?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
pnv_phb4_pec: use pnv_phb4_pec_get_phb_id() in pnv_pec_dt_xscom()
[qemu.git] / target / ppc / int_helper.c
blob9bc327bcba5acc4af871717295157b2d77fae61e
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 2.1 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
20 #include "qemu/osdep.h"
21 #include "cpu.h"
22 #include "internal.h"
23 #include "qemu/host-utils.h"
24 #include "qemu/main-loop.h"
25 #include "qemu/log.h"
26 #include "exec/helper-proto.h"
27 #include "crypto/aes.h"
28 #include "fpu/softfloat.h"
29 #include "qapi/error.h"
30 #include "qemu/guest-random.h"
31
32 #include "helper_regs.h"
33 /*****************************************************************************/
34 /* Fixed point operations helpers */
35
36 static inline void helper_update_ov_legacy(CPUPPCState *env, int ov)
37 {
38 if (unlikely(ov)) {
39 env->so = env->ov = 1;
40 } else {
41 env->ov = 0;
42 }
43 }
44
45 target_ulong helper_divweu(CPUPPCState *env, target_ulong ra, target_ulong rb,
46 uint32_t oe)
47 {
48 uint64_t rt = 0;
49 int overflow = 0;
50
51 uint64_t dividend = (uint64_t)ra << 32;
52 uint64_t divisor = (uint32_t)rb;
53
54 if (unlikely(divisor == 0)) {
55 overflow = 1;
56 } else {
57 rt = dividend / divisor;
58 overflow = rt > UINT32_MAX;
59 }
60
61 if (unlikely(overflow)) {
62 rt = 0; /* Undefined */
63 }
64
65 if (oe) {
66 helper_update_ov_legacy(env, overflow);
67 }
68
69 return (target_ulong)rt;
70 }
71
72 target_ulong helper_divwe(CPUPPCState *env, target_ulong ra, target_ulong rb,
73 uint32_t oe)
74 {
75 int64_t rt = 0;
76 int overflow = 0;
77
78 int64_t dividend = (int64_t)ra << 32;
79 int64_t divisor = (int64_t)((int32_t)rb);
80
81 if (unlikely((divisor == 0) ||
82 ((divisor == -1ull) && (dividend == INT64_MIN)))) {
83 overflow = 1;
84 } else {
85 rt = dividend / divisor;
86 overflow = rt != (int32_t)rt;
87 }
88
89 if (unlikely(overflow)) {
90 rt = 0; /* Undefined */
91 }
92
93 if (oe) {
94 helper_update_ov_legacy(env, overflow);
95 }
96
97 return (target_ulong)rt;
98 }
99
100 #if defined(TARGET_PPC64)
101
102 uint64_t helper_divdeu(CPUPPCState *env, uint64_t ra, uint64_t rb, uint32_t oe)
103 {
104 uint64_t rt = 0;
105 int overflow = 0;
106
107 if (unlikely(rb == 0 || ra >= rb)) {
108 overflow = 1;
109 rt = 0; /* Undefined */
110 } else {
111 divu128(&rt, &ra, rb);
112 }
113
114 if (oe) {
115 helper_update_ov_legacy(env, overflow);
116 }
117
118 return rt;
119 }
120
121 uint64_t helper_divde(CPUPPCState *env, uint64_t rau, uint64_t rbu, uint32_t oe)
122 {
123 uint64_t rt = 0;
124 int64_t ra = (int64_t)rau;
125 int64_t rb = (int64_t)rbu;
126 int overflow = 0;
127
128 if (unlikely(rb == 0 || uabs64(ra) >= uabs64(rb))) {
129 overflow = 1;
130 rt = 0; /* Undefined */
131 } else {
132 divs128(&rt, &ra, rb);
133 }
134
135 if (oe) {
136 helper_update_ov_legacy(env, overflow);
137 }
138
139 return rt;
140 }
141
142 #endif
143
144
145 #if defined(TARGET_PPC64)
146 /* if x = 0xab, returns 0xababababababababa */
147 #define pattern(x) (((x) & 0xff) * (~(target_ulong)0 / 0xff))
148
149 /*
150 * subtract 1 from each byte, and with inverse, check if MSB is set at each
151 * byte.
152 * i.e. ((0x00 - 0x01) & ~(0x00)) & 0x80
153 * (0xFF & 0xFF) & 0x80 = 0x80 (zero found)
154 */
155 #define haszero(v) (((v) - pattern(0x01)) & ~(v) & pattern(0x80))
156
157 /* When you XOR the pattern and there is a match, that byte will be zero */
158 #define hasvalue(x, n) (haszero((x) ^ pattern(n)))
159
160 uint32_t helper_cmpeqb(target_ulong ra, target_ulong rb)
161 {
162 return hasvalue(rb, ra) ? CRF_GT : 0;
163 }
164
165 #undef pattern
166 #undef haszero
167 #undef hasvalue
168
169 /*
170 * Return a random number.
171 */
172 uint64_t helper_darn32(void)
173 {
174 Error *err = NULL;
175 uint32_t ret;
176
177 if (qemu_guest_getrandom(&ret, sizeof(ret), &err) < 0) {
178 qemu_log_mask(LOG_UNIMP, "darn: Crypto failure: %s",
179 error_get_pretty(err));
180 error_free(err);
181 return -1;
182 }
183
184 return ret;
185 }
186
187 uint64_t helper_darn64(void)
188 {
189 Error *err = NULL;
190 uint64_t ret;
191
192 if (qemu_guest_getrandom(&ret, sizeof(ret), &err) < 0) {
193 qemu_log_mask(LOG_UNIMP, "darn: Crypto failure: %s",
194 error_get_pretty(err));
195 error_free(err);
196 return -1;
197 }
198
199 return ret;
200 }
201
202 uint64_t helper_bpermd(uint64_t rs, uint64_t rb)
203 {
204 int i;
205 uint64_t ra = 0;
206
207 for (i = 0; i < 8; i++) {
208 int index = (rs >> (i * 8)) & 0xFF;
209 if (index < 64) {
210 if (rb & PPC_BIT(index)) {
211 ra |= 1 << i;
212 }
213 }
214 }
215 return ra;
216 }
217
218 #endif
219
220 target_ulong helper_cmpb(target_ulong rs, target_ulong rb)
221 {
222 target_ulong mask = 0xff;
223 target_ulong ra = 0;
224 int i;
225
226 for (i = 0; i < sizeof(target_ulong); i++) {
227 if ((rs & mask) == (rb & mask)) {
228 ra |= mask;
229 }
230 mask <<= 8;
231 }
232 return ra;
233 }
234
235 /* shift right arithmetic helper */
236 target_ulong helper_sraw(CPUPPCState *env, target_ulong value,
237 target_ulong shift)
238 {
239 int32_t ret;
240
241 if (likely(!(shift & 0x20))) {
242 if (likely((uint32_t)shift != 0)) {
243 shift &= 0x1f;
244 ret = (int32_t)value >> shift;
245 if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) {
246 env->ca32 = env->ca = 0;
247 } else {
248 env->ca32 = env->ca = 1;
249 }
250 } else {
251 ret = (int32_t)value;
252 env->ca32 = env->ca = 0;
253 }
254 } else {
255 ret = (int32_t)value >> 31;
256 env->ca32 = env->ca = (ret != 0);
257 }
258 return (target_long)ret;
259 }
260
261 #if defined(TARGET_PPC64)
262 target_ulong helper_srad(CPUPPCState *env, target_ulong value,
263 target_ulong shift)
264 {
265 int64_t ret;
266
267 if (likely(!(shift & 0x40))) {
268 if (likely((uint64_t)shift != 0)) {
269 shift &= 0x3f;
270 ret = (int64_t)value >> shift;
271 if (likely(ret >= 0 || (value & ((1ULL << shift) - 1)) == 0)) {
272 env->ca32 = env->ca = 0;
273 } else {
274 env->ca32 = env->ca = 1;
275 }
276 } else {
277 ret = (int64_t)value;
278 env->ca32 = env->ca = 0;
279 }
280 } else {
281 ret = (int64_t)value >> 63;
282 env->ca32 = env->ca = (ret != 0);
283 }
284 return ret;
285 }
286 #endif
287
288 #if defined(TARGET_PPC64)
289 target_ulong helper_popcntb(target_ulong val)
290 {
291 /* Note that we don't fold past bytes */
292 val = (val & 0x5555555555555555ULL) + ((val >> 1) &
293 0x5555555555555555ULL);
294 val = (val & 0x3333333333333333ULL) + ((val >> 2) &
295 0x3333333333333333ULL);
296 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
297 0x0f0f0f0f0f0f0f0fULL);
298 return val;
299 }
300
301 target_ulong helper_popcntw(target_ulong val)
302 {
303 /* Note that we don't fold past words. */
304 val = (val & 0x5555555555555555ULL) + ((val >> 1) &
305 0x5555555555555555ULL);
306 val = (val & 0x3333333333333333ULL) + ((val >> 2) &
307 0x3333333333333333ULL);
308 val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >> 4) &
309 0x0f0f0f0f0f0f0f0fULL);
310 val = (val & 0x00ff00ff00ff00ffULL) + ((val >> 8) &
311 0x00ff00ff00ff00ffULL);
312 val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) &
313 0x0000ffff0000ffffULL);
314 return val;
315 }
316 #else
317 target_ulong helper_popcntb(target_ulong val)
318 {
319 /* Note that we don't fold past bytes */
320 val = (val & 0x55555555) + ((val >> 1) & 0x55555555);
321 val = (val & 0x33333333) + ((val >> 2) & 0x33333333);
322 val = (val & 0x0f0f0f0f) + ((val >> 4) & 0x0f0f0f0f);
323 return val;
324 }
325 #endif
326
327 uint64_t helper_CFUGED(uint64_t src, uint64_t mask)
328 {
329 /*
330 * Instead of processing the mask bit-by-bit from the most significant to
331 * the least significant bit, as described in PowerISA, we'll handle it in
332 * blocks of 'n' zeros/ones from LSB to MSB. To avoid the decision to use
333 * ctz or cto, we negate the mask at the end of the loop.
334 */
335 target_ulong m, left = 0, right = 0;
336 unsigned int n, i = 64;
337 bool bit = false; /* tracks if we are processing zeros or ones */
338
339 if (mask == 0 || mask == -1) {
340 return src;
341 }
342
343 /* Processes the mask in blocks, from LSB to MSB */
344 while (i) {
345 /* Find how many bits we should take */
346 n = ctz64(mask);
347 if (n > i) {
348 n = i;
349 }
350
351 /*
352 * Extracts 'n' trailing bits of src and put them on the leading 'n'
353 * bits of 'right' or 'left', pushing down the previously extracted
354 * values.
355 */
356 m = (1ll << n) - 1;
357 if (bit) {
358 right = ror64(right | (src & m), n);
359 } else {
360 left = ror64(left | (src & m), n);
361 }
362
363 /*
364 * Discards the processed bits from 'src' and 'mask'. Note that we are
365 * removing 'n' trailing zeros from 'mask', but the logical shift will
366 * add 'n' leading zeros back, so the population count of 'mask' is kept
367 * the same.
368 */
369 src >>= n;
370 mask >>= n;
371 i -= n;
372 bit = !bit;
373 mask = ~mask;
374 }
375
376 /*
377 * At the end, right was ror'ed ctpop(mask) times. To put it back in place,
378 * we'll shift it more 64-ctpop(mask) times.
