xref: /openbmc/qemu/target/i386/ops_sse.h (revision da278d58)
1 /*
2  *  MMX/3DNow!/SSE/SSE2/SSE3/SSSE3/SSE4/PNI support
3  *
4  *  Copyright (c) 2005 Fabrice Bellard
5  *  Copyright (c) 2008 Intel Corporation  <andrew.zaborowski@intel.com>
6  *
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.
11  *
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.
16  *
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/>.
19  */
20 
21 #include "crypto/aes.h"
22 
23 #if SHIFT == 0
24 #define Reg MMXReg
25 #define XMM_ONLY(...)
26 #define B(n) MMX_B(n)
27 #define W(n) MMX_W(n)
28 #define L(n) MMX_L(n)
29 #define Q(n) MMX_Q(n)
30 #define SUFFIX _mmx
31 #else
32 #define Reg ZMMReg
33 #define XMM_ONLY(...) __VA_ARGS__
34 #define B(n) ZMM_B(n)
35 #define W(n) ZMM_W(n)
36 #define L(n) ZMM_L(n)
37 #define Q(n) ZMM_Q(n)
38 #define SUFFIX _xmm
39 #endif
40 
41 void glue(helper_psrlw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
42 {
43     int shift;
44 
45     if (s->Q(0) > 15) {
46         d->Q(0) = 0;
47 #if SHIFT == 1
48         d->Q(1) = 0;
49 #endif
50     } else {
51         shift = s->B(0);
52         d->W(0) >>= shift;
53         d->W(1) >>= shift;
54         d->W(2) >>= shift;
55         d->W(3) >>= shift;
56 #if SHIFT == 1
57         d->W(4) >>= shift;
58         d->W(5) >>= shift;
59         d->W(6) >>= shift;
60         d->W(7) >>= shift;
61 #endif
62     }
63 }
64 
65 void glue(helper_psraw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
66 {
67     int shift;
68 
69     if (s->Q(0) > 15) {
70         shift = 15;
71     } else {
72         shift = s->B(0);
73     }
74     d->W(0) = (int16_t)d->W(0) >> shift;
75     d->W(1) = (int16_t)d->W(1) >> shift;
76     d->W(2) = (int16_t)d->W(2) >> shift;
77     d->W(3) = (int16_t)d->W(3) >> shift;
78 #if SHIFT == 1
79     d->W(4) = (int16_t)d->W(4) >> shift;
80     d->W(5) = (int16_t)d->W(5) >> shift;
81     d->W(6) = (int16_t)d->W(6) >> shift;
82     d->W(7) = (int16_t)d->W(7) >> shift;
83 #endif
84 }
85 
86 void glue(helper_psllw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
87 {
88     int shift;
89 
90     if (s->Q(0) > 15) {
91         d->Q(0) = 0;
92 #if SHIFT == 1
93         d->Q(1) = 0;
94 #endif
95     } else {
96         shift = s->B(0);
97         d->W(0) <<= shift;
98         d->W(1) <<= shift;
99         d->W(2) <<= shift;
100         d->W(3) <<= shift;
101 #if SHIFT == 1
102         d->W(4) <<= shift;
103         d->W(5) <<= shift;
104         d->W(6) <<= shift;
105         d->W(7) <<= shift;
106 #endif
107     }
108 }
109 
110 void glue(helper_psrld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
111 {
112     int shift;
113 
114     if (s->Q(0) > 31) {
115         d->Q(0) = 0;
116 #if SHIFT == 1
117         d->Q(1) = 0;
118 #endif
119     } else {
120         shift = s->B(0);
121         d->L(0) >>= shift;
122         d->L(1) >>= shift;
123 #if SHIFT == 1
124         d->L(2) >>= shift;
125         d->L(3) >>= shift;
126 #endif
127     }
128 }
129 
130 void glue(helper_psrad, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
131 {
132     int shift;
133 
134     if (s->Q(0) > 31) {
135         shift = 31;
136     } else {
137         shift = s->B(0);
138     }
139     d->L(0) = (int32_t)d->L(0) >> shift;
140     d->L(1) = (int32_t)d->L(1) >> shift;
141 #if SHIFT == 1
142     d->L(2) = (int32_t)d->L(2) >> shift;
143     d->L(3) = (int32_t)d->L(3) >> shift;
144 #endif
145 }
146 
147 void glue(helper_pslld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
148 {
149     int shift;
150 
151     if (s->Q(0) > 31) {
152         d->Q(0) = 0;
153 #if SHIFT == 1
154         d->Q(1) = 0;
155 #endif
156     } else {
157         shift = s->B(0);
158         d->L(0) <<= shift;
159         d->L(1) <<= shift;
160 #if SHIFT == 1
161         d->L(2) <<= shift;
162         d->L(3) <<= shift;
163 #endif
164     }
165 }
166 
167 void glue(helper_psrlq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
168 {
169     int shift;
170 
171     if (s->Q(0) > 63) {
172         d->Q(0) = 0;
173 #if SHIFT == 1
174         d->Q(1) = 0;
175 #endif
176     } else {
177         shift = s->B(0);
178         d->Q(0) >>= shift;
179 #if SHIFT == 1
180         d->Q(1) >>= shift;
181 #endif
182     }
183 }
184 
185 void glue(helper_psllq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
186 {
187     int shift;
188 
189     if (s->Q(0) > 63) {
190         d->Q(0) = 0;
191 #if SHIFT == 1
192         d->Q(1) = 0;
193 #endif
194     } else {
195         shift = s->B(0);
196         d->Q(0) <<= shift;
197 #if SHIFT == 1
198         d->Q(1) <<= shift;
199 #endif
200     }
201 }
202 
203 #if SHIFT == 1
204 void glue(helper_psrldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
205 {
206     int shift, i;
207 
208     shift = s->L(0);
209     if (shift > 16) {
210         shift = 16;
211     }
212     for (i = 0; i < 16 - shift; i++) {
213         d->B(i) = d->B(i + shift);
214     }
215     for (i = 16 - shift; i < 16; i++) {
216         d->B(i) = 0;
217     }
218 }
219 
220 void glue(helper_pslldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
221 {
222     int shift, i;
223 
224     shift = s->L(0);
225     if (shift > 16) {
226         shift = 16;
227     }
228     for (i = 15; i >= shift; i--) {
229         d->B(i) = d->B(i - shift);
230     }
231     for (i = 0; i < shift; i++) {
232         d->B(i) = 0;
233     }
234 }
235 #endif
236 
237 #define SSE_HELPER_B(name, F)                                   \
238     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
239     {                                                           \
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));                          \
248         XMM_ONLY(                                               \
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));              \
257                                                         )       \
258             }
259 
260 #define SSE_HELPER_W(name, F)                                   \
261     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
262     {                                                           \
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));                          \
267         XMM_ONLY(                                               \
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));                 \
272                                                         )       \
273             }
274 
275 #define SSE_HELPER_L(name, F)                                   \
276     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
277     {                                                           \
278         d->L(0) = F(d->L(0), s->L(0));                          \
279         d->L(1) = F(d->L(1), s->L(1));                          \
280         XMM_ONLY(                                               \
281                  d->L(2) = F(d->L(2), s->L(2));                 \
282                  d->L(3) = F(d->L(3), s->L(3));                 \
283                                                         )       \
284             }
285 
286 #define SSE_HELPER_Q(name, F)                                   \
287     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
288     {                                                           \
289         d->Q(0) = F(d->Q(0), s->Q(0));                          \
290         XMM_ONLY(                                               \
291                  d->Q(1) = F(d->Q(1), s->Q(1));                 \
292                                                         )       \
293             }
294 
295 #if SHIFT == 0
296 static inline int satub(int x)
297 {
298     if (x < 0) {
299         return 0;
300     } else if (x > 255) {
301         return 255;
302     } else {
303         return x;
304     }
305 }
306 
307 static inline int satuw(int x)
308 {
309     if (x < 0) {
310         return 0;
311     } else if (x > 65535) {
312         return 65535;
313     } else {
314         return x;
315     }
316 }
317 
318 static inline int satsb(int x)
319 {
320     if (x < -128) {
321         return -128;
322     } else if (x > 127) {
323         return 127;
324     } else {
325         return x;
326     }
327 }
328 
329 static inline int satsw(int x)
330 {
331     if (x < -32768) {
332         return -32768;
333     } else if (x > 32767) {
334         return 32767;
335     } else {
336         return x;
337     }
338 }
339 
340 #define FADD(a, b) ((a) + (b))
341 #define FADDUB(a, b) satub((a) + (b))
342 #define FADDUW(a, b) satuw((a) + (b))
343 #define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
344 #define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
345 
346 #define FSUB(a, b) ((a) - (b))
347 #define FSUBUB(a, b) satub((a) - (b))
348 #define FSUBUW(a, b) satuw((a) - (b))
349 #define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
350 #define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
351 #define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
352 #define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
353 #define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
354 #define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
355 
356 #define FAND(a, b) ((a) & (b))
357 #define FANDN(a, b) ((~(a)) & (b))
358 #define FOR(a, b) ((a) | (b))
359 #define FXOR(a, b) ((a) ^ (b))
360 
361 #define FCMPGTB(a, b) ((int8_t)(a) > (int8_t)(b) ? -1 : 0)
362 #define FCMPGTW(a, b) ((int16_t)(a) > (int16_t)(b) ? -1 : 0)
363 #define FCMPGTL(a, b) ((int32_t)(a) > (int32_t)(b) ? -1 : 0)
364 #define FCMPEQ(a, b) ((a) == (b) ? -1 : 0)
365 
366 #define FMULLW(a, b) ((a) * (b))
367 #define FMULHRW(a, b) (((int16_t)(a) * (int16_t)(b) + 0x8000) >> 16)
368 #define FMULHUW(a, b) ((a) * (b) >> 16)
369 #define FMULHW(a, b) ((int16_t)(a) * (int16_t)(b) >> 16)
370 
371 #define FAVG(a, b) (((a) + (b) + 1) >> 1)
372 #endif
373 
374 SSE_HELPER_B(helper_paddb, FADD)
375 SSE_HELPER_W(helper_paddw, FADD)
376 SSE_HELPER_L(helper_paddl, FADD)
377 SSE_HELPER_Q(helper_paddq, FADD)
378 
379 SSE_HELPER_B(helper_psubb, FSUB)
380 SSE_HELPER_W(helper_psubw, FSUB)
381 SSE_HELPER_L(helper_psubl, FSUB)
382 SSE_HELPER_Q(helper_psubq, FSUB)
383 
384 SSE_HELPER_B(helper_paddusb, FADDUB)
385 SSE_HELPER_B(helper_paddsb, FADDSB)
386 SSE_HELPER_B(helper_psubusb, FSUBUB)
387 SSE_HELPER_B(helper_psubsb, FSUBSB)
388 
389 SSE_HELPER_W(helper_paddusw, FADDUW)
390 SSE_HELPER_W(helper_paddsw, FADDSW)
391 SSE_HELPER_W(helper_psubusw, FSUBUW)
392 SSE_HELPER_W(helper_psubsw, FSUBSW)
393 
394 SSE_HELPER_B(helper_pminub, FMINUB)
395 SSE_HELPER_B(helper_pmaxub, FMAXUB)
