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, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 #define XMM_ONLY(x...)
31 #define XMM_ONLY(x...) x
39 void glue(helper_psrlw, SUFFIX)(Reg *d, Reg *s)
64 void glue(helper_psraw, SUFFIX)(Reg *d, Reg *s)
73 d->W(0) = (int16_t)d->W(0) >> shift;
74 d->W(1) = (int16_t)d->W(1) >> shift;
75 d->W(2) = (int16_t)d->W(2) >> shift;
76 d->W(3) = (int16_t)d->W(3) >> shift;
78 d->W(4) = (int16_t)d->W(4) >> shift;
79 d->W(5) = (int16_t)d->W(5) >> shift;
80 d->W(6) = (int16_t)d->W(6) >> shift;
81 d->W(7) = (int16_t)d->W(7) >> shift;
85 void glue(helper_psllw, SUFFIX)(Reg *d, Reg *s)
110 void glue(helper_psrld, SUFFIX)(Reg *d, Reg *s)
131 void glue(helper_psrad, SUFFIX)(Reg *d, Reg *s)
140 d->L(0) = (int32_t)d->L(0) >> shift;
141 d->L(1) = (int32_t)d->L(1) >> shift;
143 d->L(2) = (int32_t)d->L(2) >> shift;
144 d->L(3) = (int32_t)d->L(3) >> shift;
148 void glue(helper_pslld, SUFFIX)(Reg *d, Reg *s)
169 void glue(helper_psrlq, SUFFIX)(Reg *d, Reg *s)
188 void glue(helper_psllq, SUFFIX)(Reg *d, Reg *s)
208 void glue(helper_psrldq, SUFFIX)(Reg *d, Reg *s)
215 for(i = 0; i < 16 - shift; i++)
216 d->B(i) = d->B(i + shift);
217 for(i = 16 - shift; i < 16; i++)
222 void glue(helper_pslldq, SUFFIX)(Reg *d, Reg *s)
229 for(i = 15; i >= shift; i--)
230 d->B(i) = d->B(i - shift);
231 for(i = 0; i < shift; i++)
237 #define SSE_HELPER_B(name, F)\
238 void glue(name, SUFFIX) (Reg *d, Reg *s)\
240 d->B(0) = F(d->B(0), s->B(0));\
241 d->B(1) = F(d->B(1), s->B(1));\
242 d->B(2) = F(d->B(2), s->B(2));\
243 d->B(3) = F(d->B(3), s->B(3));\
244 d->B(4) = F(d->B(4), s->B(4));\
245 d->B(5) = F(d->B(5), s->B(5));\
246 d->B(6) = F(d->B(6), s->B(6));\
247 d->B(7) = F(d->B(7), s->B(7));\
249 d->B(8) = F(d->B(8), s->B(8));\
250 d->B(9) = F(d->B(9), s->B(9));\
251 d->B(10) = F(d->B(10), s->B(10));\
252 d->B(11) = F(d->B(11), s->B(11));\
253 d->B(12) = F(d->B(12), s->B(12));\
254 d->B(13) = F(d->B(13), s->B(13));\
255 d->B(14) = F(d->B(14), s->B(14));\
256 d->B(15) = F(d->B(15), s->B(15));\
260 #define SSE_HELPER_W(name, F)\
261 void glue(name, SUFFIX) (Reg *d, Reg *s)\
263 d->W(0) = F(d->W(0), s->W(0));\
264 d->W(1) = F(d->W(1), s->W(1));\
265 d->W(2) = F(d->W(2), s->W(2));\
266 d->W(3) = F(d->W(3), s->W(3));\
268 d->W(4) = F(d->W(4), s->W(4));\
269 d->W(5) = F(d->W(5), s->W(5));\
270 d->W(6) = F(d->W(6), s->W(6));\
271 d->W(7) = F(d->W(7), s->W(7));\
275 #define SSE_HELPER_L(name, F)\
276 void glue(name, SUFFIX) (Reg *d, Reg *s)\
278 d->L(0) = F(d->L(0), s->L(0));\
279 d->L(1) = F(d->L(1), s->L(1));\
281 d->L(2) = F(d->L(2), s->L(2));\
282 d->L(3) = F(d->L(3), s->L(3));\
286 #define SSE_HELPER_Q(name, F)\
287 void glue(name, SUFFIX) (Reg *d, Reg *s)\
289 d->Q(0) = F(d->Q(0), s->Q(0));\
291 d->Q(1) = F(d->Q(1), s->Q(1));\
296 static inline int satub(int x)
306 static inline int satuw(int x)
316 static inline int satsb(int x)
326 static inline int satsw(int x)
336 #define FADD(a, b) ((a) + (b))
337 #define FADDUB(a, b) satub((a) + (b))
338 #define FADDUW(a, b) satuw((a) + (b))
339 #define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
340 #define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
342 #define FSUB(a, b) ((a) - (b))
343 #define FSUBUB(a, b) satub((a) - (b))
344 #define FSUBUW(a, b) satuw((a) - (b))
345 #define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
346 #define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
347 #define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
348 #define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
349 #define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
350 #define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
352 #define FAND(a, b) (a) & (b)
353 #define FANDN(a, b) ((~(a)) & (b))
354 #define FOR(a, b) (a) | (b)
355 #define FXOR(a, b) (a) ^ (b)
357 #define FCMPGTB(a, b) (int8_t)(a) > (int8_t)(b) ? -1 : 0
358 #define FCMPGTW(a, b) (int16_t)(a) > (int16_t)(b) ? -1 : 0
359 #define FCMPGTL(a, b) (int32_t)(a) > (int32_t)(b) ? -1 : 0
360 #define FCMPEQ(a, b) (a) == (b) ? -1 : 0
362 #define FMULLW(a, b) (a) * (b)
363 #define FMULHRW(a, b) ((int16_t)(a) * (int16_t)(b) + 0x8000) >> 16
364 #define FMULHUW(a, b) (a) * (b) >> 16
365 #define FMULHW(a, b) (int16_t)(a) * (int16_t)(b) >> 16
367 #define FAVG(a, b) ((a) + (b) + 1) >> 1
370 SSE_HELPER_B(helper_paddb, FADD)
371 SSE_HELPER_W(helper_paddw, FADD)
372 SSE_HELPER_L(helper_paddl, FADD)
373 SSE_HELPER_Q(helper_paddq, FADD)
375 SSE_HELPER_B(helper_psubb, FSUB)
376 SSE_HELPER_W(helper_psubw, FSUB)
377 SSE_HELPER_L(helper_psubl, FSUB)
378 SSE_HELPER_Q(helper_psubq, FSUB)
380 SSE_HELPER_B(helper_paddusb, FADDUB)
381 SSE_HELPER_B(helper_paddsb, FADDSB)
382 SSE_HELPER_B(helper_psubusb, FSUBUB)
383 SSE_HELPER_B(helper_psubsb, FSUBSB)
385 SSE_HELPER_W(helper_paddusw, FADDUW)
386 SSE_HELPER_W(helper_paddsw, FADDSW)
387 SSE_HELPER_W(helper_psubusw, FSUBUW)
388 SSE_HELPER_W(helper_psubsw, FSUBSW)
390 SSE_HELPER_B(helper_pminub, FMINUB)
391 SSE_HELPER_B(helper_pmaxub, FMAXUB)
393 SSE_HELPER_W(helper_pminsw, FMINSW)
394 SSE_HELPER_W(helper_pmaxsw, FMAXSW)
396 SSE_HELPER_Q(helper_pand, FAND)
397 SSE_HELPER_Q(helper_pandn, FANDN)
398 SSE_HELPER_Q(helper_por, FOR)
399 SSE_HELPER_Q(helper_pxor, FXOR)
401 SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
402 SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
403 SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
405 SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
406 SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
407 SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
409 SSE_HELPER_W(helper_pmullw, FMULLW)
411 SSE_HELPER_W(helper_pmulhrw, FMULHRW)
413 SSE_HELPER_W(helper_pmulhuw, FMULHUW)
414 SSE_HELPER_W(helper_pmulhw, FMULHW)
