u.l = val; |
| 3199 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3200 |
if (unlikely(float32_is_nan(u.f)))
|
| 3201 |
return 0; |
| 3202 |
tmp = uint64_to_float32(1ULL << 32, &env->vec_status); |
| 3203 |
u.f = float32_mul(u.f, tmp, &env->vec_status); |
| 3204 |
|
| 3205 |
return float32_to_int32(u.f, &env->vec_status);
|
| 3206 |
} |
| 3207 |
|
| 3208 |
static always_inline uint32_t efsctuf (uint32_t val)
|
| 3209 |
{
|
| 3210 |
CPU_FloatU u; |
| 3211 |
float32 tmp; |
| 3212 |
|
| 3213 |
u.l = val; |
| 3214 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3215 |
if (unlikely(float32_is_nan(u.f)))
|
| 3216 |
return 0; |
| 3217 |
tmp = uint64_to_float32(1ULL << 32, &env->vec_status); |
| 3218 |
u.f = float32_mul(u.f, tmp, &env->vec_status); |
| 3219 |
|
| 3220 |
return float32_to_uint32(u.f, &env->vec_status);
|
| 3221 |
} |
| 3222 |
|
| 3223 |
#define HELPER_SPE_SINGLE_CONV(name) \
|
| 3224 |
uint32_t helper_e##name (uint32_t val) \ |
| 3225 |
{ \
|
| 3226 |
return e##name(val); \ |
| 3227 |
} |
| 3228 |
/* efscfsi */
|
| 3229 |
HELPER_SPE_SINGLE_CONV(fscfsi); |
| 3230 |
/* efscfui */
|
| 3231 |
HELPER_SPE_SINGLE_CONV(fscfui); |
| 3232 |
/* efscfuf */
|
| 3233 |
HELPER_SPE_SINGLE_CONV(fscfuf); |
| 3234 |
/* efscfsf */
|
| 3235 |
HELPER_SPE_SINGLE_CONV(fscfsf); |
| 3236 |
/* efsctsi */
|
| 3237 |
HELPER_SPE_SINGLE_CONV(fsctsi); |
| 3238 |
/* efsctui */
|
| 3239 |
HELPER_SPE_SINGLE_CONV(fsctui); |
| 3240 |
/* efsctsiz */
|
| 3241 |
HELPER_SPE_SINGLE_CONV(fsctsiz); |
| 3242 |
/* efsctuiz */
|
| 3243 |
HELPER_SPE_SINGLE_CONV(fsctuiz); |
| 3244 |
/* efsctsf */
|
| 3245 |
HELPER_SPE_SINGLE_CONV(fsctsf); |
| 3246 |
/* efsctuf */
|
| 3247 |
HELPER_SPE_SINGLE_CONV(fsctuf); |
| 3248 |
|
| 3249 |
#define HELPER_SPE_VECTOR_CONV(name) \
|
| 3250 |
uint64_t helper_ev##name (uint64_t val) \ |
| 3251 |
{ \
|
| 3252 |
return ((uint64_t)e##name(val >> 32) << 32) | \ |
| 3253 |
(uint64_t)e##name(val); \ |
| 3254 |
} |
| 3255 |
/* evfscfsi */
|
| 3256 |
HELPER_SPE_VECTOR_CONV(fscfsi); |
| 3257 |
/* evfscfui */
|
| 3258 |
HELPER_SPE_VECTOR_CONV(fscfui); |
| 3259 |
/* evfscfuf */
|
| 3260 |
HELPER_SPE_VECTOR_CONV(fscfuf); |
| 3261 |
/* evfscfsf */
|
| 3262 |
HELPER_SPE_VECTOR_CONV(fscfsf); |
| 3263 |
/* evfsctsi */
|
| 3264 |
HELPER_SPE_VECTOR_CONV(fsctsi); |
| 3265 |
/* evfsctui */
|
| 3266 |
HELPER_SPE_VECTOR_CONV(fsctui); |
| 3267 |
/* evfsctsiz */
|
| 3268 |
HELPER_SPE_VECTOR_CONV(fsctsiz); |
| 3269 |
/* evfsctuiz */
|
| 3270 |
HELPER_SPE_VECTOR_CONV(fsctuiz); |
| 3271 |
/* evfsctsf */
|
| 3272 |
HELPER_SPE_VECTOR_CONV(fsctsf); |
| 3273 |
/* evfsctuf */
|
| 3274 |
HELPER_SPE_VECTOR_CONV(fsctuf); |
| 3275 |
|
| 3276 |
/* Single-precision floating-point arithmetic */
|
| 3277 |
static always_inline uint32_t efsadd (uint32_t op1, uint32_t op2)
|
| 3278 |
{
|
| 3279 |
CPU_FloatU u1, u2; |
| 3280 |
u1.l = op1; |
| 3281 |
u2.l = op2; |
| 3282 |
u1.f = float32_add(u1.f, u2.f, &env->vec_status); |
| 3283 |
return u1.l;
|
| 3284 |
} |
| 3285 |
|
| 3286 |
static always_inline uint32_t efssub (uint32_t op1, uint32_t op2)
|
| 3287 |
{
|
| 3288 |
CPU_FloatU u1, u2; |
| 3289 |
u1.l = op1; |
| 3290 |
u2.l = op2; |
| 3291 |
u1.f = float32_sub(u1.f, u2.f, &env->vec_status); |
| 3292 |
return u1.l;
|
| 3293 |
} |
| 3294 |
|
| 3295 |
static always_inline uint32_t efsmul (uint32_t op1, uint32_t op2)
|
| 3296 |
{
|
| 3297 |
CPU_FloatU u1, u2; |
| 3298 |
u1.l = op1; |
| 3299 |
u2.l = op2; |
| 3300 |
u1.f = float32_mul(u1.f, u2.f, &env->vec_status); |
| 3301 |
return u1.l;
|
| 3302 |
} |
| 3303 |
|
| 3304 |
static always_inline uint32_t efsdiv (uint32_t op1, uint32_t op2)
|
| 3305 |
{
|
| 3306 |
CPU_FloatU u1, u2; |
| 3307 |
u1.l = op1; |
| 3308 |
u2.l = op2; |
| 3309 |
u1.f = float32_div(u1.f, u2.f, &env->vec_status); |
| 3310 |
return u1.l;
|
| 3311 |
} |
| 3312 |
|
| 3313 |
#define HELPER_SPE_SINGLE_ARITH(name) \
|
| 3314 |
uint32_t helper_e##name (uint32_t op1, uint32_t op2) \ |
| 3315 |
{ \
|
| 3316 |
return e##name(op1, op2); \ |
| 3317 |
} |
| 3318 |
/* efsadd */
|
| 3319 |
HELPER_SPE_SINGLE_ARITH(fsadd); |
| 3320 |