379 */
380 if (bit) {
381 n = ctpop64(mask);
382 } else {
383 n = 64 - ctpop64(mask);
384 }
385
386 return left | (right >> n);
387 }
388
389 uint64_t helper_PDEPD(uint64_t src, uint64_t mask)
390 {
391 int i, o;
392 uint64_t result = 0;
393
394 if (mask == -1) {
395 return src;
396 }
397
398 for (i = 0; mask != 0; i++) {
399 o = ctz64(mask);
400 mask &= mask - 1;
401 result |= ((src >> i) & 1) << o;
402 }
403
404 return result;
405 }
406
407 uint64_t helper_PEXTD(uint64_t src, uint64_t mask)
408 {
409 int i, o;
410 uint64_t result = 0;
411
412 if (mask == -1) {
413 return src;
414 }
415
416 for (o = 0; mask != 0; o++) {
417 i = ctz64(mask);
418 mask &= mask - 1;
419 result |= ((src >> i) & 1) << o;
420 }
421
422 return result;
423 }
424
425 /*****************************************************************************/
426 /* PowerPC 601 specific instructions (POWER bridge) */
427 target_ulong helper_div(CPUPPCState *env, target_ulong arg1, target_ulong arg2)
428 {
429 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
430
431 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
432 (int32_t)arg2 == 0) {
433 env->spr[SPR_MQ] = 0;
434 return INT32_MIN;
435 } else {
436 env->spr[SPR_MQ] = tmp % arg2;
437 return tmp / (int32_t)arg2;
438 }
439 }
440
441 target_ulong helper_divo(CPUPPCState *env, target_ulong arg1,
442 target_ulong arg2)
443 {
444 uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
445
446 if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
447 (int32_t)arg2 == 0) {
448 env->so = env->ov = 1;
449 env->spr[SPR_MQ] = 0;
450 return INT32_MIN;
451 } else {
452 env->spr[SPR_MQ] = tmp % arg2;
453 tmp /= (int32_t)arg2;
454 if ((int32_t)tmp != tmp) {
455 env->so = env->ov = 1;
456 } else {
457 env->ov = 0;
458 }
459 return tmp;
460 }
461 }
462
463 target_ulong helper_divs(CPUPPCState *env, target_ulong arg1,
464 target_ulong arg2)
465 {
466 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
467 (int32_t)arg2 == 0) {
468 env->spr[SPR_MQ] = 0;
469 return INT32_MIN;
470 } else {
471 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
472 return (int32_t)arg1 / (int32_t)arg2;
473 }
474 }
475
476 target_ulong helper_divso(CPUPPCState *env, target_ulong arg1,
477 target_ulong arg2)
478 {
479 if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
480 (int32_t)arg2 == 0) {
481 env->so = env->ov = 1;
482 env->spr[SPR_MQ] = 0;
483 return INT32_MIN;
484 } else {
485 env->ov = 0;
486 env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
487 return (int32_t)arg1 / (int32_t)arg2;
488 }
489 }
490
491 /*****************************************************************************/
492 /* 602 specific instructions */
493 /* mfrom is the most crazy instruction ever seen, imho ! */
494 /* Real implementation uses a ROM table. Do the same */
495 /*
496 * Extremely decomposed:
497 * -arg / 256
498 * return 256 * log10(10 + 1.0) + 0.5
499 */
500 #if !defined(CONFIG_USER_ONLY)
501 target_ulong helper_602_mfrom(target_ulong arg)
502 {
503 if (likely(arg < 602)) {
504 #include "mfrom_table.c.inc"
505 return mfrom_ROM_table[arg];
506 } else {
507 return 0;
508 }
509 }
510 #endif
511
512 /*****************************************************************************/
513 /* Altivec extension helpers */
514 #if defined(HOST_WORDS_BIGENDIAN)
515 #define VECTOR_FOR_INORDER_I(index, element) \
516 for (index = 0; index < ARRAY_SIZE(r->element); index++)
517 #else
518 #define VECTOR_FOR_INORDER_I(index, element) \
519 for (index = ARRAY_SIZE(r->element) - 1; index >= 0; index--)
520 #endif
521
522 /* Saturating arithmetic helpers. */
523 #define SATCVT(from, to, from_type, to_type, min, max) \
524 static inline to_type cvt##from##to(from_type x, int *sat) \
525 { \
526 to_type r; \
527 \
528 if (x < (from_type)min) { \
529 r = min; \
530 *sat = 1; \
531 } else if (x > (from_type)max) { \
532 r = max; \
533 *sat = 1; \
534 } else { \
535 r = x; \
536 } \
537 return r; \
538 }
539 #define SATCVTU(from, to, from_type, to_type, min, max) \
540 static inline to_type cvt##from##to(from_type x, int *sat) \
541 { \
542 to_type r; \
543 \
544 if (x > (from_type)max) { \
545 r = max; \
546 *sat = 1; \
547 } else { \
548 r = x; \
549 } \
550 return r; \
551 }
552 SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX)
553 SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX)
554 SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX)
555
556 SATCVTU(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX)
557 SATCVTU(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX)
558 SATCVTU(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX)
559 SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX)
560 SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX)
561 SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX)
562 #undef SATCVT
563 #undef SATCVTU
564
565 void helper_mtvscr(CPUPPCState *env, uint32_t vscr)
566 {
567 ppc_store_vscr(env, vscr);
568 }
569
570 uint32_t helper_mfvscr(CPUPPCState *env)
571 {
572 return ppc_get_vscr(env);
573 }
574
575 static inline void set_vscr_sat(CPUPPCState *env)
576 {
577 /* The choice of non-zero value is arbitrary. */
578 env->vscr_sat.u32[0] = 1;
579 }
580
581 void helper_vaddcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
582 {
583 int i;
584
585 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
586 r->u32[i] = ~a->u32[i] < b->u32[i];
587 }
588 }
589
590 /* vprtybw */
591 void helper_vprtybw(ppc_avr_t *r, ppc_avr_t *b)
592 {
593 int i;
594 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
595 uint64_t res = b->u32[i] ^ (b->u32[i] >> 16);
596 res ^= res >> 8;
597 r->u32[i] = res & 1;
598 }
599 }
600
601 /* vprtybd */
602 void helper_vprtybd(ppc_avr_t *r, ppc_avr_t *b)
603 {
604 int i;
605 for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
606 uint64_t res = b->u64[i] ^ (b->u64[i] >> 32);
607 res ^= res >> 16;
608 res ^= res >> 8;
609 r->u64[i] = res & 1;
610 }
611 }
612
613 /* vprtybq */
614 void helper_vprtybq(ppc_avr_t *r, ppc_avr_t *b)
615 {
616 uint64_t res = b->u64[0] ^ b->u64[1];
617 res ^= res >> 32;
618 res ^= res >> 16;
619 res ^= res >> 8;
620 r->VsrD(1) = res & 1;
621 r->VsrD(0) = 0;
622 }
623
624 #define VARITHFP(suffix, func) \
625 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
626 ppc_avr_t *b) \
627 { \
628 int i; \
629 \
630 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
631 r->f32[i] = func(a->f32[i], b->f32[i], &env->vec_status); \
632 } \
633 }
634 VARITHFP(addfp, float32_add)
635 VARITHFP(subfp, float32_sub)
636 VARITHFP(minfp, float32_min)
637 VARITHFP(maxfp, float32_max)
638 #undef VARITHFP
639
640 #define VARITHFPFMA(suffix, type) \
641 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
642 ppc_avr_t *b, ppc_avr_t *c) \
643 { \
644 int i; \
645 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
646 r->f32[i] = float32_muladd(a->f32[i], c->f32[i], b->f32[i], \
647 type, &env->vec_status); \
648 } \
649 }
650 VARITHFPFMA(maddfp, 0);
651 VARITHFPFMA(nmsubfp, float_muladd_negate_result | float_muladd_negate_c);
652 #undef VARITHFPFMA
653
654 #define VARITHSAT_CASE(type, op, cvt, element) \
655 { \
656 type result = (type)a->element[i] op (type)b->element[i]; \
657 r->element[i] = cvt(result, &sat); \
658 }
659
660 #define VARITHSAT_DO(name, op, optype, cvt, element) \
661 void helper_v##name(ppc_avr_t *r, ppc_avr_t *vscr_sat, \
662 ppc_avr_t *a, ppc_avr_t *b, uint32_t desc) \
663 { \
664 int sat = 0; \
665 int i; \
666 \
667 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
668 VARITHSAT_CASE(optype, op, cvt, element); \
669 } \
670 if (sat) { \
671 vscr_sat->u32[0] = 1; \
672 } \
673 }
674 #define VARITHSAT_SIGNED(suffix, element, optype, cvt) \
675 VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \
676 VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element)
677 #define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \
678 VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \
679 VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element)
680 VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb)
681 VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh)
682 VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw)
683 VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub)
684 VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh)
685 VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw)
686 #undef VARITHSAT_CASE
687 #undef VARITHSAT_DO
688 #undef VARITHSAT_SIGNED
689 #undef VARITHSAT_UNSIGNED
690
691 #define VAVG_DO(name, element, etype) \
692 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
693 { \
694 int i; \
695 \
696 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
697 etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \
698 r->element[i] = x >> 1; \
699 } \
700 }
701
702 #define VAVG(type, signed_element, signed_type, unsigned_element, \
703 unsigned_type) \
704 VAVG_DO(avgs##type, signed_element, signed_type) \
705 VAVG_DO(avgu##type, unsigned_element, unsigned_type)
706 VAVG(b, s8, int16_t, u8, uint16_t)
707 VAVG(h, s16, int32_t, u16, uint32_t)
708 VAVG(w, s32, int64_t, u32, uint64_t)
709 #undef VAVG_DO
710 #undef VAVG
711
712 #define VABSDU_DO(name, element) \
713 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
714 { \
715 int i; \
716 \
717 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
718 r->element[i] = (a->element[i] > b->element[i]) ? \
719 (a->element[i] - b->element[i]) : \
720 (b->element[i] - a->element[i]); \
721 } \
722 }
723
724 /*
725 * VABSDU - Vector absolute difference unsigned
726 * name - instruction mnemonic suffix (b: byte, h: halfword, w: word)
727 * element - element type to access from vector
728 */
729 #define VABSDU(type, element) \
730 VABSDU_DO(absdu##type, element)
731 VABSDU(b, u8)
732 VABSDU(h, u16)
733 VABSDU(w, u32)
734 #undef VABSDU_DO
735 #undef VABSDU
736
737 #define VCF(suffix, cvt, element) \
738 void helper_vcf##suffix(CPUPPCState *env, ppc_avr_t *r, \
739 ppc_avr_t *b, uint32_t uim) \
740 { \
741 int i; \
742 \
743 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
744 float32 t = cvt(b->element[i], &env->vec_status); \
745 r->f32[i] = float32_scalbn(t, -uim, &env->vec_status); \
746 } \
747 }
748 VCF(ux, uint32_to_float32, u32)
749 VCF(sx, int32_to_float32, s32)
750 #undef VCF
751
752 #define VCMP_DO(suffix, compare, element, record) \
753 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \
754 ppc_avr_t *a, ppc_avr_t *b) \
755 { \
756 uint64_t ones = (uint64_t)-1; \
757 uint64_t all = ones; \
758 uint64_t none = 0; \
759 int i; \
760 \
761 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
762 uint64_t result = (a->element[i] compare b->element[i] ? \
763 ones : 0x0); \
764 switch (sizeof(a->element[0])) { \
765 case 8: \
766 r->u64[i] = result; \
767 break; \
768 case 4: \
769 r->u32[i] = result; \
770 break; \
