2
* MMX/3DNow!/SSE/SSE2/SSE3/SSSE3/SSE4/PNI support
4
* Copyright (c) 2005 Fabrice Bellard
5
* Copyright (c) 2008 Intel Corporation <andrew.zaborowski@intel.com>
7
* This library is free software; you can redistribute it and/or
8
* modify it under the terms of the GNU Lesser General Public
9
* License as published by the Free Software Foundation; either
10
* version 2 of the License, or (at your option) any later version.
12
* This library is distributed in the hope that it will be useful,
13
* but WITHOUT ANY WARRANTY; without even the implied warranty of
14
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15
* Lesser General Public License for more details.
17
* You should have received a copy of the GNU Lesser General Public
18
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
30
#define XMM_ONLY(...) __VA_ARGS__
38
void glue(helper_psrlw, SUFFIX)(Reg *d, Reg *s)
62
void glue(helper_psraw, SUFFIX)(Reg *d, Reg *s)
71
d->W(0) = (int16_t)d->W(0) >> shift;
72
d->W(1) = (int16_t)d->W(1) >> shift;
73
d->W(2) = (int16_t)d->W(2) >> shift;
74
d->W(3) = (int16_t)d->W(3) >> shift;
76
d->W(4) = (int16_t)d->W(4) >> shift;
77
d->W(5) = (int16_t)d->W(5) >> shift;
78
d->W(6) = (int16_t)d->W(6) >> shift;
79
d->W(7) = (int16_t)d->W(7) >> shift;
83
void glue(helper_psllw, SUFFIX)(Reg *d, Reg *s)
107
void glue(helper_psrld, SUFFIX)(Reg *d, Reg *s)
127
void glue(helper_psrad, SUFFIX)(Reg *d, Reg *s)
136
d->L(0) = (int32_t)d->L(0) >> shift;
137
d->L(1) = (int32_t)d->L(1) >> shift;
139
d->L(2) = (int32_t)d->L(2) >> shift;
140
d->L(3) = (int32_t)d->L(3) >> shift;
144
void glue(helper_pslld, SUFFIX)(Reg *d, Reg *s)
164
void glue(helper_psrlq, SUFFIX)(Reg *d, Reg *s)
182
void glue(helper_psllq, SUFFIX)(Reg *d, Reg *s)
201
void glue(helper_psrldq, SUFFIX)(Reg *d, Reg *s)
208
for(i = 0; i < 16 - shift; i++)
209
d->B(i) = d->B(i + shift);
210
for(i = 16 - shift; i < 16; i++)
214
void glue(helper_pslldq, SUFFIX)(Reg *d, Reg *s)
221
for(i = 15; i >= shift; i--)
222
d->B(i) = d->B(i - shift);
223
for(i = 0; i < shift; i++)
228
#define SSE_HELPER_B(name, F)\
229
void glue(name, SUFFIX) (Reg *d, Reg *s)\
231
d->B(0) = F(d->B(0), s->B(0));\
232
d->B(1) = F(d->B(1), s->B(1));\
233
d->B(2) = F(d->B(2), s->B(2));\
234
d->B(3) = F(d->B(3), s->B(3));\
235
d->B(4) = F(d->B(4), s->B(4));\
236
d->B(5) = F(d->B(5), s->B(5));\
237
d->B(6) = F(d->B(6), s->B(6));\
238
d->B(7) = F(d->B(7), s->B(7));\
240
d->B(8) = F(d->B(8), s->B(8));\
241
d->B(9) = F(d->B(9), s->B(9));\
242
d->B(10) = F(d->B(10), s->B(10));\
243
d->B(11) = F(d->B(11), s->B(11));\
244
d->B(12) = F(d->B(12), s->B(12));\
245
d->B(13) = F(d->B(13), s->B(13));\
246
d->B(14) = F(d->B(14), s->B(14));\
247
d->B(15) = F(d->B(15), s->B(15));\
251
#define SSE_HELPER_W(name, F)\
252
void glue(name, SUFFIX) (Reg *d, Reg *s)\
254
d->W(0) = F(d->W(0), s->W(0));\
255
d->W(1) = F(d->W(1), s->W(1));\
256
d->W(2) = F(d->W(2), s->W(2));\
257
d->W(3) = F(d->W(3), s->W(3));\
259
d->W(4) = F(d->W(4), s->W(4));\
260
d->W(5) = F(d->W(5), s->W(5));\
261
d->W(6) = F(d->W(6), s->W(6));\
262
d->W(7) = F(d->W(7), s->W(7));\
266
#define SSE_HELPER_L(name, F)\
267
void glue(name, SUFFIX) (Reg *d, Reg *s)\
269
d->L(0) = F(d->L(0), s->L(0));\
270
d->L(1) = F(d->L(1), s->L(1));\
272
d->L(2) = F(d->L(2), s->L(2));\
273
d->L(3) = F(d->L(3), s->L(3));\
277
#define SSE_HELPER_Q(name, F)\
278
void glue(name, SUFFIX) (Reg *d, Reg *s)\
280
d->Q(0) = F(d->Q(0), s->Q(0));\
282
d->Q(1) = F(d->Q(1), s->Q(1));\
287
static inline int satub(int x)
297
static inline int satuw(int x)
307
static inline int satsb(int x)
317
static inline int satsw(int x)
327
#define FADD(a, b) ((a) + (b))
328
#define FADDUB(a, b) satub((a) + (b))
329
#define FADDUW(a, b) satuw((a) + (b))
330
#define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
331
#define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
333
#define FSUB(a, b) ((a) - (b))
334
#define FSUBUB(a, b) satub((a) - (b))
335
#define FSUBUW(a, b) satuw((a) - (b))
336
#define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
337
#define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
338
#define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
339
#define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
340
#define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
341
#define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
343
#define FAND(a, b) (a) & (b)
344
#define FANDN(a, b) ((~(a)) & (b))
345
#define FOR(a, b) (a) | (b)
346
#define FXOR(a, b) (a) ^ (b)
348
#define FCMPGTB(a, b) (int8_t)(a) > (int8_t)(b) ? -1 : 0
349
#define FCMPGTW(a, b) (int16_t)(a) > (int16_t)(b) ? -1 : 0
350
#define FCMPGTL(a, b) (int32_t)(a) > (int32_t)(b) ? -1 : 0
351
#define FCMPEQ(a, b) (a) == (b) ? -1 : 0
353
#define FMULLW(a, b) (a) * (b)
354
#define FMULHRW(a, b) ((int16_t)(a) * (int16_t)(b) + 0x8000) >> 16
355
#define FMULHUW(a, b) (a) * (b) >> 16
356
#define FMULHW(a, b) (int16_t)(a) * (int16_t)(b) >> 16
358
#define FAVG(a, b) ((a) + (b) + 1) >> 1
361
SSE_HELPER_B(helper_paddb, FADD)
362
SSE_HELPER_W(helper_paddw, FADD)
363
SSE_HELPER_L(helper_paddl, FADD)
364
SSE_HELPER_Q(helper_paddq, FADD)
366
SSE_HELPER_B(helper_psubb, FSUB)
367
SSE_HELPER_W(helper_psubw, FSUB)
368
SSE_HELPER_L(helper_psubl, FSUB)
369
SSE_HELPER_Q(helper_psubq, FSUB)
371
SSE_HELPER_B(helper_paddusb, FADDUB)
372
SSE_HELPER_B(helper_paddsb, FADDSB)
373
SSE_HELPER_B(helper_psubusb, FSUBUB)
374
SSE_HELPER_B(helper_psubsb, FSUBSB)
376
SSE_HELPER_W(helper_paddusw, FADDUW)
377
SSE_HELPER_W(helper_paddsw, FADDSW)
378
SSE_HELPER_W(helper_psubusw, FSUBUW)
379
SSE_HELPER_W(helper_psubsw, FSUBSW)
381
SSE_HELPER_B(helper_pminub, FMINUB)
382
SSE_HELPER_B(helper_pmaxub, FMAXUB)
384
SSE_HELPER_W(helper_pminsw, FMINSW)
385
SSE_HELPER_W(helper_pmaxsw, FMAXSW)
387
SSE_HELPER_Q(helper_pand, FAND)
388
SSE_HELPER_Q(helper_pandn, FANDN)
389
SSE_HELPER_Q(helper_por, FOR)
390
SSE_HELPER_Q(helper_pxor, FXOR)
392
SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
393
SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
394
SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
