2 * PowerPC integer and vector emulation helpers for QEMU.
4 * Copyright (c) 2003-2007 Jocelyn Mayer
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
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.
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/>.
20 #include "qemu/osdep.h"
23 #include "qemu/host-utils.h"
24 #include "qemu/main-loop.h"
26 #include "exec/helper-proto.h"
27 #include "crypto/aes.h"
28 #include "fpu/softfloat.h"
29 #include "qapi/error.h"
30 #include "qemu/guest-random.h"
31 #include "tcg/tcg-gvec-desc.h"
33 #include "helper_regs.h"
34 /*****************************************************************************/
35 /* Fixed point operations helpers */
37 static inline void helper_update_ov_legacy(CPUPPCState
*env
, int ov
)
40 env
->so
= env
->ov
= 1;
46 target_ulong
helper_divweu(CPUPPCState
*env
, target_ulong ra
, target_ulong rb
,
52 uint64_t dividend
= (uint64_t)ra
<< 32;
53 uint64_t divisor
= (uint32_t)rb
;
55 if (unlikely(divisor
== 0)) {
58 rt
= dividend
/ divisor
;
59 overflow
= rt
> UINT32_MAX
;
62 if (unlikely(overflow
)) {
63 rt
= 0; /* Undefined */
67 helper_update_ov_legacy(env
, overflow
);
70 return (target_ulong
)rt
;
73 target_ulong
helper_divwe(CPUPPCState
*env
, target_ulong ra
, target_ulong rb
,
79 int64_t dividend
= (int64_t)ra
<< 32;
80 int64_t divisor
= (int64_t)((int32_t)rb
);
82 if (unlikely((divisor
== 0) ||
83 ((divisor
== -1ull) && (dividend
== INT64_MIN
)))) {
86 rt
= dividend
/ divisor
;
87 overflow
= rt
!= (int32_t)rt
;
90 if (unlikely(overflow
)) {
91 rt
= 0; /* Undefined */
95 helper_update_ov_legacy(env
, overflow
);
98 return (target_ulong
)rt
;
101 #if defined(TARGET_PPC64)
103 uint64_t helper_divdeu(CPUPPCState
*env
, uint64_t ra
, uint64_t rb
, uint32_t oe
)
108 if (unlikely(rb
== 0 || ra
>= rb
)) {
110 rt
= 0; /* Undefined */
112 divu128(&rt
, &ra
, rb
);
116 helper_update_ov_legacy(env
, overflow
);
122 uint64_t helper_divde(CPUPPCState
*env
, uint64_t rau
, uint64_t rbu
, uint32_t oe
)
125 int64_t ra
= (int64_t)rau
;
126 int64_t rb
= (int64_t)rbu
;
129 if (unlikely(rb
== 0 || uabs64(ra
) >= uabs64(rb
))) {
131 rt
= 0; /* Undefined */
133 divs128(&rt
, &ra
, rb
);
137 helper_update_ov_legacy(env
, overflow
);
146 #if defined(TARGET_PPC64)
147 /* if x = 0xab, returns 0xababababababababa */
148 #define pattern(x) (((x) & 0xff) * (~(target_ulong)0 / 0xff))
151 * subtract 1 from each byte, and with inverse, check if MSB is set at each
153 * i.e. ((0x00 - 0x01) & ~(0x00)) & 0x80
154 * (0xFF & 0xFF) & 0x80 = 0x80 (zero found)
156 #define haszero(v) (((v) - pattern(0x01)) & ~(v) & pattern(0x80))
158 /* When you XOR the pattern and there is a match, that byte will be zero */
159 #define hasvalue(x, n) (haszero((x) ^ pattern(n)))
161 uint32_t helper_cmpeqb(target_ulong ra
, target_ulong rb
)
163 return hasvalue(rb
, ra
) ? CRF_GT
: 0;
171 * Return a random number.
173 uint64_t helper_darn32(void)
178 if (qemu_guest_getrandom(&ret
, sizeof(ret
), &err
) < 0) {
179 qemu_log_mask(LOG_UNIMP
, "darn: Crypto failure: %s",
180 error_get_pretty(err
));
188 uint64_t helper_darn64(void)
193 if (qemu_guest_getrandom(&ret
, sizeof(ret
), &err
) < 0) {
194 qemu_log_mask(LOG_UNIMP
, "darn: Crypto failure: %s",
195 error_get_pretty(err
));
203 uint64_t helper_bpermd(uint64_t rs
, uint64_t rb
)
208 for (i
= 0; i
< 8; i
++) {
209 int index
= (rs
>> (i
* 8)) & 0xFF;
211 if (rb
& PPC_BIT(index
)) {
221 target_ulong
helper_cmpb(target_ulong rs
, target_ulong rb
)
223 target_ulong mask
= 0xff;
227 for (i
= 0; i
< sizeof(target_ulong
); i
++) {
228 if ((rs
& mask
) == (rb
& mask
)) {
236 /* shift right arithmetic helper */
237 target_ulong
helper_sraw(CPUPPCState
*env
, target_ulong value
,
242 if (likely(!(shift
& 0x20))) {
243 if (likely((uint32_t)shift
!= 0)) {
245 ret
= (int32_t)value
>> shift
;
246 if (likely(ret
>= 0 || (value
& ((1 << shift
) - 1)) == 0)) {
247 env
->ca32
= env
->ca
= 0;
249 env
->ca32
= env
->ca
= 1;
252 ret
= (int32_t)value
;
253 env
->ca32
= env
->ca
= 0;
256 ret
= (int32_t)value
>> 31;
257 env
->ca32
= env
->ca
= (ret
!= 0);
259 return (target_long
)ret
;
262 #if defined(TARGET_PPC64)
263 target_ulong
helper_srad(CPUPPCState
*env
, target_ulong value
,
268 if (likely(!(shift
& 0x40))) {
269 if (likely((uint64_t)shift
!= 0)) {
271 ret
= (int64_t)value
>> shift
;
272 if (likely(ret
>= 0 || (value
& ((1ULL << shift
) - 1)) == 0)) {
273 env
->ca32
= env
->ca
= 0;
275 env
->ca32
= env
->ca
= 1;
278 ret
= (int64_t)value
;
279 env
->ca32
= env
->ca
= 0;
282 ret
= (int64_t)value
>> 63;
283 env
->ca32
= env
->ca
= (ret
!= 0);
289 #if defined(TARGET_PPC64)
290 target_ulong
helper_popcntb(target_ulong val
)
292 /* Note that we don't fold past bytes */
293 val
= (val
& 0x5555555555555555ULL
) + ((val
>> 1) &
294 0x5555555555555555ULL
);
295 val
= (val
& 0x3333333333333333ULL
) + ((val
>> 2) &
296 0x3333333333333333ULL
);
297 val
= (val
& 0x0f0f0f0f0f0f0f0fULL
) + ((val
>> 4) &
298 0x0f0f0f0f0f0f0f0fULL
);
302 target_ulong
helper_popcntw(target_ulong val
)
304 /* Note that we don't fold past words. */
305 val
= (val
& 0x5555555555555555ULL
) + ((val
>> 1) &
306 0x5555555555555555ULL
);
307 val
= (val
& 0x3333333333333333ULL
) + ((val
>> 2) &
308 0x3333333333333333ULL
);
309 val
= (val
& 0x0f0f0f0f0f0f0f0fULL
) + ((val
>> 4) &
310 0x0f0f0f0f0f0f0f0fULL
);
311 val
= (val
& 0x00ff00ff00ff00ffULL
) + ((val
>> 8) &
312 0x00ff00ff00ff00ffULL
);
313 val
= (val
& 0x0000ffff0000ffffULL
) + ((val
>> 16) &
314 0x0000ffff0000ffffULL
);
318 target_ulong
helper_popcntb(target_ulong val
)
320 /* Note that we don't fold past bytes */
321 val
= (val
& 0x55555555) + ((val
>> 1) & 0x55555555);
322 val
= (val
& 0x33333333) + ((val
>> 2) & 0x33333333);
323 val
= (val
& 0x0f0f0f0f) + ((val
>> 4) & 0x0f0f0f0f);
328 uint64_t helper_CFUGED(uint64_t src
, uint64_t mask
)
331 * Instead of processing the mask bit-by-bit from the most significant to
332 * the least significant bit, as described in PowerISA, we'll handle it in
333 * blocks of 'n' zeros/ones from LSB to MSB. To avoid the decision to use
334 * ctz or cto, we negate the mask at the end of the loop.
336 target_ulong m
, left
= 0, right
= 0;
337 unsigned int n
, i
= 64;
338 bool bit
= false; /* tracks if we are processing zeros or ones */
340 if (mask
== 0 || mask
== -1) {
344 /* Processes the mask in blocks, from LSB to MSB */
346 /* Find how many bits we should take */
353 * Extracts 'n' trailing bits of src and put them on the leading 'n'
354 * bits of 'right' or 'left', pushing down the previously extracted
359 right
= ror64(right
| (src
& m
), n
);
361 left
= ror64(left
| (src
& m
), n
);
365 * Discards the processed bits from 'src' and 'mask'. Note that we are
366 * removing 'n' trailing zeros from 'mask', but the logical shift will
367 * add 'n' leading zeros back, so the population count of 'mask' is kept
378 * At the end, right was ror'ed ctpop(mask) times. To put it back in place,
379 * we'll shift it more 64-ctpop(mask) times.