/* efssub */
|
| 3321 |
HELPER_SPE_SINGLE_ARITH(fssub); |
| 3322 |
/* efsmul */
|
| 3323 |
HELPER_SPE_SINGLE_ARITH(fsmul); |
| 3324 |
/* efsdiv */
|
| 3325 |
HELPER_SPE_SINGLE_ARITH(fsdiv); |
| 3326 |
|
| 3327 |
#define HELPER_SPE_VECTOR_ARITH(name) \
|
| 3328 |
uint64_t helper_ev##name (uint64_t op1, uint64_t op2) \ |
| 3329 |
{ \
|
| 3330 |
return ((uint64_t)e##name(op1 >> 32, op2 >> 32) << 32) | \ |
| 3331 |
(uint64_t)e##name(op1, op2); \ |
| 3332 |
} |
| 3333 |
/* evfsadd */
|
| 3334 |
HELPER_SPE_VECTOR_ARITH(fsadd); |
| 3335 |
/* evfssub */
|
| 3336 |
HELPER_SPE_VECTOR_ARITH(fssub); |
| 3337 |
/* evfsmul */
|
| 3338 |
HELPER_SPE_VECTOR_ARITH(fsmul); |
| 3339 |
/* evfsdiv */
|
| 3340 |
HELPER_SPE_VECTOR_ARITH(fsdiv); |
| 3341 |
|
| 3342 |
/* Single-precision floating-point comparisons */
|
| 3343 |
static always_inline uint32_t efststlt (uint32_t op1, uint32_t op2)
|
| 3344 |
{
|
| 3345 |
CPU_FloatU u1, u2; |
| 3346 |
u1.l = op1; |
| 3347 |
u2.l = op2; |
| 3348 |
return float32_lt(u1.f, u2.f, &env->vec_status) ? 4 : 0; |
| 3349 |
} |
| 3350 |
|
| 3351 |
static always_inline uint32_t efststgt (uint32_t op1, uint32_t op2)
|
| 3352 |
{
|
| 3353 |
CPU_FloatU u1, u2; |
| 3354 |
u1.l = op1; |
| 3355 |
u2.l = op2; |
| 3356 |
return float32_le(u1.f, u2.f, &env->vec_status) ? 0 : 4; |
| 3357 |
} |
| 3358 |
|
| 3359 |
static always_inline uint32_t efststeq (uint32_t op1, uint32_t op2)
|
| 3360 |
{
|
| 3361 |
CPU_FloatU u1, u2; |
| 3362 |
u1.l = op1; |
| 3363 |
u2.l = op2; |
| 3364 |
return float32_eq(u1.f, u2.f, &env->vec_status) ? 4 : 0; |
| 3365 |
} |
| 3366 |
|
| 3367 |
static always_inline uint32_t efscmplt (uint32_t op1, uint32_t op2)
|
| 3368 |
{
|
| 3369 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
| 3370 |
return efststlt(op1, op2);
|
| 3371 |
} |
| 3372 |
|
| 3373 |
static always_inline uint32_t efscmpgt (uint32_t op1, uint32_t op2)
|
| 3374 |
{
|
| 3375 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
| 3376 |
return efststgt(op1, op2);
|
| 3377 |
} |
| 3378 |
|
| 3379 |
static always_inline uint32_t efscmpeq (uint32_t op1, uint32_t op2)
|
| 3380 |
{
|
| 3381 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
| 3382 |
return efststeq(op1, op2);
|
| 3383 |
} |
| 3384 |
|
| 3385 |
#define HELPER_SINGLE_SPE_CMP(name) \
|
| 3386 |
uint32_t helper_e##name (uint32_t op1, uint32_t op2) \ |
| 3387 |
{ \
|
| 3388 |
return e##name(op1, op2) << 2; \ |
| 3389 |
} |
| 3390 |
/* efststlt */
|
| 3391 |
HELPER_SINGLE_SPE_CMP(fststlt); |
| 3392 |
/* efststgt */
|
| 3393 |
HELPER_SINGLE_SPE_CMP(fststgt); |
| 3394 |
/* efststeq */
|
| 3395 |
HELPER_SINGLE_SPE_CMP(fststeq); |
| 3396 |
/* efscmplt */
|
| 3397 |
HELPER_SINGLE_SPE_CMP(fscmplt); |
| 3398 |
/* efscmpgt */
|
| 3399 |
HELPER_SINGLE_SPE_CMP(fscmpgt); |
| 3400 |
/* efscmpeq */
|
| 3401 |
HELPER_SINGLE_SPE_CMP(fscmpeq); |
| 3402 |
|
| 3403 |
static always_inline uint32_t evcmp_merge (int t0, int t1) |
| 3404 |
{
|
| 3405 |
return (t0 << 3) | (t1 << 2) | ((t0 | t1) << 1) | (t0 & t1); |
| 3406 |
} |
| 3407 |
|
| 3408 |
#define HELPER_VECTOR_SPE_CMP(name) \
|
| 3409 |
uint32_t helper_ev##name (uint64_t op1, uint64_t op2) \ |
| 3410 |
{ \
|
| 3411 |
return evcmp_merge(e##name(op1 >> 32, op2 >> 32), e##name(op1, op2)); \ |
| 3412 |
} |
| 3413 |
/* evfststlt */
|
| 3414 |
HELPER_VECTOR_SPE_CMP(fststlt); |
| 3415 |
/* evfststgt */
|
| 3416 |
HELPER_VECTOR_SPE_CMP(fststgt); |
| 3417 |
/* evfststeq */
|
| 3418 |
HELPER_VECTOR_SPE_CMP(fststeq); |
| 3419 |
/* evfscmplt */
|
| 3420 |
HELPER_VECTOR_SPE_CMP(fscmplt); |
| 3421 |
/* evfscmpgt */
|
| 3422 |
HELPER_VECTOR_SPE_CMP(fscmpgt); |
| 3423 |
/* evfscmpeq */
|
| 3424 |
HELPER_VECTOR_SPE_CMP(fscmpeq); |
| 3425 |
|
| 3426 |
/* Double-precision floating-point conversion */
|
| 3427 |
uint64_t helper_efdcfsi (uint32_t val) |
| 3428 |
{
|
| 3429 |
CPU_DoubleU u; |
| 3430 |
|
| 3431 |
u.d = int32_to_float64(val, &env->vec_status); |
| 3432 |
|
| 3433 |
return u.ll;
|
| 3434 |
} |
| 3435 |
|
| 3436 |
uint64_t helper_efdcfsid (uint64_t val) |
| 3437 |
{
|
| 3438 |
CPU_DoubleU u; |
| 3439 |
|
| 3440 |