771 case 2: \
772 r->u16[i] = result; \
773 break; \
774 case 1: \
775 r->u8[i] = result; \
776 break; \
777 } \
778 all &= result; \
779 none |= result; \
780 } \
781 if (record) { \
782 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
783 } \
784 }
785 #define VCMP(suffix, compare, element) \
786 VCMP_DO(suffix, compare, element, 0) \
787 VCMP_DO(suffix##_dot, compare, element, 1)
788 VCMP(equb, ==, u8)
789 VCMP(equh, ==, u16)
790 VCMP(equw, ==, u32)
791 VCMP(equd, ==, u64)
792 VCMP(gtub, >, u8)
793 VCMP(gtuh, >, u16)
794 VCMP(gtuw, >, u32)
795 VCMP(gtud, >, u64)
796 VCMP(gtsb, >, s8)
797 VCMP(gtsh, >, s16)
798 VCMP(gtsw, >, s32)
799 VCMP(gtsd, >, s64)
800 #undef VCMP_DO
801 #undef VCMP
802
803 #define VCMPNE_DO(suffix, element, etype, cmpzero, record) \
804 void helper_vcmpne##suffix(CPUPPCState *env, ppc_avr_t *r, \
805 ppc_avr_t *a, ppc_avr_t *b) \
806 { \
807 etype ones = (etype)-1; \
808 etype all = ones; \
809 etype result, none = 0; \
810 int i; \
811 \
812 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
813 if (cmpzero) { \
814 result = ((a->element[i] == 0) \
815 || (b->element[i] == 0) \
816 || (a->element[i] != b->element[i]) ? \
817 ones : 0x0); \
818 } else { \
819 result = (a->element[i] != b->element[i]) ? ones : 0x0; \
820 } \
821 r->element[i] = result; \
822 all &= result; \
823 none |= result; \
824 } \
825 if (record) { \
826 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
827 } \
828 }
829
830 /*
831 * VCMPNEZ - Vector compare not equal to zero
832 * suffix - instruction mnemonic suffix (b: byte, h: halfword, w: word)
833 * element - element type to access from vector
834 */
835 #define VCMPNE(suffix, element, etype, cmpzero) \
836 VCMPNE_DO(suffix, element, etype, cmpzero, 0) \
837 VCMPNE_DO(suffix##_dot, element, etype, cmpzero, 1)
838 VCMPNE(zb, u8, uint8_t, 1)
839 VCMPNE(zh, u16, uint16_t, 1)
840 VCMPNE(zw, u32, uint32_t, 1)
841 VCMPNE(b, u8, uint8_t, 0)
842 VCMPNE(h, u16, uint16_t, 0)
843 VCMPNE(w, u32, uint32_t, 0)
844 #undef VCMPNE_DO
845 #undef VCMPNE
846
847 #define VCMPFP_DO(suffix, compare, order, record) \
848 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \
849 ppc_avr_t *a, ppc_avr_t *b) \
850 { \
851 uint32_t ones = (uint32_t)-1; \
852 uint32_t all = ones; \
853 uint32_t none = 0; \
854 int i; \
855 \
856 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
857 uint32_t result; \
858 FloatRelation rel = \
859 float32_compare_quiet(a->f32[i], b->f32[i], \
860 &env->vec_status); \
861 if (rel == float_relation_unordered) { \
862 result = 0; \
863 } else if (rel compare order) { \
864 result = ones; \
865 } else { \
866 result = 0; \
867 } \
868 r->u32[i] = result; \
869 all &= result; \
870 none |= result; \
871 } \
872 if (record) { \
873 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
874 } \
875 }
876 #define VCMPFP(suffix, compare, order) \
877 VCMPFP_DO(suffix, compare, order, 0) \
878 VCMPFP_DO(suffix##_dot, compare, order, 1)
879 VCMPFP(eqfp, ==, float_relation_equal)
880 VCMPFP(gefp, !=, float_relation_less)
881 VCMPFP(gtfp, ==, float_relation_greater)
882 #undef VCMPFP_DO
883 #undef VCMPFP
884
885 static inline void vcmpbfp_internal(CPUPPCState *env, ppc_avr_t *r,
886 ppc_avr_t *a, ppc_avr_t *b, int record)
887 {
888 int i;
889 int all_in = 0;
890
891 for (i = 0; i < ARRAY_SIZE(r->f32); i++) {
892 FloatRelation le_rel = float32_compare_quiet(a->f32[i], b->f32[i],
893 &env->vec_status);
894 if (le_rel == float_relation_unordered) {
895 r->u32[i] = 0xc0000000;
896 all_in = 1;
897 } else {
898 float32 bneg = float32_chs(b->f32[i]);
899 FloatRelation ge_rel = float32_compare_quiet(a->f32[i], bneg,
900 &env->vec_status);
901 int le = le_rel != float_relation_greater;
902 int ge = ge_rel != float_relation_less;
903
904 r->u32[i] = ((!le) << 31) | ((!ge) << 30);
905 all_in |= (!le | !ge);
906 }
907 }
908 if (record) {
909 env->crf[6] = (all_in == 0) << 1;
910 }
911 }
912
913 void helper_vcmpbfp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
914 {
915 vcmpbfp_internal(env, r, a, b, 0);
916 }
917
918 void helper_vcmpbfp_dot(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
919 ppc_avr_t *b)
920 {
921 vcmpbfp_internal(env, r, a, b, 1);
922 }
923
924 #define VCT(suffix, satcvt, element) \
925 void helper_vct##suffix(CPUPPCState *env, ppc_avr_t *r, \
926 ppc_avr_t *b, uint32_t uim) \
927 { \
928 int i; \
929 int sat = 0; \
930 float_status s = env->vec_status; \
931 \
932 set_float_rounding_mode(float_round_to_zero, &s); \
933 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
934 if (float32_is_any_nan(b->f32[i])) { \
935 r->element[i] = 0; \
936 } else { \
937 float64 t = float32_to_float64(b->f32[i], &s); \
938 int64_t j; \
939 \
940 t = float64_scalbn(t, uim, &s); \
941 j = float64_to_int64(t, &s); \
942 r->element[i] = satcvt(j, &sat); \
943 } \
944 } \
945 if (sat) { \
946 set_vscr_sat(env); \
947 } \
948 }
949 VCT(uxs, cvtsduw, u32)
950 VCT(sxs, cvtsdsw, s32)
951 #undef VCT
952
953 target_ulong helper_vclzlsbb(ppc_avr_t *r)
954 {
955 target_ulong count = 0;
956 int i;
957 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
958 if (r->VsrB(i) & 0x01) {
959 break;
960 }
961 count++;
962 }
963 return count;
964 }
965
966 target_ulong helper_vctzlsbb(ppc_avr_t *r)
967 {
968 target_ulong count = 0;
969 int i;
970 for (i = ARRAY_SIZE(r->u8) - 1; i >= 0; i--) {
971 if (r->VsrB(i) & 0x01) {
972 break;
973 }
974 count++;
975 }
976 return count;
977 }
978
979 void helper_vmhaddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
980 ppc_avr_t *b, ppc_avr_t *c)
981 {
982 int sat = 0;
983 int i;
984
985 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
986 int32_t prod = a->s16[i] * b->s16[i];
987 int32_t t = (int32_t)c->s16[i] + (prod >> 15);
988
989 r->s16[i] = cvtswsh(t, &sat);
990 }
991
992 if (sat) {
993 set_vscr_sat(env);
994 }
995 }
996
997 void helper_vmhraddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
998 ppc_avr_t *b, ppc_avr_t *c)
999 {
1000 int sat = 0;
1001 int i;
1002
1003 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1004 int32_t prod = a->s16[i] * b->s16[i] + 0x00004000;
1005 int32_t t = (int32_t)c->s16[i] + (prod >> 15);
1006 r->s16[i] = cvtswsh(t, &sat);
1007 }
1008
1009 if (sat) {
1010 set_vscr_sat(env);
1011 }
1012 }
1013
1014 void helper_vmladduhm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
1015 {
1016 int i;
1017
1018 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1019 int32_t prod = a->s16[i] * b->s16[i];
1020 r->s16[i] = (int16_t) (prod + c->s16[i]);
1021 }
1022 }
1023
1024 #define VMRG_DO(name, element, access, ofs) \
1025 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1026 { \
1027 ppc_avr_t result; \
1028 int i, half = ARRAY_SIZE(r->element) / 2; \
1029 \
1030 for (i = 0; i < half; i++) { \
1031 result.access(i * 2 + 0) = a->access(i + ofs); \
1032 result.access(i * 2 + 1) = b->access(i + ofs); \
1033 } \
1034 *r = result; \
1035 }
1036
1037 #define VMRG(suffix, element, access) \
1038 VMRG_DO(mrgl##suffix, element, access, half) \
1039 VMRG_DO(mrgh##suffix, element, access, 0)
1040 VMRG(b, u8, VsrB)
1041 VMRG(h, u16, VsrH)
1042 VMRG(w, u32, VsrW)
1043 #undef VMRG_DO
1044 #undef VMRG
1045
1046 void helper_vmsummbm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1047 ppc_avr_t *b, ppc_avr_t *c)
1048 {
1049 int32_t prod[16];
1050 int i;
1051
1052 for (i = 0; i < ARRAY_SIZE(r->s8); i++) {
1053 prod[i] = (int32_t)a->s8[i] * b->u8[i];
1054 }
1055
1056 VECTOR_FOR_INORDER_I(i, s32) {
1057 r->s32[i] = c->s32[i] + prod[4 * i] + prod[4 * i + 1] +
1058 prod[4 * i + 2] + prod[4 * i + 3];
1059 }
1060 }
1061
1062 void helper_vmsumshm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1063 ppc_avr_t *b, ppc_avr_t *c)
1064 {
1065 int32_t prod[8];
1066 int i;
1067
1068 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1069 prod[i] = a->s16[i] * b->s16[i];
1070 }
1071
1072 VECTOR_FOR_INORDER_I(i, s32) {
1073 r->s32[i] = c->s32[i] + prod[2 * i] + prod[2 * i + 1];
1074 }
1075 }
1076
1077 void helper_vmsumshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1078 ppc_avr_t *b, ppc_avr_t *c)
1079 {
1080 int32_t prod[8];
1081 int i;
1082 int sat = 0;
1083
1084 for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1085 prod[i] = (int32_t)a->s16[i] * b->s16[i];
1086 }
1087
1088 VECTOR_FOR_INORDER_I(i, s32) {
1089 int64_t t = (int64_t)c->s32[i] + prod[2 * i] + prod[2 * i + 1];
1090
1091 r->u32[i] = cvtsdsw(t, &sat);
1092 }
1093
1094 if (sat) {
1095 set_vscr_sat(env);
1096 }
1097 }
1098
1099 void helper_vmsumubm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1100 ppc_avr_t *b, ppc_avr_t *c)
1101 {
1102 uint16_t prod[16];
1103 int i;
1104
1105 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1106 prod[i] = a->u8[i] * b->u8[i];
1107 }
1108
1109 VECTOR_FOR_INORDER_I(i, u32) {
1110 r->u32[i] = c->u32[i] + prod[4 * i] + prod[4 * i + 1] +
1111 prod[4 * i + 2] + prod[4 * i + 3];
1112 }
1113 }
1114
1115 void helper_vmsumuhm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1116 ppc_avr_t *b, ppc_avr_t *c)
1117 {
1118 uint32_t prod[8];
1119 int i;
1120
1121 for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
1122 prod[i] = a->u16[i] * b->u16[i];
1123 }
1124
1125 VECTOR_FOR_INORDER_I(i, u32) {
1126 r->u32[i] = c->u32[i] + prod[2 * i] + prod[2 * i + 1];
1127 }
1128 }
1129
1130 void helper_vmsumuhs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1131 ppc_avr_t *b, ppc_avr_t *c)
1132 {
1133 uint32_t prod[8];
1134 int i;
1135 int sat = 0;
1136
1137 for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
1138 prod[i] = a->u16[i] * b->u16[i];
1139 }
1140
1141 VECTOR_FOR_INORDER_I(i, s32) {
1142 uint64_t t = (uint64_t)c->u32[i] + prod[2 * i] + prod[2 * i + 1];
1143
1144 r->u32[i] = cvtuduw(t, &sat);
1145 }
1146
1147 if (sat) {
1148 set_vscr_sat(env);
1149 }
1150 }
1151
1152 #define VMUL_DO_EVN(name, mul_element, mul_access, prod_access, cast) \
1153 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1154 { \
1155 int i; \
1156 \
1157 for (i = 0; i < ARRAY_SIZE(r->mul_element); i += 2) { \
1158 r->prod_access(i >> 1) = (cast)a->mul_access(i) * \
1159 (cast)b->mul_access(i); \
1160 } \
1161 }
1162
1163 #define VMUL_DO_ODD(name, mul_element, mul_access, prod_access, cast) \
1164 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1165 { \
1166 int i; \
1167 \
1168 for (i = 0; i < ARRAY_SIZE(r->mul_element); i += 2) { \
1169 r->prod_access(i >> 1) = (cast)a->mul_access(i + 1) * \
1170 (cast)b->mul_access(i + 1); \
1171 } \
1172 }
1173
1174 #define VMUL(suffix, mul_element, mul_access, prod_access, cast) \
1175 VMUL_DO_EVN(mule##suffix, mul_element, mul_access, prod_access, cast) \
1176 VMUL_DO_ODD(mulo##suffix, mul_element, mul_access, prod_access, cast)
1177 VMUL(sb, s8, VsrSB, VsrSH, int16_t)
1178 VMUL(sh, s16, VsrSH, VsrSW, int32_t)
1179 VMUL(sw, s32, VsrSW, VsrSD, int64_t)
