?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tcg-ppc64: bswap64 rotates output 32 bits
[qemu/ar7.git] / target-ppc / int_helper.c
blobe50bdd20ecd2f31d70d59d4fc768cd14601cd334
1 /*
2 * PowerPC integer and vector emulation helpers for QEMU.
3 *
4 * Copyright (c) 2003-2007 Jocelyn Mayer
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
18 */
19 #include "cpu.h"
20 #include "qemu/host-utils.h"
21 #include "helper.h"
22
23 #include "helper_regs.h"
24 /*****************************************************************************/
25 /* Fixed point operations helpers */
26 #if defined(TARGET_PPC64)
27
28 uint64_t helper_mulldo(CPUPPCState *env, uint64_t arg1, uint64_t arg2)
29 {
30 int64_t th;
31 uint64_t tl;
32
33 muls64(&tl, (uint64_t *)&th, arg1, arg2);
34 /* If th != 0 && th != -1, then we had an overflow */
35 if (likely((uint64_t)(th + 1) <= 1)) {
36 env->ov = 0;
37 } else {
38 env->so = env->ov = 1;
39 }
40 return (int64_t)tl;
41 }
42 #endif
43
44 target_ulong helper_cntlzw(target_ulong t)
45 {
46 return clz32(t);
47 }
48
49 #if defined(TARGET_PPC64)
50 target_ulong helper_cntlzd(target_ulong t)
51 {
52 return clz64(t);
53 }
54 #endif
55
56 target_ulong helper_cmpb(target_ulong rs, target_ulong rb)
57 {
58 target_ulong mask = 0xff;
59 target_ulong ra = 0;
60 int i;
61
62 for (i = 0; i < sizeof(target_ulong); i++) {
63 if ((rs & mask) == (rb & mask)) {
64 ra |= mask;
65 }
66 mask <<= 8;
67 }
68 return ra;
69 }
70
71 /* shift right arithmetic helper */
72 target_ulong helper_sraw(CPUPPCState *env, target_ulong value,
73 target_ulong shift)
74 {
75 int32_t ret;
76
77 if (likely(!(shift & 0x20))) {
78 if (likely((uint32_t)shift != 0)) {
79 shift &= 0x1f;
80 ret = (int32_t)value >> shift;
81 if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) {
82 env->ca = 0;
83 } else {
84 env->ca = 1;
85 }
86 } else {
87 ret = (int32_t)value;
88 env->ca = 0;
89 }
90 } else {
91 ret = (int32_t)value >> 31;
92 env->ca = (ret != 0);
93 }
94 return (target_long)ret;
95 }
96
97 #if defined(TARGET_PPC64)
98 target_ulong helper_srad(CPUPPCState *env, target_ulong value,
99 target_ulong shift)
100 {
101 int64_t ret;
102
103 if (likely(!(shift & 0x40))) {
104 if (likely((uint64_t)shift != 0)) {
105 shift &= 0x3f;
106 ret = (int64_t)value >> shift;
107 if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) {
108 env->ca = 0;
109 } else {
110 env->ca = 1;
111 }
112 } else {
113 ret = (int64_t)value;
114 env->ca = 0;
115 }
116 } else {
117 ret = (int64_t)value >> 63;
118 env->ca = (ret != 0);
119 }
120 return ret;
121 }
122 #endif
123
124 #if defined(TARGET_PPC64)
125 target_ulong helper_popcntb(target_ulong val)
126 {
127 val = (val & 0x5555555555555555ULL) + ((val >> 1) &
128 0x5555555555555555ULL);
129 val = (val & 0x3333333333333333ULL) + ((val >> 2) &
130 0x3333333333333333ULL);
131 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
132 0x0f0f0f0f0f0f0f0fULL);
133 return val;
134 }
135
136 target_ulong helper_popcntw(target_ulong val)
137 {
138 val = (val & 0x5555555555555555ULL) + ((val >> 1) &
139 0x5555555555555555ULL);
140 val = (val & 0x3333333333333333ULL) + ((val >> 2) &
141 0x3333333333333333ULL);
142 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
143 0x0f0f0f0f0f0f0f0fULL);
144 val = (val & 0x00ff00ff00ff00ffULL) + ((val >> 8) &
145 0x00ff00ff00ff00ffULL);
146 val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) &
147 0x0000ffff0000ffffULL);
148 return val;
149 }
150
151 target_ulong helper_popcntd(target_ulong val)
152 {
153 return ctpop64(val);
154 }
155 #else
156 target_ulong helper_popcntb(target_ulong val)
157 {
158 val = (val & 0x55555555) + ((val >> 1) & 0x55555555);
159 val = (val & 0x33333333) + ((val >> 2) & 0x33333333);
160 val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f);
161 return val;
162 }
163
164 target_ulong helper_popcntw(target_ulong val)
165 {
166 val = (val & 0x55555555) + ((val >> 1) & 0x55555555);
167 val = (val & 0x33333333) + ((val >> 2) & 0x33333333);
168 val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f);
169 val = (val & 0x00ff00ff) + ((val >> 8) & 0x00ff00ff);
170 val = (val & 0x0000ffff) + ((val >> 16) & 0x0000ffff);
171 return val;
172 }
173 #endif
174
175 /*****************************************************************************/
176 /* PowerPC 601 specific instructions (POWER bridge) */
177 target_ulong helper_div(CPUPPCState *env, target_ulong arg1, target_ulong arg2)
178 {
179 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
180
181 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
182 (int32_t)arg2 == 0) {
183 env->spr[SPR_MQ] = 0;
184 return INT32_MIN;
185 } else {
186 env->spr[SPR_MQ] = tmp % arg2;
187 return tmp / (int32_t)arg2;
188 }
189 }
190
191 target_ulong helper_divo(CPUPPCState *env, target_ulong arg1,
192 target_ulong arg2)
193 {
