root / target-ppc / op_helper.c @ 6ce0ca12
History | View | Annotate | Download (124.4 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, write to the Free Software
|
18 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301 USA
|
19 |
*/
|
20 |
#include <string.h> |
21 |
#include "exec.h" |
22 |
#include "host-utils.h" |
23 |
#include "helper.h" |
24 |
|
25 |
#include "helper_regs.h" |
26 |
|
27 |
//#define DEBUG_OP
|
28 |
//#define DEBUG_EXCEPTIONS
|
29 |
//#define DEBUG_SOFTWARE_TLB
|
30 |
|
31 |
#ifdef DEBUG_SOFTWARE_TLB
|
32 |
# define LOG_SWTLB(...) qemu_log(__VA_ARGS__)
|
33 |
#else
|
34 |
# define LOG_SWTLB(...) do { } while (0) |
35 |
#endif
|
36 |
|
37 |
|
38 |
/*****************************************************************************/
|
39 |
/* Exceptions processing helpers */
|
40 |
|
41 |
void helper_raise_exception_err (uint32_t exception, uint32_t error_code)
|
42 |
{ |
43 |
#if 0
|
44 |
printf("Raise exception %3x code : %d\n", exception, error_code);
|
45 |
#endif
|
46 |
env->exception_index = exception; |
47 |
env->error_code = error_code; |
48 |
cpu_loop_exit(); |
49 |
} |
50 |
|
51 |
void helper_raise_exception (uint32_t exception)
|
52 |
{ |
53 |
helper_raise_exception_err(exception, 0);
|
54 |
} |
55 |
|
56 |
/*****************************************************************************/
|
57 |
/* Registers load and stores */
|
58 |
target_ulong helper_load_cr (void)
|
59 |
{ |
60 |
return (env->crf[0] << 28) | |
61 |
(env->crf[1] << 24) | |
62 |
(env->crf[2] << 20) | |
63 |
(env->crf[3] << 16) | |
64 |
(env->crf[4] << 12) | |
65 |
(env->crf[5] << 8) | |
66 |
(env->crf[6] << 4) | |
67 |
(env->crf[7] << 0); |
68 |
} |
69 |
|
70 |
void helper_store_cr (target_ulong val, uint32_t mask)
|
71 |
{ |
72 |
int i, sh;
|
73 |
|
74 |
for (i = 0, sh = 7; i < 8; i++, sh--) { |
75 |
if (mask & (1 << sh)) |
76 |
env->crf[i] = (val >> (sh * 4)) & 0xFUL; |
77 |
} |
78 |
} |
79 |
|
80 |
/*****************************************************************************/
|
81 |
/* SPR accesses */
|
82 |
void helper_load_dump_spr (uint32_t sprn)
|
83 |
{ |
84 |
qemu_log("Read SPR %d %03x => " ADDRX "\n", |
85 |
sprn, sprn, env->spr[sprn]); |
86 |
} |
87 |
|
88 |
void helper_store_dump_spr (uint32_t sprn)
|
89 |
{ |
90 |
qemu_log("Write SPR %d %03x <= " ADDRX "\n", |
91 |
sprn, sprn, env->spr[sprn]); |
92 |
} |
93 |
|
94 |
target_ulong helper_load_tbl (void)
|
95 |
{ |
96 |
return cpu_ppc_load_tbl(env);
|
97 |
} |
98 |
|
99 |
target_ulong helper_load_tbu (void)
|
100 |
{ |
101 |
return cpu_ppc_load_tbu(env);
|
102 |
} |
103 |
|
104 |
target_ulong helper_load_atbl (void)
|
105 |
{ |
106 |
return cpu_ppc_load_atbl(env);
|
107 |
} |
108 |
|
109 |
target_ulong helper_load_atbu (void)
|
110 |
{ |
111 |
return cpu_ppc_load_atbu(env);
|
112 |
} |
113 |
|
114 |
target_ulong helper_load_601_rtcl (void)
|
115 |
{ |
116 |
return cpu_ppc601_load_rtcl(env);
|
117 |
} |
118 |
|
119 |
target_ulong helper_load_601_rtcu (void)
|
120 |
{ |
121 |
return cpu_ppc601_load_rtcu(env);
|
122 |
} |
123 |
|
124 |
#if !defined(CONFIG_USER_ONLY)
|
125 |
#if defined (TARGET_PPC64)
|
126 |
void helper_store_asr (target_ulong val)
|
127 |
{ |
128 |
ppc_store_asr(env, val); |
129 |
} |
130 |
#endif
|
131 |
|
132 |
void helper_store_sdr1 (target_ulong val)
|
133 |
{ |
134 |
ppc_store_sdr1(env, val); |
135 |
} |
136 |
|
137 |
void helper_store_tbl (target_ulong val)
|
138 |
{ |
139 |
cpu_ppc_store_tbl(env, val); |
140 |
} |
141 |
|
142 |
void helper_store_tbu (target_ulong val)
|
143 |
{ |
144 |
cpu_ppc_store_tbu(env, val); |
145 |
} |
146 |
|
147 |
void helper_store_atbl (target_ulong val)
|
148 |
{ |
149 |
cpu_ppc_store_atbl(env, val); |
150 |
} |
151 |
|
152 |
void helper_store_atbu (target_ulong val)
|
153 |
{ |
154 |
cpu_ppc_store_atbu(env, val); |
155 |
} |
156 |
|
157 |
void helper_store_601_rtcl (target_ulong val)
|
158 |
{ |
159 |
cpu_ppc601_store_rtcl(env, val); |
160 |
} |
161 |
|
162 |
void helper_store_601_rtcu (target_ulong val)
|
163 |
{ |
164 |
cpu_ppc601_store_rtcu(env, val); |
165 |
} |
166 |
|
167 |
target_ulong helper_load_decr (void)
|
168 |
{ |
169 |
return cpu_ppc_load_decr(env);
|
170 |
} |
171 |
|
172 |
void helper_store_decr (target_ulong val)
|
173 |
{ |
174 |
cpu_ppc_store_decr(env, val); |
175 |
} |
176 |
|
177 |
void helper_store_hid0_601 (target_ulong val)
|
178 |
{ |
179 |
target_ulong hid0; |
180 |
|
181 |
hid0 = env->spr[SPR_HID0]; |
182 |
if ((val ^ hid0) & 0x00000008) { |
183 |
/* Change current endianness */
|
184 |
env->hflags &= ~(1 << MSR_LE);
|
185 |
env->hflags_nmsr &= ~(1 << MSR_LE);
|
186 |
env->hflags_nmsr |= (1 << MSR_LE) & (((val >> 3) & 1) << MSR_LE); |
187 |
env->hflags |= env->hflags_nmsr; |
188 |
qemu_log("%s: set endianness to %c => " ADDRX "\n", |
189 |
__func__, val & 0x8 ? 'l' : 'b', env->hflags); |
190 |
} |
191 |
env->spr[SPR_HID0] = (uint32_t)val; |
192 |
} |
193 |
|
194 |
void helper_store_403_pbr (uint32_t num, target_ulong value)
|
195 |
{ |
196 |
if (likely(env->pb[num] != value)) {
|
197 |
env->pb[num] = value; |
198 |
/* Should be optimized */
|
199 |
tlb_flush(env, 1);
|
200 |
} |
201 |
} |
202 |
|
203 |
target_ulong helper_load_40x_pit (void)
|
204 |
{ |
205 |
return load_40x_pit(env);
|
206 |
} |
207 |
|
208 |
void helper_store_40x_pit (target_ulong val)
|
209 |
{ |
210 |
store_40x_pit(env, val); |
211 |
} |
212 |
|
213 |
void helper_store_40x_dbcr0 (target_ulong val)
|
214 |
{ |
215 |
store_40x_dbcr0(env, val); |
216 |
} |
217 |
|
218 |
void helper_store_40x_sler (target_ulong val)
|
219 |
{ |
220 |
store_40x_sler(env, val); |
221 |
} |
222 |
|
223 |
void helper_store_booke_tcr (target_ulong val)
|
224 |
{ |
225 |
store_booke_tcr(env, val); |
226 |
} |
227 |
|
228 |
void helper_store_booke_tsr (target_ulong val)
|
229 |
{ |
230 |
store_booke_tsr(env, val); |
231 |
} |
232 |
|
233 |
void helper_store_ibatu (uint32_t nr, target_ulong val)
|
234 |
{ |
235 |
ppc_store_ibatu(env, nr, val); |
236 |
} |
237 |
|
238 |
void helper_store_ibatl (uint32_t nr, target_ulong val)
|
239 |
{ |
240 |
ppc_store_ibatl(env, nr, val); |
241 |
} |
242 |
|
243 |
void helper_store_dbatu (uint32_t nr, target_ulong val)
|
244 |
{ |
245 |
ppc_store_dbatu(env, nr, val); |
246 |
} |
247 |
|
248 |
void helper_store_dbatl (uint32_t nr, target_ulong val)
|
249 |
{ |
250 |
ppc_store_dbatl(env, nr, val); |
251 |
} |
252 |
|
253 |
void helper_store_601_batl (uint32_t nr, target_ulong val)
|
254 |
{ |
255 |
ppc_store_ibatl_601(env, nr, val); |
256 |
} |
257 |
|
258 |
void helper_store_601_batu (uint32_t nr, target_ulong val)
|
259 |
{ |
260 |
ppc_store_ibatu_601(env, nr, val); |
261 |
} |
262 |
#endif
|
263 |
|
264 |
/*****************************************************************************/
|
265 |
/* Memory load and stores */
|
266 |
|
267 |
static always_inline target_ulong addr_add(target_ulong addr, target_long arg)
|
268 |
{ |
269 |
#if defined(TARGET_PPC64)
|
270 |
if (!msr_sf)
|
271 |
return (uint32_t)(addr + arg);
|
272 |
else
|
273 |
#endif
|
274 |
return addr + arg;
|
275 |
} |
276 |
|
277 |
void helper_lmw (target_ulong addr, uint32_t reg)
|
278 |
{ |
279 |
for (; reg < 32; reg++) { |
280 |
if (msr_le)
|
281 |
env->gpr[reg] = bswap32(ldl(addr)); |
282 |
else
|
283 |
env->gpr[reg] = ldl(addr); |
284 |
addr = addr_add(addr, 4);
|
285 |
} |
286 |
} |
287 |
|
288 |
void helper_stmw (target_ulong addr, uint32_t reg)
|
289 |
{ |
290 |
for (; reg < 32; reg++) { |
291 |
if (msr_le)
|
292 |
stl(addr, bswap32((uint32_t)env->gpr[reg])); |
293 |
else
|
294 |
stl(addr, (uint32_t)env->gpr[reg]); |
295 |
addr = addr_add(addr, 4);
|
296 |
} |
297 |
} |
298 |
|
299 |
void helper_lsw(target_ulong addr, uint32_t nb, uint32_t reg)
|
300 |
{ |
301 |
int sh;
|
302 |
for (; nb > 3; nb -= 4) { |
303 |
env->gpr[reg] = ldl(addr); |
304 |
reg = (reg + 1) % 32; |
305 |
addr = addr_add(addr, 4);
|
306 |
} |
307 |
if (unlikely(nb > 0)) { |
308 |
env->gpr[reg] = 0;
|
309 |
for (sh = 24; nb > 0; nb--, sh -= 8) { |
310 |
env->gpr[reg] |= ldub(addr) << sh; |
311 |
addr = addr_add(addr, 1);
|
312 |
} |
313 |
} |
314 |
} |
315 |
/* PPC32 specification says we must generate an exception if
|
316 |
* rA is in the range of registers to be loaded.
|
317 |
* In an other hand, IBM says this is valid, but rA won't be loaded.
|
318 |
* For now, I'll follow the spec...
|
319 |
*/
|
320 |
void helper_lswx(target_ulong addr, uint32_t reg, uint32_t ra, uint32_t rb)
|
321 |
{ |
322 |
if (likely(xer_bc != 0)) { |
323 |
if (unlikely((ra != 0 && reg < ra && (reg + xer_bc) > ra) || |
324 |
(reg < rb && (reg + xer_bc) > rb))) { |
325 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
326 |
POWERPC_EXCP_INVAL | |
327 |
POWERPC_EXCP_INVAL_LSWX); |
328 |
} else {
|
329 |
helper_lsw(addr, xer_bc, reg); |
330 |
} |
331 |
} |
332 |
} |
333 |
|
334 |
void helper_stsw(target_ulong addr, uint32_t nb, uint32_t reg)
|
335 |
{ |
336 |
int sh;
|
337 |
for (; nb > 3; nb -= 4) { |
338 |
stl(addr, env->gpr[reg]); |
339 |
reg = (reg + 1) % 32; |
340 |
addr = addr_add(addr, 4);
|
341 |
} |
342 |
if (unlikely(nb > 0)) { |
343 |
for (sh = 24; nb > 0; nb--, sh -= 8) { |
344 |
stb(addr, (env->gpr[reg] >> sh) & 0xFF);
|
345 |
addr = addr_add(addr, 1);
|
346 |
} |
347 |
} |
348 |
} |
349 |
|
350 |
static void do_dcbz(target_ulong addr, int dcache_line_size) |
351 |
{ |
352 |
addr &= ~(dcache_line_size - 1);
|
353 |
int i;
|
354 |
for (i = 0 ; i < dcache_line_size ; i += 4) { |
355 |
stl(addr + i , 0);
|
356 |
} |
357 |
if (env->reserve == addr)
|
358 |
env->reserve = (target_ulong)-1ULL;
|
359 |
} |
360 |
|
361 |
void helper_dcbz(target_ulong addr)
|
362 |
{ |
363 |
do_dcbz(addr, env->dcache_line_size); |
364 |
} |
365 |
|
366 |
void helper_dcbz_970(target_ulong addr)
|
367 |
{ |
368 |
if (((env->spr[SPR_970_HID5] >> 7) & 0x3) == 1) |
369 |
do_dcbz(addr, 32);
|
370 |
else
|
371 |
do_dcbz(addr, env->dcache_line_size); |
372 |
} |
373 |
|
374 |
void helper_icbi(target_ulong addr)
|
375 |
{ |
376 |
uint32_t tmp; |
377 |
|
378 |
addr &= ~(env->dcache_line_size - 1);
|
379 |
/* Invalidate one cache line :
|
380 |
* PowerPC specification says this is to be treated like a load
|
381 |
* (not a fetch) by the MMU. To be sure it will be so,
|
382 |
* do the load "by hand".
|
383 |
*/
|
384 |
tmp = ldl(addr); |
385 |
tb_invalidate_page_range(addr, addr + env->icache_line_size); |
386 |
} |
387 |
|
388 |
// XXX: to be tested
|
389 |
target_ulong helper_lscbx (target_ulong addr, uint32_t reg, uint32_t ra, uint32_t rb) |
390 |
{ |
391 |
int i, c, d;
|
392 |
d = 24;
|
393 |
for (i = 0; i < xer_bc; i++) { |
394 |
c = ldub(addr); |
395 |
addr = addr_add(addr, 1);
|
396 |
/* ra (if not 0) and rb are never modified */
|
397 |
if (likely(reg != rb && (ra == 0 || reg != ra))) { |
398 |
env->gpr[reg] = (env->gpr[reg] & ~(0xFF << d)) | (c << d);
|
399 |
} |
400 |
if (unlikely(c == xer_cmp))
|
401 |
break;
|
402 |
if (likely(d != 0)) { |
403 |
d -= 8;
|
404 |
} else {
|
405 |
d = 24;
|
406 |
reg++; |
407 |
reg = reg & 0x1F;
|
408 |
} |
409 |
} |
410 |
return i;
|
411 |
} |
412 |
|
413 |
/*****************************************************************************/
|
414 |
/* Fixed point operations helpers */
|
415 |
#if defined(TARGET_PPC64)
|
416 |
|
417 |
/* multiply high word */
|
418 |
uint64_t helper_mulhd (uint64_t arg1, uint64_t arg2) |
419 |
{ |
420 |
uint64_t tl, th; |
421 |
|
422 |
muls64(&tl, &th, arg1, arg2); |
423 |
return th;
|
424 |
} |
425 |
|
426 |
/* multiply high word unsigned */
|
427 |
uint64_t helper_mulhdu (uint64_t arg1, uint64_t arg2) |
428 |
{ |
429 |
uint64_t tl, th; |
430 |
|
431 |
mulu64(&tl, &th, arg1, arg2); |
432 |
return th;
|
433 |
} |
434 |
|
435 |
uint64_t helper_mulldo (uint64_t arg1, uint64_t arg2) |
436 |
{ |
437 |
int64_t th; |
438 |
uint64_t tl; |
439 |
|
440 |
muls64(&tl, (uint64_t *)&th, arg1, arg2); |
441 |
/* If th != 0 && th != -1, then we had an overflow */
|
442 |
if (likely((uint64_t)(th + 1) <= 1)) { |
443 |
env->xer &= ~(1 << XER_OV);
|
444 |
} else {
|
445 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
446 |
} |
447 |
return (int64_t)tl;
|
448 |
} |
449 |
#endif
|
450 |
|
451 |
target_ulong helper_cntlzw (target_ulong t) |
452 |
{ |
453 |
return clz32(t);
|
454 |
} |
455 |
|
456 |
#if defined(TARGET_PPC64)
|
457 |
target_ulong helper_cntlzd (target_ulong t) |
458 |
{ |
459 |
return clz64(t);
|
460 |
} |
461 |
#endif
|
462 |
|
463 |
/* shift right arithmetic helper */
|
464 |
target_ulong helper_sraw (target_ulong value, target_ulong shift) |
465 |
{ |
466 |
int32_t ret; |
467 |
|
468 |
if (likely(!(shift & 0x20))) { |
469 |
if (likely((uint32_t)shift != 0)) { |
470 |
shift &= 0x1f;
|
471 |
ret = (int32_t)value >> shift; |
472 |
if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) { |
473 |
env->xer &= ~(1 << XER_CA);
|
474 |
} else {
|
475 |
env->xer |= (1 << XER_CA);
|
476 |
} |
477 |
} else {
|
478 |
ret = (int32_t)value; |
479 |
env->xer &= ~(1 << XER_CA);
|
480 |
} |
481 |
} else {
|
482 |
ret = (int32_t)value >> 31;
|
483 |
if (ret) {
|
484 |
env->xer |= (1 << XER_CA);
|
485 |
} else {
|
486 |
env->xer &= ~(1 << XER_CA);
|
487 |
} |
488 |
} |
489 |
return (target_long)ret;
|
490 |
} |
491 |
|
492 |
#if defined(TARGET_PPC64)
|
493 |
target_ulong helper_srad (target_ulong value, target_ulong shift) |
494 |
{ |
495 |
int64_t ret; |
496 |
|
497 |
if (likely(!(shift & 0x40))) { |
498 |
if (likely((uint64_t)shift != 0)) { |
499 |
shift &= 0x3f;
|
500 |
ret = (int64_t)value >> shift; |
501 |
if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) { |
502 |
env->xer &= ~(1 << XER_CA);
|
503 |
} else {
|
504 |
env->xer |= (1 << XER_CA);
|
505 |
} |
506 |
} else {
|
507 |
ret = (int64_t)value; |
508 |
env->xer &= ~(1 << XER_CA);
|
509 |
} |
510 |
} else {
|
511 |
ret = (int64_t)value >> 63;
|
512 |
if (ret) {
|
513 |
env->xer |= (1 << XER_CA);
|
514 |
} else {
|
515 |
env->xer &= ~(1 << XER_CA);
|
516 |
} |
517 |
} |
518 |
return ret;
|
519 |
} |
520 |
#endif
|
521 |
|
522 |
target_ulong helper_popcntb (target_ulong val) |
523 |
{ |
524 |
val = (val & 0x55555555) + ((val >> 1) & 0x55555555); |
525 |
val = (val & 0x33333333) + ((val >> 2) & 0x33333333); |
526 |
val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f); |
527 |
return val;
|
528 |
} |
529 |
|
530 |
#if defined(TARGET_PPC64)
|
531 |
target_ulong helper_popcntb_64 (target_ulong val) |
532 |
{ |
533 |
val = (val & 0x5555555555555555ULL) + ((val >> 1) & 0x5555555555555555ULL); |
534 |
val = (val & 0x3333333333333333ULL) + ((val >> 2) & 0x3333333333333333ULL); |
535 |
val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) & 0x0f0f0f0f0f0f0f0fULL); |
536 |
return val;
|
537 |
} |
538 |
#endif
|
539 |
|
540 |
/*****************************************************************************/
|
541 |
/* Floating point operations helpers */
|
542 |
uint64_t helper_float32_to_float64(uint32_t arg) |
543 |
{ |
544 |
CPU_FloatU f; |
545 |
CPU_DoubleU d; |
546 |
f.l = arg; |
547 |
d.d = float32_to_float64(f.f, &env->fp_status); |
548 |
return d.ll;
|
549 |
} |
550 |
|
551 |
uint32_t helper_float64_to_float32(uint64_t arg) |
552 |
{ |
553 |
CPU_FloatU f; |
554 |
CPU_DoubleU d; |
555 |
d.ll = arg; |
556 |
f.f = float64_to_float32(d.d, &env->fp_status); |
557 |
return f.l;
|
558 |
} |
559 |
|
560 |
static always_inline int isden (float64 d) |
561 |
{ |
562 |
CPU_DoubleU u; |
563 |
|
564 |
u.d = d; |
565 |
|
566 |
return ((u.ll >> 52) & 0x7FF) == 0; |
567 |
} |
568 |
|
569 |
uint32_t helper_compute_fprf (uint64_t arg, uint32_t set_fprf) |
570 |
{ |
571 |
CPU_DoubleU farg; |
572 |
int isneg;
|
573 |
int ret;
|
574 |
farg.ll = arg; |
575 |
isneg = float64_is_neg(farg.d); |
576 |
if (unlikely(float64_is_nan(farg.d))) {
|
577 |
if (float64_is_signaling_nan(farg.d)) {
|
578 |
/* Signaling NaN: flags are undefined */
|
579 |
ret = 0x00;
|
580 |
} else {
|
581 |
/* Quiet NaN */
|
582 |
ret = 0x11;
|
583 |
} |
584 |
} else if (unlikely(float64_is_infinity(farg.d))) { |
585 |
/* +/- infinity */
|
586 |
if (isneg)
|
587 |
ret = 0x09;
|
588 |
else
|
589 |
ret = 0x05;
|
590 |
} else {
|
591 |
if (float64_is_zero(farg.d)) {
|
592 |
/* +/- zero */
|
593 |
if (isneg)
|
594 |
ret = 0x12;
|
595 |
else
|
596 |
ret = 0x02;
|
597 |
} else {
|
598 |
if (isden(farg.d)) {
|
599 |
/* Denormalized numbers */
|
600 |
ret = 0x10;
|
601 |
} else {
|
602 |
/* Normalized numbers */
|
603 |
ret = 0x00;
|
604 |
} |
605 |
if (isneg) {
|
606 |
ret |= 0x08;
|
607 |
} else {
|
608 |
ret |= 0x04;
|
609 |
} |
610 |
} |
611 |
} |
612 |
if (set_fprf) {
|
613 |
/* We update FPSCR_FPRF */
|
614 |
env->fpscr &= ~(0x1F << FPSCR_FPRF);
|
615 |
env->fpscr |= ret << FPSCR_FPRF; |
616 |
} |
617 |
/* We just need fpcc to update Rc1 */
|
618 |
return ret & 0xF; |
619 |
} |
620 |
|
621 |
/* Floating-point invalid operations exception */