396 
397 SSE_HELPER_W(helper_pminsw, FMINSW)
398 SSE_HELPER_W(helper_pmaxsw, FMAXSW)
399 
400 SSE_HELPER_Q(helper_pand, FAND)
401 SSE_HELPER_Q(helper_pandn, FANDN)
402 SSE_HELPER_Q(helper_por, FOR)
403 SSE_HELPER_Q(helper_pxor, FXOR)
404 
405 SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
406 SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
407 SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
408 
409 SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
410 SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
411 SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
412 
413 SSE_HELPER_W(helper_pmullw, FMULLW)
414 #if SHIFT == 0
415 SSE_HELPER_W(helper_pmulhrw, FMULHRW)
416 #endif
417 SSE_HELPER_W(helper_pmulhuw, FMULHUW)
418 SSE_HELPER_W(helper_pmulhw, FMULHW)
419 
420 SSE_HELPER_B(helper_pavgb, FAVG)
421 SSE_HELPER_W(helper_pavgw, FAVG)
422 
423 void glue(helper_pmuludq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
424 {
425     d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
426 #if SHIFT == 1
427     d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
428 #endif
429 }
430 
431 void glue(helper_pmaddwd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
432 {
433     int i;
434 
435     for (i = 0; i < (2 << SHIFT); i++) {
436         d->L(i) = (int16_t)s->W(2 * i) * (int16_t)d->W(2 * i) +
437             (int16_t)s->W(2 * i + 1) * (int16_t)d->W(2 * i + 1);
438     }
439 }
440 
441 #if SHIFT == 0
442 static inline int abs1(int a)
443 {
444     if (a < 0) {
445         return -a;
446     } else {
447         return a;
448     }
449 }
450 #endif
451 void glue(helper_psadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
452 {
453     unsigned int val;
454 
455     val = 0;
456     val += abs1(d->B(0) - s->B(0));
457     val += abs1(d->B(1) - s->B(1));
458     val += abs1(d->B(2) - s->B(2));
459     val += abs1(d->B(3) - s->B(3));
460     val += abs1(d->B(4) - s->B(4));
461     val += abs1(d->B(5) - s->B(5));
462     val += abs1(d->B(6) - s->B(6));
463     val += abs1(d->B(7) - s->B(7));
464     d->Q(0) = val;
465 #if SHIFT == 1
466     val = 0;
467     val += abs1(d->B(8) - s->B(8));
468     val += abs1(d->B(9) - s->B(9));
469     val += abs1(d->B(10) - s->B(10));
470     val += abs1(d->B(11) - s->B(11));
471     val += abs1(d->B(12) - s->B(12));
472     val += abs1(d->B(13) - s->B(13));
473     val += abs1(d->B(14) - s->B(14));
474     val += abs1(d->B(15) - s->B(15));
475     d->Q(1) = val;
476 #endif
477 }
478 
479 void glue(helper_maskmov, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
480                                   target_ulong a0)
481 {
482     int i;
483 
484     for (i = 0; i < (8 << SHIFT); i++) {
485         if (s->B(i) & 0x80) {
486             cpu_stb_data_ra(env, a0 + i, d->B(i), GETPC());
487         }
488     }
489 }
490 
491 void glue(helper_movl_mm_T0, SUFFIX)(Reg *d, uint32_t val)
492 {
493     d->L(0) = val;
494     d->L(1) = 0;
495 #if SHIFT == 1
496     d->Q(1) = 0;
497 #endif
498 }
499 
500 #ifdef TARGET_X86_64
501 void glue(helper_movq_mm_T0, SUFFIX)(Reg *d, uint64_t val)
502 {
503     d->Q(0) = val;
504 #if SHIFT == 1
505     d->Q(1) = 0;
506 #endif
507 }
508 #endif
509 
510 #if SHIFT == 0
511 void glue(helper_pshufw, SUFFIX)(Reg *d, Reg *s, int order)
512 {
513     Reg r;
514 
515     r.W(0) = s->W(order & 3);
516     r.W(1) = s->W((order >> 2) & 3);
517     r.W(2) = s->W((order >> 4) & 3);
518     r.W(3) = s->W((order >> 6) & 3);
519     *d = r;
520 }
521 #else
522 void helper_shufps(Reg *d, Reg *s, int order)
523 {
524     Reg r;
525 
526     r.L(0) = d->L(order & 3);
527     r.L(1) = d->L((order >> 2) & 3);
528     r.L(2) = s->L((order >> 4) & 3);
529     r.L(3) = s->L((order >> 6) & 3);
530     *d = r;
531 }
532 
533 void helper_shufpd(Reg *d, Reg *s, int order)
534 {
535     Reg r;
536 
537     r.Q(0) = d->Q(order & 1);
538     r.Q(1) = s->Q((order >> 1) & 1);
539     *d = r;
540 }
541 
542 void glue(helper_pshufd, SUFFIX)(Reg *d, Reg *s, int order)
543 {
544     Reg r;
545 
546     r.L(0) = s->L(order & 3);
547     r.L(1) = s->L((order >> 2) & 3);
548     r.L(2) = s->L((order >> 4) & 3);
549     r.L(3) = s->L((order >> 6) & 3);
550     *d = r;
551 }
552 
553 void glue(helper_pshuflw, SUFFIX)(Reg *d, Reg *s, int order)
554 {
555     Reg r;
556 
557     r.W(0) = s->W(order & 3);
558     r.W(1) = s->W((order >> 2) & 3);
559     r.W(2) = s->W((order >> 4) & 3);
560     r.W(3) = s->W((order >> 6) & 3);
561     r.Q(1) = s->Q(1);
562     *d = r;
563 }
564 
565 void glue(helper_pshufhw, SUFFIX)(Reg *d, Reg *s, int order)
566 {
567     Reg r;
568 
569     r.Q(0) = s->Q(0);
570     r.W(4) = s->W(4 + (order & 3));
571     r.W(5) = s->W(4 + ((order >> 2) & 3));
572     r.W(6) = s->W(4 + ((order >> 4) & 3));
573     r.W(7) = s->W(4 + ((order >> 6) & 3));
574     *d = r;
575 }
576 #endif
577 
578 #if SHIFT == 1
579 /* FPU ops */
580 /* XXX: not accurate */
581 
582 #define SSE_HELPER_S(name, F)                                           \
583     void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s)        \
584     {                                                                   \
585         d->ZMM_S(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
586         d->ZMM_S(1) = F(32, d->ZMM_S(1), s->ZMM_S(1));                  \
587         d->ZMM_S(2) = F(32, d->ZMM_S(2), s->ZMM_S(2));                  \
588         d->ZMM_S(3) = F(32, d->ZMM_S(3), s->ZMM_S(3));                  \
589     }                                                                   \
590                                                                         \
591     void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s)        \
592     {                                                                   \
593         d->ZMM_S(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
594     }                                                                   \
595                                                                         \
596     void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s)        \
597     {                                                                   \
598         d->ZMM_D(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
599         d->ZMM_D(1) = F(64, d->ZMM_D(1), s->ZMM_D(1));                  \
600     }                                                                   \
601                                                                         \
602     void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s)        \
603     {                                                                   \
604         d->ZMM_D(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
605     }
606 
607 #define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
608 #define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
609 #define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
610 #define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
611 #define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
612 
613 /* Note that the choice of comparison op here is important to get the
614  * special cases right: for min and max Intel specifies that (-0,0),
615  * (NaN, anything) and (anything, NaN) return the second argument.
616  */
617 #define FPU_MIN(size, a, b)                                     \
618     (float ## size ## _lt(a, b, &env->sse_status) ? (a) : (b))
619 #define FPU_MAX(size, a, b)                                     \
620     (float ## size ## _lt(b, a, &env->sse_status) ? (a) : (b))
621 
622 SSE_HELPER_S(add, FPU_ADD)
623 SSE_HELPER_S(sub, FPU_SUB)
624 SSE_HELPER_S(mul, FPU_MUL)
625 SSE_HELPER_S(div, FPU_DIV)
626 SSE_HELPER_S(min, FPU_MIN)
627 SSE_HELPER_S(max, FPU_MAX)
628 SSE_HELPER_S(sqrt, FPU_SQRT)
629 
630 
631 /* float to float conversions */
632 void helper_cvtps2pd(CPUX86State *env, Reg *d, Reg *s)
633 {
634     float32 s0, s1;
635 
636     s0 = s->ZMM_S(0);
637     s1 = s->ZMM_S(1);
638     d->ZMM_D(0) = float32_to_float64(s0, &env->sse_status);
639     d->ZMM_D(1) = float32_to_float64(s1, &env->sse_status);
640 }
641 
642 void helper_cvtpd2ps(CPUX86State *env, Reg *d, Reg *s)
643 {
644     d->ZMM_S(0) = float64_to_float32(s->ZMM_D(0), &env->sse_status);
645     d->ZMM_S(1) = float64_to_float32(s->ZMM_D(1), &env->sse_status);
646     d->Q(1) = 0;
647 }
648 
649 void helper_cvtss2sd(CPUX86State *env, Reg *d, Reg *s)
650 {
651     d->ZMM_D(0) = float32_to_float64(s->ZMM_S(0), &env->sse_status);
652 }
653 
654 void helper_cvtsd2ss(CPUX86State *env, Reg *d, Reg *s)
655 {
656     d->ZMM_S(0) = float64_to_float32(s->ZMM_D(0), &env->sse_status);
657 }
658 
659 /* integer to float */
660 void helper_cvtdq2ps(CPUX86State *env, Reg *d, Reg *s)
661 {
662     d->ZMM_S(0) = int32_to_float32(s->ZMM_L(0), &env->sse_status);
663     d->ZMM_S(1) = int32_to_float32(s->ZMM_L(1), &env->sse_status);
664     d->ZMM_S(2) = int32_to_float32(s->ZMM_L(2), &env->sse_status);
665     d->ZMM_S(3) = int32_to_float32(s->ZMM_L(3), &env->sse_status);
666 }
667 
668 void helper_cvtdq2pd(CPUX86State *env, Reg *d, Reg *s)
669 {
670     int32_t l0, l1;
671 
672     l0 = (int32_t)s->ZMM_L(0);
673     l1 = (int32_t)s->ZMM_L(1);
674     d->ZMM_D(0) = int32_to_float64(l0, &env->sse_status);
675     d->ZMM_D(1) = int32_to_float64(l1, &env->sse_status);
676 }
677 
678 void helper_cvtpi2ps(CPUX86State *env, ZMMReg *d, MMXReg *s)
679 {
680     d->ZMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
681     d->ZMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
682 }
683 
684 void helper_cvtpi2pd(CPUX86State *env, ZMMReg *d, MMXReg *s)
685 {
686     d->ZMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
687     d->ZMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
688 }
689 
690 void helper_cvtsi2ss(CPUX86State *env, ZMMReg *d, uint32_t val)
691 {
692     d->ZMM_S(0) = int32_to_float32(val, &env->sse_status);
693 }
694 
695 void helper_cvtsi2sd(CPUX86State *env, ZMMReg *d, uint32_t val)
696 {
697     d->ZMM_D(0) = int32_to_float64(val, &env->sse_status);
698 }
699 
700 #ifdef TARGET_X86_64
701 void helper_cvtsq2ss(CPUX86State *env, ZMMReg *d, uint64_t val)
702 {
703     d->ZMM_S(0) = int64_to_float32(val, &env->sse_status);
704 }
705 
706 void helper_cvtsq2sd(CPUX86State *env, ZMMReg *d, uint64_t val)
707 {
708     d->ZMM_D(0) = int64_to_float64(val, &env->sse_status);
709 }
710 #endif
711 
712 /* float to integer */
713 
714 /*
715  * x86 mandates that we return the indefinite integer value for the result
716  * of any float-to-integer conversion that raises the 'invalid' exception.