194 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
195
196 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
197 (int32_t)arg2 == 0) {
198 env->so = env->ov = 1;
199 env->spr[SPR_MQ] = 0;
200 return INT32_MIN;
201 } else {
202 env->spr[SPR_MQ] = tmp % arg2;
203 tmp /= (int32_t)arg2;
204 if ((int32_t)tmp != tmp) {
205 env->so = env->ov = 1;
206 } else {
207 env->ov = 0;
208 }
209 return tmp;
210 }
211 }
212
213 target_ulong helper_divs(CPUPPCState *env, target_ulong arg1,
214 target_ulong arg2)
215 {
216 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
217 (int32_t)arg2 == 0) {
218 env->spr[SPR_MQ] = 0;
219 return INT32_MIN;
220 } else {
221 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
222 return (int32_t)arg1 / (int32_t)arg2;
223 }
224 }
225
226 target_ulong helper_divso(CPUPPCState *env, target_ulong arg1,
227 target_ulong arg2)
228 {
229 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
230 (int32_t)arg2 == 0) {
231 env->so = env->ov = 1;
232 env->spr[SPR_MQ] = 0;
233 return INT32_MIN;
234 } else {
235 env->ov = 0;
236 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
237 return (int32_t)arg1 / (int32_t)arg2;
238 }
239 }
240
241 /*****************************************************************************/
242 /* 602 specific instructions */
243 /* mfrom is the most crazy instruction ever seen, imho ! */
244 /* Real implementation uses a ROM table. Do the same */
245 /* Extremely decomposed:
246 * -arg / 256
247 * return 256 * log10(10 + 1.0) + 0.5
248 */
249 #if !defined(CONFIG_USER_ONLY)
250 target_ulong helper_602_mfrom(target_ulong arg)
251 {
252 if (likely(arg < 602)) {
253 #include "mfrom_table.c"
254 return mfrom_ROM_table[arg];
255 } else {
256 return 0;
257 }
258 }
259 #endif
260
261 /*****************************************************************************/
262 /* Altivec extension helpers */
263 #if defined(HOST_WORDS_BIGENDIAN)
264 #define HI_IDX 0
265 #define LO_IDX 1
266 #else
267 #define HI_IDX 1
268 #define LO_IDX 0
269 #endif
270
271 #if defined(HOST_WORDS_BIGENDIAN)
272 #define VECTOR_FOR_INORDER_I(index, element) \
273 for (index = 0; index < ARRAY_SIZE(r->element); index++)
274 #else
275 #define VECTOR_FOR_INORDER_I(index, element) \
276 for (index = ARRAY_SIZE(r->element)-1; index >= 0; index--)
277 #endif
278
279 /* Saturating arithmetic helpers. */
280 #define SATCVT(from, to, from_type, to_type, min, max) \
281 static inline to_type cvt##from##to(from_type x, int *sat) \
282 { \
283 to_type r; \
284 \
285 if (x < (from_type)min) { \
286 r = min; \
287 *sat = 1; \
288 } else if (x > (from_type)max) { \
289 r = max; \
290 *sat = 1; \
291 } else { \
292 r = x; \
293 } \
294 return r; \
295 }
296 #define SATCVTU(from, to, from_type, to_type, min, max) \
297 static inline to_type cvt##from##to(from_type x, int *sat) \
298 { \
299 to_type r; \
300 \
301 if (x > (from_type)max) { \
302 r = max; \
303 *sat = 1; \
304 } else { \
305 r = x; \
306 } \
307 return r; \
308 }
309 SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX)
310 SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX)
311 SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX)
312
313 SATCVTU(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX)
314 SATCVTU(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX)
315 SATCVTU(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX)
316 SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX)
317 SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX)
318 SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX)
319 #undef SATCVT
320 #undef SATCVTU
321
322 void helper_lvsl(ppc_avr_t *r, target_ulong sh)
323 {
324 int i, j = (sh & 0xf);
325
326 VECTOR_FOR_INORDER_I(i, u8) {
327 r->u8[i] = j++;
328 }
329 }
330
331 void helper_lvsr(ppc_avr_t *r, target_ulong sh)
332 {
333 int i, j = 0x10 - (sh & 0xf);
334
335 VECTOR_FOR_INORDER_I(i, u8) {
336 r->u8[i] = j++;
337 }
338 }
339
340 void helper_mtvscr(CPUPPCState *env, ppc_avr_t *r)
341 {
342 #if defined(HOST_WORDS_BIGENDIAN)
343 env->vscr = r->u32[3];
344 #else
345 env->vscr = r->u32[0];
346 #endif
347 set_flush_to_zero(vscr_nj, &env->vec_status);
348 }
349
350 void helper_vaddcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
351 {
352 int i;
353
354 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
355 r->u32[i] = ~a->u32[i] < b->u32[i];
356 }
357 }
358
359 #define VARITH_DO(name, op, element) \
360 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
361 { \
362 int i; \
363 \
364 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
365 r->element[i] = a->element[i] op b->element[i]; \
366 } \
367 }
368 #define VARITH(suffix, element) \
369 VARITH_DO(add##suffix, +, element) \
370 VARITH_DO(sub##suffix, -, element)
371 VARITH(ubm, u8)
372 VARITH(uhm, u16)
373 VARITH(uwm, u32)
374 #undef VARITH_DO
375 #undef VARITH