416 SSE_HELPER_B(helper_pavgb, FAVG)
417 SSE_HELPER_W(helper_pavgw, FAVG)
419 void glue(helper_pmuludq, SUFFIX) (Reg *d, Reg *s)
421 d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
423 d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
427 void glue(helper_pmaddwd, SUFFIX) (Reg *d, Reg *s)
431 for(i = 0; i < (2 << SHIFT); i++) {
432 d->L(i) = (int16_t)s->W(2*i) * (int16_t)d->W(2*i) +
433 (int16_t)s->W(2*i+1) * (int16_t)d->W(2*i+1);
439 static inline int abs1(int a)
447 void glue(helper_psadbw, SUFFIX) (Reg *d, Reg *s)
452 val += abs1(d->B(0) - s->B(0));
453 val += abs1(d->B(1) - s->B(1));
454 val += abs1(d->B(2) - s->B(2));
455 val += abs1(d->B(3) - s->B(3));
456 val += abs1(d->B(4) - s->B(4));
457 val += abs1(d->B(5) - s->B(5));
458 val += abs1(d->B(6) - s->B(6));
459 val += abs1(d->B(7) - s->B(7));
463 val += abs1(d->B(8) - s->B(8));
464 val += abs1(d->B(9) - s->B(9));
465 val += abs1(d->B(10) - s->B(10));
466 val += abs1(d->B(11) - s->B(11));
467 val += abs1(d->B(12) - s->B(12));
468 val += abs1(d->B(13) - s->B(13));
469 val += abs1(d->B(14) - s->B(14));
470 val += abs1(d->B(15) - s->B(15));
475 void glue(helper_maskmov, SUFFIX) (Reg *d, Reg *s, target_ulong a0)
478 for(i = 0; i < (8 << SHIFT); i++) {
480 stb(a0 + i, d->B(i));
485 void glue(helper_movl_mm_T0, SUFFIX) (Reg *d, uint32_t val)
495 void glue(helper_movq_mm_T0, SUFFIX) (Reg *d, uint64_t val)
505 void glue(helper_pshufw, SUFFIX) (Reg *d, Reg *s, int order)
508 r.W(0) = s->W(order & 3);
509 r.W(1) = s->W((order >> 2) & 3);
510 r.W(2) = s->W((order >> 4) & 3);
511 r.W(3) = s->W((order >> 6) & 3);
515 void helper_shufps(Reg *d, Reg *s, int order)
518 r.L(0) = d->L(order & 3);
519 r.L(1) = d->L((order >> 2) & 3);
520 r.L(2) = s->L((order >> 4) & 3);
521 r.L(3) = s->L((order >> 6) & 3);
525 void helper_shufpd(Reg *d, Reg *s, int order)
528 r.Q(0) = d->Q(order & 1);
529 r.Q(1) = s->Q((order >> 1) & 1);
533 void glue(helper_pshufd, SUFFIX) (Reg *d, Reg *s, int order)
536 r.L(0) = s->L(order & 3);
537 r.L(1) = s->L((order >> 2) & 3);
538 r.L(2) = s->L((order >> 4) & 3);
539 r.L(3) = s->L((order >> 6) & 3);
543 void glue(helper_pshuflw, SUFFIX) (Reg *d, Reg *s, int order)
546 r.W(0) = s->W(order & 3);
547 r.W(1) = s->W((order >> 2) & 3);
548 r.W(2) = s->W((order >> 4) & 3);
549 r.W(3) = s->W((order >> 6) & 3);
554 void glue(helper_pshufhw, SUFFIX) (Reg *d, Reg *s, int order)
558 r.W(4) = s->W(4 + (order & 3));
559 r.W(5) = s->W(4 + ((order >> 2) & 3));
560 r.W(6) = s->W(4 + ((order >> 4) & 3));
561 r.W(7) = s->W(4 + ((order >> 6) & 3));
568 /* XXX: not accurate */
570 #define SSE_HELPER_S(name, F)\
571 void helper_ ## name ## ps (Reg *d, Reg *s)\
573 d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
574 d->XMM_S(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
575 d->XMM_S(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
576 d->XMM_S(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
579 void helper_ ## name ## ss (Reg *d, Reg *s)\
581 d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
583 void helper_ ## name ## pd (Reg *d, Reg *s)\
585 d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
586 d->XMM_D(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
589 void helper_ ## name ## sd (Reg *d, Reg *s)\
591 d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
594 #define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
595 #define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
596 #define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
597 #define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
598 #define FPU_MIN(size, a, b) (a) < (b) ? (a) : (b)
599 #define FPU_MAX(size, a, b) (a) > (b) ? (a) : (b)
600 #define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
602 SSE_HELPER_S(add, FPU_ADD)
603 SSE_HELPER_S(sub, FPU_SUB)
604 SSE_HELPER_S(mul, FPU_MUL)
605 SSE_HELPER_S(div, FPU_DIV)
606 SSE_HELPER_S(min, FPU_MIN)
607 SSE_HELPER_S(max, FPU_MAX)
608 SSE_HELPER_S(sqrt, FPU_SQRT)
611 /* float to float conversions */
612 void helper_cvtps2pd(Reg *d, Reg *s)
617 d->XMM_D(0) = float32_to_float64(s0, &env->sse_status);
618 d->XMM_D(1) = float32_to_float64(s1, &env->sse_status);
621 void helper_cvtpd2ps(Reg *d, Reg *s)
623 d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
624 d->XMM_S(1) = float64_to_float32(s->XMM_D(1), &env->sse_status);
628 void helper_cvtss2sd(Reg *d, Reg *s)
630 d->XMM_D(0) = float32_to_float64(s->XMM_S(0), &env->sse_status);
633 void helper_cvtsd2ss(Reg *d, Reg *s)
635 d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
638 /* integer to float */
639 void helper_cvtdq2ps(Reg *d, Reg *s)
641 d->XMM_S(0) = int32_to_float32(s->XMM_L(0), &env->sse_status);
642 d->XMM_S(1) = int32_to_float32(s->XMM_L(1), &env->sse_status);
643 d->XMM_S(2) = int32_to_float32(s->XMM_L(2), &env->sse_status);
644 d->XMM_S(3) = int32_to_float32(s->XMM_L(3), &env->sse_status);
647 void helper_cvtdq2pd(Reg *d, Reg *s)
650 l0 = (int32_t)s->XMM_L(0);
651 l1 = (int32_t)s->XMM_L(1);
652 d->XMM_D(0) = int32_to_float64(l0, &env->sse_status);
653 d->XMM_D(1) = int32_to_float64(l1, &env->sse_status);
656 void helper_cvtpi2ps(XMMReg *d, MMXReg *s)
658 d->XMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
659 d->XMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
662 void helper_cvtpi2pd(XMMReg *d, MMXReg *s)
664 d->XMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
665 d->XMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
668 void helper_cvtsi2ss(XMMReg *d, uint32_t val)
670 d->XMM_S(0) = int32_to_float32(val, &env->sse_status);
673 void helper_cvtsi2sd(XMMReg *d, uint32_t val)
675 d->XMM_D(0) = int32_to_float64(val, &env->sse_status);
679 void helper_cvtsq2ss(XMMReg *d, uint64_t val)