717  * Wrap the softfloat functions to get this behaviour.
718  */
719 #define WRAP_FLOATCONV(RETTYPE, FN, FLOATTYPE, INDEFVALUE)              \
720     static inline RETTYPE x86_##FN(FLOATTYPE a, float_status *s)        \
721     {                                                                   \
722         int oldflags, newflags;                                         \
723         RETTYPE r;                                                      \
724                                                                         \
725         oldflags = get_float_exception_flags(s);                        \
726         set_float_exception_flags(0, s);                                \
727         r = FN(a, s);                                                   \
728         newflags = get_float_exception_flags(s);                        \
729         if (newflags & float_flag_invalid) {                            \
730             r = INDEFVALUE;                                             \
731         }                                                               \
732         set_float_exception_flags(newflags | oldflags, s);              \
733         return r;                                                       \
734     }
735 
736 WRAP_FLOATCONV(int32_t, float32_to_int32, float32, INT32_MIN)
737 WRAP_FLOATCONV(int32_t, float32_to_int32_round_to_zero, float32, INT32_MIN)
738 WRAP_FLOATCONV(int32_t, float64_to_int32, float64, INT32_MIN)
739 WRAP_FLOATCONV(int32_t, float64_to_int32_round_to_zero, float64, INT32_MIN)
740 WRAP_FLOATCONV(int64_t, float32_to_int64, float32, INT64_MIN)
741 WRAP_FLOATCONV(int64_t, float32_to_int64_round_to_zero, float32, INT64_MIN)
742 WRAP_FLOATCONV(int64_t, float64_to_int64, float64, INT64_MIN)
743 WRAP_FLOATCONV(int64_t, float64_to_int64_round_to_zero, float64, INT64_MIN)
744 
745 void helper_cvtps2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
746 {
747     d->ZMM_L(0) = x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
748     d->ZMM_L(1) = x86_float32_to_int32(s->ZMM_S(1), &env->sse_status);
749     d->ZMM_L(2) = x86_float32_to_int32(s->ZMM_S(2), &env->sse_status);
750     d->ZMM_L(3) = x86_float32_to_int32(s->ZMM_S(3), &env->sse_status);
751 }
752 
753 void helper_cvtpd2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
754 {
755     d->ZMM_L(0) = x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
756     d->ZMM_L(1) = x86_float64_to_int32(s->ZMM_D(1), &env->sse_status);
757     d->ZMM_Q(1) = 0;
758 }
759 
760 void helper_cvtps2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
761 {
762     d->MMX_L(0) = x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
763     d->MMX_L(1) = x86_float32_to_int32(s->ZMM_S(1), &env->sse_status);
764 }
765 
766 void helper_cvtpd2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
767 {
768     d->MMX_L(0) = x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
769     d->MMX_L(1) = x86_float64_to_int32(s->ZMM_D(1), &env->sse_status);
770 }
771 
772 int32_t helper_cvtss2si(CPUX86State *env, ZMMReg *s)
773 {
774     return x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
775 }
776 
777 int32_t helper_cvtsd2si(CPUX86State *env, ZMMReg *s)
778 {
779     return x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
780 }
781 
782 #ifdef TARGET_X86_64
783 int64_t helper_cvtss2sq(CPUX86State *env, ZMMReg *s)
784 {
785     return x86_float32_to_int64(s->ZMM_S(0), &env->sse_status);
786 }
787 
788 int64_t helper_cvtsd2sq(CPUX86State *env, ZMMReg *s)
789 {
790     return x86_float64_to_int64(s->ZMM_D(0), &env->sse_status);
791 }
792 #endif
793 
794 /* float to integer truncated */
795 void helper_cvttps2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
796 {
797     d->ZMM_L(0) = x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
798     d->ZMM_L(1) = x86_float32_to_int32_round_to_zero(s->ZMM_S(1), &env->sse_status);
799     d->ZMM_L(2) = x86_float32_to_int32_round_to_zero(s->ZMM_S(2), &env->sse_status);
800     d->ZMM_L(3) = x86_float32_to_int32_round_to_zero(s->ZMM_S(3), &env->sse_status);
801 }
802 
803 void helper_cvttpd2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
804 {
805     d->ZMM_L(0) = x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
806     d->ZMM_L(1) = x86_float64_to_int32_round_to_zero(s->ZMM_D(1), &env->sse_status);
807     d->ZMM_Q(1) = 0;
808 }
809 
810 void helper_cvttps2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
811 {
812     d->MMX_L(0) = x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
813     d->MMX_L(1) = x86_float32_to_int32_round_to_zero(s->ZMM_S(1), &env->sse_status);
814 }
815 
816 void helper_cvttpd2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
817 {
818     d->MMX_L(0) = x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
819     d->MMX_L(1) = x86_float64_to_int32_round_to_zero(s->ZMM_D(1), &env->sse_status);
820 }
821 
822 int32_t helper_cvttss2si(CPUX86State *env, ZMMReg *s)
823 {
824     return x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
825 }
826 
827 int32_t helper_cvttsd2si(CPUX86State *env, ZMMReg *s)
828 {
829     return x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
830 }
831 
832 #ifdef TARGET_X86_64
833 int64_t helper_cvttss2sq(CPUX86State *env, ZMMReg *s)
834 {
835     return x86_float32_to_int64_round_to_zero(s->ZMM_S(0), &env->sse_status);
836 }
837 
838 int64_t helper_cvttsd2sq(CPUX86State *env, ZMMReg *s)
839 {
840     return x86_float64_to_int64_round_to_zero(s->ZMM_D(0), &env->sse_status);
841 }
842 #endif
843 
844 void helper_rsqrtps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
845 {
846     d->ZMM_S(0) = float32_div(float32_one,
847                               float32_sqrt(s->ZMM_S(0), &env->sse_status),
848                               &env->sse_status);
849     d->ZMM_S(1) = float32_div(float32_one,
850                               float32_sqrt(s->ZMM_S(1), &env->sse_status),
851                               &env->sse_status);
852     d->ZMM_S(2) = float32_div(float32_one,
853                               float32_sqrt(s->ZMM_S(2), &env->sse_status),
854                               &env->sse_status);
855     d->ZMM_S(3) = float32_div(float32_one,
856                               float32_sqrt(s->ZMM_S(3), &env->sse_status),
857                               &env->sse_status);
858 }
859 
860 void helper_rsqrtss(CPUX86State *env, ZMMReg *d, ZMMReg *s)
861 {
862     d->ZMM_S(0) = float32_div(float32_one,
863                               float32_sqrt(s->ZMM_S(0), &env->sse_status),
864                               &env->sse_status);
865 }
866 
867 void helper_rcpps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
868 {
869     d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
870     d->ZMM_S(1) = float32_div(float32_one, s->ZMM_S(1), &env->sse_status);
871     d->ZMM_S(2) = float32_div(float32_one, s->ZMM_S(2), &env->sse_status);
872     d->ZMM_S(3) = float32_div(float32_one, s->ZMM_S(3), &env->sse_status);
873 }
874 
875 void helper_rcpss(CPUX86State *env, ZMMReg *d, ZMMReg *s)
876 {
877     d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
878 }
879 
880 static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
881 {
882     uint64_t mask;
883 
884     if (len == 0) {
885         mask = ~0LL;
886     } else {
887         mask = (1ULL << len) - 1;
888     }
889     return (src >> shift) & mask;
890 }
891 
892 void helper_extrq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
893 {
894     d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), s->ZMM_B(1), s->ZMM_B(0));
895 }
896 
897 void helper_extrq_i(CPUX86State *env, ZMMReg *d, int index, int length)
898 {
899     d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), index, length);
900 }
901 
902 static inline uint64_t helper_insertq(uint64_t src, int shift, int len)
903 {
904     uint64_t mask;
905 
906     if (len == 0) {
907         mask = ~0ULL;
908     } else {
909         mask = (1ULL << len) - 1;
910     }
911     return (src & ~(mask << shift)) | ((src & mask) << shift);
912 }
913 
914 void helper_insertq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
915 {
916     d->ZMM_Q(0) = helper_insertq(s->ZMM_Q(0), s->ZMM_B(9), s->ZMM_B(8));
917 }
918 
919 void helper_insertq_i(CPUX86State *env, ZMMReg *d, int index, int length)
920 {
921     d->ZMM_Q(0) = helper_insertq(d->ZMM_Q(0), index, length);
922 }
923 
924 void helper_haddps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
925 {
926     ZMMReg r;
927 
928     r.ZMM_S(0) = float32_add(d->ZMM_S(0), d->ZMM_S(1), &env->sse_status);
929     r.ZMM_S(1) = float32_add(d->ZMM_S(2), d->ZMM_S(3), &env->sse_status);
930     r.ZMM_S(2) = float32_add(s->ZMM_S(0), s->ZMM_S(1), &env->sse_status);
931     r.ZMM_S(3) = float32_add(s->ZMM_S(2), s->ZMM_S(3), &env->sse_status);
932     *d = r;
933 }
934 
935 void helper_haddpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
936 {
937     ZMMReg r;
938 
939     r.ZMM_D(0) = float64_add(d->ZMM_D(0), d->ZMM_D(1), &env->sse_status);
940     r.ZMM_D(1) = float64_add(s->ZMM_D(0), s->ZMM_D(1), &env->sse_status);
941     *d = r;
942 }
943 
944 void helper_hsubps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
945 {
946     ZMMReg r;
947 
948     r.ZMM_S(0) = float32_sub(d->ZMM_S(0), d->ZMM_S(1), &env->sse_status);
949     r.ZMM_S(1) = float32_sub(d->ZMM_S(2), d->ZMM_S(3), &env->sse_status);
950     r.ZMM_S(2) = float32_sub(s->ZMM_S(0), s->ZMM_S(1), &env->sse_status);
951     r.ZMM_S(3) = float32_sub(s->ZMM_S(2), s->ZMM_S(3), &env->sse_status);
952     *d = r;
953 }
954 
955 void helper_hsubpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
956 {
957     ZMMReg r;
958 
959     r.ZMM_D(0) = float64_sub(d->ZMM_D(0), d->ZMM_D(1), &env->sse_status);
960     r.ZMM_D(1) = float64_sub(s->ZMM_D(0), s->ZMM_D(1), &env->sse_status);
961     *d = r;
962 }
963 
964 void helper_addsubps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
965 {
966     d->ZMM_S(0) = float32_sub(d->ZMM_S(0), s->ZMM_S(0), &env->sse_status);
967     d->ZMM_S(1) = float32_add(d->ZMM_S(1), s->ZMM_S(1), &env->sse_status);
968     d->ZMM_S(2) = float32_sub(d->ZMM_S(2), s->ZMM_S(2), &env->sse_status);
969     d->ZMM_S(3) = float32_add(d->ZMM_S(3), s->ZMM_S(3), &env->sse_status);
970 }
971 
972 void helper_addsubpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
973 {
974     d->ZMM_D(0) = float64_sub(d->ZMM_D(0), s->ZMM_D(0), &env->sse_status);
975     d->ZMM_D(1) = float64_add(d->ZMM_D(1), s->ZMM_D(1), &env->sse_status);
976 }
977 
978 /* XXX: unordered */
979 #define SSE_HELPER_CMP(name, F)                                         \
980     void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s)        \
981     {                                                                   \