396
SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
397
SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
398
SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
400
SSE_HELPER_W(helper_pmullw, FMULLW)
402
SSE_HELPER_W(helper_pmulhrw, FMULHRW)
404
SSE_HELPER_W(helper_pmulhuw, FMULHUW)
405
SSE_HELPER_W(helper_pmulhw, FMULHW)
407
SSE_HELPER_B(helper_pavgb, FAVG)
408
SSE_HELPER_W(helper_pavgw, FAVG)
410
void glue(helper_pmuludq, SUFFIX) (Reg *d, Reg *s)
412
d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
414
d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
418
void glue(helper_pmaddwd, SUFFIX) (Reg *d, Reg *s)
422
for(i = 0; i < (2 << SHIFT); i++) {
423
d->L(i) = (int16_t)s->W(2*i) * (int16_t)d->W(2*i) +
424
(int16_t)s->W(2*i+1) * (int16_t)d->W(2*i+1);
429
static inline int abs1(int a)
437
void glue(helper_psadbw, SUFFIX) (Reg *d, Reg *s)
442
val += abs1(d->B(0) - s->B(0));
443
val += abs1(d->B(1) - s->B(1));
444
val += abs1(d->B(2) - s->B(2));
445
val += abs1(d->B(3) - s->B(3));
446
val += abs1(d->B(4) - s->B(4));
447
val += abs1(d->B(5) - s->B(5));
448
val += abs1(d->B(6) - s->B(6));
449
val += abs1(d->B(7) - s->B(7));
453
val += abs1(d->B(8) - s->B(8));
454
val += abs1(d->B(9) - s->B(9));
455
val += abs1(d->B(10) - s->B(10));
456
val += abs1(d->B(11) - s->B(11));
457
val += abs1(d->B(12) - s->B(12));
458
val += abs1(d->B(13) - s->B(13));
459
val += abs1(d->B(14) - s->B(14));
460
val += abs1(d->B(15) - s->B(15));
465
void glue(helper_maskmov, SUFFIX) (Reg *d, Reg *s, target_ulong a0)
468
for(i = 0; i < (8 << SHIFT); i++) {
470
stb(a0 + i, d->B(i));
474
void glue(helper_movl_mm_T0, SUFFIX) (Reg *d, uint32_t val)
484
void glue(helper_movq_mm_T0, SUFFIX) (Reg *d, uint64_t val)
494
void glue(helper_pshufw, SUFFIX) (Reg *d, Reg *s, int order)
497
r.W(0) = s->W(order & 3);
498
r.W(1) = s->W((order >> 2) & 3);
499
r.W(2) = s->W((order >> 4) & 3);
500
r.W(3) = s->W((order >> 6) & 3);
504
void helper_shufps(Reg *d, Reg *s, int order)
507
r.L(0) = d->L(order & 3);
508
r.L(1) = d->L((order >> 2) & 3);
509
r.L(2) = s->L((order >> 4) & 3);
510
r.L(3) = s->L((order >> 6) & 3);
514
void helper_shufpd(Reg *d, Reg *s, int order)
517
r.Q(0) = d->Q(order & 1);
518
r.Q(1) = s->Q((order >> 1) & 1);
522
void glue(helper_pshufd, SUFFIX) (Reg *d, Reg *s, int order)
525
r.L(0) = s->L(order & 3);
526
r.L(1) = s->L((order >> 2) & 3);
527
r.L(2) = s->L((order >> 4) & 3);
528
r.L(3) = s->L((order >> 6) & 3);
532
void glue(helper_pshuflw, SUFFIX) (Reg *d, Reg *s, int order)
535
r.W(0) = s->W(order & 3);
536
r.W(1) = s->W((order >> 2) & 3);
537
r.W(2) = s->W((order >> 4) & 3);
538
r.W(3) = s->W((order >> 6) & 3);
543
void glue(helper_pshufhw, SUFFIX) (Reg *d, Reg *s, int order)
547
r.W(4) = s->W(4 + (order & 3));
548
r.W(5) = s->W(4 + ((order >> 2) & 3));
549
r.W(6) = s->W(4 + ((order >> 4) & 3));
550
r.W(7) = s->W(4 + ((order >> 6) & 3));
557
/* XXX: not accurate */
559
#define SSE_HELPER_S(name, F)\
560
void helper_ ## name ## ps (Reg *d, Reg *s)\
562
d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
563
d->XMM_S(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
564
d->XMM_S(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
565
d->XMM_S(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
568
void helper_ ## name ## ss (Reg *d, Reg *s)\
570
d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
572
void helper_ ## name ## pd (Reg *d, Reg *s)\
574
d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
575
d->XMM_D(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
578
void helper_ ## name ## sd (Reg *d, Reg *s)\
580
d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
583
#define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
584
#define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
585
#define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
586
#define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
587
#define FPU_MIN(size, a, b) (a) < (b) ? (a) : (b)
588
#define FPU_MAX(size, a, b) (a) > (b) ? (a) : (b)
589
#define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
591
SSE_HELPER_S(add, FPU_ADD)
592
SSE_HELPER_S(sub, FPU_SUB)
593
SSE_HELPER_S(mul, FPU_MUL)
594
SSE_HELPER_S(div, FPU_DIV)
595
SSE_HELPER_S(min, FPU_MIN)
596
SSE_HELPER_S(max, FPU_MAX)
597
SSE_HELPER_S(sqrt, FPU_SQRT)
600
/* float to float conversions */
601
void helper_cvtps2pd(Reg *d, Reg *s)
606
d->XMM_D(0) = float32_to_float64(s0, &env->sse_status);
607
d->XMM_D(1) = float32_to_float64(s1, &env->sse_status);
610
void helper_cvtpd2ps(Reg *d, Reg *s)
612
d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
613
d->XMM_S(1) = float64_to_float32(s->XMM_D(1), &env->sse_status);
617
void helper_cvtss2sd(Reg *d, Reg *s)
619
d->XMM_D(0) = float32_to_float64(s->XMM_S(0), &env->sse_status);
622
void helper_cvtsd2ss(Reg *d, Reg *s)
624
d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
627
/* integer to float */
628
void helper_cvtdq2ps(Reg *d, Reg *s)
630
d->XMM_S(0) = int32_to_float32(s->XMM_L(0), &env->sse_status);
631
d->XMM_S(1) = int32_to_float32(s->XMM_L(1), &env->sse_status);
632
d->XMM_S(2) = int32_to_float32(s->XMM_L(2), &env->sse_status);
633
d->XMM_S(3) = int32_to_float32(s->XMM_L(3), &env->sse_status);
636
void helper_cvtdq2pd(Reg *d, Reg *s)
639
l0 = (int32_t)s->XMM_L(0);
640
l1 = (int32_t)s->XMM_L(1);
641
d->XMM_D(0) = int32_to_float64(l0, &env->sse_status);
642
d->XMM_D(1) = int32_to_float64(l1, &env->sse_status);
645
void helper_cvtpi2ps(XMMReg *d, MMXReg *s)
647
d->XMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
648
d->XMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
651
void helper_cvtpi2pd(XMMReg *d, MMXReg *s)
653
d->XMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
654
d->XMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
657
void helper_cvtsi2ss(XMMReg *d, uint32_t val)
659
d->XMM_S(0) = int32_to_float32(val, &env->sse_status);
662
void helper_cvtsi2sd(XMMReg *d, uint32_t val)
664
d->XMM_D(0) = int32_to_float64(val, &env->sse_status);
668
void helper_cvtsq2ss(XMMReg *d, uint64_t val)