|
622 |
static always_inline uint64_t fload_invalid_op_excp (int op) |
623 |
{ |
624 |
uint64_t ret = 0;
|
625 |
int ve;
|
626 |
|
627 |
ve = fpscr_ve; |
628 |
switch (op) {
|
629 |
case POWERPC_EXCP_FP_VXSNAN:
|
630 |
env->fpscr |= 1 << FPSCR_VXSNAN;
|
631 |
break;
|
632 |
case POWERPC_EXCP_FP_VXSOFT:
|
633 |
env->fpscr |= 1 << FPSCR_VXSOFT;
|
634 |
break;
|
635 |
case POWERPC_EXCP_FP_VXISI:
|
636 |
/* Magnitude subtraction of infinities */
|
637 |
env->fpscr |= 1 << FPSCR_VXISI;
|
638 |
goto update_arith;
|
639 |
case POWERPC_EXCP_FP_VXIDI:
|
640 |
/* Division of infinity by infinity */
|
641 |
env->fpscr |= 1 << FPSCR_VXIDI;
|
642 |
goto update_arith;
|
643 |
case POWERPC_EXCP_FP_VXZDZ:
|
644 |
/* Division of zero by zero */
|
645 |
env->fpscr |= 1 << FPSCR_VXZDZ;
|
646 |
goto update_arith;
|
647 |
case POWERPC_EXCP_FP_VXIMZ:
|
648 |
/* Multiplication of zero by infinity */
|
649 |
env->fpscr |= 1 << FPSCR_VXIMZ;
|
650 |
goto update_arith;
|
651 |
case POWERPC_EXCP_FP_VXVC:
|
652 |
/* Ordered comparison of NaN */
|
653 |
env->fpscr |= 1 << FPSCR_VXVC;
|
654 |
env->fpscr &= ~(0xF << FPSCR_FPCC);
|
655 |
env->fpscr |= 0x11 << FPSCR_FPCC;
|
656 |
/* We must update the target FPR before raising the exception */
|
657 |
if (ve != 0) { |
658 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
659 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_VXVC; |
660 |
/* Update the floating-point enabled exception summary */
|
661 |
env->fpscr |= 1 << FPSCR_FEX;
|
662 |
/* Exception is differed */
|
663 |
ve = 0;
|
664 |
} |
665 |
break;
|
666 |
case POWERPC_EXCP_FP_VXSQRT:
|
667 |
/* Square root of a negative number */
|
668 |
env->fpscr |= 1 << FPSCR_VXSQRT;
|
669 |
update_arith:
|
670 |
env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI)); |
671 |
if (ve == 0) { |
672 |
/* Set the result to quiet NaN */
|
673 |
ret = 0xFFF8000000000000ULL;
|
674 |
env->fpscr &= ~(0xF << FPSCR_FPCC);
|
675 |
env->fpscr |= 0x11 << FPSCR_FPCC;
|
676 |
} |
677 |
break;
|
678 |
case POWERPC_EXCP_FP_VXCVI:
|
679 |
/* Invalid conversion */
|
680 |
env->fpscr |= 1 << FPSCR_VXCVI;
|
681 |
env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI)); |
682 |
if (ve == 0) { |
683 |
/* Set the result to quiet NaN */
|
684 |
ret = 0xFFF8000000000000ULL;
|
685 |
env->fpscr &= ~(0xF << FPSCR_FPCC);
|
686 |
env->fpscr |= 0x11 << FPSCR_FPCC;
|
687 |
} |
688 |
break;
|
689 |
} |
690 |
/* Update the floating-point invalid operation summary */
|
691 |
env->fpscr |= 1 << FPSCR_VX;
|
692 |
/* Update the floating-point exception summary */
|
693 |
env->fpscr |= 1 << FPSCR_FX;
|
694 |
if (ve != 0) { |
695 |
/* Update the floating-point enabled exception summary */
|
696 |
env->fpscr |= 1 << FPSCR_FEX;
|
697 |
if (msr_fe0 != 0 || msr_fe1 != 0) |
698 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_FP | op); |
699 |
} |
700 |
return ret;
|
701 |
} |
702 |
|
703 |
static always_inline void float_zero_divide_excp (void) |
704 |
{ |
705 |
env->fpscr |= 1 << FPSCR_ZX;
|
706 |
env->fpscr &= ~((1 << FPSCR_FR) | (1 << FPSCR_FI)); |
707 |
/* Update the floating-point exception summary */
|
708 |
env->fpscr |= 1 << FPSCR_FX;
|
709 |
if (fpscr_ze != 0) { |
710 |
/* Update the floating-point enabled exception summary */
|
711 |
env->fpscr |= 1 << FPSCR_FEX;
|
712 |
if (msr_fe0 != 0 || msr_fe1 != 0) { |
713 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
714 |
POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX); |
715 |
} |
716 |
} |
717 |
} |
718 |
|
719 |
static always_inline void float_overflow_excp (void) |
720 |
{ |
721 |
env->fpscr |= 1 << FPSCR_OX;
|
722 |
/* Update the floating-point exception summary */
|
723 |
env->fpscr |= 1 << FPSCR_FX;
|
724 |
if (fpscr_oe != 0) { |
725 |
/* XXX: should adjust the result */
|
726 |
/* Update the floating-point enabled exception summary */
|
727 |
env->fpscr |= 1 << FPSCR_FEX;
|
728 |
/* We must update the target FPR before raising the exception */
|
729 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
730 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX; |
731 |
} else {
|
732 |
env->fpscr |= 1 << FPSCR_XX;
|
733 |
env->fpscr |= 1 << FPSCR_FI;
|
734 |
} |
735 |
} |
736 |
|
737 |
static always_inline void float_underflow_excp (void) |
738 |
{ |
739 |
env->fpscr |= 1 << FPSCR_UX;
|
740 |
/* Update the floating-point exception summary */
|
741 |
env->fpscr |= 1 << FPSCR_FX;
|
742 |
if (fpscr_ue != 0) { |
743 |
/* XXX: should adjust the result */
|
744 |
/* Update the floating-point enabled exception summary */
|
745 |
env->fpscr |= 1 << FPSCR_FEX;
|
746 |
/* We must update the target FPR before raising the exception */
|
747 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
748 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX; |
749 |
} |
750 |
} |
751 |
|
752 |
static always_inline void float_inexact_excp (void) |
753 |
{ |
754 |
env->fpscr |= 1 << FPSCR_XX;
|
755 |
/* Update the floating-point exception summary */
|
756 |
env->fpscr |= 1 << FPSCR_FX;
|
757 |
if (fpscr_xe != 0) { |
758 |
/* Update the floating-point enabled exception summary */
|
759 |
env->fpscr |= 1 << FPSCR_FEX;
|
760 |
/* We must update the target FPR before raising the exception */
|
761 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
762 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX; |
763 |
} |
764 |
} |
765 |
|
766 |
static always_inline void fpscr_set_rounding_mode (void) |
767 |
{ |
768 |
int rnd_type;
|
769 |
|
770 |
/* Set rounding mode */
|
771 |
switch (fpscr_rn) {
|
772 |
case 0: |
773 |
/* Best approximation (round to nearest) */
|
774 |
rnd_type = float_round_nearest_even; |
775 |
break;
|
776 |
case 1: |
777 |
/* Smaller magnitude (round toward zero) */
|
778 |
rnd_type = float_round_to_zero; |
779 |
break;
|
780 |
case 2: |
781 |
/* Round toward +infinite */
|
782 |
rnd_type = float_round_up; |
783 |
break;
|
784 |
default:
|
785 |
case 3: |
786 |
/* Round toward -infinite */
|
787 |
rnd_type = float_round_down; |
788 |
break;
|
789 |
} |
790 |
set_float_rounding_mode(rnd_type, &env->fp_status); |
791 |
} |
792 |
|
793 |
void helper_fpscr_clrbit (uint32_t bit)
|
794 |
{ |
795 |
int prev;
|
796 |
|
797 |
prev = (env->fpscr >> bit) & 1;
|
798 |
env->fpscr &= ~(1 << bit);
|
799 |
if (prev == 1) { |
800 |
switch (bit) {
|
801 |
case FPSCR_RN1:
|
802 |
case FPSCR_RN:
|
803 |
fpscr_set_rounding_mode(); |
804 |
break;
|
805 |
default:
|
806 |
break;
|
807 |
} |
808 |
} |
809 |
} |
810 |
|
811 |
void helper_fpscr_setbit (uint32_t bit)
|
812 |
{ |
813 |
int prev;
|
814 |
|
815 |
prev = (env->fpscr >> bit) & 1;
|
816 |
env->fpscr |= 1 << bit;
|
817 |
if (prev == 0) { |
818 |
switch (bit) {
|
819 |
case FPSCR_VX:
|
820 |
env->fpscr |= 1 << FPSCR_FX;
|
821 |
if (fpscr_ve)
|
822 |
goto raise_ve;
|
823 |
case FPSCR_OX:
|
824 |
env->fpscr |= 1 << FPSCR_FX;
|
825 |
if (fpscr_oe)
|
826 |
goto raise_oe;
|
827 |
break;
|
828 |
case FPSCR_UX:
|
829 |
env->fpscr |= 1 << FPSCR_FX;
|
830 |
if (fpscr_ue)
|
831 |
goto raise_ue;
|
832 |
break;
|
833 |
case FPSCR_ZX:
|
834 |
env->fpscr |= 1 << FPSCR_FX;
|
835 |
if (fpscr_ze)
|
836 |
goto raise_ze;
|
837 |
break;
|
838 |
case FPSCR_XX:
|
839 |
env->fpscr |= 1 << FPSCR_FX;
|
840 |
if (fpscr_xe)
|
841 |
goto raise_xe;
|
842 |
break;
|
843 |
case FPSCR_VXSNAN:
|
844 |
case FPSCR_VXISI:
|
845 |
case FPSCR_VXIDI:
|
846 |
case FPSCR_VXZDZ:
|
847 |
case FPSCR_VXIMZ:
|
848 |
case FPSCR_VXVC:
|
849 |
case FPSCR_VXSOFT:
|
850 |
case FPSCR_VXSQRT:
|
851 |
case FPSCR_VXCVI:
|
852 |
env->fpscr |= 1 << FPSCR_VX;
|
853 |
env->fpscr |= 1 << FPSCR_FX;
|
854 |
if (fpscr_ve != 0) |
855 |
goto raise_ve;
|
856 |
break;
|
857 |
case FPSCR_VE:
|
858 |
if (fpscr_vx != 0) { |
859 |
raise_ve:
|
860 |
env->error_code = POWERPC_EXCP_FP; |
861 |
if (fpscr_vxsnan)
|
862 |
env->error_code |= POWERPC_EXCP_FP_VXSNAN; |
863 |
if (fpscr_vxisi)
|
864 |
env->error_code |= POWERPC_EXCP_FP_VXISI; |
865 |
if (fpscr_vxidi)
|
866 |
env->error_code |= POWERPC_EXCP_FP_VXIDI; |
867 |
if (fpscr_vxzdz)
|
868 |
env->error_code |= POWERPC_EXCP_FP_VXZDZ; |
869 |
if (fpscr_vximz)
|
870 |
env->error_code |= POWERPC_EXCP_FP_VXIMZ; |
871 |
if (fpscr_vxvc)
|
872 |
env->error_code |= POWERPC_EXCP_FP_VXVC; |
873 |
if (fpscr_vxsoft)
|
874 |
env->error_code |= POWERPC_EXCP_FP_VXSOFT; |
875 |
if (fpscr_vxsqrt)
|
876 |
env->error_code |= POWERPC_EXCP_FP_VXSQRT; |
877 |
if (fpscr_vxcvi)
|
878 |
env->error_code |= POWERPC_EXCP_FP_VXCVI; |
879 |
goto raise_excp;
|
880 |
} |
881 |
break;
|
882 |
case FPSCR_OE:
|
883 |
if (fpscr_ox != 0) { |
884 |
raise_oe:
|
885 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_OX; |
886 |
goto raise_excp;
|
887 |
} |
888 |
break;
|
889 |
case FPSCR_UE:
|
890 |
if (fpscr_ux != 0) { |
891 |
raise_ue:
|
892 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_UX; |
893 |
goto raise_excp;
|
894 |
} |
895 |
break;
|
896 |
case FPSCR_ZE:
|
897 |
if (fpscr_zx != 0) { |
898 |
raise_ze:
|
899 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_ZX; |
900 |
goto raise_excp;
|
901 |
} |
902 |
break;
|
903 |
case FPSCR_XE:
|
904 |
if (fpscr_xx != 0) { |
905 |
raise_xe:
|
906 |
env->error_code = POWERPC_EXCP_FP | POWERPC_EXCP_FP_XX; |
907 |
goto raise_excp;
|
908 |
} |
909 |
break;
|
910 |
case FPSCR_RN1:
|
911 |
case FPSCR_RN:
|
912 |
fpscr_set_rounding_mode(); |
913 |
break;
|
914 |
default:
|
915 |
break;
|
916 |
raise_excp:
|
917 |
/* Update the floating-point enabled exception summary */
|
918 |
env->fpscr |= 1 << FPSCR_FEX;
|
919 |
/* We have to update Rc1 before raising the exception */
|
920 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
921 |
break;
|
922 |
} |
923 |
} |
924 |
} |
925 |
|
926 |
void helper_store_fpscr (uint64_t arg, uint32_t mask)
|
927 |
{ |
928 |
/*
|
929 |
* We use only the 32 LSB of the incoming fpr
|
930 |
*/
|
931 |
uint32_t prev, new; |
932 |
int i;
|
933 |
|
934 |
prev = env->fpscr; |
935 |
new = (uint32_t)arg; |
936 |
new &= ~0x60000000;
|
937 |
new |= prev & 0x60000000;
|
938 |
for (i = 0; i < 8; i++) { |
939 |
if (mask & (1 << i)) { |
940 |
env->fpscr &= ~(0xF << (4 * i)); |
941 |
env->fpscr |= new & (0xF << (4 * i)); |
942 |
} |
943 |
} |
944 |
/* Update VX and FEX */
|
945 |
if (fpscr_ix != 0) |
946 |
env->fpscr |= 1 << FPSCR_VX;
|
947 |
else
|
948 |
env->fpscr &= ~(1 << FPSCR_VX);
|
949 |
if ((fpscr_ex & fpscr_eex) != 0) { |
950 |
env->fpscr |= 1 << FPSCR_FEX;
|
951 |
env->exception_index = POWERPC_EXCP_PROGRAM; |
952 |
/* XXX: we should compute it properly */
|
953 |
env->error_code = POWERPC_EXCP_FP; |
954 |
} |
955 |
else
|
956 |
env->fpscr &= ~(1 << FPSCR_FEX);
|
957 |
fpscr_set_rounding_mode(); |
958 |
} |
959 |
|
960 |
void helper_float_check_status (void) |
961 |
{ |
962 |
#ifdef CONFIG_SOFTFLOAT
|
963 |
if (env->exception_index == POWERPC_EXCP_PROGRAM &&
|
964 |
(env->error_code & POWERPC_EXCP_FP)) { |
965 |
/* Differred floating-point exception after target FPR update */
|
966 |
if (msr_fe0 != 0 || msr_fe1 != 0) |
967 |
helper_raise_exception_err(env->exception_index, env->error_code); |
968 |
} else {
|
969 |
int status = get_float_exception_flags(&env->fp_status);
|
970 |
if (status & float_flag_divbyzero) {
|
971 |
float_zero_divide_excp(); |
972 |
} else if (status & float_flag_overflow) { |
973 |
float_overflow_excp(); |
974 |
} else if (status & float_flag_underflow) { |
975 |
float_underflow_excp(); |
976 |
} else if (status & float_flag_inexact) { |
977 |
float_inexact_excp(); |
978 |
} |
979 |
} |
980 |
#else
|
981 |
if (env->exception_index == POWERPC_EXCP_PROGRAM &&
|
982 |
(env->error_code & POWERPC_EXCP_FP)) { |
983 |
/* Differred floating-point exception after target FPR update */
|
984 |
if (msr_fe0 != 0 || msr_fe1 != 0) |
985 |
helper_raise_exception_err(env->exception_index, env->error_code); |
986 |
} |
987 |
#endif
|
988 |
} |
989 |
|
990 |
#ifdef CONFIG_SOFTFLOAT
|
991 |
void helper_reset_fpstatus (void) |
992 |
{ |
993 |
set_float_exception_flags(0, &env->fp_status);
|
994 |
} |
995 |
#endif
|
996 |
|
997 |
/* fadd - fadd. */
|
998 |
uint64_t helper_fadd (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 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1006 |
float64_is_signaling_nan(farg2.d))) { |
1007 |
/* sNaN addition */
|
1008 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1009 |
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) && |
1010 |
float64_is_neg(farg1.d) != float64_is_neg(farg2.d))) { |
1011 |
/* Magnitude subtraction of infinities */
|
1012 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
1013 |
} else {
|
1014 |
farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status); |
1015 |
} |
1016 |
#else
|
1017 |
farg1.d = float64_add(farg1.d, farg2.d, &env->fp_status); |
1018 |
#endif
|
1019 |
return farg1.ll;
|
1020 |
} |
1021 |
|
1022 |
/* fsub - fsub. */
|
1023 |
uint64_t helper_fsub (uint64_t arg1, uint64_t arg2) |
1024 |
{ |
1025 |
CPU_DoubleU farg1, farg2; |
1026 |
|
1027 |
farg1.ll = arg1; |
1028 |
farg2.ll = arg2; |
1029 |
#if USE_PRECISE_EMULATION
|
1030 |
{ |
1031 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1032 |
float64_is_signaling_nan(farg2.d))) { |
1033 |
/* sNaN subtraction */
|
1034 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1035 |
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d) && |
1036 |
float64_is_neg(farg1.d) == float64_is_neg(farg2.d))) { |
1037 |
/* Magnitude subtraction of infinities */
|
1038 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
1039 |
} else {
|
1040 |
farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status); |
1041 |
} |
1042 |
} |
1043 |
#else
|
1044 |
farg1.d = float64_sub(farg1.d, farg2.d, &env->fp_status); |
1045 |
#endif
|
1046 |
return farg1.ll;
|
1047 |
} |
1048 |
|
1049 |
/* fmul - fmul. */
|
1050 |
uint64_t helper_fmul (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 multiplication */
|
1060 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1061 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
1062 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { |
1063 |
/* Multiplication of zero by infinity */
|
1064 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
1065 |
} else {
|
1066 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
1067 |
} |
1068 |
#else
|
1069 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
1070 |
#endif
|
1071 |
return farg1.ll;
|
1072 |
} |
1073 |
|
1074 |
/* fdiv - fdiv. */
|
1075 |
uint64_t helper_fdiv (uint64_t arg1, uint64_t arg2) |
1076 |
{ |
1077 |
CPU_DoubleU farg1, farg2; |
1078 |
|
1079 |
farg1.ll = arg1; |
1080 |
farg2.ll = arg2; |
1081 |
#if USE_PRECISE_EMULATION
|
1082 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1083 |
float64_is_signaling_nan(farg2.d))) { |
1084 |
/* sNaN division */
|
1085 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1086 |
} else if (unlikely(float64_is_infinity(farg1.d) && float64_is_infinity(farg2.d))) { |
1087 |
/* Division of infinity by infinity */
|
1088 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIDI); |
1089 |
} else if (unlikely(float64_is_zero(farg1.d) && float64_is_zero(farg2.d))) { |
1090 |
/* Division of zero by zero */
|
1091 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXZDZ); |
1092 |
} else {
|
1093 |
farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status); |
1094 |
} |
1095 |
#else
|
1096 |
farg1.d = float64_div(farg1.d, farg2.d, &env->fp_status); |
1097 |
#endif
|
1098 |
return farg1.ll;
|
1099 |
} |
1100 |
|
1101 |
/* fabs */
|
1102 |
uint64_t helper_fabs (uint64_t arg) |
1103 |
{ |
1104 |
CPU_DoubleU farg; |
1105 |
|
1106 |
farg.ll = arg; |
1107 |
farg.d = float64_abs(farg.d); |
1108 |
return farg.ll;
|
1109 |
} |
1110 |
|
1111 |
/* fnabs */
|
1112 |
uint64_t helper_fnabs (uint64_t arg) |
1113 |
{ |
1114 |
CPU_DoubleU farg; |
1115 |
|
1116 |
farg.ll = arg; |
1117 |
farg.d = float64_abs(farg.d); |
1118 |
farg.d = float64_chs(farg.d); |
1119 |
return farg.ll;
|
1120 |
} |
1121 |
|
1122 |
/* fneg */
|
1123 |
uint64_t helper_fneg (uint64_t arg) |
1124 |
{ |
1125 |
CPU_DoubleU farg; |
1126 |
|
1127 |
farg.ll = arg; |
1128 |
farg.d = float64_chs(farg.d); |
1129 |
return farg.ll;
|
1130 |
} |
1131 |
|
1132 |
/* fctiw - fctiw. */
|
1133 |
uint64_t helper_fctiw (uint64_t arg) |
1134 |
{ |
1135 |
CPU_DoubleU farg; |
1136 |
farg.ll = arg; |
1137 |
|
1138 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1139 |
/* sNaN conversion */
|
1140 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
1141 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
1142 |
/* qNan / infinity conversion */
|
1143 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
1144 |
} else {
|
1145 |
farg.ll = float64_to_int32(farg.d, &env->fp_status); |
1146 |
#if USE_PRECISE_EMULATION
|
1147 |
/* XXX: higher bits are not supposed to be significant.