982         d->ZMM_L(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
983         d->ZMM_L(1) = F(32, d->ZMM_S(1), s->ZMM_S(1));                  \
984         d->ZMM_L(2) = F(32, d->ZMM_S(2), s->ZMM_S(2));                  \
985         d->ZMM_L(3) = F(32, d->ZMM_S(3), s->ZMM_S(3));                  \
986     }                                                                   \
987                                                                         \
988     void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s)        \
989     {                                                                   \
990         d->ZMM_L(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
991     }                                                                   \
992                                                                         \
993     void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s)        \
994     {                                                                   \
995         d->ZMM_Q(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
996         d->ZMM_Q(1) = F(64, d->ZMM_D(1), s->ZMM_D(1));                  \
997     }                                                                   \
998                                                                         \
999     void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s)        \
1000     {                                                                   \
1001         d->ZMM_Q(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
1002     }
1003 
1004 #define FPU_CMPEQ(size, a, b)                                           \
1005     (float ## size ## _eq_quiet(a, b, &env->sse_status) ? -1 : 0)
1006 #define FPU_CMPLT(size, a, b)                                           \
1007     (float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0)
1008 #define FPU_CMPLE(size, a, b)                                           \
1009     (float ## size ## _le(a, b, &env->sse_status) ? -1 : 0)
1010 #define FPU_CMPUNORD(size, a, b)                                        \
1011     (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? -1 : 0)
1012 #define FPU_CMPNEQ(size, a, b)                                          \
1013     (float ## size ## _eq_quiet(a, b, &env->sse_status) ? 0 : -1)
1014 #define FPU_CMPNLT(size, a, b)                                          \
1015     (float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1)
1016 #define FPU_CMPNLE(size, a, b)                                          \
1017     (float ## size ## _le(a, b, &env->sse_status) ? 0 : -1)
1018 #define FPU_CMPORD(size, a, b)                                          \
1019     (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? 0 : -1)
1020 
1021 SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
1022 SSE_HELPER_CMP(cmplt, FPU_CMPLT)
1023 SSE_HELPER_CMP(cmple, FPU_CMPLE)
1024 SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
1025 SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
1026 SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
1027 SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
1028 SSE_HELPER_CMP(cmpord, FPU_CMPORD)
1029 
1030 static const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
1031 
1032 void helper_ucomiss(CPUX86State *env, Reg *d, Reg *s)
1033 {
1034     FloatRelation ret;
1035     float32 s0, s1;
1036 
1037     s0 = d->ZMM_S(0);
1038     s1 = s->ZMM_S(0);
1039     ret = float32_compare_quiet(s0, s1, &env->sse_status);
1040     CC_SRC = comis_eflags[ret + 1];
1041 }
1042 
1043 void helper_comiss(CPUX86State *env, Reg *d, Reg *s)
1044 {
1045     FloatRelation ret;
1046     float32 s0, s1;
1047 
1048     s0 = d->ZMM_S(0);
1049     s1 = s->ZMM_S(0);
1050     ret = float32_compare(s0, s1, &env->sse_status);
1051     CC_SRC = comis_eflags[ret + 1];
1052 }
1053 
1054 void helper_ucomisd(CPUX86State *env, Reg *d, Reg *s)
1055 {
1056     FloatRelation ret;
1057     float64 d0, d1;
1058 
1059     d0 = d->ZMM_D(0);
1060     d1 = s->ZMM_D(0);
1061     ret = float64_compare_quiet(d0, d1, &env->sse_status);
1062     CC_SRC = comis_eflags[ret + 1];
1063 }
1064 
1065 void helper_comisd(CPUX86State *env, Reg *d, Reg *s)
1066 {
1067     FloatRelation ret;
1068     float64 d0, d1;
1069 
1070     d0 = d->ZMM_D(0);
1071     d1 = s->ZMM_D(0);
1072     ret = float64_compare(d0, d1, &env->sse_status);
1073     CC_SRC = comis_eflags[ret + 1];
1074 }
1075 
1076 uint32_t helper_movmskps(CPUX86State *env, Reg *s)
1077 {
1078     int b0, b1, b2, b3;
1079 
1080     b0 = s->ZMM_L(0) >> 31;
1081     b1 = s->ZMM_L(1) >> 31;
1082     b2 = s->ZMM_L(2) >> 31;
1083     b3 = s->ZMM_L(3) >> 31;
1084     return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
1085 }
1086 
1087 uint32_t helper_movmskpd(CPUX86State *env, Reg *s)
1088 {
1089     int b0, b1;
1090 
1091     b0 = s->ZMM_L(1) >> 31;
1092     b1 = s->ZMM_L(3) >> 31;
1093     return b0 | (b1 << 1);
1094 }
1095 
1096 #endif
1097 
1098 uint32_t glue(helper_pmovmskb, SUFFIX)(CPUX86State *env, Reg *s)
1099 {
1100     uint32_t val;
1101 
1102     val = 0;
1103     val |= (s->B(0) >> 7);
1104     val |= (s->B(1) >> 6) & 0x02;
1105     val |= (s->B(2) >> 5) & 0x04;
1106     val |= (s->B(3) >> 4) & 0x08;
1107     val |= (s->B(4) >> 3) & 0x10;
1108     val |= (s->B(5) >> 2) & 0x20;
1109     val |= (s->B(6) >> 1) & 0x40;
1110     val |= (s->B(7)) & 0x80;
1111 #if SHIFT == 1
1112     val |= (s->B(8) << 1) & 0x0100;
1113     val |= (s->B(9) << 2) & 0x0200;
1114     val |= (s->B(10) << 3) & 0x0400;
1115     val |= (s->B(11) << 4) & 0x0800;
1116     val |= (s->B(12) << 5) & 0x1000;
1117     val |= (s->B(13) << 6) & 0x2000;
1118     val |= (s->B(14) << 7) & 0x4000;
1119     val |= (s->B(15) << 8) & 0x8000;
1120 #endif
1121     return val;
1122 }
1123 
1124 void glue(helper_packsswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1125 {
1126     Reg r;
1127 
1128     r.B(0) = satsb((int16_t)d->W(0));
1129     r.B(1) = satsb((int16_t)d->W(1));
1130     r.B(2) = satsb((int16_t)d->W(2));
1131     r.B(3) = satsb((int16_t)d->W(3));
1132 #if SHIFT == 1
1133     r.B(4) = satsb((int16_t)d->W(4));
1134     r.B(5) = satsb((int16_t)d->W(5));
1135     r.B(6) = satsb((int16_t)d->W(6));
1136     r.B(7) = satsb((int16_t)d->W(7));
1137 #endif
1138     r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1139     r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1140     r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1141     r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1142 #if SHIFT == 1
1143     r.B(12) = satsb((int16_t)s->W(4));
1144     r.B(13) = satsb((int16_t)s->W(5));
1145     r.B(14) = satsb((int16_t)s->W(6));
1146     r.B(15) = satsb((int16_t)s->W(7));
1147 #endif
1148     *d = r;
1149 }
1150 
1151 void glue(helper_packuswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1152 {
1153     Reg r;
1154 
1155     r.B(0) = satub((int16_t)d->W(0));
1156     r.B(1) = satub((int16_t)d->W(1));
1157     r.B(2) = satub((int16_t)d->W(2));
1158     r.B(3) = satub((int16_t)d->W(3));
1159 #if SHIFT == 1
1160     r.B(4) = satub((int16_t)d->W(4));
1161     r.B(5) = satub((int16_t)d->W(5));
1162     r.B(6) = satub((int16_t)d->W(6));
1163     r.B(7) = satub((int16_t)d->W(7));
1164 #endif
1165     r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1166     r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1167     r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1168     r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1169 #if SHIFT == 1
1170     r.B(12) = satub((int16_t)s->W(4));
1171     r.B(13) = satub((int16_t)s->W(5));
1172     r.B(14) = satub((int16_t)s->W(6));
1173     r.B(15) = satub((int16_t)s->W(7));
1174 #endif
1175     *d = r;
1176 }
1177 
1178 void glue(helper_packssdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1179 {
1180     Reg r;
1181 
1182     r.W(0) = satsw(d->L(0));
1183     r.W(1) = satsw(d->L(1));
1184 #if SHIFT == 1
1185     r.W(2) = satsw(d->L(2));
1186     r.W(3) = satsw(d->L(3));
1187 #endif
1188     r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1189     r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1190 #if SHIFT == 1
1191     r.W(6) = satsw(s->L(2));
1192     r.W(7) = satsw(s->L(3));
1193 #endif
1194     *d = r;
1195 }
1196 
1197 #define UNPCK_OP(base_name, base)                                       \
1198                                                                         \
1199     void glue(helper_punpck ## base_name ## bw, SUFFIX)(CPUX86State *env,\
1200                                                         Reg *d, Reg *s) \
1201     {                                                                   \
1202         Reg r;                                                          \
1203                                                                         \
1204         r.B(0) = d->B((base << (SHIFT + 2)) + 0);                       \
1205         r.B(1) = s->B((base << (SHIFT + 2)) + 0);                       \
1206         r.B(2) = d->B((base << (SHIFT + 2)) + 1);                       \
1207         r.B(3) = s->B((base << (SHIFT + 2)) + 1);                       \
1208         r.B(4) = d->B((base << (SHIFT + 2)) + 2);                       \
1209         r.B(5) = s->B((base << (SHIFT + 2)) + 2);                       \
1210         r.B(6) = d->B((base << (SHIFT + 2)) + 3);                       \
1211         r.B(7) = s->B((base << (SHIFT + 2)) + 3);                       \
1212         XMM_ONLY(                                                       \
1213                  r.B(8) = d->B((base << (SHIFT + 2)) + 4);              \
1214                  r.B(9) = s->B((base << (SHIFT + 2)) + 4);              \
1215                  r.B(10) = d->B((base << (SHIFT + 2)) + 5);             \
1216                  r.B(11) = s->B((base << (SHIFT + 2)) + 5);             \
1217                  r.B(12) = d->B((base << (SHIFT + 2)) + 6);             \
1218                  r.B(13) = s->B((base << (SHIFT + 2)) + 6);             \
1219                  r.B(14) = d->B((base << (SHIFT + 2)) + 7);             \
1220                  r.B(15) = s->B((base << (SHIFT + 2)) + 7);             \
1221                                                                       ) \
1222             *d = r;                                                     \
1223     }                                                                   \
1224                                                                         \
1225     void glue(helper_punpck ## base_name ## wd, SUFFIX)(CPUX86State *env,\
1226                                                         Reg *d, Reg *s) \
1227     {                                                                   \