u.d = int64_to_float64(val, &env->vec_status); |
| 3441 |
|
| 3442 |
return u.ll;
|
| 3443 |
} |
| 3444 |
|
| 3445 |
uint64_t helper_efdcfui (uint32_t val) |
| 3446 |
{
|
| 3447 |
CPU_DoubleU u; |
| 3448 |
|
| 3449 |
u.d = uint32_to_float64(val, &env->vec_status); |
| 3450 |
|
| 3451 |
return u.ll;
|
| 3452 |
} |
| 3453 |
|
| 3454 |
uint64_t helper_efdcfuid (uint64_t val) |
| 3455 |
{
|
| 3456 |
CPU_DoubleU u; |
| 3457 |
|
| 3458 |
u.d = uint64_to_float64(val, &env->vec_status); |
| 3459 |
|
| 3460 |
return u.ll;
|
| 3461 |
} |
| 3462 |
|
| 3463 |
uint32_t helper_efdctsi (uint64_t val) |
| 3464 |
{
|
| 3465 |
CPU_DoubleU u; |
| 3466 |
|
| 3467 |
u.ll = val; |
| 3468 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3469 |
if (unlikely(float64_is_nan(u.d)))
|
| 3470 |
return 0; |
| 3471 |
|
| 3472 |
return float64_to_int32(u.d, &env->vec_status);
|
| 3473 |
} |
| 3474 |
|
| 3475 |
uint32_t helper_efdctui (uint64_t val) |
| 3476 |
{
|
| 3477 |
CPU_DoubleU u; |
| 3478 |
|
| 3479 |
u.ll = val; |
| 3480 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3481 |
if (unlikely(float64_is_nan(u.d)))
|
| 3482 |
return 0; |
| 3483 |
|
| 3484 |
return float64_to_uint32(u.d, &env->vec_status);
|
| 3485 |
} |
| 3486 |
|
| 3487 |
uint32_t helper_efdctsiz (uint64_t val) |
| 3488 |
{
|
| 3489 |
CPU_DoubleU u; |
| 3490 |
|
| 3491 |
u.ll = val; |
| 3492 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3493 |
if (unlikely(float64_is_nan(u.d)))
|
| 3494 |
return 0; |
| 3495 |
|
| 3496 |
return float64_to_int32_round_to_zero(u.d, &env->vec_status);
|
| 3497 |
} |
| 3498 |
|
| 3499 |
uint64_t helper_efdctsidz (uint64_t val) |
| 3500 |
{
|
| 3501 |
CPU_DoubleU u; |
| 3502 |
|
| 3503 |
u.ll = val; |
| 3504 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3505 |
if (unlikely(float64_is_nan(u.d)))
|
| 3506 |
return 0; |
| 3507 |
|
| 3508 |
return float64_to_int64_round_to_zero(u.d, &env->vec_status);
|
| 3509 |
} |
| 3510 |
|
| 3511 |
uint32_t helper_efdctuiz (uint64_t val) |
| 3512 |
{
|
| 3513 |
CPU_DoubleU u; |
| 3514 |
|
| 3515 |
u.ll = val; |
| 3516 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3517 |
if (unlikely(float64_is_nan(u.d)))
|
| 3518 |
return 0; |
| 3519 |
|
| 3520 |
return float64_to_uint32_round_to_zero(u.d, &env->vec_status);
|
| 3521 |
} |
| 3522 |
|
| 3523 |
uint64_t helper_efdctuidz (uint64_t val) |
| 3524 |
{
|
| 3525 |
CPU_DoubleU u; |
| 3526 |
|
| 3527 |
u.ll = val; |
| 3528 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3529 |
if (unlikely(float64_is_nan(u.d)))
|
| 3530 |
return 0; |
| 3531 |
|
| 3532 |
return float64_to_uint64_round_to_zero(u.d, &env->vec_status);
|
| 3533 |
} |
| 3534 |
|
| 3535 |
uint64_t helper_efdcfsf (uint32_t val) |
| 3536 |
{
|
| 3537 |
CPU_DoubleU u; |
| 3538 |
float64 tmp; |
| 3539 |
|
| 3540 |
u.d = int32_to_float64(val, &env->vec_status); |
| 3541 |
tmp = int64_to_float64(1ULL << 32, &env->vec_status); |
| 3542 |
u.d = float64_div(u.d, tmp, &env->vec_status); |
| 3543 |
|
| 3544 |
return u.ll;
|
| 3545 |
} |
| 3546 |
|
| 3547 |
uint64_t helper_efdcfuf (uint32_t val) |
| 3548 |
{
|
| 3549 |
CPU_DoubleU u; |
| 3550 |
float64 tmp; |
| 3551 |
|
| 3552 |
u.d = uint32_to_float64(val, &env->vec_status); |
| 3553 |
tmp = int64_to_float64(1ULL << 32, &env->vec_status); |
| 3554 |
u.d = float64_div(u.d, tmp, &env->vec_status); |
| 3555 |
|
| 3556 |
return u.ll;
|
| 3557 |
} |
| 3558 |
|
| 3559 |
uint32_t helper_efdctsf (uint64_t val) |
| 3560 |
{
|
| 3561 |
CPU_DoubleU u; |
| 3562 |
float64 tmp; |
| 3563 |
|
| 3564 |
u.ll = val; |
| 3565 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3566 |
if (unlikely(float64_is_nan(u.d)))
|
| 3567 |
return 0; |
| 3568 |
tmp = uint64_to_float64(1ULL << 32, &env->vec_status); |
| 3569 |
u.d = float64_mul(u.d, tmp, &env->vec_status); |
| 3570 |
|
| 3571 |
return float64_to_int32(u.d, &env->vec_status);
|
| 3572 |
} |
| 3573 |
|
| 3574 |
uint32_t helper_efdctuf (uint64_t val) |
| 3575 |
{
|
| 3576 |
CPU_DoubleU u; |
| 3577 |
float64 tmp; |
| 3578 |
|
| 3579 |
u.ll = val; |
| 3580 |
/* NaN are not treated the same way IEEE 754 does */