|
1148 |
* to make tests easier, return the same as a real PowerPC 750
|
1149 |
*/
|
1150 |
farg.ll |= 0xFFF80000ULL << 32; |
1151 |
#endif
|
1152 |
} |
1153 |
return farg.ll;
|
1154 |
} |
1155 |
|
1156 |
/* fctiwz - fctiwz. */
|
1157 |
uint64_t helper_fctiwz (uint64_t arg) |
1158 |
{ |
1159 |
CPU_DoubleU farg; |
1160 |
farg.ll = arg; |
1161 |
|
1162 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1163 |
/* sNaN conversion */
|
1164 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
1165 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
1166 |
/* qNan / infinity conversion */
|
1167 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
1168 |
} else {
|
1169 |
farg.ll = float64_to_int32_round_to_zero(farg.d, &env->fp_status); |
1170 |
#if USE_PRECISE_EMULATION
|
1171 |
/* XXX: higher bits are not supposed to be significant.
|
1172 |
* to make tests easier, return the same as a real PowerPC 750
|
1173 |
*/
|
1174 |
farg.ll |= 0xFFF80000ULL << 32; |
1175 |
#endif
|
1176 |
} |
1177 |
return farg.ll;
|
1178 |
} |
1179 |
|
1180 |
#if defined(TARGET_PPC64)
|
1181 |
/* fcfid - fcfid. */
|
1182 |
uint64_t helper_fcfid (uint64_t arg) |
1183 |
{ |
1184 |
CPU_DoubleU farg; |
1185 |
farg.d = int64_to_float64(arg, &env->fp_status); |
1186 |
return farg.ll;
|
1187 |
} |
1188 |
|
1189 |
/* fctid - fctid. */
|
1190 |
uint64_t helper_fctid (uint64_t arg) |
1191 |
{ |
1192 |
CPU_DoubleU farg; |
1193 |
farg.ll = arg; |
1194 |
|
1195 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1196 |
/* sNaN conversion */
|
1197 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
1198 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
1199 |
/* qNan / infinity conversion */
|
1200 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
1201 |
} else {
|
1202 |
farg.ll = float64_to_int64(farg.d, &env->fp_status); |
1203 |
} |
1204 |
return farg.ll;
|
1205 |
} |
1206 |
|
1207 |
/* fctidz - fctidz. */
|
1208 |
uint64_t helper_fctidz (uint64_t arg) |
1209 |
{ |
1210 |
CPU_DoubleU farg; |
1211 |
farg.ll = arg; |
1212 |
|
1213 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1214 |
/* sNaN conversion */
|
1215 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
1216 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
1217 |
/* qNan / infinity conversion */
|
1218 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
1219 |
} else {
|
1220 |
farg.ll = float64_to_int64_round_to_zero(farg.d, &env->fp_status); |
1221 |
} |
1222 |
return farg.ll;
|
1223 |
} |
1224 |
|
1225 |
#endif
|
1226 |
|
1227 |
static always_inline uint64_t do_fri (uint64_t arg, int rounding_mode) |
1228 |
{ |
1229 |
CPU_DoubleU farg; |
1230 |
farg.ll = arg; |
1231 |
|
1232 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1233 |
/* sNaN round */
|
1234 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | POWERPC_EXCP_FP_VXCVI); |
1235 |
} else if (unlikely(float64_is_nan(farg.d) || float64_is_infinity(farg.d))) { |
1236 |
/* qNan / infinity round */
|
1237 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXCVI); |
1238 |
} else {
|
1239 |
set_float_rounding_mode(rounding_mode, &env->fp_status); |
1240 |
farg.ll = float64_round_to_int(farg.d, &env->fp_status); |
1241 |
/* Restore rounding mode from FPSCR */
|
1242 |
fpscr_set_rounding_mode(); |
1243 |
} |
1244 |
return farg.ll;
|
1245 |
} |
1246 |
|
1247 |
uint64_t helper_frin (uint64_t arg) |
1248 |
{ |
1249 |
return do_fri(arg, float_round_nearest_even);
|
1250 |
} |
1251 |
|
1252 |
uint64_t helper_friz (uint64_t arg) |
1253 |
{ |
1254 |
return do_fri(arg, float_round_to_zero);
|
1255 |
} |
1256 |
|
1257 |
uint64_t helper_frip (uint64_t arg) |
1258 |
{ |
1259 |
return do_fri(arg, float_round_up);
|
1260 |
} |
1261 |
|
1262 |
uint64_t helper_frim (uint64_t arg) |
1263 |
{ |
1264 |
return do_fri(arg, float_round_down);
|
1265 |
} |
1266 |
|
1267 |
/* fmadd - fmadd. */
|
1268 |
uint64_t helper_fmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
1269 |
{ |
1270 |
CPU_DoubleU farg1, farg2, farg3; |
1271 |
|
1272 |
farg1.ll = arg1; |
1273 |
farg2.ll = arg2; |
1274 |
farg3.ll = arg3; |
1275 |
#if USE_PRECISE_EMULATION
|
1276 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1277 |
float64_is_signaling_nan(farg2.d) || |
1278 |
float64_is_signaling_nan(farg3.d))) { |
1279 |
/* sNaN operation */
|
1280 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1281 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
1282 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { |
1283 |
/* Multiplication of zero by infinity */
|
1284 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
1285 |
} else {
|
1286 |
#ifdef FLOAT128
|
1287 |
/* This is the way the PowerPC specification defines it */
|
1288 |
float128 ft0_128, ft1_128; |
1289 |
|
1290 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
1291 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
1292 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
1293 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
1294 |
float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) { |
1295 |
/* Magnitude subtraction of infinities */
|
1296 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
1297 |
} else {
|
1298 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
1299 |
ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status); |
1300 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
1301 |
} |
1302 |
#else
|
1303 |
/* This is OK on x86 hosts */
|
1304 |
farg1.d = (farg1.d * farg2.d) + farg3.d; |
1305 |
#endif
|
1306 |
} |
1307 |
#else
|
1308 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
1309 |
farg1.d = float64_add(farg1.d, farg3.d, &env->fp_status); |
1310 |
#endif
|
1311 |
return farg1.ll;
|
1312 |
} |
1313 |
|
1314 |
/* fmsub - fmsub. */
|
1315 |
uint64_t helper_fmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
1316 |
{ |
1317 |
CPU_DoubleU farg1, farg2, farg3; |
1318 |
|
1319 |
farg1.ll = arg1; |
1320 |
farg2.ll = arg2; |
1321 |
farg3.ll = arg3; |
1322 |
#if USE_PRECISE_EMULATION
|
1323 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1324 |
float64_is_signaling_nan(farg2.d) || |
1325 |
float64_is_signaling_nan(farg3.d))) { |
1326 |
/* sNaN operation */
|
1327 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1328 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
1329 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { |
1330 |
/* Multiplication of zero by infinity */
|
1331 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
1332 |
} else {
|
1333 |
#ifdef FLOAT128
|
1334 |
/* This is the way the PowerPC specification defines it */
|
1335 |
float128 ft0_128, ft1_128; |
1336 |
|
1337 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
1338 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
1339 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
1340 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
1341 |
float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) { |
1342 |
/* Magnitude subtraction of infinities */
|
1343 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
1344 |
} else {
|
1345 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
1346 |
ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status); |
1347 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
1348 |
} |
1349 |
#else
|
1350 |
/* This is OK on x86 hosts */
|
1351 |
farg1.d = (farg1.d * farg2.d) - farg3.d; |
1352 |
#endif
|
1353 |
} |
1354 |
#else
|
1355 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
1356 |
farg1.d = float64_sub(farg1.d, farg3.d, &env->fp_status); |
1357 |
#endif
|
1358 |
return farg1.ll;
|
1359 |
} |
1360 |
|
1361 |
/* fnmadd - fnmadd. */
|
1362 |
uint64_t helper_fnmadd (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
1363 |
{ |
1364 |
CPU_DoubleU farg1, farg2, farg3; |
1365 |
|
1366 |
farg1.ll = arg1; |
1367 |
farg2.ll = arg2; |
1368 |
farg3.ll = arg3; |
1369 |
|
1370 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1371 |
float64_is_signaling_nan(farg2.d) || |
1372 |
float64_is_signaling_nan(farg3.d))) { |
1373 |
/* sNaN operation */
|
1374 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1375 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
1376 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { |
1377 |
/* Multiplication of zero by infinity */
|
1378 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
1379 |
} else {
|
1380 |
#if USE_PRECISE_EMULATION
|
1381 |
#ifdef FLOAT128
|
1382 |
/* This is the way the PowerPC specification defines it */
|
1383 |
float128 ft0_128, ft1_128; |
1384 |
|
1385 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
1386 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
1387 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
1388 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
1389 |
float128_is_neg(ft0_128) != float64_is_neg(farg3.d))) { |
1390 |
/* Magnitude subtraction of infinities */
|
1391 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
1392 |
} else {
|
1393 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
1394 |
ft0_128 = float128_add(ft0_128, ft1_128, &env->fp_status); |
1395 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
1396 |
} |
1397 |
#else
|
1398 |
/* This is OK on x86 hosts */
|
1399 |
farg1.d = (farg1.d * farg2.d) + farg3.d; |
1400 |
#endif
|
1401 |
#else
|
1402 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
1403 |
farg1.d = float64_add(farg1.d, farg3.d, &env->fp_status); |
1404 |
#endif
|
1405 |
if (likely(!float64_is_nan(farg1.d)))
|
1406 |
farg1.d = float64_chs(farg1.d); |
1407 |
} |
1408 |
return farg1.ll;
|
1409 |
} |
1410 |
|
1411 |
/* fnmsub - fnmsub. */
|
1412 |
uint64_t helper_fnmsub (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
1413 |
{ |
1414 |
CPU_DoubleU farg1, farg2, farg3; |
1415 |
|
1416 |
farg1.ll = arg1; |
1417 |
farg2.ll = arg2; |
1418 |
farg3.ll = arg3; |
1419 |
|
1420 |
if (unlikely(float64_is_signaling_nan(farg1.d) ||
|
1421 |
float64_is_signaling_nan(farg2.d) || |
1422 |
float64_is_signaling_nan(farg3.d))) { |
1423 |
/* sNaN operation */
|
1424 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1425 |
} else if (unlikely((float64_is_infinity(farg1.d) && float64_is_zero(farg2.d)) || |
1426 |
(float64_is_zero(farg1.d) && float64_is_infinity(farg2.d)))) { |
1427 |
/* Multiplication of zero by infinity */
|
1428 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXIMZ); |
1429 |
} else {
|
1430 |
#if USE_PRECISE_EMULATION
|
1431 |
#ifdef FLOAT128
|
1432 |
/* This is the way the PowerPC specification defines it */
|
1433 |
float128 ft0_128, ft1_128; |
1434 |
|
1435 |
ft0_128 = float64_to_float128(farg1.d, &env->fp_status); |
1436 |
ft1_128 = float64_to_float128(farg2.d, &env->fp_status); |
1437 |
ft0_128 = float128_mul(ft0_128, ft1_128, &env->fp_status); |
1438 |
if (unlikely(float128_is_infinity(ft0_128) && float64_is_infinity(farg3.d) &&
|
1439 |
float128_is_neg(ft0_128) == float64_is_neg(farg3.d))) { |
1440 |
/* Magnitude subtraction of infinities */
|
1441 |
farg1.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXISI); |
1442 |
} else {
|
1443 |
ft1_128 = float64_to_float128(farg3.d, &env->fp_status); |
1444 |
ft0_128 = float128_sub(ft0_128, ft1_128, &env->fp_status); |
1445 |
farg1.d = float128_to_float64(ft0_128, &env->fp_status); |
1446 |
} |
1447 |
#else
|
1448 |
/* This is OK on x86 hosts */
|
1449 |
farg1.d = (farg1.d * farg2.d) - farg3.d; |
1450 |
#endif
|
1451 |
#else
|
1452 |
farg1.d = float64_mul(farg1.d, farg2.d, &env->fp_status); |
1453 |
farg1.d = float64_sub(farg1.d, farg3.d, &env->fp_status); |
1454 |
#endif
|
1455 |
if (likely(!float64_is_nan(farg1.d)))
|
1456 |
farg1.d = float64_chs(farg1.d); |
1457 |
} |
1458 |
return farg1.ll;
|
1459 |
} |
1460 |
|
1461 |
/* frsp - frsp. */
|
1462 |
uint64_t helper_frsp (uint64_t arg) |
1463 |
{ |
1464 |
CPU_DoubleU farg; |
1465 |
float32 f32; |
1466 |
farg.ll = arg; |
1467 |
|
1468 |
#if USE_PRECISE_EMULATION
|
1469 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1470 |
/* sNaN square root */
|
1471 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1472 |
} else {
|
1473 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
1474 |
farg.d = float32_to_float64(f32, &env->fp_status); |
1475 |
} |
1476 |
#else
|
1477 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
1478 |
farg.d = float32_to_float64(f32, &env->fp_status); |
1479 |
#endif
|
1480 |
return farg.ll;
|
1481 |
} |
1482 |
|
1483 |
/* fsqrt - fsqrt. */
|
1484 |
uint64_t helper_fsqrt (uint64_t arg) |
1485 |
{ |
1486 |
CPU_DoubleU farg; |
1487 |
farg.ll = arg; |
1488 |
|
1489 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1490 |
/* sNaN square root */
|
1491 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1492 |
} else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) { |
1493 |
/* Square root of a negative nonzero number */
|
1494 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT); |
1495 |
} else {
|
1496 |
farg.d = float64_sqrt(farg.d, &env->fp_status); |
1497 |
} |
1498 |
return farg.ll;
|
1499 |
} |
1500 |
|
1501 |
/* fre - fre. */
|
1502 |
uint64_t helper_fre (uint64_t arg) |
1503 |
{ |
1504 |
CPU_DoubleU farg; |
1505 |
farg.ll = arg; |
1506 |
|
1507 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1508 |
/* sNaN reciprocal */
|
1509 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1510 |
} else {
|
1511 |
farg.d = float64_div(float64_one, farg.d, &env->fp_status); |
1512 |
} |
1513 |
return farg.d;
|
1514 |
} |
1515 |
|
1516 |
/* fres - fres. */
|
1517 |
uint64_t helper_fres (uint64_t arg) |
1518 |
{ |
1519 |
CPU_DoubleU farg; |
1520 |
float32 f32; |
1521 |
farg.ll = arg; |
1522 |
|
1523 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1524 |
/* sNaN reciprocal */
|
1525 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1526 |
} else {
|
1527 |
farg.d = float64_div(float64_one, farg.d, &env->fp_status); |
1528 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
1529 |
farg.d = float32_to_float64(f32, &env->fp_status); |
1530 |
} |
1531 |
return farg.ll;
|
1532 |
} |
1533 |
|
1534 |
/* frsqrte - frsqrte. */
|
1535 |
uint64_t helper_frsqrte (uint64_t arg) |
1536 |
{ |
1537 |
CPU_DoubleU farg; |
1538 |
float32 f32; |
1539 |
farg.ll = arg; |
1540 |
|
1541 |
if (unlikely(float64_is_signaling_nan(farg.d))) {
|
1542 |
/* sNaN reciprocal square root */
|
1543 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1544 |
} else if (unlikely(float64_is_neg(farg.d) && !float64_is_zero(farg.d))) { |
1545 |
/* Reciprocal square root of a negative nonzero number */
|
1546 |
farg.ll = fload_invalid_op_excp(POWERPC_EXCP_FP_VXSQRT); |
1547 |
} else {
|
1548 |
farg.d = float64_sqrt(farg.d, &env->fp_status); |
1549 |
farg.d = float64_div(float64_one, farg.d, &env->fp_status); |
1550 |
f32 = float64_to_float32(farg.d, &env->fp_status); |
1551 |
farg.d = float32_to_float64(f32, &env->fp_status); |
1552 |
} |
1553 |
return farg.ll;
|
1554 |
} |
1555 |
|
1556 |
/* fsel - fsel. */
|
1557 |
uint64_t helper_fsel (uint64_t arg1, uint64_t arg2, uint64_t arg3) |
1558 |
{ |
1559 |
CPU_DoubleU farg1; |
1560 |
|
1561 |
farg1.ll = arg1; |
1562 |
|
1563 |
if ((!float64_is_neg(farg1.d) || float64_is_zero(farg1.d)) && !float64_is_nan(farg1.d))
|
1564 |
return arg2;
|
1565 |
else
|
1566 |
return arg3;
|
1567 |
} |
1568 |
|
1569 |
void helper_fcmpu (uint64_t arg1, uint64_t arg2, uint32_t crfD)
|
1570 |
{ |
1571 |
CPU_DoubleU farg1, farg2; |
1572 |
uint32_t ret = 0;
|
1573 |
farg1.ll = arg1; |
1574 |
farg2.ll = arg2; |
1575 |
|
1576 |
if (unlikely(float64_is_nan(farg1.d) ||
|
1577 |
float64_is_nan(farg2.d))) { |
1578 |
ret = 0x01UL;
|
1579 |
} else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) { |
1580 |
ret = 0x08UL;
|
1581 |
} else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) { |
1582 |
ret = 0x04UL;
|
1583 |
} else {
|
1584 |
ret = 0x02UL;
|
1585 |
} |
1586 |
|
1587 |
env->fpscr &= ~(0x0F << FPSCR_FPRF);
|
1588 |
env->fpscr |= ret << FPSCR_FPRF; |
1589 |
env->crf[crfD] = ret; |
1590 |
if (unlikely(ret == 0x01UL |
1591 |
&& (float64_is_signaling_nan(farg1.d) || |
1592 |
float64_is_signaling_nan(farg2.d)))) { |
1593 |
/* sNaN comparison */
|
1594 |
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN); |
1595 |
} |
1596 |
} |
1597 |
|
1598 |
void helper_fcmpo (uint64_t arg1, uint64_t arg2, uint32_t crfD)
|
1599 |
{ |
1600 |
CPU_DoubleU farg1, farg2; |
1601 |
uint32_t ret = 0;
|
1602 |
farg1.ll = arg1; |
1603 |
farg2.ll = arg2; |
1604 |
|
1605 |
if (unlikely(float64_is_nan(farg1.d) ||
|
1606 |
float64_is_nan(farg2.d))) { |
1607 |
ret = 0x01UL;
|
1608 |
} else if (float64_lt(farg1.d, farg2.d, &env->fp_status)) { |
1609 |
ret = 0x08UL;
|
1610 |
} else if (!float64_le(farg1.d, farg2.d, &env->fp_status)) { |
1611 |
ret = 0x04UL;
|
1612 |
} else {
|
1613 |
ret = 0x02UL;
|
1614 |
} |
1615 |
|
1616 |
env->fpscr &= ~(0x0F << FPSCR_FPRF);
|
1617 |
env->fpscr |= ret << FPSCR_FPRF; |
1618 |
env->crf[crfD] = ret; |
1619 |
if (unlikely (ret == 0x01UL)) { |
1620 |
if (float64_is_signaling_nan(farg1.d) ||
|
1621 |
float64_is_signaling_nan(farg2.d)) { |
1622 |
/* sNaN comparison */
|
1623 |
fload_invalid_op_excp(POWERPC_EXCP_FP_VXSNAN | |
1624 |
POWERPC_EXCP_FP_VXVC); |
1625 |
} else {
|
1626 |
/* qNaN comparison */
|
1627 |
fload_invalid_op_excp(POWERPC_EXCP_FP_VXVC); |
1628 |
} |
1629 |
} |
1630 |
} |
1631 |
|
1632 |
#if !defined (CONFIG_USER_ONLY)
|
1633 |
void helper_store_msr (target_ulong val)
|
1634 |
{ |
1635 |
val = hreg_store_msr(env, val, 0);
|
1636 |
if (val != 0) { |
1637 |
env->interrupt_request |= CPU_INTERRUPT_EXITTB; |
1638 |
helper_raise_exception(val); |
1639 |
} |
1640 |
} |
1641 |
|
1642 |
static always_inline void do_rfi (target_ulong nip, target_ulong msr, |
1643 |
target_ulong msrm, int keep_msrh)
|
1644 |
{ |
1645 |
#if defined(TARGET_PPC64)
|
1646 |
if (msr & (1ULL << MSR_SF)) { |
1647 |
nip = (uint64_t)nip; |
1648 |
msr &= (uint64_t)msrm; |
1649 |
} else {
|
1650 |
nip = (uint32_t)nip; |
1651 |
msr = (uint32_t)(msr & msrm); |
1652 |
if (keep_msrh)
|
1653 |
msr |= env->msr & ~((uint64_t)0xFFFFFFFF);
|
1654 |
} |
1655 |
#else
|
1656 |
nip = (uint32_t)nip; |
1657 |
msr &= (uint32_t)msrm; |
1658 |
#endif
|
1659 |
/* XXX: beware: this is false if VLE is supported */
|
1660 |
env->nip = nip & ~((target_ulong)0x00000003);
|
1661 |
hreg_store_msr(env, msr, 1);
|
1662 |
#if defined (DEBUG_OP)
|
1663 |
cpu_dump_rfi(env->nip, env->msr); |
1664 |
#endif
|
1665 |
/* No need to raise an exception here,
|
1666 |
* as rfi is always the last insn of a TB
|
1667 |
*/
|
1668 |
env->interrupt_request |= CPU_INTERRUPT_EXITTB; |
1669 |
} |
1670 |
|
1671 |
void helper_rfi (void) |
1672 |
{ |
1673 |
do_rfi(env->spr[SPR_SRR0], env->spr[SPR_SRR1], |
1674 |
~((target_ulong)0xFFFF0000), 1); |
1675 |
} |
1676 |
|
1677 |
#if defined(TARGET_PPC64)
|
1678 |
void helper_rfid (void) |
1679 |
{ |
1680 |
do_rfi(env->spr[SPR_SRR0], env->spr[SPR_SRR1], |
1681 |
~((target_ulong)0xFFFF0000), 0); |
1682 |
} |
1683 |
|
1684 |
void helper_hrfid (void) |
1685 |
{ |
1686 |
do_rfi(env->spr[SPR_HSRR0], env->spr[SPR_HSRR1], |
1687 |
~((target_ulong)0xFFFF0000), 0); |
1688 |
} |
1689 |
#endif
|
1690 |
#endif
|
1691 |
|
1692 |
void helper_tw (target_ulong arg1, target_ulong arg2, uint32_t flags)
|
1693 |
{ |
1694 |
if (!likely(!(((int32_t)arg1 < (int32_t)arg2 && (flags & 0x10)) || |
1695 |
((int32_t)arg1 > (int32_t)arg2 && (flags & 0x08)) ||
|
1696 |
((int32_t)arg1 == (int32_t)arg2 && (flags & 0x04)) ||
|
1697 |
((uint32_t)arg1 < (uint32_t)arg2 && (flags & 0x02)) ||
|
1698 |
((uint32_t)arg1 > (uint32_t)arg2 && (flags & 0x01))))) {
|
1699 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_TRAP); |
1700 |
} |
1701 |
} |
1702 |
|
1703 |
#if defined(TARGET_PPC64)
|
1704 |
void helper_td (target_ulong arg1, target_ulong arg2, uint32_t flags)
|
1705 |
{ |
1706 |
if (!likely(!(((int64_t)arg1 < (int64_t)arg2 && (flags & 0x10)) || |
1707 |
((int64_t)arg1 > (int64_t)arg2 && (flags & 0x08)) ||
|
1708 |
((int64_t)arg1 == (int64_t)arg2 && (flags & 0x04)) ||
|
1709 |
((uint64_t)arg1 < (uint64_t)arg2 && (flags & 0x02)) ||
|
1710 |
((uint64_t)arg1 > (uint64_t)arg2 && (flags & 0x01)))))
|
1711 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, POWERPC_EXCP_TRAP); |
1712 |
} |
1713 |
#endif
|
1714 |
|
1715 |
/*****************************************************************************/
|
1716 |
/* PowerPC 601 specific instructions (POWER bridge) */
|
1717 |
|
1718 |
target_ulong helper_clcs (uint32_t arg) |
1719 |
{ |
1720 |
switch (arg) {
|
1721 |
case 0x0CUL: |
1722 |
/* Instruction cache line size */
|
1723 |
return env->icache_line_size;
|
1724 |
break;
|
1725 |
case 0x0DUL: |
1726 |
/* Data cache line size */
|
1727 |
return env->dcache_line_size;
|
1728 |
break;
|
1729 |
case 0x0EUL: |
1730 |
/* Minimum cache line size */
|
1731 |
return (env->icache_line_size < env->dcache_line_size) ?
