xref: /openbmc/qemu/target/i386/ops_sse.h (revision b91a0fa7)
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 2.1 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     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
847     d->ZMM_S(0) = float32_div(float32_one,
848                               float32_sqrt(s->ZMM_S(0), &env->sse_status),
849                               &env->sse_status);
850     d->ZMM_S(1) = float32_div(float32_one,
851                               float32_sqrt(s->ZMM_S(1), &env->sse_status),
852                               &env->sse_status);
853     d->ZMM_S(2) = float32_div(float32_one,
854                               float32_sqrt(s->ZMM_S(2), &env->sse_status),
855                               &env->sse_status);
856     d->ZMM_S(3) = float32_div(float32_one,
857                               float32_sqrt(s->ZMM_S(3), &env->sse_status),
858                               &env->sse_status);
859     set_float_exception_flags(old_flags, &env->sse_status);
860 }
861 
862 void helper_rsqrtss(CPUX86State *env, ZMMReg *d, ZMMReg *s)
863 {
864     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
865     d->ZMM_S(0) = float32_div(float32_one,
866                               float32_sqrt(s->ZMM_S(0), &env->sse_status),
867                               &env->sse_status);
868     set_float_exception_flags(old_flags, &env->sse_status);
869 }
870 
871 void helper_rcpps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
872 {
873     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
874     d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
875     d->ZMM_S(1) = float32_div(float32_one, s->ZMM_S(1), &env->sse_status);
876     d->ZMM_S(2) = float32_div(float32_one, s->ZMM_S(2), &env->sse_status);
877     d->ZMM_S(3) = float32_div(float32_one, s->ZMM_S(3), &env->sse_status);
878     set_float_exception_flags(old_flags, &env->sse_status);
879 }
880 
881 void helper_rcpss(CPUX86State *env, ZMMReg *d, ZMMReg *s)
882 {
883     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
884     d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
885     set_float_exception_flags(old_flags, &env->sse_status);
886 }
887 
888 static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
889 {
890     uint64_t mask;
891 
892     if (len == 0) {
893         mask = ~0LL;
894     } else {
895         mask = (1ULL << len) - 1;
896     }
897     return (src >> shift) & mask;
898 }
899 
900 void helper_extrq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
901 {
902     d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), s->ZMM_B(1), s->ZMM_B(0));
903 }
904 
905 void helper_extrq_i(CPUX86State *env, ZMMReg *d, int index, int length)
906 {
907     d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), index, length);
908 }
909 
910 static inline uint64_t helper_insertq(uint64_t src, int shift, int len)
911 {
912     uint64_t mask;
913 
914     if (len == 0) {
915         mask = ~0ULL;
916     } else {
917         mask = (1ULL << len) - 1;
918     }
919     return (src & ~(mask << shift)) | ((src & mask) << shift);
920 }
921 
922 void helper_insertq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
923 {
924     d->ZMM_Q(0) = helper_insertq(s->ZMM_Q(0), s->ZMM_B(9), s->ZMM_B(8));
925 }
926 
927 void helper_insertq_i(CPUX86State *env, ZMMReg *d, int index, int length)
928 {
929     d->ZMM_Q(0) = helper_insertq(d->ZMM_Q(0), index, length);
930 }
931 
932 void helper_haddps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
933 {
934     ZMMReg r;
935 
936     r.ZMM_S(0) = float32_add(d->ZMM_S(0), d->ZMM_S(1), &env->sse_status);
937     r.ZMM_S(1) = float32_add(d->ZMM_S(2), d->ZMM_S(3), &env->sse_status);
938     r.ZMM_S(2) = float32_add(s->ZMM_S(0), s->ZMM_S(1), &env->sse_status);
939     r.ZMM_S(3) = float32_add(s->ZMM_S(2), s->ZMM_S(3), &env->sse_status);
940     *d = r;
941 }
942 
943 void helper_haddpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
944 {
945     ZMMReg r;
946 
947     r.ZMM_D(0) = float64_add(d->ZMM_D(0), d->ZMM_D(1), &env->sse_status);
948     r.ZMM_D(1) = float64_add(s->ZMM_D(0), s->ZMM_D(1), &env->sse_status);
949     *d = r;
950 }
951 
952 void helper_hsubps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
953 {
954     ZMMReg r;
955 
956     r.ZMM_S(0) = float32_sub(d->ZMM_S(0), d->ZMM_S(1), &env->sse_status);
957     r.ZMM_S(1) = float32_sub(d->ZMM_S(2), d->ZMM_S(3), &env->sse_status);
958     r.ZMM_S(2) = float32_sub(s->ZMM_S(0), s->ZMM_S(1), &env->sse_status);
959     r.ZMM_S(3) = float32_sub(s->ZMM_S(2), s->ZMM_S(3), &env->sse_status);
960     *d = r;
961 }
962 
963 void helper_hsubpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
964 {
965     ZMMReg r;
966 
967     r.ZMM_D(0) = float64_sub(d->ZMM_D(0), d->ZMM_D(1), &env->sse_status);
968     r.ZMM_D(1) = float64_sub(s->ZMM_D(0), s->ZMM_D(1), &env->sse_status);
969     *d = r;
970 }
971 
972 void helper_addsubps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
973 {
974     d->ZMM_S(0) = float32_sub(d->ZMM_S(0), s->ZMM_S(0), &env->sse_status);
975     d->ZMM_S(1) = float32_add(d->ZMM_S(1), s->ZMM_S(1), &env->sse_status);
976     d->ZMM_S(2) = float32_sub(d->ZMM_S(2), s->ZMM_S(2), &env->sse_status);
977     d->ZMM_S(3) = float32_add(d->ZMM_S(3), s->ZMM_S(3), &env->sse_status);
978 }
979 
980 void helper_addsubpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
981 {
982     d->ZMM_D(0) = float64_sub(d->ZMM_D(0), s->ZMM_D(0), &env->sse_status);
983     d->ZMM_D(1) = float64_add(d->ZMM_D(1), s->ZMM_D(1), &env->sse_status);
984 }
985 
986 /* XXX: unordered */
987 #define SSE_HELPER_CMP(name, F)                                         \
988     void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s)        \
989     {                                                                   \
990         d->ZMM_L(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
991         d->ZMM_L(1) = F(32, d->ZMM_S(1), s->ZMM_S(1));                  \
992         d->ZMM_L(2) = F(32, d->ZMM_S(2), s->ZMM_S(2));                  \
993         d->ZMM_L(3) = F(32, d->ZMM_S(3), s->ZMM_S(3));                  \
994     }                                                                   \
995                                                                         \
996     void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s)        \
997     {                                                                   \
998         d->ZMM_L(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
999     }                                                                   \
1000                                                                         \
1001     void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s)        \
1002     {                                                                   \
1003         d->ZMM_Q(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
1004         d->ZMM_Q(1) = F(64, d->ZMM_D(1), s->ZMM_D(1));                  \
1005     }                                                                   \
1006                                                                         \
1007     void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s)        \
1008     {                                                                   \
1009         d->ZMM_Q(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
1010     }
1011 
1012 #define FPU_CMPEQ(size, a, b)                                           \