681 d->XMM_S(0) = int64_to_float32(val, &env->sse_status);
684 void helper_cvtsq2sd(XMMReg *d, uint64_t val)
686 d->XMM_D(0) = int64_to_float64(val, &env->sse_status);
690 /* float to integer */
691 void helper_cvtps2dq(XMMReg *d, XMMReg *s)
693 d->XMM_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
694 d->XMM_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
695 d->XMM_L(2) = float32_to_int32(s->XMM_S(2), &env->sse_status);
696 d->XMM_L(3) = float32_to_int32(s->XMM_S(3), &env->sse_status);
699 void helper_cvtpd2dq(XMMReg *d, XMMReg *s)
701 d->XMM_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
702 d->XMM_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
706 void helper_cvtps2pi(MMXReg *d, XMMReg *s)
708 d->MMX_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
709 d->MMX_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
712 void helper_cvtpd2pi(MMXReg *d, XMMReg *s)
714 d->MMX_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
715 d->MMX_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
718 int32_t helper_cvtss2si(XMMReg *s)
720 return float32_to_int32(s->XMM_S(0), &env->sse_status);
723 int32_t helper_cvtsd2si(XMMReg *s)
725 return float64_to_int32(s->XMM_D(0), &env->sse_status);
729 int64_t helper_cvtss2sq(XMMReg *s)
731 return float32_to_int64(s->XMM_S(0), &env->sse_status);
734 int64_t helper_cvtsd2sq(XMMReg *s)
736 return float64_to_int64(s->XMM_D(0), &env->sse_status);
740 /* float to integer truncated */
741 void helper_cvttps2dq(XMMReg *d, XMMReg *s)
743 d->XMM_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
744 d->XMM_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
745 d->XMM_L(2) = float32_to_int32_round_to_zero(s->XMM_S(2), &env->sse_status);
746 d->XMM_L(3) = float32_to_int32_round_to_zero(s->XMM_S(3), &env->sse_status);
749 void helper_cvttpd2dq(XMMReg *d, XMMReg *s)
751 d->XMM_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
752 d->XMM_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
756 void helper_cvttps2pi(MMXReg *d, XMMReg *s)
758 d->MMX_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
759 d->MMX_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
762 void helper_cvttpd2pi(MMXReg *d, XMMReg *s)
764 d->MMX_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
765 d->MMX_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
768 int32_t helper_cvttss2si(XMMReg *s)
770 return float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
773 int32_t helper_cvttsd2si(XMMReg *s)
775 return float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
779 int64_t helper_cvttss2sq(XMMReg *s)
781 return float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
784 int64_t helper_cvttsd2sq(XMMReg *s)
786 return float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
790 void helper_rsqrtps(XMMReg *d, XMMReg *s)
792 d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
793 d->XMM_S(1) = approx_rsqrt(s->XMM_S(1));
794 d->XMM_S(2) = approx_rsqrt(s->XMM_S(2));
795 d->XMM_S(3) = approx_rsqrt(s->XMM_S(3));
798 void helper_rsqrtss(XMMReg *d, XMMReg *s)
800 d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
803 void helper_rcpps(XMMReg *d, XMMReg *s)
805 d->XMM_S(0) = approx_rcp(s->XMM_S(0));
806 d->XMM_S(1) = approx_rcp(s->XMM_S(1));
807 d->XMM_S(2) = approx_rcp(s->XMM_S(2));
808 d->XMM_S(3) = approx_rcp(s->XMM_S(3));
811 void helper_rcpss(XMMReg *d, XMMReg *s)
813 d->XMM_S(0) = approx_rcp(s->XMM_S(0));
816 void helper_haddps(XMMReg *d, XMMReg *s)
819 r.XMM_S(0) = d->XMM_S(0) + d->XMM_S(1);
820 r.XMM_S(1) = d->XMM_S(2) + d->XMM_S(3);
821 r.XMM_S(2) = s->XMM_S(0) + s->XMM_S(1);
822 r.XMM_S(3) = s->XMM_S(2) + s->XMM_S(3);
826 void helper_haddpd(XMMReg *d, XMMReg *s)
829 r.XMM_D(0) = d->XMM_D(0) + d->XMM_D(1);
830 r.XMM_D(1) = s->XMM_D(0) + s->XMM_D(1);
834 void helper_hsubps(XMMReg *d, XMMReg *s)
837 r.XMM_S(0) = d->XMM_S(0) - d->XMM_S(1);
838 r.XMM_S(1) = d->XMM_S(2) - d->XMM_S(3);
839 r.XMM_S(2) = s->XMM_S(0) - s->XMM_S(1);
840 r.XMM_S(3) = s->XMM_S(2) - s->XMM_S(3);
844 void helper_hsubpd(XMMReg *d, XMMReg *s)
847 r.XMM_D(0) = d->XMM_D(0) - d->XMM_D(1);
848 r.XMM_D(1) = s->XMM_D(0) - s->XMM_D(1);
852 void helper_addsubps(XMMReg *d, XMMReg *s)
854 d->XMM_S(0) = d->XMM_S(0) - s->XMM_S(0);
855 d->XMM_S(1) = d->XMM_S(1) + s->XMM_S(1);
856 d->XMM_S(2) = d->XMM_S(2) - s->XMM_S(2);
857 d->XMM_S(3) = d->XMM_S(3) + s->XMM_S(3);
860 void helper_addsubpd(XMMReg *d, XMMReg *s)
862 d->XMM_D(0) = d->XMM_D(0) - s->XMM_D(0);
863 d->XMM_D(1) = d->XMM_D(1) + s->XMM_D(1);
867 #define SSE_HELPER_CMP(name, F)\
868 void helper_ ## name ## ps (Reg *d, Reg *s)\
870 d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
871 d->XMM_L(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
872 d->XMM_L(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
873 d->XMM_L(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
876 void helper_ ## name ## ss (Reg *d, Reg *s)\
878 d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
880 void helper_ ## name ## pd (Reg *d, Reg *s)\
882 d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
883 d->XMM_Q(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
886 void helper_ ## name ## sd (Reg *d, Reg *s)\
888 d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
891 #define FPU_CMPEQ(size, a, b) float ## size ## _eq(a, b, &env->sse_status) ? -1 : 0
892 #define FPU_CMPLT(size, a, b) float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0
893 #define FPU_CMPLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? -1 : 0
894 #define FPU_CMPUNORD(size, a, b) float ## size ## _unordered(a, b, &env->sse_status) ? - 1 : 0
895 #define FPU_CMPNEQ(size, a, b) float ## size ## _eq(a, b, &env->sse_status) ? 0 : -1
896 #define FPU_CMPNLT(size, a, b) float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1