384 n
= 64 - ctpop64(mask
);
387 return left
| (right
>> n
);
390 uint64_t helper_PDEPD(uint64_t src
, uint64_t mask
)
399 for (i
= 0; mask
!= 0; i
++) {
402 result
|= ((src
>> i
) & 1) << o
;
408 uint64_t helper_PEXTD(uint64_t src
, uint64_t mask
)
417 for (o
= 0; mask
!= 0; o
++) {
420 result
|= ((src
>> i
) & 1) << o
;
426 /*****************************************************************************/
427 /* Altivec extension helpers */
429 #define VECTOR_FOR_INORDER_I(index, element) \
430 for (index = 0; index < ARRAY_SIZE(r->element); index++)
432 #define VECTOR_FOR_INORDER_I(index, element) \
433 for (index = ARRAY_SIZE(r->element) - 1; index >= 0; index--)
436 /* Saturating arithmetic helpers. */
437 #define SATCVT(from, to, from_type, to_type, min, max) \
438 static inline to_type cvt##from##to(from_type x, int *sat) \
442 if (x < (from_type)min) { \
445 } else if (x > (from_type)max) { \
453 #define SATCVTU(from, to, from_type, to_type, min, max) \
454 static inline to_type cvt##from##to(from_type x, int *sat) \
458 if (x > (from_type)max) { \
466 SATCVT(sh
, sb
, int16_t, int8_t, INT8_MIN
, INT8_MAX
)
467 SATCVT(sw
, sh
, int32_t, int16_t, INT16_MIN
, INT16_MAX
)
468 SATCVT(sd
, sw
, int64_t, int32_t, INT32_MIN
, INT32_MAX
)
470 SATCVTU(uh
, ub
, uint16_t, uint8_t, 0, UINT8_MAX
)
471 SATCVTU(uw
, uh
, uint32_t, uint16_t, 0, UINT16_MAX
)
472 SATCVTU(ud
, uw
, uint64_t, uint32_t, 0, UINT32_MAX
)
473 SATCVT(sh
, ub
, int16_t, uint8_t, 0, UINT8_MAX
)
474 SATCVT(sw
, uh
, int32_t, uint16_t, 0, UINT16_MAX
)
475 SATCVT(sd
, uw
, int64_t, uint32_t, 0, UINT32_MAX
)
479 void helper_mtvscr(CPUPPCState
*env
, uint32_t vscr
)
481 ppc_store_vscr(env
, vscr
);
484 uint32_t helper_mfvscr(CPUPPCState
*env
)
486 return ppc_get_vscr(env
);
489 static inline void set_vscr_sat(CPUPPCState
*env
)
491 /* The choice of non-zero value is arbitrary. */
492 env
->vscr_sat
.u32
[0] = 1;
495 void helper_vaddcuw(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
499 for (i
= 0; i
< ARRAY_SIZE(r
->u32
); i
++) {
500 r
->u32
[i
] = ~a
->u32
[i
] < b
->u32
[i
];
505 void helper_vprtybw(ppc_avr_t
*r
, ppc_avr_t
*b
)
508 for (i
= 0; i
< ARRAY_SIZE(r
->u32
); i
++) {
509 uint64_t res
= b
->u32
[i
] ^ (b
->u32
[i
] >> 16);
516 void helper_vprtybd(ppc_avr_t
*r
, ppc_avr_t
*b
)
519 for (i
= 0; i
< ARRAY_SIZE(r
->u64
); i
++) {
520 uint64_t res
= b
->u64
[i
] ^ (b
->u64
[i
] >> 32);
528 void helper_vprtybq(ppc_avr_t
*r
, ppc_avr_t
*b
)
530 uint64_t res
= b
->u64
[0] ^ b
->u64
[1];
534 r
->VsrD(1) = res
& 1;
538 #define VARITHFP(suffix, func) \
539 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
544 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
545 r->f32[i] = func(a->f32[i], b->f32[i], &env->vec_status); \
548 VARITHFP(addfp
, float32_add
)
549 VARITHFP(subfp
, float32_sub
)
550 VARITHFP(minfp
, float32_min
)
551 VARITHFP(maxfp
, float32_max
)
554 #define VARITHFPFMA(suffix, type) \
555 void helper_v##suffix(CPUPPCState *env, ppc_avr_t *r, ppc_avr_t *a, \
556 ppc_avr_t *b, ppc_avr_t *c) \
559 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
560 r->f32[i] = float32_muladd(a->f32[i], c->f32[i], b->f32[i], \
561 type, &env->vec_status); \
564 VARITHFPFMA(maddfp
, 0);
565 VARITHFPFMA(nmsubfp
, float_muladd_negate_result
| float_muladd_negate_c
);
568 #define VARITHSAT_CASE(type, op, cvt, element) \
570 type result = (type)a->element[i] op (type)b->element[i]; \
571 r->element[i] = cvt(result, &sat); \
574 #define VARITHSAT_DO(name, op, optype, cvt, element) \
575 void helper_v##name(ppc_avr_t *r, ppc_avr_t *vscr_sat, \
576 ppc_avr_t *a, ppc_avr_t *b, uint32_t desc) \
581 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
582 VARITHSAT_CASE(optype, op, cvt, element); \
585 vscr_sat->u32[0] = 1; \
588 #define VARITHSAT_SIGNED(suffix, element, optype, cvt) \
589 VARITHSAT_DO(adds##suffix##s, +, optype, cvt, element) \
590 VARITHSAT_DO(subs##suffix##s, -, optype, cvt, element)
591 #define VARITHSAT_UNSIGNED(suffix, element, optype, cvt) \
592 VARITHSAT_DO(addu##suffix##s, +, optype, cvt, element) \
593 VARITHSAT_DO(subu##suffix##s, -, optype, cvt, element)
594 VARITHSAT_SIGNED(b
, s8
, int16_t, cvtshsb
)
595 VARITHSAT_SIGNED(h
, s16
, int32_t, cvtswsh
)
596 VARITHSAT_SIGNED(w
, s32
, int64_t, cvtsdsw
)
597 VARITHSAT_UNSIGNED(b
, u8
, uint16_t, cvtshub
)
598 VARITHSAT_UNSIGNED(h
, u16
, uint32_t, cvtswuh
)
599 VARITHSAT_UNSIGNED(w
, u32
, uint64_t, cvtsduw
)
600 #undef VARITHSAT_CASE
602 #undef VARITHSAT_SIGNED
603 #undef VARITHSAT_UNSIGNED
605 #define VAVG_DO(name, element, etype) \
606 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
610 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
611 etype x = (etype)a->element[i] + (etype)b->element[i] + 1; \
612 r->element[i] = x >> 1; \
616 #define VAVG(type, signed_element, signed_type, unsigned_element, \
618 VAVG_DO(avgs##type, signed_element, signed_type) \
619 VAVG_DO(avgu##type, unsigned_element, unsigned_type)
620 VAVG(b
, s8
, int16_t, u8
, uint16_t)
621 VAVG(h
, s16
, int32_t, u16
, uint32_t)
622 VAVG(w
, s32
, int64_t, u32
, uint64_t)
626 #define VABSDU_DO(name, element) \
627 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
631 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
632 r->element[i] = (a->element[i] > b->element[i]) ? \
633 (a->element[i] - b->element[i]) : \
634 (b->element[i] - a->element[i]); \
639 * VABSDU - Vector absolute difference unsigned
640 * name - instruction mnemonic suffix (b: byte, h: halfword, w: word)
641 * element - element type to access from vector
643 #define VABSDU(type, element) \
644 VABSDU_DO(absdu##type, element)
651 #define VCF(suffix, cvt, element) \
652 void helper_vcf##suffix(CPUPPCState *env, ppc_avr_t *r, \
653 ppc_avr_t *b, uint32_t uim) \
657 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
658 float32 t = cvt(b->element[i], &env->vec_status); \
659 r->f32[i] = float32_scalbn(t, -uim, &env->vec_status); \
662 VCF(ux
, uint32_to_float32
, u32
)
663 VCF(sx
, int32_to_float32
, s32
)
666 #define VCMPNEZ(NAME, ELEM) \
667 void helper_##NAME(ppc_vsr_t *t, ppc_vsr_t *a, ppc_vsr_t *b, uint32_t desc) \
669 for (int i = 0; i < ARRAY_SIZE(t->ELEM); i++) { \
670 t->ELEM[i] = ((a->ELEM[i] == 0) || (b->ELEM[i] == 0) || \
671 (a->ELEM[i] != b->ELEM[i])) ? -1 : 0; \
674 VCMPNEZ(VCMPNEZB
, u8
)
675 VCMPNEZ(VCMPNEZH
, u16
)
676 VCMPNEZ(VCMPNEZW
, u32
)
679 #define VCMPFP_DO(suffix, compare, order, record) \
680 void helper_vcmp##suffix(CPUPPCState *env, ppc_avr_t *r, \
681 ppc_avr_t *a, ppc_avr_t *b) \
683 uint32_t ones = (uint32_t)-1; \
684 uint32_t all = ones; \
688 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
690 FloatRelation rel = \
691 float32_compare_quiet(a->f32[i], b->f32[i], \
693 if (rel == float_relation_unordered) { \
695 } else if (rel compare order) { \
700 r->u32[i] = result; \
705 env->crf[6] = ((all != 0) << 3) | ((none == 0) << 1); \
708 #define VCMPFP(suffix, compare, order) \
709 VCMPFP_DO(suffix, compare, order, 0) \
710 VCMPFP_DO(suffix##_dot, compare, order, 1)
711 VCMPFP(eqfp
, ==, float_relation_equal
)
712 VCMPFP(gefp
, !=, float_relation_less
)
713 VCMPFP(gtfp
, ==, float_relation_greater
)
717 static inline void vcmpbfp_internal(CPUPPCState
*env
, ppc_avr_t
*r
,
718 ppc_avr_t
*a
, ppc_avr_t
*b
, int record
)
723 for (i
= 0; i
< ARRAY_SIZE(r
->f32
); i
++) {
724 FloatRelation le_rel
= float32_compare_quiet(a
->f32
[i
], b
->f32
[i
],
726 if (le_rel
== float_relation_unordered
) {
727 r
->u32
[i
] = 0xc0000000;
730 float32 bneg
= float32_chs(b
->f32
[i
]);
731 FloatRelation ge_rel
= float32_compare_quiet(a
->f32
[i
], bneg
,
733 int le
= le_rel
!= float_relation_greater
;
734 int ge
= ge_rel
!= float_relation_less
;
736 r
->u32
[i
] = ((!le
) << 31) | ((!ge
) << 30);
737 all_in
|= (!le
| !ge
);
741 env
->crf
[6] = (all_in
== 0) << 1;
745 void helper_vcmpbfp(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
747 vcmpbfp_internal(env
, r
, a
, b
, 0);
750 void helper_vcmpbfp_dot(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
,
753 vcmpbfp_internal(env
, r
, a
, b
, 1);
756 #define VCT(suffix, satcvt, element) \
757 void helper_vct##suffix(CPUPPCState *env, ppc_avr_t *r, \
758 ppc_avr_t *b, uint32_t uim) \
762 float_status s = env->vec_status; \
764 set_float_rounding_mode(float_round_to_zero, &s); \
765 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
766 if (float32_is_any_nan(b->f32[i])) { \
769 float64 t = float32_to_float64(b->f32[i], &s); \
772 t = float64_scalbn(t, uim, &s); \