1013     (float ## size ## _eq_quiet(a, b, &env->sse_status) ? -1 : 0)
1014 #define FPU_CMPLT(size, a, b)                                           \
1015     (float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0)
1016 #define FPU_CMPLE(size, a, b)                                           \
1017     (float ## size ## _le(a, b, &env->sse_status) ? -1 : 0)
1018 #define FPU_CMPUNORD(size, a, b)                                        \
1019     (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? -1 : 0)
1020 #define FPU_CMPNEQ(size, a, b)                                          \
1021     (float ## size ## _eq_quiet(a, b, &env->sse_status) ? 0 : -1)
1022 #define FPU_CMPNLT(size, a, b)                                          \
1023     (float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1)
1024 #define FPU_CMPNLE(size, a, b)                                          \
1025     (float ## size ## _le(a, b, &env->sse_status) ? 0 : -1)
1026 #define FPU_CMPORD(size, a, b)                                          \
1027     (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? 0 : -1)
1028 
1029 SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
1030 SSE_HELPER_CMP(cmplt, FPU_CMPLT)
1031 SSE_HELPER_CMP(cmple, FPU_CMPLE)
1032 SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
1033 SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
1034 SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
1035 SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
1036 SSE_HELPER_CMP(cmpord, FPU_CMPORD)
1037 
1038 static const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
1039 
1040 void helper_ucomiss(CPUX86State *env, Reg *d, Reg *s)
1041 {
1042     FloatRelation ret;
1043     float32 s0, s1;
1044 
1045     s0 = d->ZMM_S(0);
1046     s1 = s->ZMM_S(0);
1047     ret = float32_compare_quiet(s0, s1, &env->sse_status);
1048     CC_SRC = comis_eflags[ret + 1];
1049 }
1050 
1051 void helper_comiss(CPUX86State *env, Reg *d, Reg *s)
1052 {
1053     FloatRelation ret;
1054     float32 s0, s1;
1055 
1056     s0 = d->ZMM_S(0);
1057     s1 = s->ZMM_S(0);
1058     ret = float32_compare(s0, s1, &env->sse_status);
1059     CC_SRC = comis_eflags[ret + 1];
1060 }
1061 
1062 void helper_ucomisd(CPUX86State *env, Reg *d, Reg *s)
1063 {
1064     FloatRelation ret;
1065     float64 d0, d1;
1066 
1067     d0 = d->ZMM_D(0);
1068     d1 = s->ZMM_D(0);
1069     ret = float64_compare_quiet(d0, d1, &env->sse_status);
1070     CC_SRC = comis_eflags[ret + 1];
1071 }
1072 
1073 void helper_comisd(CPUX86State *env, Reg *d, Reg *s)
1074 {
1075     FloatRelation ret;
1076     float64 d0, d1;
1077 
1078     d0 = d->ZMM_D(0);
1079     d1 = s->ZMM_D(0);
1080     ret = float64_compare(d0, d1, &env->sse_status);
1081     CC_SRC = comis_eflags[ret + 1];
1082 }
1083 
1084 uint32_t helper_movmskps(CPUX86State *env, Reg *s)
1085 {
1086     int b0, b1, b2, b3;
1087 
1088     b0 = s->ZMM_L(0) >> 31;
1089     b1 = s->ZMM_L(1) >> 31;
1090     b2 = s->ZMM_L(2) >> 31;
1091     b3 = s->ZMM_L(3) >> 31;
1092     return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
1093 }
1094 
1095 uint32_t helper_movmskpd(CPUX86State *env, Reg *s)
1096 {
1097     int b0, b1;
1098 
1099     b0 = s->ZMM_L(1) >> 31;
1100     b1 = s->ZMM_L(3) >> 31;
1101     return b0 | (b1 << 1);
1102 }
1103 
1104 #endif
1105 
1106 uint32_t glue(helper_pmovmskb, SUFFIX)(CPUX86State *env, Reg *s)
1107 {
1108     uint32_t val;
1109 
1110     val = 0;
1111     val |= (s->B(0) >> 7);
1112     val |= (s->B(1) >> 6) & 0x02;
1113     val |= (s->B(2) >> 5) & 0x04;
1114     val |= (s->B(3) >> 4) & 0x08;
1115     val |= (s->B(4) >> 3) & 0x10;
1116     val |= (s->B(5) >> 2) & 0x20;
1117     val |= (s->B(6) >> 1) & 0x40;
1118     val |= (s->B(7)) & 0x80;
1119 #if SHIFT == 1
1120     val |= (s->B(8) << 1) & 0x0100;
1121     val |= (s->B(9) << 2) & 0x0200;
1122     val |= (s->B(10) << 3) & 0x0400;
1123     val |= (s->B(11) << 4) & 0x0800;
1124     val |= (s->B(12) << 5) & 0x1000;
1125     val |= (s->B(13) << 6) & 0x2000;
1126     val |= (s->B(14) << 7) & 0x4000;
1127     val |= (s->B(15) << 8) & 0x8000;
1128 #endif
1129     return val;
1130 }
1131 
1132 void glue(helper_packsswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1133 {
1134     Reg r;
1135 
1136     r.B(0) = satsb((int16_t)d->W(0));
1137     r.B(1) = satsb((int16_t)d->W(1));
1138     r.B(2) = satsb((int16_t)d->W(2));
1139     r.B(3) = satsb((int16_t)d->W(3));
1140 #if SHIFT == 1
1141     r.B(4) = satsb((int16_t)d->W(4));
1142     r.B(5) = satsb((int16_t)d->W(5));
1143     r.B(6) = satsb((int16_t)d->W(6));
1144     r.B(7) = satsb((int16_t)d->W(7));
1145 #endif
1146     r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1147     r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1148     r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1149     r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1150 #if SHIFT == 1
1151     r.B(12) = satsb((int16_t)s->W(4));
1152     r.B(13) = satsb((int16_t)s->W(5));
1153     r.B(14) = satsb((int16_t)s->W(6));
1154     r.B(15) = satsb((int16_t)s->W(7));
1155 #endif
1156     *d = r;
1157 }
1158 
1159 void glue(helper_packuswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1160 {
1161     Reg r;
1162 
1163     r.B(0) = satub((int16_t)d->W(0));
1164     r.B(1) = satub((int16_t)d->W(1));
1165     r.B(2) = satub((int16_t)d->W(2));
1166     r.B(3) = satub((int16_t)d->W(3));
1167 #if SHIFT == 1
1168     r.B(4) = satub((int16_t)d->W(4));
1169     r.B(5) = satub((int16_t)d->W(5));
1170     r.B(6) = satub((int16_t)d->W(6));
1171     r.B(7) = satub((int16_t)d->W(7));
1172 #endif
1173     r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1174     r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1175     r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1176     r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1177 #if SHIFT == 1
1178     r.B(12) = satub((int16_t)s->W(4));
1179     r.B(13) = satub((int16_t)s->W(5));
1180     r.B(14) = satub((int16_t)s->W(6));
1181     r.B(15) = satub((int16_t)s->W(7));
1182 #endif
1183     *d = r;
1184 }
1185 
1186 void glue(helper_packssdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1187 {
1188     Reg r;
1189 
1190     r.W(0) = satsw(d->L(0));
1191     r.W(1) = satsw(d->L(1));
1192 #if SHIFT == 1
1193     r.W(2) = satsw(d->L(2));
1194     r.W(3) = satsw(d->L(3));
1195 #endif
1196     r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1197     r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1198 #if SHIFT == 1
1199     r.W(6) = satsw(s->L(2));
1200     r.W(7) = satsw(s->L(3));
1201 #endif
1202     *d = r;
1203 }
1204 
1205 #define UNPCK_OP(base_name, base)                                       \
1206                                                                         \
1207     void glue(helper_punpck ## base_name ## bw, SUFFIX)(CPUX86State *env,\
1208                                                         Reg *d, Reg *s) \
1209     {                                                                   \
1210         Reg r;                                                          \
1211                                                                         \
1212         r.B(0) = d->B((base << (SHIFT + 2)) + 0);                       \
1213         r.B(1) = s->B((base << (SHIFT + 2)) + 0);                       \
1214         r.B(2) = d->B((base << (SHIFT + 2)) + 1);                       \
1215         r.B(3) = s->B((base << (SHIFT + 2)) + 1);                       \