670
d->XMM_S(0) = int64_to_float32(val, &env->sse_status);
673
void helper_cvtsq2sd(XMMReg *d, uint64_t val)
675
d->XMM_D(0) = int64_to_float64(val, &env->sse_status);
679
/* float to integer */
680
void helper_cvtps2dq(XMMReg *d, XMMReg *s)
682
d->XMM_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
683
d->XMM_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
684
d->XMM_L(2) = float32_to_int32(s->XMM_S(2), &env->sse_status);
685
d->XMM_L(3) = float32_to_int32(s->XMM_S(3), &env->sse_status);
688
void helper_cvtpd2dq(XMMReg *d, XMMReg *s)
690
d->XMM_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
691
d->XMM_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
695
void helper_cvtps2pi(MMXReg *d, XMMReg *s)
697
d->MMX_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
698
d->MMX_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
701
void helper_cvtpd2pi(MMXReg *d, XMMReg *s)
703
d->MMX_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
704
d->MMX_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
707
int32_t helper_cvtss2si(XMMReg *s)
709
return float32_to_int32(s->XMM_S(0), &env->sse_status);
712
int32_t helper_cvtsd2si(XMMReg *s)
714
return float64_to_int32(s->XMM_D(0), &env->sse_status);
718
int64_t helper_cvtss2sq(XMMReg *s)
720
return float32_to_int64(s->XMM_S(0), &env->sse_status);
723
int64_t helper_cvtsd2sq(XMMReg *s)
725
return float64_to_int64(s->XMM_D(0), &env->sse_status);
729
/* float to integer truncated */
730
void helper_cvttps2dq(XMMReg *d, XMMReg *s)
732
d->XMM_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
733
d->XMM_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
734
d->XMM_L(2) = float32_to_int32_round_to_zero(s->XMM_S(2), &env->sse_status);
735
d->XMM_L(3) = float32_to_int32_round_to_zero(s->XMM_S(3), &env->sse_status);
738
void helper_cvttpd2dq(XMMReg *d, XMMReg *s)
740
d->XMM_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
741
d->XMM_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
745
void helper_cvttps2pi(MMXReg *d, XMMReg *s)
747
d->MMX_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
748
d->MMX_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
751
void helper_cvttpd2pi(MMXReg *d, XMMReg *s)
753
d->MMX_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
754
d->MMX_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
757
int32_t helper_cvttss2si(XMMReg *s)
759
return float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
762
int32_t helper_cvttsd2si(XMMReg *s)
764
return float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
768
int64_t helper_cvttss2sq(XMMReg *s)
770
return float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
773
int64_t helper_cvttsd2sq(XMMReg *s)
775
return float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
779
void helper_rsqrtps(XMMReg *d, XMMReg *s)
781
d->XMM_S(0) = float32_div(float32_one,
782
float32_sqrt(s->XMM_S(0), &env->sse_status),
784
d->XMM_S(1) = float32_div(float32_one,
785
float32_sqrt(s->XMM_S(1), &env->sse_status),
787
d->XMM_S(2) = float32_div(float32_one,
788
float32_sqrt(s->XMM_S(2), &env->sse_status),
790
d->XMM_S(3) = float32_div(float32_one,
791
float32_sqrt(s->XMM_S(3), &env->sse_status),
795
void helper_rsqrtss(XMMReg *d, XMMReg *s)
797
d->XMM_S(0) = float32_div(float32_one,
798
float32_sqrt(s->XMM_S(0), &env->sse_status),
802
void helper_rcpps(XMMReg *d, XMMReg *s)
804
d->XMM_S(0) = float32_div(float32_one, s->XMM_S(0), &env->sse_status);
805
d->XMM_S(1) = float32_div(float32_one, s->XMM_S(1), &env->sse_status);
806
d->XMM_S(2) = float32_div(float32_one, s->XMM_S(2), &env->sse_status);
807
d->XMM_S(3) = float32_div(float32_one, s->XMM_S(3), &env->sse_status);
810
void helper_rcpss(XMMReg *d, XMMReg *s)
812
d->XMM_S(0) = float32_div(float32_one, s->XMM_S(0), &env->sse_status);
815
static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
822
mask = (1ULL << len) - 1;
824
return (src >> shift) & mask;
827
void helper_extrq_r(XMMReg *d, XMMReg *s)
829
d->XMM_Q(0) = helper_extrq(d->XMM_Q(0), s->XMM_B(1), s->XMM_B(0));
832
void helper_extrq_i(XMMReg *d, int index, int length)
834
d->XMM_Q(0) = helper_extrq(d->XMM_Q(0), index, length);
837
static inline uint64_t helper_insertq(uint64_t src, int shift, int len)
844
mask = (1ULL << len) - 1;
846
return (src & ~(mask << shift)) | ((src & mask) << shift);
849
void helper_insertq_r(XMMReg *d, XMMReg *s)
851
d->XMM_Q(0) = helper_insertq(s->XMM_Q(0), s->XMM_B(9), s->XMM_B(8));
854
void helper_insertq_i(XMMReg *d, int index, int length)
856
d->XMM_Q(0) = helper_insertq(d->XMM_Q(0), index, length);
859
void helper_haddps(XMMReg *d, XMMReg *s)
862
r.XMM_S(0) = float32_add(d->XMM_S(0), d->XMM_S(1), &env->sse_status);
863
r.XMM_S(1) = float32_add(d->XMM_S(2), d->XMM_S(3), &env->sse_status);
864
r.XMM_S(2) = float32_add(s->XMM_S(0), s->XMM_S(1), &env->sse_status);
865
r.XMM_S(3) = float32_add(s->XMM_S(2), s->XMM_S(3), &env->sse_status);
869
void helper_haddpd(XMMReg *d, XMMReg *s)
872
r.XMM_D(0) = float64_add(d->XMM_D(0), d->XMM_D(1), &env->sse_status);
873
r.XMM_D(1) = float64_add(s->XMM_D(0), s->XMM_D(1), &env->sse_status);
877
void helper_hsubps(XMMReg *d, XMMReg *s)
880
r.XMM_S(0) = float32_sub(d->XMM_S(0), d->XMM_S(1), &env->sse_status);
881
r.XMM_S(1) = float32_sub(d->XMM_S(2), d->XMM_S(3), &env->sse_status);
882
r.XMM_S(2) = float32_sub(s->XMM_S(0), s->XMM_S(1), &env->sse_status);
883
r.XMM_S(3) = float32_sub(s->XMM_S(2), s->XMM_S(3), &env->sse_status);
887
void helper_hsubpd(XMMReg *d, XMMReg *s)
890
r.XMM_D(0) = float64_sub(d->XMM_D(0), d->XMM_D(1), &env->sse_status);
891
r.XMM_D(1) = float64_sub(s->XMM_D(0), s->XMM_D(1), &env->sse_status);
895
void helper_addsubps(XMMReg *d, XMMReg *s)
897
d->XMM_S(0) = float32_sub(d->XMM_S(0), s->XMM_S(0), &env->sse_status);
898
d->XMM_S(1) = float32_add(d->XMM_S(1), s->XMM_S(1), &env->sse_status);
899
d->XMM_S(2) = float32_sub(d->XMM_S(2), s->XMM_S(2), &env->sse_status);
900
d->XMM_S(3) = float32_add(d->XMM_S(3), s->XMM_S(3), &env->sse_status);
903
void helper_addsubpd(XMMReg *d, XMMReg *s)