773 j = float64_to_int64(t, &s); \
774 r->element[i] = satcvt(j, &sat); \
781 VCT(uxs
, cvtsduw
, u32
)
782 VCT(sxs
, cvtsdsw
, s32
)
785 typedef int64_t do_ger(uint32_t, uint32_t, uint32_t);
787 static int64_t ger_rank8(uint32_t a
, uint32_t b
, uint32_t mask
)
790 for (int i
= 0; i
< 8; i
++, mask
>>= 1) {
792 psum
+= (int64_t)sextract32(a
, 4 * i
, 4) * sextract32(b
, 4 * i
, 4);
798 static int64_t ger_rank4(uint32_t a
, uint32_t b
, uint32_t mask
)
801 for (int i
= 0; i
< 4; i
++, mask
>>= 1) {
803 psum
+= sextract32(a
, 8 * i
, 8) * (int64_t)extract32(b
, 8 * i
, 8);
809 static int64_t ger_rank2(uint32_t a
, uint32_t b
, uint32_t mask
)
812 for (int i
= 0; i
< 2; i
++, mask
>>= 1) {
814 psum
+= (int64_t)sextract32(a
, 16 * i
, 16) *
815 sextract32(b
, 16 * i
, 16);
821 static void xviger(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
, ppc_acc_t
*at
,
822 uint32_t mask
, bool sat
, bool acc
, do_ger ger
)
824 uint8_t pmsk
= FIELD_EX32(mask
, GER_MSK
, PMSK
),
825 xmsk
= FIELD_EX32(mask
, GER_MSK
, XMSK
),
826 ymsk
= FIELD_EX32(mask
, GER_MSK
, YMSK
);
827 uint8_t xmsk_bit
, ymsk_bit
;
830 for (i
= 0, xmsk_bit
= 1 << 3; i
< 4; i
++, xmsk_bit
>>= 1) {
831 for (j
= 0, ymsk_bit
= 1 << 3; j
< 4; j
++, ymsk_bit
>>= 1) {
832 if ((xmsk_bit
& xmsk
) && (ymsk_bit
& ymsk
)) {
833 psum
= ger(a
->VsrW(i
), b
->VsrW(j
), pmsk
);
835 psum
+= at
[i
].VsrSW(j
);
837 if (sat
&& psum
> INT32_MAX
) {
839 at
[i
].VsrSW(j
) = INT32_MAX
;
840 } else if (sat
&& psum
< INT32_MIN
) {
842 at
[i
].VsrSW(j
) = INT32_MIN
;
844 at
[i
].VsrSW(j
) = (int32_t) psum
;
854 void helper_XVI4GER8(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
855 ppc_acc_t
*at
, uint32_t mask
)
857 xviger(env
, a
, b
, at
, mask
, false, false, ger_rank8
);
861 void helper_XVI4GER8PP(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
862 ppc_acc_t
*at
, uint32_t mask
)
864 xviger(env
, a
, b
, at
, mask
, false, true, ger_rank8
);
868 void helper_XVI8GER4(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
869 ppc_acc_t
*at
, uint32_t mask
)
871 xviger(env
, a
, b
, at
, mask
, false, false, ger_rank4
);
875 void helper_XVI8GER4PP(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
876 ppc_acc_t
*at
, uint32_t mask
)
878 xviger(env
, a
, b
, at
, mask
, false, true, ger_rank4
);
882 void helper_XVI8GER4SPP(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
883 ppc_acc_t
*at
, uint32_t mask
)
885 xviger(env
, a
, b
, at
, mask
, true, true, ger_rank4
);
889 void helper_XVI16GER2(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
890 ppc_acc_t
*at
, uint32_t mask
)
892 xviger(env
, a
, b
, at
, mask
, false, false, ger_rank2
);
896 void helper_XVI16GER2S(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
897 ppc_acc_t
*at
, uint32_t mask
)
899 xviger(env
, a
, b
, at
, mask
, true, false, ger_rank2
);
903 void helper_XVI16GER2PP(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
904 ppc_acc_t
*at
, uint32_t mask
)
906 xviger(env
, a
, b
, at
, mask
, false, true, ger_rank2
);
910 void helper_XVI16GER2SPP(CPUPPCState
*env
, ppc_vsr_t
*a
, ppc_vsr_t
*b
,
911 ppc_acc_t
*at
, uint32_t mask
)
913 xviger(env
, a
, b
, at
, mask
, true, true, ger_rank2
);
916 target_ulong
helper_vclzlsbb(ppc_avr_t
*r
)
918 target_ulong count
= 0;
920 for (i
= 0; i
< ARRAY_SIZE(r
->u8
); i
++) {
921 if (r
->VsrB(i
) & 0x01) {
929 target_ulong
helper_vctzlsbb(ppc_avr_t
*r
)
931 target_ulong count
= 0;
933 for (i
= ARRAY_SIZE(r
->u8
) - 1; i
>= 0; i
--) {
934 if (r
->VsrB(i
) & 0x01) {
942 void helper_vmhaddshs(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
,
943 ppc_avr_t
*b
, ppc_avr_t
*c
)
948 for (i
= 0; i
< ARRAY_SIZE(r
->s16
); i
++) {
949 int32_t prod
= a
->s16
[i
] * b
->s16
[i
];
950 int32_t t
= (int32_t)c
->s16
[i
] + (prod
>> 15);
952 r
->s16
[i
] = cvtswsh(t
, &sat
);
960 void helper_vmhraddshs(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
,
961 ppc_avr_t
*b
, ppc_avr_t
*c
)
966 for (i
= 0; i
< ARRAY_SIZE(r
->s16
); i
++) {
967 int32_t prod
= a
->s16
[i
] * b
->s16
[i
] + 0x00004000;
968 int32_t t
= (int32_t)c
->s16
[i
] + (prod
>> 15);
969 r
->s16
[i
] = cvtswsh(t
, &sat
);
977 void helper_vmladduhm(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
981 for (i
= 0; i
< ARRAY_SIZE(r
->s16
); i
++) {
982 int32_t prod
= a
->s16
[i
] * b
->s16
[i
];
983 r
->s16
[i
] = (int16_t) (prod
+ c
->s16
[i
]);
987 #define VMRG_DO(name, element, access, ofs) \
988 void helper_v##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
991 int i, half = ARRAY_SIZE(r->element) / 2; \
993 for (i = 0; i < half; i++) { \
994 result.access(i * 2 + 0) = a->access(i + ofs); \
995 result.access(i * 2 + 1) = b->access(i + ofs); \
1000 #define VMRG(suffix, element, access) \
1001 VMRG_DO(mrgl##suffix, element, access, half) \
1002 VMRG_DO(mrgh##suffix, element, access, 0)
1009 void helper_VMSUMMBM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
1014 for (i
= 0; i
< ARRAY_SIZE(r
->s8
); i
++) {
1015 prod
[i
] = (int32_t)a
->s8
[i
] * b
->u8
[i
];
1018 VECTOR_FOR_INORDER_I(i
, s32
) {
1019 r
->s32
[i
] = c
->s32
[i
] + prod
[4 * i
] + prod
[4 * i
+ 1] +
1020 prod
[4 * i
+ 2] + prod
[4 * i
+ 3];
1024 void helper_VMSUMSHM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
1029 for (i
= 0; i
< ARRAY_SIZE(r
->s16
); i
++) {
1030 prod
[i
] = a
->s16
[i
] * b
->s16
[i
];
1033 VECTOR_FOR_INORDER_I(i
, s32
) {
1034 r
->s32
[i
] = c
->s32
[i
] + prod
[2 * i
] + prod
[2 * i
+ 1];
1038 void helper_VMSUMSHS(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
,
1039 ppc_avr_t
*b
, ppc_avr_t
*c
)
1045 for (i
= 0; i
< ARRAY_SIZE(r
->s16
); i
++) {
1046 prod
[i
] = (int32_t)a
->s16
[i
] * b
->s16
[i
];
1049 VECTOR_FOR_INORDER_I(i
, s32
) {
1050 int64_t t
= (int64_t)c
->s32
[i
] + prod
[2 * i
] + prod
[2 * i
+ 1];
1052 r
->u32
[i
] = cvtsdsw(t
, &sat
);
1060 void helper_VMSUMUBM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
1065 for (i
= 0; i
< ARRAY_SIZE(r
->u8
); i
++) {
1066 prod
[i
] = a
->u8
[i
] * b
->u8
[i
];
1069 VECTOR_FOR_INORDER_I(i
, u32
) {
1070 r
->u32
[i
] = c
->u32
[i
] + prod
[4 * i
] + prod
[4 * i
+ 1] +
1071 prod
[4 * i
+ 2] + prod
[4 * i
+ 3];
1075 void helper_VMSUMUHM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
1080 for (i
= 0; i
< ARRAY_SIZE(r
->u16
); i
++) {
1081 prod
[i
] = a
->u16
[i
] * b
->u16
[i
];
1084 VECTOR_FOR_INORDER_I(i
, u32
) {
1085 r
->u32
[i
] = c
->u32
[i
] + prod
[2 * i
] + prod
[2 * i
+ 1];
1089 void helper_VMSUMUHS(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
,
1090 ppc_avr_t
*b
, ppc_avr_t
*c
)
1096 for (i
= 0; i
< ARRAY_SIZE(r
->u16
); i
++) {
1097 prod
[i
] = a
->u16
[i
] * b
->u16
[i
];
1100 VECTOR_FOR_INORDER_I(i
, s32
) {
1101 uint64_t t
= (uint64_t)c
->u32
[i
] + prod
[2 * i
] + prod
[2 * i
+ 1];
1103 r
->u32
[i
] = cvtuduw(t
, &sat
);
1111 #define VMUL_DO_EVN(name, mul_element, mul_access, prod_access, cast) \
1112 void helper_V##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1116 for (i = 0; i < ARRAY_SIZE(r->mul_element); i += 2) { \
1117 r->prod_access(i >> 1) = (cast)a->mul_access(i) * \
1118 (cast)b->mul_access(i); \
1122 #define VMUL_DO_ODD(name, mul_element, mul_access, prod_access, cast) \
1123 void helper_V##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1127 for (i = 0; i < ARRAY_SIZE(r->mul_element); i += 2) { \
1128 r->prod_access(i >> 1) = (cast)a->mul_access(i + 1) * \
1129 (cast)b->mul_access(i + 1); \
1133 #define VMUL(suffix, mul_element, mul_access, prod_access, cast) \
1134 VMUL_DO_EVN(MULE##suffix, mul_element, mul_access, prod_access, cast) \
1135 VMUL_DO_ODD(MULO##suffix, mul_element, mul_access, prod_access, cast)
1136 VMUL(SB
, s8
, VsrSB
, VsrSH
, int16_t)
1137 VMUL(SH
, s16
, VsrSH
, VsrSW
, int32_t)
1138 VMUL(SW
, s32
, VsrSW
, VsrSD
, int64_t)
1139 VMUL(UB
, u8
, VsrB
, VsrH
, uint16_t)
1140 VMUL(UH
, u16
, VsrH
, VsrW
, uint32_t)
1141 VMUL(UW
, u32
, VsrW
, VsrD
, uint64_t)
1146 void helper_XXPERMX(ppc_vsr_t
*t
, ppc_vsr_t
*s0
, ppc_vsr_t
*s1
, ppc_vsr_t
*pcv
,
1150 ppc_vsr_t tmp
= { .u64
= {0, 0} };
1152 for (i
= 0; i
< ARRAY_SIZE(t
->u8
); i
++) {
1153 if ((pcv
->VsrB(i
) >> 5) == uim
) {
1154 idx
= pcv
->VsrB(i
) & 0x1f;
1155 if (idx
< ARRAY_SIZE(t
->u8
)) {
1156 tmp
.VsrB(i
) = s0
->VsrB(idx
);
1158 tmp
.VsrB(i
) = s1
->VsrB(idx
- ARRAY_SIZE(t
->u8
));
1166 void helper_VDIVSQ(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1168 Int128 neg1
= int128_makes64(-1);
1169 Int128 int128_min
= int128_make128(0, INT64_MIN
);
1170 if (likely(int128_nz(b
->s128
) &&
1171 (int128_ne(a
->s128