1216         r.B(4) = d->B((base << (SHIFT + 2)) + 2);                       \
1217         r.B(5) = s->B((base << (SHIFT + 2)) + 2);                       \
1218         r.B(6) = d->B((base << (SHIFT + 2)) + 3);                       \
1219         r.B(7) = s->B((base << (SHIFT + 2)) + 3);                       \
1220         XMM_ONLY(                                                       \
1221                  r.B(8) = d->B((base << (SHIFT + 2)) + 4);              \
1222                  r.B(9) = s->B((base << (SHIFT + 2)) + 4);              \
1223                  r.B(10) = d->B((base << (SHIFT + 2)) + 5);             \
1224                  r.B(11) = s->B((base << (SHIFT + 2)) + 5);             \
1225                  r.B(12) = d->B((base << (SHIFT + 2)) + 6);             \
1226                  r.B(13) = s->B((base << (SHIFT + 2)) + 6);             \
1227                  r.B(14) = d->B((base << (SHIFT + 2)) + 7);             \
1228                  r.B(15) = s->B((base << (SHIFT + 2)) + 7);             \
1229                                                                       ) \
1230             *d = r;                                                     \
1231     }                                                                   \
1232                                                                         \
1233     void glue(helper_punpck ## base_name ## wd, SUFFIX)(CPUX86State *env,\
1234                                                         Reg *d, Reg *s) \
1235     {                                                                   \
1236         Reg r;                                                          \
1237                                                                         \
1238         r.W(0) = d->W((base << (SHIFT + 1)) + 0);                       \
1239         r.W(1) = s->W((base << (SHIFT + 1)) + 0);                       \
1240         r.W(2) = d->W((base << (SHIFT + 1)) + 1);                       \
1241         r.W(3) = s->W((base << (SHIFT + 1)) + 1);                       \
1242         XMM_ONLY(                                                       \
1243                  r.W(4) = d->W((base << (SHIFT + 1)) + 2);              \
1244                  r.W(5) = s->W((base << (SHIFT + 1)) + 2);              \
1245                  r.W(6) = d->W((base << (SHIFT + 1)) + 3);              \
1246                  r.W(7) = s->W((base << (SHIFT + 1)) + 3);              \
1247                                                                       ) \
1248             *d = r;                                                     \
1249     }                                                                   \
1250                                                                         \
1251     void glue(helper_punpck ## base_name ## dq, SUFFIX)(CPUX86State *env,\
1252                                                         Reg *d, Reg *s) \
1253     {                                                                   \
1254         Reg r;                                                          \
1255                                                                         \
1256         r.L(0) = d->L((base << SHIFT) + 0);                             \
1257         r.L(1) = s->L((base << SHIFT) + 0);                             \
1258         XMM_ONLY(                                                       \
1259                  r.L(2) = d->L((base << SHIFT) + 1);                    \
1260                  r.L(3) = s->L((base << SHIFT) + 1);                    \
1261                                                                       ) \
1262             *d = r;                                                     \
1263     }                                                                   \
1264                                                                         \
1265     XMM_ONLY(                                                           \
1266              void glue(helper_punpck ## base_name ## qdq, SUFFIX)(CPUX86State \
1267                                                                   *env, \
1268                                                                   Reg *d, \
1269                                                                   Reg *s) \
1270              {                                                          \
1271                  Reg r;                                                 \
1272                                                                         \
1273                  r.Q(0) = d->Q(base);                                   \
1274                  r.Q(1) = s->Q(base);                                   \
1275                  *d = r;                                                \
1276              }                                                          \
1277                                                                         )
1278 
1279 UNPCK_OP(l, 0)
1280 UNPCK_OP(h, 1)
1281 
1282 /* 3DNow! float ops */
1283 #if SHIFT == 0
1284 void helper_pi2fd(CPUX86State *env, MMXReg *d, MMXReg *s)
1285 {
1286     d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
1287     d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
1288 }
1289 
1290 void helper_pi2fw(CPUX86State *env, MMXReg *d, MMXReg *s)
1291 {
1292     d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
1293     d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
1294 }
1295 
1296 void helper_pf2id(CPUX86State *env, MMXReg *d, MMXReg *s)
1297 {
1298     d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
1299     d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
1300 }
1301 
1302 void helper_pf2iw(CPUX86State *env, MMXReg *d, MMXReg *s)
1303 {
1304     d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0),
1305                                                        &env->mmx_status));
1306     d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1),
1307                                                        &env->mmx_status));
1308 }
1309 
1310 void helper_pfacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1311 {
1312     MMXReg r;
1313 
1314     r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1315     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1316     *d = r;
1317 }
1318 
1319 void helper_pfadd(CPUX86State *env, MMXReg *d, MMXReg *s)
1320 {
1321     d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1322     d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1323 }
1324 
1325 void helper_pfcmpeq(CPUX86State *env, MMXReg *d, MMXReg *s)
1326 {
1327     d->MMX_L(0) = float32_eq_quiet(d->MMX_S(0), s->MMX_S(0),
1328                                    &env->mmx_status) ? -1 : 0;
1329     d->MMX_L(1) = float32_eq_quiet(d->MMX_S(1), s->MMX_S(1),
1330                                    &env->mmx_status) ? -1 : 0;
1331 }
1332 
1333 void helper_pfcmpge(CPUX86State *env, MMXReg *d, MMXReg *s)
1334 {
1335     d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0),
1336                              &env->mmx_status) ? -1 : 0;
1337     d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1),
1338                              &env->mmx_status) ? -1 : 0;
1339 }
1340 
1341 void helper_pfcmpgt(CPUX86State *env, MMXReg *d, MMXReg *s)
1342 {
1343     d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0),
1344                              &env->mmx_status) ? -1 : 0;
1345     d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1),