|
| 3581 |
if (unlikely(float64_is_nan(u.d)))
|
| 3582 |
return 0; |
| 3583 |
tmp = uint64_to_float64(1ULL << 32, &env->vec_status); |
| 3584 |
u.d = float64_mul(u.d, tmp, &env->vec_status); |
| 3585 |
|
| 3586 |
return float64_to_uint32(u.d, &env->vec_status);
|
| 3587 |
} |
| 3588 |
|
| 3589 |
uint32_t helper_efscfd (uint64_t val) |
| 3590 |
{
|
| 3591 |
CPU_DoubleU u1; |
| 3592 |
CPU_FloatU u2; |
| 3593 |
|
| 3594 |
u1.ll = val; |
| 3595 |
u2.f = float64_to_float32(u1.d, &env->vec_status); |
| 3596 |
|
| 3597 |
return u2.l;
|
| 3598 |
} |
| 3599 |
|
| 3600 |
uint64_t helper_efdcfs (uint32_t val) |
| 3601 |
{
|
| 3602 |
CPU_DoubleU u2; |
| 3603 |
CPU_FloatU u1; |
| 3604 |
|
| 3605 |
u1.l = val; |
| 3606 |
u2.d = float32_to_float64(u1.f, &env->vec_status); |
| 3607 |
|
| 3608 |
return u2.ll;
|
| 3609 |
} |
| 3610 |
|
| 3611 |
/* Double precision fixed-point arithmetic */
|
| 3612 |
uint64_t helper_efdadd (uint64_t op1, uint64_t op2) |
| 3613 |
{
|
| 3614 |
CPU_DoubleU u1, u2; |
| 3615 |
u1.ll = op1; |
| 3616 |
u2.ll = op2; |
| 3617 |
u1.d = float64_add(u1.d, u2.d, &env->vec_status); |
| 3618 |
return u1.ll;
|
| 3619 |
} |
| 3620 |
|
| 3621 |
uint64_t helper_efdsub (uint64_t op1, uint64_t op2) |
| 3622 |
{
|
| 3623 |
CPU_DoubleU u1, u2; |
| 3624 |
u1.ll = op1; |
| 3625 |
u2.ll = op2; |
| 3626 |
u1.d = float64_sub(u1.d, u2.d, &env->vec_status); |
| 3627 |
return u1.ll;
|
| 3628 |
} |
| 3629 |
|
| 3630 |
uint64_t helper_efdmul (uint64_t op1, uint64_t op2) |
| 3631 |
{
|
| 3632 |
CPU_DoubleU u1, u2; |
| 3633 |
u1.ll = op1; |
| 3634 |
u2.ll = op2; |
| 3635 |
u1.d = float64_mul(u1.d, u2.d, &env->vec_status); |
| 3636 |
return u1.ll;
|
| 3637 |
} |
| 3638 |
|
| 3639 |
uint64_t helper_efddiv (uint64_t op1, uint64_t op2) |
| 3640 |
{
|
| 3641 |
CPU_DoubleU u1, u2; |
| 3642 |
u1.ll = op1; |
| 3643 |
u2.ll = op2; |
| 3644 |
u1.d = float64_div(u1.d, u2.d, &env->vec_status); |
| 3645 |
return u1.ll;
|
| 3646 |
} |
| 3647 |
|
| 3648 |
/* Double precision floating point helpers */
|
| 3649 |
uint32_t helper_efdtstlt (uint64_t op1, uint64_t op2) |
| 3650 |
{
|
| 3651 |
CPU_DoubleU u1, u2; |
| 3652 |
u1.ll = op1; |
| 3653 |
u2.ll = op2; |
| 3654 |
return float64_lt(u1.d, u2.d, &env->vec_status) ? 4 : 0; |
| 3655 |
} |
| 3656 |
|
| 3657 |
uint32_t helper_efdtstgt (uint64_t op1, uint64_t op2) |
| 3658 |
{
|
| 3659 |
CPU_DoubleU u1, u2; |
| 3660 |
u1.ll = op1; |
| 3661 |
u2.ll = op2; |
| 3662 |
return float64_le(u1.d, u2.d, &env->vec_status) ? 0 : 4; |
| 3663 |
} |
| 3664 |
|
| 3665 |
uint32_t helper_efdtsteq (uint64_t op1, uint64_t op2) |
| 3666 |
{
|
| 3667 |
CPU_DoubleU u1, u2; |
| 3668 |
u1.ll = op1; |
| 3669 |
u2.ll = op2; |
| 3670 |
return float64_eq(u1.d, u2.d, &env->vec_status) ? 4 : 0; |
| 3671 |
} |
| 3672 |
|
| 3673 |
uint32_t helper_efdcmplt (uint64_t op1, uint64_t op2) |
| 3674 |
{
|
| 3675 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
| 3676 |
return helper_efdtstlt(op1, op2);
|
| 3677 |
} |
| 3678 |
|
| 3679 |
uint32_t helper_efdcmpgt (uint64_t op1, uint64_t op2) |
| 3680 |
{
|
| 3681 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
| 3682 |
return helper_efdtstgt(op1, op2);
|
| 3683 |
} |
| 3684 |
|
| 3685 |
uint32_t helper_efdcmpeq (uint64_t op1, uint64_t op2) |
| 3686 |
{
|
| 3687 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
| 3688 |
return helper_efdtsteq(op1, op2);
|
| 3689 |
} |
| 3690 |
|
| 3691 |
/*****************************************************************************/
|
| 3692 |
/* Softmmu support */
|
| 3693 |
#if !defined (CONFIG_USER_ONLY)
|
| 3694 |
|
| 3695 |
#define MMUSUFFIX _mmu
|
| 3696 |
|
| 3697 |
#define SHIFT 0 |
| 3698 |
#include "softmmu_template.h" |
| 3699 |
|
| 3700 |
#define SHIFT 1 |
| 3701 |
#include "softmmu_template.h" |
| 3702 |
|
| 3703 |
#define SHIFT 2 |
| 3704 |
#include "softmmu_template.h" |
| 3705 |
|
| 3706 |
#define SHIFT 3 |
| 3707 |
#include "softmmu_template.h" |
| 3708 |
|
| 3709 |
/* try to fill the TLB and return an exception if error. If retaddr is
|
| 3710 |
NULL, it means that the function was called in C code (i.e. not