1180 VMUL(ub, u8, VsrB, VsrH, uint16_t)
1181 VMUL(uh, u16, VsrH, VsrW, uint32_t)
1182 VMUL(uw, u32, VsrW, VsrD, uint64_t)
1183 #undef VMUL_DO_EVN
1184 #undef VMUL_DO_ODD
1185 #undef VMUL
1186
1187 void helper_vmulhsw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1188 {
1189 int i;
1190
1191 for (i = 0; i < 4; i++) {
1192 r->s32[i] = (int32_t)(((int64_t)a->s32[i] * (int64_t)b->s32[i]) >> 32);
1193 }
1194 }
1195
1196 void helper_vmulhuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1197 {
1198 int i;
1199
1200 for (i = 0; i < 4; i++) {
1201 r->u32[i] = (uint32_t)(((uint64_t)a->u32[i] *
1202 (uint64_t)b->u32[i]) >> 32);
1203 }
1204 }
1205
1206 void helper_vmulhsd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1207 {
1208 uint64_t discard;
1209
1210 muls64(&discard, &r->u64[0], a->s64[0], b->s64[0]);
1211 muls64(&discard, &r->u64[1], a->s64[1], b->s64[1]);
1212 }
1213
1214 void helper_vmulhud(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1215 {
1216 uint64_t discard;
1217
1218 mulu64(&discard, &r->u64[0], a->u64[0], b->u64[0]);
1219 mulu64(&discard, &r->u64[1], a->u64[1], b->u64[1]);
1220 }
1221
1222 void helper_vperm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1223 ppc_avr_t *c)
1224 {
1225 ppc_avr_t result;
1226 int i;
1227
1228 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1229 int s = c->VsrB(i) & 0x1f;
1230 int index = s & 0xf;
1231
1232 if (s & 0x10) {
1233 result.VsrB(i) = b->VsrB(index);
1234 } else {
1235 result.VsrB(i) = a->VsrB(index);
1236 }
1237 }
1238 *r = result;
1239 }
1240
1241 void helper_vpermr(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1242 ppc_avr_t *c)
1243 {
1244 ppc_avr_t result;
1245 int i;
1246
1247 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1248 int s = c->VsrB(i) & 0x1f;
1249 int index = 15 - (s & 0xf);
1250
1251 if (s & 0x10) {
1252 result.VsrB(i) = a->VsrB(index);
1253 } else {
1254 result.VsrB(i) = b->VsrB(index);
1255 }
1256 }
1257 *r = result;
1258 }
1259
1260 #if defined(HOST_WORDS_BIGENDIAN)
1261 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[(i)])
1262 #define VBPERMD_INDEX(i) (i)
1263 #define VBPERMQ_DW(index) (((index) & 0x40) != 0)
1264 #define EXTRACT_BIT(avr, i, index) (extract64((avr)->u64[i], index, 1))
1265 #else
1266 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[15 - (i)])
1267 #define VBPERMD_INDEX(i) (1 - i)
1268 #define VBPERMQ_DW(index) (((index) & 0x40) == 0)
1269 #define EXTRACT_BIT(avr, i, index) \
1270 (extract64((avr)->u64[1 - i], 63 - index, 1))
1271 #endif
1272
1273 void helper_vbpermd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1274 {
1275 int i, j;
1276 ppc_avr_t result = { .u64 = { 0, 0 } };
1277 VECTOR_FOR_INORDER_I(i, u64) {
1278 for (j = 0; j < 8; j++) {
1279 int index = VBPERMQ_INDEX(b, (i * 8) + j);
1280 if (index < 64 && EXTRACT_BIT(a, i, index)) {
1281 result.u64[VBPERMD_INDEX(i)] |= (0x80 >> j);
1282 }
1283 }
1284 }
1285 *r = result;
1286 }
1287
1288 void helper_vbpermq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1289 {
1290 int i;
1291 uint64_t perm = 0;
1292
1293 VECTOR_FOR_INORDER_I(i, u8) {
1294 int index = VBPERMQ_INDEX(b, i);
1295
1296 if (index < 128) {
1297 uint64_t mask = (1ull << (63 - (index & 0x3F)));
1298 if (a->u64[VBPERMQ_DW(index)] & mask) {
1299 perm |= (0x8000 >> i);
1300 }
1301 }
1302 }
1303
1304 r->VsrD(0) = perm;
1305 r->VsrD(1) = 0;
1306 }
1307
1308 #undef VBPERMQ_INDEX
1309 #undef VBPERMQ_DW
1310
1311 #define PMSUM(name, srcfld, trgfld, trgtyp) \
1312 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1313 { \
1314 int i, j; \
1315 trgtyp prod[sizeof(ppc_avr_t) / sizeof(a->srcfld[0])]; \
1316 \
1317 VECTOR_FOR_INORDER_I(i, srcfld) { \
1318 prod[i] = 0; \
1319 for (j = 0; j < sizeof(a->srcfld[0]) * 8; j++) { \
1320 if (a->srcfld[i] & (1ull << j)) { \
1321 prod[i] ^= ((trgtyp)b->srcfld[i] << j); \
1322 } \
1323 } \
1324 } \
1325 \
1326 VECTOR_FOR_INORDER_I(i, trgfld) { \
1327 r->trgfld[i] = prod[2 * i] ^ prod[2 * i + 1]; \
1328 } \
1329 }
1330
1331 PMSUM(vpmsumb, u8, u16, uint16_t)
1332 PMSUM(vpmsumh, u16, u32, uint32_t)
1333 PMSUM(vpmsumw, u32, u64, uint64_t)
1334
1335 void helper_vpmsumd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1336 {
1337
1338 #ifdef CONFIG_INT128
1339 int i, j;
1340 __uint128_t prod[2];
1341
1342 VECTOR_FOR_INORDER_I(i, u64) {
1343 prod[i] = 0;
1344 for (j = 0; j < 64; j++) {
1345 if (a->u64[i] & (1ull << j)) {
1346 prod[i] ^= (((__uint128_t)b->u64[i]) << j);
1347 }
1348 }
1349 }
1350
1351 r->u128 = prod[0] ^ prod[1];
1352
1353 #else
1354 int i, j;
1355 ppc_avr_t prod[2];
1356
1357 VECTOR_FOR_INORDER_I(i, u64) {
1358 prod[i].VsrD(1) = prod[i].VsrD(0) = 0;
1359 for (j = 0; j < 64; j++) {
1360 if (a->u64[i] & (1ull << j)) {
1361 ppc_avr_t bshift;
1362 if (j == 0) {
1363 bshift.VsrD(0) = 0;
1364 bshift.VsrD(1) = b->u64[i];
1365 } else {
1366 bshift.VsrD(0) = b->u64[i] >> (64 - j);
1367 bshift.VsrD(1) = b->u64[i] << j;
1368 }
1369 prod[i].VsrD(1) ^= bshift.VsrD(1);
1370 prod[i].VsrD(0) ^= bshift.VsrD(0);
1371 }
1372 }
1373 }
1374
1375 r->VsrD(1) = prod[0].VsrD(1) ^ prod[1].VsrD(1);
1376 r->VsrD(0) = prod[0].VsrD(0) ^ prod[1].VsrD(0);
1377 #endif
1378 }
1379
1380
1381 #if defined(HOST_WORDS_BIGENDIAN)
1382 #define PKBIG 1
1383 #else
1384 #define PKBIG 0
1385 #endif
1386 void helper_vpkpx(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1387 {
1388 int i, j;
1389 ppc_avr_t result;
1390 #if defined(HOST_WORDS_BIGENDIAN)
1391 const ppc_avr_t *x[2] = { a, b };
1392 #else
1393 const ppc_avr_t *x[2] = { b, a };
1394 #endif
1395
1396 VECTOR_FOR_INORDER_I(i, u64) {
1397 VECTOR_FOR_INORDER_I(j, u32) {
1398 uint32_t e = x[i]->u32[j];
1399
1400 result.u16[4 * i + j] = (((e >> 9) & 0xfc00) |
1401 ((e >> 6) & 0x3e0) |
1402 ((e >> 3) & 0x1f));
1403 }
1404 }
1405 *r = result;
1406 }
1407
1408 #define VPK(suffix, from, to, cvt, dosat) \
1409 void helper_vpk##suffix(CPUPPCState *env, ppc_avr_t *r, \
1410 ppc_avr_t *a, ppc_avr_t *b) \
1411 { \
1412 int i; \
1413 int sat = 0; \
1414 ppc_avr_t result; \
1415 ppc_avr_t *a0 = PKBIG ? a : b; \
1416 ppc_avr_t *a1 = PKBIG ? b : a; \
1417 \
1418 VECTOR_FOR_INORDER_I(i, from) { \
1419 result.to[i] = cvt(a0->from[i], &sat); \
1420 result.to[i + ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat);\
1421 } \
1422 *r = result; \
1423 if (dosat && sat) { \
1424 set_vscr_sat(env); \
1425 } \
1426 }
1427 #define I(x, y) (x)
1428 VPK(shss, s16, s8, cvtshsb, 1)
1429 VPK(shus, s16, u8, cvtshub, 1)
1430 VPK(swss, s32, s16, cvtswsh, 1)
1431 VPK(swus, s32, u16, cvtswuh, 1)
1432 VPK(sdss, s64, s32, cvtsdsw, 1)
1433 VPK(sdus, s64, u32, cvtsduw, 1)
1434 VPK(uhus, u16, u8, cvtuhub, 1)
1435 VPK(uwus, u32, u16, cvtuwuh, 1)
1436 VPK(udus, u64, u32, cvtuduw, 1)
1437 VPK(uhum, u16, u8, I, 0)
1438 VPK(uwum, u32, u16, I, 0)
1439 VPK(udum, u64, u32, I, 0)
1440 #undef I
1441 #undef VPK
1442 #undef PKBIG
1443
1444 void helper_vrefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1445 {
1446 int i;
1447
1448 for (i = 0; i < ARRAY_SIZE(r->f32); i++) {
1449 r->f32[i] = float32_div(float32_one, b->f32[i], &env->vec_status);
1450 }
1451 }
1452
1453 #define VRFI(suffix, rounding) \
1454 void helper_vrfi##suffix(CPUPPCState *env, ppc_avr_t *r, \
1455 ppc_avr_t *b) \
1456 { \
1457 int i; \
1458 float_status s = env->vec_status; \
1459 \
1460 set_float_rounding_mode(rounding, &s); \
1461 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
1462 r->f32[i] = float32_round_to_int (b->f32[i], &s); \
1463 } \
1464 }
1465 VRFI(n, float_round_nearest_even)
1466 VRFI(m, float_round_down)
1467 VRFI(p, float_round_up)
1468 VRFI(z, float_round_to_zero)
1469 #undef VRFI
1470
1471 void helper_vrsqrtefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1472 {
1473 int i;
1474
1475 for (i = 0; i < ARRAY_SIZE(r->f32); i++) {
1476 float32 t = float32_sqrt(b->f32[i], &env->vec_status);
1477
1478 r->f32[i] = float32_div(float32_one, t, &env->vec_status);
1479 }
1480 }
1481
1482 #define VRLMI(name, size, element, insert) \
1483 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1484 { \
1485 int i; \
1486 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1487 uint##size##_t src1 = a->element[i]; \
1488 uint##size##_t src2 = b->element[i]; \
1489 uint##size##_t src3 = r->element[i]; \
1490 uint##size##_t begin, end, shift, mask, rot_val; \
1491 \
1492 shift = extract##size(src2, 0, 6); \
1493 end = extract##size(src2, 8, 6); \
1494 begin = extract##size(src2, 16, 6); \
1495 rot_val = rol##size(src1, shift); \
1496 mask = mask_u##size(begin, end); \
1497 if (insert) { \
1498 r->element[i] = (rot_val & mask) | (src3 & ~mask); \
1499 } else { \
1500 r->element[i] = (rot_val & mask); \
1501 } \
1502 } \
1503 }
1504
1505 VRLMI(vrldmi, 64, u64, 1);
1506 VRLMI(vrlwmi, 32, u32, 1);
1507 VRLMI(vrldnm, 64, u64, 0);
1508 VRLMI(vrlwnm, 32, u32, 0);
1509
1510 void helper_vsel(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1511 ppc_avr_t *c)
1512 {
1513 r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]);
1514 r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]);
1515 }
1516
1517 void helper_vexptefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1518 {
1519 int i;
1520
1521 for (i = 0; i < ARRAY_SIZE(r->f32); i++) {
1522 r->f32[i] = float32_exp2(b->f32[i], &env->vec_status);
1523 }
1524 }
1525
1526 void helper_vlogefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1527 {
1528 int i;
1529
1530 for (i = 0; i < ARRAY_SIZE(r->f32); i++) {
1531 r->f32[i] = float32_log2(b->f32[i], &env->vec_status);
1532 }
1533 }
1534
1535 #define VEXTU_X_DO(name, size, left) \
1536 target_ulong glue(helper_, name)(target_ulong a, ppc_avr_t *b) \
1537 { \
1538 int index = (a & 0xf) * 8; \
1539 if (left) { \
1540 index = 128 - index - size; \
1541 } \
1542 return int128_getlo(int128_rshift(b->s128, index)) & \
1543 MAKE_64BIT_MASK(0, size); \
1544 }
1545 VEXTU_X_DO(vextublx, 8, 1)
1546 VEXTU_X_DO(vextuhlx, 16, 1)
1547 VEXTU_X_DO(vextuwlx, 32, 1)
1548 VEXTU_X_DO(vextubrx, 8, 0)
1549 VEXTU_X_DO(vextuhrx, 16, 0)
1550 VEXTU_X_DO(vextuwrx, 32, 0)
1551 #undef VEXTU_X_DO
1552
1553 void helper_vslv(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1554 {
1555 int i;
1556 unsigned int shift, bytes, size;
1557
1558 size = ARRAY_SIZE(r->u8);
1559 for (i = 0; i < size; i++) {
1560 shift = b->VsrB(i) & 0x7; /* extract shift value */
1561 bytes = (a->VsrB(i) << 8) + /* extract adjacent bytes */
1562 (((i + 1) < size) ? a->VsrB(i + 1) : 0);
1563 r->VsrB(i) = (bytes << shift) >> 8; /* shift and store result */
1564 }
1565 }
1566
1567 void helper_vsrv(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1568 {
1569 int i;
1570 unsigned int shift, bytes;
1571
1572 /*
1573 * Use reverse order, as destination and source register can be
1574 * same. Its being modified in place saving temporary, reverse
1575 * order will guarantee that computed result is not fed back.