1346                              &env->mmx_status) ? -1 : 0;
1347 }
1348 
1349 void helper_pfmax(CPUX86State *env, MMXReg *d, MMXReg *s)
1350 {
1351     if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status)) {
1352         d->MMX_S(0) = s->MMX_S(0);
1353     }
1354     if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status)) {
1355         d->MMX_S(1) = s->MMX_S(1);
1356     }
1357 }
1358 
1359 void helper_pfmin(CPUX86State *env, MMXReg *d, MMXReg *s)
1360 {
1361     if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status)) {
1362         d->MMX_S(0) = s->MMX_S(0);
1363     }
1364     if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status)) {
1365         d->MMX_S(1) = s->MMX_S(1);
1366     }
1367 }
1368 
1369 void helper_pfmul(CPUX86State *env, MMXReg *d, MMXReg *s)
1370 {
1371     d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1372     d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1373 }
1374 
1375 void helper_pfnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1376 {
1377     MMXReg r;
1378 
1379     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1380     r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1381     *d = r;
1382 }
1383 
1384 void helper_pfpnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1385 {
1386     MMXReg r;
1387 
1388     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1389     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1390     *d = r;
1391 }
1392 
1393 void helper_pfrcp(CPUX86State *env, MMXReg *d, MMXReg *s)
1394 {
1395     d->MMX_S(0) = float32_div(float32_one, s->MMX_S(0), &env->mmx_status);
1396     d->MMX_S(1) = d->MMX_S(0);
1397 }
1398 
1399 void helper_pfrsqrt(CPUX86State *env, MMXReg *d, MMXReg *s)
1400 {
1401     d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
1402     d->MMX_S(1) = float32_div(float32_one,
1403                               float32_sqrt(d->MMX_S(1), &env->mmx_status),
1404                               &env->mmx_status);
1405     d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
1406     d->MMX_L(0) = d->MMX_L(1);
1407 }
1408 
1409 void helper_pfsub(CPUX86State *env, MMXReg *d, MMXReg *s)
1410 {
1411     d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1412     d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1413 }
1414 
1415 void helper_pfsubr(CPUX86State *env, MMXReg *d, MMXReg *s)
1416 {
1417     d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
1418     d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
1419 }
1420 
1421 void helper_pswapd(CPUX86State *env, MMXReg *d, MMXReg *s)
1422 {
1423     MMXReg r;
1424 
1425     r.MMX_L(0) = s->MMX_L(1);
1426     r.MMX_L(1) = s->MMX_L(0);
1427     *d = r;
1428 }
1429 #endif
1430 
1431 /* SSSE3 op helpers */
1432 void glue(helper_pshufb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1433 {
1434     int i;
1435     Reg r;
1436 
1437     for (i = 0; i < (8 << SHIFT); i++) {
1438         r.B(i) = (s->B(i) & 0x80) ? 0 : (d->B(s->B(i) & ((8 << SHIFT) - 1)));
1439     }
1440 
1441     *d = r;
1442 }
1443 
1444 void glue(helper_phaddw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1445 {
1446 
1447     Reg r;
1448 
1449     r.W(0) = (int16_t)d->W(0) + (int16_t)d->W(1);
1450     r.W(1) = (int16_t)d->W(2) + (int16_t)d->W(3);
1451     XMM_ONLY(r.W(2) = (int16_t)d->W(4) + (int16_t)d->W(5));
1452     XMM_ONLY(r.W(3) = (int16_t)d->W(6) + (int16_t)d->W(7));
1453     r.W((2 << SHIFT) + 0) = (int16_t)s->W(0) + (int16_t)s->W(1);
1454     r.W((2 << SHIFT) + 1) = (int16_t)s->W(2) + (int16_t)s->W(3);
1455     XMM_ONLY(r.W(6) = (int16_t)s->W(4) + (int16_t)s->W(5));
1456     XMM_ONLY(r.W(7) = (int16_t)s->W(6) + (int16_t)s->W(7));
1457 
1458     *d = r;
1459 }
1460 
1461 void glue(helper_phaddd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1462 {
1463     Reg r;
1464 
1465     r.L(0) = (int32_t)d->L(0) + (int32_t)d->L(1);
1466     XMM_ONLY(r.L(1) = (int32_t)d->L(2) + (int32_t)d->L(3));
1467     r.L((1 << SHIFT) + 0) = (int32_t)s->L(0) + (int32_t)s->L(1);
1468     XMM_ONLY(r.L(3) = (int32_t)s->L(2) + (int32_t)s->L(3));
1469 
1470     *d = r;
1471 }
1472 
1473 void glue(helper_phaddsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1474 {
1475     Reg r;
1476 
1477     r.W(0) = satsw((int16_t)d->W(0) + (int16_t)d->W(1));
1478     r.W(1) = satsw((int16_t)d->W(2) + (int16_t)d->W(3));
1479     XMM_ONLY(r.W(2) = satsw((int16_t)d->W(4) + (int16_t)d->W(5)));
1480     XMM_ONLY(r.W(3) = satsw((int16_t)d->W(6) + (int16_t)d->W(7)));
1481     r.W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) + (int16_t)s->W(1));
1482     r.W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) + (int16_t)s->W(3));
1483     XMM_ONLY(r.W(6) = satsw((int16_t)s->W(4) + (int16_t)s->W(5)));
1484     XMM_ONLY(r.W(7) = satsw((int16_t)s->W(6) + (int16_t)s->W(7)));
1485 
1486     *d = r;
1487 }
1488 
1489 void glue(helper_pmaddubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1490 {
1491     d->W(0) = satsw((int8_t)s->B(0) * (uint8_t)d->B(0) +
1492                     (int8_t)s->B(1) * (uint8_t)d->B(1));
1493     d->W(1) = satsw((int8_t)s->B(2) * (uint8_t)d->B(2) +
1494                     (int8_t)s->B(3) * (uint8_t)d->B(3));
1495     d->W(2) = satsw((int8_t)s->B(4) * (uint8_t)d->B(4) +
1496                     (int8_t)s->B(5) * (uint8_t)d->B(5));
1497     d->W(3) = satsw((int8_t)s->B(6) * (uint8_t)d->B(6) +
1498                     (int8_t)s->B(7) * (uint8_t)d->B(7));
1499 #if SHIFT == 1
1500     d->W(4) = satsw((int8_t)s->B(8) * (uint8_t)d->B(8) +
1501                     (int8_t)s->B(9) * (uint8_t)d->B(9));
1502     d->W(5) = satsw((int8_t)s->B(10) * (uint8_t)d->B(10) +
1503                     (int8_t)s->B(11) * (uint8_t)d->B(11));
1504     d->W(6) = satsw((int8_t)s->B(12) * (uint8_t)d->B(12) +
1505                     (int8_t)s->B(13) * (uint8_t)d->B(13));
1506     d->W(7) = satsw((int8_t)s->B(14) * (uint8_t)d->B(14) +
1507                     (int8_t)s->B(15) * (uint8_t)d->B(15));
1508 #endif
1509 }
1510 
1511 void glue(helper_phsubw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1512 {
1513     d->W(0) = (int16_t)d->W(0) - (int16_t)d->W(1);
1514     d->W(1) = (int16_t)d->W(2) - (int16_t)d->W(3);
1515     XMM_ONLY(d->W(2) = (int16_t)d->W(4) - (int16_t)d->W(5));
1516     XMM_ONLY(d->W(3) = (int16_t)d->W(6) - (int16_t)d->W(7));
1517     d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) - (int16_t)s->W(1);
1518     d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) - (int16_t)s->W(3);
1519     XMM_ONLY(d->W(6) = (int16_t)s->W(4) - (int16_t)s->W(5));
1520     XMM_ONLY(d->W(7) = (int16_t)s->W(6) - (int16_t)s->W(7));
1521 }
1522 
1523 void glue(helper_phsubd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1524 {
1525     d->L(0) = (int32_t)d->L(0) - (int32_t)d->L(1);
1526     XMM_ONLY(d->L(1) = (int32_t)d->L(2) - (int32_t)d->L(3));
1527     d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) - (int32_t)s->L(1);
1528     XMM_ONLY(d->L(3) = (int32_t)s->L(2) - (int32_t)s->L(3));
1529 }
1530 
1531 void glue(helper_phsubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1532 {
1533     d->W(0) = satsw((int16_t)d->W(0) - (int16_t)d->W(1));
1534     d->W(1) = satsw((int16_t)d->W(2) - (int16_t)d->W(3));
1535     XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) - (int16_t)d->W(5)));
1536     XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) - (int16_t)d->W(7)));
1537     d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) - (int16_t)s->W(1));