897 #define FPU_CMPNLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? 0 : -1
898 #define FPU_CMPORD(size, a, b) float ## size ## _unordered(a, b, &env->sse_status) ? 0 : -1
900 SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
901 SSE_HELPER_CMP(cmplt, FPU_CMPLT)
902 SSE_HELPER_CMP(cmple, FPU_CMPLE)
903 SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
904 SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
905 SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
906 SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
907 SSE_HELPER_CMP(cmpord, FPU_CMPORD)
909 const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
911 void helper_ucomiss(Reg *d, Reg *s)
918 ret = float32_compare_quiet(s0, s1, &env->sse_status);
919 CC_SRC = comis_eflags[ret + 1];
923 void helper_comiss(Reg *d, Reg *s)
930 ret = float32_compare(s0, s1, &env->sse_status);
931 CC_SRC = comis_eflags[ret + 1];
935 void helper_ucomisd(Reg *d, Reg *s)
942 ret = float64_compare_quiet(d0, d1, &env->sse_status);
943 CC_SRC = comis_eflags[ret + 1];
947 void helper_comisd(Reg *d, Reg *s)
954 ret = float64_compare(d0, d1, &env->sse_status);
955 CC_SRC = comis_eflags[ret + 1];
959 uint32_t helper_movmskps(Reg *s)
962 b0 = s->XMM_L(0) >> 31;
963 b1 = s->XMM_L(1) >> 31;
964 b2 = s->XMM_L(2) >> 31;
965 b3 = s->XMM_L(3) >> 31;
966 return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
969 uint32_t helper_movmskpd(Reg *s)
972 b0 = s->XMM_L(1) >> 31;
973 b1 = s->XMM_L(3) >> 31;
974 return b0 | (b1 << 1);
979 uint32_t glue(helper_pmovmskb, SUFFIX)(Reg *s)
983 val |= (s->B(0) >> 7);
984 val |= (s->B(1) >> 6) & 0x02;
985 val |= (s->B(2) >> 5) & 0x04;
986 val |= (s->B(3) >> 4) & 0x08;
987 val |= (s->B(4) >> 3) & 0x10;
988 val |= (s->B(5) >> 2) & 0x20;
989 val |= (s->B(6) >> 1) & 0x40;
990 val |= (s->B(7)) & 0x80;
992 val |= (s->B(8) << 1) & 0x0100;
993 val |= (s->B(9) << 2) & 0x0200;
994 val |= (s->B(10) << 3) & 0x0400;
995 val |= (s->B(11) << 4) & 0x0800;
996 val |= (s->B(12) << 5) & 0x1000;
997 val |= (s->B(13) << 6) & 0x2000;
998 val |= (s->B(14) << 7) & 0x4000;
999 val |= (s->B(15) << 8) & 0x8000;
1004 void glue(helper_packsswb, SUFFIX) (Reg *d, Reg *s)
1008 r.B(0) = satsb((int16_t)d->W(0));
1009 r.B(1) = satsb((int16_t)d->W(1));
1010 r.B(2) = satsb((int16_t)d->W(2));
1011 r.B(3) = satsb((int16_t)d->W(3));
1013 r.B(4) = satsb((int16_t)d->W(4));
1014 r.B(5) = satsb((int16_t)d->W(5));
1015 r.B(6) = satsb((int16_t)d->W(6));
1016 r.B(7) = satsb((int16_t)d->W(7));
1018 r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1019 r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1020 r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1021 r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1023 r.B(12) = satsb((int16_t)s->W(4));
1024 r.B(13) = satsb((int16_t)s->W(5));
1025 r.B(14) = satsb((int16_t)s->W(6));
1026 r.B(15) = satsb((int16_t)s->W(7));
1031 void glue(helper_packuswb, SUFFIX) (Reg *d, Reg *s)
1035 r.B(0) = satub((int16_t)d->W(0));
1036 r.B(1) = satub((int16_t)d->W(1));
1037 r.B(2) = satub((int16_t)d->W(2));
1038 r.B(3) = satub((int16_t)d->W(3));
1040 r.B(4) = satub((int16_t)d->W(4));
1041 r.B(5) = satub((int16_t)d->W(5));
1042 r.B(6) = satub((int16_t)d->W(6));
1043 r.B(7) = satub((int16_t)d->W(7));
1045 r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1046 r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1047 r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1048 r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1050 r.B(12) = satub((int16_t)s->W(4));
1051 r.B(13) = satub((int16_t)s->W(5));
1052 r.B(14) = satub((int16_t)s->W(6));
1053 r.B(15) = satub((int16_t)s->W(7));
1058 void glue(helper_packssdw, SUFFIX) (Reg *d, Reg *s)
1062 r.W(0) = satsw(d->L(0));
1063 r.W(1) = satsw(d->L(1));
1065 r.W(2) = satsw(d->L(2));
1066 r.W(3) = satsw(d->L(3));
1068 r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1069 r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1071 r.W(6) = satsw(s->L(2));
1072 r.W(7) = satsw(s->L(3));
1077 #define UNPCK_OP(base_name, base) \
1079 void glue(helper_punpck ## base_name ## bw, SUFFIX) (Reg *d, Reg *s) \
1083 r.B(0) = d->B((base << (SHIFT + 2)) + 0); \
1084 r.B(1) = s->B((base << (SHIFT + 2)) + 0); \
1085 r.B(2) = d->B((base << (SHIFT + 2)) + 1); \
1086 r.B(3) = s->B((base << (SHIFT + 2)) + 1); \
1087 r.B(4) = d->B((base << (SHIFT + 2)) + 2); \
1088 r.B(5) = s->B((base << (SHIFT + 2)) + 2); \
1089 r.B(6) = d->B((base << (SHIFT + 2)) + 3); \
1090 r.B(7) = s->B((base << (SHIFT + 2)) + 3); \
1092 r.B(8) = d->B((base << (SHIFT + 2)) + 4); \
1093 r.B(9) = s->B((base << (SHIFT + 2)) + 4); \
1094 r.B(10) = d->B((base << (SHIFT + 2)) + 5); \
1095 r.B(11) = s->B((base << (SHIFT + 2)) + 5); \
1096 r.B(12) = d->B((base << (SHIFT + 2)) + 6); \
1097 r.B(13) = s->B((base << (SHIFT + 2)) + 6); \
1098 r.B(14) = d->B((base << (SHIFT + 2)) + 7); \
1099 r.B(15) = s->B((base << (SHIFT + 2)) + 7); \
1104 void glue(helper_punpck ## base_name ## wd, SUFFIX) (Reg *d, Reg *s) \
1108 r.W(0) = d->W((base << (SHIFT + 1)) + 0); \
1109 r.W(1) = s->W((base << (SHIFT + 1)) + 0); \
1110 r.W(2) = d->W((base << (SHIFT + 1)) + 1); \
1111 r.W(3) = s->W((base << (SHIFT + 1)) + 1); \
1113 r.W(4) = d->W((base << (SHIFT + 1)) + 2); \
1114 r.W(5) = s->W((base << (SHIFT + 1)) + 2); \
1115 r.W(6) = d->W((base << (SHIFT + 1)) + 3); \
1116 r.W(7) = s->W((base << (SHIFT + 1)) + 3); \
1121 void glue(helper_punpck ## base_name ## dq, SUFFIX) (Reg *d, Reg *s) \
1125 r.L(0) = d->L((base << SHIFT) + 0); \
1126 r.L(1) = s->L((base << SHIFT) + 0); \
1128 r.L(2) = d->L((base << SHIFT) + 1); \
1129 r.L(3) = s->L((base << SHIFT) + 1); \
1135 void glue(helper_punpck ## base_name ## qdq, SUFFIX) (Reg *d, Reg *s) \