1228         Reg r;                                                          \
1229                                                                         \
1230         r.W(0) = d->W((base << (SHIFT + 1)) + 0);                       \
1231         r.W(1) = s->W((base << (SHIFT + 1)) + 0);                       \
1232         r.W(2) = d->W((base << (SHIFT + 1)) + 1);                       \
1233         r.W(3) = s->W((base << (SHIFT + 1)) + 1);                       \
1234         XMM_ONLY(                                                       \
1235                  r.W(4) = d->W((base << (SHIFT + 1)) + 2);              \
1236                  r.W(5) = s->W((base << (SHIFT + 1)) + 2);              \
1237                  r.W(6) = d->W((base << (SHIFT + 1)) + 3);              \
1238                  r.W(7) = s->W((base << (SHIFT + 1)) + 3);              \
1239                                                                       ) \
1240             *d = r;                                                     \
1241     }                                                                   \
1242                                                                         \
1243     void glue(helper_punpck ## base_name ## dq, SUFFIX)(CPUX86State *env,\
1244                                                         Reg *d, Reg *s) \
1245     {                                                                   \
1246         Reg r;                                                          \
1247                                                                         \
1248         r.L(0) = d->L((base << SHIFT) + 0);                             \
1249         r.L(1) = s->L((base << SHIFT) + 0);                             \
1250         XMM_ONLY(                                                       \
1251                  r.L(2) = d->L((base << SHIFT) + 1);                    \
1252                  r.L(3) = s->L((base << SHIFT) + 1);                    \
1253                                                                       ) \
1254             *d = r;                                                     \
1255     }                                                                   \
1256                                                                         \
1257     XMM_ONLY(                                                           \
1258              void glue(helper_punpck ## base_name ## qdq, SUFFIX)(CPUX86State \
1259                                                                   *env, \
1260                                                                   Reg *d, \
1261                                                                   Reg *s) \
1262              {                                                          \
1263                  Reg r;                                                 \
1264                                                                         \
1265                  r.Q(0) = d->Q(base);                                   \
1266                  r.Q(1) = s->Q(base);                                   \
1267                  *d = r;                                                \
1268              }                                                          \
1269                                                                         )
1270 
1271 UNPCK_OP(l, 0)
1272 UNPCK_OP(h, 1)
1273 
1274 /* 3DNow! float ops */
1275 #if SHIFT == 0
1276 void helper_pi2fd(CPUX86State *env, MMXReg *d, MMXReg *s)
1277 {
1278     d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
1279     d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
1280 }
1281 
1282 void helper_pi2fw(CPUX86State *env, MMXReg *d, MMXReg *s)
1283 {
1284     d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
1285     d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
1286 }
1287 
1288 void helper_pf2id(CPUX86State *env, MMXReg *d, MMXReg *s)
1289 {
1290     d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
1291     d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
1292 }
1293 
1294 void helper_pf2iw(CPUX86State *env, MMXReg *d, MMXReg *s)
1295 {
1296     d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0),
1297                                                        &env->mmx_status));
1298     d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1),
1299                                                        &env->mmx_status));
1300 }
1301 
1302 void helper_pfacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1303 {
1304     MMXReg r;
1305 
1306     r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1307     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1308     *d = r;
1309 }
1310 
1311 void helper_pfadd(CPUX86State *env, MMXReg *d, MMXReg *s)
1312 {
1313     d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1314     d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1315 }
1316 
1317 void helper_pfcmpeq(CPUX86State *env, MMXReg *d, MMXReg *s)
1318 {
1319     d->MMX_L(0) = float32_eq_quiet(d->MMX_S(0), s->MMX_S(0),
1320                                    &env->mmx_status) ? -1 : 0;
1321     d->MMX_L(1) = float32_eq_quiet(d->MMX_S(1), s->MMX_S(1),
1322                                    &env->mmx_status) ? -1 : 0;
1323 }
1324 
1325 void helper_pfcmpge(CPUX86State *env, MMXReg *d, MMXReg *s)
1326 {
1327     d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0),
1328                              &env->mmx_status) ? -1 : 0;
1329     d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1),
1330                              &env->mmx_status) ? -1 : 0;
1331 }
1332 
1333 void helper_pfcmpgt(CPUX86State *env, MMXReg *d, MMXReg *s)
1334 {
1335     d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0),
1336                              &env->mmx_status) ? -1 : 0;
1337     d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1),
1338                              &env->mmx_status) ? -1 : 0;
1339 }
1340 
1341 void helper_pfmax(CPUX86State *env, MMXReg *d, MMXReg *s)
1342 {
1343     if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status)) {
1344         d->MMX_S(0) = s->MMX_S(0);
1345     }
1346     if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status)) {
1347         d->MMX_S(1) = s->MMX_S(1);
1348     }
1349 }
1350 
1351 void helper_pfmin(CPUX86State *env, MMXReg *d, MMXReg *s)
1352 {
1353     if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status)) {
1354         d->MMX_S(0) = s->MMX_S(0);
1355     }
1356     if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status)) {
1357         d->MMX_S(1) = s->MMX_S(1);
1358     }
1359 }
1360 
1361 void helper_pfmul(CPUX86State *env, MMXReg *d, MMXReg *s)
1362 {
1363     d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1364     d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1365 }
1366 
1367 void helper_pfnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1368 {
1369     MMXReg r;
1370 
1371     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1372     r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1373     *d = r;
1374 }
1375 
1376 void helper_pfpnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1377 {
1378     MMXReg r;
1379 
1380     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1381     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1382     *d = r;
1383 }
1384 
1385 void helper_pfrcp(CPUX86State *env, MMXReg *d, MMXReg *s)
1386 {
1387     d->MMX_S(0) = float32_div(float32_one, s->MMX_S(0), &env->mmx_status);
1388     d->MMX_S(1) = d->MMX_S(0);
1389 }
1390 
1391 void helper_pfrsqrt(CPUX86State *env, MMXReg *d, MMXReg *s)
1392 {
1393     d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
1394     d->MMX_S(1) = float32_div(float32_one,
1395                               float32_sqrt(d->MMX_S(1), &env->mmx_status),
1396                               &env->mmx_status);
1397     d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
1398     d->MMX_L(0) = d->MMX_L(1);
1399 }
1400 
1401 void helper_pfsub(CPUX86State *env, MMXReg *d, MMXReg *s)
1402 {
1403     d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1404     d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1405 }
1406 
1407 void helper_pfsubr(CPUX86State *env, MMXReg *d, MMXReg *s)
1408 {
1409     d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
1410     d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
1411 }
1412 
1413 void helper_pswapd(CPUX86State *env, MMXReg *d, MMXReg *s)
1414 {
1415     MMXReg r;
1416 
1417     r.MMX_L(0) = s->MMX_L(1);
1418     r.MMX_L(1) = s->MMX_L(0);
1419     *d = r;
1420 }
1421 #endif
1422 
1423 /* SSSE3 op helpers */
1424 void glue(helper_pshufb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1425 {
1426     int i;
1427     Reg r;
1428 
1429     for (i = 0; i < (8 << SHIFT); i++) {
1430         r.B(i) = (s->B(i) & 0x80) ? 0 : (d->B(s->B(i) & ((8 << SHIFT) - 1)));
1431     }
1432 
1433     *d = r;
1434 }
1435 
1436 void glue(helper_phaddw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1437 {
1438 
1439     Reg r;
1440 
1441     r.W(0) = (int16_t)d->W(0) + (int16_t)d->W(1);
1442     r.W(1) = (int16_t)d->W(2) + (int16_t)d->W(3);
1443     XMM_ONLY(r.W(2) = (int16_t)d->W(4) + (int16_t)d->W(5));
1444     XMM_ONLY(r.W(3) = (int16_t)d->W(6) + (int16_t)d->W(7));
1445     r.W((2 << SHIFT) + 0) = (int16_t)s->W(0) + (int16_t)s->W(1);
1446     r.W((2 << SHIFT) + 1) = (int16_t)s->W(2) + (int16_t)s->W(3);
1447     XMM_ONLY(r.W(6) = (int16_t)s->W(4) + (int16_t)s->W(5));
1448     XMM_ONLY(r.W(7) = (int16_t)s->W(6) + (int16_t)s->W(7));
1449 
1450     *d = r;
1451 }
1452 
1453 void glue(helper_phaddd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1454 {
1455     Reg r;
1456 
1457     r.L(0) = (int32_t)d->L(0) + (int32_t)d->L(1);
1458     XMM_ONLY(r.L(1) = (int32_t)d->L(2) + (int32_t)d->L(3));
1459     r.L((1 << SHIFT) + 0) = (int32_t)s->L(0) + (int32_t)s->L(1);
1460     XMM_ONLY(r.L(3) = (int32_t)s->L(2) + (int32_t)s->L(3));
1461 
1462     *d = r;
1463 }
1464 
1465 void glue(helper_phaddsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1466 {
1467     Reg r;
1468 
1469     r.W(0) = satsw((int16_t)d->W(0) + (int16_t)d->W(1));
1470     r.W(1) = satsw((int16_t)d->W(2) + (int16_t)d->W(3));
1471     XMM_ONLY(r.W(2) = satsw((int16_t)d->W(4) + (int16_t)d->W(5)));
1472     XMM_ONLY(r.W(3) = satsw((int16_t)d->W(6) + (int16_t)d->W(7)));
1473     r.W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) + (int16_t)s->W(1));
1474     r.W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) + (int16_t)s->W(3));
1475     XMM_ONLY(r.W(6) = satsw((int16_t)s->W(4) + (int16_t)s->W(5)));
1476     XMM_ONLY(r.W(7) = satsw((int16_t)s->W(6) + (int16_t)s->W(7)));
1477 
1478     *d = r;
1479 }
1480 