, int128_min
) || int128_ne(b
->s128
, neg1
)))) {
1172 t
->s128
= int128_divs(a
->s128
, b
->s128
);
1174 t
->s128
= a
->s128
; /* Undefined behavior */
1178 void helper_VDIVUQ(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1180 if (int128_nz(b
->s128
)) {
1181 t
->s128
= int128_divu(a
->s128
, b
->s128
);
1183 t
->s128
= a
->s128
; /* Undefined behavior */
1187 void helper_VDIVESD(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1192 for (i
= 0; i
< 2; i
++) {
1195 if (unlikely((high
== INT64_MIN
&& b
->s64
[i
] == -1) || !b
->s64
[i
])) {
1196 t
->s64
[i
] = a
->s64
[i
]; /* Undefined behavior */
1198 divs128(&low
, &high
, b
->s64
[i
]);
1204 void helper_VDIVEUD(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1208 for (i
= 0; i
< 2; i
++) {
1211 if (unlikely(!b
->u64
[i
])) {
1212 t
->u64
[i
] = a
->u64
[i
]; /* Undefined behavior */
1214 divu128(&low
, &high
, b
->u64
[i
]);
1220 void helper_VDIVESQ(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1223 Int128 int128_min
= int128_make128(0, INT64_MIN
);
1224 Int128 neg1
= int128_makes64(-1);
1227 low
= int128_zero();
1228 if (unlikely(!int128_nz(b
->s128
) ||
1229 (int128_eq(b
->s128
, neg1
) && int128_eq(high
, int128_min
)))) {
1230 t
->s128
= a
->s128
; /* Undefined behavior */
1232 divs256(&low
, &high
, b
->s128
);
1237 void helper_VDIVEUQ(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1242 low
= int128_zero();
1243 if (unlikely(!int128_nz(b
->s128
))) {
1244 t
->s128
= a
->s128
; /* Undefined behavior */
1246 divu256(&low
, &high
, b
->s128
);
1251 void helper_VMODSQ(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1253 Int128 neg1
= int128_makes64(-1);
1254 Int128 int128_min
= int128_make128(0, INT64_MIN
);
1255 if (likely(int128_nz(b
->s128
) &&
1256 (int128_ne(a
->s128
, int128_min
) || int128_ne(b
->s128
, neg1
)))) {
1257 t
->s128
= int128_rems(a
->s128
, b
->s128
);
1259 t
->s128
= int128_zero(); /* Undefined behavior */
1263 void helper_VMODUQ(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1265 if (likely(int128_nz(b
->s128
))) {
1266 t
->s128
= int128_remu(a
->s128
, b
->s128
);
1268 t
->s128
= int128_zero(); /* Undefined behavior */
1272 void helper_VPERM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
1277 for (i
= 0; i
< ARRAY_SIZE(r
->u8
); i
++) {
1278 int s
= c
->VsrB(i
) & 0x1f;
1279 int index
= s
& 0xf;
1282 result
.VsrB(i
) = b
->VsrB(index
);
1284 result
.VsrB(i
) = a
->VsrB(index
);
1290 void helper_VPERMR(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
1295 for (i
= 0; i
< ARRAY_SIZE(r
->u8
); i
++) {
1296 int s
= c
->VsrB(i
) & 0x1f;
1297 int index
= 15 - (s
& 0xf);
1300 result
.VsrB(i
) = a
->VsrB(index
);
1302 result
.VsrB(i
) = b
->VsrB(index
);
1308 #define XXGENPCV_BE_EXP(NAME, SZ) \
1309 void glue(helper_, glue(NAME, _be_exp))(ppc_vsr_t *t, ppc_vsr_t *b) \
1313 /* Initialize tmp with the result of an all-zeros mask */ \
1314 tmp.VsrD(0) = 0x1011121314151617; \
1315 tmp.VsrD(1) = 0x18191A1B1C1D1E1F; \
1317 /* Iterate over the most significant byte of each element */ \
1318 for (int i = 0, j = 0; i < ARRAY_SIZE(b->u8); i += SZ) { \
1319 if (b->VsrB(i) & 0x80) { \
1320 /* Update each byte of the element */ \
1321 for (int k = 0; k < SZ; k++) { \
1322 tmp.VsrB(i + k) = j + k; \
1331 #define XXGENPCV_BE_COMP(NAME, SZ) \
1332 void glue(helper_, glue(NAME, _be_comp))(ppc_vsr_t *t, ppc_vsr_t *b)\
1334 ppc_vsr_t tmp = { .u64 = { 0, 0 } }; \
1336 /* Iterate over the most significant byte of each element */ \
1337 for (int i = 0, j = 0; i < ARRAY_SIZE(b->u8); i += SZ) { \
1338 if (b->VsrB(i) & 0x80) { \
1339 /* Update each byte of the element */ \
1340 for (int k = 0; k < SZ; k++) { \
1341 tmp.VsrB(j + k) = i + k; \
1350 #define XXGENPCV_LE_EXP(NAME, SZ) \
1351 void glue(helper_, glue(NAME, _le_exp))(ppc_vsr_t *t, ppc_vsr_t *b) \
1355 /* Initialize tmp with the result of an all-zeros mask */ \
1356 tmp.VsrD(0) = 0x1F1E1D1C1B1A1918; \
1357 tmp.VsrD(1) = 0x1716151413121110; \
1359 /* Iterate over the most significant byte of each element */ \
1360 for (int i = 0, j = 0; i < ARRAY_SIZE(b->u8); i += SZ) { \
1361 /* Reverse indexing of "i" */ \
1362 const int idx = ARRAY_SIZE(b->u8) - i - SZ; \
1363 if (b->VsrB(idx) & 0x80) { \
1364 /* Update each byte of the element */ \
1365 for (int k = 0, rk = SZ - 1; k < SZ; k++, rk--) { \
1366 tmp.VsrB(idx + rk) = j + k; \
1375 #define XXGENPCV_LE_COMP(NAME, SZ) \
1376 void glue(helper_, glue(NAME, _le_comp))(ppc_vsr_t *t, ppc_vsr_t *b)\
1378 ppc_vsr_t tmp = { .u64 = { 0, 0 } }; \
1380 /* Iterate over the most significant byte of each element */ \
1381 for (int i = 0, j = 0; i < ARRAY_SIZE(b->u8); i += SZ) { \
1382 if (b->VsrB(ARRAY_SIZE(b->u8) - i - SZ) & 0x80) { \
1383 /* Update each byte of the element */ \
1384 for (int k = 0, rk = SZ - 1; k < SZ; k++, rk--) { \
1385 /* Reverse indexing of "j" */ \
1386 const int idx = ARRAY_SIZE(b->u8) - j - SZ; \
1387 tmp.VsrB(idx + rk) = i + k; \
1396 #define XXGENPCV(NAME, SZ) \
1397 XXGENPCV_BE_EXP(NAME, SZ) \
1398 XXGENPCV_BE_COMP(NAME, SZ) \
1399 XXGENPCV_LE_EXP(NAME, SZ) \
1400 XXGENPCV_LE_COMP(NAME, SZ) \
1402 XXGENPCV(XXGENPCVBM, 1)
1403 XXGENPCV(XXGENPCVHM
, 2)
1404 XXGENPCV(XXGENPCVWM
, 4)
1405 XXGENPCV(XXGENPCVDM
, 8)
1407 #undef XXGENPCV_BE_EXP
1408 #undef XXGENPCV_BE_COMP
1409 #undef XXGENPCV_LE_EXP
1410 #undef XXGENPCV_LE_COMP
1414 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[(i)])
1415 #define VBPERMD_INDEX(i) (i)
1416 #define VBPERMQ_DW(index) (((index) & 0x40) != 0)
1418 #define VBPERMQ_INDEX(avr, i) ((avr)->u8[15 - (i)])
1419 #define VBPERMD_INDEX(i) (1 - i)
1420 #define VBPERMQ_DW(index) (((index) & 0x40) == 0)
1422 #define EXTRACT_BIT(avr, i, index) \
1423 (extract64((avr)->VsrD(i), 63 - index, 1))
1425 void helper_vbpermd(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1428 ppc_avr_t result
= { .u64
= { 0, 0 } };
1429 VECTOR_FOR_INORDER_I(i
, u64
) {
1430 for (j
= 0; j
< 8; j
++) {
1431 int index
= VBPERMQ_INDEX(b
, (i
* 8) + j
);
1432 if (index
< 64 && EXTRACT_BIT(a
, i
, index
)) {
1433 result
.u64
[VBPERMD_INDEX(i
)] |= (0x80 >> j
);
1440 void helper_vbpermq(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1445 VECTOR_FOR_INORDER_I(i
, u8
) {
1446 int index
= VBPERMQ_INDEX(b
, i
);
1449 uint64_t mask
= (1ull << (63 - (index
& 0x3F)));
1450 if (a
->u64
[VBPERMQ_DW(index
)] & mask
) {
1451 perm
|= (0x8000 >> i
);
1460 #undef VBPERMQ_INDEX
1463 #define PMSUM(name, srcfld, trgfld, trgtyp) \
1464 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b) \
1467 trgtyp prod[sizeof(ppc_avr_t) / sizeof(a->srcfld[0])]; \
1469 VECTOR_FOR_INORDER_I(i, srcfld) { \
1471 for (j = 0; j < sizeof(a->srcfld[0]) * 8; j++) { \
1472 if (a->srcfld[i] & (1ull << j)) { \
1473 prod[i] ^= ((trgtyp)b->srcfld[i] << j); \
1478 VECTOR_FOR_INORDER_I(i, trgfld) { \
1479 r->trgfld[i] = prod[2 * i] ^ prod[2 * i + 1]; \
1483 PMSUM(vpmsumb
, u8
, u16
, uint16_t)
1484 PMSUM(vpmsumh
, u16
, u32
, uint32_t)
1485 PMSUM(vpmsumw
, u32
, u64
, uint64_t)
1487 void helper_VPMSUMD(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1490 Int128 tmp
, prod
[2] = {int128_zero(), int128_zero()};
1492 for (j
= 0; j
< 64; j
++) {
1493 for (i
= 0; i
< ARRAY_SIZE(r
->u64
); i
++) {
1494 if (a
->VsrD(i
) & (1ull << j
)) {
1495 tmp
= int128_make64(b
->VsrD(i
));
1496 tmp
= int128_lshift(tmp
, j
);
1497 prod
[i
] = int128_xor(prod
[i
], tmp
);
1502 r
->s128
= int128_xor(prod
[0], prod
[1]);
1510 void helper_vpkpx(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1515 const ppc_avr_t
*x
[2] = { a
, b
};
1517 const ppc_avr_t
*x
[2] = { b
, a
};
1520 VECTOR_FOR_INORDER_I(i
, u64
) {
1521 VECTOR_FOR_INORDER_I(j
, u32
) {
1522 uint32_t e
= x
[i
]->u32
[j
];
1524 result
.u16
[4 * i
+ j
] = (((e
>> 9) & 0xfc00) |
1525 ((e
>> 6) & 0x3e0) |
1532 #define VPK(suffix, from, to, cvt, dosat) \
1533 void helper_vpk##suffix(CPUPPCState *env, ppc_avr_t *r, \
1534 ppc_avr_t *a, ppc_avr_t *b) \
1539 ppc_avr_t *a0 = PKBIG ? a : b; \
1540 ppc_avr_t *a1 = PKBIG ? b : a; \
1542 VECTOR_FOR_INORDER_I(i, from) { \
1543 result.to[i] = cvt(a0->from[i], &sat); \
1544 result.to[i + ARRAY_SIZE(r->from)] = cvt(a1->from[i], &sat);\
1547 if (dosat && sat) { \
1548 set_vscr_sat(env); \
1552 VPK(shss
, s16
, s8
, cvtshsb
, 1)
1553 VPK(shus
, s16
, u8
, cvtshub
, 1)
1554 VPK(swss
, s32
, s16
, cvtswsh
, 1)
1555 VPK(swus
, s32
, u16
, cvtswuh
, 1)
1556 VPK(sdss
, s64
, s32
, cvtsdsw
, 1)
1557 VPK(sdus
, s64
, u32
, cvtsduw
, 1)
1558 VPK(uhus
, u16
, u8
, cvtuhub
, 1)
1559 VPK(uwus
, u32
, u16
, cvtuwuh