|
| 3711 |
from generated code or from helper.c) */
|
| 3712 |
/* XXX: fix it to restore all registers */
|
| 3713 |
void tlb_fill (target_ulong addr, int is_write, int mmu_idx, void *retaddr) |
| 3714 |
{
|
| 3715 |
TranslationBlock *tb; |
| 3716 |
CPUState *saved_env; |
| 3717 |
unsigned long pc; |
| 3718 |
int ret;
|
| 3719 |
|
| 3720 |
/* XXX: hack to restore env in all cases, even if not called from
|
| 3721 |
generated code */
|
| 3722 |
saved_env = env; |
| 3723 |
env = cpu_single_env; |
| 3724 |
ret = cpu_ppc_handle_mmu_fault(env, addr, is_write, mmu_idx, 1);
|
| 3725 |
if (unlikely(ret != 0)) { |
| 3726 |
if (likely(retaddr)) {
|
| 3727 |
/* now we have a real cpu fault */
|
| 3728 |
pc = (unsigned long)retaddr; |
| 3729 |
tb = tb_find_pc(pc); |
| 3730 |
if (likely(tb)) {
|
| 3731 |
/* the PC is inside the translated code. It means that we have
|
| 3732 |
a virtual CPU fault */
|
| 3733 |
cpu_restore_state(tb, env, pc, NULL);
|
| 3734 |
} |
| 3735 |
} |
| 3736 |
helper_raise_exception_err(env->exception_index, env->error_code); |
| 3737 |
} |
| 3738 |
env = saved_env; |
| 3739 |
} |
| 3740 |
|
| 3741 |
/* Segment registers load and store */
|
| 3742 |
target_ulong helper_load_sr (target_ulong sr_num) |
| 3743 |
{
|
| 3744 |
#if defined(TARGET_PPC64)
|
| 3745 |
if (env->mmu_model & POWERPC_MMU_64)
|
| 3746 |
return ppc_load_sr(env, sr_num);
|
| 3747 |
#endif
|
| 3748 |
return env->sr[sr_num];
|
| 3749 |
} |
| 3750 |
|
| 3751 |
void helper_store_sr (target_ulong sr_num, target_ulong val)
|
| 3752 |
{
|
| 3753 |
ppc_store_sr(env, sr_num, val); |
| 3754 |
} |
| 3755 |
|
| 3756 |
/* SLB management */
|
| 3757 |
#if defined(TARGET_PPC64)
|
| 3758 |
target_ulong helper_load_slb (target_ulong slb_nr) |
| 3759 |
{
|
| 3760 |
return ppc_load_slb(env, slb_nr);
|
| 3761 |
} |
| 3762 |
|
| 3763 |
void helper_store_slb (target_ulong rb, target_ulong rs)
|
| 3764 |
{
|
| 3765 |
ppc_store_slb(env, rb, rs); |
| 3766 |
} |
| 3767 |
|
| 3768 |
void helper_slbia (void) |
| 3769 |
{
|
| 3770 |
ppc_slb_invalidate_all(env); |
| 3771 |
} |
| 3772 |
|
| 3773 |
void helper_slbie (target_ulong addr)
|
| 3774 |
{
|
| 3775 |
ppc_slb_invalidate_one(env, addr); |
| 3776 |
} |
| 3777 |
|
| 3778 |
#endif /* defined(TARGET_PPC64) */ |
| 3779 |
|
| 3780 |
/* TLB management */
|
| 3781 |
void helper_tlbia (void) |
| 3782 |
{
|
| 3783 |
ppc_tlb_invalidate_all(env); |
| 3784 |
} |
| 3785 |
|
| 3786 |
void helper_tlbie (target_ulong addr)
|
| 3787 |
{
|
| 3788 |
ppc_tlb_invalidate_one(env, addr); |
| 3789 |
} |
| 3790 |
|
| 3791 |
/* Software driven TLBs management */
|
| 3792 |
/* PowerPC 602/603 software TLB load instructions helpers */
|
| 3793 |
static void do_6xx_tlb (target_ulong new_EPN, int is_code) |
| 3794 |
{
|
| 3795 |
target_ulong RPN, CMP, EPN; |
| 3796 |
int way;
|
| 3797 |
|
| 3798 |
RPN = env->spr[SPR_RPA]; |
| 3799 |
if (is_code) {
|
| 3800 |
CMP = env->spr[SPR_ICMP]; |
| 3801 |
EPN = env->spr[SPR_IMISS]; |
| 3802 |
} else {
|
| 3803 |
CMP = env->spr[SPR_DCMP]; |
| 3804 |
EPN = env->spr[SPR_DMISS]; |
| 3805 |
} |
| 3806 |
way = (env->spr[SPR_SRR1] >> 17) & 1; |
| 3807 |
LOG_SWTLB("%s: EPN " ADDRX " " ADDRX " PTE0 " ADDRX |
| 3808 |
" PTE1 " ADDRX " way %d\n", |
| 3809 |
__func__, new_EPN, EPN, CMP, RPN, way); |
| 3810 |
/* Store this TLB */
|
| 3811 |
ppc6xx_tlb_store(env, (uint32_t)(new_EPN & TARGET_PAGE_MASK), |
| 3812 |
way, is_code, CMP, RPN); |
| 3813 |
} |
| 3814 |
|
| 3815 |
void helper_6xx_tlbd (target_ulong EPN)
|
| 3816 |
{
|
| 3817 |
do_6xx_tlb(EPN, 0);
|
| 3818 |
} |
| 3819 |
|
| 3820 |
void helper_6xx_tlbi (target_ulong EPN)
|
| 3821 |
{
|
| 3822 |
do_6xx_tlb(EPN, 1);
|
| 3823 |
} |
| 3824 |
|
| 3825 |
/* PowerPC 74xx software TLB load instructions helpers */
|
| 3826 |
static void do_74xx_tlb (target_ulong new_EPN, int is_code) |
| 3827 |
{
|
| 3828 |
target_ulong RPN, CMP, EPN; |
| 3829 |
int way;
|
| 3830 |
|
| 3831 |
RPN = env->spr[SPR_PTELO]; |
| 3832 |
CMP = env->spr[SPR_PTEHI]; |
| 3833 |
EPN = env->spr[SPR_TLBMISS] & ~0x3;
|
| 3834 |
way = env->spr[SPR_TLBMISS] & 0x3;
|