905
d->XMM_D(0) = float64_sub(d->XMM_D(0), s->XMM_D(0), &env->sse_status);
906
d->XMM_D(1) = float64_add(d->XMM_D(1), s->XMM_D(1), &env->sse_status);
910
#define SSE_HELPER_CMP(name, F)\
911
void helper_ ## name ## ps (Reg *d, Reg *s)\
913
d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
914
d->XMM_L(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
915
d->XMM_L(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
916
d->XMM_L(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
919
void helper_ ## name ## ss (Reg *d, Reg *s)\
921
d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
923
void helper_ ## name ## pd (Reg *d, Reg *s)\
925
d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
926
d->XMM_Q(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
929
void helper_ ## name ## sd (Reg *d, Reg *s)\
931
d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
934
#define FPU_CMPEQ(size, a, b) float ## size ## _eq_quiet(a, b, &env->sse_status) ? -1 : 0
935
#define FPU_CMPLT(size, a, b) float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0
936
#define FPU_CMPLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? -1 : 0
937
#define FPU_CMPUNORD(size, a, b) float ## size ## _unordered_quiet(a, b, &env->sse_status) ? - 1 : 0
938
#define FPU_CMPNEQ(size, a, b) float ## size ## _eq_quiet(a, b, &env->sse_status) ? 0 : -1
939
#define FPU_CMPNLT(size, a, b) float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1
940
#define FPU_CMPNLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? 0 : -1
941
#define FPU_CMPORD(size, a, b) float ## size ## _unordered_quiet(a, b, &env->sse_status) ? 0 : -1
943
SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
944
SSE_HELPER_CMP(cmplt, FPU_CMPLT)
945
SSE_HELPER_CMP(cmple, FPU_CMPLE)
946
SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
947
SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
948
SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
949
SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
950
SSE_HELPER_CMP(cmpord, FPU_CMPORD)
952
static const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
954
void helper_ucomiss(Reg *d, Reg *s)
961
ret = float32_compare_quiet(s0, s1, &env->sse_status);
962
CC_SRC = comis_eflags[ret + 1];
965
void helper_comiss(Reg *d, Reg *s)
972
ret = float32_compare(s0, s1, &env->sse_status);
973
CC_SRC = comis_eflags[ret + 1];
976
void helper_ucomisd(Reg *d, Reg *s)
983
ret = float64_compare_quiet(d0, d1, &env->sse_status);
984
CC_SRC = comis_eflags[ret + 1];
987
void helper_comisd(Reg *d, Reg *s)
994
ret = float64_compare(d0, d1, &env->sse_status);
995
CC_SRC = comis_eflags[ret + 1];
998
uint32_t helper_movmskps(Reg *s)
1001
b0 = s->XMM_L(0) >> 31;
1002
b1 = s->XMM_L(1) >> 31;
1003
b2 = s->XMM_L(2) >> 31;
1004
b3 = s->XMM_L(3) >> 31;
1005
return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
1008
uint32_t helper_movmskpd(Reg *s)
1011
b0 = s->XMM_L(1) >> 31;
1012
b1 = s->XMM_L(3) >> 31;
1013
return b0 | (b1 << 1);
1018
uint32_t glue(helper_pmovmskb, SUFFIX)(Reg *s)
1022
val |= (s->B(0) >> 7);
1023
val |= (s->B(1) >> 6) & 0x02;
1024
val |= (s->B(2) >> 5) & 0x04;
1025
val |= (s->B(3) >> 4) & 0x08;
1026
val |= (s->B(4) >> 3) & 0x10;
1027
val |= (s->B(5) >> 2) & 0x20;
1028
val |= (s->B(6) >> 1) & 0x40;
1029
val |= (s->B(7)) & 0x80;
1031
val |= (s->B(8) << 1) & 0x0100;
1032
val |= (s->B(9) << 2) & 0x0200;
1033
val |= (s->B(10) << 3) & 0x0400;
1034
val |= (s->B(11) << 4) & 0x0800;
1035
val |= (s->B(12) << 5) & 0x1000;
1036
val |= (s->B(13) << 6) & 0x2000;
1037
val |= (s->B(14) << 7) & 0x4000;
1038
val |= (s->B(15) << 8) & 0x8000;
1043
void glue(helper_packsswb, SUFFIX) (Reg *d, Reg *s)
1047
r.B(0) = satsb((int16_t)d->W(0));
1048
r.B(1) = satsb((int16_t)d->W(1));
1049
r.B(2) = satsb((int16_t)d->W(2));
1050
r.B(3) = satsb((int16_t)d->W(3));
1052
r.B(4) = satsb((int16_t)d->W(4));
1053
r.B(5) = satsb((int16_t)d->W(5));
1054
r.B(6) = satsb((int16_t)d->W(6));
1055
r.B(7) = satsb((int16_t)d->W(7));
1057
r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1058
r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1059
r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1060
r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1062
r.B(12) = satsb((int16_t)s->W(4));
1063
r.B(13) = satsb((int16_t)s->W(5));
1064
r.B(14) = satsb((int16_t)s->W(6));
1065
r.B(15) = satsb((int16_t)s->W(7));
1070
void glue(helper_packuswb, SUFFIX) (Reg *d, Reg *s)
1074
r.B(0) = satub((int16_t)d->W(0));
1075
r.B(1) = satub((int16_t)d->W(1));
1076
r.B(2) = satub((int16_t)d->W(2));
1077
r.B(3) = satub((int16_t)d->W(3));
1079
r.B(4) = satub((int16_t)d->W(4));
1080
r.B(5) = satub((int16_t)d->W(5));
1081
r.B(6) = satub((int16_t)d->W(6));
1082
r.B(7) = satub((int16_t)d->W(7));
1084
r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1085
r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1086
r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1087
r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1089
r.B(12) = satub((int16_t)s->W(4));
1090
r.B(13) = satub((int16_t)s->W(5));
1091
r.B(14) = satub((int16_t)s->W(6));
1092
r.B(15) = satub((int16_t)s->W(7));
1097
void glue(helper_packssdw, SUFFIX) (Reg *d, Reg *s)
1101
r.W(0) = satsw(d->L(0));
1102
r.W(1) = satsw(d->L(1));
1104
r.W(2) = satsw(d->L(2));
1105
r.W(3) = satsw(d->L(3));
1107
r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1108
r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1110
r.W(6) = satsw(s->L(2));
1111
r.W(7) = satsw(s->L(3));
1116
#define UNPCK_OP(base_name, base) \
1118
void glue(helper_punpck ## base_name ## bw, SUFFIX) (Reg *d, Reg *s) \
1122
r.B(0) = d->B((base << (SHIFT + 2)) + 0); \
1123
r.B(1) = s->B((base << (SHIFT + 2)) + 0); \
1124
r.B(2) = d->B((base << (SHIFT + 2)) + 1); \
1125
r.B(3) = s->B((base << (SHIFT + 2)) + 1); \
1126
r.B(4) = d->B((base << (SHIFT + 2)) + 2); \
1127
r.B(5) = s->B((base << (SHIFT + 2)) + 2); \
1128
r.B(6) = d->B((base << (SHIFT + 2)) + 3); \
1129
r.B(7) = s->B((base << (SHIFT + 2)) + 3); \
1131
r.B(8) = d->B((base << (SHIFT + 2)) + 4); \
1132