|
1732 |
env->icache_line_size : env->dcache_line_size; |
1733 |
break;
|
1734 |
case 0x0FUL: |
1735 |
/* Maximum cache line size */
|
1736 |
return (env->icache_line_size > env->dcache_line_size) ?
|
1737 |
env->icache_line_size : env->dcache_line_size; |
1738 |
break;
|
1739 |
default:
|
1740 |
/* Undefined */
|
1741 |
return 0; |
1742 |
break;
|
1743 |
} |
1744 |
} |
1745 |
|
1746 |
target_ulong helper_div (target_ulong arg1, target_ulong arg2) |
1747 |
{ |
1748 |
uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
|
1749 |
|
1750 |
if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
1751 |
(int32_t)arg2 == 0) {
|
1752 |
env->spr[SPR_MQ] = 0;
|
1753 |
return INT32_MIN;
|
1754 |
} else {
|
1755 |
env->spr[SPR_MQ] = tmp % arg2; |
1756 |
return tmp / (int32_t)arg2;
|
1757 |
} |
1758 |
} |
1759 |
|
1760 |
target_ulong helper_divo (target_ulong arg1, target_ulong arg2) |
1761 |
{ |
1762 |
uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
|
1763 |
|
1764 |
if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
1765 |
(int32_t)arg2 == 0) {
|
1766 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
1767 |
env->spr[SPR_MQ] = 0;
|
1768 |
return INT32_MIN;
|
1769 |
} else {
|
1770 |
env->spr[SPR_MQ] = tmp % arg2; |
1771 |
tmp /= (int32_t)arg2; |
1772 |
if ((int32_t)tmp != tmp) {
|
1773 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
1774 |
} else {
|
1775 |
env->xer &= ~(1 << XER_OV);
|
1776 |
} |
1777 |
return tmp;
|
1778 |
} |
1779 |
} |
1780 |
|
1781 |
target_ulong helper_divs (target_ulong arg1, target_ulong arg2) |
1782 |
{ |
1783 |
if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
1784 |
(int32_t)arg2 == 0) {
|
1785 |
env->spr[SPR_MQ] = 0;
|
1786 |
return INT32_MIN;
|
1787 |
} else {
|
1788 |
env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2; |
1789 |
return (int32_t)arg1 / (int32_t)arg2;
|
1790 |
} |
1791 |
} |
1792 |
|
1793 |
target_ulong helper_divso (target_ulong arg1, target_ulong arg2) |
1794 |
{ |
1795 |
if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) || |
1796 |
(int32_t)arg2 == 0) {
|
1797 |
env->xer |= (1 << XER_OV) | (1 << XER_SO); |
1798 |
env->spr[SPR_MQ] = 0;
|
1799 |
return INT32_MIN;
|
1800 |
} else {
|
1801 |
env->xer &= ~(1 << XER_OV);
|
1802 |
env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2; |
1803 |
return (int32_t)arg1 / (int32_t)arg2;
|
1804 |
} |
1805 |
} |
1806 |
|
1807 |
#if !defined (CONFIG_USER_ONLY)
|
1808 |
target_ulong helper_rac (target_ulong addr) |
1809 |
{ |
1810 |
mmu_ctx_t ctx; |
1811 |
int nb_BATs;
|
1812 |
target_ulong ret = 0;
|
1813 |
|
1814 |
/* We don't have to generate many instances of this instruction,
|
1815 |
* as rac is supervisor only.
|
1816 |
*/
|
1817 |
/* XXX: FIX THIS: Pretend we have no BAT */
|
1818 |
nb_BATs = env->nb_BATs; |
1819 |
env->nb_BATs = 0;
|
1820 |
if (get_physical_address(env, &ctx, addr, 0, ACCESS_INT) == 0) |
1821 |
ret = ctx.raddr; |
1822 |
env->nb_BATs = nb_BATs; |
1823 |
return ret;
|
1824 |
} |
1825 |
|
1826 |
void helper_rfsvc (void) |
1827 |
{ |
1828 |
do_rfi(env->lr, env->ctr, 0x0000FFFF, 0); |
1829 |
} |
1830 |
#endif
|
1831 |
|
1832 |
/*****************************************************************************/
|
1833 |
/* 602 specific instructions */
|
1834 |
/* mfrom is the most crazy instruction ever seen, imho ! */
|
1835 |
/* Real implementation uses a ROM table. Do the same */
|
1836 |
/* Extremly decomposed:
|
1837 |
* -arg / 256
|
1838 |
* return 256 * log10(10 + 1.0) + 0.5
|
1839 |
*/
|
1840 |
#if !defined (CONFIG_USER_ONLY)
|
1841 |
target_ulong helper_602_mfrom (target_ulong arg) |
1842 |
{ |
1843 |
if (likely(arg < 602)) { |
1844 |
#include "mfrom_table.c" |
1845 |
return mfrom_ROM_table[arg];
|
1846 |
} else {
|
1847 |
return 0; |
1848 |
} |
1849 |
} |
1850 |
#endif
|
1851 |
|
1852 |
/*****************************************************************************/
|
1853 |
/* Embedded PowerPC specific helpers */
|
1854 |
|
1855 |
/* XXX: to be improved to check access rights when in user-mode */
|
1856 |
target_ulong helper_load_dcr (target_ulong dcrn) |
1857 |
{ |
1858 |
target_ulong val = 0;
|
1859 |
|
1860 |
if (unlikely(env->dcr_env == NULL)) { |
1861 |
qemu_log("No DCR environment\n");
|
1862 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
1863 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_INVAL_INVAL); |
1864 |
} else if (unlikely(ppc_dcr_read(env->dcr_env, dcrn, &val) != 0)) { |
1865 |
qemu_log("DCR read error %d %03x\n", (int)dcrn, (int)dcrn); |
1866 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
1867 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_PRIV_REG); |
1868 |
} |
1869 |
return val;
|
1870 |
} |
1871 |
|
1872 |
void helper_store_dcr (target_ulong dcrn, target_ulong val)
|
1873 |
{ |
1874 |
if (unlikely(env->dcr_env == NULL)) { |
1875 |
qemu_log("No DCR environment\n");
|
1876 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
1877 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_INVAL_INVAL); |
1878 |
} else if (unlikely(ppc_dcr_write(env->dcr_env, dcrn, val) != 0)) { |
1879 |
qemu_log("DCR write error %d %03x\n", (int)dcrn, (int)dcrn); |
1880 |
helper_raise_exception_err(POWERPC_EXCP_PROGRAM, |
1881 |
POWERPC_EXCP_INVAL | POWERPC_EXCP_PRIV_REG); |
1882 |
} |
1883 |
} |
1884 |
|
1885 |
#if !defined(CONFIG_USER_ONLY)
|
1886 |
void helper_40x_rfci (void) |
1887 |
{ |
1888 |
do_rfi(env->spr[SPR_40x_SRR2], env->spr[SPR_40x_SRR3], |
1889 |
~((target_ulong)0xFFFF0000), 0); |
1890 |
} |
1891 |
|
1892 |
void helper_rfci (void) |
1893 |
{ |
1894 |
do_rfi(env->spr[SPR_BOOKE_CSRR0], SPR_BOOKE_CSRR1, |
1895 |
~((target_ulong)0x3FFF0000), 0); |
1896 |
} |
1897 |
|
1898 |
void helper_rfdi (void) |
1899 |
{ |
1900 |
do_rfi(env->spr[SPR_BOOKE_DSRR0], SPR_BOOKE_DSRR1, |
1901 |
~((target_ulong)0x3FFF0000), 0); |
1902 |
} |
1903 |
|
1904 |
void helper_rfmci (void) |
1905 |
{ |
1906 |
do_rfi(env->spr[SPR_BOOKE_MCSRR0], SPR_BOOKE_MCSRR1, |
1907 |
~((target_ulong)0x3FFF0000), 0); |
1908 |
} |
1909 |
#endif
|
1910 |
|
1911 |
/* 440 specific */
|
1912 |
target_ulong helper_dlmzb (target_ulong high, target_ulong low, uint32_t update_Rc) |
1913 |
{ |
1914 |
target_ulong mask; |
1915 |
int i;
|
1916 |
|
1917 |
i = 1;
|
1918 |
for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { |
1919 |
if ((high & mask) == 0) { |
1920 |
if (update_Rc) {
|
1921 |
env->crf[0] = 0x4; |
1922 |
} |
1923 |
goto done;
|
1924 |
} |
1925 |
i++; |
1926 |
} |
1927 |
for (mask = 0xFF000000; mask != 0; mask = mask >> 8) { |
1928 |
if ((low & mask) == 0) { |
1929 |
if (update_Rc) {
|
1930 |
env->crf[0] = 0x8; |
1931 |
} |
1932 |
goto done;
|
1933 |
} |
1934 |
i++; |
1935 |
} |
1936 |
if (update_Rc) {
|
1937 |
env->crf[0] = 0x2; |
1938 |
} |
1939 |
done:
|
1940 |
env->xer = (env->xer & ~0x7F) | i;
|
1941 |
if (update_Rc) {
|
1942 |
env->crf[0] |= xer_so;
|
1943 |
} |
1944 |
return i;
|
1945 |
} |
1946 |
|
1947 |
/*****************************************************************************/
|
1948 |
/* Altivec extension helpers */
|
1949 |
#if defined(WORDS_BIGENDIAN)
|
1950 |
#define HI_IDX 0 |
1951 |
#define LO_IDX 1 |
1952 |
#else
|
1953 |
#define HI_IDX 1 |
1954 |
#define LO_IDX 0 |
1955 |
#endif
|
1956 |
|
1957 |
#if defined(WORDS_BIGENDIAN)
|
1958 |
#define VECTOR_FOR_INORDER_I(index, element) \
|
1959 |
for (index = 0; index < ARRAY_SIZE(r->element); index++) |
1960 |
#else
|
1961 |
#define VECTOR_FOR_INORDER_I(index, element) \
|
1962 |
for (index = ARRAY_SIZE(r->element)-1; index >= 0; index--) |
1963 |
#endif
|
1964 |
|
1965 |
/* If X is a NaN, store the corresponding QNaN into RESULT. Otherwise,
|
1966 |
* execute the following block. */
|
1967 |
#define DO_HANDLE_NAN(result, x) \
|
1968 |
if (float32_is_nan(x) || float32_is_signaling_nan(x)) { \
|
1969 |
CPU_FloatU __f; \ |
1970 |
__f.f = x; \ |
1971 |
__f.l = __f.l | (1 << 22); /* Set QNaN bit. */ \ |
1972 |
result = __f.f; \ |
1973 |
} else
|
1974 |
|
1975 |
#define HANDLE_NAN1(result, x) \
|
1976 |
DO_HANDLE_NAN(result, x) |
1977 |
#define HANDLE_NAN2(result, x, y) \
|
1978 |
DO_HANDLE_NAN(result, x) DO_HANDLE_NAN(result, y) |
1979 |
#define HANDLE_NAN3(result, x, y, z) \
|
1980 |
DO_HANDLE_NAN(result, x) DO_HANDLE_NAN(result, y) DO_HANDLE_NAN(result, z) |
1981 |
|
1982 |
/* Saturating arithmetic helpers. */
|
1983 |
#define SATCVT(from, to, from_type, to_type, min, max, use_min, use_max) \
|
1984 |
static always_inline to_type cvt##from##to (from_type x, int *sat) \ |
1985 |
{ \ |
1986 |
to_type r; \ |
1987 |
if (use_min && x < min) { \
|
1988 |
r = min; \ |
1989 |
*sat = 1; \
|
1990 |
} else if (use_max && x > max) { \ |
1991 |
r = max; \ |
1992 |
*sat = 1; \
|
1993 |
} else { \
|
1994 |
r = x; \ |
1995 |
} \ |
1996 |
return r; \
|
1997 |
} |
1998 |
SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX, 1, 1) |
1999 |
SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX, 1, 1) |
2000 |
SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX, 1, 1) |
2001 |
SATCVT(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX, 0, 1) |
2002 |
SATCVT(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX, 0, 1) |
2003 |
SATCVT(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX, 0, 1) |
2004 |
SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX, 1, 1) |
2005 |
SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX, 1, 1) |
2006 |
SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX, 1, 1) |
2007 |
#undef SATCVT
|
2008 |
|
2009 |
#define LVE(name, access, swap, element) \
|
2010 |
void helper_##name (ppc_avr_t *r, target_ulong addr) \ |
2011 |
{ \ |
2012 |
size_t n_elems = ARRAY_SIZE(r->element); \ |
2013 |
int adjust = HI_IDX*(n_elems-1); \ |
2014 |
int sh = sizeof(r->element[0]) >> 1; \ |
2015 |
int index = (addr & 0xf) >> sh; \ |
2016 |
if(msr_le) { \
|
2017 |
r->element[LO_IDX ? index : (adjust - index)] = swap(access(addr)); \ |
2018 |
} else { \
|
2019 |
r->element[LO_IDX ? index : (adjust - index)] = access(addr); \ |
2020 |
} \ |
2021 |
} |
2022 |
#define I(x) (x)
|
2023 |
LVE(lvebx, ldub, I, u8) |
2024 |
LVE(lvehx, lduw, bswap16, u16) |
2025 |
LVE(lvewx, ldl, bswap32, u32) |
2026 |
#undef I
|
2027 |
#undef LVE
|
2028 |
|
2029 |
void helper_lvsl (ppc_avr_t *r, target_ulong sh)
|
2030 |
{ |
2031 |
int i, j = (sh & 0xf); |
2032 |
|
2033 |
VECTOR_FOR_INORDER_I (i, u8) { |
2034 |
r->u8[i] = j++; |
2035 |
} |
2036 |
} |
2037 |
|
2038 |
void helper_lvsr (ppc_avr_t *r, target_ulong sh)
|
2039 |
{ |
2040 |
int i, j = 0x10 - (sh & 0xf); |
2041 |
|
2042 |
VECTOR_FOR_INORDER_I (i, u8) { |
2043 |
r->u8[i] = j++; |
2044 |
} |
2045 |
} |
2046 |
|
2047 |
#define STVE(name, access, swap, element) \
|
2048 |
void helper_##name (ppc_avr_t *r, target_ulong addr) \ |
2049 |
{ \ |
2050 |
size_t n_elems = ARRAY_SIZE(r->element); \ |
2051 |
int adjust = HI_IDX*(n_elems-1); \ |
2052 |
int sh = sizeof(r->element[0]) >> 1; \ |
2053 |
int index = (addr & 0xf) >> sh; \ |
2054 |
if(msr_le) { \
|
2055 |
access(addr, swap(r->element[LO_IDX ? index : (adjust - index)])); \ |
2056 |
} else { \
|
2057 |
access(addr, r->element[LO_IDX ? index : (adjust - index)]); \ |
2058 |
} \ |
2059 |
} |
2060 |
#define I(x) (x)
|
2061 |
STVE(stvebx, stb, I, u8) |
2062 |
STVE(stvehx, stw, bswap16, u16) |
2063 |
STVE(stvewx, stl, bswap32, u32) |
2064 |
#undef I
|
2065 |
#undef LVE
|
2066 |
|
2067 |
void helper_mtvscr (ppc_avr_t *r)
|
2068 |
{ |
2069 |
#if defined(WORDS_BIGENDIAN)
|
2070 |
env->vscr = r->u32[3];
|
2071 |
#else
|
2072 |
env->vscr = r->u32[0];
|
2073 |
#endif
|
2074 |
set_flush_to_zero(vscr_nj, &env->vec_status); |
2075 |
} |
2076 |
|
2077 |
void helper_vaddcuw (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2078 |
{ |
2079 |
int i;
|
2080 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { |
2081 |
r->u32[i] = ~a->u32[i] < b->u32[i]; |
2082 |
} |
2083 |
} |
2084 |
|
2085 |
#define VARITH_DO(name, op, element) \
|
2086 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2087 |
{ \ |
2088 |
int i; \
|
2089 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2090 |
r->element[i] = a->element[i] op b->element[i]; \ |
2091 |
} \ |
2092 |
} |
2093 |
#define VARITH(suffix, element) \
|
2094 |
VARITH_DO(add##suffix, +, element) \ |
2095 |
VARITH_DO(sub##suffix, -, element) |
2096 |
VARITH(ubm, u8) |
2097 |
VARITH(uhm, u16) |
2098 |
VARITH(uwm, u32) |
2099 |
#undef VARITH_DO
|
2100 |
#undef VARITH
|
2101 |
|
2102 |
#define VARITHFP(suffix, func) \
|
2103 |
void helper_v##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2104 |
{ \ |
2105 |
int i; \
|
2106 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
2107 |
HANDLE_NAN2(r->f[i], a->f[i], b->f[i]) { \ |
2108 |
r->f[i] = func(a->f[i], b->f[i], &env->vec_status); \ |
2109 |
} \ |
2110 |
} \ |
2111 |
} |
2112 |
VARITHFP(addfp, float32_add) |
2113 |
VARITHFP(subfp, float32_sub) |
2114 |
#undef VARITHFP
|
2115 |
|
2116 |
#define VARITHSAT_CASE(type, op, cvt, element) \
|
2117 |
{ \ |
2118 |
type result = (type)a->element[i] op (type)b->element[i]; \ |
2119 |
r->element[i] = cvt(result, &sat); \ |
2120 |
} |
2121 |
|
2122 |
#define VARITHSAT_DO(name, op, optype, cvt, element) \
|
2123 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2124 |
{ \ |
2125 |
int sat = 0; \ |
2126 |
int i; \
|
2127 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2128 |
switch (sizeof(r->element[0])) { \ |
2129 |
case 1: VARITHSAT_CASE(optype, op, cvt, element); break; \ |
2130 |
case 2: VARITHSAT_CASE(optype, op, cvt, element); break; \ |
2131 |
case 4: VARITHSAT_CASE(optype, op, cvt, element); break; \ |
2132 |
} \ |
2133 |
} \ |
2134 |
if (sat) { \
|
2135 |
env->vscr |= (1 << VSCR_SAT); \
|
2136 |
} \ |
2137 |
} |
2138 |
#define VARITHSAT_SIGNED(suffix, element, optype, cvt) \
|
2139 |
VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \ |
2140 |
VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element) |
2141 |
#define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \
|
2142 |
VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \ |
2143 |
VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element) |
2144 |
VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb) |
2145 |
VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh) |
2146 |
VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw) |
2147 |
VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub) |
2148 |
VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh) |
2149 |
VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw) |
2150 |
#undef VARITHSAT_CASE
|
2151 |
#undef VARITHSAT_DO
|
2152 |
#undef VARITHSAT_SIGNED
|
2153 |
#undef VARITHSAT_UNSIGNED
|
2154 |
|
2155 |
#define VAVG_DO(name, element, etype) \
|
2156 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2157 |
{ \ |
2158 |
int i; \
|
2159 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2160 |
etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \
|
2161 |
r->element[i] = x >> 1; \
|
2162 |
} \ |
2163 |
} |
2164 |
|
2165 |
#define VAVG(type, signed_element, signed_type, unsigned_element, unsigned_type) \
|
2166 |
VAVG_DO(avgs##type, signed_element, signed_type) \ |
2167 |
VAVG_DO(avgu##type, unsigned_element, unsigned_type) |
2168 |
VAVG(b, s8, int16_t, u8, uint16_t) |
2169 |
VAVG(h, s16, int32_t, u16, uint32_t) |
2170 |
VAVG(w, s32, int64_t, u32, uint64_t) |
2171 |
#undef VAVG_DO
|
2172 |
#undef VAVG
|
2173 |
|
2174 |
#define VCF(suffix, cvt, element) \
|
2175 |