, 1)
1560 VPK(udus
, u64
, u32
, cvtuduw
, 1)
1561 VPK(uhum
, u16
, u8
, I
, 0)
1562 VPK(uwum
, u32
, u16
, I
, 0)
1563 VPK(udum
, u64
, u32
, I
, 0)
1568 void helper_vrefp(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*b
)
1572 for (i
= 0; i
< ARRAY_SIZE(r
->f32
); i
++) {
1573 r
->f32
[i
] = float32_div(float32_one
, b
->f32
[i
], &env
->vec_status
);
1577 #define VRFI(suffix, rounding) \
1578 void helper_vrfi##suffix(CPUPPCState *env, ppc_avr_t *r, \
1582 float_status s = env->vec_status; \
1584 set_float_rounding_mode(rounding, &s); \
1585 for (i = 0; i < ARRAY_SIZE(r->f32); i++) { \
1586 r->f32[i] = float32_round_to_int (b->f32[i], &s); \
1589 VRFI(n
, float_round_nearest_even
)
1590 VRFI(m
, float_round_down
)
1591 VRFI(p
, float_round_up
)
1592 VRFI(z
, float_round_to_zero
)
1595 void helper_vrsqrtefp(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*b
)
1599 for (i
= 0; i
< ARRAY_SIZE(r
->f32
); i
++) {
1600 float32 t
= float32_sqrt(b
->f32
[i
], &env
->vec_status
);
1602 r
->f32
[i
] = float32_div(float32_one
, t
, &env
->vec_status
);
1606 #define VRLMI(name, size, element, insert) \
1607 void helper_##name(ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, uint32_t desc) \
1610 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1611 uint##size##_t src1 = a->element[i]; \
1612 uint##size##_t src2 = b->element[i]; \
1613 uint##size##_t src3 = r->element[i]; \
1614 uint##size##_t begin, end, shift, mask, rot_val; \
1616 shift = extract##size(src2, 0, 6); \
1617 end = extract##size(src2, 8, 6); \
1618 begin = extract##size(src2, 16, 6); \
1619 rot_val = rol##size(src1, shift); \
1620 mask = mask_u##size(begin, end); \
1622 r->element[i] = (rot_val & mask) | (src3 & ~mask); \
1624 r->element[i] = (rot_val & mask); \
1629 VRLMI(VRLDMI
, 64, u64
, 1);
1630 VRLMI(VRLWMI
, 32, u32
, 1);
1631 VRLMI(VRLDNM
, 64, u64
, 0);
1632 VRLMI(VRLWNM
, 32, u32
, 0);
1634 void helper_vexptefp(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*b
)
1638 for (i
= 0; i
< ARRAY_SIZE(r
->f32
); i
++) {
1639 r
->f32
[i
] = float32_exp2(b
->f32
[i
], &env
->vec_status
);
1643 void helper_vlogefp(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*b
)
1647 for (i
= 0; i
< ARRAY_SIZE(r
->f32
); i
++) {
1648 r
->f32
[i
] = float32_log2(b
->f32
[i
], &env
->vec_status
);
1652 #define VEXTU_X_DO(name, size, left) \
1653 target_ulong glue(helper_, name)(target_ulong a, ppc_avr_t *b) \
1655 int index = (a & 0xf) * 8; \
1657 index = 128 - index - size; \
1659 return int128_getlo(int128_rshift(b->s128, index)) & \
1660 MAKE_64BIT_MASK(0, size); \
1662 VEXTU_X_DO(vextublx
, 8, 1)
1663 VEXTU_X_DO(vextuhlx
, 16, 1)
1664 VEXTU_X_DO(vextuwlx
, 32, 1)
1665 VEXTU_X_DO(vextubrx
, 8, 0)
1666 VEXTU_X_DO(vextuhrx
, 16, 0)
1667 VEXTU_X_DO(vextuwrx
, 32, 0)
1670 void helper_vslv(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1673 unsigned int shift
, bytes
, size
;
1675 size
= ARRAY_SIZE(r
->u8
);
1676 for (i
= 0; i
< size
; i
++) {
1677 shift
= b
->VsrB(i
) & 0x7; /* extract shift value */
1678 bytes
= (a
->VsrB(i
) << 8) + /* extract adjacent bytes */
1679 (((i
+ 1) < size
) ? a
->VsrB(i
+ 1) : 0);
1680 r
->VsrB(i
) = (bytes
<< shift
) >> 8; /* shift and store result */
1684 void helper_vsrv(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1687 unsigned int shift
, bytes
;
1690 * Use reverse order, as destination and source register can be
1691 * same. Its being modified in place saving temporary, reverse
1692 * order will guarantee that computed result is not fed back.
1694 for (i
= ARRAY_SIZE(r
->u8
) - 1; i
>= 0; i
--) {
1695 shift
= b
->VsrB(i
) & 0x7; /* extract shift value */
1696 bytes
= ((i
? a
->VsrB(i
- 1) : 0) << 8) + a
->VsrB(i
);
1697 /* extract adjacent bytes */
1698 r
->VsrB(i
) = (bytes
>> shift
) & 0xFF; /* shift and store result */
1702 void helper_vsldoi(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t shift
)
1704 int sh
= shift
& 0xf;
1708 for (i
= 0; i
< ARRAY_SIZE(r
->u8
); i
++) {
1711 result
.VsrB(i
) = b
->VsrB(index
- 0x10);
1713 result
.VsrB(i
) = a
->VsrB(index
);
1719 void helper_vslo(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1721 int sh
= (b
->VsrB(0xf) >> 3) & 0xf;
1724 memmove(&r
->u8
[0], &a
->u8
[sh
], 16 - sh
);
1725 memset(&r
->u8
[16 - sh
], 0, sh
);
1727 memmove(&r
->u8
[sh
], &a
->u8
[0], 16 - sh
);
1728 memset(&r
->u8
[0], 0, sh
);
1733 #define ELEM_ADDR(VEC, IDX, SIZE) (&(VEC)->u8[IDX])
1735 #define ELEM_ADDR(VEC, IDX, SIZE) (&(VEC)->u8[15 - (IDX)] - (SIZE) + 1)
1738 #define VINSX(SUFFIX, TYPE) \
1739 void glue(glue(helper_VINS, SUFFIX), LX)(CPUPPCState *env, ppc_avr_t *t, \
1740 uint64_t val, target_ulong index) \
1742 const int maxidx = ARRAY_SIZE(t->u8) - sizeof(TYPE); \
1743 target_long idx = index; \
1745 if (idx < 0 || idx > maxidx) { \
1746 idx = idx < 0 ? sizeof(TYPE) - idx : idx; \
1747 qemu_log_mask(LOG_GUEST_ERROR, \
1748 "Invalid index for Vector Insert Element after 0x" TARGET_FMT_lx \
1749 ", RA = " TARGET_FMT_ld " > %d\n", env->nip, idx, maxidx); \
1752 memcpy(ELEM_ADDR(t, idx, sizeof(TYPE)), &src, sizeof(TYPE)); \
1762 #define VEXTDVLX(NAME, SIZE) \
1763 void helper_##NAME(CPUPPCState *env, ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, \
1764 target_ulong index) \
1766 const target_long idx = index; \
1767 ppc_avr_t tmp[2] = { *a, *b }; \
1768 memset(t, 0, sizeof(*t)); \
1769 if (idx >= 0 && idx + SIZE <= sizeof(tmp)) { \
1770 memcpy(&t->u8[ARRAY_SIZE(t->u8) / 2 - SIZE], (void *)tmp + idx, SIZE); \
1772 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for " #NAME " after 0x" \
1773 TARGET_FMT_lx ", RC = " TARGET_FMT_ld " > %d\n", \
1774 env->nip, idx < 0 ? SIZE - idx : idx, 32 - SIZE); \
1778 #define VEXTDVLX(NAME, SIZE) \
1779 void helper_##NAME(CPUPPCState *env, ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, \
1780 target_ulong index) \
1782 const target_long idx = index; \
1783 ppc_avr_t tmp[2] = { *b, *a }; \
1784 memset(t, 0, sizeof(*t)); \
1785 if (idx >= 0 && idx + SIZE <= sizeof(tmp)) { \
1786 memcpy(&t->u8[ARRAY_SIZE(t->u8) / 2], \
1787 (void *)tmp + sizeof(tmp) - SIZE - idx, SIZE); \
1789 qemu_log_mask(LOG_GUEST_ERROR, "Invalid index for " #NAME " after 0x" \
1790 TARGET_FMT_lx ", RC = " TARGET_FMT_ld " > %d\n", \
1791 env->nip, idx < 0 ? SIZE - idx : idx, 32 - SIZE); \
1795 VEXTDVLX(VEXTDUBVLX
, 1)
1796 VEXTDVLX(VEXTDUHVLX
, 2)
1797 VEXTDVLX(VEXTDUWVLX
, 4)
1798 VEXTDVLX(VEXTDDVLX
, 8)
1801 #define VEXTRACT(suffix, element) \
1802 void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1804 uint32_t es = sizeof(r->element[0]); \
1805 memmove(&r->u8[8 - es], &b->u8[index], es); \
1806 memset(&r->u8[8], 0, 8); \
1807 memset(&r->u8[0], 0, 8 - es); \
1810 #define VEXTRACT(suffix, element) \
1811 void helper_vextract##suffix(ppc_avr_t *r, ppc_avr_t *b, uint32_t index) \
1813 uint32_t es = sizeof(r->element[0]); \
1814 uint32_t s = (16 - index) - es; \
1815 memmove(&r->u8[8], &b->u8[s], es); \
1816 memset(&r->u8[0], 0, 8); \
1817 memset(&r->u8[8 + es], 0, 8 - es); \
1826 #define VSTRI(NAME, ELEM, NUM_ELEMS, LEFT) \
1827 uint32_t helper_##NAME(ppc_avr_t *t, ppc_avr_t *b) \
1829 int i, idx, crf = 0; \
1831 for (i = 0; i < NUM_ELEMS; i++) { \
1832 idx = LEFT ? i : NUM_ELEMS - i - 1; \
1833 if (b->Vsr##ELEM(idx)) { \
1834 t->Vsr##ELEM(idx) = b->Vsr##ELEM(idx); \
1841 for (; i < NUM_ELEMS; i++) { \
1842 idx = LEFT ? i : NUM_ELEMS - i - 1; \
1843 t->Vsr##ELEM(idx) = 0; \
1848 VSTRI(VSTRIBL
, B
, 16, true)
1849 VSTRI(VSTRIBR
, B
, 16, false)
1850 VSTRI(VSTRIHL
, H
, 8, true)
1851 VSTRI(VSTRIHR
, H
, 8, false)
1854 void helper_XXEXTRACTUW(ppc_vsr_t
*xt
, ppc_vsr_t
*xb
, uint32_t index
)
1857 size_t es
= sizeof(uint32_t);
1862 for (i
= 0; i
< es
; i
++, ext_index
++) {
1863 t
.VsrB(8 - es
+ i
) = xb
->VsrB(ext_index
% 16);
1869 void helper_XXINSERTW(ppc_vsr_t
*xt
, ppc_vsr_t
*xb
, uint32_t index
)
1872 size_t es
= sizeof(uint32_t);
1873 int ins_index
, i
= 0;
1876 for (i
= 0; i
< es
&& ins_index
< 16; i
++, ins_index
++) {
1877 t
.VsrB(ins_index
) = xb
->VsrB(8 - es
+ i
);
1883 void helper_XXEVAL(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
,
1887 * Instead of processing imm bit-by-bit, we'll skip the computation of
1888 * conjunctions whose corresponding bit is unset.
1890 int bit
, imm
= simd_data(desc
);
1891 Int128 conj
, disj
= int128_zero();
1893 /* Iterate over set bits from the least to the most significant bit */
1896 * Get the next bit to be processed with ctz64. Invert the result of
1897 * ctz64 to match the indexing used by PowerISA.