376
377 #define VARITHFP(suffix, func) \
378 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
379 ppc_avr_t *b) \
380 { \
381 int i; \
382 \
383 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
384 r->f[i] = func(a->f[i], b->f[i], &env->vec_status); \
385 } \
386 }
387 VARITHFP(addfp, float32_add)
388 VARITHFP(subfp, float32_sub)
389 VARITHFP(minfp, float32_min)
390 VARITHFP(maxfp, float32_max)
391 #undef VARITHFP
392
393 #define VARITHFPFMA(suffix, type) \
394 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
395 ppc_avr_t *b, ppc_avr_t *c) \
396 { \
397 int i; \
398 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
399 r->f[i] = float32_muladd(a->f[i], c->f[i], b->f[i], \
400 type, &env->vec_status); \
401 } \
402 }
403 VARITHFPFMA(maddfp, 0);
404 VARITHFPFMA(nmsubfp, float_muladd_negate_result | float_muladd_negate_c);
405 #undef VARITHFPFMA
406
407 #define VARITHSAT_CASE(type, op, cvt, element) \
408 { \
409 type result = (type)a->element[i] op (type)b->element[i]; \
410 r->element[i] = cvt(result, &sat); \
411 }
412
413 #define VARITHSAT_DO(name, op, optype, cvt, element) \
414 void helper_v##name(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
415 ppc_avr_t *b) \
416 { \
417 int sat = 0; \
418 int i; \
419 \
420 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
421 switch (sizeof(r->element[0])) { \
422 case 1: \
423 VARITHSAT_CASE(optype, op, cvt, element); \
424 break; \
425 case 2: \
426 VARITHSAT_CASE(optype, op, cvt, element); \
427 break; \
428 case 4: \
429 VARITHSAT_CASE(optype, op, cvt, element); \
430 break; \
431 } \
432 } \
433 if (sat) { \
434 env->vscr |= (1 << VSCR_SAT); \
435 } \
436 }
437 #define VARITHSAT_SIGNED(suffix, element, optype, cvt) \
438 VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \
439 VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element)
440 #define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \
441 VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \
442 VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element)
443 VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb)
444 VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh)
445 VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw)
446 VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub)
447 VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh)
448 VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw)
449 #undef VARITHSAT_CASE
450 #undef VARITHSAT_DO
451 #undef VARITHSAT_SIGNED
452 #undef VARITHSAT_UNSIGNED
453
454 #define VAVG_DO(name, element, etype) \
455 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
456 { \
457 int i; \
458 \
459 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
460 etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \
461 r->element[i] = x >> 1; \
462 } \
463 }
464
465 #define VAVG(type, signed_element, signed_type, unsigned_element, \
466 unsigned_type) \
467 VAVG_DO(avgs##type, signed_element, signed_type) \
468 VAVG_DO(avgu##type, unsigned_element, unsigned_type)
469 VAVG(b, s8, int16_t, u8, uint16_t)
470 VAVG(h, s16, int32_t, u16, uint32_t)
471 VAVG(w, s32, int64_t, u32, uint64_t)
472 #undef VAVG_DO
473 #undef VAVG
474
475 #define VCF(suffix, cvt, element) \
476 void helper_vcf##suffix(CPUPPCState *env, ppc_avr_t *r, \
477 ppc_avr_t *b, uint32_t uim) \
478 { \
479 int i; \
480 \
481 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
482 float32 t = cvt(b->element[i], &env->vec_status); \
483 r->f[i] = float32_scalbn(t, -uim, &env->vec_status); \
484 } \
485 }
486 VCF(ux, uint32_to_float32, u32)
487 VCF(sx, int32_to_float32, s32)
488 #undef VCF
489
490 #define VCMP_DO(suffix, compare, element, record) \
491 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \
492 ppc_avr_t *a, ppc_avr_t *b) \
493 { \
494 uint32_t ones = (uint32_t)-1; \
495 uint32_t all = ones; \
496 uint32_t none = 0; \
497 int i; \
498 \
499 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
500 uint32_t result = (a->element[i] compare b->element[i] ? \
501 ones : 0x0); \
502 switch (sizeof(a->element[0])) { \
503 case 4: \
504 r->u32[i] = result; \
505 break; \
506 case 2: \
507 r->u16[i] = result; \
508 break; \
509 case 1: \
510 r->u8[i] = result; \
511 break; \
512 } \
513 all &= result; \
514 none |= result; \
515 } \
516 if (record) { \
517 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
518 } \
519 }
520 #define VCMP(suffix, compare, element) \
521 VCMP_DO(suffix, compare, element, 0) \
522 VCMP_DO(suffix##_dot, compare, element, 1)
523 VCMP(equb, ==, u8)
524 VCMP(equh, ==, u16)
525 VCMP(equw, ==, u32)
526 VCMP(gtub, >, u8)
527 VCMP(gtuh, >, u16)
528 VCMP(gtuw, >, u32)
529 VCMP(gtsb, >, s8)
530 VCMP(gtsh, >, s16)
531 VCMP(gtsw, >, s32)
532 #undef VCMP_DO
533 #undef VCMP
534
535 #define VCMPFP_DO(suffix, compare, order, record) \