void helper_vcf##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t uim) \ |
2176 |
{ \ |
2177 |
int i; \
|
2178 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
2179 |
float32 t = cvt(b->element[i], &env->vec_status); \ |
2180 |
r->f[i] = float32_scalbn (t, -uim, &env->vec_status); \ |
2181 |
} \ |
2182 |
} |
2183 |
VCF(ux, uint32_to_float32, u32) |
2184 |
VCF(sx, int32_to_float32, s32) |
2185 |
#undef VCF
|
2186 |
|
2187 |
#define VCMP_DO(suffix, compare, element, record) \
|
2188 |
void helper_vcmp##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2189 |
{ \ |
2190 |
uint32_t ones = (uint32_t)-1; \
|
2191 |
uint32_t all = ones; \ |
2192 |
uint32_t none = 0; \
|
2193 |
int i; \
|
2194 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2195 |
uint32_t result = (a->element[i] compare b->element[i] ? ones : 0x0); \
|
2196 |
switch (sizeof (a->element[0])) { \ |
2197 |
case 4: r->u32[i] = result; break; \ |
2198 |
case 2: r->u16[i] = result; break; \ |
2199 |
case 1: r->u8[i] = result; break; \ |
2200 |
} \ |
2201 |
all &= result; \ |
2202 |
none |= result; \ |
2203 |
} \ |
2204 |
if (record) { \
|
2205 |
env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ |
2206 |
} \ |
2207 |
} |
2208 |
#define VCMP(suffix, compare, element) \
|
2209 |
VCMP_DO(suffix, compare, element, 0) \
|
2210 |
VCMP_DO(suffix##_dot, compare, element, 1) |
2211 |
VCMP(equb, ==, u8) |
2212 |
VCMP(equh, ==, u16) |
2213 |
VCMP(equw, ==, u32) |
2214 |
VCMP(gtub, >, u8) |
2215 |
VCMP(gtuh, >, u16) |
2216 |
VCMP(gtuw, >, u32) |
2217 |
VCMP(gtsb, >, s8) |
2218 |
VCMP(gtsh, >, s16) |
2219 |
VCMP(gtsw, >, s32) |
2220 |
#undef VCMP_DO
|
2221 |
#undef VCMP
|
2222 |
|
2223 |
#define VCMPFP_DO(suffix, compare, order, record) \
|
2224 |
void helper_vcmp##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2225 |
{ \ |
2226 |
uint32_t ones = (uint32_t)-1; \
|
2227 |
uint32_t all = ones; \ |
2228 |
uint32_t none = 0; \
|
2229 |
int i; \
|
2230 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
2231 |
uint32_t result; \ |
2232 |
int rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status); \
|
2233 |
if (rel == float_relation_unordered) { \
|
2234 |
result = 0; \
|
2235 |
} else if (rel compare order) { \ |
2236 |
result = ones; \ |
2237 |
} else { \
|
2238 |
result = 0; \
|
2239 |
} \ |
2240 |
r->u32[i] = result; \ |
2241 |
all &= result; \ |
2242 |
none |= result; \ |
2243 |
} \ |
2244 |
if (record) { \
|
2245 |
env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \ |
2246 |
} \ |
2247 |
} |
2248 |
#define VCMPFP(suffix, compare, order) \
|
2249 |
VCMPFP_DO(suffix, compare, order, 0) \
|
2250 |
VCMPFP_DO(suffix##_dot, compare, order, 1) |
2251 |
VCMPFP(eqfp, ==, float_relation_equal) |
2252 |
VCMPFP(gefp, !=, float_relation_less) |
2253 |
VCMPFP(gtfp, ==, float_relation_greater) |
2254 |
#undef VCMPFP_DO
|
2255 |
#undef VCMPFP
|
2256 |
|
2257 |
static always_inline void vcmpbfp_internal (ppc_avr_t *r, ppc_avr_t *a, |
2258 |
ppc_avr_t *b, int record)
|
2259 |
{ |
2260 |
int i;
|
2261 |
int all_in = 0; |
2262 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
2263 |
int le_rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status);
|
2264 |
if (le_rel == float_relation_unordered) {
|
2265 |
r->u32[i] = 0xc0000000;
|
2266 |
/* ALL_IN does not need to be updated here. */
|
2267 |
} else {
|
2268 |
float32 bneg = float32_chs(b->f[i]); |
2269 |
int ge_rel = float32_compare_quiet(a->f[i], bneg, &env->vec_status);
|
2270 |
int le = le_rel != float_relation_greater;
|
2271 |
int ge = ge_rel != float_relation_less;
|
2272 |
r->u32[i] = ((!le) << 31) | ((!ge) << 30); |
2273 |
all_in |= (!le | !ge); |
2274 |
} |
2275 |
} |
2276 |
if (record) {
|
2277 |
env->crf[6] = (all_in == 0) << 1; |
2278 |
} |
2279 |
} |
2280 |
|
2281 |
void helper_vcmpbfp (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2282 |
{ |
2283 |
vcmpbfp_internal(r, a, b, 0);
|
2284 |
} |
2285 |
|
2286 |
void helper_vcmpbfp_dot (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2287 |
{ |
2288 |
vcmpbfp_internal(r, a, b, 1);
|
2289 |
} |
2290 |
|
2291 |
#define VCT(suffix, satcvt, element) \
|
2292 |
void helper_vct##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t uim) \ |
2293 |
{ \ |
2294 |
int i; \
|
2295 |
int sat = 0; \ |
2296 |
float_status s = env->vec_status; \ |
2297 |
set_float_rounding_mode(float_round_to_zero, &s); \ |
2298 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
2299 |
if (float32_is_nan(b->f[i]) || \
|
2300 |
float32_is_signaling_nan(b->f[i])) { \ |
2301 |
r->element[i] = 0; \
|
2302 |
} else { \
|
2303 |
float64 t = float32_to_float64(b->f[i], &s); \ |
2304 |
int64_t j; \ |
2305 |
t = float64_scalbn(t, uim, &s); \ |
2306 |
j = float64_to_int64(t, &s); \ |
2307 |
r->element[i] = satcvt(j, &sat); \ |
2308 |
} \ |
2309 |
} \ |
2310 |
if (sat) { \
|
2311 |
env->vscr |= (1 << VSCR_SAT); \
|
2312 |
} \ |
2313 |
} |
2314 |
VCT(uxs, cvtsduw, u32) |
2315 |
VCT(sxs, cvtsdsw, s32) |
2316 |
#undef VCT
|
2317 |
|
2318 |
void helper_vmaddfp (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2319 |
{ |
2320 |
int i;
|
2321 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
2322 |
HANDLE_NAN3(r->f[i], a->f[i], b->f[i], c->f[i]) { |
2323 |
/* Need to do the computation in higher precision and round
|
2324 |
* once at the end. */
|
2325 |
float64 af, bf, cf, t; |
2326 |
af = float32_to_float64(a->f[i], &env->vec_status); |
2327 |
bf = float32_to_float64(b->f[i], &env->vec_status); |
2328 |
cf = float32_to_float64(c->f[i], &env->vec_status); |
2329 |
t = float64_mul(af, cf, &env->vec_status); |
2330 |
t = float64_add(t, bf, &env->vec_status); |
2331 |
r->f[i] = float64_to_float32(t, &env->vec_status); |
2332 |
} |
2333 |
} |
2334 |
} |
2335 |
|
2336 |
void helper_vmhaddshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2337 |
{ |
2338 |
int sat = 0; |
2339 |
int i;
|
2340 |
|
2341 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
2342 |
int32_t prod = a->s16[i] * b->s16[i]; |
2343 |
int32_t t = (int32_t)c->s16[i] + (prod >> 15);
|
2344 |
r->s16[i] = cvtswsh (t, &sat); |
2345 |
} |
2346 |
|
2347 |
if (sat) {
|
2348 |
env->vscr |= (1 << VSCR_SAT);
|
2349 |
} |
2350 |
} |
2351 |
|
2352 |
void helper_vmhraddshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2353 |
{ |
2354 |
int sat = 0; |
2355 |
int i;
|
2356 |
|
2357 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
2358 |
int32_t prod = a->s16[i] * b->s16[i] + 0x00004000;
|
2359 |
int32_t t = (int32_t)c->s16[i] + (prod >> 15);
|
2360 |
r->s16[i] = cvtswsh (t, &sat); |
2361 |
} |
2362 |
|
2363 |
if (sat) {
|
2364 |
env->vscr |= (1 << VSCR_SAT);
|
2365 |
} |
2366 |
} |
2367 |
|
2368 |
#define VMINMAX_DO(name, compare, element) \
|
2369 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2370 |
{ \ |
2371 |
int i; \
|
2372 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2373 |
if (a->element[i] compare b->element[i]) { \
|
2374 |
r->element[i] = b->element[i]; \ |
2375 |
} else { \
|
2376 |
r->element[i] = a->element[i]; \ |
2377 |
} \ |
2378 |
} \ |
2379 |
} |
2380 |
#define VMINMAX(suffix, element) \
|
2381 |
VMINMAX_DO(min##suffix, >, element) \ |
2382 |
VMINMAX_DO(max##suffix, <, element) |
2383 |
VMINMAX(sb, s8) |
2384 |
VMINMAX(sh, s16) |
2385 |
VMINMAX(sw, s32) |
2386 |
VMINMAX(ub, u8) |
2387 |
VMINMAX(uh, u16) |
2388 |
VMINMAX(uw, u32) |
2389 |
#undef VMINMAX_DO
|
2390 |
#undef VMINMAX
|
2391 |
|
2392 |
#define VMINMAXFP(suffix, rT, rF) \
|
2393 |
void helper_v##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2394 |
{ \ |
2395 |
int i; \
|
2396 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
2397 |
HANDLE_NAN2(r->f[i], a->f[i], b->f[i]) { \ |
2398 |
if (float32_lt_quiet(a->f[i], b->f[i], &env->vec_status)) { \
|
2399 |
r->f[i] = rT->f[i]; \ |
2400 |
} else { \
|
2401 |
r->f[i] = rF->f[i]; \ |
2402 |
} \ |
2403 |
} \ |
2404 |
} \ |
2405 |
} |
2406 |
VMINMAXFP(minfp, a, b) |
2407 |
VMINMAXFP(maxfp, b, a) |
2408 |
#undef VMINMAXFP
|
2409 |
|
2410 |
void helper_vmladduhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2411 |
{ |
2412 |
int i;
|
2413 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
2414 |
int32_t prod = a->s16[i] * b->s16[i]; |
2415 |
r->s16[i] = (int16_t) (prod + c->s16[i]); |
2416 |
} |
2417 |
} |
2418 |
|
2419 |
#define VMRG_DO(name, element, highp) \
|
2420 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2421 |
{ \ |
2422 |
ppc_avr_t result; \ |
2423 |
int i; \
|
2424 |
size_t n_elems = ARRAY_SIZE(r->element); \ |
2425 |
for (i = 0; i < n_elems/2; i++) { \ |
2426 |
if (highp) { \
|
2427 |
result.element[i*2+HI_IDX] = a->element[i]; \
|
2428 |
result.element[i*2+LO_IDX] = b->element[i]; \
|
2429 |
} else { \
|
2430 |
result.element[n_elems - i*2 - (1+HI_IDX)] = b->element[n_elems - i - 1]; \ |
2431 |
result.element[n_elems - i*2 - (1+LO_IDX)] = a->element[n_elems - i - 1]; \ |
2432 |
} \ |
2433 |
} \ |
2434 |
*r = result; \ |
2435 |
} |
2436 |
#if defined(WORDS_BIGENDIAN)
|
2437 |
#define MRGHI 0 |
2438 |
#define MRGLO 1 |
2439 |
#else
|
2440 |
#define MRGHI 1 |
2441 |
#define MRGLO 0 |
2442 |
#endif
|
2443 |
#define VMRG(suffix, element) \
|
2444 |
VMRG_DO(mrgl##suffix, element, MRGHI) \ |
2445 |
VMRG_DO(mrgh##suffix, element, MRGLO) |
2446 |
VMRG(b, u8) |
2447 |
VMRG(h, u16) |
2448 |
VMRG(w, u32) |
2449 |
#undef VMRG_DO
|
2450 |
#undef VMRG
|
2451 |
#undef MRGHI
|
2452 |
#undef MRGLO
|
2453 |
|
2454 |
void helper_vmsummbm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2455 |
{ |
2456 |
int32_t prod[16];
|
2457 |
int i;
|
2458 |
|
2459 |
for (i = 0; i < ARRAY_SIZE(r->s8); i++) { |
2460 |
prod[i] = (int32_t)a->s8[i] * b->u8[i]; |
2461 |
} |
2462 |
|
2463 |
VECTOR_FOR_INORDER_I(i, s32) { |
2464 |
r->s32[i] = c->s32[i] + prod[4*i] + prod[4*i+1] + prod[4*i+2] + prod[4*i+3]; |
2465 |
} |
2466 |
} |
2467 |
|
2468 |
void helper_vmsumshm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2469 |
{ |
2470 |
int32_t prod[8];
|
2471 |
int i;
|
2472 |
|
2473 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
2474 |
prod[i] = a->s16[i] * b->s16[i]; |
2475 |
} |
2476 |
|
2477 |
VECTOR_FOR_INORDER_I(i, s32) { |
2478 |
r->s32[i] = c->s32[i] + prod[2*i] + prod[2*i+1]; |
2479 |
} |
2480 |
} |
2481 |
|
2482 |
void helper_vmsumshs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2483 |
{ |
2484 |
int32_t prod[8];
|
2485 |
int i;
|
2486 |
int sat = 0; |
2487 |
|
2488 |
for (i = 0; i < ARRAY_SIZE(r->s16); i++) { |
2489 |
prod[i] = (int32_t)a->s16[i] * b->s16[i]; |
2490 |
} |
2491 |
|
2492 |
VECTOR_FOR_INORDER_I (i, s32) { |
2493 |
int64_t t = (int64_t)c->s32[i] + prod[2*i] + prod[2*i+1]; |
2494 |
r->u32[i] = cvtsdsw(t, &sat); |
2495 |
} |
2496 |
|
2497 |
if (sat) {
|
2498 |
env->vscr |= (1 << VSCR_SAT);
|
2499 |
} |
2500 |
} |
2501 |
|
2502 |
void helper_vmsumubm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2503 |
{ |
2504 |
uint16_t prod[16];
|
2505 |
int i;
|
2506 |
|
2507 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { |
2508 |
prod[i] = a->u8[i] * b->u8[i]; |
2509 |
} |
2510 |
|
2511 |
VECTOR_FOR_INORDER_I(i, u32) { |
2512 |
r->u32[i] = c->u32[i] + prod[4*i] + prod[4*i+1] + prod[4*i+2] + prod[4*i+3]; |
2513 |
} |
2514 |
} |
2515 |
|
2516 |
void helper_vmsumuhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2517 |
{ |
2518 |
uint32_t prod[8];
|
2519 |
int i;
|
2520 |
|
2521 |
for (i = 0; i < ARRAY_SIZE(r->u16); i++) { |
2522 |
prod[i] = a->u16[i] * b->u16[i]; |
2523 |
} |
2524 |
|
2525 |
VECTOR_FOR_INORDER_I(i, u32) { |
2526 |
r->u32[i] = c->u32[i] + prod[2*i] + prod[2*i+1]; |
2527 |
} |
2528 |
} |
2529 |
|
2530 |
void helper_vmsumuhs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2531 |
{ |
2532 |
uint32_t prod[8];
|
2533 |
int i;
|
2534 |
int sat = 0; |
2535 |
|
2536 |
for (i = 0; i < ARRAY_SIZE(r->u16); i++) { |
2537 |
prod[i] = a->u16[i] * b->u16[i]; |
2538 |
} |
2539 |
|
2540 |
VECTOR_FOR_INORDER_I (i, s32) { |
2541 |
uint64_t t = (uint64_t)c->u32[i] + prod[2*i] + prod[2*i+1]; |
2542 |
r->u32[i] = cvtuduw(t, &sat); |
2543 |
} |
2544 |
|
2545 |
if (sat) {
|
2546 |
env->vscr |= (1 << VSCR_SAT);
|
2547 |
} |
2548 |
} |
2549 |
|
2550 |
#define VMUL_DO(name, mul_element, prod_element, evenp) \
|
2551 |
void helper_v##name (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2552 |
{ \ |
2553 |
int i; \
|
2554 |
VECTOR_FOR_INORDER_I(i, prod_element) { \ |
2555 |
if (evenp) { \
|
2556 |
r->prod_element[i] = a->mul_element[i*2+HI_IDX] * b->mul_element[i*2+HI_IDX]; \ |
2557 |
} else { \
|
2558 |
r->prod_element[i] = a->mul_element[i*2+LO_IDX] * b->mul_element[i*2+LO_IDX]; \ |
2559 |
} \ |
2560 |
} \ |
2561 |
} |
2562 |
#define VMUL(suffix, mul_element, prod_element) \
|
2563 |
VMUL_DO(mule##suffix, mul_element, prod_element, 1) \ |
2564 |
VMUL_DO(mulo##suffix, mul_element, prod_element, 0) |
2565 |
VMUL(sb, s8, s16) |
2566 |
VMUL(sh, s16, s32) |
2567 |
VMUL(ub, u8, u16) |
2568 |
VMUL(uh, u16, u32) |
2569 |
#undef VMUL_DO
|
2570 |
#undef VMUL
|
2571 |
|
2572 |
void helper_vnmsubfp (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2573 |
{ |
2574 |
int i;
|
2575 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
2576 |
HANDLE_NAN3(r->f[i], a->f[i], b->f[i], c->f[i]) { |
2577 |
/* Need to do the computation is higher precision and round
|
2578 |
* once at the end. */
|
2579 |
float64 af, bf, cf, t; |
2580 |
af = float32_to_float64(a->f[i], &env->vec_status); |
2581 |
bf = float32_to_float64(b->f[i], &env->vec_status); |
2582 |
cf = float32_to_float64(c->f[i], &env->vec_status); |
2583 |
t = float64_mul(af, cf, &env->vec_status); |
2584 |
t = float64_sub(t, bf, &env->vec_status); |
2585 |
t = float64_chs(t); |
2586 |
r->f[i] = float64_to_float32(t, &env->vec_status); |
2587 |
} |
2588 |
} |
2589 |
} |
2590 |
|
2591 |
void helper_vperm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2592 |
{ |
2593 |
ppc_avr_t result; |
2594 |
int i;
|
2595 |
VECTOR_FOR_INORDER_I (i, u8) { |
2596 |
int s = c->u8[i] & 0x1f; |
2597 |
#if defined(WORDS_BIGENDIAN)
|
2598 |
int index = s & 0xf; |
2599 |
#else
|
2600 |
int index = 15 - (s & 0xf); |
2601 |
#endif
|
2602 |
if (s & 0x10) { |
2603 |
result.u8[i] = b->u8[index]; |
2604 |
} else {
|
2605 |
result.u8[i] = a->u8[index]; |
2606 |
} |
2607 |
} |
2608 |
*r = result; |
2609 |
} |
2610 |
|
2611 |
#if defined(WORDS_BIGENDIAN)
|
2612 |
#define PKBIG 1 |
2613 |
#else
|
2614 |
#define PKBIG 0 |
2615 |
#endif
|
2616 |
void helper_vpkpx (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2617 |
{ |
2618 |
int i, j;
|
2619 |
ppc_avr_t result; |
2620 |
#if defined(WORDS_BIGENDIAN)
|
2621 |
const ppc_avr_t *x[2] = { a, b }; |
2622 |
#else
|
2623 |
const ppc_avr_t *x[2] = { b, a }; |
2624 |
#endif
|
2625 |
|
2626 |
VECTOR_FOR_INORDER_I (i, u64) { |
2627 |
VECTOR_FOR_INORDER_I (j, u32){ |
2628 |
uint32_t e = x[i]->u32[j]; |
2629 |
result.u16[4*i+j] = (((e >> 9) & 0xfc00) | |
2630 |
((e >> 6) & 0x3e0) | |
2631 |
((e >> 3) & 0x1f)); |
2632 |
} |
2633 |
} |
2634 |
*r = result; |
2635 |
} |
2636 |
|
2637 |
#define VPK(suffix, from, to, cvt, dosat) \
|
2638 |
void helper_vpk##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2639 |
{ \ |
2640 |
int i; \
|
2641 |
int sat = 0; \ |
2642 |
ppc_avr_t result; \ |
2643 |
ppc_avr_t *a0 = PKBIG ? a : b; \ |
2644 |
ppc_avr_t *a1 = PKBIG ? b : a; \ |
2645 |
VECTOR_FOR_INORDER_I (i, from) { \ |
2646 |
result.to[i] = cvt(a0->from[i], &sat); \ |
2647 |
result.to[i+ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat); \ |
2648 |
} \ |
