xref: /openbmc/qemu/target/ppc/int_helper.c (revision f6a51c84)
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 "qemu/osdep.h"
20 #include "cpu.h"
21 #include "internal.h"
22 #include "exec/exec-all.h"
23 #include "qemu/host-utils.h"
24 #include "exec/helper-proto.h"
25 #include "crypto/aes.h"
26 
27 #include "helper_regs.h"
28 /*****************************************************************************/
29 /* Fixed point operations helpers */
30 
31 target_ulong helper_divweu(CPUPPCState *env, target_ulong ra, target_ulong rb,
32                            uint32_t oe)
33 {
34     uint64_t rt = 0;
35     int overflow = 0;
36 
37     uint64_t dividend = (uint64_t)ra << 32;
38     uint64_t divisor = (uint32_t)rb;
39 
40     if (unlikely(divisor == 0)) {
41         overflow = 1;
42     } else {
43         rt = dividend / divisor;
44         overflow = rt > UINT32_MAX;
45     }
46 
47     if (unlikely(overflow)) {
48         rt = 0; /* Undefined */
49     }
50 
51     if (oe) {
52         if (unlikely(overflow)) {
53             env->so = env->ov = 1;
54         } else {
55             env->ov = 0;
56         }
57     }
58 
59     return (target_ulong)rt;
60 }
61 
62 target_ulong helper_divwe(CPUPPCState *env, target_ulong ra, target_ulong rb,
63                           uint32_t oe)
64 {
65     int64_t rt = 0;
66     int overflow = 0;
67 
68     int64_t dividend = (int64_t)ra << 32;
69     int64_t divisor = (int64_t)((int32_t)rb);
70 
71     if (unlikely((divisor == 0) ||
72                  ((divisor == -1ull) && (dividend == INT64_MIN)))) {
73         overflow = 1;
74     } else {
75         rt = dividend / divisor;
76         overflow = rt != (int32_t)rt;
77     }
78 
79     if (unlikely(overflow)) {
80         rt = 0; /* Undefined */
81     }
82 
83     if (oe) {
84         if (unlikely(overflow)) {
85             env->so = env->ov = 1;
86         } else {
87             env->ov = 0;
88         }
89     }
90 
91     return (target_ulong)rt;
92 }
93 
94 #if defined(TARGET_PPC64)
95 
96 uint64_t helper_divdeu(CPUPPCState *env, uint64_t ra, uint64_t rb, uint32_t oe)
97 {
98     uint64_t rt = 0;
99     int overflow = 0;
100 
101     overflow = divu128(&rt, &ra, rb);
102 
103     if (unlikely(overflow)) {
104         rt = 0; /* Undefined */
105     }
106 
107     if (oe) {
108         if (unlikely(overflow)) {
109             env->so = env->ov = 1;
110         } else {
111             env->ov = 0;
112         }
113     }
114 
115     return rt;
116 }
117 
118 uint64_t helper_divde(CPUPPCState *env, uint64_t rau, uint64_t rbu, uint32_t oe)
119 {
120     int64_t rt = 0;
121     int64_t ra = (int64_t)rau;
122     int64_t rb = (int64_t)rbu;
123     int overflow = divs128(&rt, &ra, rb);
124 
125     if (unlikely(overflow)) {
126         rt = 0; /* Undefined */
127     }
128 
129     if (oe) {
130 
131         if (unlikely(overflow)) {
132             env->so = env->ov = 1;
133         } else {
134             env->ov = 0;
135         }
136     }
137 
138     return rt;
139 }
140 
141 #endif
142 
143 
144 target_ulong helper_cntlzw(target_ulong t)
145 {
146     return clz32(t);
147 }
148 
149 target_ulong helper_cnttzw(target_ulong t)
150 {
151     return ctz32(t);
152 }
153 
154 #if defined(TARGET_PPC64)
155 /* if x = 0xab, returns 0xababababababababa */
156 #define pattern(x) (((x) & 0xff) * (~(target_ulong)0 / 0xff))
157 
158 /* substract 1 from each byte, and with inverse, check if MSB is set at each
159  * byte.
160  * i.e. ((0x00 - 0x01) & ~(0x00)) & 0x80
161  *      (0xFF & 0xFF) & 0x80 = 0x80 (zero found)
162  */
163 #define haszero(v) (((v) - pattern(0x01)) & ~(v) & pattern(0x80))
164 
165 /* When you XOR the pattern and there is a match, that byte will be zero */
166 #define hasvalue(x, n)  (haszero((x) ^ pattern(n)))
167 
168 uint32_t helper_cmpeqb(target_ulong ra, target_ulong rb)
169 {
170     return hasvalue(rb, ra) ? 1 << CRF_GT : 0;
171 }
172 
173 #undef pattern
174 #undef haszero
175 #undef hasvalue
176 
177 target_ulong helper_cntlzd(target_ulong t)
178 {
179     return clz64(t);
180 }
181 
182 target_ulong helper_cnttzd(target_ulong t)
183 {
184     return ctz64(t);
185 }
186 
187 /* Return invalid random number.
188  *
189  * FIXME: Add rng backend or other mechanism to get cryptographically suitable
190  * random number
191  */
192 target_ulong helper_darn32(void)
193 {
194     return -1;
195 }
196 
197 target_ulong helper_darn64(void)
198 {
199     return -1;
200 }
201 
202 #endif
203 
204 #if defined(TARGET_PPC64)
205 
206 uint64_t helper_bpermd(uint64_t rs, uint64_t rb)
207 {
208     int i;
209     uint64_t ra = 0;
210 
211     for (i = 0; i < 8; i++) {
212         int index = (rs >> (i*8)) & 0xFF;
213         if (index < 64) {
214             if (rb & (1ull << (63-index))) {
215                 ra |= 1 << i;
216             }
217         }
218     }
219     return ra;
220 }
221 
222 #endif
223 
224 target_ulong helper_cmpb(target_ulong rs, target_ulong rb)
225 {
226     target_ulong mask = 0xff;
227     target_ulong ra = 0;
228     int i;
229 
230     for (i = 0; i < sizeof(target_ulong); i++) {
231         if ((rs & mask) == (rb & mask)) {
232             ra |= mask;
233         }
234         mask <<= 8;
235     }
236     return ra;
237 }
238 
239 /* shift right arithmetic helper */
240 target_ulong helper_sraw(CPUPPCState *env, target_ulong value,
241                          target_ulong shift)
242 {
243     int32_t ret;
244 
245     if (likely(!(shift & 0x20))) {
246         if (likely((uint32_t)shift != 0)) {
247             shift &= 0x1f;
248             ret = (int32_t)value >> shift;
249             if (likely(ret >= 0 || (value & ((1 << shift) - 1)) == 0)) {
250                 env->ca = 0;
251             } else {
252                 env->ca = 1;
253             }
254         } else {
255             ret = (int32_t)value;
256             env->ca = 0;
257         }
258     } else {
259         ret = (int32_t)value >> 31;
260         env->ca = (ret != 0);
261     }
262     return (target_long)ret;
263 }
264 
265 #if defined(TARGET_PPC64)
266 target_ulong helper_srad(CPUPPCState *env, target_ulong value,
267                          target_ulong shift)
268 {
269     int64_t ret;
270 
271     if (likely(!(shift & 0x40))) {
272         if (likely((uint64_t)shift != 0)) {
273             shift &= 0x3f;
274             ret = (int64_t)value >> shift;
275             if (likely(ret >= 0 || (value & ((1ULL << shift) - 1)) == 0)) {
276                 env->ca = 0;
277             } else {
278                 env->ca = 1;
279             }
280         } else {
281             ret = (int64_t)value;
282             env->ca = 0;
283         }
284     } else {
285         ret = (int64_t)value >> 63;
286         env->ca = (ret != 0);
287     }
288     return ret;
289 }
290 #endif
291 
292 #if defined(TARGET_PPC64)
293 target_ulong helper_popcntb(target_ulong val)
294 {
295     val = (val & 0x5555555555555555ULL) + ((val >>  1) &
296                                            0x5555555555555555ULL);
297     val = (val & 0x3333333333333333ULL) + ((val >>  2) &
298                                            0x3333333333333333ULL);
299     val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >>  4) &
300                                            0x0f0f0f0f0f0f0f0fULL);
301     return val;
302 }
303 
304 target_ulong helper_popcntw(target_ulong val)
305 {
306     val = (val & 0x5555555555555555ULL) + ((val >>  1) &
307                                            0x5555555555555555ULL);
308     val = (val & 0x3333333333333333ULL) + ((val >>  2) &
309                                            0x3333333333333333ULL);
310     val = (val & 0x0f0f0f0f0f0f0f0fULL) + ((val >>  4) &
311                                            0x0f0f0f0f0f0f0f0fULL);
312     val = (val & 0x00ff00ff00ff00ffULL) + ((val >>  8) &
313                                            0x00ff00ff00ff00ffULL);
314     val = (val & 0x0000ffff0000ffffULL) + ((val >> 16) &
315                                            0x0000ffff0000ffffULL);
316     return val;
317 }
318 
319 target_ulong helper_popcntd(target_ulong val)
320 {
321     return ctpop64(val);
322 }
323 #else
324 target_ulong helper_popcntb(target_ulong val)
325 {
326     val = (val & 0x55555555) + ((val >>  1) & 0x55555555);
327     val = (val & 0x33333333) + ((val >>  2) & 0x33333333);
328     val = (val & 0x0f0f0f0f) + ((val >>  4) & 0x0f0f0f0f);
329     return val;
330 }
331 
332 target_ulong helper_popcntw(target_ulong val)
333 {
334     val = (val & 0x55555555) + ((val >>  1) & 0x55555555);
335     val = (val & 0x33333333) + ((val >>  2) & 0x33333333);
336     val = (val & 0x0f0f0f0f) + ((val >>  4) & 0x0f0f0f0f);
337     val = (val & 0x00ff00ff) + ((val >>  8) & 0x00ff00ff);
338     val = (val & 0x0000ffff) + ((val >> 16) & 0x0000ffff);
339     return val;
340 }
341 #endif
342 
343 /*****************************************************************************/
344 /* PowerPC 601 specific instructions (POWER bridge) */
345 target_ulong helper_div(CPUPPCState *env, target_ulong arg1, target_ulong arg2)
346 {
347     uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
348 
349     if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
350         (int32_t)arg2 == 0) {
351         env->spr[SPR_MQ] = 0;
352         return INT32_MIN;
353     } else {
354         env->spr[SPR_MQ] = tmp % arg2;
355         return  tmp / (int32_t)arg2;
356     }
357 }
358 
359 target_ulong helper_divo(CPUPPCState *env, target_ulong arg1,
360                          target_ulong arg2)
361 {
362     uint64_t tmp = (uint64_t)arg1 << 32 | env->spr[SPR_MQ];
363 
364     if (((int32_t)tmp == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
365         (int32_t)arg2 == 0) {
366         env->so = env->ov = 1;
367         env->spr[SPR_MQ] = 0;
368         return INT32_MIN;
369     } else {
370         env->spr[SPR_MQ] = tmp % arg2;
371         tmp /= (int32_t)arg2;
372         if ((int32_t)tmp != tmp) {
373             env->so = env->ov = 1;
374         } else {
375             env->ov = 0;
376         }
377         return tmp;
378     }
379 }
380 
381 target_ulong helper_divs(CPUPPCState *env, target_ulong arg1,
382                          target_ulong arg2)
383 {
384     if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
385         (int32_t)arg2 == 0) {
386         env->spr[SPR_MQ] = 0;
387         return INT32_MIN;
388     } else {
389         env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
390         return (int32_t)arg1 / (int32_t)arg2;
391     }
392 }
393 
394 target_ulong helper_divso(CPUPPCState *env, target_ulong arg1,
395                           target_ulong arg2)
396 {
397     if (((int32_t)arg1 == INT32_MIN && (int32_t)arg2 == (int32_t)-1) ||
398         (int32_t)arg2 == 0) {
399         env->so = env->ov = 1;
400         env->spr[SPR_MQ] = 0;
401         return INT32_MIN;
402     } else {
403         env->ov = 0;
404         env->spr[SPR_MQ] = (int32_t)arg1 % (int32_t)arg2;
405         return (int32_t)arg1 / (int32_t)arg2;
406     }
407 }
408 
409 /*****************************************************************************/
410 /* 602 specific instructions */
411 /* mfrom is the most crazy instruction ever seen, imho ! */
412 /* Real implementation uses a ROM table. Do the same */
413 /* Extremely decomposed:
414  *                      -arg / 256
415  * return 256 * log10(10           + 1.0) + 0.5
416  */
417 #if !defined(CONFIG_USER_ONLY)
418 target_ulong helper_602_mfrom(target_ulong arg)
419 {
420     if (likely(arg < 602)) {
421 #include "mfrom_table.c"
422         return mfrom_ROM_table[arg];
423     } else {
424         return 0;
425     }
426 }
427 #endif
428 
429 /*****************************************************************************/
430 /* Altivec extension helpers */
431 #if defined(HOST_WORDS_BIGENDIAN)
432 #define HI_IDX 0
433 #define LO_IDX 1
434 #define AVRB(i) u8[i]
435 #define AVRW(i) u32[i]
436 #else
437 #define HI_IDX 1
438 #define LO_IDX 0
439 #define AVRB(i) u8[15-(i)]
440 #define AVRW(i) u32[3-(i)]
441 #endif
442 
443 #if defined(HOST_WORDS_BIGENDIAN)
444 #define VECTOR_FOR_INORDER_I(index, element)                    \
445     for (index = 0; index < ARRAY_SIZE(r->element); index++)
446 #else
447 #define VECTOR_FOR_INORDER_I(index, element)                    \
448     for (index = ARRAY_SIZE(r->element)-1; index >= 0; index--)
449 #endif
450 
451 /* Saturating arithmetic helpers.  */
452 #define SATCVT(from, to, from_type, to_type, min, max)          \
453     static inline to_type cvt##from##to(from_type x, int *sat)  \
454     {                                                           \
455         to_type r;                                              \
456                                                                 \
457         if (x < (from_type)min) {                               \
458             r = min;                                            \
459             *sat = 1;                                           \
460         } else if (x > (from_type)max) {                        \
461             r = max;                                            \
462             *sat = 1;                                           \
463         } else {                                                \
464             r = x;                                              \
465         }                                                       \
466         return r;                                               \
467     }
468 #define SATCVTU(from, to, from_type, to_type, min, max)         \
469     static inline to_type cvt##from##to(from_type x, int *sat)  \
470     {                                                           \
471         to_type r;                                              \
472                                                                 \
473         if (x > (from_type)max) {                               \
474             r = max;                                            \
475             *sat = 1;                                           \
476         } else {                                                \
477             r = x;                                              \
478         }                                                       \
479         return r;                                               \
480     }
481 SATCVT(sh, sb, int16_t, int8_t, INT8_MIN, INT8_MAX)
482 SATCVT(sw, sh, int32_t, int16_t, INT16_MIN, INT16_MAX)
483 SATCVT(sd, sw, int64_t, int32_t, INT32_MIN, INT32_MAX)
484 
485 SATCVTU(uh, ub, uint16_t, uint8_t, 0, UINT8_MAX)
486 SATCVTU(uw, uh, uint32_t, uint16_t, 0, UINT16_MAX)
487 SATCVTU(ud, uw, uint64_t, uint32_t, 0, UINT32_MAX)
488 SATCVT(sh, ub, int16_t, uint8_t, 0, UINT8_MAX)
489 SATCVT(sw, uh, int32_t, uint16_t, 0, UINT16_MAX)
490 SATCVT(sd, uw, int64_t, uint32_t, 0, UINT32_MAX)
491 #undef SATCVT
492 #undef SATCVTU
493 
494 void helper_lvsl(ppc_avr_t *r, target_ulong sh)
495 {
496     int i, j = (sh & 0xf);