536 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \
537 ppc_avr_t *a, ppc_avr_t *b) \
538 { \
539 uint32_t ones = (uint32_t)-1; \
540 uint32_t all = ones; \
541 uint32_t none = 0; \
542 int i; \
543 \
544 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
545 uint32_t result; \
546 int rel = float32_compare_quiet(a->f[i], b->f[i], \
547 &env->vec_status); \
548 if (rel == float_relation_unordered) { \
549 result = 0; \
550 } else if (rel compare order) { \
551 result = ones; \
552 } else { \
553 result = 0; \
554 } \
555 r->u32[i] = result; \
556 all &= result; \
557 none |= result; \
558 } \
559 if (record) { \
560 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
561 } \
562 }
563 #define VCMPFP(suffix, compare, order) \
564 VCMPFP_DO(suffix, compare, order, 0) \
565 VCMPFP_DO(suffix##_dot, compare, order, 1)
566 VCMPFP(eqfp, ==, float_relation_equal)
567 VCMPFP(gefp, !=, float_relation_less)
568 VCMPFP(gtfp, ==, float_relation_greater)
569 #undef VCMPFP_DO
570 #undef VCMPFP
571
572 static inline void vcmpbfp_internal(CPUPPCState *env, ppc_avr_t *r,
573 ppc_avr_t *a, ppc_avr_t *b, int record)
574 {
575 int i;
576 int all_in = 0;
577
578 for (i = 0; i < ARRAY_SIZE(r->f); i++) {
579 int le_rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status);
580 if (le_rel == float_relation_unordered) {
581 r->u32[i] = 0xc0000000;
582 /* ALL_IN does not need to be updated here. */
583 } else {
584 float32 bneg = float32_chs(b->f[i]);
585 int ge_rel = float32_compare_quiet(a->f[i], bneg, &env->vec_status);
586 int le = le_rel != float_relation_greater;
587 int ge = ge_rel != float_relation_less;
588
589 r->u32[i] = ((!le) << 31) | ((!ge) << 30);
590 all_in |= (!le | !ge);
591 }
592 }
593 if (record) {
594 env->crf[6] = (all_in == 0) << 1;
595 }
596 }
597
598 void helper_vcmpbfp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
599 {
600 vcmpbfp_internal(env, r, a, b, 0);
601 }
602
603 void helper_vcmpbfp_dot(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
604 ppc_avr_t *b)
605 {
606 vcmpbfp_internal(env, r, a, b, 1);
607 }
608
609 #define VCT(suffix, satcvt, element) \
610 void helper_vct##suffix(CPUPPCState *env, ppc_avr_t *r, \
611 ppc_avr_t *b, uint32_t uim) \
612 { \
613 int i; \
614 int sat = 0; \
615 float_status s = env->vec_status; \
616 \
617 set_float_rounding_mode(float_round_to_zero, &s); \
618 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
619 if (float32_is_any_nan(b->f[i])) { \
620 r->element[i] = 0; \
621 } else { \
622 float64 t = float32_to_float64(b->f[i], &s); \
623 int64_t j; \
624 \
625 t = float64_scalbn(t, uim, &s); \
626 j = float64_to_int64(t, &s); \
627 r->element[i] = satcvt(j, &sat); \
628 } \
629 } \
630 if (sat) { \
631 env->vscr |= (1 << VSCR_SAT); \
632 } \
633 }
634 VCT(uxs, cvtsduw, u32)
635 VCT(sxs, cvtsdsw, s32)
636 #undef VCT
637
638 void helper_vmhaddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
639 ppc_avr_t *b, ppc_avr_t *c)
640 {
641 int sat = 0;
642 int i;
643
644 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
645 int32_t prod = a->s16[i] * b->s16[i];
646 int32_t t = (int32_t)c->s16[i] + (prod >> 15);
647
648 r->s16[i] = cvtswsh(t, &sat);
649 }
650
651 if (sat) {
652 env->vscr |= (1 << VSCR_SAT);
653 }
654 }
655
656 void helper_vmhraddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
657 ppc_avr_t *b, ppc_avr_t *c)
658 {
659 int sat = 0;
660 int i;
661
662 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
663 int32_t prod = a->s16[i] * b->s16[i] + 0x00004000;
664 int32_t t = (int32_t)c->s16[i] + (prod >> 15);
665 r->s16[i] = cvtswsh(t, &sat);
666 }
667
668 if (sat) {
669 env->vscr |= (1 << VSCR_SAT);
670 }
671 }
672
673 #define VMINMAX_DO(name, compare, element) \
674 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
675 { \
676 int i; \
677 \
678 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
679 if (a->element[i] compare b->element[i]) { \
680 r->element[i] = b->element[i]; \
681 } else { \
682 r->element[i] = a->element[i]; \
683 } \
684 } \
685 }
686 #define VMINMAX(suffix, element) \
687 VMINMAX_DO(min##suffix, >, element) \
688 VMINMAX_DO(max##suffix, <, element)
689 VMINMAX(sb, s8)
690 VMINMAX(sh, s16)
691 VMINMAX(sw, s32)
692 VMINMAX(ub, u8)
693 VMINMAX(uh, u16)
694 VMINMAX(uw, u32)
695 #undef VMINMAX_DO
696 #undef VMINMAX
697
698 void helper_vmladduhm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
699 {
700 int i;
701
702 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
703 int32_t prod = a->s16[i] * b->s16[i];
704 r->s16[i] = (int16_t) (prod + c->s16[i]);
705 }
706 }
707
708 #define VMRG_DO(name, element, highp) \
709 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
710 { \
711 ppc_avr_t result; \
712 int i; \
713 size_t n_elems = ARRAY_SIZE(r->element); \
714 \
715 for (i = 0; i < n_elems / 2; i++) { \
716 if (highp) { \