1576 */
1577 for (i = ARRAY_SIZE(r->u8) - 1; i >= 0; i--) {
1578 shift = b->VsrB(i) & 0x7; /* extract shift value */
1579 bytes = ((i ? a->VsrB(i - 1) : 0) << 8) + a->VsrB(i);
1580 /* extract adjacent bytes */
1581 r->VsrB(i) = (bytes >> shift) & 0xFF; /* shift and store result */
1582 }
1583 }
1584
1585 void helper_vsldoi(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift)
1586 {
1587 int sh = shift & 0xf;
1588 int i;
1589 ppc_avr_t result;
1590
1591 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1592 int index = sh + i;
1593 if (index > 0xf) {
1594 result.VsrB(i) = b->VsrB(index - 0x10);
1595 } else {
1596 result.VsrB(i) = a->VsrB(index);
1597 }
1598 }
1599 *r = result;
1600 }
1601
1602 void helper_vslo(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1603 {
1604 int sh = (b->VsrB(0xf) >> 3) & 0xf;
1605
1606 #if defined(HOST_WORDS_BIGENDIAN)
1607 memmove(&r->u8[0], &a->u8[sh], 16 - sh);
1608 memset(&r->u8[16 - sh], 0, sh);
1609 #else
1610 memmove(&r->u8[sh], &a->u8[0], 16 - sh);
1611 memset(&r->u8[0], 0, sh);
1612 #endif
1613 }
1614
1615 #if defined(HOST_WORDS_BIGENDIAN)
1616 #define ELEM_ADDR(VEC, IDX, SIZE) (&(VEC)->u8[IDX])
1617 #else
1618 #define ELEM_ADDR(VEC, IDX, SIZE) (&(VEC)->u8[15 - (IDX)] - (SIZE) + 1)
1619 #endif
1620
1621 #define VINSX(SUFFIX, TYPE) \
1622 void glue(glue(helper_VINS, SUFFIX), LX)(CPUPPCState *env, ppc_avr_t *t, \
1623 uint64_t val, target_ulong index) \
1624 { \
1625 const int maxidx = ARRAY_SIZE(t->u8) - sizeof(TYPE); \
1626 target_long idx = index; \
1627 \
1628 if (idx < 0 || idx > maxidx) { \
1629 idx = idx < 0 ? sizeof(TYPE) - idx : idx; \
1630 qemu_log_mask(LOG_GUEST_ERROR, \
1631 "Invalid index for Vector Insert Element after 0x" TARGET_FMT_lx \
1632 ", RA = " TARGET_FMT_ld " > %d\n", env->nip, idx, maxidx); \
1633 } else { \
1634 TYPE src = val; \
1635 memcpy(ELEM_ADDR(t, idx, sizeof(TYPE)), &src, sizeof(TYPE)); \
1636 } \
1637 }
1638 VINSX(B, uint8_t)
1639 VINSX(H, uint16_t)
1640 VINSX(W, uint32_t)
1641 VINSX(D, uint64_t)
1642 #undef ELEM_ADDR
1643 #undef VINSX
1644 #if defined(HOST_WORDS_BIGENDIAN)
1645 #define VEXTDVLX(NAME, SIZE) \
1646 void helper_##NAME(CPUPPCState *env, ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, \
1647 target_ulong index) \
1648 { \
1649 const target_long idx = index; \
1650 ppc_avr_t tmp[2] = { *a, *b }; \
1651 memset(t, 0, sizeof(*t)); \
1652 if (idx >= 0 && idx + SIZE <= sizeof(tmp)) { \
1653 memcpy(&t->u8[ARRAY_SIZE(t->u8) / 2 - SIZE], (void *)tmp + idx, SIZE); \
1654 } else { \
1655 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for " #NAME " after 0x" \
1656 TARGET_FMT_lx ", RC = " TARGET_FMT_ld " > %d\n", \
1657 env->nip, idx < 0 ? SIZE - idx : idx, 32 - SIZE); \
1658 } \
1659 }
1660 #else
1661 #define VEXTDVLX(NAME, SIZE) \
1662 void helper_##NAME(CPUPPCState *env, ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, \
1663 target_ulong index) \
1664 { \
1665 const target_long idx = index; \
1666 ppc_avr_t tmp[2] = { *b, *a }; \
1667 memset(t, 0, sizeof(*t)); \
1668 if (idx >= 0 && idx + SIZE <= sizeof(tmp)) { \
1669 memcpy(&t->u8[ARRAY_SIZE(t->u8) / 2], \
1670 (void *)tmp + sizeof(tmp) - SIZE - idx, SIZE); \
1671 } else { \
1672 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for " #NAME " after 0x" \
1673 TARGET_FMT_lx ", RC = " TARGET_FMT_ld " > %d\n", \
1674 env->nip, idx < 0 ? SIZE - idx : idx, 32 - SIZE); \
1675 } \
1676 }
1677 #endif
1678 VEXTDVLX(VEXTDUBVLX, 1)
1679 VEXTDVLX(VEXTDUHVLX, 2)
1680 VEXTDVLX(VEXTDUWVLX, 4)
1681 VEXTDVLX(VEXTDDVLX, 8)
1682 #undef VEXTDVLX
1683 #if defined(HOST_WORDS_BIGENDIAN)
1684 #define VEXTRACT(suffix, element) \
1685 void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1686 { \
1687 uint32_t es = sizeof(r->element[0]); \
1688 memmove(&r->u8[8 - es], &b->u8[index], es); \
1689 memset(&r->u8[8], 0, 8); \
1690 memset(&r->u8[0], 0, 8 - es); \
1691 }
1692 #else
1693 #define VEXTRACT(suffix, element) \
1694 void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1695 { \
1696 uint32_t es = sizeof(r->element[0]); \
1697 uint32_t s = (16 - index) - es; \
1698 memmove(&r->u8[8], &b->u8[s], es); \
1699 memset(&r->u8[0], 0, 8); \
1700 memset(&r->u8[8 + es], 0, 8 - es); \
1701 }
1702 #endif
1703 VEXTRACT(ub, u8)
1704 VEXTRACT(uh, u16)
1705 VEXTRACT(uw, u32)
1706 VEXTRACT(d, u64)
1707 #undef VEXTRACT
1708
1709 void helper_xxextractuw(CPUPPCState *env, ppc_vsr_t *xt,
1710 ppc_vsr_t *xb, uint32_t index)
1711 {
1712 ppc_vsr_t t = { };
1713 size_t es = sizeof(uint32_t);
1714 uint32_t ext_index;
1715 int i;
1716
1717 ext_index = index;
1718 for (i = 0; i < es; i++, ext_index++) {
1719 t.VsrB(8 - es + i) = xb->VsrB(ext_index % 16);
1720 }
1721
1722 *xt = t;
1723 }
1724
1725 void helper_xxinsertw(CPUPPCState *env, ppc_vsr_t *xt,
1726 ppc_vsr_t *xb, uint32_t index)
1727 {
1728 ppc_vsr_t t = *xt;
1729 size_t es = sizeof(uint32_t);
1730 int ins_index, i = 0;
1731
1732 ins_index = index;
1733 for (i = 0; i < es && ins_index < 16; i++, ins_index++) {
1734 t.VsrB(ins_index) = xb->VsrB(8 - es + i);
1735 }
1736
1737 *xt = t;
1738 }
1739
1740 #define XXBLEND(name, sz) \
1741 void glue(helper_XXBLENDV, name)(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, \
1742 ppc_avr_t *c, uint32_t desc) \
1743 { \
1744 for (int i = 0; i < ARRAY_SIZE(t->glue(u, sz)); i++) { \
1745 t->glue(u, sz)[i] = (c->glue(s, sz)[i] >> (sz - 1)) ? \
1746 b->glue(u, sz)[i] : a->glue(u, sz)[i]; \
1747 } \
1748 }
1749 XXBLEND(B, 8)
1750 XXBLEND(H, 16)
1751 XXBLEND(W, 32)
1752 XXBLEND(D, 64)
1753 #undef XXBLEND
1754
1755 #define VEXT_SIGNED(name, element, cast) \
1756 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \
1757 { \
1758 int i; \
1759 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1760 r->element[i] = (cast)b->element[i]; \
1761 } \
1762 }
1763 VEXT_SIGNED(vextsb2w, s32, int8_t)
1764 VEXT_SIGNED(vextsb2d, s64, int8_t)
1765 VEXT_SIGNED(vextsh2w, s32, int16_t)
1766 VEXT_SIGNED(vextsh2d, s64, int16_t)
1767 VEXT_SIGNED(vextsw2d, s64, int32_t)
1768 #undef VEXT_SIGNED
1769
1770 #define VNEG(name, element) \
1771 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \
1772 { \
1773 int i; \
1774 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1775 r->element[i] = -b->element[i]; \
1776 } \
1777 }
1778 VNEG(vnegw, s32)
1779 VNEG(vnegd, s64)
1780 #undef VNEG
1781
1782 void helper_vsro(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1783 {
1784 int sh = (b->VsrB(0xf) >> 3) & 0xf;
1785
1786 #if defined(HOST_WORDS_BIGENDIAN)
1787 memmove(&r->u8[sh], &a->u8[0], 16 - sh);
1788 memset(&r->u8[0], 0, sh);
1789 #else
1790 memmove(&r->u8[0], &a->u8[sh], 16 - sh);
1791 memset(&r->u8[16 - sh], 0, sh);
1792 #endif
1793 }
1794
1795 void helper_vsubcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1796 {
1797 int i;
1798
1799 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
1800 r->u32[i] = a->u32[i] >= b->u32[i];
1801 }
1802 }
1803
1804 void helper_vsumsws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1805 {
1806 int64_t t;
1807 int i, upper;
1808 ppc_avr_t result;
1809 int sat = 0;
1810
1811 upper = ARRAY_SIZE(r->s32) - 1;
1812 t = (int64_t)b->VsrSW(upper);
1813 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
1814 t += a->VsrSW(i);
1815 result.VsrSW(i) = 0;
1816 }
1817 result.VsrSW(upper) = cvtsdsw(t, &sat);
1818 *r = result;
1819
1820 if (sat) {
1821 set_vscr_sat(env);
1822 }
1823 }
1824
1825 void helper_vsum2sws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1826 {
1827 int i, j, upper;
1828 ppc_avr_t result;
1829 int sat = 0;
1830
1831 upper = 1;
1832 for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
1833 int64_t t = (int64_t)b->VsrSW(upper + i * 2);
1834
1835 result.VsrD(i) = 0;
1836 for (j = 0; j < ARRAY_SIZE(r->u64); j++) {
1837 t += a->VsrSW(2 * i + j);
1838 }
1839 result.VsrSW(upper + i * 2) = cvtsdsw(t, &sat);
1840 }
1841
1842 *r = result;
1843 if (sat) {
1844 set_vscr_sat(env);
1845 }
1846 }
1847
1848 void helper_vsum4sbs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1849 {
1850 int i, j;
1851 int sat = 0;
1852
1853 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
1854 int64_t t = (int64_t)b->s32[i];
1855
1856 for (j = 0; j < ARRAY_SIZE(r->s32); j++) {
1857 t += a->s8[4 * i + j];
1858 }
1859 r->s32[i] = cvtsdsw(t, &sat);
1860 }
1861
1862 if (sat) {
1863 set_vscr_sat(env);
1864 }
1865 }
1866
1867 void helper_vsum4shs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1868 {
1869 int sat = 0;
1870 int i;
1871
1872 for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
1873 int64_t t = (int64_t)b->s32[i];
1874
1875 t += a->s16[2 * i] + a->s16[2 * i + 1];
1876 r->s32[i] = cvtsdsw(t, &sat);
1877 }
1878
1879 if (sat) {
1880 set_vscr_sat(env);
1881 }
1882 }
1883
1884 void helper_vsum4ubs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1885 {
1886 int i, j;
1887 int sat = 0;
1888
1889 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
1890 uint64_t t = (uint64_t)b->u32[i];
1891
1892 for (j = 0; j < ARRAY_SIZE(r->u32); j++) {
1893 t += a->u8[4 * i + j];
1894 }
1895 r->u32[i] = cvtuduw(t, &sat);
1896 }
1897
1898 if (sat) {
1899 set_vscr_sat(env);
1900 }
1901 }
1902
1903 #if defined(HOST_WORDS_BIGENDIAN)
1904 #define UPKHI 1
1905 #define UPKLO 0
1906 #else
1907 #define UPKHI 0
1908 #define UPKLO 1
1909 #endif
1910 #define VUPKPX(suffix, hi) \
1911 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \
1912 { \
1913 int i; \
1914 ppc_avr_t result; \
1915 \
1916 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \
1917 uint16_t e = b->u16[hi ? i : i + 4]; \
1918 uint8_t a = (e >> 15) ? 0xff : 0; \
1919 uint8_t r = (e >> 10) & 0x1f; \
1920 uint8_t g = (e >> 5) & 0x1f; \
1921 uint8_t b = e & 0x1f; \
1922 \
1923 result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \
1924 } \
1925 *r = result; \
1926 }
1927 VUPKPX(lpx, UPKLO)