497 
498     VECTOR_FOR_INORDER_I(i, u8) {
499         r->u8[i] = j++;
500     }
501 }
502 
503 void helper_lvsr(ppc_avr_t *r, target_ulong sh)
504 {
505     int i, j = 0x10 - (sh & 0xf);
506 
507     VECTOR_FOR_INORDER_I(i, u8) {
508         r->u8[i] = j++;
509     }
510 }
511 
512 void helper_mtvscr(CPUPPCState *env, ppc_avr_t *r)
513 {
514 #if defined(HOST_WORDS_BIGENDIAN)
515     env->vscr = r->u32[3];
516 #else
517     env->vscr = r->u32[0];
518 #endif
519     set_flush_to_zero(vscr_nj, &env->vec_status);
520 }
521 
522 void helper_vaddcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
523 {
524     int i;
525 
526     for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
527         r->u32[i] = ~a->u32[i] < b->u32[i];
528     }
529 }
530 
531 /* vprtybw */
532 void helper_vprtybw(ppc_avr_t *r, ppc_avr_t *b)
533 {
534     int i;
535     for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
536         uint64_t res = b->u32[i] ^ (b->u32[i] >> 16);
537         res ^= res >> 8;
538         r->u32[i] = res & 1;
539     }
540 }
541 
542 /* vprtybd */
543 void helper_vprtybd(ppc_avr_t *r, ppc_avr_t *b)
544 {
545     int i;
546     for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
547         uint64_t res = b->u64[i] ^ (b->u64[i] >> 32);
548         res ^= res >> 16;
549         res ^= res >> 8;
550         r->u64[i] = res & 1;
551     }
552 }
553 
554 /* vprtybq */
555 void helper_vprtybq(ppc_avr_t *r, ppc_avr_t *b)
556 {
557     uint64_t res = b->u64[0] ^ b->u64[1];
558     res ^= res >> 32;
559     res ^= res >> 16;
560     res ^= res >> 8;
561     r->u64[LO_IDX] = res & 1;
562     r->u64[HI_IDX] = 0;
563 }
564 
565 #define VARITH_DO(name, op, element)                                    \
566     void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)       \
567     {                                                                   \
568         int i;                                                          \
569                                                                         \
570         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
571             r->element[i] = a->element[i] op b->element[i];             \
572         }                                                               \
573     }
574 #define VARITH(suffix, element)                 \
575     VARITH_DO(add##suffix, +, element)          \
576     VARITH_DO(sub##suffix, -, element)
577 VARITH(ubm, u8)
578 VARITH(uhm, u16)
579 VARITH(uwm, u32)
580 VARITH(udm, u64)
581 VARITH_DO(muluwm, *, u32)
582 #undef VARITH_DO
583 #undef VARITH
584 
585 #define VARITHFP(suffix, func)                                          \
586     void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
587                           ppc_avr_t *b)                                 \
588     {                                                                   \
589         int i;                                                          \
590                                                                         \
591         for (i = 0; i < ARRAY_SIZE(r->f); i++) {                        \
592             r->f[i] = func(a->f[i], b->f[i], &env->vec_status);         \
593         }                                                               \
594     }
595 VARITHFP(addfp, float32_add)
596 VARITHFP(subfp, float32_sub)
597 VARITHFP(minfp, float32_min)
598 VARITHFP(maxfp, float32_max)
599 #undef VARITHFP
600 
601 #define VARITHFPFMA(suffix, type)                                       \
602     void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
603                            ppc_avr_t *b, ppc_avr_t *c)                  \
604     {                                                                   \
605         int i;                                                          \
606         for (i = 0; i < ARRAY_SIZE(r->f); i++) {                        \
607             r->f[i] = float32_muladd(a->f[i], c->f[i], b->f[i],         \
608                                      type, &env->vec_status);           \
609         }                                                               \
610     }
611 VARITHFPFMA(maddfp, 0);
612 VARITHFPFMA(nmsubfp, float_muladd_negate_result | float_muladd_negate_c);
613 #undef VARITHFPFMA
614 
615 #define VARITHSAT_CASE(type, op, cvt, element)                          \
616     {                                                                   \
617         type result = (type)a->element[i] op (type)b->element[i];       \
618         r->element[i] = cvt(result, &sat);                              \
619     }
620 
621 #define VARITHSAT_DO(name, op, optype, cvt, element)                    \
622     void helper_v##name(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,   \
623                         ppc_avr_t *b)                                   \
624     {                                                                   \
625         int sat = 0;                                                    \
626         int i;                                                          \
627                                                                         \
628         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
629             switch (sizeof(r->element[0])) {                            \
630             case 1:                                                     \
631                 VARITHSAT_CASE(optype, op, cvt, element);               \
632                 break;                                                  \
633             case 2:                                                     \
634                 VARITHSAT_CASE(optype, op, cvt, element);               \
635                 break;                                                  \
636             case 4:                                                     \
637                 VARITHSAT_CASE(optype, op, cvt, element);               \
638                 break;                                                  \
639             }                                                           \
640         }                                                               \
641         if (sat) {                                                      \
642             env->vscr |= (1 << VSCR_SAT);                               \
643         }                                                               \
644     }
645 #define VARITHSAT_SIGNED(suffix, element, optype, cvt)          \
646     VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element)      \
647     VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element)
648 #define VARITHSAT_UNSIGNED(suffix, element, optype, cvt)        \
649     VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element)      \
650     VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element)
651 VARITHSAT_SIGNED(b, s8, int16_t, cvtshsb)
652 VARITHSAT_SIGNED(h, s16, int32_t, cvtswsh)
653 VARITHSAT_SIGNED(w, s32, int64_t, cvtsdsw)
654 VARITHSAT_UNSIGNED(b, u8, uint16_t, cvtshub)
655 VARITHSAT_UNSIGNED(h, u16, uint32_t, cvtswuh)
656 VARITHSAT_UNSIGNED(w, u32, uint64_t, cvtsduw)
657 #undef VARITHSAT_CASE
658 #undef VARITHSAT_DO
659 #undef VARITHSAT_SIGNED
660 #undef VARITHSAT_UNSIGNED
661 
662 #define VAVG_DO(name, element, etype)                                   \
663     void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)       \
664     {                                                                   \
665         int i;                                                          \
666                                                                         \
667         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
668             etype x = (etype)a->element[i] + (etype)b->element[i] + 1;  \
669             r->element[i] = x >> 1;                                     \
670         }                                                               \
671     }
672 
673 #define VAVG(type, signed_element, signed_type, unsigned_element,       \
674              unsigned_type)                                             \
675     VAVG_DO(avgs##type, signed_element, signed_type)                    \
676     VAVG_DO(avgu##type, unsigned_element, unsigned_type)
677 VAVG(b, s8, int16_t, u8, uint16_t)
678 VAVG(h, s16, int32_t, u16, uint32_t)
679 VAVG(w, s32, int64_t, u32, uint64_t)
680 #undef VAVG_DO
681 #undef VAVG
682 
683 #define VABSDU_DO(name, element)                                        \
684 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)           \
685 {                                                                       \
686     int i;                                                              \
687                                                                         \
688     for (i = 0; i < ARRAY_SIZE(r->element); i++) {                      \
689         r->element[i] = (a->element[i] > b->element[i]) ?               \
690             (a->element[i] - b->element[i]) :                           \
691             (b->element[i] - a->element[i]);                            \
692     }                                                                   \
693 }
694 
695 /* VABSDU - Vector absolute difference unsigned
696  *   name    - instruction mnemonic suffix (b: byte, h: halfword, w: word)
697  *   element - element type to access from vector
698  */
699 #define VABSDU(type, element)                   \
700     VABSDU_DO(absdu##type, element)
701 VABSDU(b, u8)
702 VABSDU(h, u16)
703 VABSDU(w, u32)
704 #undef VABSDU_DO
705 #undef VABSDU
706 
707 #define VCF(suffix, cvt, element)                                       \
708     void helper_vcf##suffix(CPUPPCState *env, ppc_avr_t *r,             \
709                             ppc_avr_t *b, uint32_t uim)                 \
710     {                                                                   \
711         int i;                                                          \
712                                                                         \
713         for (i = 0; i < ARRAY_SIZE(r->f); i++) {                        \
714             float32 t = cvt(b->element[i], &env->vec_status);           \
715             r->f[i] = float32_scalbn(t, -uim, &env->vec_status);        \
716         }                                                               \
717     }
718 VCF(ux, uint32_to_float32, u32)
719 VCF(sx, int32_to_float32, s32)
720 #undef VCF
721 
722 #define VCMP_DO(suffix, compare, element, record)                       \
723     void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r,            \
724                              ppc_avr_t *a, ppc_avr_t *b)                \
725     {                                                                   \
726         uint64_t ones = (uint64_t)-1;                                   \
727         uint64_t all = ones;                                            \
728         uint64_t none = 0;                                              \
729         int i;                                                          \
730                                                                         \
731         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
732             uint64_t result = (a->element[i] compare b->element[i] ?    \
733                                ones : 0x0);                             \
734             switch (sizeof(a->element[0])) {                            \
735             case 8:                                                     \
736                 r->u64[i] = result;                                     \
737                 break;                                                  \
738             case 4:                                                     \
739                 r->u32[i] = result;                                     \
740                 break;                                                  \
741             case 2:                                                     \
742                 r->u16[i] = result;                                     \
743                 break;                                                  \
744             case 1:                                                     \
745                 r->u8[i] = result;                                      \
746                 break;                                                  \
747             }                                                           \
748             all &= result;                                              \
749             none |= result;                                             \
750         }                                                               \
751         if (record) {                                                   \
752             env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1);       \
753         }                                                               \
754     }
755 #define VCMP(suffix, compare, element)          \
756     VCMP_DO(suffix, compare, element, 0)        \
757     VCMP_DO(suffix##_dot, compare, element, 1)
758 VCMP(equb, ==, u8)
759 VCMP(equh, ==, u16)
760 VCMP(equw, ==, u32)
761 VCMP(equd, ==, u64)
762 VCMP(gtub, >, u8)
763 VCMP(gtuh, >, u16)
764 VCMP(gtuw, >, u32)
765 VCMP(gtud, >, u64)
766 VCMP(gtsb, >, s8)
767 VCMP(gtsh, >, s16)
768 VCMP(gtsw, >, s32)
769 VCMP(gtsd, >, s64)
770 #undef VCMP_DO
771 #undef VCMP
772 
773 #define VCMPNE_DO(suffix, element, etype, cmpzero, record)              \
774 void helper_vcmpne##suffix(CPUPPCState *env, ppc_avr_t *r,              \
775                             ppc_avr_t *a, ppc_avr_t *b)                 \
776 {                                                                       \
777     etype ones = (etype)-1;                                             \
778     etype all = ones;                                                   \
779     etype result, none = 0;                                             \
780     int i;                                                              \
781                                                                         \
782     for (i = 0; i < ARRAY_SIZE(r->element); i++) {                      \
783         if (cmpzero) {                                                  \
784             result = ((a->element[i] == 0)                              \
785                            || (b->element[i] == 0)                      \
786                            || (a->element[i] != b->element[i]) ?        \
787                            ones : 0x0);                                 \
788         } else {                                                        \
789             result = (a->element[i] != b->element[i]) ? ones : 0x0;     \
790         }                                                               \
791         r->element[i] = result;                                         \
792         all &= result;                                                  \
793         none |= result;                                                 \