2649 |
*r = result; \ |
2650 |
if (dosat && sat) { \
|
2651 |
env->vscr |= (1 << VSCR_SAT); \
|
2652 |
} \ |
2653 |
} |
2654 |
#define I(x, y) (x)
|
2655 |
VPK(shss, s16, s8, cvtshsb, 1)
|
2656 |
VPK(shus, s16, u8, cvtshub, 1)
|
2657 |
VPK(swss, s32, s16, cvtswsh, 1)
|
2658 |
VPK(swus, s32, u16, cvtswuh, 1)
|
2659 |
VPK(uhus, u16, u8, cvtuhub, 1)
|
2660 |
VPK(uwus, u32, u16, cvtuwuh, 1)
|
2661 |
VPK(uhum, u16, u8, I, 0)
|
2662 |
VPK(uwum, u32, u16, I, 0)
|
2663 |
#undef I
|
2664 |
#undef VPK
|
2665 |
#undef PKBIG
|
2666 |
|
2667 |
void helper_vrefp (ppc_avr_t *r, ppc_avr_t *b)
|
2668 |
{ |
2669 |
int i;
|
2670 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
2671 |
HANDLE_NAN1(r->f[i], b->f[i]) { |
2672 |
r->f[i] = float32_div(float32_one, b->f[i], &env->vec_status); |
2673 |
} |
2674 |
} |
2675 |
} |
2676 |
|
2677 |
#define VRFI(suffix, rounding) \
|
2678 |
void helper_vrfi##suffix (ppc_avr_t *r, ppc_avr_t *b) \ |
2679 |
{ \ |
2680 |
int i; \
|
2681 |
float_status s = env->vec_status; \ |
2682 |
set_float_rounding_mode(rounding, &s); \ |
2683 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { \ |
2684 |
HANDLE_NAN1(r->f[i], b->f[i]) { \ |
2685 |
r->f[i] = float32_round_to_int (b->f[i], &s); \ |
2686 |
} \ |
2687 |
} \ |
2688 |
} |
2689 |
VRFI(n, float_round_nearest_even) |
2690 |
VRFI(m, float_round_down) |
2691 |
VRFI(p, float_round_up) |
2692 |
VRFI(z, float_round_to_zero) |
2693 |
#undef VRFI
|
2694 |
|
2695 |
#define VROTATE(suffix, element) \
|
2696 |
void helper_vrl##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2697 |
{ \ |
2698 |
int i; \
|
2699 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2700 |
unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \ |
2701 |
unsigned int shift = b->element[i] & mask; \ |
2702 |
r->element[i] = (a->element[i] << shift) | (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \ |
2703 |
} \ |
2704 |
} |
2705 |
VROTATE(b, u8) |
2706 |
VROTATE(h, u16) |
2707 |
VROTATE(w, u32) |
2708 |
#undef VROTATE
|
2709 |
|
2710 |
void helper_vrsqrtefp (ppc_avr_t *r, ppc_avr_t *b)
|
2711 |
{ |
2712 |
int i;
|
2713 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
2714 |
HANDLE_NAN1(r->f[i], b->f[i]) { |
2715 |
float32 t = float32_sqrt(b->f[i], &env->vec_status); |
2716 |
r->f[i] = float32_div(float32_one, t, &env->vec_status); |
2717 |
} |
2718 |
} |
2719 |
} |
2720 |
|
2721 |
void helper_vsel (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
|
2722 |
{ |
2723 |
r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]); |
2724 |
r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]); |
2725 |
} |
2726 |
|
2727 |
void helper_vlogefp (ppc_avr_t *r, ppc_avr_t *b)
|
2728 |
{ |
2729 |
int i;
|
2730 |
for (i = 0; i < ARRAY_SIZE(r->f); i++) { |
2731 |
HANDLE_NAN1(r->f[i], b->f[i]) { |
2732 |
r->f[i] = float32_log2(b->f[i], &env->vec_status); |
2733 |
} |
2734 |
} |
2735 |
} |
2736 |
|
2737 |
#if defined(WORDS_BIGENDIAN)
|
2738 |
#define LEFT 0 |
2739 |
#define RIGHT 1 |
2740 |
#else
|
2741 |
#define LEFT 1 |
2742 |
#define RIGHT 0 |
2743 |
#endif
|
2744 |
/* The specification says that the results are undefined if all of the
|
2745 |
* shift counts are not identical. We check to make sure that they are
|
2746 |
* to conform to what real hardware appears to do. */
|
2747 |
#define VSHIFT(suffix, leftp) \
|
2748 |
void helper_vs##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2749 |
{ \ |
2750 |
int shift = b->u8[LO_IDX*0x15] & 0x7; \ |
2751 |
int doit = 1; \ |
2752 |
int i; \
|
2753 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { \ |
2754 |
doit = doit && ((b->u8[i] & 0x7) == shift); \
|
2755 |
} \ |
2756 |
if (doit) { \
|
2757 |
if (shift == 0) { \ |
2758 |
*r = *a; \ |
2759 |
} else if (leftp) { \ |
2760 |
uint64_t carry = a->u64[LO_IDX] >> (64 - shift); \
|
2761 |
r->u64[HI_IDX] = (a->u64[HI_IDX] << shift) | carry; \ |
2762 |
r->u64[LO_IDX] = a->u64[LO_IDX] << shift; \ |
2763 |
} else { \
|
2764 |
uint64_t carry = a->u64[HI_IDX] << (64 - shift); \
|
2765 |
r->u64[LO_IDX] = (a->u64[LO_IDX] >> shift) | carry; \ |
2766 |
r->u64[HI_IDX] = a->u64[HI_IDX] >> shift; \ |
2767 |
} \ |
2768 |
} \ |
2769 |
} |
2770 |
VSHIFT(l, LEFT) |
2771 |
VSHIFT(r, RIGHT) |
2772 |
#undef VSHIFT
|
2773 |
#undef LEFT
|
2774 |
#undef RIGHT
|
2775 |
|
2776 |
#define VSL(suffix, element) \
|
2777 |
void helper_vsl##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2778 |
{ \ |
2779 |
int i; \
|
2780 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2781 |
unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \ |
2782 |
unsigned int shift = b->element[i] & mask; \ |
2783 |
r->element[i] = a->element[i] << shift; \ |
2784 |
} \ |
2785 |
} |
2786 |
VSL(b, u8) |
2787 |
VSL(h, u16) |
2788 |
VSL(w, u32) |
2789 |
#undef VSL
|
2790 |
|
2791 |
void helper_vsldoi (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift)
|
2792 |
{ |
2793 |
int sh = shift & 0xf; |
2794 |
int i;
|
2795 |
ppc_avr_t result; |
2796 |
|
2797 |
#if defined(WORDS_BIGENDIAN)
|
2798 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { |
2799 |
int index = sh + i;
|
2800 |
if (index > 0xf) { |
2801 |
result.u8[i] = b->u8[index-0x10];
|
2802 |
} else {
|
2803 |
result.u8[i] = a->u8[index]; |
2804 |
} |
2805 |
} |
2806 |
#else
|
2807 |
for (i = 0; i < ARRAY_SIZE(r->u8); i++) { |
2808 |
int index = (16 - sh) + i; |
2809 |
if (index > 0xf) { |
2810 |
result.u8[i] = a->u8[index-0x10];
|
2811 |
} else {
|
2812 |
result.u8[i] = b->u8[index]; |
2813 |
} |
2814 |
} |
2815 |
#endif
|
2816 |
*r = result; |
2817 |
} |
2818 |
|
2819 |
void helper_vslo (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2820 |
{ |
2821 |
int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf; |
2822 |
|
2823 |
#if defined (WORDS_BIGENDIAN)
|
2824 |
memmove (&r->u8[0], &a->u8[sh], 16-sh); |
2825 |
memset (&r->u8[16-sh], 0, sh); |
2826 |
#else
|
2827 |
memmove (&r->u8[sh], &a->u8[0], 16-sh); |
2828 |
memset (&r->u8[0], 0, sh); |
2829 |
#endif
|
2830 |
} |
2831 |
|
2832 |
/* Experimental testing shows that hardware masks the immediate. */
|
2833 |
#define _SPLAT_MASKED(element) (splat & (ARRAY_SIZE(r->element) - 1)) |
2834 |
#if defined(WORDS_BIGENDIAN)
|
2835 |
#define SPLAT_ELEMENT(element) _SPLAT_MASKED(element)
|
2836 |
#else
|
2837 |
#define SPLAT_ELEMENT(element) (ARRAY_SIZE(r->element)-1 - _SPLAT_MASKED(element)) |
2838 |
#endif
|
2839 |
#define VSPLT(suffix, element) \
|
2840 |
void helper_vsplt##suffix (ppc_avr_t *r, ppc_avr_t *b, uint32_t splat) \ |
2841 |
{ \ |
2842 |
uint32_t s = b->element[SPLAT_ELEMENT(element)]; \ |
2843 |
int i; \
|
2844 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2845 |
r->element[i] = s; \ |
2846 |
} \ |
2847 |
} |
2848 |
VSPLT(b, u8) |
2849 |
VSPLT(h, u16) |
2850 |
VSPLT(w, u32) |
2851 |
#undef VSPLT
|
2852 |
#undef SPLAT_ELEMENT
|
2853 |
#undef _SPLAT_MASKED
|
2854 |
|
2855 |
#define VSPLTI(suffix, element, splat_type) \
|
2856 |
void helper_vspltis##suffix (ppc_avr_t *r, uint32_t splat) \ |
2857 |
{ \ |
2858 |
splat_type x = (int8_t)(splat << 3) >> 3; \ |
2859 |
int i; \
|
2860 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2861 |
r->element[i] = x; \ |
2862 |
} \ |
2863 |
} |
2864 |
VSPLTI(b, s8, int8_t) |
2865 |
VSPLTI(h, s16, int16_t) |
2866 |
VSPLTI(w, s32, int32_t) |
2867 |
#undef VSPLTI
|
2868 |
|
2869 |
#define VSR(suffix, element) \
|
2870 |
void helper_vsr##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \ |
2871 |
{ \ |
2872 |
int i; \
|
2873 |
for (i = 0; i < ARRAY_SIZE(r->element); i++) { \ |
2874 |
unsigned int mask = ((1 << (3 + (sizeof (a->element[0]) >> 1))) - 1); \ |
2875 |
unsigned int shift = b->element[i] & mask; \ |
2876 |
r->element[i] = a->element[i] >> shift; \ |
2877 |
} \ |
2878 |
} |
2879 |
VSR(ab, s8) |
2880 |
VSR(ah, s16) |
2881 |
VSR(aw, s32) |
2882 |
VSR(b, u8) |
2883 |
VSR(h, u16) |
2884 |
VSR(w, u32) |
2885 |
#undef VSR
|
2886 |
|
2887 |
void helper_vsro (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2888 |
{ |
2889 |
int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf; |
2890 |
|
2891 |
#if defined (WORDS_BIGENDIAN)
|
2892 |
memmove (&r->u8[sh], &a->u8[0], 16-sh); |
2893 |
memset (&r->u8[0], 0, sh); |
2894 |
#else
|
2895 |
memmove (&r->u8[0], &a->u8[sh], 16-sh); |
2896 |
memset (&r->u8[16-sh], 0, sh); |
2897 |
#endif
|
2898 |
} |
2899 |
|
2900 |
void helper_vsubcuw (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2901 |
{ |
2902 |
int i;
|
2903 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { |
2904 |
r->u32[i] = a->u32[i] >= b->u32[i]; |
2905 |
} |
2906 |
} |
2907 |
|
2908 |
void helper_vsumsws (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2909 |
{ |
2910 |
int64_t t; |
2911 |
int i, upper;
|
2912 |
ppc_avr_t result; |
2913 |
int sat = 0; |
2914 |
|
2915 |
#if defined(WORDS_BIGENDIAN)
|
2916 |
upper = ARRAY_SIZE(r->s32)-1;
|
2917 |
#else
|
2918 |
upper = 0;
|
2919 |
#endif
|
2920 |
t = (int64_t)b->s32[upper]; |
2921 |
for (i = 0; i < ARRAY_SIZE(r->s32); i++) { |
2922 |
t += a->s32[i]; |
2923 |
result.s32[i] = 0;
|
2924 |
} |
2925 |
result.s32[upper] = cvtsdsw(t, &sat); |
2926 |
*r = result; |
2927 |
|
2928 |
if (sat) {
|
2929 |
env->vscr |= (1 << VSCR_SAT);
|
2930 |
} |
2931 |
} |
2932 |
|
2933 |
void helper_vsum2sws (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2934 |
{ |
2935 |
int i, j, upper;
|
2936 |
ppc_avr_t result; |
2937 |
int sat = 0; |
2938 |
|
2939 |
#if defined(WORDS_BIGENDIAN)
|
2940 |
upper = 1;
|
2941 |
#else
|
2942 |
upper = 0;
|
2943 |
#endif
|
2944 |
for (i = 0; i < ARRAY_SIZE(r->u64); i++) { |
2945 |
int64_t t = (int64_t)b->s32[upper+i*2];
|
2946 |
result.u64[i] = 0;
|
2947 |
for (j = 0; j < ARRAY_SIZE(r->u64); j++) { |
2948 |
t += a->s32[2*i+j];
|
2949 |
} |
2950 |
result.s32[upper+i*2] = cvtsdsw(t, &sat);
|
2951 |
} |
2952 |
|
2953 |
*r = result; |
2954 |
if (sat) {
|
2955 |
env->vscr |= (1 << VSCR_SAT);
|
2956 |
} |
2957 |
} |
2958 |
|
2959 |
void helper_vsum4sbs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2960 |
{ |
2961 |
int i, j;
|
2962 |
int sat = 0; |
2963 |
|
2964 |
for (i = 0; i < ARRAY_SIZE(r->s32); i++) { |
2965 |
int64_t t = (int64_t)b->s32[i]; |
2966 |
for (j = 0; j < ARRAY_SIZE(r->s32); j++) { |
2967 |
t += a->s8[4*i+j];
|
2968 |
} |
2969 |
r->s32[i] = cvtsdsw(t, &sat); |
2970 |
} |
2971 |
|
2972 |
if (sat) {
|
2973 |
env->vscr |= (1 << VSCR_SAT);
|
2974 |
} |
2975 |
} |
2976 |
|
2977 |
void helper_vsum4shs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2978 |
{ |
2979 |
int sat = 0; |
2980 |
int i;
|
2981 |
|
2982 |
for (i = 0; i < ARRAY_SIZE(r->s32); i++) { |
2983 |
int64_t t = (int64_t)b->s32[i]; |
2984 |
t += a->s16[2*i] + a->s16[2*i+1]; |
2985 |
r->s32[i] = cvtsdsw(t, &sat); |
2986 |
} |
2987 |
|
2988 |
if (sat) {
|
2989 |
env->vscr |= (1 << VSCR_SAT);
|
2990 |
} |
2991 |
} |
2992 |
|
2993 |
void helper_vsum4ubs (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
|
2994 |
{ |
2995 |
int i, j;
|
2996 |
int sat = 0; |
2997 |
|
2998 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { |
2999 |
uint64_t t = (uint64_t)b->u32[i]; |
3000 |
for (j = 0; j < ARRAY_SIZE(r->u32); j++) { |
3001 |
t += a->u8[4*i+j];
|
3002 |
} |
3003 |
r->u32[i] = cvtuduw(t, &sat); |
3004 |
} |
3005 |
|
3006 |
if (sat) {
|
3007 |
env->vscr |= (1 << VSCR_SAT);
|
3008 |
} |
3009 |
} |
3010 |
|
3011 |
#if defined(WORDS_BIGENDIAN)
|
3012 |
#define UPKHI 1 |
3013 |
#define UPKLO 0 |
3014 |
#else
|
3015 |
#define UPKHI 0 |
3016 |
#define UPKLO 1 |
3017 |
#endif
|
3018 |
#define VUPKPX(suffix, hi) \
|
3019 |
void helper_vupk##suffix (ppc_avr_t *r, ppc_avr_t *b) \ |
3020 |
{ \ |
3021 |
int i; \
|
3022 |
ppc_avr_t result; \ |
3023 |
for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \ |
3024 |
uint16_t e = b->u16[hi ? i : i+4]; \
|
3025 |
uint8_t a = (e >> 15) ? 0xff : 0; \ |
3026 |
uint8_t r = (e >> 10) & 0x1f; \ |
3027 |
uint8_t g = (e >> 5) & 0x1f; \ |
3028 |
uint8_t b = e & 0x1f; \
|
3029 |
result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \ |
3030 |
} \ |
3031 |
*r = result; \ |
3032 |
} |
3033 |
VUPKPX(lpx, UPKLO) |
3034 |
VUPKPX(hpx, UPKHI) |
3035 |
#undef VUPKPX
|
3036 |
|
3037 |
#define VUPK(suffix, unpacked, packee, hi) \
|
3038 |
void helper_vupk##suffix (ppc_avr_t *r, ppc_avr_t *b) \ |
3039 |
{ \ |
3040 |
int i; \
|
3041 |
ppc_avr_t result; \ |
3042 |
if (hi) { \
|
3043 |
for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \ |
3044 |
result.unpacked[i] = b->packee[i]; \ |
3045 |
} \ |
3046 |
} else { \
|
3047 |
for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); i++) { \
|
3048 |
result.unpacked[i-ARRAY_SIZE(r->unpacked)] = b->packee[i]; \ |
3049 |
} \ |
3050 |
} \ |
3051 |
*r = result; \ |
3052 |
} |
3053 |
VUPK(hsb, s16, s8, UPKHI) |
3054 |
VUPK(hsh, s32, s16, UPKHI) |
3055 |
VUPK(lsb, s16, s8, UPKLO) |
3056 |
VUPK(lsh, s32, s16, UPKLO) |
3057 |
#undef VUPK
|
3058 |
#undef UPKHI
|
3059 |
#undef UPKLO
|
3060 |
|
3061 |
#undef DO_HANDLE_NAN
|
3062 |
#undef HANDLE_NAN1
|
3063 |
#undef HANDLE_NAN2
|
3064 |
#undef HANDLE_NAN3
|
3065 |
#undef VECTOR_FOR_INORDER_I
|
3066 |
#undef HI_IDX
|
3067 |
#undef LO_IDX
|
3068 |
|
3069 |
/*****************************************************************************/
|
3070 |
/* SPE extension helpers */
|
3071 |
/* Use a table to make this quicker */
|
3072 |
static uint8_t hbrev[16] = { |
3073 |
0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE, |
3074 |
0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF, |
3075 |
}; |
3076 |
|
3077 |
static always_inline uint8_t byte_reverse (uint8_t val)
|
3078 |
{ |
3079 |
return hbrev[val >> 4] | (hbrev[val & 0xF] << 4); |
3080 |
} |
3081 |
|
3082 |
static always_inline uint32_t word_reverse (uint32_t val)
|
3083 |
{ |
3084 |
return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) | |
3085 |
(byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24); |
3086 |
} |
3087 |
|
3088 |
#define MASKBITS 16 // Random value - to be fixed (implementation dependant) |
3089 |
target_ulong helper_brinc (target_ulong arg1, target_ulong arg2) |
3090 |
{ |
3091 |
uint32_t a, b, d, mask; |
3092 |
|
3093 |
mask = UINT32_MAX >> (32 - MASKBITS);
|
3094 |
a = arg1 & mask; |
3095 |
b = arg2 & mask; |
3096 |
d = word_reverse(1 + word_reverse(a | ~b));
|
3097 |
return (arg1 & ~mask) | (d & b);
|
3098 |
} |
3099 |
|
3100 |
uint32_t helper_cntlsw32 (uint32_t val) |
3101 |
{ |
3102 |
if (val & 0x80000000) |
3103 |
return clz32(~val);
|
3104 |
else
|
3105 |
return clz32(val);
|
3106 |
} |
3107 |
|
3108 |
uint32_t helper_cntlzw32 (uint32_t val) |
3109 |
{ |
3110 |
return clz32(val);
|
3111 |
} |
3112 |
|
3113 |
/* Single-precision floating-point conversions */
|
3114 |
static always_inline uint32_t efscfsi (uint32_t val)
|
3115 |
{ |
3116 |
CPU_FloatU u; |
3117 |
|
3118 |
u.f = int32_to_float32(val, &env->vec_status); |
3119 |
|
3120 |
return u.l;
|
3121 |
} |
3122 |
|
3123 |
static always_inline uint32_t efscfui (uint32_t val)
|
3124 |
{ |
3125 |
CPU_FloatU u; |
3126 |
|
3127 |
u.f = uint32_to_float32(val, &env->vec_status); |
3128 |
|
3129 |
return u.l;
|
3130 |
} |
3131 |
|
3132 |
static always_inline int32_t efsctsi (uint32_t val)
|
3133 |
{ |
3134 |
CPU_FloatU u; |
3135 |
|
3136 |
u.l = val; |
3137 |
/* NaN are not treated the same way IEEE 754 does */
|
3138 |
if (unlikely(float32_is_nan(u.f)))
|
3139 |
return 0; |
3140 |
|
3141 |
return float32_to_int32(u.f, &env->vec_status);
|
3142 |
} |
3143 |
|
3144 |
static always_inline uint32_t efsctui (uint32_t val)