1538     d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) - (int16_t)s->W(3));
1539     XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) - (int16_t)s->W(5)));
1540     XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) - (int16_t)s->W(7)));
1541 }
1542 
1543 #define FABSB(_, x) (x > INT8_MAX  ? -(int8_t)x : x)
1544 #define FABSW(_, x) (x > INT16_MAX ? -(int16_t)x : x)
1545 #define FABSL(_, x) (x > INT32_MAX ? -(int32_t)x : x)
1546 SSE_HELPER_B(helper_pabsb, FABSB)
1547 SSE_HELPER_W(helper_pabsw, FABSW)
1548 SSE_HELPER_L(helper_pabsd, FABSL)
1549 
1550 #define FMULHRSW(d, s) (((int16_t) d * (int16_t)s + 0x4000) >> 15)
1551 SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
1552 
1553 #define FSIGNB(d, s) (s <= INT8_MAX  ? s ? d : 0 : -(int8_t)d)
1554 #define FSIGNW(d, s) (s <= INT16_MAX ? s ? d : 0 : -(int16_t)d)
1555 #define FSIGNL(d, s) (s <= INT32_MAX ? s ? d : 0 : -(int32_t)d)
1556 SSE_HELPER_B(helper_psignb, FSIGNB)
1557 SSE_HELPER_W(helper_psignw, FSIGNW)
1558 SSE_HELPER_L(helper_psignd, FSIGNL)
1559 
1560 void glue(helper_palignr, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1561                                   int32_t shift)
1562 {
1563     Reg r;
1564 
1565     /* XXX could be checked during translation */
1566     if (shift >= (16 << SHIFT)) {
1567         r.Q(0) = 0;
1568         XMM_ONLY(r.Q(1) = 0);
1569     } else {
1570         shift <<= 3;
1571 #define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
1572 #if SHIFT == 0
1573         r.Q(0) = SHR(s->Q(0), shift - 0) |
1574             SHR(d->Q(0), shift -  64);
1575 #else
1576         r.Q(0) = SHR(s->Q(0), shift - 0) |
1577             SHR(s->Q(1), shift -  64) |
1578             SHR(d->Q(0), shift - 128) |
1579             SHR(d->Q(1), shift - 192);
1580         r.Q(1) = SHR(s->Q(0), shift + 64) |
1581             SHR(s->Q(1), shift -   0) |
1582             SHR(d->Q(0), shift -  64) |
1583             SHR(d->Q(1), shift - 128);
1584 #endif
1585 #undef SHR
1586     }
1587 
1588     *d = r;
1589 }
1590 
1591 #define XMM0 (env->xmm_regs[0])
1592 
1593 #if SHIFT == 1
1594 #define SSE_HELPER_V(name, elem, num, F)                                \
1595     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)           \
1596     {                                                                   \
1597         d->elem(0) = F(d->elem(0), s->elem(0), XMM0.elem(0));           \
1598         d->elem(1) = F(d->elem(1), s->elem(1), XMM0.elem(1));           \
1599         if (num > 2) {                                                  \
1600             d->elem(2) = F(d->elem(2), s->elem(2), XMM0.elem(2));       \
1601             d->elem(3) = F(d->elem(3), s->elem(3), XMM0.elem(3));       \
1602             if (num > 4) {                                              \
1603                 d->elem(4) = F(d->elem(4), s->elem(4), XMM0.elem(4));   \
1604                 d->elem(5) = F(d->elem(5), s->elem(5), XMM0.elem(5));   \
1605                 d->elem(6) = F(d->elem(6), s->elem(6), XMM0.elem(6));   \
1606                 d->elem(7) = F(d->elem(7), s->elem(7), XMM0.elem(7));   \
1607                 if (num > 8) {                                          \
1608                     d->elem(8) = F(d->elem(8), s->elem(8), XMM0.elem(8)); \
1609                     d->elem(9) = F(d->elem(9), s->elem(9), XMM0.elem(9)); \
1610                     d->elem(10) = F(d->elem(10), s->elem(10), XMM0.elem(10)); \
1611                     d->elem(11) = F(d->elem(11), s->elem(11), XMM0.elem(11)); \
1612                     d->elem(12) = F(d->elem(12), s->elem(12), XMM0.elem(12)); \
1613                     d->elem(13) = F(d->elem(13), s->elem(13), XMM0.elem(13)); \
1614                     d->elem(14) = F(d->elem(14), s->elem(14), XMM0.elem(14)); \
1615                     d->elem(15) = F(d->elem(15), s->elem(15), XMM0.elem(15)); \
1616                 }                                                       \
1617             }                                                           \
1618         }                                                               \
1619     }
1620 
1621 #define SSE_HELPER_I(name, elem, num, F)                                \
1622     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t imm) \
1623     {                                                                   \
1624         d->elem(0) = F(d->elem(0), s->elem(0), ((imm >> 0) & 1));       \
1625         d->elem(1) = F(d->elem(1), s->elem(1), ((imm >> 1) & 1));       \
1626         if (num > 2) {                                                  \
1627             d->elem(2) = F(d->elem(2), s->elem(2), ((imm >> 2) & 1));   \
1628             d->elem(3) = F(d->elem(3), s->elem(3), ((imm >> 3) & 1));   \
1629             if (num > 4) {                                              \
1630                 d->elem(4) = F(d->elem(4), s->elem(4), ((imm >> 4) & 1)); \
1631                 d->elem(5) = F(d->elem(5), s->elem(5), ((imm >> 5) & 1)); \
1632                 d->elem(6) = F(d->elem(6), s->elem(6), ((imm >> 6) & 1)); \
1633                 d->elem(7) = F(d->elem(7), s->elem(7), ((imm >> 7) & 1)); \
1634                 if (num > 8) {                                          \
1635                     d->elem(8) = F(d->elem(8), s->elem(8), ((imm >> 8) & 1)); \
1636                     d->elem(9) = F(d->elem(9), s->elem(9), ((imm >> 9) & 1)); \
1637                     d->elem(10) = F(d->elem(10), s->elem(10),           \
1638                                     ((imm >> 10) & 1));                 \
1639                     d->elem(11) = F(d->elem(11), s->elem(11),           \
1640                                     ((imm >> 11) & 1));                 \
1641                     d->elem(12) = F(d->elem(12), s->elem(12),           \
1642                                     ((imm >> 12) & 1));                 \
1643                     d->elem(13) = F(d->elem(13), s->elem(13),           \
1644                                     ((imm >> 13) & 1));                 \
1645                     d->elem(14) = F(d->elem(14), s->elem(14),           \
1646                                     ((imm >> 14) & 1));                 \
1647                     d->elem(15) = F(d->elem(15), s->elem(15),           \
1648                                     ((imm >> 15) & 1));                 \
1649                 }                                                       \
1650             }                                                           \
1651         }                                                               \
1652     }
1653 
1654 /* SSE4.1 op helpers */
1655 #define FBLENDVB(d, s, m) ((m & 0x80) ? s : d)
1656 #define FBLENDVPS(d, s, m) ((m & 0x80000000) ? s : d)
1657 #define FBLENDVPD(d, s, m) ((m & 0x8000000000000000LL) ? s : d)
1658 SSE_HELPER_V(helper_pblendvb, B, 16, FBLENDVB)
1659 SSE_HELPER_V(helper_blendvps, L, 4, FBLENDVPS)
1660 SSE_HELPER_V(helper_blendvpd, Q, 2, FBLENDVPD)
1661 
1662 void glue(helper_ptest, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1663 {
1664     uint64_t zf = (s->Q(0) &  d->Q(0)) | (s->Q(1) &  d->Q(1));
1665     uint64_t cf = (s->Q(0) & ~d->Q(0)) | (s->Q(1) & ~d->Q(1));
1666 
1667     CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
1668 }
1669 
1670 #define SSE_HELPER_F(name, elem, num, F)        \
1671     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)     \
1672     {                                           \