794     }                                                                   \
795     if (record) {                                                       \
796         env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1);           \
797     }                                                                   \
798 }
799 
800 /* VCMPNEZ - Vector compare not equal to zero
801  *   suffix  - instruction mnemonic suffix (b: byte, h: halfword, w: word)
802  *   element - element type to access from vector
803  */
804 #define VCMPNE(suffix, element, etype, cmpzero)         \
805     VCMPNE_DO(suffix, element, etype, cmpzero, 0)       \
806     VCMPNE_DO(suffix##_dot, element, etype, cmpzero, 1)
807 VCMPNE(zb, u8, uint8_t, 1)
808 VCMPNE(zh, u16, uint16_t, 1)
809 VCMPNE(zw, u32, uint32_t, 1)
810 VCMPNE(b, u8, uint8_t, 0)
811 VCMPNE(h, u16, uint16_t, 0)
812 VCMPNE(w, u32, uint32_t, 0)
813 #undef VCMPNE_DO
814 #undef VCMPNE
815 
816 #define VCMPFP_DO(suffix, compare, order, record)                       \
817     void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r,            \
818                              ppc_avr_t *a, ppc_avr_t *b)                \
819     {                                                                   \
820         uint32_t ones = (uint32_t)-1;                                   \
821         uint32_t all = ones;                                            \
822         uint32_t none = 0;                                              \
823         int i;                                                          \
824                                                                         \
825         for (i = 0; i < ARRAY_SIZE(r->f); i++) {                        \
826             uint32_t result;                                            \
827             int rel = float32_compare_quiet(a->f[i], b->f[i],           \
828                                             &env->vec_status);          \
829             if (rel == float_relation_unordered) {                      \
830                 result = 0;                                             \
831             } else if (rel compare order) {                             \
832                 result = ones;                                          \
833             } else {                                                    \
834                 result = 0;                                             \
835             }                                                           \
836             r->u32[i] = result;                                         \
837             all &= result;                                              \
838             none |= result;                                             \
839         }                                                               \
840         if (record) {                                                   \
841             env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1);       \
842         }                                                               \
843     }
844 #define VCMPFP(suffix, compare, order)          \
845     VCMPFP_DO(suffix, compare, order, 0)        \
846     VCMPFP_DO(suffix##_dot, compare, order, 1)
847 VCMPFP(eqfp, ==, float_relation_equal)
848 VCMPFP(gefp, !=, float_relation_less)
849 VCMPFP(gtfp, ==, float_relation_greater)
850 #undef VCMPFP_DO
851 #undef VCMPFP
852 
853 static inline void vcmpbfp_internal(CPUPPCState *env, ppc_avr_t *r,
854                                     ppc_avr_t *a, ppc_avr_t *b, int record)
855 {
856     int i;
857     int all_in = 0;
858 
859     for (i = 0; i < ARRAY_SIZE(r->f); i++) {
860         int le_rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status);
861         if (le_rel == float_relation_unordered) {
862             r->u32[i] = 0xc0000000;
863             all_in = 1;
864         } else {
865             float32 bneg = float32_chs(b->f[i]);
866             int ge_rel = float32_compare_quiet(a->f[i], bneg, &env->vec_status);
867             int le = le_rel != float_relation_greater;
868             int ge = ge_rel != float_relation_less;
869 
870             r->u32[i] = ((!le) << 31) | ((!ge) << 30);
871             all_in |= (!le | !ge);
872         }
873     }
874     if (record) {
875         env->crf[6] = (all_in == 0) << 1;
876     }
877 }
878 
879 void helper_vcmpbfp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
880 {
881     vcmpbfp_internal(env, r, a, b, 0);
882 }
883 
884 void helper_vcmpbfp_dot(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
885                         ppc_avr_t *b)
886 {
887     vcmpbfp_internal(env, r, a, b, 1);
888 }
889 
890 #define VCT(suffix, satcvt, element)                                    \
891     void helper_vct##suffix(CPUPPCState *env, ppc_avr_t *r,             \
892                             ppc_avr_t *b, uint32_t uim)                 \
893     {                                                                   \
894         int i;                                                          \
895         int sat = 0;                                                    \
896         float_status s = env->vec_status;                               \
897                                                                         \
898         set_float_rounding_mode(float_round_to_zero, &s);               \
899         for (i = 0; i < ARRAY_SIZE(r->f); i++) {                        \
900             if (float32_is_any_nan(b->f[i])) {                          \
901                 r->element[i] = 0;                                      \
902             } else {                                                    \
903                 float64 t = float32_to_float64(b->f[i], &s);            \
904                 int64_t j;                                              \
905                                                                         \
906                 t = float64_scalbn(t, uim, &s);                         \
907                 j = float64_to_int64(t, &s);                            \
908                 r->element[i] = satcvt(j, &sat);                        \
909             }                                                           \
910         }                                                               \
911         if (sat) {                                                      \
912             env->vscr |= (1 << VSCR_SAT);                               \
913         }                                                               \
914     }
915 VCT(uxs, cvtsduw, u32)
916 VCT(sxs, cvtsdsw, s32)
917 #undef VCT
918 
919 target_ulong helper_vclzlsbb(ppc_avr_t *r)
920 {
921     target_ulong count = 0;
922     int i;
923     VECTOR_FOR_INORDER_I(i, u8) {
924         if (r->u8[i] & 0x01) {
925             break;
926         }
927         count++;
928     }
929     return count;
930 }
931 
932 target_ulong helper_vctzlsbb(ppc_avr_t *r)
933 {
934     target_ulong count = 0;
935     int i;
936 #if defined(HOST_WORDS_BIGENDIAN)
937     for (i = ARRAY_SIZE(r->u8) - 1; i >= 0; i--) {
938 #else
939     for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
940 #endif
941         if (r->u8[i] & 0x01) {
942             break;
943         }
944         count++;
945     }
946     return count;
947 }
948 
949 void helper_vmhaddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
950                       ppc_avr_t *b, ppc_avr_t *c)
951 {
952     int sat = 0;
953     int i;
954 
955     for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
956         int32_t prod = a->s16[i] * b->s16[i];
957         int32_t t = (int32_t)c->s16[i] + (prod >> 15);
958 
959         r->s16[i] = cvtswsh(t, &sat);
960     }
961 
962     if (sat) {
963         env->vscr |= (1 << VSCR_SAT);
964     }
965 }
966 
967 void helper_vmhraddshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
968                        ppc_avr_t *b, ppc_avr_t *c)
969 {
970     int sat = 0;
971     int i;
972 
973     for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
974         int32_t prod = a->s16[i] * b->s16[i] + 0x00004000;
975         int32_t t = (int32_t)c->s16[i] + (prod >> 15);
976         r->s16[i] = cvtswsh(t, &sat);
977     }
978 
979     if (sat) {
980         env->vscr |= (1 << VSCR_SAT);
981     }
982 }
983 
984 #define VMINMAX_DO(name, compare, element)                              \
985     void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)       \
986     {                                                                   \
987         int i;                                                          \
988                                                                         \
989         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
990             if (a->element[i] compare b->element[i]) {                  \
991                 r->element[i] = b->element[i];                          \
992             } else {                                                    \
993                 r->element[i] = a->element[i];                          \
994             }                                                           \
995         }                                                               \
996     }
997 #define VMINMAX(suffix, element)                \
998     VMINMAX_DO(min##suffix, >, element)         \
999     VMINMAX_DO(max##suffix, <, element)
1000 VMINMAX(sb, s8)
1001 VMINMAX(sh, s16)
1002 VMINMAX(sw, s32)
1003 VMINMAX(sd, s64)
1004 VMINMAX(ub, u8)
1005 VMINMAX(uh, u16)
1006 VMINMAX(uw, u32)
1007 VMINMAX(ud, u64)
1008 #undef VMINMAX_DO
1009 #undef VMINMAX
1010 
1011 void helper_vmladduhm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
1012 {
1013     int i;
1014 
1015     for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1016         int32_t prod = a->s16[i] * b->s16[i];
1017         r->s16[i] = (int16_t) (prod + c->s16[i]);
1018     }
1019 }
1020 
1021 #define VMRG_DO(name, element, highp)                                   \
1022     void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)       \
1023     {                                                                   \
1024         ppc_avr_t result;                                               \
1025         int i;                                                          \
1026         size_t n_elems = ARRAY_SIZE(r->element);                        \
1027                                                                         \
1028         for (i = 0; i < n_elems / 2; i++) {                             \
1029             if (highp) {                                                \
1030                 result.element[i*2+HI_IDX] = a->element[i];             \
1031                 result.element[i*2+LO_IDX] = b->element[i];             \
1032             } else {                                                    \
1033                 result.element[n_elems - i * 2 - (1 + HI_IDX)] =        \
1034                     b->element[n_elems - i - 1];                        \
1035                 result.element[n_elems - i * 2 - (1 + LO_IDX)] =        \
1036                     a->element[n_elems - i - 1];                        \
1037             }                                                           \
1038         }                                                               \
1039         *r = result;                                                    \
1040     }
1041 #if defined(HOST_WORDS_BIGENDIAN)
1042 #define MRGHI 0
1043 #define MRGLO 1
1044 #else
1045 #define MRGHI 1
1046 #define MRGLO 0
1047 #endif
1048 #define VMRG(suffix, element)                   \
1049     VMRG_DO(mrgl##suffix, element, MRGHI)       \
1050     VMRG_DO(mrgh##suffix, element, MRGLO)
1051 VMRG(b, u8)
1052 VMRG(h, u16)
1053 VMRG(w, u32)
1054 #undef VMRG_DO
1055 #undef VMRG
1056 #undef MRGHI
1057 #undef MRGLO
1058 
1059 void helper_vmsummbm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1060                      ppc_avr_t *b, ppc_avr_t *c)
1061 {
1062     int32_t prod[16];
1063     int i;
1064 
1065     for (i = 0; i < ARRAY_SIZE(r->s8); i++) {
1066         prod[i] = (int32_t)a->s8[i] * b->u8[i];
1067     }
1068 
1069     VECTOR_FOR_INORDER_I(i, s32) {
1070         r->s32[i] = c->s32[i] + prod[4 * i] + prod[4 * i + 1] +
1071             prod[4 * i + 2] + prod[4 * i + 3];
1072     }
1073 }
1074 
1075 void helper_vmsumshm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1076                      ppc_avr_t *b, ppc_avr_t *c)
1077 {
1078     int32_t prod[8];
1079     int i;
1080 
1081     for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1082         prod[i] = a->s16[i] * b->s16[i];
1083     }
1084 
1085     VECTOR_FOR_INORDER_I(i, s32) {
1086         r->s32[i] = c->s32[i] + prod[2 * i] + prod[2 * i + 1];
1087     }
1088 }
1089 
1090 void helper_vmsumshs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1091                      ppc_avr_t *b, ppc_avr_t *c)
1092 {
1093     int32_t prod[8];
1094     int i;
1095     int sat = 0;
1096 
1097     for (i = 0; i < ARRAY_SIZE(r->s16); i++) {
1098         prod[i] = (int32_t)a->s16[i] * b->s16[i];
1099     }
1100 
1101     VECTOR_FOR_INORDER_I(i, s32) {
1102         int64_t t = (int64_t)c->s32[i] + prod[2 * i] + prod[2 * i + 1];
1103 
1104         r->u32[i] = cvtsdsw(t, &sat);
1105     }
1106 
1107     if (sat) {
1108         env->vscr |= (1 << VSCR_SAT);
1109     }
1110 }
1111 
1112 void helper_vmsumubm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1113                      ppc_avr_t *b, ppc_avr_t *c)
1114 {
1115     uint16_t prod[16];
1116     int i;
1117 
1118     for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1119         prod[i] = a->u8[i] * b->u8[i];
1120     }
1121 
1122     VECTOR_FOR_INORDER_I(i, u32) {
1123         r->u32[i] = c->u32[i] + prod[4 * i] + prod[4 * i + 1] +
1124             prod[4 * i + 2] + prod[4 * i + 3];
1125     }
1126 }
1127 
1128 void helper_vmsumuhm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1129                      ppc_avr_t *b, ppc_avr_t *c)
1130 {
1131     uint32_t prod[8];
1132     int i;
1133 
1134     for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
1135         prod[i] = a->u16[i] * b->u16[i];
1136     }
1137 
1138     VECTOR_FOR_INORDER_I(i, u32) {
1139         r->u32[i] = c->u32[i] + prod[2 * i] + prod[2 * i + 1];
1140     }
1141 }
1142 
1143 void helper_vmsumuhs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a,
1144                      ppc_avr_t *b, ppc_avr_t *c)
1145 {
1146     uint32_t prod[8];
1147     int i;
1148     int sat = 0;
1149 
1150     for (i = 0; i < ARRAY_SIZE(r->u16); i++) {
1151         prod[i] = a->u16[i] * b->u16[i];
1152     }
1153 
1154     VECTOR_FOR_INORDER_I(i, s32) {
1155         uint64_t t = (uint64_t)c->u32[i] + prod[2 * i] + prod[2 * i + 1];
1156 
1157         r->u32[i] = cvtuduw(t, &sat);
1158     }
1159 
1160     if (sat) {
1161         env->vscr |= (1 << VSCR_SAT);
1162     }
1163 }
1164 
1165 #define VMUL_DO(name, mul_element, prod_element, cast, evenp)           \