1481 void glue(helper_pmaddubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1482 {
1483     d->W(0) = satsw((int8_t)s->B(0) * (uint8_t)d->B(0) +
1484                     (int8_t)s->B(1) * (uint8_t)d->B(1));
1485     d->W(1) = satsw((int8_t)s->B(2) * (uint8_t)d->B(2) +
1486                     (int8_t)s->B(3) * (uint8_t)d->B(3));
1487     d->W(2) = satsw((int8_t)s->B(4) * (uint8_t)d->B(4) +
1488                     (int8_t)s->B(5) * (uint8_t)d->B(5));
1489     d->W(3) = satsw((int8_t)s->B(6) * (uint8_t)d->B(6) +
1490                     (int8_t)s->B(7) * (uint8_t)d->B(7));
1491 #if SHIFT == 1
1492     d->W(4) = satsw((int8_t)s->B(8) * (uint8_t)d->B(8) +
1493                     (int8_t)s->B(9) * (uint8_t)d->B(9));
1494     d->W(5) = satsw((int8_t)s->B(10) * (uint8_t)d->B(10) +
1495                     (int8_t)s->B(11) * (uint8_t)d->B(11));
1496     d->W(6) = satsw((int8_t)s->B(12) * (uint8_t)d->B(12) +
1497                     (int8_t)s->B(13) * (uint8_t)d->B(13));
1498     d->W(7) = satsw((int8_t)s->B(14) * (uint8_t)d->B(14) +
1499                     (int8_t)s->B(15) * (uint8_t)d->B(15));
1500 #endif
1501 }
1502 
1503 void glue(helper_phsubw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1504 {
1505     d->W(0) = (int16_t)d->W(0) - (int16_t)d->W(1);
1506     d->W(1) = (int16_t)d->W(2) - (int16_t)d->W(3);
1507     XMM_ONLY(d->W(2) = (int16_t)d->W(4) - (int16_t)d->W(5));
1508     XMM_ONLY(d->W(3) = (int16_t)d->W(6) - (int16_t)d->W(7));
1509     d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) - (int16_t)s->W(1);
1510     d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) - (int16_t)s->W(3);
1511     XMM_ONLY(d->W(6) = (int16_t)s->W(4) - (int16_t)s->W(5));
1512     XMM_ONLY(d->W(7) = (int16_t)s->W(6) - (int16_t)s->W(7));
1513 }
1514 
1515 void glue(helper_phsubd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1516 {
1517     d->L(0) = (int32_t)d->L(0) - (int32_t)d->L(1);
1518     XMM_ONLY(d->L(1) = (int32_t)d->L(2) - (int32_t)d->L(3));
1519     d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) - (int32_t)s->L(1);
1520     XMM_ONLY(d->L(3) = (int32_t)s->L(2) - (int32_t)s->L(3));
1521 }
1522 
1523 void glue(helper_phsubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1524 {
1525     d->W(0) = satsw((int16_t)d->W(0) - (int16_t)d->W(1));
1526     d->W(1) = satsw((int16_t)d->W(2) - (int16_t)d->W(3));
1527     XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) - (int16_t)d->W(5)));
1528     XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) - (int16_t)d->W(7)));
1529     d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) - (int16_t)s->W(1));
1530     d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) - (int16_t)s->W(3));
1531     XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) - (int16_t)s->W(5)));
1532     XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) - (int16_t)s->W(7)));
1533 }
1534 
1535 #define FABSB(_, x) (x > INT8_MAX  ? -(int8_t)x : x)
1536 #define FABSW(_, x) (x > INT16_MAX ? -(int16_t)x : x)
1537 #define FABSL(_, x) (x > INT32_MAX ? -(int32_t)x : x)
1538 SSE_HELPER_B(helper_pabsb, FABSB)
1539 SSE_HELPER_W(helper_pabsw, FABSW)
1540 SSE_HELPER_L(helper_pabsd, FABSL)
1541 
1542 #define FMULHRSW(d, s) (((int16_t) d * (int16_t)s + 0x4000) >> 15)
1543 SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
1544 
1545 #define FSIGNB(d, s) (s <= INT8_MAX  ? s ? d : 0 : -(int8_t)d)
1546 #define FSIGNW(d, s) (s <= INT16_MAX ? s ? d : 0 : -(int16_t)d)
1547 #define FSIGNL(d, s) (s <= INT32_MAX ? s ? d : 0 : -(int32_t)d)
1548 SSE_HELPER_B(helper_psignb, FSIGNB)
1549 SSE_HELPER_W(helper_psignw, FSIGNW)
1550 SSE_HELPER_L(helper_psignd, FSIGNL)
1551 
1552 void glue(helper_palignr, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1553                                   int32_t shift)
1554 {
1555     Reg r;
1556 
1557     /* XXX could be checked during translation */
1558     if (shift >= (16 << SHIFT)) {
1559         r.Q(0) = 0;
1560         XMM_ONLY(r.Q(1) = 0);
1561     } else {
1562         shift <<= 3;
1563 #define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
1564 #if SHIFT == 0
1565         r.Q(0) = SHR(s->Q(0), shift - 0) |
1566             SHR(d->Q(0), shift -  64);
1567 #else
1568         r.Q(0) = SHR(s->Q(0), shift - 0) |
1569             SHR(s->Q(1), shift -  64) |
1570             SHR(d->Q(0), shift - 128) |
1571             SHR(d->Q(1), shift - 192);
1572         r.Q(1) = SHR(s->Q(0), shift + 64) |
1573             SHR(s->Q(1), shift -   0) |
1574             SHR(d->Q(0), shift -  64) |
1575             SHR(d->Q(1), shift - 128);
1576 #endif
1577 #undef SHR
1578     }
1579 
1580     *d = r;
1581 }
1582 
1583 #define XMM0 (env->xmm_regs[0])
1584 
1585 #if SHIFT == 1
1586 #define SSE_HELPER_V(name, elem, num, F)                                \
1587     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)           \
1588     {                                                                   \
1589         d->elem(0) = F(d->elem(0), s->elem(0), XMM0.elem(0));           \
1590         d->elem(1) = F(d->elem(1), s->elem(1), XMM0.elem(1));           \
1591         if (num > 2) {                                                  \
1592             d->elem(2) = F(d->elem(2), s->elem(2), XMM0.elem(2));       \
1593             d->elem(3) = F(d->elem(3), s->elem(3), XMM0.elem(3));       \
1594             if (num > 4) {                                              \
1595                 d->elem(4) = F(d->elem(4), s->elem(4), XMM0.elem(4));   \
1596                 d->elem(5) = F(d->elem(5), s->elem(5), XMM0.elem(5));   \
1597                 d->elem(6) = F(d->elem(6), s->elem(6), XMM0.elem(6));   \
1598                 d->elem(7) = F(d->elem(7), s->elem(7), XMM0.elem(7));   \
1599                 if (num > 8) {                                          \
1600                     d->elem(8) = F(d->elem(8), s->elem(8), XMM0.elem(8)); \
1601                     d->elem(9) = F(d->elem(9), s->elem(9), XMM0.elem(9)); \
1602                     d->elem(10) = F(d->elem(10), s->elem(10), XMM0.elem(10)); \
1603                     d->elem(11) = F(d->elem(11), s->elem(11), XMM0.elem(11)); \
1604                     d->elem(12) = F(d->elem(12), s->elem(12), XMM0.elem(12)); \
1605                     d->elem(13) = F(d->elem(13), s->elem(13), XMM0.elem(13)); \
1606                     d->elem(14) = F(d->elem(14), s->elem(14), XMM0.elem(14)); \
1607                     d->elem(15) = F(d->elem(15), s->elem(15), XMM0.elem(15)); \
1608                 }                                                       \
1609             }                                                           \
1610         }                                                               \
1611     }
1612 
1613 #define SSE_HELPER_I(name, elem, num, F)                                \
1614     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t imm) \
1615     {                                                                   \
1616         d->elem(0) = F(d->elem(0), s->elem(0), ((imm >> 0) & 1));       \
1617         d->elem(1) = F(d->elem(1), s->elem(1), ((imm >> 1) & 1));       \
1618         if (num > 2) {                                                  \
1619             d->elem(2) = F(d->elem(2), s->elem(2), ((imm >> 2) & 1));   \
1620             d->elem(3) = F(d->elem(3), s->elem(3), ((imm >> 3) & 1));   \
1621             if (num > 4) {                                              \
1622                 d->elem(4) = F(d->elem(4), s->elem(4), ((imm >> 4) & 1)); \
1623                 d->elem(5) = F(d->elem(5), s->elem(5), ((imm >> 5) & 1)); \
1624                 d->elem(6) = F(d->elem(6), s->elem(6), ((imm >> 6) & 1)); \
1625                 d->elem(7) = F(d->elem(7), s->elem(7), ((imm >> 7) & 1)); \
1626                 if (num > 8) {                                          \
1627                     d->elem(8) = F(d->elem(8), s->elem(8), ((imm >> 8) & 1)); \
1628                     d->elem(9) = F(d->elem(9), s->elem(9), ((imm >> 9) & 1)); \
1629                     d->elem(10) = F(d->elem(10), s->elem(10),           \
1630                                     ((imm >> 10) & 1));                 \
1631                     d->elem(11) = F(d->elem(11), s->elem(11),           \
1632                                     ((imm >> 11) & 1));                 \
1633                     d->elem(12) = F(d->elem(12), s->elem(12),           \
1634                                     ((imm >> 12) & 1));                 \
1635                     d->elem(13) = F(d->elem(13), s->elem(13),           \
1636                                     ((imm >> 13) & 1));                 \
1637                     d->elem(14) = F(d->elem(14), s->elem(14),           \
1638                                     ((imm >> 14) & 1));                 \
1639                     d->elem(15) = F(d->elem(15), s->elem(15),           \
1640                                     ((imm >> 15) & 1));                 \
1641                 }                                                       \
1642             }                                                           \
1643         }                                                               \
1644     }
1645 
1646 /* SSE4.1 op helpers */
1647 #define FBLENDVB(d, s, m) ((m & 0x80) ? s : d)
1648 #define FBLENDVPS(d, s, m) ((m & 0x80000000) ? s : d)
1649 #define FBLENDVPD(d, s, m) ((m & 0x8000000000000000LL) ? s : d)
1650 SSE_HELPER_V(helper_pblendvb, B, 16, FBLENDVB)
1651 SSE_HELPER_V(helper_blendvps, L, 4, FBLENDVPS)
1652 SSE_HELPER_V(helper_blendvpd, Q, 2, FBLENDVPD)
1653 
1654 void glue(helper_ptest, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1655 {
1656     uint64_t zf = (s->Q(0) &  d->Q(0)) | (s->Q(1) &  d->Q(1));
1657     uint64_t cf = (s->Q(0) & ~d->Q(0)) | (s->Q(1) & ~d->Q(1));
1658 
1659     CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
1660 }
1661 
1662 #define SSE_HELPER_F(name, elem, num, F)        \
1663     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)     \
1664     {                                           \
1665         if (num > 2) {                          \
1666             if (num > 4) {                      \
1667                 d->elem(7) = F(7);              \
1668                 d->elem(6) = F(6);              \
1669                 d->elem(5) = F(5);              \
1670                 d->elem(4) = F(4);              \
1671             }                                   \
1672             d->elem(3) = F(3);                  \
1673             d->elem(2) = F(2);                  \
1674         }                                       \
1675         d->elem(1) = F(1);                      \
1676         d->elem(0) = F(0);                      \