r.B(9) = s->B((base << (SHIFT + 2)) + 4); \
1133
r.B(10) = d->B((base << (SHIFT + 2)) + 5); \
1134
r.B(11) = s->B((base << (SHIFT + 2)) + 5); \
1135
r.B(12) = d->B((base << (SHIFT + 2)) + 6); \
1136
r.B(13) = s->B((base << (SHIFT + 2)) + 6); \
1137
r.B(14) = d->B((base << (SHIFT + 2)) + 7); \
1138
r.B(15) = s->B((base << (SHIFT + 2)) + 7); \
1143
void glue(helper_punpck ## base_name ## wd, SUFFIX) (Reg *d, Reg *s) \
1147
r.W(0) = d->W((base << (SHIFT + 1)) + 0); \
1148
r.W(1) = s->W((base << (SHIFT + 1)) + 0); \
1149
r.W(2) = d->W((base << (SHIFT + 1)) + 1); \
1150
r.W(3) = s->W((base << (SHIFT + 1)) + 1); \
1152
r.W(4) = d->W((base << (SHIFT + 1)) + 2); \
1153
r.W(5) = s->W((base << (SHIFT + 1)) + 2); \
1154
r.W(6) = d->W((base << (SHIFT + 1)) + 3); \
1155
r.W(7) = s->W((base << (SHIFT + 1)) + 3); \
1160
void glue(helper_punpck ## base_name ## dq, SUFFIX) (Reg *d, Reg *s) \
1164
r.L(0) = d->L((base << SHIFT) + 0); \
1165
r.L(1) = s->L((base << SHIFT) + 0); \
1167
r.L(2) = d->L((base << SHIFT) + 1); \
1168
r.L(3) = s->L((base << SHIFT) + 1); \
1174
void glue(helper_punpck ## base_name ## qdq, SUFFIX) (Reg *d, Reg *s) \
1178
r.Q(0) = d->Q(base); \
1179
r.Q(1) = s->Q(base); \
1187
/* 3DNow! float ops */
1189
void helper_pi2fd(MMXReg *d, MMXReg *s)
1191
d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
1192
d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
1195
void helper_pi2fw(MMXReg *d, MMXReg *s)
1197
d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
1198
d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
1201
void helper_pf2id(MMXReg *d, MMXReg *s)
1203
d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
1204
d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
1207
void helper_pf2iw(MMXReg *d, MMXReg *s)
1209
d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status));
1210
d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status));
1213
void helper_pfacc(MMXReg *d, MMXReg *s)
1216
r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1217
r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1221
void helper_pfadd(MMXReg *d, MMXReg *s)
1223
d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1224
d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1227
void helper_pfcmpeq(MMXReg *d, MMXReg *s)
1229
d->MMX_L(0) = float32_eq_quiet(d->MMX_S(0), s->MMX_S(0), &env->mmx_status) ? -1 : 0;
1230
d->MMX_L(1) = float32_eq_quiet(d->MMX_S(1), s->MMX_S(1), &env->mmx_status) ? -1 : 0;
1233
void helper_pfcmpge(MMXReg *d, MMXReg *s)
1235
d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
1236
d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
1239
void helper_pfcmpgt(MMXReg *d, MMXReg *s)
1241
d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
1242
d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
1245
void helper_pfmax(MMXReg *d, MMXReg *s)
1247
if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status))
1248
d->MMX_S(0) = s->MMX_S(0);
1249
if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status))
1250
d->MMX_S(1) = s->MMX_S(1);
1253
void helper_pfmin(MMXReg *d, MMXReg *s)
1255
if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status))
1256
d->MMX_S(0) = s->MMX_S(0);
1257
if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status))
1258
d->MMX_S(1) = s->MMX_S(1);
1261
void helper_pfmul(MMXReg *d, MMXReg *s)
1263
d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1264
d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1267
void helper_pfnacc(MMXReg *d, MMXReg *s)
1270
r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1271
r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1275
void helper_pfpnacc(MMXReg *d, MMXReg *s)
1278
r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1279
r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1283
void helper_pfrcp(MMXReg *d, MMXReg *s)
1285
d->MMX_S(0) = float32_div(float32_one, s->MMX_S(0), &env->mmx_status);
1286
d->MMX_S(1) = d->MMX_S(0);
1289
void helper_pfrsqrt(MMXReg *d, MMXReg *s)
1291
d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
1292
d->MMX_S(1) = float32_div(float32_one,
1293
float32_sqrt(d->MMX_S(1), &env->mmx_status),
1295
d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
1296
d->MMX_L(0) = d->MMX_L(1);
1299
void helper_pfsub(MMXReg *d, MMXReg *s)
1301
d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1302
d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1305
void helper_pfsubr(MMXReg *d, MMXReg *s)
1307
d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
1308
d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
1311
void helper_pswapd(MMXReg *d, MMXReg *s)
1314
r.MMX_L(0) = s->MMX_L(1);
1315
r.MMX_L(1) = s->MMX_L(0);
1320
/* SSSE3 op helpers */
1321
void glue(helper_pshufb, SUFFIX) (Reg *d, Reg *s)
1326
for (i = 0; i < (8 << SHIFT); i++)
1327
r.B(i) = (s->B(i) & 0x80) ? 0 : (d->B(s->B(i) & ((8 << SHIFT) - 1)));
1332
void glue(helper_phaddw, SUFFIX) (Reg *d, Reg *s)
1334
d->W(0) = (int16_t)d->W(0) + (int16_t)d->W(1);
1335
d->W(1) = (int16_t)d->W(2) + (int16_t)d->W(3);
1336
XMM_ONLY(d->W(2) = (int16_t)d->W(4) + (int16_t)d->W(5));
1337
XMM_ONLY(d->W(3) = (int16_t)d->W(6) + (int16_t)d->W(7));
1338
d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) + (int16_t)s->W(1);
1339
d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) + (int16_t)s->W(3);
1340
XMM_ONLY(d->W(6) = (int16_t)s->W(4) + (int16_t)s->W(5));
1341
XMM_ONLY(d->W(7) = (int16_t)s->W(6) + (int16_t)s->W(7));
1344
void glue(helper_phaddd, SUFFIX) (Reg *d, Reg *s)
1346
d->L(0) = (int32_t)d->L(0) + (int32_t)d->L(1);
1347
XMM_ONLY(d->L(1) = (int32_t)d->L(2) + (int32_t)d->L(3));
1348
d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) + (int32_t)s->L(1);
1349
XMM_ONLY(d->L(3) = (int32_t)s->L(2) + (int32_t)s->L(3));
1352
void glue(helper_phaddsw, SUFFIX) (Reg *d, Reg *s)
1354
d->W(0) = satsw((int16_t)d->W(0) + (int16_t)d->W(1));
1355
d->W(1) = satsw((int16_t)d->W(2) + (int16_t)d->W(3));
1356
XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) + (int16_t)d->W(5)));
1357
XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) + (int16_t)d->W(7)));