1928 VUPKPX(hpx, UPKHI)
1929 #undef VUPKPX
1930
1931 #define VUPK(suffix, unpacked, packee, hi) \
1932 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \
1933 { \
1934 int i; \
1935 ppc_avr_t result; \
1936 \
1937 if (hi) { \
1938 for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \
1939 result.unpacked[i] = b->packee[i]; \
1940 } \
1941 } else { \
1942 for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); \
1943 i++) { \
1944 result.unpacked[i - ARRAY_SIZE(r->unpacked)] = b->packee[i]; \
1945 } \
1946 } \
1947 *r = result; \
1948 }
1949 VUPK(hsb, s16, s8, UPKHI)
1950 VUPK(hsh, s32, s16, UPKHI)
1951 VUPK(hsw, s64, s32, UPKHI)
1952 VUPK(lsb, s16, s8, UPKLO)
1953 VUPK(lsh, s32, s16, UPKLO)
1954 VUPK(lsw, s64, s32, UPKLO)
1955 #undef VUPK
1956 #undef UPKHI
1957 #undef UPKLO
1958
1959 #define VGENERIC_DO(name, element) \
1960 void helper_v##name(ppc_avr_t *r, ppc_avr_t *b) \
1961 { \
1962 int i; \
1963 \
1964 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1965 r->element[i] = name(b->element[i]); \
1966 } \
1967 }
1968
1969 #define clzb(v) ((v) ? clz32((uint32_t)(v) << 24) : 8)
1970 #define clzh(v) ((v) ? clz32((uint32_t)(v) << 16) : 16)
1971
1972 VGENERIC_DO(clzb, u8)
1973 VGENERIC_DO(clzh, u16)
1974
1975 #undef clzb
1976 #undef clzh
1977
1978 #define ctzb(v) ((v) ? ctz32(v) : 8)
1979 #define ctzh(v) ((v) ? ctz32(v) : 16)
1980 #define ctzw(v) ctz32((v))
1981 #define ctzd(v) ctz64((v))
1982
1983 VGENERIC_DO(ctzb, u8)
1984 VGENERIC_DO(ctzh, u16)
1985 VGENERIC_DO(ctzw, u32)
1986 VGENERIC_DO(ctzd, u64)
1987
1988 #undef ctzb
1989 #undef ctzh
1990 #undef ctzw
1991 #undef ctzd
1992
1993 #define popcntb(v) ctpop8(v)
1994 #define popcnth(v) ctpop16(v)
1995 #define popcntw(v) ctpop32(v)
1996 #define popcntd(v) ctpop64(v)
1997
1998 VGENERIC_DO(popcntb, u8)
1999 VGENERIC_DO(popcnth, u16)
2000 VGENERIC_DO(popcntw, u32)
2001 VGENERIC_DO(popcntd, u64)
2002
2003 #undef popcntb
2004 #undef popcnth
2005 #undef popcntw
2006 #undef popcntd
2007
2008 #undef VGENERIC_DO
2009
2010 #if defined(HOST_WORDS_BIGENDIAN)
2011 #define QW_ONE { .u64 = { 0, 1 } }
2012 #else
2013 #define QW_ONE { .u64 = { 1, 0 } }
2014 #endif
2015
2016 #ifndef CONFIG_INT128
2017
2018 static inline void avr_qw_not(ppc_avr_t *t, ppc_avr_t a)
2019 {
2020 t->u64[0] = ~a.u64[0];
2021 t->u64[1] = ~a.u64[1];
2022 }
2023
2024 static int avr_qw_cmpu(ppc_avr_t a, ppc_avr_t b)
2025 {
2026 if (a.VsrD(0) < b.VsrD(0)) {
2027 return -1;
2028 } else if (a.VsrD(0) > b.VsrD(0)) {
2029 return 1;
2030 } else if (a.VsrD(1) < b.VsrD(1)) {
2031 return -1;
2032 } else if (a.VsrD(1) > b.VsrD(1)) {
2033 return 1;
2034 } else {
2035 return 0;
2036 }
2037 }
2038
2039 static void avr_qw_add(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b)
2040 {
2041 t->VsrD(1) = a.VsrD(1) + b.VsrD(1);
2042 t->VsrD(0) = a.VsrD(0) + b.VsrD(0) +
2043 (~a.VsrD(1) < b.VsrD(1));
2044 }
2045
2046 static int avr_qw_addc(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b)
2047 {
2048 ppc_avr_t not_a;
2049 t->VsrD(1) = a.VsrD(1) + b.VsrD(1);
2050 t->VsrD(0) = a.VsrD(0) + b.VsrD(0) +
2051 (~a.VsrD(1) < b.VsrD(1));
2052 avr_qw_not(&not_a, a);
2053 return avr_qw_cmpu(not_a, b) < 0;
2054 }
2055
2056 #endif
2057
2058 void helper_vadduqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2059 {
2060 #ifdef CONFIG_INT128
2061 r->u128 = a->u128 + b->u128;
2062 #else
2063 avr_qw_add(r, *a, *b);
2064 #endif
2065 }
2066
2067 void helper_vaddeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2068 {
2069 #ifdef CONFIG_INT128
2070 r->u128 = a->u128 + b->u128 + (c->u128 & 1);
2071 #else
2072
2073 if (c->VsrD(1) & 1) {
2074 ppc_avr_t tmp;
2075
2076 tmp.VsrD(0) = 0;
2077 tmp.VsrD(1) = c->VsrD(1) & 1;
2078 avr_qw_add(&tmp, *a, tmp);
2079 avr_qw_add(r, tmp, *b);
2080 } else {
2081 avr_qw_add(r, *a, *b);
2082 }
2083 #endif
2084 }
2085
2086 void helper_vaddcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2087 {
2088 #ifdef CONFIG_INT128
2089 r->u128 = (~a->u128 < b->u128);
2090 #else
2091 ppc_avr_t not_a;
2092
2093 avr_qw_not(&not_a, *a);
2094
2095 r->VsrD(0) = 0;
2096 r->VsrD(1) = (avr_qw_cmpu(not_a, *b) < 0);
2097 #endif
2098 }
2099
2100 void helper_vaddecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2101 {
2102 #ifdef CONFIG_INT128
2103 int carry_out = (~a->u128 < b->u128);
2104 if (!carry_out && (c->u128 & 1)) {
2105 carry_out = ((a->u128 + b->u128 + 1) == 0) &&
2106 ((a->u128 != 0) || (b->u128 != 0));
2107 }
2108 r->u128 = carry_out;
2109 #else
2110
2111 int carry_in = c->VsrD(1) & 1;
2112 int carry_out = 0;
2113 ppc_avr_t tmp;
2114
2115 carry_out = avr_qw_addc(&tmp, *a, *b);
2116
2117 if (!carry_out && carry_in) {
2118 ppc_avr_t one = QW_ONE;
2119 carry_out = avr_qw_addc(&tmp, tmp, one);
2120 }
2121 r->VsrD(0) = 0;
2122 r->VsrD(1) = carry_out;
2123 #endif
2124 }
2125
2126 void helper_vsubuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2127 {
2128 #ifdef CONFIG_INT128
2129 r->u128 = a->u128 - b->u128;
2130 #else
2131 ppc_avr_t tmp;
2132 ppc_avr_t one = QW_ONE;
2133
2134 avr_qw_not(&tmp, *b);
2135 avr_qw_add(&tmp, *a, tmp);
2136 avr_qw_add(r, tmp, one);
2137 #endif
2138 }
2139
2140 void helper_vsubeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2141 {
2142 #ifdef CONFIG_INT128
2143 r->u128 = a->u128 + ~b->u128 + (c->u128 & 1);
2144 #else
2145 ppc_avr_t tmp, sum;
2146
2147 avr_qw_not(&tmp, *b);
2148 avr_qw_add(&sum, *a, tmp);
2149
2150 tmp.VsrD(0) = 0;
2151 tmp.VsrD(1) = c->VsrD(1) & 1;
2152 avr_qw_add(r, sum, tmp);
2153 #endif
2154 }
2155
2156 void helper_vsubcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2157 {
2158 #ifdef CONFIG_INT128
2159 r->u128 = (~a->u128 < ~b->u128) ||
2160 (a->u128 + ~b->u128 == (__uint128_t)-1);
2161 #else
2162 int carry = (avr_qw_cmpu(*a, *b) > 0);
2163 if (!carry) {
2164 ppc_avr_t tmp;
2165 avr_qw_not(&tmp, *b);
2166 avr_qw_add(&tmp, *a, tmp);
2167 carry = ((tmp.VsrSD(0) == -1ull) && (tmp.VsrSD(1) == -1ull));
2168 }
2169 r->VsrD(0) = 0;
2170 r->VsrD(1) = carry;
2171 #endif
2172 }
2173
2174 void helper_vsubecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2175 {
2176 #ifdef CONFIG_INT128
2177 r->u128 =
2178 (~a->u128 < ~b->u128) ||
2179 ((c->u128 & 1) && (a->u128 + ~b->u128 == (__uint128_t)-1));
2180 #else
2181 int carry_in = c->VsrD(1) & 1;
2182 int carry_out = (avr_qw_cmpu(*a, *b) > 0);
2183 if (!carry_out && carry_in) {
2184 ppc_avr_t tmp;
2185 avr_qw_not(&tmp, *b);
2186 avr_qw_add(&tmp, *a, tmp);
2187 carry_out = ((tmp.VsrD(0) == -1ull) && (tmp.VsrD(1) == -1ull));
2188 }
2189
2190 r->VsrD(0) = 0;
2191 r->VsrD(1) = carry_out;
2192 #endif
2193 }
2194
2195 #define BCD_PLUS_PREF_1 0xC
2196 #define BCD_PLUS_PREF_2 0xF
2197 #define BCD_PLUS_ALT_1 0xA
2198 #define BCD_NEG_PREF 0xD
2199 #define BCD_NEG_ALT 0xB
2200 #define BCD_PLUS_ALT_2 0xE
2201 #define NATIONAL_PLUS 0x2B
2202 #define NATIONAL_NEG 0x2D
2203
2204 #define BCD_DIG_BYTE(n) (15 - ((n) / 2))
2205
2206 static int bcd_get_sgn(ppc_avr_t *bcd)
2207 {
2208 switch (bcd->VsrB(BCD_DIG_BYTE(0)) & 0xF) {
2209 case BCD_PLUS_PREF_1:
2210 case BCD_PLUS_PREF_2:
2211 case BCD_PLUS_ALT_1:
2212 case BCD_PLUS_ALT_2:
2213 {
2214 return 1;
2215 }
2216
2217 case BCD_NEG_PREF:
2218 case BCD_NEG_ALT:
2219 {
2220 return -1;
2221 }
2222
2223 default:
2224 {
2225 return 0;
2226 }
2227 }
2228 }
2229
2230 static int bcd_preferred_sgn(int sgn, int ps)
2231 {
2232 if (sgn >= 0) {
2233 return (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2;
2234 } else {
2235 return BCD_NEG_PREF;
2236 }
2237 }
2238
2239 static uint8_t bcd_get_digit(ppc_avr_t *bcd, int n, int *invalid)
2240 {
2241 uint8_t result;
2242 if (n & 1) {
2243 result = bcd->VsrB(BCD_DIG_BYTE(n)) >> 4;
2244 } else {
2245 result = bcd->VsrB(BCD_DIG_BYTE(n)) & 0xF;
2246 }
2247
2248 if (unlikely(result > 9)) {
2249 *invalid = true;
2250 }
2251 return result;
2252 }
2253
2254 static void bcd_put_digit(ppc_avr_t *bcd, uint8_t digit, int n)
2255 {
2256 if (n & 1) {
2257 bcd->VsrB(BCD_DIG_BYTE(n)) &= 0x0F;
2258 bcd->VsrB(BCD_DIG_BYTE(n)) |= (digit << 4);
2259 } else {
2260 bcd->VsrB(BCD_DIG_BYTE(n)) &= 0xF0;
2261 bcd->VsrB(BCD_DIG_BYTE(n)) |= digit;
2262 }
2263 }
2264
2265 static bool bcd_is_valid(ppc_avr_t *bcd)
2266 {
2267 int i;
2268 int invalid = 0;
2269
2270 if (bcd_get_sgn(bcd) == 0) {
2271 return false;
2272 }
2273
2274 for (i = 1; i < 32; i++) {
2275 bcd_get_digit(bcd, i, &invalid);
2276 if (unlikely(invalid)) {
2277 return false;
2278 }
2279 }
2280 return true;
2281 }
2282
2283 static int bcd_cmp_zero(ppc_avr_t *bcd)
2284 {
2285 if (bcd->VsrD(0) == 0 && (bcd->VsrD(1) >> 4) == 0) {
2286 return CRF_EQ;
2287 } else {
2288 return (bcd_get_sgn(bcd) == 1) ? CRF_GT : CRF_LT;
2289 }
2290 }
2291
2292 static uint16_t get_national_digit(ppc_avr_t *reg, int n)
2293 {
2294 return reg->VsrH(7 - n);
2295 }
2296
2297 static void set_national_digit(ppc_avr_t *reg, uint8_t val, int n)
2298 {
2299 reg->VsrH(7 - n) = val;
2300 }
2301
2302 static int bcd_cmp_mag(ppc_avr_t *a, ppc_avr_t *b)
2303 {
2304 int i;
2305 int invalid = 0;
2306 for (i = 31; i > 0; i--) {
2307 uint8_t dig_a = bcd_get_digit(a, i, &invalid);
2308 uint8_t dig_b = bcd_get_digit(b, i, &invalid);
2309 if (unlikely(invalid)) {
2310 return 0; /* doesn't matter */
2311 } else if (dig_a > dig_b) {
2312 return 1;
2313 } else if (dig_a < dig_b) {
2314 return -1;
2315 }
2316 }
2317
2318 return 0;
2319 }
2320
2321 static int bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid,
2322 int *overflow)
2323 {
2324 int carry = 0;
2325 int i;
2326 int is_zero = 1;
2327
2328 for (i = 1; i <= 31; i++) {
2329 uint8_t digit = bcd_get_digit(a, i, invalid) +
2330 bcd_get_digit(b, i, invalid) + carry;
2331 is_zero &= (digit == 0);
2332 if (digit > 9) {
2333 carry = 1;
2334 digit -= 10;
2335 } else {
2336 carry = 0;
2337 }
2338
2339 bcd_put_digit(t, digit, i);
2340 }
2341
2342 *overflow = carry;