|
3145 |
{ |
3146 |
CPU_FloatU u; |
3147 |
|
3148 |
u.l = val; |
3149 |
/* NaN are not treated the same way IEEE 754 does */
|
3150 |
if (unlikely(float32_is_nan(u.f)))
|
3151 |
return 0; |
3152 |
|
3153 |
return float32_to_uint32(u.f, &env->vec_status);
|
3154 |
} |
3155 |
|
3156 |
static always_inline uint32_t efsctsiz (uint32_t val)
|
3157 |
{ |
3158 |
CPU_FloatU u; |
3159 |
|
3160 |
u.l = val; |
3161 |
/* NaN are not treated the same way IEEE 754 does */
|
3162 |
if (unlikely(float32_is_nan(u.f)))
|
3163 |
return 0; |
3164 |
|
3165 |
return float32_to_int32_round_to_zero(u.f, &env->vec_status);
|
3166 |
} |
3167 |
|
3168 |
static always_inline uint32_t efsctuiz (uint32_t val)
|
3169 |
{ |
3170 |
CPU_FloatU u; |
3171 |
|
3172 |
u.l = val; |
3173 |
/* NaN are not treated the same way IEEE 754 does */
|
3174 |
if (unlikely(float32_is_nan(u.f)))
|
3175 |
return 0; |
3176 |
|
3177 |
return float32_to_uint32_round_to_zero(u.f, &env->vec_status);
|
3178 |
} |
3179 |
|
3180 |
static always_inline uint32_t efscfsf (uint32_t val)
|
3181 |
{ |
3182 |
CPU_FloatU u; |
3183 |
float32 tmp; |
3184 |
|
3185 |
u.f = int32_to_float32(val, &env->vec_status); |
3186 |
tmp = int64_to_float32(1ULL << 32, &env->vec_status); |
3187 |
u.f = float32_div(u.f, tmp, &env->vec_status); |
3188 |
|
3189 |
return u.l;
|
3190 |
} |
3191 |
|
3192 |
static always_inline uint32_t efscfuf (uint32_t val)
|
3193 |
{ |
3194 |
CPU_FloatU u; |
3195 |
float32 tmp; |
3196 |
|
3197 |
u.f = uint32_to_float32(val, &env->vec_status); |
3198 |
tmp = uint64_to_float32(1ULL << 32, &env->vec_status); |
3199 |
u.f = float32_div(u.f, tmp, &env->vec_status); |
3200 |
|
3201 |
return u.l;
|
3202 |
} |
3203 |
|
3204 |
static always_inline uint32_t efsctsf (uint32_t val)
|
3205 |
{ |
3206 |
CPU_FloatU u; |
3207 |
float32 tmp; |
3208 |
|
3209 |
u.l = val; |
3210 |
/* NaN are not treated the same way IEEE 754 does */
|
3211 |
if (unlikely(float32_is_nan(u.f)))
|
3212 |
return 0; |
3213 |
tmp = uint64_to_float32(1ULL << 32, &env->vec_status); |
3214 |
u.f = float32_mul(u.f, tmp, &env->vec_status); |
3215 |
|
3216 |
return float32_to_int32(u.f, &env->vec_status);
|
3217 |
} |
3218 |
|
3219 |
static always_inline uint32_t efsctuf (uint32_t val)
|
3220 |
{ |
3221 |
CPU_FloatU u; |
3222 |
float32 tmp; |
3223 |
|
3224 |
u.l = val; |
3225 |
/* NaN are not treated the same way IEEE 754 does */
|
3226 |
if (unlikely(float32_is_nan(u.f)))
|
3227 |
return 0; |
3228 |
tmp = uint64_to_float32(1ULL << 32, &env->vec_status); |
3229 |
u.f = float32_mul(u.f, tmp, &env->vec_status); |
3230 |
|
3231 |
return float32_to_uint32(u.f, &env->vec_status);
|
3232 |
} |
3233 |
|
3234 |
#define HELPER_SPE_SINGLE_CONV(name) \
|
3235 |
uint32_t helper_e##name (uint32_t val) \ |
3236 |
{ \ |
3237 |
return e##name(val); \ |
3238 |
} |
3239 |
/* efscfsi */
|
3240 |
HELPER_SPE_SINGLE_CONV(fscfsi); |
3241 |
/* efscfui */
|
3242 |
HELPER_SPE_SINGLE_CONV(fscfui); |
3243 |
/* efscfuf */
|
3244 |
HELPER_SPE_SINGLE_CONV(fscfuf); |
3245 |
/* efscfsf */
|
3246 |
HELPER_SPE_SINGLE_CONV(fscfsf); |
3247 |
/* efsctsi */
|
3248 |
HELPER_SPE_SINGLE_CONV(fsctsi); |
3249 |
/* efsctui */
|
3250 |
HELPER_SPE_SINGLE_CONV(fsctui); |
3251 |
/* efsctsiz */
|
3252 |
HELPER_SPE_SINGLE_CONV(fsctsiz); |
3253 |
/* efsctuiz */
|
3254 |
HELPER_SPE_SINGLE_CONV(fsctuiz); |
3255 |
/* efsctsf */
|
3256 |
HELPER_SPE_SINGLE_CONV(fsctsf); |
3257 |
/* efsctuf */
|
3258 |
HELPER_SPE_SINGLE_CONV(fsctuf); |
3259 |
|
3260 |
#define HELPER_SPE_VECTOR_CONV(name) \
|
3261 |
uint64_t helper_ev##name (uint64_t val) \ |
3262 |
{ \ |
3263 |
return ((uint64_t)e##name(val >> 32) << 32) | \ |
3264 |
(uint64_t)e##name(val); \ |
3265 |
} |
3266 |
/* evfscfsi */
|
3267 |
HELPER_SPE_VECTOR_CONV(fscfsi); |
3268 |
/* evfscfui */
|
3269 |
HELPER_SPE_VECTOR_CONV(fscfui); |
3270 |
/* evfscfuf */
|
3271 |
HELPER_SPE_VECTOR_CONV(fscfuf); |
3272 |
/* evfscfsf */
|
3273 |
HELPER_SPE_VECTOR_CONV(fscfsf); |
3274 |
/* evfsctsi */
|
3275 |
HELPER_SPE_VECTOR_CONV(fsctsi); |
3276 |
/* evfsctui */
|
3277 |
HELPER_SPE_VECTOR_CONV(fsctui); |
3278 |
/* evfsctsiz */
|
3279 |
HELPER_SPE_VECTOR_CONV(fsctsiz); |
3280 |
/* evfsctuiz */
|
3281 |
HELPER_SPE_VECTOR_CONV(fsctuiz); |
3282 |
/* evfsctsf */
|
3283 |
HELPER_SPE_VECTOR_CONV(fsctsf); |
3284 |
/* evfsctuf */
|
3285 |
HELPER_SPE_VECTOR_CONV(fsctuf); |
3286 |
|
3287 |
/* Single-precision floating-point arithmetic */
|
3288 |
static always_inline uint32_t efsadd (uint32_t op1, uint32_t op2)
|
3289 |
{ |
3290 |
CPU_FloatU u1, u2; |
3291 |
u1.l = op1; |
3292 |
u2.l = op2; |
3293 |
u1.f = float32_add(u1.f, u2.f, &env->vec_status); |
3294 |
return u1.l;
|
3295 |
} |
3296 |
|
3297 |
static always_inline uint32_t efssub (uint32_t op1, uint32_t op2)
|
3298 |
{ |
3299 |
CPU_FloatU u1, u2; |
3300 |
u1.l = op1; |
3301 |
u2.l = op2; |
3302 |
u1.f = float32_sub(u1.f, u2.f, &env->vec_status); |
3303 |
return u1.l;
|
3304 |
} |
3305 |
|
3306 |
static always_inline uint32_t efsmul (uint32_t op1, uint32_t op2)
|
3307 |
{ |
3308 |
CPU_FloatU u1, u2; |
3309 |
u1.l = op1; |
3310 |
u2.l = op2; |
3311 |
u1.f = float32_mul(u1.f, u2.f, &env->vec_status); |
3312 |
return u1.l;
|
3313 |
} |
3314 |
|
3315 |
static always_inline uint32_t efsdiv (uint32_t op1, uint32_t op2)
|
3316 |
{ |
3317 |
CPU_FloatU u1, u2; |
3318 |
u1.l = op1; |
3319 |
u2.l = op2; |
3320 |
u1.f = float32_div(u1.f, u2.f, &env->vec_status); |
3321 |
return u1.l;
|
3322 |
} |
3323 |
|
3324 |
#define HELPER_SPE_SINGLE_ARITH(name) \
|
3325 |
uint32_t helper_e##name (uint32_t op1, uint32_t op2) \ |
3326 |
{ \ |
3327 |
return e##name(op1, op2); \ |
3328 |
} |
3329 |
/* efsadd */
|
3330 |
HELPER_SPE_SINGLE_ARITH(fsadd); |
3331 |
/* efssub */
|
3332 |
HELPER_SPE_SINGLE_ARITH(fssub); |
3333 |
/* efsmul */
|
3334 |
HELPER_SPE_SINGLE_ARITH(fsmul); |
3335 |
/* efsdiv */
|
3336 |
HELPER_SPE_SINGLE_ARITH(fsdiv); |
3337 |
|
3338 |
#define HELPER_SPE_VECTOR_ARITH(name) \
|
3339 |
uint64_t helper_ev##name (uint64_t op1, uint64_t op2) \ |
3340 |
{ \ |
3341 |
return ((uint64_t)e##name(op1 >> 32, op2 >> 32) << 32) | \ |
3342 |
(uint64_t)e##name(op1, op2); \ |
3343 |
} |
3344 |
/* evfsadd */
|
3345 |
HELPER_SPE_VECTOR_ARITH(fsadd); |
3346 |
/* evfssub */
|
3347 |
HELPER_SPE_VECTOR_ARITH(fssub); |
3348 |
/* evfsmul */
|
3349 |
HELPER_SPE_VECTOR_ARITH(fsmul); |
3350 |
/* evfsdiv */
|
3351 |
HELPER_SPE_VECTOR_ARITH(fsdiv); |
3352 |
|
3353 |
/* Single-precision floating-point comparisons */
|
3354 |
static always_inline uint32_t efststlt (uint32_t op1, uint32_t op2)
|
3355 |
{ |
3356 |
CPU_FloatU u1, u2; |
3357 |
u1.l = op1; |
3358 |
u2.l = op2; |
3359 |
return float32_lt(u1.f, u2.f, &env->vec_status) ? 4 : 0; |
3360 |
} |
3361 |
|
3362 |
static always_inline uint32_t efststgt (uint32_t op1, uint32_t op2)
|
3363 |
{ |
3364 |
CPU_FloatU u1, u2; |
3365 |
u1.l = op1; |
3366 |
u2.l = op2; |
3367 |
return float32_le(u1.f, u2.f, &env->vec_status) ? 0 : 4; |
3368 |
} |
3369 |
|
3370 |
static always_inline uint32_t efststeq (uint32_t op1, uint32_t op2)
|
3371 |
{ |
3372 |
CPU_FloatU u1, u2; |
3373 |
u1.l = op1; |
3374 |
u2.l = op2; |
3375 |
return float32_eq(u1.f, u2.f, &env->vec_status) ? 4 : 0; |
3376 |
} |
3377 |
|
3378 |
static always_inline uint32_t efscmplt (uint32_t op1, uint32_t op2)
|
3379 |
{ |
3380 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
3381 |
return efststlt(op1, op2);
|
3382 |
} |
3383 |
|
3384 |
static always_inline uint32_t efscmpgt (uint32_t op1, uint32_t op2)
|
3385 |
{ |
3386 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
3387 |
return efststgt(op1, op2);
|
3388 |
} |
3389 |
|
3390 |
static always_inline uint32_t efscmpeq (uint32_t op1, uint32_t op2)
|
3391 |
{ |
3392 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
3393 |
return efststeq(op1, op2);
|
3394 |
} |
3395 |
|
3396 |
#define HELPER_SINGLE_SPE_CMP(name) \
|
3397 |
uint32_t helper_e##name (uint32_t op1, uint32_t op2) \ |
3398 |
{ \ |
3399 |
return e##name(op1, op2) << 2; \ |
3400 |
} |
3401 |
/* efststlt */
|
3402 |
HELPER_SINGLE_SPE_CMP(fststlt); |
3403 |
/* efststgt */
|
3404 |
HELPER_SINGLE_SPE_CMP(fststgt); |
3405 |
/* efststeq */
|
3406 |
HELPER_SINGLE_SPE_CMP(fststeq); |
3407 |
/* efscmplt */
|
3408 |
HELPER_SINGLE_SPE_CMP(fscmplt); |
3409 |
/* efscmpgt */
|
3410 |
HELPER_SINGLE_SPE_CMP(fscmpgt); |
3411 |
/* efscmpeq */
|
3412 |
HELPER_SINGLE_SPE_CMP(fscmpeq); |
3413 |
|
3414 |
static always_inline uint32_t evcmp_merge (int t0, int t1) |
3415 |
{ |
3416 |
return (t0 << 3) | (t1 << 2) | ((t0 | t1) << 1) | (t0 & t1); |
3417 |
} |
3418 |
|
3419 |
#define HELPER_VECTOR_SPE_CMP(name) \
|
3420 |
uint32_t helper_ev##name (uint64_t op1, uint64_t op2) \ |
3421 |
{ \ |
3422 |
return evcmp_merge(e##name(op1 >> 32, op2 >> 32), e##name(op1, op2)); \ |
3423 |
} |
3424 |
/* evfststlt */
|
3425 |
HELPER_VECTOR_SPE_CMP(fststlt); |
3426 |
/* evfststgt */
|
3427 |
HELPER_VECTOR_SPE_CMP(fststgt); |
3428 |
/* evfststeq */
|
3429 |
HELPER_VECTOR_SPE_CMP(fststeq); |
3430 |
/* evfscmplt */
|
3431 |
HELPER_VECTOR_SPE_CMP(fscmplt); |
3432 |
/* evfscmpgt */
|
3433 |
HELPER_VECTOR_SPE_CMP(fscmpgt); |
3434 |
/* evfscmpeq */
|
3435 |
HELPER_VECTOR_SPE_CMP(fscmpeq); |
3436 |
|
3437 |
/* Double-precision floating-point conversion */
|
3438 |
uint64_t helper_efdcfsi (uint32_t val) |
3439 |
{ |
3440 |
CPU_DoubleU u; |
3441 |
|
3442 |
u.d = int32_to_float64(val, &env->vec_status); |
3443 |
|
3444 |
return u.ll;
|
3445 |
} |
3446 |
|
3447 |
uint64_t helper_efdcfsid (uint64_t val) |
3448 |
{ |
3449 |
CPU_DoubleU u; |
3450 |
|
3451 |
u.d = int64_to_float64(val, &env->vec_status); |
3452 |
|
3453 |
return u.ll;
|
3454 |
} |
3455 |
|
3456 |
uint64_t helper_efdcfui (uint32_t val) |
3457 |
{ |
3458 |
CPU_DoubleU u; |
3459 |
|
3460 |
u.d = uint32_to_float64(val, &env->vec_status); |
3461 |
|
3462 |
return u.ll;
|
3463 |
} |
3464 |
|
3465 |
uint64_t helper_efdcfuid (uint64_t val) |
3466 |
{ |
3467 |
CPU_DoubleU u; |
3468 |
|
3469 |
u.d = uint64_to_float64(val, &env->vec_status); |
3470 |
|
3471 |
return u.ll;
|
3472 |
} |
3473 |
|
3474 |
uint32_t helper_efdctsi (uint64_t val) |
3475 |
{ |
3476 |
CPU_DoubleU u; |
3477 |
|
3478 |
u.ll = val; |
3479 |
/* NaN are not treated the same way IEEE 754 does */
|
3480 |
if (unlikely(float64_is_nan(u.d)))
|
3481 |
return 0; |
3482 |
|
3483 |
return float64_to_int32(u.d, &env->vec_status);
|
3484 |
} |
3485 |
|
3486 |
uint32_t helper_efdctui (uint64_t val) |
3487 |
{ |
3488 |
CPU_DoubleU u; |
3489 |
|
3490 |
u.ll = val; |
3491 |
/* NaN are not treated the same way IEEE 754 does */
|
3492 |
if (unlikely(float64_is_nan(u.d)))
|
3493 |
return 0; |
3494 |
|
3495 |
return float64_to_uint32(u.d, &env->vec_status);
|
3496 |
} |
3497 |
|
3498 |
uint32_t helper_efdctsiz (uint64_t val) |
3499 |
{ |
3500 |
CPU_DoubleU u; |
3501 |
|
3502 |
u.ll = val; |
3503 |
/* NaN are not treated the same way IEEE 754 does */
|
3504 |
if (unlikely(float64_is_nan(u.d)))
|
3505 |
return 0; |
3506 |
|
3507 |
return float64_to_int32_round_to_zero(u.d, &env->vec_status);
|
3508 |
} |
3509 |
|
3510 |
uint64_t helper_efdctsidz (uint64_t val) |
3511 |
{ |
3512 |
CPU_DoubleU u; |
3513 |
|
3514 |
u.ll = val; |
3515 |
/* NaN are not treated the same way IEEE 754 does */
|
3516 |
if (unlikely(float64_is_nan(u.d)))
|
3517 |
return 0; |
3518 |
|
3519 |
return float64_to_int64_round_to_zero(u.d, &env->vec_status);
|
3520 |
} |
3521 |
|
3522 |
uint32_t helper_efdctuiz (uint64_t val) |
3523 |
{ |
3524 |
CPU_DoubleU u; |
3525 |
|
3526 |
u.ll = val; |
3527 |
/* NaN are not treated the same way IEEE 754 does */
|
3528 |
if (unlikely(float64_is_nan(u.d)))
|
3529 |
return 0; |
3530 |
|
3531 |
return float64_to_uint32_round_to_zero(u.d, &env->vec_status);
|
3532 |
} |
3533 |
|
3534 |
uint64_t helper_efdctuidz (uint64_t val) |
3535 |
{ |
3536 |
CPU_DoubleU u; |
3537 |
|
3538 |
u.ll = val; |
3539 |
/* NaN are not treated the same way IEEE 754 does */
|
3540 |
if (unlikely(float64_is_nan(u.d)))
|
3541 |
return 0; |
3542 |
|
3543 |
return float64_to_uint64_round_to_zero(u.d, &env->vec_status);
|
3544 |
} |
3545 |
|
3546 |
uint64_t helper_efdcfsf (uint32_t val) |
3547 |
{ |
3548 |
CPU_DoubleU u; |
3549 |
float64 tmp; |
3550 |
|
3551 |
u.d = int32_to_float64(val, &env->vec_status); |
3552 |
tmp = int64_to_float64(1ULL << 32, &env->vec_status); |
3553 |
u.d = float64_div(u.d, tmp, &env->vec_status); |
3554 |
|
3555 |
return u.ll;
|
3556 |
} |
3557 |
|
3558 |
uint64_t helper_efdcfuf (uint32_t val) |
3559 |
{ |
3560 |
CPU_DoubleU u; |
3561 |
float64 tmp; |
3562 |
|
3563 |
u.d = uint32_to_float64(val, &env->vec_status); |
3564 |
tmp = int64_to_float64(1ULL << 32, &env->vec_status); |
3565 |
u.d = float64_div(u.d, tmp, &env->vec_status); |
3566 |
|
3567 |
return u.ll;
|
3568 |
} |
3569 |
|
3570 |
uint32_t helper_efdctsf (uint64_t val) |
3571 |
{ |
3572 |
CPU_DoubleU u; |
3573 |
float64 tmp; |
3574 |
|
3575 |
u.ll = val; |
3576 |
/* NaN are not treated the same way IEEE 754 does */
|
3577 |
if (unlikely(float64_is_nan(u.d)))
|
3578 |
return 0; |
3579 |
tmp = uint64_to_float64(1ULL << 32, &env->vec_status); |
3580 |
u.d = float64_mul(u.d, tmp, &env->vec_status); |
3581 |
|
3582 |
return float64_to_int32(u.d, &env->vec_status);
|
3583 |
} |
3584 |
|
3585 |
uint32_t helper_efdctuf (uint64_t val) |
3586 |
{ |
3587 |
CPU_DoubleU u; |
3588 |
float64 tmp; |
3589 |
|
3590 |
u.ll = val; |
3591 |
/* NaN are not treated the same way IEEE 754 does */
|
3592 |
if (unlikely(float64_is_nan(u.d)))
|
3593 |
return 0; |
3594 |
tmp = uint64_to_float64(1ULL << 32, &env->vec_status); |
3595 |
u.d = float64_mul(u.d, tmp, &env->vec_status); |
3596 |
|
3597 |
return float64_to_uint32(u.d, &env->vec_status);
|
3598 |
} |
3599 |
|
3600 |
uint32_t helper_efscfd (uint64_t val) |
3601 |
{ |
3602 |
CPU_DoubleU u1; |
3603 |
CPU_FloatU u2; |
3604 |
|
3605 |
u1.ll = val; |
3606 |
u2.f = float64_to_float32(u1.d, &env->vec_status); |
3607 |
|
3608 |
return u2.l;
|
3609 |
} |
3610 |
|
3611 |
uint64_t helper_efdcfs (uint32_t val) |
3612 |
{ |
3613 |
CPU_DoubleU u2; |
3614 |
CPU_FloatU u1; |
3615 |
|
3616 |
u1.l = val; |
3617 |
u2.d = float32_to_float64(u1.f, &env->vec_status); |
3618 |
|
3619 |
return u2.ll;
|
3620 |
} |
3621 |
|
3622 |
/* Double precision fixed-point arithmetic */
|
3623 |
uint64_t helper_efdadd (uint64_t op1, uint64_t op2) |
3624 |
{ |
3625 |
CPU_DoubleU u1, u2; |
3626 |
u1.ll = op1; |
3627 |
u2.ll = op2; |
3628 |
u1.d = float64_add(u1.d, u2.d, &env->vec_status); |
3629 |
return u1.ll;
|
3630 |
} |
3631 |
|
3632 |
uint64_t helper_efdsub (uint64_t op1, uint64_t op2) |
3633 |
{ |
3634 |
CPU_DoubleU u1, u2; |
3635 |
u1.ll = op1; |
3636 |
u2.ll = op2; |
3637 |
u1.d = float64_sub(u1.d, u2.d, &env->vec_status); |
3638 |
return u1.ll;
|
3639 |
} |
3640 |
|
3641 |
uint64_t helper_efdmul (uint64_t op1, uint64_t op2) |
3642 |
{ |
3643 |
CPU_DoubleU u1, u2; |
3644 |
u1.ll = op1; |
3645 |
u2.ll = op2; |
3646 |
u1.d = float64_mul(u1.d, u2.d, &env->vec_status); |
3647 |
return u1.ll;
|
3648 |
} |
3649 |
|
3650 |
uint64_t helper_efddiv (uint64_t op1, uint64_t op2) |
3651 |
{ |
3652 |
CPU_DoubleU u1, u2; |
3653 |
u1.ll = op1; |
3654 |
u2.ll = op2; |
3655 |
u1.d = float64_div(u1.d, u2.d, &env->vec_status); |
3656 |
return u1.ll;
|
3657 |
} |
3658 |
|
3659 |
/* Double precision floating point helpers */
|
3660 |
uint32_t helper_efdtstlt (uint64_t op1, uint64_t op2) |
3661 |
{ |
3662 |
CPU_DoubleU u1, u2; |
3663 |
u1.ll = op1; |
3664 |