1166     void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)       \
1167     {                                                                   \
1168         int i;                                                          \
1169                                                                         \
1170         VECTOR_FOR_INORDER_I(i, prod_element) {                         \
1171             if (evenp) {                                                \
1172                 r->prod_element[i] =                                    \
1173                     (cast)a->mul_element[i * 2 + HI_IDX] *              \
1174                     (cast)b->mul_element[i * 2 + HI_IDX];               \
1175             } else {                                                    \
1176                 r->prod_element[i] =                                    \
1177                     (cast)a->mul_element[i * 2 + LO_IDX] *              \
1178                     (cast)b->mul_element[i * 2 + LO_IDX];               \
1179             }                                                           \
1180         }                                                               \
1181     }
1182 #define VMUL(suffix, mul_element, prod_element, cast)            \
1183     VMUL_DO(mule##suffix, mul_element, prod_element, cast, 1)    \
1184     VMUL_DO(mulo##suffix, mul_element, prod_element, cast, 0)
1185 VMUL(sb, s8, s16, int16_t)
1186 VMUL(sh, s16, s32, int32_t)
1187 VMUL(sw, s32, s64, int64_t)
1188 VMUL(ub, u8, u16, uint16_t)
1189 VMUL(uh, u16, u32, uint32_t)
1190 VMUL(uw, u32, u64, uint64_t)
1191 #undef VMUL_DO
1192 #undef VMUL
1193 
1194 void helper_vperm(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1195                   ppc_avr_t *c)
1196 {
1197     ppc_avr_t result;
1198     int i;
1199 
1200     VECTOR_FOR_INORDER_I(i, u8) {
1201         int s = c->u8[i] & 0x1f;
1202 #if defined(HOST_WORDS_BIGENDIAN)
1203         int index = s & 0xf;
1204 #else
1205         int index = 15 - (s & 0xf);
1206 #endif
1207 
1208         if (s & 0x10) {
1209             result.u8[i] = b->u8[index];
1210         } else {
1211             result.u8[i] = a->u8[index];
1212         }
1213     }
1214     *r = result;
1215 }
1216 
1217 void helper_vpermr(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1218                   ppc_avr_t *c)
1219 {
1220     ppc_avr_t result;
1221     int i;
1222 
1223     VECTOR_FOR_INORDER_I(i, u8) {
1224         int s = c->u8[i] & 0x1f;
1225 #if defined(HOST_WORDS_BIGENDIAN)
1226         int index = 15 - (s & 0xf);
1227 #else
1228         int index = s & 0xf;
1229 #endif
1230 
1231         if (s & 0x10) {
1232             result.u8[i] = a->u8[index];
1233         } else {
1234             result.u8[i] = b->u8[index];
1235         }
1236     }
1237     *r = result;
1238 }
1239 
1240 #if defined(HOST_WORDS_BIGENDIAN)
1241 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[(i)])
1242 #define VBPERMD_INDEX(i) (i)
1243 #define VBPERMQ_DW(index) (((index) & 0x40) != 0)
1244 #define EXTRACT_BIT(avr, i, index) (extract64((avr)->u64[i], index, 1))
1245 #else
1246 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[15-(i)])
1247 #define VBPERMD_INDEX(i) (1 - i)
1248 #define VBPERMQ_DW(index) (((index) & 0x40) == 0)
1249 #define EXTRACT_BIT(avr, i, index) \
1250         (extract64((avr)->u64[1 - i], 63 - index, 1))
1251 #endif
1252 
1253 void helper_vbpermd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1254 {
1255     int i, j;
1256     ppc_avr_t result = { .u64 = { 0, 0 } };
1257     VECTOR_FOR_INORDER_I(i, u64) {
1258         for (j = 0; j < 8; j++) {
1259             int index = VBPERMQ_INDEX(b, (i * 8) + j);
1260             if (index < 64 && EXTRACT_BIT(a, i, index)) {
1261                 result.u64[VBPERMD_INDEX(i)] |= (0x80 >> j);
1262             }
1263         }
1264     }
1265     *r = result;
1266 }
1267 
1268 void helper_vbpermq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1269 {
1270     int i;
1271     uint64_t perm = 0;
1272 
1273     VECTOR_FOR_INORDER_I(i, u8) {
1274         int index = VBPERMQ_INDEX(b, i);
1275 
1276         if (index < 128) {
1277             uint64_t mask = (1ull << (63-(index & 0x3F)));
1278             if (a->u64[VBPERMQ_DW(index)] & mask) {
1279                 perm |= (0x8000 >> i);
1280             }
1281         }
1282     }
1283 
1284     r->u64[HI_IDX] = perm;
1285     r->u64[LO_IDX] = 0;
1286 }
1287 
1288 #undef VBPERMQ_INDEX
1289 #undef VBPERMQ_DW
1290 
1291 static const uint64_t VGBBD_MASKS[256] = {
1292     0x0000000000000000ull, /* 00 */
1293     0x0000000000000080ull, /* 01 */
1294     0x0000000000008000ull, /* 02 */
1295     0x0000000000008080ull, /* 03 */
1296     0x0000000000800000ull, /* 04 */
1297     0x0000000000800080ull, /* 05 */
1298     0x0000000000808000ull, /* 06 */
1299     0x0000000000808080ull, /* 07 */
1300     0x0000000080000000ull, /* 08 */
1301     0x0000000080000080ull, /* 09 */
1302     0x0000000080008000ull, /* 0A */
1303     0x0000000080008080ull, /* 0B */
1304     0x0000000080800000ull, /* 0C */
1305     0x0000000080800080ull, /* 0D */
1306     0x0000000080808000ull, /* 0E */
1307     0x0000000080808080ull, /* 0F */
1308     0x0000008000000000ull, /* 10 */
1309     0x0000008000000080ull, /* 11 */
1310     0x0000008000008000ull, /* 12 */
1311     0x0000008000008080ull, /* 13 */
1312     0x0000008000800000ull, /* 14 */
1313     0x0000008000800080ull, /* 15 */
1314     0x0000008000808000ull, /* 16 */
1315     0x0000008000808080ull, /* 17 */
1316     0x0000008080000000ull, /* 18 */
1317     0x0000008080000080ull, /* 19 */
1318     0x0000008080008000ull, /* 1A */
1319     0x0000008080008080ull, /* 1B */
1320     0x0000008080800000ull, /* 1C */
1321     0x0000008080800080ull, /* 1D */
1322     0x0000008080808000ull, /* 1E */
1323     0x0000008080808080ull, /* 1F */
1324     0x0000800000000000ull, /* 20 */
1325     0x0000800000000080ull, /* 21 */
1326     0x0000800000008000ull, /* 22 */
1327     0x0000800000008080ull, /* 23 */
1328     0x0000800000800000ull, /* 24 */
1329     0x0000800000800080ull, /* 25 */
1330     0x0000800000808000ull, /* 26 */
1331     0x0000800000808080ull, /* 27 */
1332     0x0000800080000000ull, /* 28 */
1333     0x0000800080000080ull, /* 29 */
1334     0x0000800080008000ull, /* 2A */
1335     0x0000800080008080ull, /* 2B */
1336     0x0000800080800000ull, /* 2C */
1337     0x0000800080800080ull, /* 2D */
1338     0x0000800080808000ull, /* 2E */
1339     0x0000800080808080ull, /* 2F */
1340     0x0000808000000000ull, /* 30 */
1341     0x0000808000000080ull, /* 31 */
1342     0x0000808000008000ull, /* 32 */
1343     0x0000808000008080ull, /* 33 */
1344     0x0000808000800000ull, /* 34 */
1345     0x0000808000800080ull, /* 35 */
1346     0x0000808000808000ull, /* 36 */
1347     0x0000808000808080ull, /* 37 */
1348     0x0000808080000000ull, /* 38 */
1349     0x0000808080000080ull, /* 39 */
1350     0x0000808080008000ull, /* 3A */
1351     0x0000808080008080ull, /* 3B */
1352     0x0000808080800000ull, /* 3C */
1353     0x0000808080800080ull, /* 3D */
1354     0x0000808080808000ull, /* 3E */
1355     0x0000808080808080ull, /* 3F */
1356     0x0080000000000000ull, /* 40 */
1357     0x0080000000000080ull, /* 41 */
1358     0x0080000000008000ull, /* 42 */
1359     0x0080000000008080ull, /* 43 */
1360     0x0080000000800000ull, /* 44 */
1361     0x0080000000800080ull, /* 45 */
1362     0x0080000000808000ull, /* 46 */
1363     0x0080000000808080ull, /* 47 */
1364     0x0080000080000000ull, /* 48 */
1365     0x0080000080000080ull, /* 49 */
1366     0x0080000080008000ull, /* 4A */
1367     0x0080000080008080ull, /* 4B */
1368     0x0080000080800000ull, /* 4C */
1369     0x0080000080800080ull, /* 4D */
1370     0x0080000080808000ull, /* 4E */
1371     0x0080000080808080ull, /* 4F */
1372     0x0080008000000000ull, /* 50 */
1373     0x0080008000000080ull, /* 51 */
1374     0x0080008000008000ull, /* 52 */
1375     0x0080008000008080ull, /* 53 */
1376     0x0080008000800000ull, /* 54 */
1377     0x0080008000800080ull, /* 55 */
1378     0x0080008000808000ull, /* 56 */
1379     0x0080008000808080ull, /* 57 */
1380     0x0080008080000000ull, /* 58 */
1381     0x0080008080000080ull, /* 59 */
1382     0x0080008080008000ull, /* 5A */
1383     0x0080008080008080ull, /* 5B */
1384     0x0080008080800000ull, /* 5C */
1385     0x0080008080800080ull, /* 5D */
1386     0x0080008080808000ull, /* 5E */
1387     0x0080008080808080ull, /* 5F */
1388     0x0080800000000000ull, /* 60 */
1389     0x0080800000000080ull, /* 61 */
1390     0x0080800000008000ull, /* 62 */
1391     0x0080800000008080ull, /* 63 */
1392     0x0080800000800000ull, /* 64 */
1393     0x0080800000800080ull, /* 65 */
1394     0x0080800000808000ull, /* 66 */
1395     0x0080800000808080ull, /* 67 */
1396     0x0080800080000000ull, /* 68 */
1397     0x0080800080000080ull, /* 69 */
1398     0x0080800080008000ull, /* 6A */
1399     0x0080800080008080ull, /* 6B */
1400     0x0080800080800000ull, /* 6C */
1401     0x0080800080800080ull, /* 6D */
1402     0x0080800080808000ull, /* 6E */
1403     0x0080800080808080ull, /* 6F */
1404     0x0080808000000000ull, /* 70 */
1405     0x0080808000000080ull, /* 71 */
1406     0x0080808000008000ull, /* 72 */
1407     0x0080808000008080ull, /* 73 */
1408     0x0080808000800000ull, /* 74 */
1409     0x0080808000800080ull, /* 75 */
1410     0x0080808000808000ull, /* 76 */
1411     0x0080808000808080ull, /* 77 */
1412     0x0080808080000000ull, /* 78 */
1413     0x0080808080000080ull, /* 79 */
1414     0x0080808080008000ull, /* 7A */
1415     0x0080808080008080ull, /* 7B */
1416     0x0080808080800000ull, /* 7C */
1417     0x0080808080800080ull, /* 7D */
1418     0x0080808080808000ull, /* 7E */
1419     0x0080808080808080ull, /* 7F */
1420     0x8000000000000000ull, /* 80 */
1421     0x8000000000000080ull, /* 81 */
1422     0x8000000000008000ull, /* 82 */
1423     0x8000000000008080ull, /* 83 */
1424     0x8000000000800000ull, /* 84 */
1425     0x8000000000800080ull, /* 85 */
1426     0x8000000000808000ull, /* 86 */
1427     0x8000000000808080ull, /* 87 */
1428     0x8000000080000000ull, /* 88 */
1429     0x8000000080000080ull, /* 89 */
1430     0x8000000080008000ull, /* 8A */
1431     0x8000000080008080ull, /* 8B */
1432     0x8000000080800000ull, /* 8C */
1433     0x8000000080800080ull, /* 8D */
1434     0x8000000080808000ull, /* 8E */
1435     0x8000000080808080ull, /* 8F */
1436     0x8000008000000000ull, /* 90 */
1437     0x8000008000000080ull, /* 91 */
1438     0x8000008000008000ull, /* 92 */
1439     0x8000008000008080ull, /* 93 */
1440     0x8000008000800000ull, /* 94 */
1441     0x8000008000800080ull, /* 95 */
1442     0x8000008000808000ull, /* 96 */
1443     0x8000008000808080ull, /* 97 */
1444     0x8000008080000000ull, /* 98 */
1445     0x8000008080000080ull, /* 99 */
1446     0x8000008080008000ull, /* 9A */
1447     0x8000008080008080ull, /* 9B */
1448     0x8000008080800000ull, /* 9C */
1449     0x8000008080800080ull, /* 9D */
1450     0x8000008080808000ull, /* 9E */
1451     0x8000008080808080ull, /* 9F */
1452     0x8000800000000000ull, /* A0 */
1453     0x8000800000000080ull, /* A1 */
1454     0x8000800000008000ull, /* A2 */
1455     0x8000800000008080ull, /* A3 */
1456     0x8000800000800000ull, /* A4 */
1457     0x8000800000800080ull, /* A5 */
1458     0x8000800000808000ull, /* A6 */
1459     0x8000800000808080ull, /* A7 */
1460     0x8000800080000000ull, /* A8 */
1461     0x8000800080000080ull, /* A9 */
1462     0x8000800080008000ull, /* AA */
1463     0x8000800080008080ull, /* AB */
1464     0x8000800080800000ull, /* AC */
1465     0x8000800080800080ull, /* AD */
1466     0x8000800080808000ull, /* AE */
1467     0x8000800080808080ull, /* AF */
1468     0x8000808000000000ull, /* B0 */
1469     0x8000808000000080ull, /* B1 */
1470     0x8000808000008000ull, /* B2 */
1471     0x8000808000008080ull, /* B3 */
1472     0x8000808000800000ull, /* B4 */
1473     0x8000808000800080ull, /* B5 */
1474     0x8000808000808000ull, /* B6 */
1475     0x8000808000808080ull, /* B7 */
1476     0x8000808080000000ull, /* B8 */
1477     0x8000808080000080ull, /* B9 */
1478     0x8000808080008000ull, /* BA */
1479     0x8000808080008080ull, /* BB */
1480     0x8000808080800000ull, /* BC */
1481     0x8000808080800080ull, /* BD */
1482     0x8000808080808000ull, /* BE */
1483     0x8000808080808080ull, /* BF */
1484     0x8080000000000000ull, /* C0 */
1485     0x8080000000000080ull, /* C1 */
1486     0x8080000000008000ull, /* C2 */
1487     0x8080000000008080ull, /* C3 */
1488     0x8080000000800000ull, /* C4 */
1489     0x8080000000800080ull, /* C5 */
1490     0x8080000000808000ull, /* C6 */
1491     0x8080000000808080ull, /* C7 */
1492     0x8080000080000000ull, /* C8 */
1493     0x8080000080000080ull, /* C9 */
1494     0x8080000080008000ull, /* CA */
1495     0x8080000080008080ull, /* CB */
1496     0x8080000080800000ull, /* CC */
1497     0x8080000080800080ull, /* CD */
1498     0x8080000080808000ull, /* CE */
1499     0x8080000080808080ull, /* CF */
1500     0x8080008000000000ull, /* D0 */
1501     0x8080008000000080ull, /* D1 */
1502     0x8080008000008000ull, /* D2 */
1503     0x8080008000008080ull, /* D3 */
1504     0x8080008000800000ull, /* D4 */
1505     0x8080008000800080ull, /* D5 */
1506     0x8080008000808000ull, /* D6 */
1507     0x8080008000808080ull, /* D7 */
1508     0x8080008080000000ull, /* D8 */
1509     0x8080008080000080ull, /* D9 */
1510     0x8080008080008000ull, /* DA */
1511     0x8080008080008080ull, /* DB */
1512     0x8080008080800000ull, /* DC */
1513     0x8080008080800080ull, /* DD */
1514     0x8080008080808000ull, /* DE */
1515     0x8080008080808080ull, /* DF */
1516     0x8080800000000000ull, /* E0 */
1517     0x8080800000000080ull, /* E1 */
1518     0x8080800000008000ull, /* E2 */
1519     0x8080800000008080ull, /* E3 */
1520     0x8080800000800000ull, /* E4 */
1521     0x8080800000800080ull, /* E5 */
1522     0x8080800000808000ull, /* E6 */
1523     0x8080800000808080ull, /* E7 */
1524     0x8080800080000000ull, /* E8 */
1525     0x8080800080000080ull, /* E9 */
1526     0x8080800080008000ull, /* EA */
1527     0x8080800080008080ull, /* EB */
1528     0x8080800080800000ull, /* EC */
1529     0x8080800080800080ull, /* ED */
1530     0x8080800080808000ull, /* EE */
1531     0x8080800080808080ull, /* EF */
1532     0x8080808000000000ull, /* F0 */
1533     0x8080808000000080ull, /* F1 */
1534     0x8080808000008000ull, /* F2 */
1535     0x8080808000008080ull, /* F3 */
1536     0x8080808000800000ull, /* F4 */
1537     0x8080808000800080ull, /* F5 */
1538     0x8080808000808000ull, /* F6 */
1539     0x8080808000808080ull, /* F7 */
1540     0x8080808080000000ull, /* F8 */
1541     0x8080808080000080ull, /* F9 */
1542     0x8080808080008000ull, /* FA */
1543     0x8080808080008080ull, /* FB */
1544     0x8080808080800000ull, /* FC */
1545     0x8080808080800080ull, /* FD */
1546     0x8080808080808000ull, /* FE */
1547     0x8080808080808080ull, /* FF */