2343 return is_zero;
2344 }
2345
2346 static void bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid,
2347 int *overflow)
2348 {
2349 int carry = 0;
2350 int i;
2351
2352 for (i = 1; i <= 31; i++) {
2353 uint8_t digit = bcd_get_digit(a, i, invalid) -
2354 bcd_get_digit(b, i, invalid) + carry;
2355 if (digit & 0x80) {
2356 carry = -1;
2357 digit += 10;
2358 } else {
2359 carry = 0;
2360 }
2361
2362 bcd_put_digit(t, digit, i);
2363 }
2364
2365 *overflow = carry;
2366 }
2367
2368 uint32_t helper_bcdadd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2369 {
2370
2371 int sgna = bcd_get_sgn(a);
2372 int sgnb = bcd_get_sgn(b);
2373 int invalid = (sgna == 0) || (sgnb == 0);
2374 int overflow = 0;
2375 int zero = 0;
2376 uint32_t cr = 0;
2377 ppc_avr_t result = { .u64 = { 0, 0 } };
2378
2379 if (!invalid) {
2380 if (sgna == sgnb) {
2381 result.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(sgna, ps);
2382 zero = bcd_add_mag(&result, a, b, &invalid, &overflow);
2383 cr = (sgna > 0) ? CRF_GT : CRF_LT;
2384 } else {
2385 int magnitude = bcd_cmp_mag(a, b);
2386 if (magnitude > 0) {
2387 result.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(sgna, ps);
2388 bcd_sub_mag(&result, a, b, &invalid, &overflow);
2389 cr = (sgna > 0) ? CRF_GT : CRF_LT;
2390 } else if (magnitude < 0) {
2391 result.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(sgnb, ps);
2392 bcd_sub_mag(&result, b, a, &invalid, &overflow);
2393 cr = (sgnb > 0) ? CRF_GT : CRF_LT;
2394 } else {
2395 result.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(0, ps);
2396 cr = CRF_EQ;
2397 }
2398 }
2399 }
2400
2401 if (unlikely(invalid)) {
2402 result.VsrD(0) = result.VsrD(1) = -1;
2403 cr = CRF_SO;
2404 } else if (overflow) {
2405 cr |= CRF_SO;
2406 } else if (zero) {
2407 cr |= CRF_EQ;
2408 }
2409
2410 *r = result;
2411
2412 return cr;
2413 }
2414
2415 uint32_t helper_bcdsub(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2416 {
2417 ppc_avr_t bcopy = *b;
2418 int sgnb = bcd_get_sgn(b);
2419 if (sgnb < 0) {
2420 bcd_put_digit(&bcopy, BCD_PLUS_PREF_1, 0);
2421 } else if (sgnb > 0) {
2422 bcd_put_digit(&bcopy, BCD_NEG_PREF, 0);
2423 }
2424 /* else invalid ... defer to bcdadd code for proper handling */
2425
2426 return helper_bcdadd(r, a, &bcopy, ps);
2427 }
2428
2429 uint32_t helper_bcdcfn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2430 {
2431 int i;
2432 int cr = 0;
2433 uint16_t national = 0;
2434 uint16_t sgnb = get_national_digit(b, 0);
2435 ppc_avr_t ret = { .u64 = { 0, 0 } };
2436 int invalid = (sgnb != NATIONAL_PLUS && sgnb != NATIONAL_NEG);
2437
2438 for (i = 1; i < 8; i++) {
2439 national = get_national_digit(b, i);
2440 if (unlikely(national < 0x30 || national > 0x39)) {
2441 invalid = 1;
2442 break;
2443 }
2444
2445 bcd_put_digit(&ret, national & 0xf, i);
2446 }
2447
2448 if (sgnb == NATIONAL_PLUS) {
2449 bcd_put_digit(&ret, (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2, 0);
2450 } else {
2451 bcd_put_digit(&ret, BCD_NEG_PREF, 0);
2452 }
2453
2454 cr = bcd_cmp_zero(&ret);
2455
2456 if (unlikely(invalid)) {
2457 cr = CRF_SO;
2458 }
2459
2460 *r = ret;
2461
2462 return cr;
2463 }
2464
2465 uint32_t helper_bcdctn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2466 {
2467 int i;
2468 int cr = 0;
2469 int sgnb = bcd_get_sgn(b);
2470 int invalid = (sgnb == 0);
2471 ppc_avr_t ret = { .u64 = { 0, 0 } };
2472
2473 int ox_flag = (b->VsrD(0) != 0) || ((b->VsrD(1) >> 32) != 0);
2474
2475 for (i = 1; i < 8; i++) {
2476 set_national_digit(&ret, 0x30 + bcd_get_digit(b, i, &invalid), i);
2477
2478 if (unlikely(invalid)) {
2479 break;
2480 }
2481 }
2482 set_national_digit(&ret, (sgnb == -1) ? NATIONAL_NEG : NATIONAL_PLUS, 0);
2483
2484 cr = bcd_cmp_zero(b);
2485
2486 if (ox_flag) {
2487 cr |= CRF_SO;
2488 }
2489
2490 if (unlikely(invalid)) {
2491 cr = CRF_SO;
2492 }
2493
2494 *r = ret;
2495
2496 return cr;
2497 }
2498
2499 uint32_t helper_bcdcfz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2500 {
2501 int i;
2502 int cr = 0;
2503 int invalid = 0;
2504 int zone_digit = 0;
2505 int zone_lead = ps ? 0xF : 0x3;
2506 int digit = 0;
2507 ppc_avr_t ret = { .u64 = { 0, 0 } };
2508 int sgnb = b->VsrB(BCD_DIG_BYTE(0)) >> 4;
2509
2510 if (unlikely((sgnb < 0xA) && ps)) {
2511 invalid = 1;
2512 }
2513
2514 for (i = 0; i < 16; i++) {
2515 zone_digit = i ? b->VsrB(BCD_DIG_BYTE(i * 2)) >> 4 : zone_lead;
2516 digit = b->VsrB(BCD_DIG_BYTE(i * 2)) & 0xF;
2517 if (unlikely(zone_digit != zone_lead || digit > 0x9)) {
2518 invalid = 1;
2519 break;
2520 }
2521
2522 bcd_put_digit(&ret, digit, i + 1);
2523 }
2524
2525 if ((ps && (sgnb == 0xB || sgnb == 0xD)) ||
2526 (!ps && (sgnb & 0x4))) {
2527 bcd_put_digit(&ret, BCD_NEG_PREF, 0);
2528 } else {
2529 bcd_put_digit(&ret, BCD_PLUS_PREF_1, 0);
2530 }
2531
2532 cr = bcd_cmp_zero(&ret);
2533
2534 if (unlikely(invalid)) {
2535 cr = CRF_SO;
2536 }
2537
2538 *r = ret;
2539
2540 return cr;
2541 }
2542
2543 uint32_t helper_bcdctz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2544 {
2545 int i;
2546 int cr = 0;
2547 uint8_t digit = 0;
2548 int sgnb = bcd_get_sgn(b);
2549 int zone_lead = (ps) ? 0xF0 : 0x30;
2550 int invalid = (sgnb == 0);
2551 ppc_avr_t ret = { .u64 = { 0, 0 } };
2552
2553 int ox_flag = ((b->VsrD(0) >> 4) != 0);
2554
2555 for (i = 0; i < 16; i++) {
2556 digit = bcd_get_digit(b, i + 1, &invalid);
2557
2558 if (unlikely(invalid)) {
2559 break;
2560 }
2561
2562 ret.VsrB(BCD_DIG_BYTE(i * 2)) = zone_lead + digit;
2563 }
2564
2565 if (ps) {
2566 bcd_put_digit(&ret, (sgnb == 1) ? 0xC : 0xD, 1);
2567 } else {
2568 bcd_put_digit(&ret, (sgnb == 1) ? 0x3 : 0x7, 1);
2569 }
2570
2571 cr = bcd_cmp_zero(b);
2572
2573 if (ox_flag) {
2574 cr |= CRF_SO;
2575 }
2576
2577 if (unlikely(invalid)) {
2578 cr = CRF_SO;
2579 }
2580
2581 *r = ret;
2582
2583 return cr;
2584 }
2585
2586 /**
2587 * Compare 2 128-bit unsigned integers, passed in as unsigned 64-bit pairs
2588 *
2589 * Returns:
2590 * > 0 if ahi|alo > bhi|blo,
2591 * 0 if ahi|alo == bhi|blo,
2592 * < 0 if ahi|alo < bhi|blo
2593 */
2594 static inline int ucmp128(uint64_t alo, uint64_t ahi,
2595 uint64_t blo, uint64_t bhi)
2596 {
2597 return (ahi == bhi) ?
2598 (alo > blo ? 1 : (alo == blo ? 0 : -1)) :
2599 (ahi > bhi ? 1 : -1);
2600 }
2601
2602 uint32_t helper_bcdcfsq(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2603 {
2604 int i;
2605 int cr;
2606 uint64_t lo_value;
2607 uint64_t hi_value;
2608 uint64_t rem;
2609 ppc_avr_t ret = { .u64 = { 0, 0 } };
2610
2611 if (b->VsrSD(0) < 0) {
2612 lo_value = -b->VsrSD(1);
2613 hi_value = ~b->VsrD(0) + !lo_value;
2614 bcd_put_digit(&ret, 0xD, 0);
2615
2616 cr = CRF_LT;
2617 } else {
2618 lo_value = b->VsrD(1);
2619 hi_value = b->VsrD(0);
2620 bcd_put_digit(&ret, bcd_preferred_sgn(0, ps), 0);
2621
2622 if (hi_value == 0 && lo_value == 0) {
2623 cr = CRF_EQ;
2624 } else {
2625 cr = CRF_GT;
2626 }
2627 }
2628
2629 /*
2630 * Check src limits: abs(src) <= 10^31 - 1
2631 *
2632 * 10^31 - 1 = 0x0000007e37be2022 c0914b267fffffff
2633 */
2634 if (ucmp128(lo_value, hi_value,
2635 0xc0914b267fffffffULL, 0x7e37be2022ULL) > 0) {
2636 cr |= CRF_SO;
2637
2638 /*
2639 * According to the ISA, if src wouldn't fit in the destination
2640 * register, the result is undefined.
2641 * In that case, we leave r unchanged.
2642 */
2643 } else {
2644 rem = divu128(&lo_value, &hi_value, 1000000000000000ULL);
2645
2646 for (i = 1; i < 16; rem /= 10, i++) {
2647 bcd_put_digit(&ret, rem % 10, i);
2648 }
2649
2650 for (; i < 32; lo_value /= 10, i++) {
2651 bcd_put_digit(&ret, lo_value % 10, i);
2652 }
2653
2654 *r = ret;
2655 }
2656
2657 return cr;
2658 }
2659
2660 uint32_t helper_bcdctsq(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2661 {
2662 uint8_t i;
2663 int cr;
2664 uint64_t carry;
2665 uint64_t unused;
2666 uint64_t lo_value;
2667 uint64_t hi_value = 0;
2668 int sgnb = bcd_get_sgn(b);
2669 int invalid = (sgnb == 0);
2670
2671 lo_value = bcd_get_digit(b, 31, &invalid);
2672 for (i = 30; i > 0; i--) {
2673 mulu64(&lo_value, &carry, lo_value, 10ULL);
2674 mulu64(&hi_value, &unused, hi_value, 10ULL);
2675 lo_value += bcd_get_digit(b, i, &invalid);
2676 hi_value += carry;
2677
2678 if (unlikely(invalid)) {
2679 break;
2680 }
2681 }
2682
2683 if (sgnb == -1) {
2684 r->VsrSD(1) = -lo_value;
2685 r->VsrSD(0) = ~hi_value + !r->VsrSD(1);
2686 } else {
2687 r->VsrSD(1) = lo_value;
2688 r->VsrSD(0) = hi_value;
2689 }
2690
2691 cr = bcd_cmp_zero(b);
2692
2693 if (unlikely(invalid)) {
2694 cr = CRF_SO;
2695 }
2696
2697 return cr;
2698 }
2699
2700 uint32_t helper_bcdcpsgn(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2701 {
2702 int i;
2703 int invalid = 0;
2704
2705 if (bcd_get_sgn(a) == 0 || bcd_get_sgn(b) == 0) {
2706 return CRF_SO;
2707 }
2708
2709 *r = *a;
2710 bcd_put_digit(r, b->VsrB(BCD_DIG_BYTE(0)) & 0xF, 0);
2711
2712 for (i = 1; i < 32; i++) {
2713 bcd_get_digit(a, i, &invalid);
2714 bcd_get_digit(b, i, &invalid);
2715 if (unlikely(invalid)) {
2716 return CRF_SO;
2717 }
2718 }
2719
2720 return bcd_cmp_zero(r);
2721 }
2722
2723 uint32_t helper_bcdsetsgn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2724 {
2725 int sgnb = bcd_get_sgn(b);
2726
2727 *r = *b;
2728 bcd_put_digit(r, bcd_preferred_sgn(sgnb, ps), 0);
2729
2730 if (bcd_is_valid(b) == false) {
2731 return CRF_SO;
2732 }
2733
2734 return bcd_cmp_zero(r);
2735 }
2736
2737 uint32_t helper_bcds(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2738 {
2739 int cr;
2740 int i = a->VsrSB(7);
2741 bool ox_flag = false;
2742 int sgnb = bcd_get_sgn(b);
2743 ppc_avr_t ret = *b;
2744 ret.VsrD(1) &= ~0xf;
2745
2746 if (bcd_is_valid(b) == false) {
2747 return CRF_SO;
2748 }
2749
2750 if (unlikely(i > 31)) {
2751 i = 31;
2752 } else if (unlikely(i < -31)) {
2753 i = -31;
2754 }
2755
2756 if (i > 0) {