1899 bit
= 7 - ctzl(imm
);
1903 conj
= int128_not(a
->s128
);
1906 conj
= int128_and(conj
, b
->s128
);
1908 conj
= int128_and(conj
, int128_not(b
->s128
));
1911 conj
= int128_and(conj
, c
->s128
);
1913 conj
= int128_and(conj
, int128_not(c
->s128
));
1915 disj
= int128_or(disj
, conj
);
1917 /* Unset the least significant bit that is set */
1924 #define XXBLEND(name, sz) \
1925 void glue(helper_XXBLENDV, name)(ppc_avr_t *t, ppc_avr_t *a, ppc_avr_t *b, \
1926 ppc_avr_t *c, uint32_t desc) \
1928 for (int i = 0; i < ARRAY_SIZE(t->glue(u, sz)); i++) { \
1929 t->glue(u, sz)[i] = (c->glue(s, sz)[i] >> (sz - 1)) ? \
1930 b->glue(u, sz)[i] : a->glue(u, sz)[i]; \
1939 #define VNEG(name, element) \
1940 void helper_##name(ppc_avr_t *r, ppc_avr_t *b) \
1943 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
1944 r->element[i] = -b->element[i]; \
1951 void helper_vsro(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1953 int sh
= (b
->VsrB(0xf) >> 3) & 0xf;
1956 memmove(&r
->u8
[sh
], &a
->u8
[0], 16 - sh
);
1957 memset(&r
->u8
[0], 0, sh
);
1959 memmove(&r
->u8
[0], &a
->u8
[sh
], 16 - sh
);
1960 memset(&r
->u8
[16 - sh
], 0, sh
);
1964 void helper_vsubcuw(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1968 for (i
= 0; i
< ARRAY_SIZE(r
->u32
); i
++) {
1969 r
->u32
[i
] = a
->u32
[i
] >= b
->u32
[i
];
1973 void helper_vsumsws(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
1980 upper
= ARRAY_SIZE(r
->s32
) - 1;
1981 t
= (int64_t)b
->VsrSW(upper
);
1982 for (i
= 0; i
< ARRAY_SIZE(r
->s32
); i
++) {
1984 result
.VsrSW(i
) = 0;
1986 result
.VsrSW(upper
) = cvtsdsw(t
, &sat
);
1994 void helper_vsum2sws(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2001 for (i
= 0; i
< ARRAY_SIZE(r
->u64
); i
++) {
2002 int64_t t
= (int64_t)b
->VsrSW(upper
+ i
* 2);
2005 for (j
= 0; j
< ARRAY_SIZE(r
->u64
); j
++) {
2006 t
+= a
->VsrSW(2 * i
+ j
);
2008 result
.VsrSW(upper
+ i
* 2) = cvtsdsw(t
, &sat
);
2017 void helper_vsum4sbs(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2022 for (i
= 0; i
< ARRAY_SIZE(r
->s32
); i
++) {
2023 int64_t t
= (int64_t)b
->s32
[i
];
2025 for (j
= 0; j
< ARRAY_SIZE(r
->s32
); j
++) {
2026 t
+= a
->s8
[4 * i
+ j
];
2028 r
->s32
[i
] = cvtsdsw(t
, &sat
);
2036 void helper_vsum4shs(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2041 for (i
= 0; i
< ARRAY_SIZE(r
->s32
); i
++) {
2042 int64_t t
= (int64_t)b
->s32
[i
];
2044 t
+= a
->s16
[2 * i
] + a
->s16
[2 * i
+ 1];
2045 r
->s32
[i
] = cvtsdsw(t
, &sat
);
2053 void helper_vsum4ubs(CPUPPCState
*env
, ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2058 for (i
= 0; i
< ARRAY_SIZE(r
->u32
); i
++) {
2059 uint64_t t
= (uint64_t)b
->u32
[i
];
2061 for (j
= 0; j
< ARRAY_SIZE(r
->u32
); j
++) {
2062 t
+= a
->u8
[4 * i
+ j
];
2064 r
->u32
[i
] = cvtuduw(t
, &sat
);
2079 #define VUPKPX(suffix, hi) \
2080 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \
2085 for (i = 0; i < ARRAY_SIZE(r->u32); i++) { \
2086 uint16_t e = b->u16[hi ? i : i + 4]; \
2087 uint8_t a = (e >> 15) ? 0xff : 0; \
2088 uint8_t r = (e >> 10) & 0x1f; \
2089 uint8_t g = (e >> 5) & 0x1f; \
2090 uint8_t b = e & 0x1f; \
2092 result.u32[i] = (a << 24) | (r << 16) | (g << 8) | b; \
2100 #define VUPK(suffix, unpacked, packee, hi) \
2101 void helper_vupk##suffix(ppc_avr_t *r, ppc_avr_t *b) \
2107 for (i = 0; i < ARRAY_SIZE(r->unpacked); i++) { \
2108 result.unpacked[i] = b->packee[i]; \
2111 for (i = ARRAY_SIZE(r->unpacked); i < ARRAY_SIZE(r->packee); \
2113 result.unpacked[i - ARRAY_SIZE(r->unpacked)] = b->packee[i]; \
2118 VUPK(hsb
, s16
, s8
, UPKHI
)
2119 VUPK(hsh
, s32
, s16
, UPKHI
)
2120 VUPK(hsw
, s64
, s32
, UPKHI
)
2121 VUPK(lsb
, s16
, s8
, UPKLO
)
2122 VUPK(lsh
, s32
, s16
, UPKLO
)
2123 VUPK(lsw
, s64
, s32
, UPKLO
)
2128 #define VGENERIC_DO(name, element) \
2129 void helper_v##name(ppc_avr_t *r, ppc_avr_t *b) \
2133 for (i = 0; i < ARRAY_SIZE(r->element); i++) { \
2134 r->element[i] = name(b->element[i]); \
2138 #define clzb(v) ((v) ? clz32((uint32_t)(v) << 24) : 8)
2139 #define clzh(v) ((v) ? clz32((uint32_t)(v) << 16) : 16)
2141 VGENERIC_DO(clzb
, u8
)
2142 VGENERIC_DO(clzh
, u16
)
2147 #define ctzb(v) ((v) ? ctz32(v) : 8)
2148 #define ctzh(v) ((v) ? ctz32(v) : 16)
2149 #define ctzw(v) ctz32((v))
2150 #define ctzd(v) ctz64((v))
2152 VGENERIC_DO(ctzb
, u8
)
2153 VGENERIC_DO(ctzh
, u16
)
2154 VGENERIC_DO(ctzw
, u32
)
2155 VGENERIC_DO(ctzd
, u64
)
2162 #define popcntb(v) ctpop8(v)
2163 #define popcnth(v) ctpop16(v)
2164 #define popcntw(v) ctpop32(v)
2165 #define popcntd(v) ctpop64(v)
2167 VGENERIC_DO(popcntb
, u8
)
2168 VGENERIC_DO(popcnth
, u16
)
2169 VGENERIC_DO(popcntw
, u32
)
2170 VGENERIC_DO(popcntd
, u64
)
2179 void helper_VADDUQM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2181 r
->s128
= int128_add(a
->s128
, b
->s128
);
2184 void helper_VADDEUQM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
2186 r
->s128
= int128_add(int128_add(a
->s128
, b
->s128
),
2187 int128_make64(int128_getlo(c
->s128
) & 1));
2190 void helper_VADDCUQ(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2192 r
->VsrD(1) = int128_ult(int128_not(a
->s128
), b
->s128
);
2196 void helper_VADDECUQ(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
2198 bool carry_out
= int128_ult(int128_not(a
->s128
), b
->s128
),
2199 carry_in
= int128_getlo(c
->s128
) & 1;
2201 if (!carry_out
&& carry_in
) {
2202 carry_out
= (int128_nz(a
->s128
) || int128_nz(b
->s128
)) &&
2203 int128_eq(int128_add(a
->s128
, b
->s128
), int128_makes64(-1));
2207 r
->VsrD(1) = carry_out
;
2210 void helper_VSUBUQM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2212 r
->s128
= int128_sub(a
->s128
, b
->s128
);
2215 void helper_VSUBEUQM(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
2217 r
->s128
= int128_add(int128_add(a
->s128
, int128_not(b
->s128
)),
2218 int128_make64(int128_getlo(c
->s128
) & 1));
2221 void helper_VSUBCUQ(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2223 Int128 tmp
= int128_not(b
->s128
);
2225 r
->VsrD(1) = int128_ult(int128_not(a
->s128
), tmp
) ||
2226 int128_eq(int128_add(a
->s128
, tmp
), int128_makes64(-1));
2230 void helper_VSUBECUQ(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
2232 Int128 tmp
= int128_not(b
->s128
);
2233 bool carry_out
= int128_ult(int128_not(a
->s128
), tmp
),
2234 carry_in
= int128_getlo(c
->s128
) & 1;
2236 r
->VsrD(1) = carry_out
|| (carry_in
&& int128_eq(int128_add(a
->s128
, tmp
),
2237 int128_makes64(-1)));
2241 #define BCD_PLUS_PREF_1 0xC
2242 #define BCD_PLUS_PREF_2 0xF
2243 #define BCD_PLUS_ALT_1 0xA
2244 #define BCD_NEG_PREF 0xD
2245 #define BCD_NEG_ALT 0xB
2246 #define BCD_PLUS_ALT_2 0xE
2247 #define NATIONAL_PLUS 0x2B
2248 #define NATIONAL_NEG 0x2D
2250 #define BCD_DIG_BYTE(n) (15 - ((n) / 2))
2252 static int bcd_get_sgn(ppc_avr_t
*bcd
)
2254 switch (bcd
->VsrB(BCD_DIG_BYTE(0)) & 0xF) {
2255 case BCD_PLUS_PREF_1
:
2256 case BCD_PLUS_PREF_2
:
2257 case BCD_PLUS_ALT_1
:
2258 case BCD_PLUS_ALT_2
:
2276 static int bcd_preferred_sgn(int sgn
, int ps
)
2279 return (ps
== 0) ? BCD_PLUS_PREF_1
: BCD_PLUS_PREF_2
;
2281 return BCD_NEG_PREF