1673         if (num > 2) {                          \
1674             if (num > 4) {                      \
1675                 d->elem(7) = F(7);              \
1676                 d->elem(6) = F(6);              \
1677                 d->elem(5) = F(5);              \
1678                 d->elem(4) = F(4);              \
1679             }                                   \
1680             d->elem(3) = F(3);                  \
1681             d->elem(2) = F(2);                  \
1682         }                                       \
1683         d->elem(1) = F(1);                      \
1684         d->elem(0) = F(0);                      \
1685     }
1686 
1687 SSE_HELPER_F(helper_pmovsxbw, W, 8, (int8_t) s->B)
1688 SSE_HELPER_F(helper_pmovsxbd, L, 4, (int8_t) s->B)
1689 SSE_HELPER_F(helper_pmovsxbq, Q, 2, (int8_t) s->B)
1690 SSE_HELPER_F(helper_pmovsxwd, L, 4, (int16_t) s->W)
1691 SSE_HELPER_F(helper_pmovsxwq, Q, 2, (int16_t) s->W)
1692 SSE_HELPER_F(helper_pmovsxdq, Q, 2, (int32_t) s->L)
1693 SSE_HELPER_F(helper_pmovzxbw, W, 8, s->B)
1694 SSE_HELPER_F(helper_pmovzxbd, L, 4, s->B)
1695 SSE_HELPER_F(helper_pmovzxbq, Q, 2, s->B)
1696 SSE_HELPER_F(helper_pmovzxwd, L, 4, s->W)
1697 SSE_HELPER_F(helper_pmovzxwq, Q, 2, s->W)
1698 SSE_HELPER_F(helper_pmovzxdq, Q, 2, s->L)
1699 
1700 void glue(helper_pmuldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1701 {
1702     d->Q(0) = (int64_t)(int32_t) d->L(0) * (int32_t) s->L(0);
1703     d->Q(1) = (int64_t)(int32_t) d->L(2) * (int32_t) s->L(2);
1704 }
1705 
1706 #define FCMPEQQ(d, s) (d == s ? -1 : 0)
1707 SSE_HELPER_Q(helper_pcmpeqq, FCMPEQQ)
1708 
1709 void glue(helper_packusdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1710 {
1711     Reg r;
1712 
1713     r.W(0) = satuw((int32_t) d->L(0));
1714     r.W(1) = satuw((int32_t) d->L(1));
1715     r.W(2) = satuw((int32_t) d->L(2));
1716     r.W(3) = satuw((int32_t) d->L(3));
1717     r.W(4) = satuw((int32_t) s->L(0));
1718     r.W(5) = satuw((int32_t) s->L(1));
1719     r.W(6) = satuw((int32_t) s->L(2));
1720     r.W(7) = satuw((int32_t) s->L(3));
1721     *d = r;
1722 }
1723 
1724 #define FMINSB(d, s) MIN((int8_t)d, (int8_t)s)
1725 #define FMINSD(d, s) MIN((int32_t)d, (int32_t)s)
1726 #define FMAXSB(d, s) MAX((int8_t)d, (int8_t)s)
1727 #define FMAXSD(d, s) MAX((int32_t)d, (int32_t)s)
1728 SSE_HELPER_B(helper_pminsb, FMINSB)
1729 SSE_HELPER_L(helper_pminsd, FMINSD)
1730 SSE_HELPER_W(helper_pminuw, MIN)
1731 SSE_HELPER_L(helper_pminud, MIN)
1732 SSE_HELPER_B(helper_pmaxsb, FMAXSB)
1733 SSE_HELPER_L(helper_pmaxsd, FMAXSD)
1734 SSE_HELPER_W(helper_pmaxuw, MAX)
1735 SSE_HELPER_L(helper_pmaxud, MAX)
1736 
1737 #define FMULLD(d, s) ((int32_t)d * (int32_t)s)
1738 SSE_HELPER_L(helper_pmulld, FMULLD)
1739 
1740 void glue(helper_phminposuw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1741 {
1742     int idx = 0;
1743 
1744     if (s->W(1) < s->W(idx)) {
1745         idx = 1;
1746     }
1747     if (s->W(2) < s->W(idx)) {
1748         idx = 2;
1749     }
1750     if (s->W(3) < s->W(idx)) {
1751         idx = 3;
1752     }
1753     if (s->W(4) < s->W(idx)) {
1754         idx = 4;
1755     }
1756     if (s->W(5) < s->W(idx)) {
1757         idx = 5;
1758     }
1759     if (s->W(6) < s->W(idx)) {
1760         idx = 6;
1761     }
1762     if (s->W(7) < s->W(idx)) {
1763         idx = 7;
1764     }
1765 
1766     d->W(0) = s->W(idx);
1767     d->W(1) = idx;
1768     d->L(1) = 0;
1769     d->Q(1) = 0;
1770 }
1771 
1772 void glue(helper_roundps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1773                                   uint32_t mode)
1774 {
1775     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1776     signed char prev_rounding_mode;
1777 
1778     prev_rounding_mode = env->sse_status.float_rounding_mode;
1779     if (!(mode & (1 << 2))) {
1780         switch (mode & 3) {
1781         case 0:
1782             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1783             break;
1784         case 1:
1785             set_float_rounding_mode(float_round_down, &env->sse_status);
1786             break;
1787         case 2:
1788             set_float_rounding_mode(float_round_up, &env->sse_status);
1789             break;
1790         case 3:
1791             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1792             break;
1793         }
1794     }
1795 
1796     d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
1797     d->ZMM_S(1) = float32_round_to_int(s->ZMM_S(1), &env->sse_status);
1798     d->ZMM_S(2) = float32_round_to_int(s->ZMM_S(2), &env->sse_status);
1799     d->ZMM_S(3) = float32_round_to_int(s->ZMM_S(3), &env->sse_status);
1800 
1801     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1802         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1803                                   ~float_flag_inexact,
1804                                   &env->sse_status);
1805     }
1806     env->sse_status.float_rounding_mode = prev_rounding_mode;
1807 }
1808 
1809 void glue(helper_roundpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1810                                   uint32_t mode)
1811 {
1812     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1813     signed char prev_rounding_mode;
1814 
1815     prev_rounding_mode = env->sse_status.float_rounding_mode;
1816     if (!(mode & (1 << 2))) {
1817         switch (mode & 3) {
1818         case 0:
1819             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1820             break;
1821         case 1:
1822             set_float_rounding_mode(float_round_down, &env->sse_status);
1823             break;
1824         case 2:
1825             set_float_rounding_mode(float_round_up, &env->sse_status);
1826             break;
1827         case 3:
1828             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1829             break;
1830         }
1831     }
1832 
1833     d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
1834     d->ZMM_D(1) = float64_round_to_int(s->ZMM_D(1), &env->sse_status);
1835 
1836     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1837         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1838                                   ~float_flag_inexact,
1839                                   &env->sse_status);
1840     }
1841     env->sse_status.float_rounding_mode = prev_rounding_mode;
1842 }
1843 
1844 void glue(helper_roundss, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1845                                   uint32_t mode)
1846 {
1847     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1848     signed char prev_rounding_mode;
1849 
1850     prev_rounding_mode = env->sse_status.float_rounding_mode;
1851     if (!(mode & (1 << 2))) {
1852         switch (mode & 3) {
1853         case 0:
1854             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1855             break;
1856         case 1:
1857             set_float_rounding_mode(float_round_down, &env->sse_status);
1858             break;
1859         case 2:
1860             set_float_rounding_mode(float_round_up, &env->sse_status);
1861             break;
1862         case 3:
1863             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1864             break;
1865         }
1866     }
1867 
1868     d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
1869 
1870     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1871         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1872                                   ~float_flag_inexact,