717 result.element[i*2+HI_IDX] = a->element[i]; \
718 result.element[i*2+LO_IDX] = b->element[i]; \
719 } else { \
720 result.element[n_elems - i * 2 - (1 + HI_IDX)] = \
721 b->element[n_elems - i - 1]; \
722 result.element[n_elems - i * 2 - (1 + LO_IDX)] = \
723 a->element[n_elems - i - 1]; \
724 } \
725 } \
726 *r = result; \
727 }
728 #if defined(HOST_WORDS_BIGENDIAN)
729 #define MRGHI 0
730 #define MRGLO 1
731 #else
732 #define MRGHI 1
733 #define MRGLO 0
734 #endif
735 #define VMRG(suffix, element) \
736 VMRG_DO(mrgl##suffix, element, MRGHI) \
737 VMRG_DO(mrgh##suffix, element, MRGLO)
738 VMRG(b, u8)
739 VMRG(h, u16)
740 VMRG(w, u32)
741 #undef VMRG_DO
742 #undef VMRG
743 #undef MRGHI
744 #undef MRGLO
745
746 void helper_vmsummbm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
747 ppc_avr_t *b, ppc_avr_t *c)
748 {
749 int32_t prod[16];
750 int i;
751
752 for (i = 0; i < ARRAY_SIZE(r->s8); i++) {
753 prod[i] = (int32_t)a->s8[i] * b->u8[i];
754 }
755
756 VECTOR_FOR_INORDER_I(i, s32) {
757 r->s32[i] = c->s32[i] + prod[4 * i] + prod[4 * i + 1] +
758 prod[4 * i + 2] + prod[4 * i + 3];
759 }
760 }
761
762 void helper_vmsumshm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
763 ppc_avr_t *b, ppc_avr_t *c)
764 {
765 int32_t prod[8];
766 int i;
767
768 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
769 prod[i] = a->s16[i] * b->s16[i];
770 }
771
772 VECTOR_FOR_INORDER_I(i, s32) {
773 r->s32[i] = c->s32[i] + prod[2 * i] + prod[2 * i + 1];
774 }
775 }
776
777 void helper_vmsumshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
778 ppc_avr_t *b, ppc_avr_t *c)
779 {
780 int32_t prod[8];
781 int i;
782 int sat = 0;
783
784 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
785 prod[i] = (int32_t)a->s16[i] * b->s16[i];
786 }
787
788 VECTOR_FOR_INORDER_I(i, s32) {
789 int64_t t = (int64_t)c->s32[i] + prod[2 * i] + prod[2 * i + 1];
790
791 r->u32[i] = cvtsdsw(t, &sat);
792 }
793
794 if (sat) {
795 env->vscr |= (1 << VSCR_SAT);
796 }
797 }
798
799 void helper_vmsumubm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
800 ppc_avr_t *b, ppc_avr_t *c)
801 {
802 uint16_t prod[16];
803 int i;
804
805 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
806 prod[i] = a->u8[i] * b->u8[i];
807 }
808
809 VECTOR_FOR_INORDER_I(i, u32) {
810 r->u32[i] = c->u32[i] + prod[4 * i] + prod[4 * i + 1] +
811 prod[4 * i + 2] + prod[4 * i + 3];
812 }
813 }
814
815 void helper_vmsumuhm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
816 ppc_avr_t *b, ppc_avr_t *c)
817 {
818 uint32_t prod[8];
819 int i;
820
821 for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
822 prod[i] = a->u16[i] * b->u16[i];
823 }
824
825 VECTOR_FOR_INORDER_I(i, u32) {
826 r->u32[i] = c->u32[i] + prod[2 * i] + prod[2 * i + 1];
827 }
828 }
829
830 void helper_vmsumuhs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
831 ppc_avr_t *b, ppc_avr_t *c)
832 {
833 uint32_t prod[8];
834 int i;
835 int sat = 0;
836
837 for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
838 prod[i] = a->u16[i] * b->u16[i];
839 }
840
841 VECTOR_FOR_INORDER_I(i, s32) {
842 uint64_t t = (uint64_t)c->u32[i] + prod[2 * i] + prod[2 * i + 1];
843
844 r->u32[i] = cvtuduw(t, &sat);
845 }
846
847 if (sat) {
848 env->vscr |= (1 << VSCR_SAT);
849 }
850 }
851
852 #define VMUL_DO(name, mul_element, prod_element, evenp) \
853 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
854 { \
855 int i; \
856 \
857 VECTOR_FOR_INORDER_I(i, prod_element) { \
858 if (evenp) { \
859 r->prod_element[i] = a->mul_element[i * 2 + HI_IDX] * \
860 b->mul_element[i * 2 + HI_IDX]; \
861 } else { \
862 r->prod_element[i] = a->mul_element[i * 2 + LO_IDX] * \
863 b->mul_element[i * 2 + LO_IDX]; \
864 } \
865 } \
866 }
867 #define VMUL(suffix, mul_element, prod_element) \
868 VMUL_DO(mule##suffix, mul_element, prod_element, 1) \
869 VMUL_DO(mulo##suffix, mul_element, prod_element, 0)
870 VMUL(sb, s8, s16)
871 VMUL(sh, s16, s32)
872 VMUL(ub, u8, u16)
873 VMUL(uh, u16, u32)
874 #undef VMUL_DO
875 #undef VMUL
876
877 void helper_vperm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
878 ppc_avr_t *c)
879 {
880 ppc_avr_t result;
881 int i;
882
883 VECTOR_FOR_INORDER_I(i, u8) {
884 int s = c->u8[i] & 0x1f;
885 #if defined(HOST_WORDS_BIGENDIAN)
886 int index = s & 0xf;
887 #else
888 int index = 15 - (s & 0xf);
889 #endif
890
891 if (s & 0x10) {
892 result.u8[i] = b->u8[index];
893 } else {
894 result.u8[i] = a->u8[index];
895 }
896 }
897 *r = result;
898 }
899
900 #if defined(HOST_WORDS_BIGENDIAN)
901 #define PKBIG 1
902 #else
903 #define PKBIG 0
904 #endif
905 void helper_vpkpx(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
906 {
907 int i, j;
908 ppc_avr_t result;
909 #if defined(HOST_WORDS_BIGENDIAN)
910 const ppc_avr_t *x[2] = { a, b };
911 #else
912 const ppc_avr_t *x[2] = { b, a };
913 #endif
914