1139 r.Q(0) = d->Q(base); \
1140 r.Q(1) = s->Q(base); \
1148 /* 3DNow! float ops */
1150 void helper_pi2fd(MMXReg *d, MMXReg *s)
1152 d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
1153 d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
1156 void helper_pi2fw(MMXReg *d, MMXReg *s)
1158 d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
1159 d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
1162 void helper_pf2id(MMXReg *d, MMXReg *s)
1164 d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
1165 d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
1168 void helper_pf2iw(MMXReg *d, MMXReg *s)
1170 d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status));
1171 d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status));
1174 void helper_pfacc(MMXReg *d, MMXReg *s)
1177 r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1178 r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1182 void helper_pfadd(MMXReg *d, MMXReg *s)
1184 d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1185 d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1188 void helper_pfcmpeq(MMXReg *d, MMXReg *s)
1190 d->MMX_L(0) = float32_eq(d->MMX_S(0), s->MMX_S(0), &env->mmx_status) ? -1 : 0;
1191 d->MMX_L(1) = float32_eq(d->MMX_S(1), s->MMX_S(1), &env->mmx_status) ? -1 : 0;
1194 void helper_pfcmpge(MMXReg *d, MMXReg *s)
1196 d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
1197 d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
1200 void helper_pfcmpgt(MMXReg *d, MMXReg *s)
1202 d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
1203 d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
1206 void helper_pfmax(MMXReg *d, MMXReg *s)
1208 if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status))
1209 d->MMX_S(0) = s->MMX_S(0);
1210 if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status))
1211 d->MMX_S(1) = s->MMX_S(1);
1214 void helper_pfmin(MMXReg *d, MMXReg *s)
1216 if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status))
1217 d->MMX_S(0) = s->MMX_S(0);
1218 if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status))
1219 d->MMX_S(1) = s->MMX_S(1);
1222 void helper_pfmul(MMXReg *d, MMXReg *s)
1224 d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1225 d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1228 void helper_pfnacc(MMXReg *d, MMXReg *s)
1231 r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1232 r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1236 void helper_pfpnacc(MMXReg *d, MMXReg *s)
1239 r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1240 r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1244 void helper_pfrcp(MMXReg *d, MMXReg *s)
1246 d->MMX_S(0) = approx_rcp(s->MMX_S(0));
1247 d->MMX_S(1) = d->MMX_S(0);
1250 void helper_pfrsqrt(MMXReg *d, MMXReg *s)
1252 d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
1253 d->MMX_S(1) = approx_rsqrt(d->MMX_S(1));
1254 d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
1255 d->MMX_L(0) = d->MMX_L(1);
1258 void helper_pfsub(MMXReg *d, MMXReg *s)
1260 d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1261 d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1264 void helper_pfsubr(MMXReg *d, MMXReg *s)
1266 d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
1267 d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
1270 void helper_pswapd(MMXReg *d, MMXReg *s)
1273 r.MMX_L(0) = s->MMX_L(1);
1274 r.MMX_L(1) = s->MMX_L(0);
1279 /* SSSE3 op helpers */
1280 void glue(helper_pshufb, SUFFIX) (Reg *d, Reg *s)
1285 for (i = 0; i < (8 << SHIFT); i++)
1286 r.B(i) = (s->B(i) & 0x80) ? 0 : (d->B(s->B(i) & ((8 << SHIFT) - 1)));
1291 void glue(helper_phaddw, SUFFIX) (Reg *d, Reg *s)
1293 d->W(0) = (int16_t)d->W(0) + (int16_t)d->W(1);
1294 d->W(1) = (int16_t)d->W(2) + (int16_t)d->W(3);
1295 XMM_ONLY(d->W(2) = (int16_t)d->W(4) + (int16_t)d->W(5));
1296 XMM_ONLY(d->W(3) = (int16_t)d->W(6) + (int16_t)d->W(7));
1297 d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) + (int16_t)s->W(1);
1298 d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) + (int16_t)s->W(3);
1299 XMM_ONLY(d->W(6) = (int16_t)s->W(4) + (int16_t)s->W(5));
1300 XMM_ONLY(d->W(7) = (int16_t)s->W(6) + (int16_t)s->W(7));
1303 void glue(helper_phaddd, SUFFIX) (Reg *d, Reg *s)
1305 d->L(0) = (int32_t)d->L(0) + (int32_t)d->L(1);
1306 XMM_ONLY(d->L(1) = (int32_t)d->L(2) + (int32_t)d->L(3));
1307 d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) + (int32_t)s->L(1);
1308 XMM_ONLY(d->L(3) = (int32_t)s->L(2) + (int32_t)s->L(3));
1311 void glue(helper_phaddsw, SUFFIX) (Reg *d, Reg *s)
1313 d->W(0) = satsw((int16_t)d->W(0) + (int16_t)d->W(1));
1314 d->W(1) = satsw((int16_t)d->W(2) + (int16_t)d->W(3));
1315 XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) + (int16_t)d->W(5)));
1316 XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) + (int16_t)d->W(7)));
1317 d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) + (int16_t)s->W(1));
1318 d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) + (int16_t)s->W(3));
1319 XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) + (int16_t)s->W(5)));
1320 XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) + (int16_t)s->W(7)));
1323 void glue(helper_pmaddubsw, SUFFIX) (Reg *d, Reg *s)
1325 d->W(0) = satsw((int8_t)s->B( 0) * (uint8_t)d->B( 0) +
1326 (int8_t)s->B( 1) * (uint8_t)d->B( 1));
1327 d->W(1) = satsw((int8_t)s->B( 2) * (uint8_t)d->B( 2) +
1328 (int8_t)s->B( 3) * (uint8_t)d->B( 3));
1329 d->W(2) = satsw((int8_t)s->B( 4) * (uint8_t)d->B( 4) +
1330 (int8_t)s->B( 5) * (uint8_t)d->B( 5));
1331 d->W(3) = satsw((int8_t)s->B( 6) * (uint8_t)d->B( 6) +
1332 (int8_t)s->B( 7) * (uint8_t)d->B( 7));
1334 d->W(4) = satsw((int8_t)s->B( 8) * (uint8_t)d->B( 8) +
1335 (int8_t)s->B( 9) * (uint8_t)d->B( 9));