| 3835 |
LOG_SWTLB("%s: EPN " ADDRX " " ADDRX " PTE0 " ADDRX |
| 3836 |
" PTE1 " ADDRX " way %d\n", |
| 3837 |
__func__, new_EPN, EPN, CMP, RPN, way); |
| 3838 |
/* Store this TLB */
|
| 3839 |
ppc6xx_tlb_store(env, (uint32_t)(new_EPN & TARGET_PAGE_MASK), |
| 3840 |
way, is_code, CMP, RPN); |
| 3841 |
} |
| 3842 |
|
| 3843 |
void helper_74xx_tlbd (target_ulong EPN)
|
| 3844 |
{
|
| 3845 |
do_74xx_tlb(EPN, 0);
|
| 3846 |
} |
| 3847 |
|
| 3848 |
void helper_74xx_tlbi (target_ulong EPN)
|
| 3849 |
{
|
| 3850 |
do_74xx_tlb(EPN, 1);
|
| 3851 |
} |
| 3852 |
|
| 3853 |
static always_inline target_ulong booke_tlb_to_page_size (int size) |
| 3854 |
{
|
| 3855 |
return 1024 << (2 * size); |
| 3856 |
} |
| 3857 |
|
| 3858 |
static always_inline int booke_page_size_to_tlb (target_ulong page_size) |
| 3859 |
{
|
| 3860 |
int size;
|
| 3861 |
|
| 3862 |
switch (page_size) {
|
| 3863 |
case 0x00000400UL: |
| 3864 |
size = 0x0;
|
| 3865 |
break;
|
| 3866 |
case 0x00001000UL: |
| 3867 |
size = 0x1;
|
| 3868 |
break;
|
| 3869 |
case 0x00004000UL: |
| 3870 |
size = 0x2;
|
| 3871 |
break;
|
| 3872 |
case 0x00010000UL: |
| 3873 |
size = 0x3;
|
| 3874 |
break;
|
| 3875 |
case 0x00040000UL: |
| 3876 |
size = 0x4;
|
| 3877 |
break;
|
| 3878 |
case 0x00100000UL: |
| 3879 |
size = 0x5;
|
| 3880 |
break;
|
| 3881 |
case 0x00400000UL: |
| 3882 |
size = 0x6;
|
| 3883 |
break;
|
| 3884 |
case 0x01000000UL: |
| 3885 |
size = 0x7;
|
| 3886 |
break;
|
| 3887 |
case 0x04000000UL: |
| 3888 |
size = 0x8;
|
| 3889 |
break;
|
| 3890 |
case 0x10000000UL: |
| 3891 |
size = 0x9;
|
| 3892 |
break;
|
| 3893 |
case 0x40000000UL: |
| 3894 |
size = 0xA;
|
| 3895 |
break;
|
| 3896 |
#if defined (TARGET_PPC64)
|
| 3897 |
case 0x000100000000ULL: |
| 3898 |
size = 0xB;
|
| 3899 |
break;
|
| 3900 |
case 0x000400000000ULL: |
| 3901 |
size = 0xC;
|
| 3902 |
break;
|
| 3903 |
case 0x001000000000ULL: |
| 3904 |
size = 0xD;
|
| 3905 |
break;
|
| 3906 |
case 0x004000000000ULL: |
| 3907 |
size = 0xE;
|
| 3908 |
break;
|
| 3909 |
case 0x010000000000ULL: |
| 3910 |
size = 0xF;
|
| 3911 |
break;
|
| 3912 |
#endif
|
| 3913 |
default:
|
| 3914 |
size = -1;
|
| 3915 |
break;
|
| 3916 |
} |
| 3917 |
|
| 3918 |
return size;
|
| 3919 |
} |
| 3920 |
|
| 3921 |
/* Helpers for 4xx TLB management */
|
| 3922 |
target_ulong helper_4xx_tlbre_lo (target_ulong entry) |
| 3923 |
{
|
| 3924 |
ppcemb_tlb_t *tlb; |
| 3925 |
target_ulong ret; |
| 3926 |
int size;
|
| 3927 |
|
| 3928 |
entry &= 0x3F;
|
| 3929 |
tlb = &env->tlb[entry].tlbe; |
| 3930 |
ret = tlb->EPN; |
| 3931 |
if (tlb->prot & PAGE_VALID)
|
| 3932 |
ret |= 0x400;
|
| 3933 |
size = booke_page_size_to_tlb(tlb->size); |
| 3934 |
if (size < 0 || size > 0x7) |
| 3935 |
size = 1;
|
| 3936 |
ret |= size << 7;
|
| 3937 |
env->spr[SPR_40x_PID] = tlb->PID; |
| 3938 |
return ret;
|
| 3939 |
} |
| 3940 |
|
| 3941 |
target_ulong helper_4xx_tlbre_hi (target_ulong entry) |
| 3942 |
{
|
| 3943 |
ppcemb_tlb_t *tlb; |
| 3944 |
target_ulong ret; |
| 3945 |
|
| 3946 |
entry &= 0x3F;
|
| 3947 |
tlb = &env->tlb[entry].tlbe; |
| 3948 |
ret = tlb->RPN; |
| 3949 |
if (tlb->prot & PAGE_EXEC)
|
| 3950 |
ret |= 0x200;
|
| 3951 |
if (tlb->prot & PAGE_WRITE)
|
| 3952 |
ret |= 0x100;
|
| 3953 |
return ret;
|
| 3954 |
} |
| 3955 |
|
| 3956 |
void helper_4xx_tlbwe_hi (target_ulong entry, target_ulong val)
|
| 3957 |
{
|
| 3958 |
ppcemb_tlb_t *tlb; |
| 3959 |
target_ulong page, end; |
| 3960 |
|
| 3961 |
LOG_SWTLB("%s entry %d val " ADDRX "\n", __func__, (int)entry, val); |
| 3962 |
entry &= 0x3F;
|
| 3963 |
tlb = &env->tlb[entry].tlbe; |
| 3964 |
/* Invalidate previous TLB (if it's valid) */
|
| 3965 |
if (tlb->prot & PAGE_VALID) {
|
| 3966 |
end = tlb->EPN + tlb->size; |
| 3967 |
LOG_SWTLB("%s: invalidate old TLB %d start " ADDRX
|
| 3968 |
" end " ADDRX "\n", __func__, (int)entry, tlb->EPN, end); |
| 3969 |
for (page = tlb->EPN; page < end; page += TARGET_PAGE_SIZE)
|
| 3970 |
tlb_flush_page(env, page); |
| 3971 |
} |
| 3972 |
tlb->size = booke_tlb_to_page_size((val >> 7) & 0x7); |
| 3973 |
/* We cannot handle TLB size < TARGET_PAGE_SIZE.