1358
d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) + (int16_t)s->W(1));
1359
d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) + (int16_t)s->W(3));
1360
XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) + (int16_t)s->W(5)));
1361
XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) + (int16_t)s->W(7)));
1364
void glue(helper_pmaddubsw, SUFFIX) (Reg *d, Reg *s)
1366
d->W(0) = satsw((int8_t)s->B( 0) * (uint8_t)d->B( 0) +
1367
(int8_t)s->B( 1) * (uint8_t)d->B( 1));
1368
d->W(1) = satsw((int8_t)s->B( 2) * (uint8_t)d->B( 2) +
1369
(int8_t)s->B( 3) * (uint8_t)d->B( 3));
1370
d->W(2) = satsw((int8_t)s->B( 4) * (uint8_t)d->B( 4) +
1371
(int8_t)s->B( 5) * (uint8_t)d->B( 5));
1372
d->W(3) = satsw((int8_t)s->B( 6) * (uint8_t)d->B( 6) +
1373
(int8_t)s->B( 7) * (uint8_t)d->B( 7));
1375
d->W(4) = satsw((int8_t)s->B( 8) * (uint8_t)d->B( 8) +
1376
(int8_t)s->B( 9) * (uint8_t)d->B( 9));
1377
d->W(5) = satsw((int8_t)s->B(10) * (uint8_t)d->B(10) +
1378
(int8_t)s->B(11) * (uint8_t)d->B(11));
1379
d->W(6) = satsw((int8_t)s->B(12) * (uint8_t)d->B(12) +
1380
(int8_t)s->B(13) * (uint8_t)d->B(13));
1381
d->W(7) = satsw((int8_t)s->B(14) * (uint8_t)d->B(14) +
1382
(int8_t)s->B(15) * (uint8_t)d->B(15));
1386
void glue(helper_phsubw, SUFFIX) (Reg *d, Reg *s)
1388
d->W(0) = (int16_t)d->W(0) - (int16_t)d->W(1);
1389
d->W(1) = (int16_t)d->W(2) - (int16_t)d->W(3);
1390
XMM_ONLY(d->W(2) = (int16_t)d->W(4) - (int16_t)d->W(5));
1391
XMM_ONLY(d->W(3) = (int16_t)d->W(6) - (int16_t)d->W(7));
1392
d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) - (int16_t)s->W(1);
1393
d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) - (int16_t)s->W(3);
1394
XMM_ONLY(d->W(6) = (int16_t)s->W(4) - (int16_t)s->W(5));
1395
XMM_ONLY(d->W(7) = (int16_t)s->W(6) - (int16_t)s->W(7));
1398
void glue(helper_phsubd, SUFFIX) (Reg *d, Reg *s)
1400
d->L(0) = (int32_t)d->L(0) - (int32_t)d->L(1);
1401
XMM_ONLY(d->L(1) = (int32_t)d->L(2) - (int32_t)d->L(3));
1402
d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) - (int32_t)s->L(1);
1403
XMM_ONLY(d->L(3) = (int32_t)s->L(2) - (int32_t)s->L(3));
1406
void glue(helper_phsubsw, SUFFIX) (Reg *d, Reg *s)
1408
d->W(0) = satsw((int16_t)d->W(0) - (int16_t)d->W(1));
1409
d->W(1) = satsw((int16_t)d->W(2) - (int16_t)d->W(3));
1410
XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) - (int16_t)d->W(5)));
1411
XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) - (int16_t)d->W(7)));
1412
d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) - (int16_t)s->W(1));
1413
d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) - (int16_t)s->W(3));
1414
XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) - (int16_t)s->W(5)));
1415
XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) - (int16_t)s->W(7)));
1418
#define FABSB(_, x) x > INT8_MAX ? -(int8_t ) x : x
1419
#define FABSW(_, x) x > INT16_MAX ? -(int16_t) x : x
1420
#define FABSL(_, x) x > INT32_MAX ? -(int32_t) x : x
1421
SSE_HELPER_B(helper_pabsb, FABSB)
1422
SSE_HELPER_W(helper_pabsw, FABSW)
1423
SSE_HELPER_L(helper_pabsd, FABSL)
1425
#define FMULHRSW(d, s) ((int16_t) d * (int16_t) s + 0x4000) >> 15
1426
SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
1428
#define FSIGNB(d, s) s <= INT8_MAX ? s ? d : 0 : -(int8_t ) d
1429
#define FSIGNW(d, s) s <= INT16_MAX ? s ? d : 0 : -(int16_t) d
1430
#define FSIGNL(d, s) s <= INT32_MAX ? s ? d : 0 : -(int32_t) d
1431
SSE_HELPER_B(helper_psignb, FSIGNB)
1432
SSE_HELPER_W(helper_psignw, FSIGNW)
1433
SSE_HELPER_L(helper_psignd, FSIGNL)
1435
void glue(helper_palignr, SUFFIX) (Reg *d, Reg *s, int32_t shift)
1439
/* XXX could be checked during translation */
1440
if (shift >= (16 << SHIFT)) {
1442
XMM_ONLY(r.Q(1) = 0);
1445
#define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
1447
r.Q(0) = SHR(s->Q(0), shift - 0) |
1448
SHR(d->Q(0), shift - 64);
1450
r.Q(0) = SHR(s->Q(0), shift - 0) |
1451
SHR(s->Q(1), shift - 64) |
1452
SHR(d->Q(0), shift - 128) |
1453
SHR(d->Q(1), shift - 192);
1454
r.Q(1) = SHR(s->Q(0), shift + 64) |
1455
SHR(s->Q(1), shift - 0) |
1456
SHR(d->Q(0), shift - 64) |
1457
SHR(d->Q(1), shift - 128);
1465
#define XMM0 env->xmm_regs[0]
1468
#define SSE_HELPER_V(name, elem, num, F)\
1469
void glue(name, SUFFIX) (Reg *d, Reg *s)\
1471
d->elem(0) = F(d->elem(0), s->elem(0), XMM0.elem(0));\
1472
d->elem(1) = F(d->elem(1), s->elem(1), XMM0.elem(1));\
1474
d->elem(2) = F(d->elem(2), s->elem(2), XMM0.elem(2));\
1475
d->elem(3) = F(d->elem(3), s->elem(3), XMM0.elem(3));\
1477
d->elem(4) = F(d->elem(4), s->elem(4), XMM0.elem(4));\
1478
d->elem(5) = F(d->elem(5), s->elem(5), XMM0.elem(5));\
1479
d->elem(6) = F(d->elem(6), s->elem(6), XMM0.elem(6));\
1480
d->elem(7) = F(d->elem(7), s->elem(7), XMM0.elem(7));\
1482
d->elem(8) = F(d->elem(8), s->elem(8), XMM0.elem(8));\
1483
d->elem(9) = F(d->elem(9), s->elem(9), XMM0.elem(9));\
1484
d->elem(10) = F(d->elem(10), s->elem(10), XMM0.elem(10));\
1485
d->elem(11) = F(d->elem(11), s->elem(11), XMM0.elem(11));\
1486
d->elem(12) = F(d->elem(12), s->elem(12), XMM0.elem(12));\
1487
d->elem(13) = F(d->elem(13), s->elem(13), XMM0.elem(13));\
1488
d->elem(14) = F(d->elem(14), s->elem(14), XMM0.elem(14));\
1489
d->elem(15) = F(d->elem(15), s->elem(15), XMM0.elem(15));\
1495
#define SSE_HELPER_I(name, elem, num, F)\
1496
void glue(name, SUFFIX) (Reg *d, Reg *s, uint32_t imm)\
1498
d->elem(0) = F(d->elem(0), s->elem(0), ((imm >> 0) & 1));\
1499
d->elem(1) = F(d->elem(1), s->elem(1), ((imm >> 1) & 1));\
1501
d->elem(2) = F(d->elem(2), s->elem(2), ((imm >> 2) & 1));\
1502
d->elem(3) = F(d->elem(3), s->elem(3), ((imm >> 3) & 1));\
1504
d->elem(4) = F(d->elem(4), s->elem(4), ((imm >> 4) & 1));\
1505
d->elem(5) = F(d->elem(5), s->elem(5), ((imm >> 5) & 1));\
1506
d->elem(6) = F(d->elem(6), s->elem(6), ((imm >> 6) & 1));\
1507
d->elem(7) = F(d->elem(7), s->elem(7), ((imm >> 7) & 1));\
1509
d->elem(8) = F(d->elem(8), s->elem(8), ((imm >> 8) & 1));\
1510
d->elem(9) = F(d->elem(9), s->elem(9), ((imm >> 9) & 1));\
1511
d->elem(10) = F(d->elem(10), s->elem(10), ((imm >> 10) & 1));\