1336 d->W(5) = satsw((int8_t)s->B(10) * (uint8_t)d->B(10) +
1337 (int8_t)s->B(11) * (uint8_t)d->B(11));
1338 d->W(6) = satsw((int8_t)s->B(12) * (uint8_t)d->B(12) +
1339 (int8_t)s->B(13) * (uint8_t)d->B(13));
1340 d->W(7) = satsw((int8_t)s->B(14) * (uint8_t)d->B(14) +
1341 (int8_t)s->B(15) * (uint8_t)d->B(15));
1345 void glue(helper_phsubw, SUFFIX) (Reg *d, Reg *s)
1347 d->W(0) = (int16_t)d->W(0) - (int16_t)d->W(1);
1348 d->W(1) = (int16_t)d->W(2) - (int16_t)d->W(3);
1349 XMM_ONLY(d->W(2) = (int16_t)d->W(4) - (int16_t)d->W(5));
1350 XMM_ONLY(d->W(3) = (int16_t)d->W(6) - (int16_t)d->W(7));
1351 d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) - (int16_t)s->W(1);
1352 d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) - (int16_t)s->W(3);
1353 XMM_ONLY(d->W(6) = (int16_t)s->W(4) - (int16_t)s->W(5));
1354 XMM_ONLY(d->W(7) = (int16_t)s->W(6) - (int16_t)s->W(7));
1357 void glue(helper_phsubd, SUFFIX) (Reg *d, Reg *s)
1359 d->L(0) = (int32_t)d->L(0) - (int32_t)d->L(1);
1360 XMM_ONLY(d->L(1) = (int32_t)d->L(2) - (int32_t)d->L(3));
1361 d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) - (int32_t)s->L(1);
1362 XMM_ONLY(d->L(3) = (int32_t)s->L(2) - (int32_t)s->L(3));
1365 void glue(helper_phsubsw, SUFFIX) (Reg *d, Reg *s)
1367 d->W(0) = satsw((int16_t)d->W(0) - (int16_t)d->W(1));
1368 d->W(1) = satsw((int16_t)d->W(2) - (int16_t)d->W(3));
1369 XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) - (int16_t)d->W(5)));
1370 XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) - (int16_t)d->W(7)));
1371 d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) - (int16_t)s->W(1));
1372 d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) - (int16_t)s->W(3));
1373 XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) - (int16_t)s->W(5)));
1374 XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) - (int16_t)s->W(7)));
1377 #define FABSB(_, x) x > INT8_MAX ? -(int8_t ) x : x
1378 #define FABSW(_, x) x > INT16_MAX ? -(int16_t) x : x
1379 #define FABSL(_, x) x > INT32_MAX ? -(int32_t) x : x
1380 SSE_HELPER_B(helper_pabsb, FABSB)
1381 SSE_HELPER_W(helper_pabsw, FABSW)
1382 SSE_HELPER_L(helper_pabsd, FABSL)
1384 #define FMULHRSW(d, s) ((int16_t) d * (int16_t) s + 0x4000) >> 15
1385 SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
1387 #define FSIGNB(d, s) s <= INT8_MAX ? s ? d : 0 : -(int8_t ) d
1388 #define FSIGNW(d, s) s <= INT16_MAX ? s ? d : 0 : -(int16_t) d
1389 #define FSIGNL(d, s) s <= INT32_MAX ? s ? d : 0 : -(int32_t) d
1390 SSE_HELPER_B(helper_psignb, FSIGNB)
1391 SSE_HELPER_W(helper_psignw, FSIGNW)
1392 SSE_HELPER_L(helper_psignd, FSIGNL)
1394 void glue(helper_palignr, SUFFIX) (Reg *d, Reg *s, int32_t shift)
1398 /* XXX could be checked during translation */
1399 if (shift >= (16 << SHIFT)) {
1401 XMM_ONLY(r.Q(1) = 0);
1404 #define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
1406 r.Q(0) = SHR(s->Q(0), shift - 0) |
1407 SHR(d->Q(0), shift - 64);
1409 r.Q(0) = SHR(s->Q(0), shift - 0) |
1410 SHR(s->Q(1), shift - 64) |
1411 SHR(d->Q(0), shift - 128) |
1412 SHR(d->Q(1), shift - 192);
1413 r.Q(1) = SHR(s->Q(0), shift + 64) |
1414 SHR(s->Q(1), shift - 0) |
1415 SHR(d->Q(0), shift - 64) |
1416 SHR(d->Q(1), shift - 128);
1424 #define XMM0 env->xmm_regs[0]
1427 #define SSE_HELPER_V(name, elem, num, F)\
1428 void glue(name, SUFFIX) (Reg *d, Reg *s)\
1430 d->elem(0) = F(d->elem(0), s->elem(0), XMM0.elem(0));\
1431 d->elem(1) = F(d->elem(1), s->elem(1), XMM0.elem(1));\
1433 d->elem(2) = F(d->elem(2), s->elem(2), XMM0.elem(2));\
1434 d->elem(3) = F(d->elem(3), s->elem(3), XMM0.elem(3));\
1436 d->elem(4) = F(d->elem(4), s->elem(4), XMM0.elem(4));\
1437 d->elem(5) = F(d->elem(5), s->elem(5), XMM0.elem(5));\
1438 d->elem(6) = F(d->elem(6), s->elem(6), XMM0.elem(6));\
1439 d->elem(7) = F(d->elem(7), s->elem(7), XMM0.elem(7));\
1441 d->elem(8) = F(d->elem(8), s->elem(8), XMM0.elem(8));\
1442 d->elem(9) = F(d->elem(9), s->elem(9), XMM0.elem(9));\
1443 d->elem(10) = F(d->elem(10), s->elem(10), XMM0.elem(10));\
1444 d->elem(11) = F(d->elem(11), s->elem(11), XMM0.elem(11));\
1445 d->elem(12) = F(d->elem(12), s->elem(12), XMM0.elem(12));\
1446 d->elem(13) = F(d->elem(13), s->elem(13), XMM0.elem(13));\
1447 d->elem(14) = F(d->elem(14), s->elem(14), XMM0.elem(14));\
1448 d->elem(15) = F(d->elem(15), s->elem(15), XMM0.elem(15));\
1454 #define SSE_HELPER_I(name, elem, num, F)\
1455 void glue(name, SUFFIX) (Reg *d, Reg *s, uint32_t imm)\
1457 d->elem(0) = F(d->elem(0), s->elem(0), ((imm >> 0) & 1));\
1458 d->elem(1) = F(d->elem(1), s->elem(1), ((imm >> 1) & 1));\
1460 d->elem(2) = F(d->elem(2), s->elem(2), ((imm >> 2) & 1));\
1461 d->elem(3) = F(d->elem(3), s->elem(3), ((imm >> 3) & 1));\
1463 d->elem(4) = F(d->elem(4), s->elem(4), ((imm >> 4) & 1));\
1464 d->elem(5) = F(d->elem(5), s->elem(5), ((imm >> 5) & 1));\
1465 d->elem(6) = F(d->elem(6), s->elem(6), ((imm >> 6) & 1));\
1466 d->elem(7) = F(d->elem(7), s->elem(7), ((imm >> 7) & 1));\
1468 d->elem(8) = F(d->elem(8), s->elem(8), ((imm >> 8) & 1));\
1469 d->elem(9) = F(d->elem(9), s->elem(9), ((imm >> 9) & 1));\
1470 d->elem(10) = F(d->elem(10), s->elem(10), ((imm >> 10) & 1));\
1471 d->elem(11) = F(d->elem(11), s->elem(11), ((imm >> 11) & 1));\
1472 d->elem(12) = F(d->elem(12), s->elem(12), ((imm >> 12) & 1));\
1473 d->elem(13) = F(d->elem(13), s->elem(13), ((imm >> 13) & 1));\
1474 d->elem(14) = F(d->elem(14), s->elem(14), ((imm >> 14) & 1));\
1475 d->elem(15) = F(d->elem(15), s->elem(15), ((imm >> 15) & 1));\
1481 /* SSE4.1 op helpers */
1482 #define FBLENDVB(d, s, m) (m & 0x80) ? s : d
1483 #define FBLENDVPS(d, s, m) (m & 0x80000000) ? s : d
1484 #define FBLENDVPD(d, s, m) (m & 0x8000000000000000LL) ? s : d
1485 SSE_HELPER_V(helper_pblendvb, B, 16, FBLENDVB)