915 VECTOR_FOR_INORDER_I(i, u64) {
916 VECTOR_FOR_INORDER_I(j, u32) {
917 uint32_t e = x[i]->u32[j];
918
919 result.u16[4*i+j] = (((e >> 9) & 0xfc00) |
920 ((e >> 6) & 0x3e0) |
921 ((e >> 3) & 0x1f));
922 }
923 }
924 *r = result;
925 }
926
927 #define VPK(suffix, from, to, cvt, dosat) \
928 void helper_vpk##suffix(CPUPPCState *env, ppc_avr_t *r, \
929 ppc_avr_t *a, ppc_avr_t *b) \
930 { \
931 int i; \
932 int sat = 0; \
933 ppc_avr_t result; \
934 ppc_avr_t *a0 = PKBIG ? a : b; \
935 ppc_avr_t *a1 = PKBIG ? b : a; \
936 \
937 VECTOR_FOR_INORDER_I(i, from) { \
938 result.to[i] = cvt(a0->from[i], &sat); \
939 result.to[i+ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat); \
940 } \
941 *r = result; \
942 if (dosat && sat) { \
943 env->vscr |= (1 << VSCR_SAT); \
944 } \
945 }
946 #define I(x, y) (x)
947 VPK(shss, s16, s8, cvtshsb, 1)
948 VPK(shus, s16, u8, cvtshub, 1)
949 VPK(swss, s32, s16, cvtswsh, 1)
950 VPK(swus, s32, u16, cvtswuh, 1)
951 VPK(uhus, u16, u8, cvtuhub, 1)
952 VPK(uwus, u32, u16, cvtuwuh, 1)
953 VPK(uhum, u16, u8, I, 0)
954 VPK(uwum, u32, u16, I, 0)
955 #undef I
956 #undef VPK
957 #undef PKBIG
958
959 void helper_vrefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
960 {
961 int i;
962
963 for (i = 0; i < ARRAY_SIZE(r->f); i++) {
964 r->f[i] = float32_div(float32_one, b->f[i], &env->vec_status);
965 }
966 }
967
968 #define VRFI(suffix, rounding) \
969 void helper_vrfi##suffix(CPUPPCState *env, ppc_avr_t *r, \
970 ppc_avr_t *b) \
971 { \
972 int i; \
973 float_status s = env->vec_status; \
974 \
975 set_float_rounding_mode(rounding, &s); \
976 for (i = 0; i < ARRAY_SIZE(r->f); i++) { \
977 r->f[i] = float32_round_to_int (b->f[i], &s); \
978 } \
979 }
980 VRFI(n, float_round_nearest_even)
981 VRFI(m, float_round_down)
982 VRFI(p, float_round_up)
983 VRFI(z, float_round_to_zero)
984 #undef VRFI
985
986 #define VROTATE(suffix, element) \
987 void helper_vrl##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
988 { \
989 int i; \
990 \
991 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
992 unsigned int mask = ((1 << \
993 (3 + (sizeof(a->element[0]) >> 1))) \
994 - 1); \
995 unsigned int shift = b->element[i] & mask; \
996 r->element[i] = (a->element[i] << shift) | \
997 (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \
998 } \
999 }
1000 VROTATE(b, u8)
1001 VROTATE(h, u16)
1002 VROTATE(w, u32)
1003 #undef VROTATE
1004
1005 void helper_vrsqrtefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1006 {
1007 int i;
1008
1009 for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1010 float32 t = float32_sqrt(b->f[i], &env->vec_status);
1011
1012 r->f[i] = float32_div(float32_one, t, &env->vec_status);
1013 }
1014 }
1015
1016 void helper_vsel(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1017 ppc_avr_t *c)
1018 {
1019 r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]);
1020 r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]);
1021 }
1022
1023 void helper_vexptefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1024 {
1025 int i;
1026
1027 for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1028 r->f[i] = float32_exp2(b->f[i], &env->vec_status);
1029 }
1030 }
1031
1032 void helper_vlogefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1033 {
1034 int i;
1035
1036 for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1037 r->f[i] = float32_log2(b->f[i], &env->vec_status);
1038 }
1039 }
1040
1041 #if defined(HOST_WORDS_BIGENDIAN)
1042 #define LEFT 0
1043 #define RIGHT 1
1044 #else
1045 #define LEFT 1
1046 #define RIGHT 0
1047 #endif
1048 /* The specification says that the results are undefined if all of the
1049 * shift counts are not identical. We check to make sure that they are
1050 * to conform to what real hardware appears to do. */
1051 #define VSHIFT(suffix, leftp) \
1052 void helper_vs##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1053 { \
1054 int shift = b->u8[LO_IDX*15] & 0x7; \
1055 int doit = 1; \
1056 int i; \
1057 \
1058 for (i = 0; i < ARRAY_SIZE(r->u8); i++) { \
1059 doit = doit && ((b->u8[i] & 0x7) == shift); \
1060 } \
1061 if (doit) { \
1062 if (shift == 0) { \
1063 *r = *a; \
1064 } else if (leftp) { \
1065 uint64_t carry = a->u64[LO_IDX] >> (64 - shift); \
1066 \
1067 r->u64[HI_IDX] = (a->u64[HI_IDX] << shift) | carry; \
1068 r->u64[LO_IDX] = a->u64[LO_IDX] << shift; \
1069 } else { \
1070 uint64_t carry = a->u64[HI_IDX] << (64 - shift); \
1071 \
1072 r->u64[LO_IDX] = (a->u64[LO_IDX] >> shift) | carry; \
1073 r->u64[HI_IDX] = a->u64[HI_IDX] >> shift; \
1074 } \
1075 } \
1076 }
1077 VSHIFT(l, LEFT)
1078 VSHIFT(r, RIGHT)
1079 #undef VSHIFT
1080 #undef LEFT
1081 #undef RIGHT
1082
1083 #define VSL(suffix, element) \
1084 void helper_vsl##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1085 { \
1086 int i; \
1087 \