1512
d->elem(11) = F(d->elem(11), s->elem(11), ((imm >> 11) & 1));\
1513
d->elem(12) = F(d->elem(12), s->elem(12), ((imm >> 12) & 1));\
1514
d->elem(13) = F(d->elem(13), s->elem(13), ((imm >> 13) & 1));\
1515
d->elem(14) = F(d->elem(14), s->elem(14), ((imm >> 14) & 1));\
1516
d->elem(15) = F(d->elem(15), s->elem(15), ((imm >> 15) & 1));\
1522
/* SSE4.1 op helpers */
1523
#define FBLENDVB(d, s, m) (m & 0x80) ? s : d
1524
#define FBLENDVPS(d, s, m) (m & 0x80000000) ? s : d
1525
#define FBLENDVPD(d, s, m) (m & 0x8000000000000000LL) ? s : d
1526
SSE_HELPER_V(helper_pblendvb, B, 16, FBLENDVB)
1527
SSE_HELPER_V(helper_blendvps, L, 4, FBLENDVPS)
1528
SSE_HELPER_V(helper_blendvpd, Q, 2, FBLENDVPD)
1530
void glue(helper_ptest, SUFFIX) (Reg *d, Reg *s)
1532
uint64_t zf = (s->Q(0) & d->Q(0)) | (s->Q(1) & d->Q(1));
1533
uint64_t cf = (s->Q(0) & ~d->Q(0)) | (s->Q(1) & ~d->Q(1));
1535
CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
1538
#define SSE_HELPER_F(name, elem, num, F)\
1539
void glue(name, SUFFIX) (Reg *d, Reg *s)\
1555
SSE_HELPER_F(helper_pmovsxbw, W, 8, (int8_t) s->B)
1556
SSE_HELPER_F(helper_pmovsxbd, L, 4, (int8_t) s->B)
1557
SSE_HELPER_F(helper_pmovsxbq, Q, 2, (int8_t) s->B)
1558
SSE_HELPER_F(helper_pmovsxwd, L, 4, (int16_t) s->W)
1559
SSE_HELPER_F(helper_pmovsxwq, Q, 2, (int16_t) s->W)
1560
SSE_HELPER_F(helper_pmovsxdq, Q, 2, (int32_t) s->L)
1561
SSE_HELPER_F(helper_pmovzxbw, W, 8, s->B)
1562
SSE_HELPER_F(helper_pmovzxbd, L, 4, s->B)
1563
SSE_HELPER_F(helper_pmovzxbq, Q, 2, s->B)
1564
SSE_HELPER_F(helper_pmovzxwd, L, 4, s->W)
1565
SSE_HELPER_F(helper_pmovzxwq, Q, 2, s->W)
1566
SSE_HELPER_F(helper_pmovzxdq, Q, 2, s->L)
1568
void glue(helper_pmuldq, SUFFIX) (Reg *d, Reg *s)
1570
d->Q(0) = (int64_t) (int32_t) d->L(0) * (int32_t) s->L(0);
1571
d->Q(1) = (int64_t) (int32_t) d->L(2) * (int32_t) s->L(2);
1574
#define FCMPEQQ(d, s) d == s ? -1 : 0
1575
SSE_HELPER_Q(helper_pcmpeqq, FCMPEQQ)
1577
void glue(helper_packusdw, SUFFIX) (Reg *d, Reg *s)
1579
d->W(0) = satuw((int32_t) d->L(0));
1580
d->W(1) = satuw((int32_t) d->L(1));
1581
d->W(2) = satuw((int32_t) d->L(2));
1582
d->W(3) = satuw((int32_t) d->L(3));
1583
d->W(4) = satuw((int32_t) s->L(0));
1584
d->W(5) = satuw((int32_t) s->L(1));
1585
d->W(6) = satuw((int32_t) s->L(2));
1586
d->W(7) = satuw((int32_t) s->L(3));
1589
#define FMINSB(d, s) MIN((int8_t) d, (int8_t) s)
1590
#define FMINSD(d, s) MIN((int32_t) d, (int32_t) s)
1591
#define FMAXSB(d, s) MAX((int8_t) d, (int8_t) s)
1592
#define FMAXSD(d, s) MAX((int32_t) d, (int32_t) s)
1593
SSE_HELPER_B(helper_pminsb, FMINSB)
1594
SSE_HELPER_L(helper_pminsd, FMINSD)
1595
SSE_HELPER_W(helper_pminuw, MIN)
1596
SSE_HELPER_L(helper_pminud, MIN)
1597
SSE_HELPER_B(helper_pmaxsb, FMAXSB)
1598
SSE_HELPER_L(helper_pmaxsd, FMAXSD)
1599
SSE_HELPER_W(helper_pmaxuw, MAX)
1600
SSE_HELPER_L(helper_pmaxud, MAX)
1602
#define FMULLD(d, s) (int32_t) d * (int32_t) s
1603
SSE_HELPER_L(helper_pmulld, FMULLD)
1605
void glue(helper_phminposuw, SUFFIX) (Reg *d, Reg *s)
1609
if (s->W(1) < s->W(idx))
1611
if (s->W(2) < s->W(idx))
1613
if (s->W(3) < s->W(idx))
1615
if (s->W(4) < s->W(idx))
1617
if (s->W(5) < s->W(idx))
1619
if (s->W(6) < s->W(idx))
1621
if (s->W(7) < s->W(idx))
1627
d->W(0) = s->W(idx);
1630
void glue(helper_roundps, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1632
signed char prev_rounding_mode;
1634
prev_rounding_mode = env->sse_status.float_rounding_mode;
1635
if (!(mode & (1 << 2)))
1638
set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1641
set_float_rounding_mode(float_round_down, &env->sse_status);
1644
set_float_rounding_mode(float_round_up, &env->sse_status);
1647
set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1651
d->L(0) = float64_round_to_int(s->L(0), &env->sse_status);
1652
d->L(1) = float64_round_to_int(s->L(1), &env->sse_status);
1653
d->L(2) = float64_round_to_int(s->L(2), &env->sse_status);
1654
d->L(3) = float64_round_to_int(s->L(3), &env->sse_status);
1657
if (mode & (1 << 3))
1658
set_float_exception_flags(
1659
get_float_exception_flags(&env->sse_status) &
1660
~float_flag_inexact,
1663
env->sse_status.float_rounding_mode = prev_rounding_mode;
1666
void glue(helper_roundpd, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1668
signed char prev_rounding_mode;
1670
prev_rounding_mode = env->sse_status.float_rounding_mode;
1671
if (!(mode & (1 << 2)))
1674
set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1677
set_float_rounding_mode(float_round_down, &env->sse_status);
1680
set_float_rounding_mode(float_round_up, &env->sse_status);
1683
set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1687
d->Q(0) = float64_round_to_int(s->Q(0), &env->sse_status);
1688
d->Q(1) = float64_round_to_int(s->Q(1), &env->sse_status);
1691
if (mode & (1 << 3))
1692
set_float_exception_flags(
1693
get_float_exception_flags(&env->sse_status) &
1694
~float_flag_inexact,
1697
env->sse_status.float_rounding_mode = prev_rounding_mode;
1700
void glue(helper_roundss, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1702
signed char prev_rounding_mode;
1704
prev_rounding_mode = env->sse_status.float_rounding_mode;
1705
if (!(mode & (1 << 2)))
1708
set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1711
set_float_rounding_mode(float_round_down, &env->sse_status);
1714
set_float_rounding_mode(float_round_up, &env->sse_status);
1717
set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1721
d->L(0) = float64_round_to_int(s->L(0), &env->sse_status);
1724
if (mode & (1 << 3))
1725
set_float_exception_flags(
1726
get_float_exception_flags(&env->sse_status) &
1727
~float_flag_inexact,
1730
env->sse_status.float_rounding_mode = prev_rounding_mode;
1733
void glue(helper_roundsd, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1735
signed char prev_rounding_mode;
1737
prev_rounding_mode = env->sse_status.float_rounding_mode;
1738
if (!(mode & (1 << 2)))
1741
set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1744
set_float_rounding_mode(float_round_down, &env->sse_status);
1747
set_float_rounding_mode(float_round_up, &env->sse_status);