1548 };
1549 
1550 void helper_vgbbd(ppc_avr_t *r, ppc_avr_t *b)
1551 {
1552     int i;
1553     uint64_t t[2] = { 0, 0 };
1554 
1555     VECTOR_FOR_INORDER_I(i, u8) {
1556 #if defined(HOST_WORDS_BIGENDIAN)
1557         t[i>>3] |= VGBBD_MASKS[b->u8[i]] >> (i & 7);
1558 #else
1559         t[i>>3] |= VGBBD_MASKS[b->u8[i]] >> (7-(i & 7));
1560 #endif
1561     }
1562 
1563     r->u64[0] = t[0];
1564     r->u64[1] = t[1];
1565 }
1566 
1567 #define PMSUM(name, srcfld, trgfld, trgtyp)                   \
1568 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)  \
1569 {                                                             \
1570     int i, j;                                                 \
1571     trgtyp prod[sizeof(ppc_avr_t)/sizeof(a->srcfld[0])];      \
1572                                                               \
1573     VECTOR_FOR_INORDER_I(i, srcfld) {                         \
1574         prod[i] = 0;                                          \
1575         for (j = 0; j < sizeof(a->srcfld[0]) * 8; j++) {      \
1576             if (a->srcfld[i] & (1ull<<j)) {                   \
1577                 prod[i] ^= ((trgtyp)b->srcfld[i] << j);       \
1578             }                                                 \
1579         }                                                     \
1580     }                                                         \
1581                                                               \
1582     VECTOR_FOR_INORDER_I(i, trgfld) {                         \
1583         r->trgfld[i] = prod[2*i] ^ prod[2*i+1];               \
1584     }                                                         \
1585 }
1586 
1587 PMSUM(vpmsumb, u8, u16, uint16_t)
1588 PMSUM(vpmsumh, u16, u32, uint32_t)
1589 PMSUM(vpmsumw, u32, u64, uint64_t)
1590 
1591 void helper_vpmsumd(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1592 {
1593 
1594 #ifdef CONFIG_INT128
1595     int i, j;
1596     __uint128_t prod[2];
1597 
1598     VECTOR_FOR_INORDER_I(i, u64) {
1599         prod[i] = 0;
1600         for (j = 0; j < 64; j++) {
1601             if (a->u64[i] & (1ull<<j)) {
1602                 prod[i] ^= (((__uint128_t)b->u64[i]) << j);
1603             }
1604         }
1605     }
1606 
1607     r->u128 = prod[0] ^ prod[1];
1608 
1609 #else
1610     int i, j;
1611     ppc_avr_t prod[2];
1612 
1613     VECTOR_FOR_INORDER_I(i, u64) {
1614         prod[i].u64[LO_IDX] = prod[i].u64[HI_IDX] = 0;
1615         for (j = 0; j < 64; j++) {
1616             if (a->u64[i] & (1ull<<j)) {
1617                 ppc_avr_t bshift;
1618                 if (j == 0) {
1619                     bshift.u64[HI_IDX] = 0;
1620                     bshift.u64[LO_IDX] = b->u64[i];
1621                 } else {
1622                     bshift.u64[HI_IDX] = b->u64[i] >> (64-j);
1623                     bshift.u64[LO_IDX] = b->u64[i] << j;
1624                 }
1625                 prod[i].u64[LO_IDX] ^= bshift.u64[LO_IDX];
1626                 prod[i].u64[HI_IDX] ^= bshift.u64[HI_IDX];
1627             }
1628         }
1629     }
1630 
1631     r->u64[LO_IDX] = prod[0].u64[LO_IDX] ^ prod[1].u64[LO_IDX];
1632     r->u64[HI_IDX] = prod[0].u64[HI_IDX] ^ prod[1].u64[HI_IDX];
1633 #endif
1634 }
1635 
1636 
1637 #if defined(HOST_WORDS_BIGENDIAN)
1638 #define PKBIG 1
1639 #else
1640 #define PKBIG 0
1641 #endif
1642 void helper_vpkpx(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1643 {
1644     int i, j;
1645     ppc_avr_t result;
1646 #if defined(HOST_WORDS_BIGENDIAN)
1647     const ppc_avr_t *x[2] = { a, b };
1648 #else
1649     const ppc_avr_t *x[2] = { b, a };
1650 #endif
1651 
1652     VECTOR_FOR_INORDER_I(i, u64) {
1653         VECTOR_FOR_INORDER_I(j, u32) {
1654             uint32_t e = x[i]->u32[j];
1655 
1656             result.u16[4*i+j] = (((e >> 9) & 0xfc00) |
1657                                  ((e >> 6) & 0x3e0) |
1658                                  ((e >> 3) & 0x1f));
1659         }
1660     }
1661     *r = result;
1662 }
1663 
1664 #define VPK(suffix, from, to, cvt, dosat)                               \
1665     void helper_vpk##suffix(CPUPPCState *env, ppc_avr_t *r,             \
1666                             ppc_avr_t *a, ppc_avr_t *b)                 \
1667     {                                                                   \
1668         int i;                                                          \
1669         int sat = 0;                                                    \
1670         ppc_avr_t result;                                               \
1671         ppc_avr_t *a0 = PKBIG ? a : b;                                  \
1672         ppc_avr_t *a1 = PKBIG ? b : a;                                  \
1673                                                                         \
1674         VECTOR_FOR_INORDER_I(i, from) {                                 \
1675             result.to[i] = cvt(a0->from[i], &sat);                      \
1676             result.to[i+ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat);  \
1677         }                                                               \
1678         *r = result;                                                    \
1679         if (dosat && sat) {                                             \
1680             env->vscr |= (1 << VSCR_SAT);                               \
1681         }                                                               \
1682     }
1683 #define I(x, y) (x)
1684 VPK(shss, s16, s8, cvtshsb, 1)
1685 VPK(shus, s16, u8, cvtshub, 1)
1686 VPK(swss, s32, s16, cvtswsh, 1)
1687 VPK(swus, s32, u16, cvtswuh, 1)
1688 VPK(sdss, s64, s32, cvtsdsw, 1)
1689 VPK(sdus, s64, u32, cvtsduw, 1)
1690 VPK(uhus, u16, u8, cvtuhub, 1)
1691 VPK(uwus, u32, u16, cvtuwuh, 1)
1692 VPK(udus, u64, u32, cvtuduw, 1)
1693 VPK(uhum, u16, u8, I, 0)
1694 VPK(uwum, u32, u16, I, 0)
1695 VPK(udum, u64, u32, I, 0)
1696 #undef I
1697 #undef VPK
1698 #undef PKBIG
1699 
1700 void helper_vrefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1701 {
1702     int i;
1703 
1704     for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1705         r->f[i] = float32_div(float32_one, b->f[i], &env->vec_status);
1706     }
1707 }
1708 
1709 #define VRFI(suffix, rounding)                                  \
1710     void helper_vrfi##suffix(CPUPPCState *env, ppc_avr_t *r,    \
1711                              ppc_avr_t *b)                      \
1712     {                                                           \
1713         int i;                                                  \
1714         float_status s = env->vec_status;                       \
1715                                                                 \
1716         set_float_rounding_mode(rounding, &s);                  \
1717         for (i = 0; i < ARRAY_SIZE(r->f); i++) {                \
1718             r->f[i] = float32_round_to_int (b->f[i], &s);       \
1719         }                                                       \
1720     }
1721 VRFI(n, float_round_nearest_even)
1722 VRFI(m, float_round_down)
1723 VRFI(p, float_round_up)
1724 VRFI(z, float_round_to_zero)
1725 #undef VRFI
1726 
1727 #define VROTATE(suffix, element, mask)                                  \
1728     void helper_vrl##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)   \
1729     {                                                                   \
1730         int i;                                                          \
1731                                                                         \
1732         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
1733             unsigned int shift = b->element[i] & mask;                  \
1734             r->element[i] = (a->element[i] << shift) |                  \
1735                 (a->element[i] >> (sizeof(a->element[0]) * 8 - shift)); \
1736         }                                                               \
1737     }
1738 VROTATE(b, u8, 0x7)
1739 VROTATE(h, u16, 0xF)
1740 VROTATE(w, u32, 0x1F)
1741 VROTATE(d, u64, 0x3F)
1742 #undef VROTATE
1743 
1744 void helper_vrsqrtefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1745 {
1746     int i;
1747 
1748     for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1749         float32 t = float32_sqrt(b->f[i], &env->vec_status);
1750 
1751         r->f[i] = float32_div(float32_one, t, &env->vec_status);
1752     }
1753 }
1754 
1755 #define VRLMI(name, size, element, insert)                            \
1756 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)          \
1757 {                                                                     \
1758     int i;                                                            \
1759     for (i = 0; i < ARRAY_SIZE(r->element); i++) {                    \
1760         uint##size##_t src1 = a->element[i];                          \
1761         uint##size##_t src2 = b->element[i];                          \
1762         uint##size##_t src3 = r->element[i];                          \
1763         uint##size##_t begin, end, shift, mask, rot_val;              \
1764                                                                       \
1765         shift = extract##size(src2, 0, 6);                            \
1766         end   = extract##size(src2, 8, 6);                            \
1767         begin = extract##size(src2, 16, 6);                           \
1768         rot_val = rol##size(src1, shift);                             \
1769         mask = mask_u##size(begin, end);                              \
1770         if (insert) {                                                 \
1771             r->element[i] = (rot_val & mask) | (src3 & ~mask);        \
1772         } else {                                                      \
1773             r->element[i] = (rot_val & mask);                         \
1774         }                                                             \
1775     }                                                                 \
1776 }
1777 
1778 VRLMI(vrldmi, 64, u64, 1);
1779 VRLMI(vrlwmi, 32, u32, 1);
1780 VRLMI(vrldnm, 64, u64, 0);
1781 VRLMI(vrlwnm, 32, u32, 0);
1782 
1783 void helper_vsel(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b,
1784                  ppc_avr_t *c)
1785 {
1786     r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]);
1787     r->u64[1] = (a->u64[1] & ~c->u64[1]) | (b->u64[1] & c->u64[1]);
1788 }
1789 
1790 void helper_vexptefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1791 {
1792     int i;
1793 
1794     for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1795         r->f[i] = float32_exp2(b->f[i], &env->vec_status);
1796     }
1797 }
1798 
1799 void helper_vlogefp(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *b)
1800 {
1801     int i;
1802 
1803     for (i = 0; i < ARRAY_SIZE(r->f); i++) {
1804         r->f[i] = float32_log2(b->f[i], &env->vec_status);
1805     }
1806 }
1807 
1808 /* The specification says that the results are undefined if all of the
1809  * shift counts are not identical.  We check to make sure that they are
1810  * to conform to what real hardware appears to do.  */
1811 #define VSHIFT(suffix, leftp)                                           \
1812     void helper_vs##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)    \
1813     {                                                                   \
1814         int shift = b->u8[LO_IDX*15] & 0x7;                             \
1815         int doit = 1;                                                   \
1816         int i;                                                          \
1817                                                                         \
1818         for (i = 0; i < ARRAY_SIZE(r->u8); i++) {                       \
1819             doit = doit && ((b->u8[i] & 0x7) == shift);                 \
1820         }                                                               \
1821         if (doit) {                                                     \
1822             if (shift == 0) {                                           \
1823                 *r = *a;                                                \
1824             } else if (leftp) {                                         \
1825                 uint64_t carry = a->u64[LO_IDX] >> (64 - shift);        \
1826                                                                         \
1827                 r->u64[HI_IDX] = (a->u64[HI_IDX] << shift) | carry;     \
1828                 r->u64[LO_IDX] = a->u64[LO_IDX] << shift;               \
1829             } else {                                                    \
1830                 uint64_t carry = a->u64[HI_IDX] << (64 - shift);        \
1831                                                                         \
1832                 r->u64[LO_IDX] = (a->u64[LO_IDX] >> shift) | carry;     \
1833                 r->u64[HI_IDX] = a->u64[HI_IDX] >> shift;               \
1834             }                                                           \
1835         }                                                               \
1836     }
1837 VSHIFT(l, 1)
1838 VSHIFT(r, 0)
1839 #undef VSHIFT
1840 
1841 #define VSL(suffix, element, mask)                                      \
1842     void helper_vsl##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)   \
1843     {                                                                   \
1844         int i;                                                          \
1845                                                                         \
1846         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
1847             unsigned int shift = b->element[i] & mask;                  \
1848                                                                         \
1849             r->element[i] = a->element[i] << shift;                     \
1850         }                                                               \
1851     }
1852 VSL(b, u8, 0x7)
1853 VSL(h, u16, 0x0F)
1854 VSL(w, u32, 0x1F)
1855 VSL(d, u64, 0x3F)
1856 #undef VSL
1857 
1858 void helper_vslv(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1859 {
1860     int i;
1861     unsigned int shift, bytes, size;
1862 
1863     size = ARRAY_SIZE(r->u8);
1864     for (i = 0; i < size; i++) {
1865         shift = b->u8[i] & 0x7;             /* extract shift value */
1866         bytes = (a->u8[i] << 8) +             /* extract adjacent bytes */
1867             (((i + 1) < size) ? a->u8[i + 1] : 0);
1868         r->u8[i] = (bytes << shift) >> 8;   /* shift and store result */
1869     }
1870 }
1871 
1872 void helper_vsrv(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1873 {
1874     int i;
1875     unsigned int shift, bytes;
1876 
1877     /* Use reverse order, as destination and source register can be same. Its
1878      * being modified in place saving temporary, reverse order will guarantee
1879      * that computed result is not fed back.