1088 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1089 unsigned int mask = ((1 << \
1090 (3 + (sizeof(a->element[0]) >> 1))) \
1091 - 1); \
1092 unsigned int shift = b->element[i] & mask; \
1093 \
1094 r->element[i] = a->element[i] << shift; \
1095 } \
1096 }
1097 VSL(b, u8)
1098 VSL(h, u16)
1099 VSL(w, u32)
1100 #undef VSL
1101
1102 void helper_vsldoi(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift)
1103 {
1104 int sh = shift & 0xf;
1105 int i;
1106 ppc_avr_t result;
1107
1108 #if defined(HOST_WORDS_BIGENDIAN)
1109 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1110 int index = sh + i;
1111 if (index > 0xf) {
1112 result.u8[i] = b->u8[index - 0x10];
1113 } else {
1114 result.u8[i] = a->u8[index];
1115 }
1116 }
1117 #else
1118 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1119 int index = (16 - sh) + i;
1120 if (index > 0xf) {
1121 result.u8[i] = a->u8[index - 0x10];
1122 } else {
1123 result.u8[i] = b->u8[index];
1124 }
1125 }
1126 #endif
1127 *r = result;
1128 }
1129
1130 void helper_vslo(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1131 {
1132 int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf;
1133
1134 #if defined(HOST_WORDS_BIGENDIAN)
1135 memmove(&r->u8[0], &a->u8[sh], 16 - sh);
1136 memset(&r->u8[16-sh], 0, sh);
1137 #else
1138 memmove(&r->u8[sh], &a->u8[0], 16 - sh);
1139 memset(&r->u8[0], 0, sh);
1140 #endif
1141 }
1142
1143 /* Experimental testing shows that hardware masks the immediate. */
1144 #define _SPLAT_MASKED(element) (splat & (ARRAY_SIZE(r->element) - 1))
1145 #if defined(HOST_WORDS_BIGENDIAN)
1146 #define SPLAT_ELEMENT(element) _SPLAT_MASKED(element)
1147 #else
1148 #define SPLAT_ELEMENT(element) \
1149 (ARRAY_SIZE(r->element) - 1 - _SPLAT_MASKED(element))
1150 #endif
1151 #define VSPLT(suffix, element) \
1152 void helper_vsplt##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t splat) \
1153 { \
1154 uint32_t s = b->element[SPLAT_ELEMENT(element)]; \
1155 int i; \
1156 \
1157 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1158 r->element[i] = s; \
1159 } \
1160 }
1161 VSPLT(b, u8)
1162 VSPLT(h, u16)
1163 VSPLT(w, u32)
1164 #undef VSPLT
1165 #undef SPLAT_ELEMENT
1166 #undef _SPLAT_MASKED
1167
1168 #define VSPLTI(suffix, element, splat_type) \
1169 void helper_vspltis##suffix(ppc_avr_t *r, uint32_t splat) \
1170 { \
1171 splat_type x = (int8_t)(splat << 3) >> 3; \
1172 int i; \
1173 \
1174 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1175 r->element[i] = x; \
1176 } \
1177 }
1178 VSPLTI(b, s8, int8_t)
1179 VSPLTI(h, s16, int16_t)
1180 VSPLTI(w, s32, int32_t)
1181 #undef VSPLTI
1182
1183 #define VSR(suffix, element) \
1184 void helper_vsr##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1185 { \
1186 int i; \
1187 \
1188 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1189 unsigned int mask = ((1 << \
1190 (3 + (sizeof(a->element[0]) >> 1))) \
1191 - 1); \
1192 unsigned int shift = b->element[i] & mask; \
1193 \
1194 r->element[i] = a->element[i] >> shift; \
1195 } \
1196 }
1197 VSR(ab, s8)
1198 VSR(ah, s16)
1199 VSR(aw, s32)
1200 VSR(b, u8)
1201 VSR(h, u16)
1202 VSR(w, u32)
1203 #undef VSR
1204
1205 void helper_vsro(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1206 {
1207 int sh = (b->u8[LO_IDX * 0xf] >> 3) & 0xf;
1208
1209 #if defined(HOST_WORDS_BIGENDIAN)
1210 memmove(&r->u8[sh], &a->u8[0], 16 - sh);
1211 memset(&r->u8[0], 0, sh);
1212 #else
1213 memmove(&r->u8[0], &a->u8[sh], 16 - sh);
1214 memset(&r->u8[16 - sh], 0, sh);
1215 #endif
1216 }
1217
1218 void helper_vsubcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1219 {
1220 int i;
1221
1222 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
1223 r->u32[i] = a->u32[i] >= b->u32[i];
1224 }
1225 }
1226
1227 void helper_vsumsws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1228 {
1229 int64_t t;
1230 int i, upper;
1231 ppc_avr_t result;
1232 int sat = 0;
1233
1234 #if defined(HOST_WORDS_BIGENDIAN)
1235 upper = ARRAY_SIZE(r->s32)-1;
1236 #else
1237 upper = 0;
1238 #endif
1239 t = (int64_t)b->s32[upper];
1240 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
1241 t += a->s32[i];
1242 result.s32[i] = 0;
1243 }
1244 result.s32[upper] = cvtsdsw(t, &sat);
1245 *r = result;
1246
1247 if (sat) {
1248 env->vscr |= (1 << VSCR_SAT);
1249 }
1250 }
1251
1252 void helper_vsum2sws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1253 {
1254 int i, j, upper;
1255 ppc_avr_t result;
1256 int sat = 0;
1257
1258 #if defined(HOST_WORDS_BIGENDIAN)
1259 upper = 1;
1260 #else
1261 upper = 0;
1262 #endif
1263 for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
1264 int64_t t = (int64_t)b->s32[upper + i * 2];
1265
1266 result.u64[i] = 0;
1267 for (j = 0; j < ARRAY_SIZE(r->u64); j++) {
1268 t += a->s32[2 * i + j];
1269 }
1270 result.s32[upper + i * 2] = cvtsdsw(t, &sat);
1271 }
1272
1273 *r = result;
1274 if (sat) {