u2.ll = op2; |
3665 |
return float64_lt(u1.d, u2.d, &env->vec_status) ? 4 : 0; |
3666 |
} |
3667 |
|
3668 |
uint32_t helper_efdtstgt (uint64_t op1, uint64_t op2) |
3669 |
{ |
3670 |
CPU_DoubleU u1, u2; |
3671 |
u1.ll = op1; |
3672 |
u2.ll = op2; |
3673 |
return float64_le(u1.d, u2.d, &env->vec_status) ? 0 : 4; |
3674 |
} |
3675 |
|
3676 |
uint32_t helper_efdtsteq (uint64_t op1, uint64_t op2) |
3677 |
{ |
3678 |
CPU_DoubleU u1, u2; |
3679 |
u1.ll = op1; |
3680 |
u2.ll = op2; |
3681 |
return float64_eq(u1.d, u2.d, &env->vec_status) ? 4 : 0; |
3682 |
} |
3683 |
|
3684 |
uint32_t helper_efdcmplt (uint64_t op1, uint64_t op2) |
3685 |
{ |
3686 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
3687 |
return helper_efdtstlt(op1, op2);
|
3688 |
} |
3689 |
|
3690 |
uint32_t helper_efdcmpgt (uint64_t op1, uint64_t op2) |
3691 |
{ |
3692 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
3693 |
return helper_efdtstgt(op1, op2);
|
3694 |
} |
3695 |
|
3696 |
uint32_t helper_efdcmpeq (uint64_t op1, uint64_t op2) |
3697 |
{ |
3698 |
/* XXX: TODO: test special values (NaN, infinites, ...) */
|
3699 |
return helper_efdtsteq(op1, op2);
|
3700 |
} |
3701 |
|
3702 |
/*****************************************************************************/
|
3703 |
/* Softmmu support */
|
3704 |
#if !defined (CONFIG_USER_ONLY)
|
3705 |
|
3706 |
#define MMUSUFFIX _mmu
|
3707 |
|
3708 |
#define SHIFT 0 |
3709 |
#include "softmmu_template.h" |
3710 |
|
3711 |
#define SHIFT 1 |
3712 |
#include "softmmu_template.h" |
3713 |
|
3714 |
#define SHIFT 2 |
3715 |
#include "softmmu_template.h" |
3716 |
|
3717 |
#define SHIFT 3 |
3718 |
#include "softmmu_template.h" |
3719 |
|
3720 |
/* try to fill the TLB and return an exception if error. If retaddr is
|
3721 |
NULL, it means that the function was called in C code (i.e. not
|
3722 |
from generated code or from helper.c) */
|
3723 |
/* XXX: fix it to restore all registers */
|
3724 |
void tlb_fill (target_ulong addr, int is_write, int mmu_idx, void *retaddr) |
3725 |
{ |
3726 |
TranslationBlock *tb; |
3727 |
CPUState *saved_env; |
3728 |
unsigned long pc; |
3729 |
int ret;
|
3730 |
|
3731 |
/* XXX: hack to restore env in all cases, even if not called from
|
3732 |
generated code */
|
3733 |
saved_env = env; |
3734 |
env = cpu_single_env; |
3735 |
ret = cpu_ppc_handle_mmu_fault(env, addr, is_write, mmu_idx, 1);
|
3736 |
if (unlikely(ret != 0)) { |
3737 |
if (likely(retaddr)) {
|
3738 |
/* now we have a real cpu fault */
|
3739 |
pc = (unsigned long)retaddr; |
3740 |
tb = tb_find_pc(pc); |
3741 |
if (likely(tb)) {
|
3742 |
/* the PC is inside the translated code. It means that we have
|
3743 |
a virtual CPU fault */
|
3744 |
cpu_restore_state(tb, env, pc, NULL);
|
3745 |
} |
3746 |
} |
3747 |
helper_raise_exception_err(env->exception_index, env->error_code); |
3748 |
} |
3749 |
env = saved_env; |
3750 |
} |
3751 |
|
3752 |
/* Segment registers load and store */
|
3753 |
target_ulong helper_load_sr (target_ulong sr_num) |
3754 |
{ |
3755 |
#if defined(TARGET_PPC64)
|
3756 |
if (env->mmu_model & POWERPC_MMU_64)
|
3757 |
return ppc_load_sr(env, sr_num);
|
3758 |
#endif
|
3759 |
return env->sr[sr_num];
|
3760 |
} |
3761 |
|
3762 |
void helper_store_sr (target_ulong sr_num, target_ulong val)
|
3763 |
{ |
3764 |
ppc_store_sr(env, sr_num, val); |
3765 |
} |
3766 |
|
3767 |
/* SLB management */
|
3768 |
#if defined(TARGET_PPC64)
|
3769 |
target_ulong helper_load_slb (target_ulong slb_nr) |
3770 |
{ |
3771 |
return ppc_load_slb(env, slb_nr);
|
3772 |
} |
3773 |
|
3774 |
void helper_store_slb (target_ulong rb, target_ulong rs)
|
3775 |
{ |
3776 |
ppc_store_slb(env, rb, rs); |
3777 |
} |
3778 |
|
3779 |
void helper_slbia (void) |
3780 |
{ |
3781 |
ppc_slb_invalidate_all(env); |
3782 |
} |
3783 |
|
3784 |
void helper_slbie (target_ulong addr)
|
3785 |
{ |
3786 |
ppc_slb_invalidate_one(env, addr); |
3787 |
} |
3788 |
|
3789 |
#endif /* defined(TARGET_PPC64) */ |
3790 |
|
3791 |
/* TLB management */
|
3792 |
void helper_tlbia (void) |
3793 |
{ |
3794 |
ppc_tlb_invalidate_all(env); |
3795 |
} |
3796 |
|
3797 |
void helper_tlbie (target_ulong addr)
|
3798 |
{ |
3799 |
ppc_tlb_invalidate_one(env, addr); |
3800 |
} |
3801 |
|
3802 |
/* Software driven TLBs management */
|
3803 |
/* PowerPC 602/603 software TLB load instructions helpers */
|
3804 |
static void do_6xx_tlb (target_ulong new_EPN, int is_code) |
3805 |
{ |
3806 |
target_ulong RPN, CMP, EPN; |
3807 |
int way;
|
3808 |
|
3809 |
RPN = env->spr[SPR_RPA]; |
3810 |
if (is_code) {
|
3811 |
CMP = env->spr[SPR_ICMP]; |
3812 |
EPN = env->spr[SPR_IMISS]; |
3813 |
} else {
|
3814 |
CMP = env->spr[SPR_DCMP]; |
3815 |
EPN = env->spr[SPR_DMISS]; |
3816 |
} |
3817 |
way = (env->spr[SPR_SRR1] >> 17) & 1; |
3818 |
LOG_SWTLB("%s: EPN " ADDRX " " ADDRX " PTE0 " ADDRX |
3819 |
" PTE1 " ADDRX " way %d\n", |
3820 |
__func__, new_EPN, EPN, CMP, RPN, way); |
3821 |
/* Store this TLB */
|
3822 |
ppc6xx_tlb_store(env, (uint32_t)(new_EPN & TARGET_PAGE_MASK), |
3823 |
way, is_code, CMP, RPN); |
3824 |
} |
3825 |
|
3826 |
void helper_6xx_tlbd (target_ulong EPN)
|
3827 |
{ |
3828 |
do_6xx_tlb(EPN, 0);
|
3829 |
} |
3830 |
|
3831 |
void helper_6xx_tlbi (target_ulong EPN)
|
3832 |
{ |
3833 |
do_6xx_tlb(EPN, 1);
|
3834 |
} |
3835 |
|
3836 |
/* PowerPC 74xx software TLB load instructions helpers */
|
3837 |
static void do_74xx_tlb (target_ulong new_EPN, int is_code) |
3838 |
{ |
3839 |
target_ulong RPN, CMP, EPN; |
3840 |
int way;
|
3841 |
|
3842 |
RPN = env->spr[SPR_PTELO]; |
3843 |
CMP = env->spr[SPR_PTEHI]; |
3844 |
EPN = env->spr[SPR_TLBMISS] & ~0x3;
|
3845 |
way = env->spr[SPR_TLBMISS] & 0x3;
|
3846 |
LOG_SWTLB("%s: EPN " ADDRX " " ADDRX " PTE0 " ADDRX |
3847 |
" PTE1 " ADDRX " way %d\n", |
3848 |
__func__, new_EPN, EPN, CMP, RPN, way); |
3849 |
/* Store this TLB */
|
3850 |
ppc6xx_tlb_store(env, (uint32_t)(new_EPN & TARGET_PAGE_MASK), |
3851 |
way, is_code, CMP, RPN); |
3852 |
} |
3853 |
|
3854 |
void helper_74xx_tlbd (target_ulong EPN)
|
3855 |
{ |
3856 |
do_74xx_tlb(EPN, 0);
|
3857 |
} |
3858 |
|
3859 |
void helper_74xx_tlbi (target_ulong EPN)
|
3860 |
{ |
3861 |
do_74xx_tlb(EPN, 1);
|
3862 |
} |
3863 |
|
3864 |
static always_inline target_ulong booke_tlb_to_page_size (int size) |
3865 |
{ |
3866 |
return 1024 << (2 * size); |
3867 |
} |
3868 |
|
3869 |
static always_inline int booke_page_size_to_tlb (target_ulong page_size) |
3870 |
{ |
3871 |
int size;
|
3872 |
|
3873 |
switch (page_size) {
|
3874 |
case 0x00000400UL: |
3875 |
size = 0x0;
|
3876 |
break;
|
3877 |
case 0x00001000UL: |
3878 |
size = 0x1;
|
3879 |
break;
|
3880 |
case 0x00004000UL: |
3881 |
size = 0x2;
|
3882 |
break;
|
3883 |
case 0x00010000UL: |
3884 |
size = 0x3;
|
3885 |
break;
|
3886 |
case 0x00040000UL: |
3887 |
size = 0x4;
|
3888 |
break;
|
3889 |
case 0x00100000UL: |
3890 |
size = 0x5;
|
3891 |
break;
|
3892 |
case 0x00400000UL: |
3893 |
size = 0x6;
|
3894 |
break;
|
3895 |
case 0x01000000UL: |
3896 |
size = 0x7;
|
3897 |
break;
|
3898 |
case 0x04000000UL: |
3899 |
size = 0x8;
|
3900 |
break;
|
3901 |
case 0x10000000UL: |
3902 |
size = 0x9;
|
3903 |
break;
|
3904 |
case 0x40000000UL: |
3905 |
size = 0xA;
|
3906 |
break;
|
3907 |
#if defined (TARGET_PPC64)
|
3908 |
case 0x000100000000ULL: |
3909 |
size = 0xB;
|
3910 |
break;
|
3911 |
case 0x000400000000ULL: |
3912 |
size = 0xC;
|
3913 |
break;
|
3914 |
case 0x001000000000ULL: |
3915 |
size = 0xD;
|
3916 |
break;
|
3917 |
case 0x004000000000ULL: |
3918 |
size = 0xE;
|
3919 |
break;
|
3920 |
case 0x010000000000ULL: |
3921 |
size = 0xF;
|
3922 |
break;
|
3923 |
#endif
|
3924 |
default:
|
3925 |
size = -1;
|
3926 |
break;
|
3927 |
} |
3928 |
|
3929 |
return size;
|
3930 |
} |
3931 |
|
3932 |
/* Helpers for 4xx TLB management */
|
3933 |
target_ulong helper_4xx_tlbre_lo (target_ulong entry) |
3934 |
{ |
3935 |
ppcemb_tlb_t *tlb; |
3936 |
target_ulong ret; |
3937 |
int size;
|
3938 |
|
3939 |
entry &= 0x3F;
|
3940 |
tlb = &env->tlb[entry].tlbe; |
3941 |
ret = tlb->EPN; |
3942 |
if (tlb->prot & PAGE_VALID)
|
3943 |
ret |= 0x400;
|
3944 |
size = booke_page_size_to_tlb(tlb->size); |
3945 |
if (size < 0 || size > 0x7) |
3946 |
size = 1;
|
3947 |
ret |= size << 7;
|
3948 |
env->spr[SPR_40x_PID] = tlb->PID; |
3949 |
return ret;
|
3950 |
} |
3951 |
|
3952 |
target_ulong helper_4xx_tlbre_hi (target_ulong entry) |
3953 |
{ |
3954 |
ppcemb_tlb_t *tlb; |
3955 |
target_ulong ret; |
3956 |
|
3957 |
entry &= 0x3F;
|
3958 |
tlb = &env->tlb[entry].tlbe; |
3959 |
ret = tlb->RPN; |
3960 |
if (tlb->prot & PAGE_EXEC)
|
3961 |
ret |= 0x200;
|
3962 |
if (tlb->prot & PAGE_WRITE)
|
3963 |
ret |= 0x100;
|
3964 |
return ret;
|
3965 |
} |
3966 |
|
3967 |
void helper_4xx_tlbwe_hi (target_ulong entry, target_ulong val)
|
3968 |
{ |
3969 |
ppcemb_tlb_t *tlb; |
3970 |
target_ulong page, end; |
3971 |
|
3972 |
LOG_SWTLB("%s entry %d val " ADDRX "\n", __func__, (int)entry, val); |
3973 |
entry &= 0x3F;
|
3974 |
tlb = &env->tlb[entry].tlbe; |
3975 |
/* Invalidate previous TLB (if it's valid) */
|
3976 |
if (tlb->prot & PAGE_VALID) {
|
3977 |
end = tlb->EPN + tlb->size; |
3978 |
LOG_SWTLB("%s: invalidate old TLB %d start " ADDRX
|
3979 |
" end " ADDRX "\n", __func__, (int)entry, tlb->EPN, end); |
3980 |
for (page = tlb->EPN; page < end; page += TARGET_PAGE_SIZE)
|
3981 |
tlb_flush_page(env, page); |
3982 |
} |
3983 |
tlb->size = booke_tlb_to_page_size((val >> 7) & 0x7); |
3984 |
/* We cannot handle TLB size < TARGET_PAGE_SIZE.
|
3985 |
* If this ever occurs, one should use the ppcemb target instead
|
3986 |
* of the ppc or ppc64 one
|
3987 |
*/
|
3988 |
if ((val & 0x40) && tlb->size < TARGET_PAGE_SIZE) { |
3989 |
cpu_abort(env, "TLB size " TARGET_FMT_lu " < %u " |
3990 |
"are not supported (%d)\n",
|
3991 |
tlb->size, TARGET_PAGE_SIZE, (int)((val >> 7) & 0x7)); |
3992 |
} |
3993 |
tlb->EPN = val & ~(tlb->size - 1);
|
3994 |
if (val & 0x40) |
3995 |
tlb->prot |= PAGE_VALID; |
3996 |
else
|
3997 |
tlb->prot &= ~PAGE_VALID; |
3998 |
if (val & 0x20) { |
3999 |
/* XXX: TO BE FIXED */
|
4000 |
cpu_abort(env, "Little-endian TLB entries are not supported by now\n");
|
4001 |
} |
4002 |
tlb->PID = env->spr[SPR_40x_PID]; /* PID */
|
4003 |
tlb->attr = val & 0xFF;
|
4004 |
LOG_SWTLB("%s: set up TLB %d RPN " PADDRX " EPN " ADDRX |
4005 |
" size " ADDRX " prot %c%c%c%c PID %d\n", __func__, |
4006 |
(int)entry, tlb->RPN, tlb->EPN, tlb->size,
|
4007 |
tlb->prot & PAGE_READ ? 'r' : '-', |
4008 |
tlb->prot & PAGE_WRITE ? 'w' : '-', |
4009 |
tlb->prot & PAGE_EXEC ? 'x' : '-', |
4010 |
tlb->prot & PAGE_VALID ? 'v' : '-', (int)tlb->PID); |
4011 |
/* Invalidate new TLB (if valid) */
|
4012 |
if (tlb->prot & PAGE_VALID) {
|
4013 |
end = tlb->EPN + tlb->size; |
4014 |
LOG_SWTLB("%s: invalidate TLB %d start " ADDRX
|
4015 |
" end " ADDRX "\n", __func__, (int)entry, tlb->EPN, end); |
4016 |
for (page = tlb->EPN; page < end; page += TARGET_PAGE_SIZE)
|
4017 |
tlb_flush_page(env, page); |
4018 |
} |
4019 |
} |
4020 |
|
4021 |
void helper_4xx_tlbwe_lo (target_ulong entry, target_ulong val)
|
4022 |
{ |
4023 |
ppcemb_tlb_t *tlb; |
4024 |
|
4025 |
LOG_SWTLB("%s entry %i val " ADDRX "\n", __func__, (int)entry, val); |
4026 |
entry &= 0x3F;
|
4027 |
tlb = &env->tlb[entry].tlbe; |
4028 |
tlb->RPN = val & 0xFFFFFC00;
|
4029 |
tlb->prot = PAGE_READ; |
4030 |
if (val & 0x200) |
4031 |
tlb->prot |= PAGE_EXEC; |
4032 |
if (val & 0x100) |
4033 |
tlb->prot |= PAGE_WRITE; |
4034 |
LOG_SWTLB("%s: set up TLB %d RPN " PADDRX " EPN " ADDRX |
4035 |
" size " ADDRX " prot %c%c%c%c PID %d\n", __func__, |
4036 |
(int)entry, tlb->RPN, tlb->EPN, tlb->size,
|
4037 |
tlb->prot & PAGE_READ ? 'r' : '-', |
4038 |
tlb->prot & PAGE_WRITE ? 'w' : '-', |
4039 |
tlb->prot & PAGE_EXEC ? 'x' : '-', |
4040 |
tlb->prot & PAGE_VALID ? 'v' : '-', (int)tlb->PID); |
4041 |
} |
4042 |
|
4043 |
target_ulong helper_4xx_tlbsx (target_ulong address) |
4044 |
{ |
4045 |
return ppcemb_tlb_search(env, address, env->spr[SPR_40x_PID]);
|
4046 |
} |
4047 |
|
4048 |
/* PowerPC 440 TLB management */
|
4049 |
void helper_440_tlbwe (uint32_t word, target_ulong entry, target_ulong value)
|
4050 |
{ |
4051 |
ppcemb_tlb_t *tlb; |
4052 |
target_ulong EPN, RPN, size; |
4053 |
int do_flush_tlbs;
|
4054 |
|
4055 |
LOG_SWTLB("%s word %d entry %d value " ADDRX "\n", |
4056 |
__func__, word, (int)entry, value);
|
4057 |
do_flush_tlbs = 0;
|
4058 |
entry &= 0x3F;
|
4059 |
tlb = &env->tlb[entry].tlbe; |
4060 |
switch (word) {
|
4061 |
default:
|
4062 |
/* Just here to please gcc */
|
4063 |
case 0: |
4064 |
EPN = value & 0xFFFFFC00;
|
4065 |
if ((tlb->prot & PAGE_VALID) && EPN != tlb->EPN)
|
4066 |
do_flush_tlbs = 1;
|
4067 |
tlb->EPN = EPN; |
4068 |
size = booke_tlb_to_page_size((value >> 4) & 0xF); |
4069 |
if ((tlb->prot & PAGE_VALID) && tlb->size < size)
|
4070 |
do_flush_tlbs = 1;
|
4071 |
tlb->size = size; |
4072 |
tlb->attr &= ~0x1;
|
4073 |
tlb->attr |= (value >> 8) & 1; |
4074 |
if (value & 0x200) { |
4075 |
tlb->prot |= PAGE_VALID; |
4076 |
} else {
|
4077 |
if (tlb->prot & PAGE_VALID) {
|
4078 |
tlb->prot &= ~PAGE_VALID; |
4079 |
do_flush_tlbs = 1;
|
4080 |
} |
4081 |
} |
4082 |
tlb->PID = env->spr[SPR_440_MMUCR] & 0x000000FF;
|
4083 |
if (do_flush_tlbs)
|
4084 |
tlb_flush(env, 1);
|
4085 |
break;
|
4086 |
case 1: |
4087 |
RPN = value & 0xFFFFFC0F;
|
4088 |
if ((tlb->prot & PAGE_VALID) && tlb->RPN != RPN)
|
4089 |
tlb_flush(env, 1);
|
4090 |
tlb->RPN = RPN; |
4091 |
break;
|
4092 |
case 2: |
4093 |
tlb->attr = (tlb->attr & 0x1) | (value & 0x0000FF00); |
4094 |
tlb->prot = tlb->prot & PAGE_VALID; |
4095 |
if (value & 0x1) |
4096 |
tlb->prot |= PAGE_READ << 4;
|
4097 |
if (value & 0x2) |
4098 |
tlb->prot |= PAGE_WRITE << 4;
|
4099 |
if (value & 0x4) |
4100 |
tlb->prot |= PAGE_EXEC << 4;
|
4101 |
if (value & 0x8) |
4102 |
tlb->prot |= PAGE_READ; |
4103 |
if (value & 0x10) |
4104 |
tlb->prot |= PAGE_WRITE; |
4105 |
if (value & 0x20) |
4106 |
tlb->prot |= PAGE_EXEC; |
4107 |
break;
|
4108 |
} |
4109 |
} |
4110 |
|
4111 |
target_ulong helper_440_tlbre (uint32_t word, target_ulong entry) |
4112 |
{ |
4113 |
ppcemb_tlb_t *tlb; |
4114 |
target_ulong ret; |
4115 |
int size;
|
4116 |
|
4117 |
entry &= 0x3F;
|
4118 |
tlb = &env->tlb[entry].tlbe; |
4119 |
switch (word) {
|
4120 |
default:
|
4121 |
/* Just here to please gcc */
|
4122 |
case 0: |
4123 |
ret = tlb->EPN; |
4124 |
size = booke_page_size_to_tlb(tlb->size); |
4125 |
if (size < 0 || size > 0xF) |
4126 |
size = 1;
|
4127 |
ret |= size << 4;
|
4128 |
if (tlb->attr & 0x1) |
4129 |
ret |= 0x100;
|
4130 |
if (tlb->prot & PAGE_VALID)
|
4131 |
ret |= 0x200;
|
4132 |
env->spr[SPR_440_MMUCR] &= ~0x000000FF;
|
4133 |
env->spr[SPR_440_MMUCR] |= tlb->PID; |
4134 |
break;
|
4135 |
case 1: |
4136 |
ret = tlb->RPN; |
4137 |
break;
|
4138 |
case 2: |
4139 |
ret = tlb->attr & ~0x1;
|
4140 |
if (tlb->prot & (PAGE_READ << 4)) |
4141 |
ret |= 0x1;
|
4142 |
if (tlb->prot & (PAGE_WRITE << 4)) |
4143 |
ret |= 0x2;
|
4144 |
if (tlb->prot & (PAGE_EXEC << 4)) |
4145 |
ret |= 0x4;
|
4146 |
if (tlb->prot & PAGE_READ)
|
4147 |
ret |= 0x8;
|
4148 |
if (tlb->prot & PAGE_WRITE)
|
4149 |
ret |= 0x10;
|
4150 |
if (tlb->prot & PAGE_EXEC)
|
4151 |
ret |= 0x20;
|
4152 |
break;
|
4153 |
} |
4154 |
return ret;
|
4155 |
} |
4156 |
|
4157 |
target_ulong helper_440_tlbsx (target_ulong address) |
4158 |
{ |
4159 |
return ppcemb_tlb_search(env, address, env->spr[SPR_440_MMUCR] & 0xFF); |
4160 |
} |
4161 |
|
4162 |
#endif /* !CONFIG_USER_ONLY */ |