1873                                   &env->sse_status);
1874     }
1875     env->sse_status.float_rounding_mode = prev_rounding_mode;
1876 }
1877 
1878 void glue(helper_roundsd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1879                                   uint32_t mode)
1880 {
1881     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1882     signed char prev_rounding_mode;
1883 
1884     prev_rounding_mode = env->sse_status.float_rounding_mode;
1885     if (!(mode & (1 << 2))) {
1886         switch (mode & 3) {
1887         case 0:
1888             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1889             break;
1890         case 1:
1891             set_float_rounding_mode(float_round_down, &env->sse_status);
1892             break;
1893         case 2:
1894             set_float_rounding_mode(float_round_up, &env->sse_status);
1895             break;
1896         case 3:
1897             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1898             break;
1899         }
1900     }
1901 
1902     d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
1903 
1904     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1905         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1906                                   ~float_flag_inexact,
1907                                   &env->sse_status);
1908     }
1909     env->sse_status.float_rounding_mode = prev_rounding_mode;
1910 }
1911 
1912 #define FBLENDP(d, s, m) (m ? s : d)
1913 SSE_HELPER_I(helper_blendps, L, 4, FBLENDP)
1914 SSE_HELPER_I(helper_blendpd, Q, 2, FBLENDP)
1915 SSE_HELPER_I(helper_pblendw, W, 8, FBLENDP)
1916 
1917 void glue(helper_dpps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1918 {
1919     float32 iresult = float32_zero;
1920 
1921     if (mask & (1 << 4)) {
1922         iresult = float32_add(iresult,
1923                               float32_mul(d->ZMM_S(0), s->ZMM_S(0),
1924                                           &env->sse_status),
1925                               &env->sse_status);
1926     }
1927     if (mask & (1 << 5)) {
1928         iresult = float32_add(iresult,
1929                               float32_mul(d->ZMM_S(1), s->ZMM_S(1),
1930                                           &env->sse_status),
1931                               &env->sse_status);
1932     }
1933     if (mask & (1 << 6)) {
1934         iresult = float32_add(iresult,
1935                               float32_mul(d->ZMM_S(2), s->ZMM_S(2),
1936                                           &env->sse_status),
1937                               &env->sse_status);
1938     }
1939     if (mask & (1 << 7)) {
1940         iresult = float32_add(iresult,
1941                               float32_mul(d->ZMM_S(3), s->ZMM_S(3),
1942                                           &env->sse_status),
1943                               &env->sse_status);
1944     }
1945     d->ZMM_S(0) = (mask & (1 << 0)) ? iresult : float32_zero;
1946     d->ZMM_S(1) = (mask & (1 << 1)) ? iresult : float32_zero;
1947     d->ZMM_S(2) = (mask & (1 << 2)) ? iresult : float32_zero;
1948     d->ZMM_S(3) = (mask & (1 << 3)) ? iresult : float32_zero;
1949 }
1950 
1951 void glue(helper_dppd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1952 {
1953     float64 iresult = float64_zero;
1954 
1955     if (mask & (1 << 4)) {
1956         iresult = float64_add(iresult,
1957                               float64_mul(d->ZMM_D(0), s->ZMM_D(0),
1958                                           &env->sse_status),
1959                               &env->sse_status);
1960     }
1961     if (mask & (1 << 5)) {
1962         iresult = float64_add(iresult,
1963                               float64_mul(d->ZMM_D(1), s->ZMM_D(1),
1964                                           &env->sse_status),
1965                               &env->sse_status);
1966     }
1967     d->ZMM_D(0) = (mask & (1 << 0)) ? iresult : float64_zero;
1968     d->ZMM_D(1) = (mask & (1 << 1)) ? iresult : float64_zero;
1969 }
1970 
1971 void glue(helper_mpsadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1972                                   uint32_t offset)
1973 {
1974     int s0 = (offset & 3) << 2;
1975     int d0 = (offset & 4) << 0;
1976     int i;
1977     Reg r;
1978 
1979     for (i = 0; i < 8; i++, d0++) {
1980         r.W(i) = 0;
1981         r.W(i) += abs1(d->B(d0 + 0) - s->B(s0 + 0));
1982         r.W(i) += abs1(d->B(d0 + 1) - s->B(s0 + 1));
1983         r.W(i) += abs1(d->B(d0 + 2) - s->B(s0 + 2));
1984         r.W(i) += abs1(d->B(d0 + 3) - s->B(s0 + 3));
1985     }
1986 
1987     *d = r;
1988 }
1989 
1990 /* SSE4.2 op helpers */
1991 #define FCMPGTQ(d, s) ((int64_t)d > (int64_t)s ? -1 : 0)
1992 SSE_HELPER_Q(helper_pcmpgtq, FCMPGTQ)
1993 
1994 static inline int pcmp_elen(CPUX86State *env, int reg, uint32_t ctrl)
1995 {
1996     int val;
1997 
1998     /* Presence of REX.W is indicated by a bit higher than 7 set */
1999     if (ctrl >> 8) {
2000         val = abs1((int64_t)env->regs[reg]);
2001     } else {
2002         val = abs1((int32_t)env->regs[reg]);
2003     }
2004 
2005     if (ctrl & 1) {
2006         if (val > 8) {
2007             return 8;
2008         }
2009     } else {
2010         if (val > 16) {
2011             return 16;
2012         }
2013     }
2014     return val;
2015 }
2016 
2017 static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
2018 {
2019     int val = 0;
2020 
2021     if (ctrl & 1) {
2022         while (val < 8 && r->W(val)) {
2023             val++;
2024         }
2025     } else {
2026         while (val < 16 && r->B(val)) {
2027             val++;
2028         }
2029     }
2030 
2031     return val;
2032 }
2033 
2034 static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
2035 {
2036     switch ((ctrl >> 0) & 3) {
2037     case 0:
2038         return r->B(i);
2039     case 1:
2040         return r->W(i);
2041     case 2:
2042         return (int8_t)r->B(i);
2043     case 3:
2044     default:
2045         return (int16_t)r->W(i);
2046     }
2047 }
2048 
2049 static inline unsigned pcmpxstrx(CPUX86State *env, Reg *d, Reg *s,
2050                                  int8_t ctrl, int valids, int validd)
2051 {
2052     unsigned int res = 0;
2053     int v;
2054     int j, i;
2055     int upper = (ctrl & 1) ? 7 : 15;
2056 
2057     valids--;
2058     validd--;
2059 
2060     CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
2061 
2062     switch ((ctrl >> 2) & 3) {
2063     case 0:
2064         for (j = valids; j >= 0; j--) {
2065             res <<= 1;
2066             v = pcmp_val(s, ctrl, j);
2067             for (i = validd; i >= 0; i--) {
2068                 res |= (v == pcmp_val(d, ctrl, i));
2069             }
2070         }
2071         break;
2072     case 1:
2073         for (j = valids; j >= 0; j--) {
2074             res <<= 1;
2075             v = pcmp_val(s, ctrl, j);
2076             for (i = ((validd - 1) | 1); i >= 0; i -= 2) {
2077                 res |= (pcmp_val(d, ctrl, i - 0) >= v &&
2078                         pcmp_val(d, ctrl, i - 1) <= v);
2079             }
2080         }
2081         break;
2082     case 2:
2083         res = (1 << (upper - MAX(valids, validd))) - 1;
2084         res <<= MAX(valids, validd) - MIN(valids, validd);
2085         for (i = MIN(valids, validd); i >= 0; i--) {
2086             res <<= 1;
2087             v = pcmp_val(s, ctrl, i);