1486 SSE_HELPER_V(helper_blendvps, L, 4, FBLENDVPS)
1487 SSE_HELPER_V(helper_blendvpd, Q, 2, FBLENDVPD)
1489 void glue(helper_ptest, SUFFIX) (Reg *d, Reg *s)
1491 uint64_t zf = (s->Q(0) & d->Q(0)) | (s->Q(1) & d->Q(1));
1492 uint64_t cf = (s->Q(0) & ~d->Q(0)) | (s->Q(1) & ~d->Q(1));
1494 CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
1497 #define SSE_HELPER_F(name, elem, num, F)\
1498 void glue(name, SUFFIX) (Reg *d, Reg *s)\
1514 SSE_HELPER_F(helper_pmovsxbw, W, 8, (int8_t) s->B)
1515 SSE_HELPER_F(helper_pmovsxbd, L, 4, (int8_t) s->B)
1516 SSE_HELPER_F(helper_pmovsxbq, Q, 2, (int8_t) s->B)
1517 SSE_HELPER_F(helper_pmovsxwd, L, 4, (int16_t) s->W)
1518 SSE_HELPER_F(helper_pmovsxwq, Q, 2, (int16_t) s->W)
1519 SSE_HELPER_F(helper_pmovsxdq, Q, 2, (int32_t) s->L)
1520 SSE_HELPER_F(helper_pmovzxbw, W, 8, s->B)
1521 SSE_HELPER_F(helper_pmovzxbd, L, 4, s->B)
1522 SSE_HELPER_F(helper_pmovzxbq, Q, 2, s->B)
1523 SSE_HELPER_F(helper_pmovzxwd, L, 4, s->W)
1524 SSE_HELPER_F(helper_pmovzxwq, Q, 2, s->W)
1525 SSE_HELPER_F(helper_pmovzxdq, Q, 2, s->L)
1527 void glue(helper_pmuldq, SUFFIX) (Reg *d, Reg *s)
1529 d->Q(0) = (int64_t) (int32_t) d->L(0) * (int32_t) s->L(0);
1530 d->Q(1) = (int64_t) (int32_t) d->L(2) * (int32_t) s->L(2);
1533 #define FCMPEQQ(d, s) d == s ? -1 : 0
1534 SSE_HELPER_Q(helper_pcmpeqq, FCMPEQQ)
1536 void glue(helper_packusdw, SUFFIX) (Reg *d, Reg *s)
1538 d->W(0) = satuw((int32_t) d->L(0));
1539 d->W(1) = satuw((int32_t) d->L(1));
1540 d->W(2) = satuw((int32_t) d->L(2));
1541 d->W(3) = satuw((int32_t) d->L(3));
1542 d->W(4) = satuw((int32_t) s->L(0));
1543 d->W(5) = satuw((int32_t) s->L(1));
1544 d->W(6) = satuw((int32_t) s->L(2));
1545 d->W(7) = satuw((int32_t) s->L(3));
1548 #define FMINSB(d, s) MIN((int8_t) d, (int8_t) s)
1549 #define FMINSD(d, s) MIN((int32_t) d, (int32_t) s)
1550 #define FMAXSB(d, s) MAX((int8_t) d, (int8_t) s)
1551 #define FMAXSD(d, s) MAX((int32_t) d, (int32_t) s)
1552 SSE_HELPER_B(helper_pminsb, FMINSB)
1553 SSE_HELPER_L(helper_pminsd, FMINSD)
1554 SSE_HELPER_W(helper_pminuw, MIN)
1555 SSE_HELPER_L(helper_pminud, MIN)
1556 SSE_HELPER_B(helper_pmaxsb, FMAXSB)
1557 SSE_HELPER_L(helper_pmaxsd, FMAXSD)
1558 SSE_HELPER_W(helper_pmaxuw, MAX)
1559 SSE_HELPER_L(helper_pmaxud, MAX)
1561 #define FMULLD(d, s) (int32_t) d * (int32_t) s
1562 SSE_HELPER_L(helper_pmulld, FMULLD)
1564 void glue(helper_phminposuw, SUFFIX) (Reg *d, Reg *s)
1568 if (s->W(1) < s->W(idx))
1570 if (s->W(2) < s->W(idx))
1572 if (s->W(3) < s->W(idx))
1574 if (s->W(4) < s->W(idx))
1576 if (s->W(5) < s->W(idx))
1578 if (s->W(6) < s->W(idx))
1580 if (s->W(7) < s->W(idx))
1586 d->W(0) = s->W(idx);
1589 void glue(helper_roundps, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1591 signed char prev_rounding_mode;
1593 prev_rounding_mode = env->sse_status.float_rounding_mode;
1594 if (!(mode & (1 << 2)))
1597 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1600 set_float_rounding_mode(float_round_down, &env->sse_status);
1603 set_float_rounding_mode(float_round_up, &env->sse_status);
1606 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1610 d->L(0) = float64_round_to_int(s->L(0), &env->sse_status);
1611 d->L(1) = float64_round_to_int(s->L(1), &env->sse_status);
1612 d->L(2) = float64_round_to_int(s->L(2), &env->sse_status);
1613 d->L(3) = float64_round_to_int(s->L(3), &env->sse_status);
1616 if (mode & (1 << 3))
1617 set_float_exception_flags(
1618 get_float_exception_flags(&env->sse_status) &
1619 ~float_flag_inexact,
1622 env->sse_status.float_rounding_mode = prev_rounding_mode;
1625 void glue(helper_roundpd, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1627 signed char prev_rounding_mode;
1629 prev_rounding_mode = env->sse_status.float_rounding_mode;
1630 if (!(mode & (1 << 2)))
1633 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1636 set_float_rounding_mode(float_round_down, &env->sse_status);
1639 set_float_rounding_mode(float_round_up, &env->sse_status);
1642 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1646 d->Q(0) = float64_round_to_int(s->Q(0), &env->sse_status);
1647 d->Q(1) = float64_round_to_int(s->Q(1), &env->sse_status);
1650 if (mode & (1 << 3))
1651 set_float_exception_flags(
1652 get_float_exception_flags(&env->sse_status) &
1653 ~float_flag_inexact,
1656 env->sse_status.float_rounding_mode = prev_rounding_mode;
1659 void glue(helper_roundss, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1661 signed char prev_rounding_mode;
1663 prev_rounding_mode = env->sse_status.float_rounding_mode;
1664 if (!(mode & (1 << 2)))
1667 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1670 set_float_rounding_mode(float_round_down, &env->sse_status);
1673 set_float_rounding_mode(float_round_up, &env->sse_status);
1676 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1680 d->L(0) = float64_round_to_int(s->L(0), &env->sse_status);
1683 if (mode & (1 << 3))
1684 set_float_exception_flags(
1685 get_float_exception_flags(&env->sse_status) &
1686 ~float_flag_inexact,
1689 env->sse_status.float_rounding_mode = prev_rounding_mode;
1692 void glue(helper_roundsd, SUFFIX) (Reg *d, Reg *s, uint32_t mode)
1694 signed char prev_rounding_mode;
1696 prev_rounding_mode = env->sse_status.float_rounding_mode;
1697 if (!(mode & (1 << 2)))
1700 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1703 set_float_rounding_mode(float_round_down, &env->sse_status);
1706 set_float_rounding_mode(float_round_up, &env->sse_status);
1709 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1713 d->Q(0) = float64_round_to_int(s->Q(0), &env->sse_status);
1716 if (mode & (1 << 3))
1717 set_float_exception_flags(
1718 get_float_exception_flags(&env->sse_status) &
1719 ~float_flag_inexact,