|
| 3974 |
* If this ever occurs, one should use the ppcemb target instead
|
| 3975 |
* of the ppc or ppc64 one
|
| 3976 |
*/
|
| 3977 |
if ((val & 0x40) && tlb->size < TARGET_PAGE_SIZE) { |
| 3978 |
cpu_abort(env, "TLB size " TARGET_FMT_lu " < %u " |
| 3979 |
"are not supported (%d)\n",
|
| 3980 |
tlb->size, TARGET_PAGE_SIZE, (int)((val >> 7) & 0x7)); |
| 3981 |
} |
| 3982 |
tlb->EPN = val & ~(tlb->size - 1);
|
| 3983 |
if (val & 0x40) |
| 3984 |
tlb->prot |= PAGE_VALID; |
| 3985 |
else
|
| 3986 |
tlb->prot &= ~PAGE_VALID; |
| 3987 |
if (val & 0x20) { |
| 3988 |
/* XXX: TO BE FIXED */
|
| 3989 |
cpu_abort(env, "Little-endian TLB entries are not supported by now\n");
|
| 3990 |
} |
| 3991 |
tlb->PID = env->spr[SPR_40x_PID]; /* PID */
|
| 3992 |
tlb->attr = val & 0xFF;
|
| 3993 |
LOG_SWTLB("%s: set up TLB %d RPN " PADDRX " EPN " ADDRX |
| 3994 |
" size " ADDRX " prot %c%c%c%c PID %d\n", __func__, |
| 3995 |
(int)entry, tlb->RPN, tlb->EPN, tlb->size,
|
| 3996 |
tlb->prot & PAGE_READ ? 'r' : '-', |
| 3997 |
tlb->prot & PAGE_WRITE ? 'w' : '-', |
| 3998 |
tlb->prot & PAGE_EXEC ? 'x' : '-', |
| 3999 |
tlb->prot & PAGE_VALID ? 'v' : '-', (int)tlb->PID); |
| 4000 |
/* Invalidate new TLB (if valid) */
|
| 4001 |
if (tlb->prot & PAGE_VALID) {
|
| 4002 |
end = tlb->EPN + tlb->size; |
| 4003 |
LOG_SWTLB("%s: invalidate TLB %d start " ADDRX
|
| 4004 |
" end " ADDRX "\n", __func__, (int)entry, tlb->EPN, end); |
| 4005 |
for (page = tlb->EPN; page < end; page += TARGET_PAGE_SIZE)
|
| 4006 |
tlb_flush_page(env, page); |
| 4007 |
} |
| 4008 |
} |
| 4009 |
|
| 4010 |
void helper_4xx_tlbwe_lo (target_ulong entry, target_ulong val)
|
| 4011 |
{
|
| 4012 |
ppcemb_tlb_t *tlb; |
| 4013 |
|
| 4014 |
LOG_SWTLB("%s entry %i val " ADDRX "\n", __func__, (int)entry, val); |
| 4015 |
entry &= 0x3F;
|
| 4016 |
tlb = &env->tlb[entry].tlbe; |
| 4017 |
tlb->RPN = val & 0xFFFFFC00;
|
| 4018 |
tlb->prot = PAGE_READ; |
| 4019 |
if (val & 0x200) |
| 4020 |
tlb->prot |= PAGE_EXEC; |
| 4021 |
if (val & 0x100) |
| 4022 |
tlb->prot |= PAGE_WRITE; |
| 4023 |
LOG_SWTLB("%s: set up TLB %d RPN " PADDRX " EPN " ADDRX |
| 4024 |
" size " ADDRX " prot %c%c%c%c PID %d\n", __func__, |
| 4025 |
(int)entry, tlb->RPN, tlb->EPN, tlb->size,
|
| 4026 |
tlb->prot & PAGE_READ ? 'r' : '-', |
| 4027 |
tlb->prot & PAGE_WRITE ? 'w' : '-', |
| 4028 |
tlb->prot & PAGE_EXEC ? 'x' : '-', |
| 4029 |
tlb->prot & PAGE_VALID ? 'v' : '-', (int)tlb->PID); |
| 4030 |
} |
| 4031 |
|
| 4032 |
target_ulong helper_4xx_tlbsx (target_ulong address) |
| 4033 |
{
|
| 4034 |
return ppcemb_tlb_search(env, address, env->spr[SPR_40x_PID]);
|
| 4035 |
} |
| 4036 |
|
| 4037 |
/* PowerPC 440 TLB management */
|
| 4038 |
void helper_440_tlbwe (uint32_t word, target_ulong entry, target_ulong value)
|
| 4039 |
{
|
| 4040 |
ppcemb_tlb_t *tlb; |
| 4041 |
target_ulong EPN, RPN, size; |
| 4042 |
int do_flush_tlbs;
|
| 4043 |
|
| 4044 |
LOG_SWTLB("%s word %d entry %d value " ADDRX "\n", |
| 4045 |
__func__, word, (int)entry, value);
|
| 4046 |
do_flush_tlbs = 0;
|
| 4047 |
entry &= 0x3F;
|
| 4048 |
tlb = &env->tlb[entry].tlbe; |
| 4049 |
switch (word) {
|
| 4050 |
default:
|
| 4051 |