2088             res |= (v == pcmp_val(d, ctrl, i));
2089         }
2090         break;
2091     case 3:
2092         if (validd == -1) {
2093             res = (2 << upper) - 1;
2094             break;
2095         }
2096         for (j = valids == upper ? valids : valids - validd; j >= 0; j--) {
2097             res <<= 1;
2098             v = 1;
2099             for (i = MIN(valids - j, validd); i >= 0; i--) {
2100                 v &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
2101             }
2102             res |= v;
2103         }
2104         break;
2105     }
2106 
2107     switch ((ctrl >> 4) & 3) {
2108     case 1:
2109         res ^= (2 << upper) - 1;
2110         break;
2111     case 3:
2112         res ^= (1 << (valids + 1)) - 1;
2113         break;
2114     }
2115 
2116     if (res) {
2117         CC_SRC |= CC_C;
2118     }
2119     if (res & 1) {
2120         CC_SRC |= CC_O;
2121     }
2122 
2123     return res;
2124 }
2125 
2126 void glue(helper_pcmpestri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2127                                     uint32_t ctrl)
2128 {
2129     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2130                                  pcmp_elen(env, R_EDX, ctrl),
2131                                  pcmp_elen(env, R_EAX, ctrl));
2132 
2133     if (res) {
2134         env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2135     } else {
2136         env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2137     }
2138 }
2139 
2140 void glue(helper_pcmpestrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2141                                     uint32_t ctrl)
2142 {
2143     int i;
2144     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2145                                  pcmp_elen(env, R_EDX, ctrl),
2146                                  pcmp_elen(env, R_EAX, ctrl));
2147 
2148     if ((ctrl >> 6) & 1) {
2149         if (ctrl & 1) {
2150             for (i = 0; i < 8; i++, res >>= 1) {
2151                 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2152             }
2153         } else {
2154             for (i = 0; i < 16; i++, res >>= 1) {
2155                 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2156             }
2157         }
2158     } else {
2159         env->xmm_regs[0].Q(1) = 0;
2160         env->xmm_regs[0].Q(0) = res;
2161     }
2162 }
2163 
2164 void glue(helper_pcmpistri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2165                                     uint32_t ctrl)
2166 {
2167     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2168                                  pcmp_ilen(s, ctrl),
2169                                  pcmp_ilen(d, ctrl));
2170 
2171     if (res) {
2172         env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2173     } else {
2174         env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2175     }
2176 }
2177 
2178 void glue(helper_pcmpistrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2179                                     uint32_t ctrl)
2180 {
2181     int i;
2182     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2183                                  pcmp_ilen(s, ctrl),
2184                                  pcmp_ilen(d, ctrl));
2185 
2186     if ((ctrl >> 6) & 1) {
2187         if (ctrl & 1) {
2188             for (i = 0; i < 8; i++, res >>= 1) {
2189                 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2190             }
2191         } else {
2192             for (i = 0; i < 16; i++, res >>= 1) {
2193                 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2194             }
2195         }
2196     } else {
2197         env->xmm_regs[0].Q(1) = 0;
2198         env->xmm_regs[0].Q(0) = res;
2199     }
2200 }
2201 
2202 #define CRCPOLY        0x1edc6f41
2203 #define CRCPOLY_BITREV 0x82f63b78
2204 target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
2205 {
2206     target_ulong crc = (msg & ((target_ulong) -1 >>
2207                                (TARGET_LONG_BITS - len))) ^ crc1;
2208 
2209     while (len--) {
2210         crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
2211     }
2212 
2213     return crc;
2214 }
2215 
2216 void glue(helper_pclmulqdq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2217                                     uint32_t ctrl)
2218 {
2219     uint64_t ah, al, b, resh, resl;
2220 
2221     ah = 0;
2222     al = d->Q((ctrl & 1) != 0);
2223     b = s->Q((ctrl & 16) != 0);
2224     resh = resl = 0;
2225 
2226     while (b) {
2227         if (b & 1) {
2228             resl ^= al;
2229             resh ^= ah;
2230         }
2231         ah = (ah << 1) | (al >> 63);
2232         al <<= 1;
2233         b >>= 1;
2234     }
2235 
2236     d->Q(0) = resl;
2237     d->Q(1) = resh;
2238 }
2239 
2240 void glue(helper_aesdec, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2241 {
2242     int i;
2243     Reg st = *d;
2244     Reg rk = *s;
2245 
2246     for (i = 0 ; i < 4 ; i++) {
2247         d->L(i) = rk.L(i) ^ bswap32(AES_Td0[st.B(AES_ishifts[4*i+0])] ^
2248                                     AES_Td1[st.B(AES_ishifts[4*i+1])] ^
2249                                     AES_Td2[st.B(AES_ishifts[4*i+2])] ^
2250                                     AES_Td3[st.B(AES_ishifts[4*i+3])]);
2251     }
2252 }
2253 
2254 void glue(helper_aesdeclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2255 {
2256     int i;
2257     Reg st = *d;
2258     Reg rk = *s;
2259 
2260     for (i = 0; i < 16; i++) {
2261         d->B(i) = rk.B(i) ^ (AES_isbox[st.B(AES_ishifts[i])]);
2262     }
2263 }
2264 
2265 void glue(helper_aesenc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2266 {
2267     int i;
2268     Reg st = *d;
2269     Reg rk = *s;
2270 
2271     for (i = 0 ; i < 4 ; i++) {
2272         d->L(i) = rk.L(i) ^ bswap32(AES_Te0[st.B(AES_shifts[4*i+0])] ^
2273                                     AES_Te1[st.B(AES_shifts[4*i+1])] ^
2274                                     AES_Te2[st.B(AES_shifts[4*i+2])] ^
2275                                     AES_Te3[st.B(AES_shifts[4*i+3])]);
2276     }
2277 }
2278 
2279 void glue(helper_aesenclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2280 {
2281     int i;
2282     Reg st = *d;
2283     Reg rk = *s;
2284 
2285     for (i = 0; i < 16; i++) {
2286         d->B(i) = rk.B(i) ^ (AES_sbox[st.B(AES_shifts[i])]);
2287     }
2288 
2289 }
2290 
2291 void glue(helper_aesimc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2292 {
2293     int i;
2294     Reg tmp = *s;
2295 
2296     for (i = 0 ; i < 4 ; i++) {
2297         d->L(i) = bswap32(AES_imc[tmp.B(4*i+0)][0] ^
2298                           AES_imc[tmp.B(4*i+1)][1] ^
2299                           AES_imc[tmp.B(4*i+2)][2] ^
2300                           AES_imc[tmp.B(4*i+3)][3]);
2301     }
2302 }
2303 
2304 void glue(helper_aeskeygenassist, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2305                                           uint32_t ctrl)
2306 {
2307     int i;
2308     Reg tmp = *s;
2309 
2310     for (i = 0 ; i < 4 ; i++) {
2311         d->B(i) = AES_sbox[tmp.B(i + 4)];
2312         d->B(i + 8) = AES_sbox[tmp.B(i + 12)];
2313     }
2314     d->L(1) = (d->L(0) << 24 | d->L(0) >> 8) ^ ctrl;
2315     d->L(3) = (d->L(2) << 24 | d->L(2) >> 8) ^ ctrl;
2316 }
2317 #endif
2318 
2319 #undef SHIFT
2320 #undef XMM_ONLY
2321 #undef Reg
2322 #undef B
2323 #undef W
2324 #undef L
2325 #undef Q
2326 #undef SUFFIX
2327