1750
set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1754
d->Q(0) = float64_round_to_int(s->Q(0), &env->sse_status);
1757
if (mode & (1 << 3))
1758
set_float_exception_flags(
1759
get_float_exception_flags(&env->sse_status) &
1760
~float_flag_inexact,
1763
env->sse_status.float_rounding_mode = prev_rounding_mode;
1766
#define FBLENDP(d, s, m) m ? s : d
1767
SSE_HELPER_I(helper_blendps, L, 4, FBLENDP)
1768
SSE_HELPER_I(helper_blendpd, Q, 2, FBLENDP)
1769
SSE_HELPER_I(helper_pblendw, W, 8, FBLENDP)
1771
void glue(helper_dpps, SUFFIX) (Reg *d, Reg *s, uint32_t mask)
1773
float32 iresult = 0 /*float32_zero*/;
1775
if (mask & (1 << 4))
1776
iresult = float32_add(iresult,
1777
float32_mul(d->L(0), s->L(0), &env->sse_status),
1779
if (mask & (1 << 5))
1780
iresult = float32_add(iresult,
1781
float32_mul(d->L(1), s->L(1), &env->sse_status),
1783
if (mask & (1 << 6))
1784
iresult = float32_add(iresult,
1785
float32_mul(d->L(2), s->L(2), &env->sse_status),
1787
if (mask & (1 << 7))
1788
iresult = float32_add(iresult,
1789
float32_mul(d->L(3), s->L(3), &env->sse_status),
1791
d->L(0) = (mask & (1 << 0)) ? iresult : 0 /*float32_zero*/;
1792
d->L(1) = (mask & (1 << 1)) ? iresult : 0 /*float32_zero*/;
1793
d->L(2) = (mask & (1 << 2)) ? iresult : 0 /*float32_zero*/;
1794
d->L(3) = (mask & (1 << 3)) ? iresult : 0 /*float32_zero*/;
1797
void glue(helper_dppd, SUFFIX) (Reg *d, Reg *s, uint32_t mask)
1799
float64 iresult = 0 /*float64_zero*/;
1801
if (mask & (1 << 4))
1802
iresult = float64_add(iresult,
1803
float64_mul(d->Q(0), s->Q(0), &env->sse_status),
1805
if (mask & (1 << 5))
1806
iresult = float64_add(iresult,
1807
float64_mul(d->Q(1), s->Q(1), &env->sse_status),
1809
d->Q(0) = (mask & (1 << 0)) ? iresult : 0 /*float64_zero*/;
1810
d->Q(1) = (mask & (1 << 1)) ? iresult : 0 /*float64_zero*/;
1813
void glue(helper_mpsadbw, SUFFIX) (Reg *d, Reg *s, uint32_t offset)
1815
int s0 = (offset & 3) << 2;
1816
int d0 = (offset & 4) << 0;
1820
for (i = 0; i < 8; i++, d0++) {
1822
r.W(i) += abs1(d->B(d0 + 0) - s->B(s0 + 0));
1823
r.W(i) += abs1(d->B(d0 + 1) - s->B(s0 + 1));
1824
r.W(i) += abs1(d->B(d0 + 2) - s->B(s0 + 2));
1825
r.W(i) += abs1(d->B(d0 + 3) - s->B(s0 + 3));
1831
/* SSE4.2 op helpers */
1832
/* it's unclear whether signed or unsigned */
1833
#define FCMPGTQ(d, s) d > s ? -1 : 0
1834
SSE_HELPER_Q(helper_pcmpgtq, FCMPGTQ)
1836
static inline int pcmp_elen(int reg, uint32_t ctrl)
1840
/* Presence of REX.W is indicated by a bit higher than 7 set */
1842
val = abs1((int64_t) env->regs[reg]);
1844
val = abs1((int32_t) env->regs[reg]);
1856
static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
1861
while (val < 8 && r->W(val))
1864
while (val < 16 && r->B(val))
1870
static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
1872
switch ((ctrl >> 0) & 3) {
1878
return (int8_t) r->B(i);
1881
return (int16_t) r->W(i);
1885
static inline unsigned pcmpxstrx(Reg *d, Reg *s,
1886
int8_t ctrl, int valids, int validd)
1888
unsigned int res = 0;
1891
int upper = (ctrl & 1) ? 7 : 15;
1896
CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
1898
switch ((ctrl >> 2) & 3) {
1900
for (j = valids; j >= 0; j--) {
1902
v = pcmp_val(s, ctrl, j);
1903
for (i = validd; i >= 0; i--)
1904
res |= (v == pcmp_val(d, ctrl, i));
1908
for (j = valids; j >= 0; j--) {
1910
v = pcmp_val(s, ctrl, j);
1911
for (i = ((validd - 1) | 1); i >= 0; i -= 2)
1912
res |= (pcmp_val(d, ctrl, i - 0) <= v &&
1913
pcmp_val(d, ctrl, i - 1) >= v);
1917
res = (2 << (upper - MAX(valids, validd))) - 1;
1918
res <<= MAX(valids, validd) - MIN(valids, validd);
1919
for (i = MIN(valids, validd); i >= 0; i--) {
1921
v = pcmp_val(s, ctrl, i);
1922
res |= (v == pcmp_val(d, ctrl, i));
1926
for (j = valids - validd; j >= 0; j--) {
1929
for (i = MIN(upper - j, validd); i >= 0; i--)
1930
res &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
1935
switch ((ctrl >> 4) & 3) {
1937
res ^= (2 << upper) - 1;
1940
res ^= (2 << valids) - 1;
1952
static inline int rffs1(unsigned int val)
1956
for (hi = sizeof(val) * 4; hi; hi /= 2)
1965
static inline int ffs1(unsigned int val)
1969
for (hi = sizeof(val) * 4; hi; hi /= 2)
1978
void glue(helper_pcmpestri, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
1980
unsigned int res = pcmpxstrx(d, s, ctrl,
1981
pcmp_elen(R_EDX, ctrl),
1982
pcmp_elen(R_EAX, ctrl));
1985
env->regs[R_ECX] = ((ctrl & (1 << 6)) ? rffs1 : ffs1)(res) - 1;
1987
env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
1990
void glue(helper_pcmpestrm, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
1993
unsigned int res = pcmpxstrx(d, s, ctrl,
1994
pcmp_elen(R_EDX, ctrl),
1995
pcmp_elen(R_EAX, ctrl));
1997
if ((ctrl >> 6) & 1) {
1999
for (i = 0; i < 8; i++, res >>= 1) {
2000
d->W(i) = (res & 1) ? ~0 : 0;
2003
for (i = 0; i < 16; i++, res >>= 1) {
2004
d->B(i) = (res & 1) ? ~0 : 0;
2012
void glue(helper_pcmpistri, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
2014
unsigned int res = pcmpxstrx(d, s, ctrl,
2016
pcmp_ilen(d, ctrl));
2019
env->regs[R_ECX] = ((ctrl & (1 << 6)) ? rffs1 : ffs1)(res) - 1;
2021
env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2024
void glue(helper_pcmpistrm, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
2027
unsigned int res = pcmpxstrx(d, s, ctrl,
2029
pcmp_ilen(d, ctrl));
2031
if ((ctrl >> 6) & 1) {
2033
for (i = 0; i < 8; i++, res >>= 1) {
2034
d->W(i) = (res & 1) ? ~0 : 0;
2037
for (i = 0; i < 16; i++, res >>= 1) {
2038
d->B(i) = (res & 1) ? ~0 : 0;
2046
#define CRCPOLY 0x1edc6f41
2047
#define CRCPOLY_BITREV 0x82f63b78
2048
target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
2050
target_ulong crc = (msg & ((target_ulong) -1 >>
2051
(TARGET_LONG_BITS - len))) ^ crc1;
2054
crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
2059
#define POPMASK(i) ((target_ulong) -1 / ((1LL << (1 << i)) + 1))
2060
#define POPCOUNT(n, i) (n & POPMASK(i)) + ((n >> (1 << i)) & POPMASK(i))
2061
target_ulong helper_popcnt(target_ulong n, uint32_t type)
2063
CC_SRC = n ? 0 : CC_Z;
2073
#ifndef TARGET_X86_64
2079
return POPCOUNT(n, 5);