1275 env->vscr |= (1 << VSCR_SAT);
1276 }
1277 }
1278
1279 void helper_vsum4sbs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1280 {
1281 int i, j;
1282 int sat = 0;
1283
1284 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
1285 int64_t t = (int64_t)b->s32[i];
1286
1287 for (j = 0; j < ARRAY_SIZE(r->s32); j++) {
1288 t += a->s8[4 * i + j];
1289 }
1290 r->s32[i] = cvtsdsw(t, &sat);
1291 }
1292
1293 if (sat) {
1294 env->vscr |= (1 << VSCR_SAT);
1295 }
1296 }
1297
1298 void helper_vsum4shs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1299 {
1300 int sat = 0;
1301 int i;
1302
1303 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
1304 int64_t t = (int64_t)b->s32[i];
1305
1306 t += a->s16[2 * i] + a->s16[2 * i + 1];
1307 r->s32[i] = cvtsdsw(t, &sat);
1308 }
1309
1310 if (sat) {
1311 env->vscr |= (1 << VSCR_SAT);
1312 }
1313 }
1314
1315 void helper_vsum4ubs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1316 {
1317 int i, j;
1318 int sat = 0;
1319
1320 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
1321 uint64_t t = (uint64_t)b->u32[i];
1322
1323 for (j = 0; j < ARRAY_SIZE(r->u32); j++) {
1324 t += a->u8[4 * i + j];
1325 }
1326 r->u32[i] = cvtuduw(t, &sat);
1327 }
1328
1329 if (sat) {
1330 env->vscr |= (1 << VSCR_SAT);
1331 }
1332 }
1333
1334 #if defined(HOST_WORDS_BIGENDIAN)
1335 #define UPKHI 1
1336 #define UPKLO 0
1337 #else
1338 #define UPKHI 0
1339 #define UPKLO 1
1340 #endif
1341 #define VUPKPX(suffix, hi) \
1342 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \
1343 { \
1344 int i; \
1345 ppc_avr_t result; \
1346 \
1347 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \
1348 uint16_t e = b->u16[hi ? i : i+4]; \
1349 uint8_t a = (e >> 15) ? 0xff : 0; \
1350 uint8_t r = (e >> 10) & 0x1f; \
1351 uint8_t g = (e >> 5) & 0x1f; \
1352 uint8_t b = e & 0x1f; \
1353 \
1354 result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \
1355 } \
1356 *r = result; \
1357 }
1358 VUPKPX(lpx, UPKLO)
1359 VUPKPX(hpx, UPKHI)
1360 #undef VUPKPX
1361
1362 #define VUPK(suffix, unpacked, packee, hi) \
1363 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \
1364 { \
1365 int i; \
1366 ppc_avr_t result; \
1367 \
1368 if (hi) { \
1369 for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \
1370 result.unpacked[i] = b->packee[i]; \
1371 } \
1372 } else { \
1373 for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); \
1374 i++) { \
1375 result.unpacked[i - ARRAY_SIZE(r->unpacked)] = b->packee[i]; \
1376 } \
1377 } \
1378 *r = result; \
1379 }
1380 VUPK(hsb, s16, s8, UPKHI)
1381 VUPK(hsh, s32, s16, UPKHI)
1382 VUPK(lsb, s16, s8, UPKLO)
1383 VUPK(lsh, s32, s16, UPKLO)
1384 #undef VUPK
1385 #undef UPKHI
1386 #undef UPKLO
1387
1388 #undef VECTOR_FOR_INORDER_I
1389 #undef HI_IDX
1390 #undef LO_IDX
1391
1392 /*****************************************************************************/
1393 /* SPE extension helpers */
1394 /* Use a table to make this quicker */
1395 static const uint8_t hbrev[16] = {
1396 0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE,
1397 0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF,
1398 };
1399
1400 static inline uint8_t byte_reverse(uint8_t val)
1401 {
1402 return hbrev[val >> 4] | (hbrev[val & 0xF] << 4);
1403 }
1404
1405 static inline uint32_t word_reverse(uint32_t val)
1406 {
1407 return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) |
1408 (byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24);
1409 }
1410
1411 #define MASKBITS 16 /* Random value - to be fixed (implementation dependent) */
1412 target_ulong helper_brinc(target_ulong arg1, target_ulong arg2)
1413 {
1414 uint32_t a, b, d, mask;
1415
1416 mask = UINT32_MAX >> (32 - MASKBITS);
1417 a = arg1 & mask;
1418 b = arg2 & mask;
1419 d = word_reverse(1 + word_reverse(a | ~b));
1420 return (arg1 & ~mask) | (d & b);
1421 }
1422
1423 uint32_t helper_cntlsw32(uint32_t val)
1424 {
1425 if (val & 0x80000000) {
1426 return clz32(~val);
1427 } else {
1428 return clz32(val);
1429 }
1430 }
1431
1432 uint32_t helper_cntlzw32(uint32_t val)
1433 {
1434 return clz32(val);
1435 }
1436
1437 /* 440 specific */
1438 target_ulong helper_dlmzb(CPUPPCState *env, target_ulong high,
1439 target_ulong low, uint32_t update_Rc)
1440 {
1441 target_ulong mask;
1442 int i;
1443
1444 i = 1;
1445 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
1446 if ((high & mask) == 0) {
1447 if (update_Rc) {
1448 env->crf[0] = 0x4;
1449 }
1450 goto done;
1451 }
1452 i++;
1453 }
1454 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
1455 if ((low & mask) == 0) {
1456 if (update_Rc) {
1457 env->crf[0] = 0x8;
1458 }
1459 goto done;
1460 }
1461 i++;
1462 }
1463 if (update_Rc) {
1464 env->crf[0] = 0x2;
1465 }
1466 done:
1467 env->xer = (env->xer & ~0x7F) | i;
1468 if (update_Rc) {
1469 env->crf[0] |= xer_so;
1470 }
1471 return i;
1472 }
AR7 emulation for QEMU