1880      */
1881     for (i = ARRAY_SIZE(r->u8) - 1; i >= 0; i--) {
1882         shift = b->u8[i] & 0x7;                 /* extract shift value */
1883         bytes = ((i ? a->u8[i - 1] : 0) << 8) + a->u8[i];
1884                                                 /* extract adjacent bytes */
1885         r->u8[i] = (bytes >> shift) & 0xFF;     /* shift and store result */
1886     }
1887 }
1888 
1889 void helper_vsldoi(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t shift)
1890 {
1891     int sh = shift & 0xf;
1892     int i;
1893     ppc_avr_t result;
1894 
1895 #if defined(HOST_WORDS_BIGENDIAN)
1896     for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1897         int index = sh + i;
1898         if (index > 0xf) {
1899             result.u8[i] = b->u8[index - 0x10];
1900         } else {
1901             result.u8[i] = a->u8[index];
1902         }
1903     }
1904 #else
1905     for (i = 0; i < ARRAY_SIZE(r->u8); i++) {
1906         int index = (16 - sh) + i;
1907         if (index > 0xf) {
1908             result.u8[i] = a->u8[index - 0x10];
1909         } else {
1910             result.u8[i] = b->u8[index];
1911         }
1912     }
1913 #endif
1914     *r = result;
1915 }
1916 
1917 void helper_vslo(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
1918 {
1919     int sh = (b->u8[LO_IDX*0xf] >> 3) & 0xf;
1920 
1921 #if defined(HOST_WORDS_BIGENDIAN)
1922     memmove(&r->u8[0], &a->u8[sh], 16 - sh);
1923     memset(&r->u8[16-sh], 0, sh);
1924 #else
1925     memmove(&r->u8[sh], &a->u8[0], 16 - sh);
1926     memset(&r->u8[0], 0, sh);
1927 #endif
1928 }
1929 
1930 /* Experimental testing shows that hardware masks the immediate.  */
1931 #define _SPLAT_MASKED(element) (splat & (ARRAY_SIZE(r->element) - 1))
1932 #if defined(HOST_WORDS_BIGENDIAN)
1933 #define SPLAT_ELEMENT(element) _SPLAT_MASKED(element)
1934 #else
1935 #define SPLAT_ELEMENT(element)                                  \
1936     (ARRAY_SIZE(r->element) - 1 - _SPLAT_MASKED(element))
1937 #endif
1938 #define VSPLT(suffix, element)                                          \
1939     void helper_vsplt##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t splat) \
1940     {                                                                   \
1941         uint32_t s = b->element[SPLAT_ELEMENT(element)];                \
1942         int i;                                                          \
1943                                                                         \
1944         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
1945             r->element[i] = s;                                          \
1946         }                                                               \
1947     }
1948 VSPLT(b, u8)
1949 VSPLT(h, u16)
1950 VSPLT(w, u32)
1951 #undef VSPLT
1952 #undef SPLAT_ELEMENT
1953 #undef _SPLAT_MASKED
1954 #if defined(HOST_WORDS_BIGENDIAN)
1955 #define VINSERT(suffix, element)                                            \
1956     void helper_vinsert##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1957     {                                                                       \
1958         memmove(&r->u8[index], &b->u8[8 - sizeof(r->element)],              \
1959                sizeof(r->element[0]));                                      \
1960     }
1961 #else
1962 #define VINSERT(suffix, element)                                            \
1963     void helper_vinsert##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1964     {                                                                       \
1965         uint32_t d = (16 - index) - sizeof(r->element[0]);                  \
1966         memmove(&r->u8[d], &b->u8[8], sizeof(r->element[0]));               \
1967     }
1968 #endif
1969 VINSERT(b, u8)
1970 VINSERT(h, u16)
1971 VINSERT(w, u32)
1972 VINSERT(d, u64)
1973 #undef VINSERT
1974 #if defined(HOST_WORDS_BIGENDIAN)
1975 #define VEXTRACT(suffix, element)                                            \
1976     void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1977     {                                                                        \
1978         uint32_t es = sizeof(r->element[0]);                                 \
1979         memmove(&r->u8[8 - es], &b->u8[index], es);                          \
1980         memset(&r->u8[8], 0, 8);                                             \
1981         memset(&r->u8[0], 0, 8 - es);                                        \
1982     }
1983 #else
1984 #define VEXTRACT(suffix, element)                                            \
1985     void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1986     {                                                                        \
1987         uint32_t es = sizeof(r->element[0]);                                 \
1988         uint32_t s = (16 - index) - es;                                      \
1989         memmove(&r->u8[8], &b->u8[s], es);                                   \
1990         memset(&r->u8[0], 0, 8);                                             \
1991         memset(&r->u8[8 + es], 0, 8 - es);                                   \
1992     }
1993 #endif
1994 VEXTRACT(ub, u8)
1995 VEXTRACT(uh, u16)
1996 VEXTRACT(uw, u32)
1997 VEXTRACT(d, u64)
1998 #undef VEXTRACT
1999 
2000 #define VEXT_SIGNED(name, element, mask, cast, recast)              \
2001 void helper_##name(ppc_avr_t *r, ppc_avr_t *b)                      \
2002 {                                                                   \
2003     int i;                                                          \
2004     VECTOR_FOR_INORDER_I(i, element) {                              \
2005         r->element[i] = (recast)((cast)(b->element[i] & mask));     \
2006     }                                                               \
2007 }
2008 VEXT_SIGNED(vextsb2w, s32, UINT8_MAX, int8_t, int32_t)
2009 VEXT_SIGNED(vextsb2d, s64, UINT8_MAX, int8_t, int64_t)
2010 VEXT_SIGNED(vextsh2w, s32, UINT16_MAX, int16_t, int32_t)
2011 VEXT_SIGNED(vextsh2d, s64, UINT16_MAX, int16_t, int64_t)
2012 VEXT_SIGNED(vextsw2d, s64, UINT32_MAX, int32_t, int64_t)
2013 #undef VEXT_SIGNED
2014 
2015 #define VNEG(name, element)                                         \
2016 void helper_##name(ppc_avr_t *r, ppc_avr_t *b)                      \
2017 {                                                                   \
2018     int i;                                                          \
2019     VECTOR_FOR_INORDER_I(i, element) {                              \
2020         r->element[i] = -b->element[i];                             \
2021     }                                                               \
2022 }
2023 VNEG(vnegw, s32)
2024 VNEG(vnegd, s64)
2025 #undef VNEG
2026 
2027 #define VSPLTI(suffix, element, splat_type)                     \
2028     void helper_vspltis##suffix(ppc_avr_t *r, uint32_t splat)   \
2029     {                                                           \
2030         splat_type x = (int8_t)(splat << 3) >> 3;               \
2031         int i;                                                  \
2032                                                                 \
2033         for (i = 0; i < ARRAY_SIZE(r->element); i++) {          \
2034             r->element[i] = x;                                  \
2035         }                                                       \
2036     }
2037 VSPLTI(b, s8, int8_t)
2038 VSPLTI(h, s16, int16_t)
2039 VSPLTI(w, s32, int32_t)
2040 #undef VSPLTI
2041 
2042 #define VSR(suffix, element, mask)                                      \
2043     void helper_vsr##suffix(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)   \
2044     {                                                                   \
2045         int i;                                                          \
2046                                                                         \
2047         for (i = 0; i < ARRAY_SIZE(r->element); i++) {                  \
2048             unsigned int shift = b->element[i] & mask;                  \
2049             r->element[i] = a->element[i] >> shift;                     \
2050         }                                                               \
2051     }
2052 VSR(ab, s8, 0x7)
2053 VSR(ah, s16, 0xF)
2054 VSR(aw, s32, 0x1F)
2055 VSR(ad, s64, 0x3F)
2056 VSR(b, u8, 0x7)
2057 VSR(h, u16, 0xF)
2058 VSR(w, u32, 0x1F)
2059 VSR(d, u64, 0x3F)
2060 #undef VSR
2061 
2062 void helper_vsro(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2063 {
2064     int sh = (b->u8[LO_IDX * 0xf] >> 3) & 0xf;
2065 
2066 #if defined(HOST_WORDS_BIGENDIAN)
2067     memmove(&r->u8[sh], &a->u8[0], 16 - sh);
2068     memset(&r->u8[0], 0, sh);
2069 #else
2070     memmove(&r->u8[0], &a->u8[sh], 16 - sh);
2071     memset(&r->u8[16 - sh], 0, sh);
2072 #endif
2073 }
2074 
2075 void helper_vsubcuw(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2076 {
2077     int i;
2078 
2079     for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
2080         r->u32[i] = a->u32[i] >= b->u32[i];
2081     }
2082 }
2083 
2084 void helper_vsumsws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2085 {
2086     int64_t t;
2087     int i, upper;
2088     ppc_avr_t result;
2089     int sat = 0;
2090 
2091 #if defined(HOST_WORDS_BIGENDIAN)
2092     upper = ARRAY_SIZE(r->s32)-1;
2093 #else
2094     upper = 0;
2095 #endif
2096     t = (int64_t)b->s32[upper];
2097     for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
2098         t += a->s32[i];
2099         result.s32[i] = 0;
2100     }
2101     result.s32[upper] = cvtsdsw(t, &sat);
2102     *r = result;
2103 
2104     if (sat) {
2105         env->vscr |= (1 << VSCR_SAT);
2106     }
2107 }
2108 
2109 void helper_vsum2sws(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2110 {
2111     int i, j, upper;
2112     ppc_avr_t result;
2113     int sat = 0;
2114 
2115 #if defined(HOST_WORDS_BIGENDIAN)
2116     upper = 1;
2117 #else
2118     upper = 0;
2119 #endif
2120     for (i = 0; i < ARRAY_SIZE(r->u64); i++) {
2121         int64_t t = (int64_t)b->s32[upper + i * 2];
2122 
2123         result.u64[i] = 0;
2124         for (j = 0; j < ARRAY_SIZE(r->u64); j++) {
2125             t += a->s32[2 * i + j];
2126         }
2127         result.s32[upper + i * 2] = cvtsdsw(t, &sat);
2128     }
2129 
2130     *r = result;
2131     if (sat) {
2132         env->vscr |= (1 << VSCR_SAT);
2133     }
2134 }
2135 
2136 void helper_vsum4sbs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2137 {
2138     int i, j;
2139     int sat = 0;
2140 
2141     for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
2142         int64_t t = (int64_t)b->s32[i];
2143 
2144         for (j = 0; j < ARRAY_SIZE(r->s32); j++) {
2145             t += a->s8[4 * i + j];
2146         }
2147         r->s32[i] = cvtsdsw(t, &sat);
2148     }
2149 
2150     if (sat) {
2151         env->vscr |= (1 << VSCR_SAT);
2152     }
2153 }
2154 
2155 void helper_vsum4shs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2156 {
2157     int sat = 0;
2158     int i;
2159 
2160     for (i = 0; i < ARRAY_SIZE(r->s32); i++) {
2161         int64_t t = (int64_t)b->s32[i];
2162 
2163         t += a->s16[2 * i] + a->s16[2 * i + 1];
2164         r->s32[i] = cvtsdsw(t, &sat);
2165     }
2166 
2167     if (sat) {
2168         env->vscr |= (1 << VSCR_SAT);
2169     }
2170 }
2171 
2172 void helper_vsum4ubs(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2173 {
2174     int i, j;
2175     int sat = 0;
2176 
2177     for (i = 0; i < ARRAY_SIZE(r->u32); i++) {
2178         uint64_t t = (uint64_t)b->u32[i];
2179 
2180         for (j = 0; j < ARRAY_SIZE(r->u32); j++) {
2181             t += a->u8[4 * i + j];
2182         }
2183         r->u32[i] = cvtuduw(t, &sat);
2184     }
2185 
2186     if (sat) {
2187         env->vscr |= (1 << VSCR_SAT);
2188     }
2189 }
2190 
2191 #if defined(HOST_WORDS_BIGENDIAN)
2192 #define UPKHI 1
2193 #define UPKLO 0
2194 #else
2195 #define UPKHI 0
2196 #define UPKLO 1
2197 #endif
2198 #define VUPKPX(suffix, hi)                                              \
2199     void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b)                \
2200     {                                                                   \
2201         int i;                                                          \
2202         ppc_avr_t result;                                               \
2203                                                                         \
2204         for (i = 0; i < ARRAY_SIZE(r->u32); i++) {                      \
2205             uint16_t e = b->u16[hi ? i : i+4];                          \
2206             uint8_t a = (e >> 15) ? 0xff : 0;                           \
2207             uint8_t r = (e >> 10) & 0x1f;                               \
2208             uint8_t g = (e >> 5) & 0x1f;                                \
2209             uint8_t b = e & 0x1f;                                       \
2210                                                                         \
2211             result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b;       \
2212         }                                                               \
2213         *r = result;                                                    \
2214     }
2215 VUPKPX(lpx, UPKLO)
2216 VUPKPX(hpx, UPKHI)
2217 #undef VUPKPX
2218 
2219 #define VUPK(suffix, unpacked, packee, hi)                              \
2220     void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b)                \
2221     {                                                                   \
2222         int i;                                                          \