1722 env->sse_status.float_rounding_mode = prev_rounding_mode;
1725 #define FBLENDP(d, s, m) m ? s : d
1726 SSE_HELPER_I(helper_blendps, L, 4, FBLENDP)
1727 SSE_HELPER_I(helper_blendpd, Q, 2, FBLENDP)
1728 SSE_HELPER_I(helper_pblendw, W, 8, FBLENDP)
1730 void glue(helper_dpps, SUFFIX) (Reg *d, Reg *s, uint32_t mask)
1732 float32 iresult = 0 /*float32_zero*/;
1734 if (mask & (1 << 4))
1735 iresult = float32_add(iresult,
1736 float32_mul(d->L(0), s->L(0), &env->sse_status),
1738 if (mask & (1 << 5))
1739 iresult = float32_add(iresult,
1740 float32_mul(d->L(1), s->L(1), &env->sse_status),
1742 if (mask & (1 << 6))
1743 iresult = float32_add(iresult,
1744 float32_mul(d->L(2), s->L(2), &env->sse_status),
1746 if (mask & (1 << 7))
1747 iresult = float32_add(iresult,
1748 float32_mul(d->L(3), s->L(3), &env->sse_status),
1750 d->L(0) = (mask & (1 << 0)) ? iresult : 0 /*float32_zero*/;
1751 d->L(1) = (mask & (1 << 1)) ? iresult : 0 /*float32_zero*/;
1752 d->L(2) = (mask & (1 << 2)) ? iresult : 0 /*float32_zero*/;
1753 d->L(3) = (mask & (1 << 3)) ? iresult : 0 /*float32_zero*/;
1756 void glue(helper_dppd, SUFFIX) (Reg *d, Reg *s, uint32_t mask)
1758 float64 iresult = 0 /*float64_zero*/;
1760 if (mask & (1 << 4))
1761 iresult = float64_add(iresult,
1762 float64_mul(d->Q(0), s->Q(0), &env->sse_status),
1764 if (mask & (1 << 5))
1765 iresult = float64_add(iresult,
1766 float64_mul(d->Q(1), s->Q(1), &env->sse_status),
1768 d->Q(0) = (mask & (1 << 0)) ? iresult : 0 /*float64_zero*/;
1769 d->Q(1) = (mask & (1 << 1)) ? iresult : 0 /*float64_zero*/;
1772 void glue(helper_mpsadbw, SUFFIX) (Reg *d, Reg *s, uint32_t offset)
1774 int s0 = (offset & 3) << 2;
1775 int d0 = (offset & 4) << 0;
1779 for (i = 0; i < 8; i++, d0++) {
1781 r.W(i) += abs1(d->B(d0 + 0) - s->B(s0 + 0));
1782 r.W(i) += abs1(d->B(d0 + 1) - s->B(s0 + 1));
1783 r.W(i) += abs1(d->B(d0 + 2) - s->B(s0 + 2));
1784 r.W(i) += abs1(d->B(d0 + 3) - s->B(s0 + 3));
1790 /* SSE4.2 op helpers */
1791 /* it's unclear whether signed or unsigned */
1792 #define FCMPGTQ(d, s) d > s ? -1 : 0
1793 SSE_HELPER_Q(helper_pcmpgtq, FCMPGTQ)
1795 static inline int pcmp_elen(int reg, uint32_t ctrl)
1799 /* Presence of REX.W is indicated by a bit higher than 7 set */
1801 val = abs1((int64_t) env->regs[reg]);
1803 val = abs1((int32_t) env->regs[reg]);
1815 static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
1820 while (val < 8 && r->W(val))
1823 while (val < 16 && r->B(val))
1829 static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
1831 switch ((ctrl >> 0) & 3) {
1837 return (int8_t) r->B(i);
1840 return (int16_t) r->W(i);
1844 static inline unsigned pcmpxstrx(Reg *d, Reg *s,
1845 int8_t ctrl, int valids, int validd)
1847 unsigned int res = 0;
1850 int upper = (ctrl & 1) ? 7 : 15;
1855 CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
1857 switch ((ctrl >> 2) & 3) {
1859 for (j = valids; j >= 0; j--) {
1861 v = pcmp_val(s, ctrl, j);
1862 for (i = validd; i >= 0; i--)
1863 res |= (v == pcmp_val(d, ctrl, i));
1867 for (j = valids; j >= 0; j--) {
1869 v = pcmp_val(s, ctrl, j);
1870 for (i = ((validd - 1) | 1); i >= 0; i -= 2)
1871 res |= (pcmp_val(d, ctrl, i - 0) <= v &&
1872 pcmp_val(d, ctrl, i - 1) >= v);
1876 res = (2 << (upper - MAX(valids, validd))) - 1;
1877 res <<= MAX(valids, validd) - MIN(valids, validd);
1878 for (i = MIN(valids, validd); i >= 0; i--) {
1880 v = pcmp_val(s, ctrl, i);
1881 res |= (v == pcmp_val(d, ctrl, i));
1885 for (j = valids - validd; j >= 0; j--) {
1888 for (i = MIN(upper - j, validd); i >= 0; i--)
1889 res &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
1894 switch ((ctrl >> 4) & 3) {
1896 res ^= (2 << upper) - 1;
1899 res ^= (2 << valids) - 1;
1911 static inline int rffs1(unsigned int val)
1915 for (hi = sizeof(val) * 4; hi; hi /= 2)
1924 static inline int ffs1(unsigned int val)
1928 for (hi = sizeof(val) * 4; hi; hi /= 2)
1937 void glue(helper_pcmpestri, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
1939 unsigned int res = pcmpxstrx(d, s, ctrl,
1940 pcmp_elen(R_EDX, ctrl),
1941 pcmp_elen(R_EAX, ctrl));
1944 env->regs[R_ECX] = ((ctrl & (1 << 6)) ? rffs1 : ffs1)(res) - 1;
1946 env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
1949 void glue(helper_pcmpestrm, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
1952 unsigned int res = pcmpxstrx(d, s, ctrl,
1953 pcmp_elen(R_EDX, ctrl),
1954 pcmp_elen(R_EAX, ctrl));
1956 if ((ctrl >> 6) & 1) {
1958 for (i = 0; i <= 8; i--, res >>= 1)
1959 d->W(i) = (res & 1) ? ~0 : 0;
1961 for (i = 0; i <= 16; i--, res >>= 1)
1962 d->B(i) = (res & 1) ? ~0 : 0;
1969 void glue(helper_pcmpistri, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
1971 unsigned int res = pcmpxstrx(d, s, ctrl,
1973 pcmp_ilen(d, ctrl));
1976 env->regs[R_ECX] = ((ctrl & (1 << 6)) ? rffs1 : ffs1)(res) - 1;
1978 env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
1981 void glue(helper_pcmpistrm, SUFFIX) (Reg *d, Reg *s, uint32_t ctrl)
1984 unsigned int res = pcmpxstrx(d, s, ctrl,
1986 pcmp_ilen(d, ctrl));
1988 if ((ctrl >> 6) & 1) {
1990 for (i = 0; i <= 8; i--, res >>= 1)
1991 d->W(i) = (res & 1) ? ~0 : 0;
1993 for (i = 0; i <= 16; i--, res >>= 1)
1994 d->B(i) = (res & 1) ? ~0 : 0;
2001 #define CRCPOLY 0x1edc6f41
2002 #define CRCPOLY_BITREV 0x82f63b78
2003 target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
2005 target_ulong crc = (msg & ((target_ulong) -1 >>
2006 (TARGET_LONG_BITS - len))) ^ crc1;
2009 crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
2014 #define POPMASK(i) ((target_ulong) -1 / ((1LL << (1 << i)) + 1))
2015 #define POPCOUNT(n, i) (n & POPMASK(i)) + ((n >> (1 << i)) & POPMASK(i))
2016 target_ulong helper_popcnt(target_ulong n, uint32_t type)
2018 CC_SRC = n ? 0 : CC_Z;
2028 #ifndef TARGET_X86_64
2034 return POPCOUNT(n, 5);