;
2285 static uint8_t bcd_get_digit(ppc_avr_t
*bcd
, int n
, int *invalid
)
2289 result
= bcd
->VsrB(BCD_DIG_BYTE(n
)) >> 4;
2291 result
= bcd
->VsrB(BCD_DIG_BYTE(n
)) & 0xF;
2294 if (unlikely(result
> 9)) {
2300 static void bcd_put_digit(ppc_avr_t
*bcd
, uint8_t digit
, int n
)
2303 bcd
->VsrB(BCD_DIG_BYTE(n
)) &= 0x0F;
2304 bcd
->VsrB(BCD_DIG_BYTE(n
)) |= (digit
<< 4);
2306 bcd
->VsrB(BCD_DIG_BYTE(n
)) &= 0xF0;
2307 bcd
->VsrB(BCD_DIG_BYTE(n
)) |= digit
;
2311 static bool bcd_is_valid(ppc_avr_t
*bcd
)
2316 if (bcd_get_sgn(bcd
) == 0) {
2320 for (i
= 1; i
< 32; i
++) {
2321 bcd_get_digit(bcd
, i
, &invalid
);
2322 if (unlikely(invalid
)) {
2329 static int bcd_cmp_zero(ppc_avr_t
*bcd
)
2331 if (bcd
->VsrD(0) == 0 && (bcd
->VsrD(1) >> 4) == 0) {
2334 return (bcd_get_sgn(bcd
) == 1) ? CRF_GT
: CRF_LT
;
2338 static uint16_t get_national_digit(ppc_avr_t
*reg
, int n
)
2340 return reg
->VsrH(7 - n
);
2343 static void set_national_digit(ppc_avr_t
*reg
, uint8_t val
, int n
)
2345 reg
->VsrH(7 - n
) = val
;
2348 static int bcd_cmp_mag(ppc_avr_t
*a
, ppc_avr_t
*b
)
2352 for (i
= 31; i
> 0; i
--) {
2353 uint8_t dig_a
= bcd_get_digit(a
, i
, &invalid
);
2354 uint8_t dig_b
= bcd_get_digit(b
, i
, &invalid
);
2355 if (unlikely(invalid
)) {
2356 return 0; /* doesn't matter */
2357 } else if (dig_a
> dig_b
) {
2359 } else if (dig_a
< dig_b
) {
2367 static int bcd_add_mag(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
, int *invalid
,
2374 for (i
= 1; i
<= 31; i
++) {
2375 uint8_t digit
= bcd_get_digit(a
, i
, invalid
) +
2376 bcd_get_digit(b
, i
, invalid
) + carry
;
2377 is_zero
&= (digit
== 0);
2385 bcd_put_digit(t
, digit
, i
);
2392 static void bcd_sub_mag(ppc_avr_t
*t
, ppc_avr_t
*a
, ppc_avr_t
*b
, int *invalid
,
2398 for (i
= 1; i
<= 31; i
++) {
2399 uint8_t digit
= bcd_get_digit(a
, i
, invalid
) -
2400 bcd_get_digit(b
, i
, invalid
) + carry
;
2408 bcd_put_digit(t
, digit
, i
);
2414 uint32_t helper_bcdadd(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2417 int sgna
= bcd_get_sgn(a
);
2418 int sgnb
= bcd_get_sgn(b
);
2419 int invalid
= (sgna
== 0) || (sgnb
== 0);
2423 ppc_avr_t result
= { .u64
= { 0, 0 } };
2427 result
.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(sgna
, ps
);
2428 zero
= bcd_add_mag(&result
, a
, b
, &invalid
, &overflow
);
2429 cr
= (sgna
> 0) ? CRF_GT
: CRF_LT
;
2431 int magnitude
= bcd_cmp_mag(a
, b
);
2432 if (magnitude
> 0) {
2433 result
.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(sgna
, ps
);
2434 bcd_sub_mag(&result
, a
, b
, &invalid
, &overflow
);
2435 cr
= (sgna
> 0) ? CRF_GT
: CRF_LT
;
2436 } else if (magnitude
< 0) {
2437 result
.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(sgnb
, ps
);
2438 bcd_sub_mag(&result
, b
, a
, &invalid
, &overflow
);
2439 cr
= (sgnb
> 0) ? CRF_GT
: CRF_LT
;
2441 result
.VsrB(BCD_DIG_BYTE(0)) = bcd_preferred_sgn(0, ps
);
2447 if (unlikely(invalid
)) {
2448 result
.VsrD(0) = result
.VsrD(1) = -1;
2450 } else if (overflow
) {
2461 uint32_t helper_bcdsub(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2463 ppc_avr_t bcopy
= *b
;
2464 int sgnb
= bcd_get_sgn(b
);
2466 bcd_put_digit(&bcopy
, BCD_PLUS_PREF_1
, 0);
2467 } else if (sgnb
> 0) {
2468 bcd_put_digit(&bcopy
, BCD_NEG_PREF
, 0);
2470 /* else invalid ... defer to bcdadd code for proper handling */
2472 return helper_bcdadd(r
, a
, &bcopy
, ps
);
2475 uint32_t helper_bcdcfn(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2479 uint16_t national
= 0;
2480 uint16_t sgnb
= get_national_digit(b
, 0);
2481 ppc_avr_t ret
= { .u64
= { 0, 0 } };
2482 int invalid
= (sgnb
!= NATIONAL_PLUS
&& sgnb
!= NATIONAL_NEG
);
2484 for (i
= 1; i
< 8; i
++) {
2485 national
= get_national_digit(b
, i
);
2486 if (unlikely(national
< 0x30 || national
> 0x39)) {
2491 bcd_put_digit(&ret
, national
& 0xf, i
);
2494 if (sgnb
== NATIONAL_PLUS
) {
2495 bcd_put_digit(&ret
, (ps
== 0) ? BCD_PLUS_PREF_1
: BCD_PLUS_PREF_2
, 0);
2497 bcd_put_digit(&ret
, BCD_NEG_PREF
, 0);
2500 cr
= bcd_cmp_zero(&ret
);
2502 if (unlikely(invalid
)) {
2511 uint32_t helper_bcdctn(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2515 int sgnb
= bcd_get_sgn(b
);
2516 int invalid
= (sgnb
== 0);
2517 ppc_avr_t ret
= { .u64
= { 0, 0 } };
2519 int ox_flag
= (b
->VsrD(0) != 0) || ((b
->VsrD(1) >> 32) != 0);
2521 for (i
= 1; i
< 8; i
++) {
2522 set_national_digit(&ret
, 0x30 + bcd_get_digit(b
, i
, &invalid
), i
);
2524 if (unlikely(invalid
)) {
2528 set_national_digit(&ret
, (sgnb
== -1) ? NATIONAL_NEG
: NATIONAL_PLUS
, 0);
2530 cr
= bcd_cmp_zero(b
);
2536 if (unlikely(invalid
)) {
2545 uint32_t helper_bcdcfz(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2551 int zone_lead
= ps
? 0xF : 0x3;
2553 ppc_avr_t ret
= { .u64
= { 0, 0 } };
2554 int sgnb
= b
->VsrB(BCD_DIG_BYTE(0)) >> 4;
2556 if (unlikely((sgnb
< 0xA) && ps
)) {
2560 for (i
= 0; i
< 16; i
++) {
2561 zone_digit
= i
? b
->VsrB(BCD_DIG_BYTE(i
* 2)) >> 4 : zone_lead
;
2562 digit
= b
->VsrB(BCD_DIG_BYTE(i
* 2)) & 0xF;
2563 if (unlikely(zone_digit
!= zone_lead
|| digit
> 0x9)) {
2568 bcd_put_digit(&ret
, digit
, i
+ 1);
2571 if ((ps
&& (sgnb
== 0xB || sgnb
== 0xD)) ||
2572 (!ps
&& (sgnb
& 0x4))) {
2573 bcd_put_digit(&ret
, BCD_NEG_PREF
, 0);
2575 bcd_put_digit(&ret
, BCD_PLUS_PREF_1
, 0);
2578 cr
= bcd_cmp_zero(&ret
);
2580 if (unlikely(invalid
)) {
2589 uint32_t helper_bcdctz(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2594 int sgnb
= bcd_get_sgn(b
);
2595 int zone_lead
= (ps
) ? 0xF0 : 0x30;
2596 int invalid
= (sgnb
== 0);
2597 ppc_avr_t ret
= { .u64
= { 0, 0 } };
2599 int ox_flag
= ((b
->VsrD(0) >> 4) != 0);
2601 for (i
= 0; i
< 16; i
++) {
2602 digit
= bcd_get_digit(b
, i
+ 1, &invalid
);
2604 if (unlikely(invalid
)) {
2608 ret
.VsrB(BCD_DIG_BYTE(i
* 2)) = zone_lead
+ digit
;
2612 bcd_put_digit(&ret
, (sgnb
== 1) ? 0xC : 0xD, 1);
2614 bcd_put_digit(&ret
, (sgnb
== 1) ? 0x3 : 0x7, 1);
2617 cr
= bcd_cmp_zero(b
);
2623 if (unlikely(invalid
)) {
2633 * Compare 2 128-bit unsigned integers, passed in as unsigned 64-bit pairs
2636 * > 0 if ahi|alo > bhi|blo,
2637 * 0 if ahi|alo == bhi|blo,
2638 * < 0 if ahi|alo < bhi|blo
2640 static inline int ucmp128(uint64_t alo
, uint64_t ahi
,
2641 uint64_t blo
, uint64_t bhi
)
2643 return (ahi
== bhi
) ?
2644 (alo
> blo
? 1 : (alo
== blo
? 0 : -1)) :
2645 (ahi
> bhi
? 1 : -1);
2648 uint32_t helper_bcdcfsq(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2655 ppc_avr_t ret
= { .u64
= { 0, 0 } };
2657 if (b
->VsrSD(0) < 0) {
2658 lo_value
= -b
->VsrSD(1);
2659 hi_value
= ~b
->VsrD(0) + !lo_value
;
2660 bcd_put_digit(&ret
, 0xD, 0);
2664 lo_value
= b
->VsrD(1);
2665 hi_value
= b
->VsrD(0);
2666 bcd_put_digit(&ret
, bcd_preferred_sgn(0, ps
), 0);
2668 if (hi_value
== 0 && lo_value
== 0) {
2676 * Check src limits: abs(src) <= 10^31 - 1
2678 * 10^31 - 1 = 0x0000007e37be2022 c0914b267fffffff
2680 if (ucmp128(lo_value
, hi_value
,
2681 0xc0914b267fffffffULL
, 0x7e37be2022ULL
) > 0) {
2685 * According to the ISA, if src wouldn't fit in the destination
2686 * register, the result is undefined.
2687 * In that case, we leave r unchanged.