2223         ppc_avr_t result;                                               \
2224                                                                         \
2225         if (hi) {                                                       \
2226             for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) {             \
2227                 result.unpacked[i] = b->packee[i];                      \
2228             }                                                           \
2229         } else {                                                        \
2230             for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); \
2231                  i++) {                                                 \
2232                 result.unpacked[i - ARRAY_SIZE(r->unpacked)] = b->packee[i]; \
2233             }                                                           \
2234         }                                                               \
2235         *r = result;                                                    \
2236     }
2237 VUPK(hsb, s16, s8, UPKHI)
2238 VUPK(hsh, s32, s16, UPKHI)
2239 VUPK(hsw, s64, s32, UPKHI)
2240 VUPK(lsb, s16, s8, UPKLO)
2241 VUPK(lsh, s32, s16, UPKLO)
2242 VUPK(lsw, s64, s32, UPKLO)
2243 #undef VUPK
2244 #undef UPKHI
2245 #undef UPKLO
2246 
2247 #define VGENERIC_DO(name, element)                                      \
2248     void helper_v##name(ppc_avr_t *r, ppc_avr_t *b)                     \
2249     {                                                                   \
2250         int i;                                                          \
2251                                                                         \
2252         VECTOR_FOR_INORDER_I(i, element) {                              \
2253             r->element[i] = name(b->element[i]);                        \
2254         }                                                               \
2255     }
2256 
2257 #define clzb(v) ((v) ? clz32((uint32_t)(v) << 24) : 8)
2258 #define clzh(v) ((v) ? clz32((uint32_t)(v) << 16) : 16)
2259 #define clzw(v) clz32((v))
2260 #define clzd(v) clz64((v))
2261 
2262 VGENERIC_DO(clzb, u8)
2263 VGENERIC_DO(clzh, u16)
2264 VGENERIC_DO(clzw, u32)
2265 VGENERIC_DO(clzd, u64)
2266 
2267 #undef clzb
2268 #undef clzh
2269 #undef clzw
2270 #undef clzd
2271 
2272 #define ctzb(v) ((v) ? ctz32(v) : 8)
2273 #define ctzh(v) ((v) ? ctz32(v) : 16)
2274 #define ctzw(v) ctz32((v))
2275 #define ctzd(v) ctz64((v))
2276 
2277 VGENERIC_DO(ctzb, u8)
2278 VGENERIC_DO(ctzh, u16)
2279 VGENERIC_DO(ctzw, u32)
2280 VGENERIC_DO(ctzd, u64)
2281 
2282 #undef ctzb
2283 #undef ctzh
2284 #undef ctzw
2285 #undef ctzd
2286 
2287 #define popcntb(v) ctpop8(v)
2288 #define popcnth(v) ctpop16(v)
2289 #define popcntw(v) ctpop32(v)
2290 #define popcntd(v) ctpop64(v)
2291 
2292 VGENERIC_DO(popcntb, u8)
2293 VGENERIC_DO(popcnth, u16)
2294 VGENERIC_DO(popcntw, u32)
2295 VGENERIC_DO(popcntd, u64)
2296 
2297 #undef popcntb
2298 #undef popcnth
2299 #undef popcntw
2300 #undef popcntd
2301 
2302 #undef VGENERIC_DO
2303 
2304 #if defined(HOST_WORDS_BIGENDIAN)
2305 #define QW_ONE { .u64 = { 0, 1 } }
2306 #else
2307 #define QW_ONE { .u64 = { 1, 0 } }
2308 #endif
2309 
2310 #ifndef CONFIG_INT128
2311 
2312 static inline void avr_qw_not(ppc_avr_t *t, ppc_avr_t a)
2313 {
2314     t->u64[0] = ~a.u64[0];
2315     t->u64[1] = ~a.u64[1];
2316 }
2317 
2318 static int avr_qw_cmpu(ppc_avr_t a, ppc_avr_t b)
2319 {
2320     if (a.u64[HI_IDX] < b.u64[HI_IDX]) {
2321         return -1;
2322     } else if (a.u64[HI_IDX] > b.u64[HI_IDX]) {
2323         return 1;
2324     } else if (a.u64[LO_IDX] < b.u64[LO_IDX]) {
2325         return -1;
2326     } else if (a.u64[LO_IDX] > b.u64[LO_IDX]) {
2327         return 1;
2328     } else {
2329         return 0;
2330     }
2331 }
2332 
2333 static void avr_qw_add(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b)
2334 {
2335     t->u64[LO_IDX] = a.u64[LO_IDX] + b.u64[LO_IDX];
2336     t->u64[HI_IDX] = a.u64[HI_IDX] + b.u64[HI_IDX] +
2337                      (~a.u64[LO_IDX] < b.u64[LO_IDX]);
2338 }
2339 
2340 static int avr_qw_addc(ppc_avr_t *t, ppc_avr_t a, ppc_avr_t b)
2341 {
2342     ppc_avr_t not_a;
2343     t->u64[LO_IDX] = a.u64[LO_IDX] + b.u64[LO_IDX];
2344     t->u64[HI_IDX] = a.u64[HI_IDX] + b.u64[HI_IDX] +
2345                      (~a.u64[LO_IDX] < b.u64[LO_IDX]);
2346     avr_qw_not(&not_a, a);
2347     return avr_qw_cmpu(not_a, b) < 0;
2348 }
2349 
2350 #endif
2351 
2352 void helper_vadduqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2353 {
2354 #ifdef CONFIG_INT128
2355     r->u128 = a->u128 + b->u128;
2356 #else
2357     avr_qw_add(r, *a, *b);
2358 #endif
2359 }
2360 
2361 void helper_vaddeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2362 {
2363 #ifdef CONFIG_INT128
2364     r->u128 = a->u128 + b->u128 + (c->u128 & 1);
2365 #else
2366 
2367     if (c->u64[LO_IDX] & 1) {
2368         ppc_avr_t tmp;
2369 
2370         tmp.u64[HI_IDX] = 0;
2371         tmp.u64[LO_IDX] = c->u64[LO_IDX] & 1;
2372         avr_qw_add(&tmp, *a, tmp);
2373         avr_qw_add(r, tmp, *b);
2374     } else {
2375         avr_qw_add(r, *a, *b);
2376     }
2377 #endif
2378 }
2379 
2380 void helper_vaddcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2381 {
2382 #ifdef CONFIG_INT128
2383     r->u128 = (~a->u128 < b->u128);
2384 #else
2385     ppc_avr_t not_a;
2386 
2387     avr_qw_not(&not_a, *a);
2388 
2389     r->u64[HI_IDX] = 0;
2390     r->u64[LO_IDX] = (avr_qw_cmpu(not_a, *b) < 0);
2391 #endif
2392 }
2393 
2394 void helper_vaddecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2395 {
2396 #ifdef CONFIG_INT128
2397     int carry_out = (~a->u128 < b->u128);
2398     if (!carry_out && (c->u128 & 1)) {
2399         carry_out = ((a->u128 + b->u128 + 1) == 0) &&
2400                     ((a->u128 != 0) || (b->u128 != 0));
2401     }
2402     r->u128 = carry_out;
2403 #else
2404 
2405     int carry_in = c->u64[LO_IDX] & 1;
2406     int carry_out = 0;
2407     ppc_avr_t tmp;
2408 
2409     carry_out = avr_qw_addc(&tmp, *a, *b);
2410 
2411     if (!carry_out && carry_in) {
2412         ppc_avr_t one = QW_ONE;
2413         carry_out = avr_qw_addc(&tmp, tmp, one);
2414     }
2415     r->u64[HI_IDX] = 0;
2416     r->u64[LO_IDX] = carry_out;
2417 #endif
2418 }
2419 
2420 void helper_vsubuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2421 {
2422 #ifdef CONFIG_INT128
2423     r->u128 = a->u128 - b->u128;
2424 #else
2425     ppc_avr_t tmp;
2426     ppc_avr_t one = QW_ONE;
2427 
2428     avr_qw_not(&tmp, *b);
2429     avr_qw_add(&tmp, *a, tmp);
2430     avr_qw_add(r, tmp, one);
2431 #endif
2432 }
2433 
2434 void helper_vsubeuqm(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2435 {
2436 #ifdef CONFIG_INT128
2437     r->u128 = a->u128 + ~b->u128 + (c->u128 & 1);
2438 #else
2439     ppc_avr_t tmp, sum;
2440 
2441     avr_qw_not(&tmp, *b);
2442     avr_qw_add(&sum, *a, tmp);
2443 
2444     tmp.u64[HI_IDX] = 0;
2445     tmp.u64[LO_IDX] = c->u64[LO_IDX] & 1;
2446     avr_qw_add(r, sum, tmp);
2447 #endif
2448 }
2449 
2450 void helper_vsubcuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2451 {
2452 #ifdef CONFIG_INT128
2453     r->u128 = (~a->u128 < ~b->u128) ||
2454                  (a->u128 + ~b->u128 == (__uint128_t)-1);
2455 #else
2456     int carry = (avr_qw_cmpu(*a, *b) > 0);
2457     if (!carry) {
2458         ppc_avr_t tmp;
2459         avr_qw_not(&tmp, *b);
2460         avr_qw_add(&tmp, *a, tmp);
2461         carry = ((tmp.s64[HI_IDX] == -1ull) && (tmp.s64[LO_IDX] == -1ull));
2462     }
2463     r->u64[HI_IDX] = 0;
2464     r->u64[LO_IDX] = carry;
2465 #endif
2466 }
2467 
2468 void helper_vsubecuq(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
2469 {
2470 #ifdef CONFIG_INT128
2471     r->u128 =
2472         (~a->u128 < ~b->u128) ||
2473         ((c->u128 & 1) && (a->u128 + ~b->u128 == (__uint128_t)-1));
2474 #else
2475     int carry_in = c->u64[LO_IDX] & 1;
2476     int carry_out = (avr_qw_cmpu(*a, *b) > 0);
2477     if (!carry_out && carry_in) {
2478         ppc_avr_t tmp;
2479         avr_qw_not(&tmp, *b);
2480         avr_qw_add(&tmp, *a, tmp);
2481         carry_out = ((tmp.u64[HI_IDX] == -1ull) && (tmp.u64[LO_IDX] == -1ull));
2482     }
2483 
2484     r->u64[HI_IDX] = 0;
2485     r->u64[LO_IDX] = carry_out;
2486 #endif
2487 }
2488 
2489 #define BCD_PLUS_PREF_1 0xC
2490 #define BCD_PLUS_PREF_2 0xF
2491 #define BCD_PLUS_ALT_1  0xA
2492 #define BCD_NEG_PREF    0xD
2493 #define BCD_NEG_ALT     0xB
2494 #define BCD_PLUS_ALT_2  0xE
2495 #define NATIONAL_PLUS   0x2B
2496 #define NATIONAL_NEG    0x2D
2497 
2498 #if defined(HOST_WORDS_BIGENDIAN)
2499 #define BCD_DIG_BYTE(n) (15 - (n/2))
2500 #else
2501 #define BCD_DIG_BYTE(n) (n/2)
2502 #endif
2503 
2504 static int bcd_get_sgn(ppc_avr_t *bcd)
2505 {
2506     switch (bcd->u8[BCD_DIG_BYTE(0)] & 0xF) {
2507     case BCD_PLUS_PREF_1:
2508     case BCD_PLUS_PREF_2:
2509     case BCD_PLUS_ALT_1:
2510     case BCD_PLUS_ALT_2:
2511     {
2512         return 1;
2513     }
2514 
2515     case BCD_NEG_PREF:
2516     case BCD_NEG_ALT:
2517     {
2518         return -1;
2519     }
2520 
2521     default:
2522     {
2523         return 0;
2524     }
2525     }
2526 }
2527 
2528 static int bcd_preferred_sgn(int sgn, int ps)
2529 {
2530     if (sgn >= 0) {
2531         return (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2;
2532     } else {
2533         return BCD_NEG_PREF;
2534     }
2535 }
2536 
2537 static uint8_t bcd_get_digit(ppc_avr_t *bcd, int n, int *invalid)
2538 {
2539     uint8_t result;
2540     if (n & 1) {
2541         result = bcd->u8[BCD_DIG_BYTE(n)] >> 4;
2542     } else {
2543        result = bcd->u8[BCD_DIG_BYTE(n)] & 0xF;
2544     }
2545 
2546     if (unlikely(result > 9)) {
2547         *invalid = true;
2548     }
2549     return result;
2550 }
2551 
2552 static void bcd_put_digit(ppc_avr_t *bcd, uint8_t digit, int n)
2553 {
2554     if (n & 1) {
2555         bcd->u8[BCD_DIG_BYTE(n)] &= 0x0F;
2556         bcd->u8[BCD_DIG_BYTE(n)] |= (digit<<4);
2557     } else {
2558         bcd->u8[BCD_DIG_BYTE(n)] &= 0xF0;
2559         bcd->u8[BCD_DIG_BYTE(n)] |= digit;
2560     }
2561 }
2562 
2563 static int bcd_cmp_zero(ppc_avr_t *bcd)
2564 {
2565     if (bcd->u64[HI_IDX] == 0 && (bcd->u64[LO_IDX] >> 4) == 0) {
2566         return 1 << CRF_EQ;
2567     } else {
2568         return (bcd_get_sgn(bcd) == 1) ? 1 << CRF_GT : 1 << CRF_LT;
2569     }
2570 }
2571 
2572 static uint16_t get_national_digit(ppc_avr_t *reg, int n)
2573 {
2574 #if defined(HOST_WORDS_BIGENDIAN)
2575     return reg->u16[7 - n];
2576 #else
2577     return reg->u16[n];
2578 #endif
2579 }
2580 
2581 static void set_national_digit(ppc_avr_t *reg, uint8_t val, int n)
2582 {
2583 #if defined(HOST_WORDS_BIGENDIAN)
2584     reg->u16[7 - n] = val;
2585 #else
2586     reg->u16[n] = val;
2587 #endif
2588 }
2589 
2590 static int bcd_cmp_mag(ppc_avr_t *a, ppc_avr_t *b)
2591 {
2592     int i;
2593     int invalid = 0;
2594     for (i = 31; i > 0; i--) {
2595         uint8_t dig_a = bcd_get_digit(a, i, &invalid);
2596         uint8_t dig_b = bcd_get_digit(b, i, &invalid);
2597         if (unlikely(invalid)) {
2598             return 0; /* doesn't matter */
2599         } else if (dig_a > dig_b) {
2600             return 1;
2601         } else if (dig_a < dig_b) {
2602             return -1;
2603         }
2604     }
2605 
2606     return 0;
2607 }
2608 
2609 static int bcd_add_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid,
2610                        int *overflow)
2611 {
2612     int carry = 0;
2613     int i;
2614     int is_zero = 1;
2615     for (i = 1; i <= 31; i++) {
2616         uint8_t digit = bcd_get_digit(a, i, invalid) +
2617                         bcd_get_digit(b, i, invalid) + carry;
2618         is_zero &= (digit == 0);
2619         if (digit > 9) {
2620             carry = 1;
2621             digit -= 10;
2622         } else {
2623             carry = 0;
2624         }
2625 
2626         bcd_put_digit(t, digit, i);
2627 
2628         if (unlikely(*invalid)) {
2629             return -1;
2630         }
2631     }
2632 
2633     *overflow = carry;
2634     return is_zero;
2635 }
2636 
2637 static int bcd_sub_mag(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, int *invalid,
2638                        int *overflow)
2639 {
2640     int carry = 0;
2641     int i;
2642     int is_zero = 1;
2643     for (i = 1; i <= 31; i++) {
2644         uint8_t digit = bcd_get_digit(a, i, invalid) -
2645                         bcd_get_digit(b, i, invalid) + carry;
2646         is_zero &= (digit == 0);
2647         if (digit & 0x80) {
2648             carry = -1;
2649             digit += 10;
2650         } else {
2651             carry = 0;
2652         }
2653 
2654         bcd_put_digit(t, digit, i);
2655 
2656         if (unlikely(*invalid)) {
2657             return -1;
2658         }
2659     }
2660 
2661     *overflow = carry;
2662     return is_zero;
2663 }
2664 
2665 uint32_t helper_bcdadd(ppc_avr_t *r,  ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2666 {
2667 
2668     int sgna = bcd_get_sgn(a);
2669     int sgnb = bcd_get_sgn(b);
2670     int invalid = (sgna == 0) || (sgnb == 0);
2671     int overflow = 0;
2672     int zero = 0;
2673     uint32_t cr = 0;
2674     ppc_avr_t result = { .u64 = { 0, 0 } };
2675 
2676     if (!invalid) {
2677         if (sgna == sgnb) {
2678             result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps);
2679             zero = bcd_add_mag(&result, a, b, &invalid, &overflow);
2680             cr = (sgna > 0) ? 1 << CRF_GT : 1 << CRF_LT;
2681         } else if (bcd_cmp_mag(a, b) > 0) {
2682             result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgna, ps);