/* Just here to please gcc */
|
| 4052 |
case 0: |
| 4053 |
EPN = value & 0xFFFFFC00;
|
| 4054 |
if ((tlb->prot & PAGE_VALID) && EPN != tlb->EPN)
|
| 4055 |
do_flush_tlbs = 1;
|
| 4056 |
tlb->EPN = EPN; |
| 4057 |
size = booke_tlb_to_page_size((value >> 4) & 0xF); |
| 4058 |
if ((tlb->prot & PAGE_VALID) && tlb->size < size)
|
| 4059 |
do_flush_tlbs = 1;
|
| 4060 |
tlb->size = size; |
| 4061 |
tlb->attr &= ~0x1;
|
| 4062 |
tlb->attr |= (value >> 8) & 1; |
| 4063 |
if (value & 0x200) { |
| 4064 |
tlb->prot |= PAGE_VALID; |
| 4065 |
} else {
|
| 4066 |
if (tlb->prot & PAGE_VALID) {
|
| 4067 |
tlb->prot &= ~PAGE_VALID; |
| 4068 |
do_flush_tlbs = 1;
|
| 4069 |
} |
| 4070 |
} |
| 4071 |
tlb->PID = env->spr[SPR_440_MMUCR] & 0x000000FF;
|
| 4072 |
if (do_flush_tlbs)
|
| 4073 |
tlb_flush(env, 1);
|
| 4074 |
break;
|
| 4075 |
case 1: |
| 4076 |
RPN = value & 0xFFFFFC0F;
|
| 4077 |
if ((tlb->prot & PAGE_VALID) && tlb->RPN != RPN)
|
| 4078 |
tlb_flush(env, 1);
|
| 4079 |
tlb->RPN = RPN; |
| 4080 |
break;
|
| 4081 |
case 2: |
| 4082 |
tlb->attr = (tlb->attr & 0x1) | (value & 0x0000FF00); |
| 4083 |
tlb->prot = tlb->prot & PAGE_VALID; |
| 4084 |
if (value & 0x1) |
| 4085 |
tlb->prot |= PAGE_READ << 4;
|
| 4086 |
if (value & 0x2) |
| 4087 |
tlb->prot |= PAGE_WRITE << 4;
|
| 4088 |
if (value & 0x4) |
| 4089 |
tlb->prot |= PAGE_EXEC << 4;
|
| 4090 |
if (value & 0x8) |
| 4091 |
tlb->prot |= PAGE_READ; |
| 4092 |
if (value & 0x10) |
| 4093 |
tlb->prot |= PAGE_WRITE; |
| 4094 |
if (value & 0x20) |
| 4095 |
tlb->prot |= PAGE_EXEC; |
| 4096 |
break;
|
| 4097 |
} |
| 4098 |
} |
| 4099 |
|
| 4100 |
target_ulong helper_440_tlbre (uint32_t word, target_ulong entry) |
| 4101 |
{
|
| 4102 |
ppcemb_tlb_t *tlb; |
| 4103 |
target_ulong ret; |
| 4104 |
int size;
|
| 4105 |
|
| 4106 |
entry &= 0x3F;
|
| 4107 |
tlb = &env->tlb[entry].tlbe; |
| 4108 |
switch (word) {
|
| 4109 |
default:
|
| 4110 |
/* Just here to please gcc */
|
| 4111 |
case 0: |
| 4112 |
ret = tlb->EPN; |
| 4113 |
size = booke_page_size_to_tlb(tlb->size); |
| 4114 |
if (size < 0 || size > 0xF) |
| 4115 |
size = 1;
|
| 4116 |
ret |= size << 4;
|
| 4117 |
if (tlb->attr & 0x1) |
| 4118 |
ret |= 0x100;
|
| 4119 |
if (tlb->prot & PAGE_VALID)
|
| 4120 |
ret |= 0x200;
|
| 4121 |
env->spr[SPR_440_MMUCR] &= ~0x000000FF;
|
| 4122 |
env->spr[SPR_440_MMUCR] |= tlb->PID; |
| 4123 |
break;
|
| 4124 |
case 1: |
| 4125 |
ret = tlb->RPN; |
| 4126 |
break;
|
| 4127 |
case 2: |
| 4128 |
ret = tlb->attr & ~0x1;
|
| 4129 |
if (tlb->prot & (PAGE_READ << 4)) |
| 4130 |
ret |= 0x1;
|
| 4131 |
if (tlb->prot & (PAGE_WRITE << 4)) |
| 4132 |
ret |= 0x2;
|
| 4133 |
if (tlb->prot & (PAGE_EXEC << 4)) |
| 4134 |
ret |= 0x4;
|
| 4135 |
if (tlb->prot & PAGE_READ)
|
| 4136 |
ret |= 0x8;
|
| 4137 |
if (tlb->prot & PAGE_WRITE)
|
| 4138 |
ret |= 0x10;
|
| 4139 |
if (tlb->prot & PAGE_EXEC)
|
| 4140 |
ret |= 0x20;
|
| 4141 |
break;
|
| 4142 |
} |
| 4143 |
return ret;
|
| 4144 |
} |
| 4145 |
|
| 4146 |
target_ulong helper_440_tlbsx (target_ulong address) |
| 4147 |
{
|
| 4148 |
return ppcemb_tlb_search(env, address, env->spr[SPR_440_MMUCR] & 0xFF); |
| 4149 |
} |
| 4150 |
|
| 4151 |
#endif /* !CONFIG_USER_ONLY */ |