1677     }
1678 
1679 SSE_HELPER_F(helper_pmovsxbw, W, 8, (int8_t) s->B)
1680 SSE_HELPER_F(helper_pmovsxbd, L, 4, (int8_t) s->B)
1681 SSE_HELPER_F(helper_pmovsxbq, Q, 2, (int8_t) s->B)
1682 SSE_HELPER_F(helper_pmovsxwd, L, 4, (int16_t) s->W)
1683 SSE_HELPER_F(helper_pmovsxwq, Q, 2, (int16_t) s->W)
1684 SSE_HELPER_F(helper_pmovsxdq, Q, 2, (int32_t) s->L)
1685 SSE_HELPER_F(helper_pmovzxbw, W, 8, s->B)
1686 SSE_HELPER_F(helper_pmovzxbd, L, 4, s->B)
1687 SSE_HELPER_F(helper_pmovzxbq, Q, 2, s->B)
1688 SSE_HELPER_F(helper_pmovzxwd, L, 4, s->W)
1689 SSE_HELPER_F(helper_pmovzxwq, Q, 2, s->W)
1690 SSE_HELPER_F(helper_pmovzxdq, Q, 2, s->L)
1691 
1692 void glue(helper_pmuldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1693 {
1694     d->Q(0) = (int64_t)(int32_t) d->L(0) * (int32_t) s->L(0);
1695     d->Q(1) = (int64_t)(int32_t) d->L(2) * (int32_t) s->L(2);
1696 }
1697 
1698 #define FCMPEQQ(d, s) (d == s ? -1 : 0)
1699 SSE_HELPER_Q(helper_pcmpeqq, FCMPEQQ)
1700 
1701 void glue(helper_packusdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1702 {
1703     Reg r;
1704 
1705     r.W(0) = satuw((int32_t) d->L(0));
1706     r.W(1) = satuw((int32_t) d->L(1));
1707     r.W(2) = satuw((int32_t) d->L(2));
1708     r.W(3) = satuw((int32_t) d->L(3));
1709     r.W(4) = satuw((int32_t) s->L(0));
1710     r.W(5) = satuw((int32_t) s->L(1));
1711     r.W(6) = satuw((int32_t) s->L(2));
1712     r.W(7) = satuw((int32_t) s->L(3));
1713     *d = r;
1714 }
1715 
1716 #define FMINSB(d, s) MIN((int8_t)d, (int8_t)s)
1717 #define FMINSD(d, s) MIN((int32_t)d, (int32_t)s)
1718 #define FMAXSB(d, s) MAX((int8_t)d, (int8_t)s)
1719 #define FMAXSD(d, s) MAX((int32_t)d, (int32_t)s)
1720 SSE_HELPER_B(helper_pminsb, FMINSB)
1721 SSE_HELPER_L(helper_pminsd, FMINSD)
1722 SSE_HELPER_W(helper_pminuw, MIN)
1723 SSE_HELPER_L(helper_pminud, MIN)
1724 SSE_HELPER_B(helper_pmaxsb, FMAXSB)
1725 SSE_HELPER_L(helper_pmaxsd, FMAXSD)
1726 SSE_HELPER_W(helper_pmaxuw, MAX)
1727 SSE_HELPER_L(helper_pmaxud, MAX)
1728 
1729 #define FMULLD(d, s) ((int32_t)d * (int32_t)s)
1730 SSE_HELPER_L(helper_pmulld, FMULLD)
1731 
1732 void glue(helper_phminposuw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1733 {
1734     int idx = 0;
1735 
1736     if (s->W(1) < s->W(idx)) {
1737         idx = 1;
1738     }
1739     if (s->W(2) < s->W(idx)) {
1740         idx = 2;
1741     }
1742     if (s->W(3) < s->W(idx)) {
1743         idx = 3;
1744     }
1745     if (s->W(4) < s->W(idx)) {
1746         idx = 4;
1747     }
1748     if (s->W(5) < s->W(idx)) {
1749         idx = 5;
1750     }
1751     if (s->W(6) < s->W(idx)) {
1752         idx = 6;
1753     }
1754     if (s->W(7) < s->W(idx)) {
1755         idx = 7;
1756     }
1757 
1758     d->W(0) = s->W(idx);
1759     d->W(1) = idx;
1760     d->L(1) = 0;
1761     d->Q(1) = 0;
1762 }
1763 
1764 void glue(helper_roundps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1765                                   uint32_t mode)
1766 {
1767     signed char prev_rounding_mode;
1768 
1769     prev_rounding_mode = env->sse_status.float_rounding_mode;
1770     if (!(mode & (1 << 2))) {
1771         switch (mode & 3) {
1772         case 0:
1773             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1774             break;
1775         case 1:
1776             set_float_rounding_mode(float_round_down, &env->sse_status);
1777             break;
1778         case 2:
1779             set_float_rounding_mode(float_round_up, &env->sse_status);
1780             break;
1781         case 3:
1782             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1783             break;
1784         }
1785     }
1786 
1787     d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
1788     d->ZMM_S(1) = float32_round_to_int(s->ZMM_S(1), &env->sse_status);
1789     d->ZMM_S(2) = float32_round_to_int(s->ZMM_S(2), &env->sse_status);
1790     d->ZMM_S(3) = float32_round_to_int(s->ZMM_S(3), &env->sse_status);
1791 
1792 #if 0 /* TODO */
1793     if (mode & (1 << 3)) {
1794         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1795                                   ~float_flag_inexact,
1796                                   &env->sse_status);
1797     }
1798 #endif
1799     env->sse_status.float_rounding_mode = prev_rounding_mode;
1800 }
1801 
1802 void glue(helper_roundpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1803                                   uint32_t mode)
1804 {
1805     signed char prev_rounding_mode;
1806 
1807     prev_rounding_mode = env->sse_status.float_rounding_mode;
1808     if (!(mode & (1 << 2))) {
1809         switch (mode & 3) {
1810         case 0:
1811             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1812             break;
1813         case 1:
1814             set_float_rounding_mode(float_round_down, &env->sse_status);
1815             break;
1816         case 2:
1817             set_float_rounding_mode(float_round_up, &env->sse_status);
1818             break;
1819         case 3:
1820             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1821             break;
1822         }
1823     }
1824 
1825     d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
1826     d->ZMM_D(1) = float64_round_to_int(s->ZMM_D(1), &env->sse_status);
1827 
1828 #if 0 /* TODO */
1829     if (mode & (1 << 3)) {
1830         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1831                                   ~float_flag_inexact,
1832                                   &env->sse_status);
1833     }
1834 #endif
1835     env->sse_status.float_rounding_mode = prev_rounding_mode;
1836 }
1837 
1838 void glue(helper_roundss, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1839                                   uint32_t mode)
1840 {
1841     signed char prev_rounding_mode;
1842 
1843     prev_rounding_mode = env->sse_status.float_rounding_mode;
1844     if (!(mode & (1 << 2))) {
1845         switch (mode & 3) {
1846         case 0:
1847             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1848             break;
1849         case 1:
1850             set_float_rounding_mode(float_round_down, &env->sse_status);
1851             break;
1852         case 2:
1853             set_float_rounding_mode(float_round_up, &env->sse_status);
1854             break;
1855         case 3:
1856             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1857             break;
1858         }
1859     }
1860 
1861     d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
1862 
1863 #if 0 /* TODO */
1864     if (mode & (1 << 3)) {
1865         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1866                                   ~float_flag_inexact,
1867                                   &env->sse_status);
1868     }
1869 #endif
1870     env->sse_status.float_rounding_mode = prev_rounding_mode;
1871 }
1872 
1873 void glue(helper_roundsd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1874                                   uint32_t mode)
1875 {
1876     signed char prev_rounding_mode;
1877 
1878     prev_rounding_mode = env->sse_status.float_rounding_mode;
1879     if (!(mode & (1 << 2))) {
1880         switch (mode & 3) {
1881         case 0:
1882             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1883             break;
1884         case 1:
1885             set_float_rounding_mode(float_round_down, &env->sse_status);
1886             break;
1887         case 2:
1888             set_float_rounding_mode(float_round_up, &env->sse_status);
1889             break;
1890         case 3:
1891             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1892             break;
1893         }
1894     }
1895 
1896     d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
1897 
1898 #if 0 /* TODO */
1899     if (mode & (1 << 3)) {
1900         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1901                                   ~float_flag_inexact,
1902                                   &env->sse_status);
1903     }
1904 #endif
1905     env->sse_status.float_rounding_mode = prev_rounding_mode;
1906 }
1907 
1908 #define FBLENDP(d, s, m) (m ? s : d)
1909 SSE_HELPER_I(helper_blendps, L, 4, FBLENDP)
1910 SSE_HELPER_I(helper_blendpd, Q, 2, FBLENDP)
1911 SSE_HELPER_I(helper_pblendw, W, 8, FBLENDP)
1912 
1913 void glue(helper_dpps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1914 {
1915     float32 iresult = float32_zero;
1916 
1917     if (mask & (1 << 4)) {
1918         iresult = float32_add(iresult,
1919                               float32_mul(d->ZMM_S(0), s->ZMM_S(0),
1920                                           &env->sse_status),
1921                               &env->sse_status);
1922     }
1923     if (mask & (1 << 5)) {
1924         iresult = float32_add(iresult,
1925                               float32_mul(d->ZMM_S(1), s->ZMM_S(1),
1926                                           &env->sse_status),
1927                               &env->sse_status);
1928     }
1929     if (mask & (1 << 6)) {
1930         iresult = float32_add(iresult,
1931                               float32_mul(d->ZMM_S(2), s->ZMM_S(2),
1932                                           &env->sse_status),
1933                               &env->sse_status);
1934     }
1935     if (mask & (1 << 7)) {
1936         iresult = float32_add(iresult,
1937                               float32_mul(d->ZMM_S(3), s->ZMM_S(3),
1938                                           &env->sse_status),
1939                               &env->sse_status);
1940     }
1941     d->ZMM_S(0) = (mask & (1 << 0)) ? iresult : float32_zero;
1942     d->ZMM_S(1) = (mask & (1 << 1)) ? iresult : float32_zero;
1943     d->ZMM_S(2) = (mask & (1 << 2)) ? iresult : float32_zero;
1944     d->ZMM_S(3) = (mask & (1 << 3)) ? iresult : float32_zero;
1945 }
1946 
1947 void glue(helper_dppd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1948 {
1949     float64 iresult = float64_zero;
1950 
1951     if (mask & (1 << 4)) {
1952         iresult = float64_add(iresult,
1953                               float64_mul(d->ZMM_D(0), s->ZMM_D(0),
1954                                           &env->sse_status),
1955                               &env->sse_status);
1956     }
1957     if (mask & (1 << 5)) {
1958         iresult = float64_add(iresult,
1959                               float64_mul(d->ZMM_D(1), s->ZMM_D(1),
1960                                           &env->sse_status),
1961                               &env->sse_status);
1962     }
1963     d->ZMM_D(0) = (mask & (1 << 0)) ? iresult : float64_zero;
1964     d->ZMM_D(1) = (mask & (1 << 1)) ? iresult : float64_zero;
1965 }
1966 
1967 void glue(helper_mpsadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1968                                   uint32_t offset)