2683             zero = bcd_sub_mag(&result, a, b, &invalid, &overflow);
2684             cr = (sgna > 0) ? 1 << CRF_GT : 1 << CRF_LT;
2685         } else {
2686             result.u8[BCD_DIG_BYTE(0)] = bcd_preferred_sgn(sgnb, ps);
2687             zero = bcd_sub_mag(&result, b, a, &invalid, &overflow);
2688             cr = (sgnb > 0) ? 1 << CRF_GT : 1 << CRF_LT;
2689         }
2690     }
2691 
2692     if (unlikely(invalid)) {
2693         result.u64[HI_IDX] = result.u64[LO_IDX] = -1;
2694         cr = 1 << CRF_SO;
2695     } else if (overflow) {
2696         cr |= 1 << CRF_SO;
2697     } else if (zero) {
2698         cr = 1 << CRF_EQ;
2699     }
2700 
2701     *r = result;
2702 
2703     return cr;
2704 }
2705 
2706 uint32_t helper_bcdsub(ppc_avr_t *r,  ppc_avr_t *a, ppc_avr_t *b, uint32_t ps)
2707 {
2708     ppc_avr_t bcopy = *b;
2709     int sgnb = bcd_get_sgn(b);
2710     if (sgnb < 0) {
2711         bcd_put_digit(&bcopy, BCD_PLUS_PREF_1, 0);
2712     } else if (sgnb > 0) {
2713         bcd_put_digit(&bcopy, BCD_NEG_PREF, 0);
2714     }
2715     /* else invalid ... defer to bcdadd code for proper handling */
2716 
2717     return helper_bcdadd(r, a, &bcopy, ps);
2718 }
2719 
2720 uint32_t helper_bcdcfn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2721 {
2722     int i;
2723     int cr = 0;
2724     uint16_t national = 0;
2725     uint16_t sgnb = get_national_digit(b, 0);
2726     ppc_avr_t ret = { .u64 = { 0, 0 } };
2727     int invalid = (sgnb != NATIONAL_PLUS && sgnb != NATIONAL_NEG);
2728 
2729     for (i = 1; i < 8; i++) {
2730         national = get_national_digit(b, i);
2731         if (unlikely(national < 0x30 || national > 0x39)) {
2732             invalid = 1;
2733             break;
2734         }
2735 
2736         bcd_put_digit(&ret, national & 0xf, i);
2737     }
2738 
2739     if (sgnb == NATIONAL_PLUS) {
2740         bcd_put_digit(&ret, (ps == 0) ? BCD_PLUS_PREF_1 : BCD_PLUS_PREF_2, 0);
2741     } else {
2742         bcd_put_digit(&ret, BCD_NEG_PREF, 0);
2743     }
2744 
2745     cr = bcd_cmp_zero(&ret);
2746 
2747     if (unlikely(invalid)) {
2748         cr = 1 << CRF_SO;
2749     }
2750 
2751     *r = ret;
2752 
2753     return cr;
2754 }
2755 
2756 uint32_t helper_bcdctn(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2757 {
2758     int i;
2759     int cr = 0;
2760     int sgnb = bcd_get_sgn(b);
2761     int invalid = (sgnb == 0);
2762     ppc_avr_t ret = { .u64 = { 0, 0 } };
2763 
2764     int ox_flag = (b->u64[HI_IDX] != 0) || ((b->u64[LO_IDX] >> 32) != 0);
2765 
2766     for (i = 1; i < 8; i++) {
2767         set_national_digit(&ret, 0x30 + bcd_get_digit(b, i, &invalid), i);
2768 
2769         if (unlikely(invalid)) {
2770             break;
2771         }
2772     }
2773     set_national_digit(&ret, (sgnb == -1) ? NATIONAL_NEG : NATIONAL_PLUS, 0);
2774 
2775     cr = bcd_cmp_zero(b);
2776 
2777     if (ox_flag) {
2778         cr |= 1 << CRF_SO;
2779     }
2780 
2781     if (unlikely(invalid)) {
2782         cr = 1 << CRF_SO;
2783     }
2784 
2785     *r = ret;
2786 
2787     return cr;
2788 }
2789 
2790 uint32_t helper_bcdcfz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2791 {
2792     int i;
2793     int cr = 0;
2794     int invalid = 0;
2795     int zone_digit = 0;
2796     int zone_lead = ps ? 0xF : 0x3;
2797     int digit = 0;
2798     ppc_avr_t ret = { .u64 = { 0, 0 } };
2799     int sgnb = b->u8[BCD_DIG_BYTE(0)] >> 4;
2800 
2801     if (unlikely((sgnb < 0xA) && ps)) {
2802         invalid = 1;
2803     }
2804 
2805     for (i = 0; i < 16; i++) {
2806         zone_digit = (i * 2) ? b->u8[BCD_DIG_BYTE(i * 2)] >> 4 : zone_lead;
2807         digit = b->u8[BCD_DIG_BYTE(i * 2)] & 0xF;
2808         if (unlikely(zone_digit != zone_lead || digit > 0x9)) {
2809             invalid = 1;
2810             break;
2811         }
2812 
2813         bcd_put_digit(&ret, digit, i + 1);
2814     }
2815 
2816     if ((ps && (sgnb == 0xB || sgnb == 0xD)) ||
2817             (!ps && (sgnb & 0x4))) {
2818         bcd_put_digit(&ret, BCD_NEG_PREF, 0);
2819     } else {
2820         bcd_put_digit(&ret, BCD_PLUS_PREF_1, 0);
2821     }
2822 
2823     cr = bcd_cmp_zero(&ret);
2824 
2825     if (unlikely(invalid)) {
2826         cr = 1 << CRF_SO;
2827     }
2828 
2829     *r = ret;
2830 
2831     return cr;
2832 }
2833 
2834 uint32_t helper_bcdctz(ppc_avr_t *r, ppc_avr_t *b, uint32_t ps)
2835 {
2836     int i;
2837     int cr = 0;
2838     uint8_t digit = 0;
2839     int sgnb = bcd_get_sgn(b);
2840     int zone_lead = (ps) ? 0xF0 : 0x30;
2841     int invalid = (sgnb == 0);
2842     ppc_avr_t ret = { .u64 = { 0, 0 } };
2843 
2844     int ox_flag = ((b->u64[HI_IDX] >> 4) != 0);
2845 
2846     for (i = 0; i < 16; i++) {
2847         digit = bcd_get_digit(b, i + 1, &invalid);
2848 
2849         if (unlikely(invalid)) {
2850             break;
2851         }
2852 
2853         ret.u8[BCD_DIG_BYTE(i * 2)] = zone_lead + digit;
2854     }
2855 
2856     if (ps) {
2857         bcd_put_digit(&ret, (sgnb == 1) ? 0xC : 0xD, 1);
2858     } else {
2859         bcd_put_digit(&ret, (sgnb == 1) ? 0x3 : 0x7, 1);
2860     }
2861 
2862     cr = bcd_cmp_zero(b);
2863 
2864     if (ox_flag) {
2865         cr |= 1 << CRF_SO;
2866     }
2867 
2868     if (unlikely(invalid)) {
2869         cr = 1 << CRF_SO;
2870     }
2871 
2872     *r = ret;
2873 
2874     return cr;
2875 }
2876 
2877 void helper_vsbox(ppc_avr_t *r, ppc_avr_t *a)
2878 {
2879     int i;
2880     VECTOR_FOR_INORDER_I(i, u8) {
2881         r->u8[i] = AES_sbox[a->u8[i]];
2882     }
2883 }
2884 
2885 void helper_vcipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2886 {
2887     ppc_avr_t result;
2888     int i;
2889 
2890     VECTOR_FOR_INORDER_I(i, u32) {
2891         result.AVRW(i) = b->AVRW(i) ^
2892             (AES_Te0[a->AVRB(AES_shifts[4*i + 0])] ^
2893              AES_Te1[a->AVRB(AES_shifts[4*i + 1])] ^
2894              AES_Te2[a->AVRB(AES_shifts[4*i + 2])] ^
2895              AES_Te3[a->AVRB(AES_shifts[4*i + 3])]);
2896     }
2897     *r = result;
2898 }
2899 
2900 void helper_vcipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2901 {
2902     ppc_avr_t result;
2903     int i;
2904 
2905     VECTOR_FOR_INORDER_I(i, u8) {
2906         result.AVRB(i) = b->AVRB(i) ^ (AES_sbox[a->AVRB(AES_shifts[i])]);
2907     }
2908     *r = result;
2909 }
2910 
2911 void helper_vncipher(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2912 {
2913     /* This differs from what is written in ISA V2.07.  The RTL is */
2914     /* incorrect and will be fixed in V2.07B.                      */
2915     int i;
2916     ppc_avr_t tmp;
2917 
2918     VECTOR_FOR_INORDER_I(i, u8) {
2919         tmp.AVRB(i) = b->AVRB(i) ^ AES_isbox[a->AVRB(AES_ishifts[i])];
2920     }
2921 
2922     VECTOR_FOR_INORDER_I(i, u32) {
2923         r->AVRW(i) =
2924             AES_imc[tmp.AVRB(4*i + 0)][0] ^
2925             AES_imc[tmp.AVRB(4*i + 1)][1] ^
2926             AES_imc[tmp.AVRB(4*i + 2)][2] ^
2927             AES_imc[tmp.AVRB(4*i + 3)][3];
2928     }
2929 }
2930 
2931 void helper_vncipherlast(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)
2932 {
2933     ppc_avr_t result;
2934     int i;
2935 
2936     VECTOR_FOR_INORDER_I(i, u8) {
2937         result.AVRB(i) = b->AVRB(i) ^ (AES_isbox[a->AVRB(AES_ishifts[i])]);
2938     }
2939     *r = result;
2940 }
2941 
2942 #define ROTRu32(v, n) (((v) >> (n)) | ((v) << (32-n)))
2943 #if defined(HOST_WORDS_BIGENDIAN)
2944 #define EL_IDX(i) (i)
2945 #else
2946 #define EL_IDX(i) (3 - (i))
2947 #endif
2948 
2949 void helper_vshasigmaw(ppc_avr_t *r,  ppc_avr_t *a, uint32_t st_six)
2950 {
2951     int st = (st_six & 0x10) != 0;
2952     int six = st_six & 0xF;
2953     int i;
2954 
2955     VECTOR_FOR_INORDER_I(i, u32) {
2956         if (st == 0) {
2957             if ((six & (0x8 >> i)) == 0) {
2958                 r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 7) ^
2959                                     ROTRu32(a->u32[EL_IDX(i)], 18) ^
2960                                     (a->u32[EL_IDX(i)] >> 3);
2961             } else { /* six.bit[i] == 1 */
2962                 r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 17) ^
2963                                     ROTRu32(a->u32[EL_IDX(i)], 19) ^
2964                                     (a->u32[EL_IDX(i)] >> 10);
2965             }
2966         } else { /* st == 1 */
2967             if ((six & (0x8 >> i)) == 0) {
2968                 r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 2) ^
2969                                     ROTRu32(a->u32[EL_IDX(i)], 13) ^
2970                                     ROTRu32(a->u32[EL_IDX(i)], 22);
2971             } else { /* six.bit[i] == 1 */
2972                 r->u32[EL_IDX(i)] = ROTRu32(a->u32[EL_IDX(i)], 6) ^
2973                                     ROTRu32(a->u32[EL_IDX(i)], 11) ^
2974                                     ROTRu32(a->u32[EL_IDX(i)], 25);
2975             }
2976         }
2977     }
2978 }
2979 
2980 #undef ROTRu32
2981 #undef EL_IDX
2982 
2983 #define ROTRu64(v, n) (((v) >> (n)) | ((v) << (64-n)))
2984 #if defined(HOST_WORDS_BIGENDIAN)
2985 #define EL_IDX(i) (i)
2986 #else
2987 #define EL_IDX(i) (1 - (i))
2988 #endif
2989 
2990 void helper_vshasigmad(ppc_avr_t *r,  ppc_avr_t *a, uint32_t st_six)
2991 {
2992     int st = (st_six & 0x10) != 0;
2993     int six = st_six & 0xF;
2994     int i;
2995 
2996     VECTOR_FOR_INORDER_I(i, u64) {
2997         if (st == 0) {
2998             if ((six & (0x8 >> (2*i))) == 0) {
2999                 r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 1) ^
3000                                     ROTRu64(a->u64[EL_IDX(i)], 8) ^
3001                                     (a->u64[EL_IDX(i)] >> 7);
3002             } else { /* six.bit[2*i] == 1 */
3003                 r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 19) ^
3004                                     ROTRu64(a->u64[EL_IDX(i)], 61) ^
3005                                     (a->u64[EL_IDX(i)] >> 6);
3006             }
3007         } else { /* st == 1 */
3008             if ((six & (0x8 >> (2*i))) == 0) {
3009                 r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 28) ^
3010                                     ROTRu64(a->u64[EL_IDX(i)], 34) ^
3011                                     ROTRu64(a->u64[EL_IDX(i)], 39);
3012             } else { /* six.bit[2*i] == 1 */
3013                 r->u64[EL_IDX(i)] = ROTRu64(a->u64[EL_IDX(i)], 14) ^
3014                                     ROTRu64(a->u64[EL_IDX(i)], 18) ^
3015                                     ROTRu64(a->u64[EL_IDX(i)], 41);
3016             }
3017         }
3018     }
3019 }
3020 
3021 #undef ROTRu64
3022 #undef EL_IDX
3023 
3024 void helper_vpermxor(ppc_avr_t *r,  ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
3025 {
3026     ppc_avr_t result;
3027     int i;
3028 
3029     VECTOR_FOR_INORDER_I(i, u8) {
3030         int indexA = c->u8[i] >> 4;
3031         int indexB = c->u8[i] & 0xF;
3032 #if defined(HOST_WORDS_BIGENDIAN)
3033         result.u8[i] = a->u8[indexA] ^ b->u8[indexB];
3034 #else
3035         result.u8[i] = a->u8[15-indexA] ^ b->u8[15-indexB];
3036 #endif
3037     }
3038     *r = result;
3039 }
3040 
3041 #undef VECTOR_FOR_INORDER_I
3042 #undef HI_IDX
3043 #undef LO_IDX
3044 
3045 /*****************************************************************************/
3046 /* SPE extension helpers */
3047 /* Use a table to make this quicker */
3048 static const uint8_t hbrev[16] = {
3049     0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE,
3050     0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF,
3051 };
3052 
3053 static inline uint8_t byte_reverse(uint8_t val)
3054 {
3055     return hbrev[val >> 4] | (hbrev[val & 0xF] << 4);
3056 }
3057 
3058 static inline uint32_t word_reverse(uint32_t val)
3059 {
3060     return byte_reverse(val >> 24) | (byte_reverse(val >> 16) << 8) |
3061         (byte_reverse(val >> 8) << 16) | (byte_reverse(val) << 24);
3062 }
3063 
3064 #define MASKBITS 16 /* Random value - to be fixed (implementation dependent) */
3065 target_ulong helper_brinc(target_ulong arg1, target_ulong arg2)
3066 {
3067     uint32_t a, b, d, mask;
3068 
3069     mask = UINT32_MAX >> (32 - MASKBITS);
3070     a = arg1 & mask;
3071     b = arg2 & mask;
3072     d = word_reverse(1 + word_reverse(a | ~b));
3073     return (arg1 & ~mask) | (d & b);
3074 }
3075 
3076 uint32_t helper_cntlsw32(uint32_t val)
3077 {
3078     if (val & 0x80000000) {
3079         return clz32(~val);
3080     } else {
3081         return clz32(val);
3082     }
3083 }
3084 
3085 uint32_t helper_cntlzw32(uint32_t val)
3086 {
3087     return clz32(val);
3088 }
3089 
3090 /* 440 specific */
3091 target_ulong helper_dlmzb(CPUPPCState *env, target_ulong high,
3092                           target_ulong low, uint32_t update_Rc)
3093 {
3094     target_ulong mask;
3095     int i;
3096 
3097     i = 1;
3098     for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
3099         if ((high & mask) == 0) {
3100             if (update_Rc) {
3101                 env->crf[0] = 0x4;
3102             }
3103             goto done;
3104         }
3105         i++;
3106     }
3107     for (mask = 0xFF000000; mask != 0; mask = mask >> 8) {
3108         if ((low & mask) == 0) {
3109             if (update_Rc) {
3110                 env->crf[0] = 0x8;
3111             }
3112             goto done;
3113         }
3114         i++;
3115     }
3116     i = 8;
3117     if (update_Rc) {
3118         env->crf[0] = 0x2;
3119     }
3120  done:
3121     env->xer = (env->xer & ~0x7F) | i;
3122     if (update_Rc) {
3123         env->crf[0] |= xer_so;
3124     }
3125     return i;
3126 }
3127