2757 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag);
2758 } else {
2759 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4);
2760 }
2761 bcd_put_digit(&ret, bcd_preferred_sgn(sgnb, ps), 0);
2762
2763 *r = ret;
2764
2765 cr = bcd_cmp_zero(r);
2766 if (ox_flag) {
2767 cr |= CRF_SO;
2768 }
2769
2770 return cr;
2771 }
2772
2773 uint32_t helper_bcdus(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2774 {
2775 int cr;
2776 int i;
2777 int invalid = 0;
2778 bool ox_flag = false;
2779 ppc_avr_t ret = *b;
2780
2781 for (i = 0; i < 32; i++) {
2782 bcd_get_digit(b, i, &invalid);
2783
2784 if (unlikely(invalid)) {
2785 return CRF_SO;
2786 }
2787 }
2788
2789 i = a->VsrSB(7);
2790 if (i >= 32) {
2791 ox_flag = true;
2792 ret.VsrD(1) = ret.VsrD(0) = 0;
2793 } else if (i <= -32) {
2794 ret.VsrD(1) = ret.VsrD(0) = 0;
2795 } else if (i > 0) {
2796 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag);
2797 } else {
2798 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4);
2799 }
2800 *r = ret;
2801
2802 cr = bcd_cmp_zero(r);
2803 if (ox_flag) {
2804 cr |= CRF_SO;
2805 }
2806
2807 return cr;
2808 }
2809
2810 uint32_t helper_bcdsr(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2811 {
2812 int cr;
2813 int unused = 0;
2814 int invalid = 0;
2815 bool ox_flag = false;
2816 int sgnb = bcd_get_sgn(b);
2817 ppc_avr_t ret = *b;
2818 ret.VsrD(1) &= ~0xf;
2819
2820 int i = a->VsrSB(7);
2821 ppc_avr_t bcd_one;
2822
2823 bcd_one.VsrD(0) = 0;
2824 bcd_one.VsrD(1) = 0x10;
2825
2826 if (bcd_is_valid(b) == false) {
2827 return CRF_SO;
2828 }
2829
2830 if (unlikely(i > 31)) {
2831 i = 31;
2832 } else if (unlikely(i < -31)) {
2833 i = -31;
2834 }
2835
2836 if (i > 0) {
2837 ulshift(&ret.VsrD(1), &ret.VsrD(0), i * 4, &ox_flag);
2838 } else {
2839 urshift(&ret.VsrD(1), &ret.VsrD(0), -i * 4);
2840
2841 if (bcd_get_digit(&ret, 0, &invalid) >= 5) {
2842 bcd_add_mag(&ret, &ret, &bcd_one, &invalid, &unused);
2843 }
2844 }
2845 bcd_put_digit(&ret, bcd_preferred_sgn(sgnb, ps), 0);
2846
2847 cr = bcd_cmp_zero(&ret);
2848 if (ox_flag) {
2849 cr |= CRF_SO;
2850 }
2851 *r = ret;
2852
2853 return cr;
2854 }
2855
2856 uint32_t helper_bcdtrunc(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2857 {
2858 uint64_t mask;
2859 uint32_t ox_flag = 0;
2860 int i = a->VsrSH(3) + 1;
2861 ppc_avr_t ret = *b;
2862
2863 if (bcd_is_valid(b) == false) {
2864 return CRF_SO;
2865 }
2866
2867 if (i > 16 && i < 32) {
2868 mask = (uint64_t)-1 >> (128 - i * 4);
2869 if (ret.VsrD(0) & ~mask) {
2870 ox_flag = CRF_SO;
2871 }
2872
2873 ret.VsrD(0) &= mask;
2874 } else if (i >= 0 && i <= 16) {
2875 mask = (uint64_t)-1 >> (64 - i * 4);
2876 if (ret.VsrD(0) || (ret.VsrD(1) & ~mask)) {
2877 ox_flag = CRF_SO;
2878 }
2879
2880 ret.VsrD(1) &= mask;
2881 ret.VsrD(0) = 0;
2882 }
2883 bcd_put_digit(&ret, bcd_preferred_sgn(bcd_get_sgn(b), ps), 0);
2884 *r = ret;
2885
2886 return bcd_cmp_zero(&ret) | ox_flag;
2887 }
2888
2889 uint32_t helper_bcdutrunc(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2890 {
2891 int i;
2892 uint64_t mask;
2893 uint32_t ox_flag = 0;
2894 int invalid = 0;
2895 ppc_avr_t ret = *b;
2896
2897 for (i = 0; i < 32; i++) {
2898 bcd_get_digit(b, i, &invalid);
2899
2900 if (unlikely(invalid)) {
2901 return CRF_SO;
2902 }
2903 }
2904
2905 i = a->VsrSH(3);
2906 if (i > 16 && i < 33) {
2907 mask = (uint64_t)-1 >> (128 - i * 4);
2908 if (ret.VsrD(0) & ~mask) {
2909 ox_flag = CRF_SO;
2910 }
2911
2912 ret.VsrD(0) &= mask;
2913 } else if (i > 0 && i <= 16) {
2914 mask = (uint64_t)-1 >> (64 - i * 4);
2915 if (ret.VsrD(0) || (ret.VsrD(1) & ~mask)) {
2916 ox_flag = CRF_SO;
2917 }
2918
2919 ret.VsrD(1) &= mask;
2920 ret.VsrD(0) = 0;
2921 } else if (i == 0) {
2922 if (ret.VsrD(0) || ret.VsrD(1)) {
2923 ox_flag = CRF_SO;
2924 }
2925 ret.VsrD(0) = ret.VsrD(1) = 0;
2926 }
2927
2928 *r = ret;
2929 if (r->VsrD(0) == 0 && r->VsrD(1) == 0) {
2930 return ox_flag | CRF_EQ;
2931 }
2932
2933 return ox_flag | CRF_GT;
2934 }
2935
2936 void helper_vsbox(ppc_avr_t *r, ppc_avr_t *a)
2937 {
2938 int i;
2939 VECTOR_FOR_INORDER_I(i, u8) {
2940 r->u8[i] = AES_sbox[a->u8[i]];
2941 }
2942 }
2943
2944 void helper_vcipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2945 {
2946 ppc_avr_t result;
2947 int i;
2948
2949 VECTOR_FOR_INORDER_I(i, u32) {
2950 result.VsrW(i) = b->VsrW(i) ^
2951 (AES_Te0[a->VsrB(AES_shifts[4 * i + 0])] ^
2952 AES_Te1[a->VsrB(AES_shifts[4 * i + 1])] ^
2953 AES_Te2[a->VsrB(AES_shifts[4 * i + 2])] ^
2954 AES_Te3[a->VsrB(AES_shifts[4 * i + 3])]);
2955 }
2956 *r = result;
2957 }
2958
2959 void helper_vcipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2960 {
2961 ppc_avr_t result;
2962 int i;
2963
2964 VECTOR_FOR_INORDER_I(i, u8) {
2965 result.VsrB(i) = b->VsrB(i) ^ (AES_sbox[a->VsrB(AES_shifts[i])]);
2966 }
2967 *r = result;
2968 }
2969
2970 void helper_vncipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2971 {
2972 /* This differs from what is written in ISA V2.07. The RTL is */
2973 /* incorrect and will be fixed in V2.07B. */
2974 int i;
2975 ppc_avr_t tmp;
2976
2977 VECTOR_FOR_INORDER_I(i, u8) {
2978 tmp.VsrB(i) = b->VsrB(i) ^ AES_isbox[a->VsrB(AES_ishifts[i])];
2979 }
2980
2981 VECTOR_FOR_INORDER_I(i, u32) {
2982 r->VsrW(i) =
2983 AES_imc[tmp.VsrB(4 * i + 0)][0] ^
2984 AES_imc[tmp.VsrB(4 * i + 1)][1] ^
2985 AES_imc[tmp.VsrB(4 * i + 2)][2] ^
2986 AES_imc[tmp.VsrB(4 * i + 3)][3];
2987 }
2988 }
2989
2990 void helper_vncipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2991 {
2992 ppc_avr_t result;
2993 int i;
2994
2995 VECTOR_FOR_INORDER_I(i, u8) {
2996 result.VsrB(i) = b->VsrB(i) ^ (AES_isbox[a->VsrB(AES_ishifts[i])]);
2997 }
2998 *r = result;
2999 }
3000
3001 void helper_vshasigmaw(ppc_avr_t *r, ppc_avr_t *a, uint32_t st_six)
3002 {
3003 int st = (st_six & 0x10) != 0;
3004 int six = st_six & 0xF;
3005 int i;
3006
3007 for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
3008 if (st == 0) {
3009 if ((six & (0x8 >> i)) == 0) {
3010 r->VsrW(i) = ror32(a->VsrW(i), 7) ^
3011 ror32(a->VsrW(i), 18) ^
3012 (a->VsrW(i) >> 3);
3013 } else { /* six.bit[i] == 1 */
3014 r->VsrW(i) = ror32(a->VsrW(i), 17) ^
3015 ror32(a->VsrW(i), 19) ^
3016 (a->VsrW(i) >> 10);
3017 }
3018 } else { /* st == 1 */
3019 if ((six & (0x8 >> i)) == 0) {
3020 r->VsrW(i) = ror32(a->VsrW(i), 2) ^
3021 ror32(a->VsrW(i), 13) ^
3022 ror32(a->VsrW(i), 22);
3023 } else { /* six.bit[i] == 1 */
3024 r->VsrW(i) = ror32(a->VsrW(i), 6) ^
3025 ror32(a->VsrW(i), 11) ^
3026 ror32(a->VsrW(i), 25);
3027 }
3028 }
3029 }
3030 }
3031
3032 void helper_vshasigmad(ppc_avr_t *r, ppc_avr_t *a, uint32_t st_six)
3033 {
3034 int st = (st_six & 0x10) != 0;
3035 int six = st_six & 0xF;
3036 int i;
3037
3038 for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
3039 if (st == 0) {
3040 if ((six & (0x8 >> (2 * i))) == 0) {
3041 r->VsrD(i) = ror64(a->VsrD(i), 1) ^
3042 ror64(a->VsrD(i), 8) ^
3043 (a->VsrD(i) >> 7);
3044 } else { /* six.bit[2*i] == 1 */
3045 r->VsrD(i) = ror64(a->VsrD(i), 19) ^
3046 ror64(a->VsrD(i), 61) ^
3047 (a->VsrD(i) >> 6);
3048 }
3049 } else { /* st == 1 */
3050 if ((six & (0x8 >> (2 * i))) == 0) {
3051 r->VsrD(i) = ror64(a->VsrD(i), 28) ^
3052 ror64(a->VsrD(i), 34) ^
3053 ror64(a->VsrD(i), 39);
3054 } else { /* six.bit[2*i] == 1 */
3055 r->VsrD(i) = ror64(a->VsrD(i), 14) ^
3056 ror64(a->VsrD(i), 18) ^
3057 ror64(a->VsrD(i), 41);
3058 }
3059 }
3060 }
3061 }
3062
3063 void helper_vpermxor(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
3064 {
3065 ppc_avr_t result;
3066 int i;
3067
3068 for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
3069 int indexA = c->VsrB(i) >> 4;
3070 int indexB = c->VsrB(i) & 0xF;
3071
3072 result.VsrB(i) = a->VsrB(indexA) ^ b->VsrB(indexB);
3073 }
3074 *r = result;
3075 }
3076
3077 #undef VECTOR_FOR_INORDER_I
3078
3079 /*****************************************************************************/
3080 /* SPE extension helpers */
3081 /* Use a table to make this quicker */
3082 static const uint8_t hbrev[16] = {
3083 0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE,
3084 0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF,
3085 };
3086
3087 static inline uint8_t byte_reverse(uint8_t val)
3088 {
3089 return hbrev[val >> 4] | (hbrev[val & 0xF] << 4);
3090 }
3091
3092 static inline uint32_t word_reverse(uint32_t val)
3093 {
3094 return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) |
3095 (byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24);
3096 }
3097
3098 #define MASKBITS 16 /* Random value - to be fixed (implementation dependent) */
3099 target_ulong helper_brinc(target_ulong arg1, target_ulong arg2)
3100 {
3101 uint32_t a, b, d, mask;
3102
3103 mask = UINT32_MAX >> (32 - MASKBITS);
3104 a = arg1 & mask;
3105 b = arg2 & mask;
3106 d = word_reverse(1 + word_reverse(a | ~b));
3107 return (arg1 & ~mask) | (d & b);
3108 }
3109
3110 uint32_t helper_cntlsw32(uint32_t val)
3111 {
3112 if (val & 0x80000000) {
3113 return clz32(~val);
3114 } else {
3115 return clz32(val);
3116 }
3117 }
3118
3119 uint32_t helper_cntlzw32(uint32_t val)
3120 {
3121 return clz32(val);
3122 }
3123
3124 /* 440 specific */
3125 target_ulong helper_dlmzb(CPUPPCState *env, target_ulong high,
3126 target_ulong low, uint32_t update_Rc)
3127 {
3128 target_ulong mask;
3129 int i;
3130
3131 i = 1;
3132 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
3133 if ((high & mask) == 0) {
3134 if (update_Rc) {
3135 env->crf[0] = 0x4;
3136 }
3137 goto done;
3138 }
3139 i++;
3140 }
3141 for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
3142 if ((low & mask) == 0) {
3143 if (update_Rc) {
3144 env->crf[0] = 0x8;
3145 }
3146 goto done;
3147 }
3148 i++;
3149 }
3150 i = 8;
3151 if (update_Rc) {
3152 env->crf[0] = 0x2;
3153 }
3154 done:
3155 env->xer = (env->xer & ~0x7F) | i;
3156 if (update_Rc) {
3157 env->crf[0] |= xer_so;
3158 }
3159 return i;
3160 }
QEmu