1969 {
1970     int s0 = (offset & 3) << 2;
1971     int d0 = (offset & 4) << 0;
1972     int i;
1973     Reg r;
1974 
1975     for (i = 0; i < 8; i++, d0++) {
1976         r.W(i) = 0;
1977         r.W(i) += abs1(d->B(d0 + 0) - s->B(s0 + 0));
1978         r.W(i) += abs1(d->B(d0 + 1) - s->B(s0 + 1));
1979         r.W(i) += abs1(d->B(d0 + 2) - s->B(s0 + 2));
1980         r.W(i) += abs1(d->B(d0 + 3) - s->B(s0 + 3));
1981     }
1982 
1983     *d = r;
1984 }
1985 
1986 /* SSE4.2 op helpers */
1987 #define FCMPGTQ(d, s) ((int64_t)d > (int64_t)s ? -1 : 0)
1988 SSE_HELPER_Q(helper_pcmpgtq, FCMPGTQ)
1989 
1990 static inline int pcmp_elen(CPUX86State *env, int reg, uint32_t ctrl)
1991 {
1992     int val;
1993 
1994     /* Presence of REX.W is indicated by a bit higher than 7 set */
1995     if (ctrl >> 8) {
1996         val = abs1((int64_t)env->regs[reg]);
1997     } else {
1998         val = abs1((int32_t)env->regs[reg]);
1999     }
2000 
2001     if (ctrl & 1) {
2002         if (val > 8) {
2003             return 8;
2004         }
2005     } else {
2006         if (val > 16) {
2007             return 16;
2008         }
2009     }
2010     return val;
2011 }
2012 
2013 static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
2014 {
2015     int val = 0;
2016 
2017     if (ctrl & 1) {
2018         while (val < 8 && r->W(val)) {
2019             val++;
2020         }
2021     } else {
2022         while (val < 16 && r->B(val)) {
2023             val++;
2024         }
2025     }
2026 
2027     return val;
2028 }
2029 
2030 static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
2031 {
2032     switch ((ctrl >> 0) & 3) {
2033     case 0:
2034         return r->B(i);
2035     case 1:
2036         return r->W(i);
2037     case 2:
2038         return (int8_t)r->B(i);
2039     case 3:
2040     default:
2041         return (int16_t)r->W(i);
2042     }
2043 }
2044 
2045 static inline unsigned pcmpxstrx(CPUX86State *env, Reg *d, Reg *s,
2046                                  int8_t ctrl, int valids, int validd)
2047 {
2048     unsigned int res = 0;
2049     int v;
2050     int j, i;
2051     int upper = (ctrl & 1) ? 7 : 15;
2052 
2053     valids--;
2054     validd--;
2055 
2056     CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
2057 
2058     switch ((ctrl >> 2) & 3) {
2059     case 0:
2060         for (j = valids; j >= 0; j--) {
2061             res <<= 1;
2062             v = pcmp_val(s, ctrl, j);
2063             for (i = validd; i >= 0; i--) {
2064                 res |= (v == pcmp_val(d, ctrl, i));
2065             }
2066         }
2067         break;
2068     case 1:
2069         for (j = valids; j >= 0; j--) {
2070             res <<= 1;
2071             v = pcmp_val(s, ctrl, j);
2072             for (i = ((validd - 1) | 1); i >= 0; i -= 2) {
2073                 res |= (pcmp_val(d, ctrl, i - 0) >= v &&
2074                         pcmp_val(d, ctrl, i - 1) <= v);
2075             }
2076         }
2077         break;
2078     case 2:
2079         res = (1 << (upper - MAX(valids, validd))) - 1;
2080         res <<= MAX(valids, validd) - MIN(valids, validd);
2081         for (i = MIN(valids, validd); i >= 0; i--) {
2082             res <<= 1;
2083             v = pcmp_val(s, ctrl, i);
2084             res |= (v == pcmp_val(d, ctrl, i));
2085         }
2086         break;
2087     case 3:
2088         if (validd == -1) {
2089             res = (2 << upper) - 1;
2090             break;
2091         }
2092         for (j = valids - validd; j >= 0; j--) {
2093             res <<= 1;
2094             v = 1;
2095             for (i = validd; i >= 0; i--) {
2096                 v &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
2097             }
2098             res |= v;
2099         }
2100         break;
2101     }
2102 
2103     switch ((ctrl >> 4) & 3) {
2104     case 1:
2105         res ^= (2 << upper) - 1;
2106         break;
2107     case 3:
2108         res ^= (1 << (valids + 1)) - 1;
2109         break;
2110     }
2111 
2112     if (res) {
2113         CC_SRC |= CC_C;
2114     }
2115     if (res & 1) {
2116         CC_SRC |= CC_O;
2117     }
2118 
2119     return res;
2120 }
2121 
2122 void glue(helper_pcmpestri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2123                                     uint32_t ctrl)
2124 {
2125     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2126                                  pcmp_elen(env, R_EDX, ctrl),
2127                                  pcmp_elen(env, R_EAX, ctrl));
2128 
2129     if (res) {
2130         env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2131     } else {
2132         env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2133     }
2134 }
2135 
2136 void glue(helper_pcmpestrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2137                                     uint32_t ctrl)
2138 {
2139     int i;
2140     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2141                                  pcmp_elen(env, R_EDX, ctrl),
2142                                  pcmp_elen(env, R_EAX, ctrl));
2143 
2144     if ((ctrl >> 6) & 1) {
2145         if (ctrl & 1) {
2146             for (i = 0; i < 8; i++, res >>= 1) {
2147                 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2148             }
2149         } else {
2150             for (i = 0; i < 16; i++, res >>= 1) {
2151                 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2152             }
2153         }
2154     } else {
2155         env->xmm_regs[0].Q(1) = 0;
2156         env->xmm_regs[0].Q(0) = res;
2157     }
2158 }
2159 
2160 void glue(helper_pcmpistri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2161                                     uint32_t ctrl)
2162 {
2163     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2164                                  pcmp_ilen(s, ctrl),
2165                                  pcmp_ilen(d, ctrl));
2166 
2167     if (res) {
2168         env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2169     } else {
2170         env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2171     }
2172 }
2173 
2174 void glue(helper_pcmpistrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2175                                     uint32_t ctrl)
2176 {
2177     int i;
2178     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2179                                  pcmp_ilen(s, ctrl),
2180                                  pcmp_ilen(d, ctrl));
2181 
2182     if ((ctrl >> 6) & 1) {
2183         if (ctrl & 1) {
2184             for (i = 0; i < 8; i++, res >>= 1) {
2185                 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2186             }
2187         } else {
2188             for (i = 0; i < 16; i++, res >>= 1) {
2189                 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2190             }
2191         }
2192     } else {
2193         env->xmm_regs[0].Q(1) = 0;
2194         env->xmm_regs[0].Q(0) = res;
2195     }
2196 }
2197 
2198 #define CRCPOLY        0x1edc6f41
2199 #define CRCPOLY_BITREV 0x82f63b78
2200 target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
2201 {
2202     target_ulong crc = (msg & ((target_ulong) -1 >>
2203                                (TARGET_LONG_BITS - len))) ^ crc1;
2204 
2205     while (len--) {
2206         crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
2207     }
2208 
2209     return crc;
2210 }
2211 
2212 void glue(helper_pclmulqdq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2213                                     uint32_t ctrl)
2214 {
2215     uint64_t ah, al, b, resh, resl;
2216 
2217     ah = 0;
2218     al = d->Q((ctrl & 1) != 0);
2219     b = s->Q((ctrl & 16) != 0);
2220     resh = resl = 0;
2221 
2222     while (b) {
2223         if (b & 1) {
2224             resl ^= al;
2225             resh ^= ah;
2226         }
2227         ah = (ah << 1) | (al >> 63);
2228         al <<= 1;
2229         b >>= 1;
2230     }
2231 
2232     d->Q(0) = resl;
2233     d->Q(1) = resh;
2234 }
2235 
2236 void glue(helper_aesdec, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2237 {
2238     int i;
2239     Reg st = *d;
2240     Reg rk = *s;
2241 
2242     for (i = 0 ; i < 4 ; i++) {
2243         d->L(i) = rk.L(i) ^ bswap32(AES_Td0[st.B(AES_ishifts[4*i+0])] ^
2244                                     AES_Td1[st.B(AES_ishifts[4*i+1])] ^
2245                                     AES_Td2[st.B(AES_ishifts[4*i+2])] ^
2246                                     AES_Td3[st.B(AES_ishifts[4*i+3])]);
2247     }
2248 }
2249 
2250 void glue(helper_aesdeclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2251 {
2252     int i;
2253     Reg st = *d;
2254     Reg rk = *s;
2255 
2256     for (i = 0; i < 16; i++) {
2257         d->B(i) = rk.B(i) ^ (AES_isbox[st.B(AES_ishifts[i])]);
2258     }
2259 }
2260 
2261 void glue(helper_aesenc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2262 {
2263     int i;
2264     Reg st = *d;
2265     Reg rk = *s;
2266 
2267     for (i = 0 ; i < 4 ; i++) {
2268         d->L(i) = rk.L(i) ^ bswap32(AES_Te0[st.B(AES_shifts[4*i+0])] ^
2269                                     AES_Te1[st.B(AES_shifts[4*i+1])] ^
2270                                     AES_Te2[st.B(AES_shifts[4*i+2])] ^
2271                                     AES_Te3[st.B(AES_shifts[4*i+3])]);
2272     }
2273 }
2274 
2275 void glue(helper_aesenclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2276 {
2277     int i;
2278     Reg st = *d;
2279     Reg rk = *s;
2280 
2281     for (i = 0; i < 16; i++) {
2282         d->B(i) = rk.B(i) ^ (AES_sbox[st.B(AES_shifts[i])]);
2283     }
2284 
2285 }
2286 
2287 void glue(helper_aesimc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2288 {
2289     int i;
2290     Reg tmp = *s;
2291 
2292     for (i = 0 ; i < 4 ; i++) {
2293         d->L(i) = bswap32(AES_imc[tmp.B(4*i+0)][0] ^
2294                           AES_imc[tmp.B(4*i+1)][1] ^
2295                           AES_imc[tmp.B(4*i+2)][2] ^
2296                           AES_imc[tmp.B(4*i+3)][3]);
2297     }
2298 }
2299 
2300 void glue(helper_aeskeygenassist, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2301                                           uint32_t ctrl)
2302 {
2303     int i;
2304     Reg tmp = *s;
2305 
2306     for (i = 0 ; i < 4 ; i++) {
2307         d->B(i) = AES_sbox[tmp.B(i + 4)];
2308         d->B(i + 8) = AES_sbox[tmp.B(i + 12)];
2309     }
2310     d->L(1) = (d->L(0) << 24 | d->L(0) >> 8) ^ ctrl;
2311     d->L(3) = (d->L(2) << 24 | d->L(2) >> 8) ^ ctrl;
2312 }
2313 #endif
2314 
2315 #undef SHIFT
2316 #undef XMM_ONLY
2317 #undef Reg
2318 #undef B
2319 #undef W
2320 #undef L
2321 #undef Q
2322 #undef SUFFIX
2323