2690 rem
= divu128(&lo_value
, &hi_value
, 1000000000000000ULL);
2692 for (i
= 1; i
< 16; rem
/= 10, i
++) {
2693 bcd_put_digit(&ret
, rem
% 10, i
);
2696 for (; i
< 32; lo_value
/= 10, i
++) {
2697 bcd_put_digit(&ret
, lo_value
% 10, i
);
2706 uint32_t helper_bcdctsq(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2713 uint64_t hi_value
= 0;
2714 int sgnb
= bcd_get_sgn(b
);
2715 int invalid
= (sgnb
== 0);
2717 lo_value
= bcd_get_digit(b
, 31, &invalid
);
2718 for (i
= 30; i
> 0; i
--) {
2719 mulu64(&lo_value
, &carry
, lo_value
, 10ULL);
2720 mulu64(&hi_value
, &unused
, hi_value
, 10ULL);
2721 lo_value
+= bcd_get_digit(b
, i
, &invalid
);
2724 if (unlikely(invalid
)) {
2730 r
->VsrSD(1) = -lo_value
;
2731 r
->VsrSD(0) = ~hi_value
+ !r
->VsrSD(1);
2733 r
->VsrSD(1) = lo_value
;
2734 r
->VsrSD(0) = hi_value
;
2737 cr
= bcd_cmp_zero(b
);
2739 if (unlikely(invalid
)) {
2746 uint32_t helper_bcdcpsgn(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2751 if (bcd_get_sgn(a
) == 0 || bcd_get_sgn(b
) == 0) {
2756 bcd_put_digit(r
, b
->VsrB(BCD_DIG_BYTE(0)) & 0xF, 0);
2758 for (i
= 1; i
< 32; i
++) {
2759 bcd_get_digit(a
, i
, &invalid
);
2760 bcd_get_digit(b
, i
, &invalid
);
2761 if (unlikely(invalid
)) {
2766 return bcd_cmp_zero(r
);
2769 uint32_t helper_bcdsetsgn(ppc_avr_t
*r
, ppc_avr_t
*b
, uint32_t ps
)
2771 int sgnb
= bcd_get_sgn(b
);
2774 bcd_put_digit(r
, bcd_preferred_sgn(sgnb
, ps
), 0);
2776 if (bcd_is_valid(b
) == false) {
2780 return bcd_cmp_zero(r
);
2783 uint32_t helper_bcds(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2786 int i
= a
->VsrSB(7);
2787 bool ox_flag
= false;
2788 int sgnb
= bcd_get_sgn(b
);
2790 ret
.VsrD(1) &= ~0xf;
2792 if (bcd_is_valid(b
) == false) {
2796 if (unlikely(i
> 31)) {
2798 } else if (unlikely(i
< -31)) {
2803 ulshift(&ret
.VsrD(1), &ret
.VsrD(0), i
* 4, &ox_flag
);
2805 urshift(&ret
.VsrD(1), &ret
.VsrD(0), -i
* 4);
2807 bcd_put_digit(&ret
, bcd_preferred_sgn(sgnb
, ps
), 0);
2811 cr
= bcd_cmp_zero(r
);
2819 uint32_t helper_bcdus(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2824 bool ox_flag
= false;
2827 for (i
= 0; i
< 32; i
++) {
2828 bcd_get_digit(b
, i
, &invalid
);
2830 if (unlikely(invalid
)) {
2838 ret
.VsrD(1) = ret
.VsrD(0) = 0;
2839 } else if (i
<= -32) {
2840 ret
.VsrD(1) = ret
.VsrD(0) = 0;
2842 ulshift(&ret
.VsrD(1), &ret
.VsrD(0), i
* 4, &ox_flag
);
2844 urshift(&ret
.VsrD(1), &ret
.VsrD(0), -i
* 4);
2848 cr
= bcd_cmp_zero(r
);
2856 uint32_t helper_bcdsr(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2861 bool ox_flag
= false;
2862 int sgnb
= bcd_get_sgn(b
);
2864 ret
.VsrD(1) &= ~0xf;
2866 int i
= a
->VsrSB(7);
2869 bcd_one
.VsrD(0) = 0;
2870 bcd_one
.VsrD(1) = 0x10;
2872 if (bcd_is_valid(b
) == false) {
2876 if (unlikely(i
> 31)) {
2878 } else if (unlikely(i
< -31)) {
2883 ulshift(&ret
.VsrD(1), &ret
.VsrD(0), i
* 4, &ox_flag
);
2885 urshift(&ret
.VsrD(1), &ret
.VsrD(0), -i
* 4);
2887 if (bcd_get_digit(&ret
, 0, &invalid
) >= 5) {
2888 bcd_add_mag(&ret
, &ret
, &bcd_one
, &invalid
, &unused
);
2891 bcd_put_digit(&ret
, bcd_preferred_sgn(sgnb
, ps
), 0);
2893 cr
= bcd_cmp_zero(&ret
);
2902 uint32_t helper_bcdtrunc(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2905 uint32_t ox_flag
= 0;
2906 int i
= a
->VsrSH(3) + 1;
2909 if (bcd_is_valid(b
) == false) {
2913 if (i
> 16 && i
< 32) {
2914 mask
= (uint64_t)-1 >> (128 - i
* 4);
2915 if (ret
.VsrD(0) & ~mask
) {
2919 ret
.VsrD(0) &= mask
;
2920 } else if (i
>= 0 && i
<= 16) {
2921 mask
= (uint64_t)-1 >> (64 - i
* 4);
2922 if (ret
.VsrD(0) || (ret
.VsrD(1) & ~mask
)) {
2926 ret
.VsrD(1) &= mask
;
2929 bcd_put_digit(&ret
, bcd_preferred_sgn(bcd_get_sgn(b
), ps
), 0);
2932 return bcd_cmp_zero(&ret
) | ox_flag
;
2935 uint32_t helper_bcdutrunc(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, uint32_t ps
)
2939 uint32_t ox_flag
= 0;
2943 for (i
= 0; i
< 32; i
++) {
2944 bcd_get_digit(b
, i
, &invalid
);
2946 if (unlikely(invalid
)) {
2952 if (i
> 16 && i
< 33) {
2953 mask
= (uint64_t)-1 >> (128 - i
* 4);
2954 if (ret
.VsrD(0) & ~mask
) {
2958 ret
.VsrD(0) &= mask
;
2959 } else if (i
> 0 && i
<= 16) {
2960 mask
= (uint64_t)-1 >> (64 - i
* 4);
2961 if (ret
.VsrD(0) || (ret
.VsrD(1) & ~mask
)) {
2965 ret
.VsrD(1) &= mask
;
2967 } else if (i
== 0) {
2968 if (ret
.VsrD(0) || ret
.VsrD(1)) {
2971 ret
.VsrD(0) = ret
.VsrD(1) = 0;
2975 if (r
->VsrD(0) == 0 && r
->VsrD(1) == 0) {
2976 return ox_flag
| CRF_EQ
;
2979 return ox_flag
| CRF_GT
;
2982 void helper_vsbox(ppc_avr_t
*r
, ppc_avr_t
*a
)
2985 VECTOR_FOR_INORDER_I(i
, u8
) {
2986 r
->u8
[i
] = AES_sbox
[a
->u8
[i
]];
2990 void helper_vcipher(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
2995 VECTOR_FOR_INORDER_I(i
, u32
) {
2996 result
.VsrW(i
) = b
->VsrW(i
) ^
2997 (AES_Te0
[a
->VsrB(AES_shifts
[4 * i
+ 0])] ^
2998 AES_Te1
[a
->VsrB(AES_shifts
[4 * i
+ 1])] ^
2999 AES_Te2
[a
->VsrB(AES_shifts
[4 * i
+ 2])] ^
3000 AES_Te3
[a
->VsrB(AES_shifts
[4 * i
+ 3])]);
3005 void helper_vcipherlast(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
3010 VECTOR_FOR_INORDER_I(i
, u8
) {
3011 result
.VsrB(i
) = b
->VsrB(i
) ^ (AES_sbox
[a
->VsrB(AES_shifts
[i
])]);
3016 void helper_vncipher(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
3018 /* This differs from what is written in ISA V2.07. The RTL is */
3019 /* incorrect and will be fixed in V2.07B. */
3023 VECTOR_FOR_INORDER_I(i
, u8
) {
3024 tmp
.VsrB(i
) = b
->VsrB(i
) ^ AES_isbox
[a
->VsrB(AES_ishifts
[i
])];
3027 VECTOR_FOR_INORDER_I(i
, u32
) {
3029 AES_imc
[tmp
.VsrB(4 * i
+ 0)][0] ^
3030 AES_imc
[tmp
.VsrB(4 * i
+ 1)][1] ^
3031 AES_imc
[tmp
.VsrB(4 * i
+ 2)][2] ^
3032 AES_imc
[tmp
.VsrB(4 * i
+ 3)][3];
3036 void helper_vncipherlast(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
)
3041 VECTOR_FOR_INORDER_I(i
, u8
) {
3042 result
.VsrB(i
) = b
->VsrB(i
) ^ (AES_isbox
[a
->VsrB(AES_ishifts
[i
])]);
3047 void helper_vshasigmaw(ppc_avr_t
*r
, ppc_avr_t
*a
, uint32_t st_six
)
3049 int st
= (st_six
& 0x10) != 0;
3050 int six
= st_six
& 0xF;
3053 for (i
= 0; i
< ARRAY_SIZE(r
->u32
); i
++) {
3055 if ((six
& (0x8 >> i
)) == 0) {
3056 r
->VsrW(i
) = ror32(a
->VsrW(i
), 7) ^
3057 ror32(a
->VsrW(i
), 18) ^
3059 } else { /* six.bit[i] == 1 */
3060 r
->VsrW(i
) = ror32(a
->VsrW(i
), 17) ^
3061 ror32(a
->VsrW(i
), 19) ^
3064 } else { /* st == 1 */
3065 if ((six
& (0x8 >> i
)) == 0) {
3066 r
->VsrW(i
) = ror32(a
->VsrW(i
), 2) ^
3067 ror32(a
->VsrW(i
), 13) ^
3068 ror32(a
->VsrW(i
), 22);
3069 } else { /* six.bit[i] == 1 */
3070 r
->VsrW(i
) = ror32(a
->VsrW(i
), 6) ^
3071 ror32(a
->VsrW(i
), 11) ^
3072 ror32(a
->VsrW(i
), 25);
3078 void helper_vshasigmad(ppc_avr_t
*r
, ppc_avr_t
*a
, uint32_t st_six
)
3080 int st
= (st_six
& 0x10) != 0;
3081 int six
= st_six
& 0xF;
3084 for (i
= 0; i
< ARRAY_SIZE(r
->u64
); i
++) {
3086 if ((six
& (0x8 >> (2 * i
))) == 0) {
3087 r
->VsrD(i
) = ror64(a
->VsrD(i
), 1) ^
3088 ror64(a
->VsrD(i
), 8) ^
3090 } else { /* six.bit[2*i] == 1 */
3091 r
->VsrD(i
) = ror64(a
->VsrD(i
), 19) ^
3092 ror64(a
->VsrD(i
), 61) ^
3095 } else { /* st == 1 */
3096 if ((six
& (0x8 >> (2 * i
))) == 0) {
3097 r
->VsrD(i
) = ror64(a
->VsrD(i
), 28) ^
3098 ror64(a
->VsrD(i
), 34) ^
3099 ror64(a
->VsrD(i
), 39);
3100 } else { /* six.bit[2*i] == 1 */
3101 r
->VsrD(i
) = ror64(a
->VsrD(i
), 14) ^
3102 ror64(a
->VsrD(i
), 18) ^
3103 ror64(a
->VsrD(i
), 41);
3109 void helper_vpermxor(ppc_avr_t
*r
, ppc_avr_t
*a
, ppc_avr_t
*b
, ppc_avr_t
*c
)
3114 for (i
= 0; i
< ARRAY_SIZE(r
->u8
); i
++) {
3115 int indexA
= c
->VsrB(i
) >> 4;
3116 int indexB
= c
->VsrB(i
) & 0xF;
3118 result
.VsrB(i
) = a
->VsrB(indexA
) ^ b
->VsrB(indexB
);
3123 #undef VECTOR_FOR_INORDER_I
3125 /*****************************************************************************/
3126 /* SPE extension helpers */
3127 /* Use a table to make this quicker */
3128 static const uint8_t hbrev
[16] = {
3129 0x0, 0x8, 0x4, 0xC, 0x2, 0xA, 0x6, 0xE,
3130 0x1, 0x9, 0x5, 0xD, 0x3, 0xB, 0x7, 0xF,
3133 static inline uint8_t byte_reverse(uint8_t val
)
3135 return hbrev
[val
>> 4] | (hbrev
[val
& 0xF] << 4);
3138 static inline uint32_t word_reverse(uint32_t val
)
3140 return byte_reverse(val
>> 24) | (byte_reverse(val
>> 16) << 8) |
3141 (byte_reverse(val
>> 8) << 16) | (byte_reverse(val
) << 24);
3144 #define MASKBITS 16 /* Random value - to be fixed (implementation dependent) */
3145 target_ulong
helper_brinc(target_ulong arg1
, target_ulong arg2
)
3147 uint32_t a
, b
, d
, mask
;
3149 mask
= UINT32_MAX
>> (32 - MASKBITS
);
3152 d
= word_reverse(1 + word_reverse(a
| ~b
));
3153 return (arg1
& ~mask
) | (d
& b
);
3156 uint32_t helper_cntlsw32(uint32_t val
)
3158 if (val
& 0x80000000) {
3165 uint32_t helper_cntlzw32(uint32_t val
)
3171 target_ulong
helper_dlmzb(CPUPPCState
*env
, target_ulong high
,
3172 target_ulong low
, uint32_t update_Rc
)
3178 for (mask
= 0xFF000000; mask
!= 0; mask
= mask
>> 8) {
3179 if ((high
& mask
) == 0) {
3187 for (mask
= 0xFF000000; mask
!= 0; mask
= mask
>> 8) {
3188 if ((low
& mask
) == 0) {
3201 env
->xer
= (env
->xer
& ~0x7F) | i
;
3203 env
->crf
[0] |= xer_so
;